UTILIZATION OF COMPUTERS FOR MANAGEMENT AND DEVELOPMENT OF TRANSPORTATION SYSTEMS: PAPERS AND PROCEEDINGS OF A U.S.-U.S.S.R. SEMINAR, MOSCOW 1975
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP79-00798A000200020005-9
Release Decision:
RIFPUB
Original Classification:
K
Document Page Count:
423
Document Creation Date:
December 12, 2016
Document Release Date:
December 6, 2000
Sequence Number:
5
Case Number:
Publication Date:
January 1, 1975
Content Type:
REPORT
File:
Attachment | Size |
---|---|
![]() | 15.56 MB |
Body:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Utilization of Computers for
the Management and Development
of Transportation Systems:
Papers and Proceedings of a
U.S.-U.S.S.R. Seminar, Moscow
1975
State Dept. declassification & release instructions on file
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Utili7 ti.on of Co::puters for
the 2iar'~a~;c~tLC It and Developra_,.ut
of Transportation Systcirs:
Papers and Proceedings of a
U.S.-U.S.S.R. Scsni.nar, Moscow
Sydney Shulr;an, Administrative Editor
Translation and publication of this work was supported by grant
No. GJ-41942 DCR 74-3.3801.A02 from the National. Science Foundation
to the National Bureau of Economic Research, Tnc. However, any
opinions, findings, conclusions or recor..?.acndaticits expressed here-
in are those of the authors and do not necessarily reflect the
views of the National Science Foundation or the National Bureau.
Since the present volume is a record of conference proceedings,
it has been exempted from the rules governing submission of manu-
scripts to, and critical review by, the Board of Directors of the
National Bureau.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
CONTENTS
INTRODUCTION ....................?.........??..?................ i
Harvey McMains
AN OVF:IRVIt:61 OF U.S. TRANSPOR:TATI.ON: SIMILARITIES AND
CONTRA:>Tc, 1'IT}i THE. U.S. S. P .. ................................... 1
Rolland Hurter
FREIGHT Dt:"ODELTING: A POLICY-SENSITIVE APPROACH........... 25
Marc N. Terziev, Mushe Yen-40hiva tnd Paul 0. Roberts
IWEDS /3W PRIORITIES IN RESEARCH ON RAIL SI:P.VICE RELIABILITY... 81
Jo.erh H. Su.`l.'nan
' i . ' ?1'i." d1'P)' ctif.l,~"H'i.rS TO TkAVHI, 1'MAIil) LuR. t:cASTI i ( .?????.?????????? 15
\.u ME dt; a
? Marvin L. M.-inhei.r.
EYAMPLES OF COIIPUT);I: API'LIC4MUNS ICI THE J^~::SPOI.1'r~i.'IOI
FIELD .......................................................... 193
Faye C. Johnson
CO IT'UTY11 APPLICATION IN THE ALLOCATION OF AIRLINE RESOUI CF.S....
223
Morton Ehrlich
A SOFTWARE SYSTF'?1 FOR URBAN
Y.obez t Y. Dial
TRANSPORTATION PLANNING............
271
AN EVALUATION OF TILE AIR QUALITY I"TACTS OF TRANSPORTATION
CONTROL POLICIES IN U. S . URBAN AREAS ........................... 288
Gregory K. Ingram
THE INTRODUCiTON OF MATIIF2.fATICAL-ECOL O'IIC METHODS AND COMPUTER
TECIP.NOLOGY IN PLANNING AND MA AGING SOVIET TRANSPORTATION...... 336
Boris S. Nozin
OPTIMIZING MODELS FOR PLANNING THE OPERATION AND DEVELOPMENT
OF A TRANSPORT NET'v'OPJ'. ......................................... 343
I. T. Kozlov
METHODS OF FIVE.-YEAR PLANNING OF TRANSPORT-ECONOMIC CONNECTIONS:
THEORETICAL DEVELOPI-T.:TFS AND EXPERIENCE WITH PRACTICAL APPLICA-
TIONS ........................................................... 353
P. I. Mokrousova and Z. I. Ffozgrina
PROBLEMS OF OPTIMAL PLANTING AND MANAGEMENT OF AUTOBUS
TRANSPORT IN THE GEORGIAN SSR ................................... 366
G. G. Tsonaja
AUTOMATION OF BOOKING AND RESERVATION OPERATIONS
ON SOVIET RAILROADS .................. .......... ......'......'..... 377
B. E. Marchuk
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
THE "SIREN" SYSTEM: A NATIONWIDE AUTOMATIC CONTROL SYSTEM
FOR BOOKING AND RESERVING SEATS ON DOMESTIC AIRLINES............ 386
V.A. Thozhiksshvili et al.
INFORMATION MODELING FOR MANAGING SOVIET MARITIME
TRANSPORTATION .................................................. 405
V. S. Bondarenko
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
i
INTRODUCTION
The U.S.-U.S.S.R. Conference on Utilization of Computers for
the Management and Development of Transportation Systems took
place in the Soviet Union on June 28-July 10, 1975. Fifteen papers
were presented at the conference and are included in these proceed-
ings. The purpose of the conference was to share concepts and
techniques being used in the Soviet Union and the United States
in the planning, design, implementation and use of computer models,
simulations and programs for managing and operating transportation
systems. There were open and frank discussions in regard to all
the material presented. Both sides agreed that the next activity
would be the development of mutually beneficial, longer term joint
research.
Harvey McMains
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
AN OVERVIEW OF U.S. TRANSPORTATION:
SIMILARITIES AND CONTRASTS WITH THE USSR
Holland Hunter
This brief essay offers a primarily statistical sketch of the
United States transportation system as it has evolved in the last
quarter century or so. It is intended to set the stage for the research
papers that follow, papers describing ways in which systems analysis
aided by computers is being used in United States transportation. The
sketch highlights some features of freight and passenger transportation
that distinguish the United States situation from that of other countries,
in particular that of the USSR. Both the U.S. and the USSR are large
countries with large economies and large transportation systems. The
contrasts, however, are as interesting as the similarities, and in our
view much can be learned by examining both to see what they imply for the
improved performance of each system.
Intercity Freight and Passenger Transportation
Over the last half century, United States intercity transportation
has been marked primarily by a decline in the relative role of the
railroads, matched by the steady rise of other carriers. For freight
transportation, trends over the last thirty-five years are shown in table
1. Railroads accounted for two-thirds of the intercity freight traffic
during the 1939-1948 decade; in recent years this share has fallen below
40%. Pipelines and intercity trucks have rapidly and steadily increased
their share, each now accounting for over 22% of total intercity freight
traffic. Freighters and barges on the Great Lakes, the Mississippi-Ohio-
Missouri River system, and the coastal waterways of the Atlantic, Gulf,
and Pacific Coasts have steadily raised the volume of their traffic and
* pAgggyS0 FqAraWj%"28NJjYJ 9 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 CIA-RDP79-00798A000200020005-9
Table 1
Domestic Intercity Freight Traffic, U.S., by Carrier,
Five-Year Totals, 1939-43 through 1969-73, in-I illions
of Metric Ton-I.Ei'lometers
Motor Inland Oil
Railroads Vehicles Waterways Pipelines Airways, Total
1939-43. 4070 457 912 520 - 5989
1944-48 5202 621 1032 848 1 7704
1949-53 4631 1295 1177 1057 3 8163
1954-58 4513 1745 1507 1527 4 9296
1959-63 4438 2197 1583 1723 7 9948
1964-68 5354 2746 1993 2424 16 12524
1969-73 5742 3264 2386 3319 26 14737
Percent Shares
1939-43 68.0 7.6 15.7 8.7 - 100
1944-48 67.5 8.1 13.4 11.0 - 100
1949-53 56.7 15.9 14.4 13.0 - 100
1954-58 48.6 18.8 16.2 16.4 - 3.00
1959-63 44.6 22.1 15.9 17.3 .1 100
1964-68 42.7 21.9 15.9 19.4 .1 100
1969-73 39.0 22.1 16.2 22.5 .2 100
Source: Compiled from annual data in ton-miles (1 metric ton - km - 1.46 short ton-mile)
in U.S. Bureau of the Census, Historical Stat. of the U.S. (1960), Series Q1-11
(adjusted for coverage changes); Statistical Abstract of the U.S: 1974, p. 547;
and U.S. Interstate Commerce Commission, 88th Annual Report, p. 120.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
maintained their share of about one-sixth of the total.
United States intercity passenger travel is heavily dominated by
the passenger automobile, as shown in table 2. Since 1950, privately
owned automobiles have accounted for at least 86% of all intercity
passenger travel; their share in 1960 reached 90% but since then has
fallen back slightly. The balance is handled by airways, buses, rail-
roads, and internal waterways, with the air share steadily rising to
over 10% and the share of buses and railroads shrinking. The absolute
volume of bus travel has been fairly steady, but rail passenger travel
has fallen by two-thirds over the last quarter century. Today, air
transport is clearly the dominant form of public intercity passenger
transport, accounting for well over 70% of intercity passenger travel not
done by automobile.
The revenues earned by carriers for all their domestic freight and
passenger service are displayed in table 3. Trends reflect the physical
changes already noted but with significant distinctions. The railroad
share of revenue, for example, is less than their share of freight, since
railroads carry a good deal of low-value freight. The truck share of
revenues, conversely, is higher than their share of freight, and, recently,
intercity trucking has taken in more than 42% of all transportation
revenues. The airline share of total passenger and freight revenues has
grown rapidly from 4% to 20%. Pipeline revenues, though they have tripled
over the last quarter century, are still only 3% of all transportation
revenue since costs and charges for this form of movement are extremely
low. Low per-unit costs also characterize internal waterway freight
traffic, where the revenue share has been falling though the volume of
traffic continues to grow and the freight share is stable.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 CIA-RDP79-00798A000200020005-9
Table 2 4
Domestic Intercity Passenger Taffic, U.S., by Carrier,
Selected Years, 1950-1973, in billions of Passenger-
kilometers
Private
Automobiles
Airways
Buses
Railroads
. Inland
Waterways
Total
1950
704.9
16.1
41.8
51.5
1.9
816.2
1955
1025.1
37.0
40.2
46.7
2.7
1151.7
1960
1136.2
54.7
30.6
35.4
4.3
1261.2
1965
1316.4
93.3
38.6
29.0
5.0
1482.3
1970
1651.2
191.5
40.2
27.7
6.4
1907.0
1973
Percent Shares
1950
86.4
2.0
5.1
6.3
0.2
100
1955
89.0
3.2
3.5
4.1
0.2
100
1960
90.1
4.3
2.4
2.8
0.4
100
1965
88.8
6.3
2.6
2.0
0.3
100
1970'
86.6
10.0
2.1
0.9
0.4
100
1973
Source: Passenger-tile data from U.S. Dept. of Commerce, Bureau of the Census,
Statistical' Abstract of the U.S., 1975, p. 562.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Table 3
Revenues From Domestic Freight and Passenger Traffic,
U.S., by Carrier, Selected Years, 1950-1973, in millions
of Dollars
Railroads Trucks
Airlines
Pipelines Buses Waterlines
Total
1950
10,147
3,737
558
442 539 330
15,753
1955
10,831
5,535
1215
678 552 452
19,263
1960
10,203
7,214
2129
770 667 427
21,410
1965
11,054
10,068
3609
904 885 426
26,946
1970
12,824
14,585
7131
1188 1062 502
37,292
1973
15,864
20,800
9605
1446 1135 615
49,465
Percent Shares
1950
64.4
23.7
3.6
2.8
3.4 2.1
100
1955
56.2
28.7
6.3
3.5
2.9 2.4
100
1960
47.7
33.7
9.9
3.6
3.1 2.0
100
1965
41.0
37.4
13.4
3.3
3.3 1.6
100
1970
34.4
39.1
19.1
3.2
2.9 1.3
100
1973
32.1
42.1
19.4
2.9
2.3 1.2
100
Source:
Statistical Abstract of the U.S., 1975, p.
560.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Table 4 presents summary data for 1973, showing not only the ton-
kilometers of freight handled by each of the four major carriers but also
their tons originated and the resulting quotient showing each carrier's
average length of haul. Intercity freight typically moves for long
distances in the United States, reflecting the continental dimensions
of the economy. Railroads and pipelines have an average length of haul
above the overall national average, while waterways and highways show a
somewhat lower figure. These crude averages, both in aggregate and for
the individual modes, summarize an underlying distribution including a
great deal of short-haul movement offset by significant amounts of
extremely long-haul traffic. Strictly local movement by trucks in urban
areas is specifically excluded from these statistics.
The massive expansion of automotive passenger and freight trans-
portation is dramatically symbolized by the data in table 5 showing the
route length of railroads and highways in the United States over the
period 1890-1973. Initially, the total length of the railroad network
was greater than the total length of surfaced highways, but by 1920 the
highway network exceeded the railroad system in length. After 1930,
U.S. railroads began a gradual abandonment of little-used roadway; the
1973 length of the railroad system is below that of 1910. In sharp
contrast, however, the length of surfaced highways almost quadrupled from
1930 to 1973. It will be seen that the total length of all highways
outside towns and cities has grown only modestly; the basic change has
involved upgrading through providing existing roads with all-weather
surfaces. These summary figures also understate the qualitative improve-
ment that has come with modern highway construction, especially with the
69,000 kA Ppe'Qed8Fd R 9#"MX1;filel9 $ - DP3Mt-W?98,4bbof66610gdg_luger
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Table 4
Freight Shipments and Average Length of Haul, U.S., 1973,
by Carrier, in Kilometers and millions of Metric Tons
Ton-kilometers
(billions)
Tons Originated
(millions)
Average Haul
(kilometers)
Railroads
1,253
1,466
855
Highways
737
1,833
402 (a)
Waterways
523
902
580
Pipelines
740
1,091
.678
Four-carrier Total
3,253
5,292
615
(a) Class I and II common and contract carriers, 1971. See ICC, Trans.
Stat. of the U.S., part 7, release 2, pp. 48, 54, 164.
Source: U.S. Dept. of Commerce, Bureau of the Census, Stat. Abstract of the
U.S., 1975, pp. 580-81, 586, 596.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Length of Railroads and Highways, U.S., Selected
Years, 1890-1973, in Thousands of Kilometers
Highways (outside towns and cities)
Railroads Surfaced Unsurfaced Total
1890 251 NA NA NA
1900 311 248 (a) NA NA
1910 388 328 NA NA
1920 418 594 4113 4707 (b)
1930 418 1117 3725 4842
1940 396 2156 2956 4812
1950 381 2702 2110 4812
1960 370 3484 1531 5015
1970 354 3880 1220 5100
1973 348 3920 1191. 5111
(a) 1904
(b) 1921
Source: U.S. Dept. of Commerce, Bureau of the Census, Historical
Stat. of the U.S. to 1957, pp. 429, 458; Stat. Abstract
of the U.S., 1975, pp. 564, 581.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
and freight movement by highway has been substantially enhanced by this
national grid of limited-access, divided-lane superhighways. Of course,
at the same time, these improvements in the highway system have been
occurring, corresponding technological improvements have occurred in
railroading (e.g., centralized traffic control and improvements in the
geometry and weight of trackage), so that even though railroad system
kilometers have contracted, the total capability of the railroad system
has remained at least constant and probably enlarged.
The usage of roads and highways outside cities in the United States
is divided roughly three to one between passenger automobiles and trucks,
with buses accounting for a negligible share of total vehicle kilometers.
The figures in table 6 show that, measured in billions of vehicle-
kilometers per year, the volume of use by passenger cars in 1973 was
3.5 times as large as in 1940, bus use had doubled, and truck use was
five times as great. This growth in usage went well beyond the expansion
in nonurban highway mileage, so that intercity traffic density per route
kilometer more than doubled. As a crude overall average, in 1940 one
could see thirteen vehicles per hour pass one point on the network every
hour every day through the year. By 1973 this figure had risen to
twenty-eight. The truck share of highway usage was gradually expanding
at the expense of buses and passenger cars and, though the growth of
bus usage appeared to be leveling off, growth in truck and passenger
car use of the highways was vigorous through 1973. Higher fuel prices
may check this growth somewhat, but the forces pressing for further
growth seem very strong.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
16'
INTERCITY MOTORWVEHICLE MOVEMENT AND DENSITY,
U.S., SELECTED YEARS, 1940-1973
(in billions of vehicle-kilometers and hourly vehicle-kilometers
per-kilometer of surfaced highway)
,Passenger
Buses Trucks
Cars
Total
Hourly Number
of Vehicles
1940
193.9
2.3
48.6
244.8
13
1950
291.4
3.4
91.4
386.2
16
1960
488.1
3.7
131.5
623.3
20
1970
654.0
4.5
215.8
874.3
26
1973
715.0
4.7
247.8
967.5
28
Percent Shares
1940
79.2
0.9
19.9
100
1950
75.4
0.9
23.7
100
1960
78.2
0.6
21.2
100
1970
74.8
0.5
24.7
100
1973
73.9
0.5
25.6
100
Source:
Derived 'from U.S. Department of Commerce,
Bureau of the Census, Stat. Abstract 175,
pp. 571 and 564
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Urban Transportation
In the last forty years urban passenger transportation in the
United States has shifted markedly away from public carriers toward
the private automobile. Table 7 shows the number of passengers carried
by the major means of mass transit at various times between 1930 and
1973. On subways and elevated electric trains, the number dropped from
13 billion annually to less than 2 billion. Railroad commutation passen-
gers dropped from 435 to 183 million per year. Passengers using buses
increased markedly from 1930 to 1950, but their number has since been
cut in half. The trolley bus similarly showed a rapid expansion up to
1950 and an equally sharp decline since then. Overall, the annual number
of passenger trips made on these four major means of mass transit has
dropped from 17.5 billion in 1950 to less than 7 billion in 1973. Much
public policy has been directed toward promoting the use of mass transit,
but the United States urban population has stubbornly continued to prefer
the private passenger automobile.
The census of 1970 asked a 15% sample of urban and rural households
about how working members of each household got to work. Their answers
are summarized in table 8. Two-thirds of the labor force rode to work
in automobiles that they drove, and more than 10% rode as passengers.
Those who walked to work made up 7.4%, and 9% used bus, street car,
subway, elevated train, railroad, or taxicab. A few used other means,
and about 4% of the urban labor force (over 12% of the rural labor force)
worked at home. This pattern reflects in part the way that residences
and work places have been dispersed in postwar U.S. metropolitan regions;
it is increasingly inconvenient or even impossible to get from home to
work and back by means other than the passenger automobile.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200Q20005-9
URBAN PASSENGER TRAFFIC, U.S., BY PUBLIC CARRIER,
SELECTED YEARS, 1930-1973
'(in millions of passengers carried)
Subway and
Elevated
Electric RR
Railroad
Commutation
Motor
Bus
Trolley
Bus
Total
1930
13,072
435
2,479
16
16,002
1940
8,325
229
4,239
534
13,327
1950
6,168
?277
9,420
1,658
17,523
1960
2,313
203
6,425
657
9,598
1970
2,116
206
5,034
182
7,538
1973
1,921
183
4,642
97
6,843
Sources: U.S. Dept. of Commerce, Bureau of the Census,
Hist. Stat. of the U.S., p. 464;
Stat. Abstract '75, p. 579; and (for railroads)
Assoc. of Am Railroads, Railroad Trans., A Statistical Record
1921-63 (Wash., 1965), p. 24; Stat. of Railroads of Class I
(Aug. 1974), p. 7.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
TABLE 8
MEANS OF TRANSPORTATION TO WORK, U.S., 1970
URBAN AND RURAL WORKERS
(in millions of workers 16 years old or over)
Urban
Rural
Passenger automobile
As driver
38,133
12,565
50,698
As passenger
6,702
2,322
9,024
On foot
4,506
1,183
5,689
Bus or streetcar
4,116
129
4,245
Subway or elevated train
1,762
6
1,768
Railroad
473
29
502
Taxicab
278
18
296
Other, or worked at home
2,258
2,371
4,629
Total
58,228
18,623
76,851
Percent Shares
Passenger automobile
As driver
65.5
67.5
66.0
As passenger
11.5
12.5
11.7
On foot
7.7
6.3
7.4
Bus or streetcar
7.1
0.7
5.5
Subway or elevated train
3.0
--
2.3
Railroad
0.8
0.2
0.7
Taxicab
0.5
0.1
0.4
Other, or worked at home
3.9
12.7
6.0
Total
100.0
100.0
100.0
Source: U. S. Department of Commerce, Bureau of the Census, Stat. Abstract '75
p. 578
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The streets and highways of U.S. cities and towns are dominated by
passenger automobiles, as shown in the data of table 9. Automatic
counting devices and inspectors at selected locations regularly measure
the number of vehicles of each type using the roads. Passenger cars
accounted for 86% of all vehicle-kilometers in 1940, and their 84% share
in 1973 was only slightly lower. The truck share rose from 13% to 16%,
while the share of buses dropped from slightly under 1% to only 0.3%.
The absolute volume of bus movement has stopped growing. Though the
total length of urban streets and highways has expanded, traffic density
(measured as the hourly number of vehicles passing a given point) has
roughly doubled since 1940.
Some Basic Dimensions of Soviet Transportation
The statistics above were presented in metric tons and kilometers
to facilitate comparisons with analogous Soviet measures of transporta-
tion. Here a few Soviet statistics are presented to illustrate some
basic contrasts between the transportation systems of the two economies.
Table 10, similar in form to table 4, shows for 1973 the level of
activity of the four major freight carriers. It is immediately apparent
that railroads dominate Soviet freight transportation to a greater
extent than has been true in the United States for many decades. Soviet
railroads carry more than twice as much freight as U.S. railroads for
approximately the same average distance. The Soviet figure for truck.
traffic relates mainly to local rather than intercity movement (the
average haul in the USSR is 16 kilometers compared to 402), so for
comparability most of this Soviet truck traffic should be excluded from
a comparison of intercity traffic. Conversely, a portion of Soviet
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
URBAN MOTOR-VEHICLE MOVEMENT AND DENSITY,
U.S., SELECTED YEARS, 1940-1973
(in billions of vehicle-kilometers and hourly vehicle-kilometers
per kilometer of surfaced highway)
Passenger
Cars Buses Trucks
Total
Hourly Number
of Vehicles
1940
207.8
1.9
31.7
241.4
N.A.
1950
293.7
3.2
54.4
351.3
77
1960
458.3
3.4
71.9
533.6
88
1970
795.8 '
3.5
129.7
929.0
117
1973
953.0
3.4
182.0
1,138.4
128
Percent Shares
1940
86.1
0.8
13.1.
100
1950
83.6
0.9
15.5
100
1960
85.9
0.6
13.5
100
1970
85.7
0.4
13.9
100
1973
83.7
0.3
16.0
100
Source:
Derived from U. S. Dept. of Commerce,
Bureau of the Census, Stat. Abstract '75,
pp. 571 and 564.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Table 10
Freight Traffic, Shipments, and Average Length of Haul,
USSR,.1973, by Carrier, in Billions of Metric Ton-Kilo-
meters , Millions of. Metric Tons, and yalcTneters
Ton-Kilometers
(Billions)
Tons Originated
'(Millions)
Average Haul
"(Kilometers)
Railroads
2,958
3,346
884
Highways
284
18,244
16
Waterways
190
419
453
Pipelines
439
421
1,043
Four-Carrier Total
3,871
22,430
173
Source: TsSU, Narkhoz SSSR 1973, pp. 503, 504, 510, 516-17
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
17
maritime freight traffic is domestic, moving between Soviet ports, and
should be included, though it is not covered in table 10.
Soviet internal waterways are not as favorably located (e.g., the
Volga River system does not directly connect Moscow with the eastern
Ukraine or the Urals, nor is there a good water connection between the
Urals and western Siberia) as those in the United States, and the volume
of Soviet waterways freight traffic is thus only about a third of the
U.S. level. Soviet pipelines, though less well developed than U.S.
pipelines, are already moving almost 40% as much as U.S. pipelines over
longer average distances. The aggregate volume of Soviet freight traffic,
using comparable coverage, would probably be some 10Z-15% greater than
U.S. intercity freight traffic. This aggregate measure is hard to
interpret, since it covers differing commodity structures in the two
countries' freight traffic, reflecting in turn the differences between
the two economies in their output composition.
Soviet intercity passenger traffic is only about one-third as
great as passenger movement in the United States. The difference lies
mainly, of course, in the absence of appreciable intercity travel by
passenger automobile (it is not yet even estimated in Soviet transpor-
tation"statistics). In marked contrast to U.S. railroads, Soviet rail-
roads are still carrying an increasing volume of intercity passenger
traffic; in 1973 they handled almost six times as many passenger-
kilometers as U.S. railroads handled in 1950. Soviet bus traffic,
similarly, in 1973 was six times the 1950 U.S. level. Soviet air traffic
has been growing very rapidly and, in 1973, reached about 40% of the
1973 U.S. level. Modest movement on inland waterways is about the same
in both. economies. While this is not the place to examine them in detail,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
DOMESTIC INTERCITY PASSENGER TRAFFIC, USSR,
BY CARRIER, SELECTED YEARS, 1950-1973
(in billions of passenger-kilometers)
Railroads
Buses
Airways Inland Waterways
Sea
Total
1950
88.0
5.2
1.2
2.7
1.2
98.3
1955
141.4
20.9
2.8
3.6
1.5
170.2
1960
. 170.8
61.0
12.1
4.3
1.3
249.5
1965
201.6
120.5
38.1
4.9
1.5
366.6
1970
265.4
202.5
78.2
S.4
1.6
553.1
1973
296.6
253.9
98.8
5.9
1.9
657,1
Percent Shares
1950
89.5
5.3
1.2
2.8
1.2
100
1955
83.1
12.3
1.6
2.1
0.9
100
1960
68.5
24.5
4.8
1.7
0.5
100
1965
55.0
32.9
10.4
1.3
0.4
100
1970
48.0
36.6
14.1
1.0
0.3
100
1973
45.2
38.6
15.0
0.9
0.3
100
Source:
TsSU, Narkhoz '73, p.
500
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
major issues of national policy surround the question of whether the
relative absence of intercity passenger automobile movement in the
USSR can, or should, continue in the future.
Table 12 shows the length of railroads and highways in the USSR
in a format similar to that of table 5 above. One notes that, though
the Soviet railroad network is very large, it has only about one-third
the length of the U.S. system before the latter began contracting. Most
observers would say that the U.S. system was overbuilt, but perhaps this
crude comparison suggests that additional railroad mileage will be useful
for the Soviet economy. As to highways, the contrast is most striking:
in 1973, the U.S. had 3.6 times as long a road system as the USSR, and
for surfaced highways the ratio was 6.6 to 1. Russian and Soviet writers
have been commenting on the problem of "roadlessness" in Russia for at
least a century and, clearly, much remains to be done. Foreign visitors
riding along the skeletal existing paved highway system are unlikely to
be aware of the lack of paved roads throughout most of the country.
The data on Soviet urban passenger traffic in table 13 show how
rapidly it has increased during the postwar period. The total number of
passengers carried per year has risen from 9 to 53 billion, almost all
moved by traditional carriers. The growth constrasts sharply with the
U.S. decline shown in table 7 (the latter, of course, confined to the
minority who do not travel by passenger automobile). In the USSR since
1950, the number of passengers carried by autobus has risen very rapidly,
as has the much smaller number who use taxis. In 1950, streetcars
dominated the scene, but their volume of traffic has been declining since
1965 and now accounts for only 15% of the total. Trolley bus, subway,
and railway commutation traffic has steadily increased though not to the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
TABLE 12
LENGTH OF RAILROADS AND HIGHWAYS, USSR,
SELECTED YEARS, 1940-1973
(in thousands of kilometers)
Highways
Surfaced
Unsurfaced
1940
106
143
1,388
1,531
1950
117
177
1,373
1,550
1960
126
271
1,095
1,366
1970
135
512
852
1,364
1973
137
598
800
1,398
Sources:
TsSU, Transport
i Sviaz
SSSR
(1972).
pp.89, 262
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Auto
Bus
1950 1,001
1955 4,294
1960 10,797
1965 17,771
1970 25,901
1973 30,458
1950 11.5
1955 28.8
1960 43.3
1965 51.2
1970 56.2
1973 57.6
URBAN PASSENGER TRAFFIC, USSR, BY CARRIER
SELECTED YEARS, 1950-1973
(in millions of passengers carried)
Street
car
Trolley
Bus Subway
Railway
Taxi
Total
5',157
945
629
955
43
8,730
6,367
1,858
937
1,392
45
14,893
7,842
3,055'
1,148
1,713
389
24,944
8,242
4,298
1,652
2,049
717
34,729
7,962
6,122
2,294
2,616
1,144
46,039
7,998
7,298
2,727
2,970
1,461
52,912
Percent Shares
59.1
10.8
7.2
10.9
0.5
100
42.8
12.5
6.3
9.3
0.3
100
31.4
12.2
4.6
6.9
1.6
100
23.7
12.4
4.7
5.9
2.1
100
17.3
13.3
5.0
5.7,
'
2.5
100
15.1
13.8
5.1
5.6
2.8
100
Sources: TsSu, Transpoii.. i Sviaz,.(1972) pp. 99-100, 244, 246-47, 256-57;
Narkhoz 173, pp 504, 522-23, 525,
Nark hoz '58, p. 588 (for autobus 1950 and 1955).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
same extent as bus traffic. These summary data are not conclusive, but
they suggest that urban public transit carriers in the USSR have a record
of healthy growth.
Analytic Contrasts Between United States and Soviet Transportation.
Behind these statistics for the two continental economies lie
several fundamental contrasts in the conditions facing transportation
agencies and the users of their services. It is worth noting them
briefly for the light they shed on the problems that systems analysis
can deal with. The most important concerns relations between supply and
demand.
In the United States, the demand for transportation confronts an
ample supply of transportation capacity. Almost everywhere, in fact,
increments of transportation demand could be quickly accommodated by at
least one carrier, and rival transportation agencies are eager to handle
more traffic. In the USSR, however, transport demand presses hard on
transport supply, both freight and passenger, and added demand is not
easily accommodated. The Soviet government did not inherit an over-
built railroad system, nor has it put major resources into a well-
developed system of local, district, and interregional paved highways.
Soviet transport capacity has barely kept up with burgeoning demand.
In the United States, the typical producing enterprise can choose
among multiple sources of supply for its inputs, multiple routes linking
the supplier with the enterprise, and multiple carriers to provide the
freight service. The availability of alternatives gives the enterprise
considerable leverage in obtaining good service. Prompt and flexible
adjustments are facilitated since bottlenecks at particular points can
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
be bypassed. Speed and reliability of service are encouraged. There
may be costs imposed on the overall economy by excess capacity in the
transportation sector, not continuously and fully utilized, but these
costs may be more than offset by benefits to the overall economy in
reduced inventories, faster production, improved product quality, and
speedier adaptation to customer needs.
Several of the following research papers focus on factors that
enter into the economic decisions made by shippers under these conditions.
United States transportation agencies need accurate information on what
their customers want, since without this information the carrier may not
obtain and ratain the shipper's orders. The factors that enter into
travelers' choices among modes of travel and specific passenger carriers
are the object of similar analysis. Railroads, airlines, bus lines, and
trucking firms in the United States are not free to set the conditions
under which they offer service, solely on the basis of their own con-
venience and internal cost-minimizing considerations. The choices made
by shippers and travelers have a decisive influence on the growth and
welfare of the transport sector.
Other papers concentrate on ways that systems analysis is used by
United"States transportation agencies to minimize their operating costs
and provide efficient service. Mr. Johnson's paper shows how railroads
use computers to keep track of freight cars and how trucking firms match
trucks with shipment flows. Mr. Ehrlich's paper shows how Eastern
Airlines uses an elegant systems analysis to find a cost-minimizing
solution to the problem of assigning aircraft and flight crews.to a
pattern of flights offered the public, a pattern which has already been
carefully tailored to fit the evidence on what the public desires.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 24
The delegation of United States transportation specialists was
impressed in July 1975 by what they heard and saw concerning the ways
that Soviet transportation agencies are using systems analysis and
computers. Clearly, simlar problems lead to similar answers. Soviet
analysts are searching for cost-minimizing solutions in the internal
operations of the carrier. Railroads plan their heavy freight traffic
flows, and the Moscow subway carefully monitors its heavy passenger
traffic. Aeroflot uses its ticketing and reservation service to mini-
mize empty-seat-kilometers, and the maritime fleet keeps track of its
farfiung merchant marine. Intracity Soviet truck traffic is centrally
organized in interesting ways that go beyond the decentralized operations
of United States trucking firms, suggesting that Soviet methods of
trucking might be a fruitful topic for joint analysis.
In spite of the differences between the USSR and the United States
in balances of supply and demand for freight and passenger transportation,
it seems clear that a great many problems of efficient operation for
each means of transportation can usefully be studied by very similar
methods of computer-aided systems analysis. Here an exchange of operat-
ing experiences and a comparison of analytic methods can be illuminating
for experts on both sides. In addition, it seems to us that further
study of the factors influencing the demand for transportation, both in
the minds of shippers and in the minds of travelers, can serve the
interests of the Soviet economy even though intercarrier competition is
absent. Long-run growth and efficiency in any economy require that
economic activity should respond correctly to the objective needs-of
consumers and producers, and for the transportation sector this means
that demand-sensitive analysis is necessary, no matter who own the means
of product`u roved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
FREIGHT DEMAND MODELLING: A POLICY SENSITIVE APPROACH
Mark N. Terziev, Moshe Ben-Akiva, and Paul 0. Roberts
1. INTRODUCTION
Determining the volume of cargo that will flow in a given freight
market is the starting point for any quantitative analysis of freight
policy. This is true whether the issues being addressed are those of
a carrier who would like to consider changes in a particular pricing
policy or a government attempting to justify the major capital expendi-
ture for a new facility. In spite of what appears to be an obvious need
for analysis tools, very little work has been done to provide such a
capability in the freight area.
One may account for this situation in several ways. First, there
has not been data available to use in model development. This is not
a condition that should be allowed to continue to exist since the data
could be obtained once the needs are known precisely. Second, there
has not been a clearly articulated statement of purpose for such a
model. This is understandable in the light of rising government interest
in the problems of the freight sector. Finally, the lack of an adequate
theoretical framework has hindered both the collection of data and the
statement of purpose for the appropriate analytical tools. This lack
is one which we intend to address directly in this writing.
Demand Supply Equilibrium in Freight Markets
First, it is appropriate to attempt to clear up the second deficien-
cy, the lack of a clearly stated purpose. This can best be done by
putting the demand forecasting element into perspective relative to the
demand-supply equilibrium of freight markets. The shipper of freight
makes a number of decisions which influence the size and nature of the
* Director, Center for Transportation Studies and Professor, M.I.T.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
market between any two points. It is useful to take the viewpoint of
the shipper in constructing the demand element of an equilibrium model
of the system. The attributes of the commodity being shipped and the
characteristics of the shipper and the markets in which he is engaged
will be important, as will information concerning the transportation
system available to the shipper in making his decisions.
These characteristics of the transportation system are influenced
directly by the decisions made by the carriers competing in a given
freight market. The size and type of equipment, its scheduling, reli-
ability, etc., all determine directly, along with the tariff, the level
of service which the shipper will receive and ultimately influence his
demand for service. The supply element of the system treats the decisions
made by the carrier in his offerings.
Equilibrium occurs in the system because of the range of choice
that exists and the feedback nature of the information between supply
and demand. Thus, as the tariff for transport by one mode increases
relative to alternative modes there is diversion to the lower cost modes
by those shippers most sensitive to shipment costs. The output of the
equilibrium process includes the volumes of shipments by the different
modes of freight transport and the levels of service experienced by the
shippers at these shipment levels. (See figure 1.1.)
The Role of Government in the Process
The process is influenced by government in a variety of ways, all
on the supply side. Government typically controls entry, mergers, owner-
ship, finances, rates, and routes by means of the regulatory agencies.
Through the executive and administrative agencies taxes, tax credits,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
FIGURE 1.1
NATURE OF THE EQUILIBRIUM PROCESS
COMMODITY,
SHIPPER AND.
MARKET
ATTRIBUTES
MANAGEMENT
OF CARRIER
PRODUCTION
PROCESS
N'
DEMAND FOR TRANSPORT
SERVICES
SUPPLY OF TRANSPORT
SERVICES
LEVELS
OF
LSERVICE
EQUILIBRIUM
PROCESS
TRANSPORT
VOLUME
LEVELS
GOVERNMENT, TAXATION
REGULATION AND
INFRASTRU?TUBE
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
tolls, subsidies, environmental and safety standards are set and transport
infrastructure is provided. In those efforts government is attempting
to act in a manner that is equitable and in the best interests of its
constituency, within a rather broad interpretation of the meaning of this
phrase.
Making the Models Policy Sensitive
For any set of models to be policy sensitive they must incorporate
the variables reflected in the policies of interest. Thus, if the policy
involves the impact of a change. in equipment on the competitive advantage
of a particular mode, the supply models must be able to represent these
changes in the level of service as they will eventually be perceived by
the shipper. If the policy involves a change in pricing, the level of
service vector must reflect the new tariff as seen by the shipper. A
variety of supply side models may be needed to explore the full range of
policies of interest since the control variables employed directly by
the government and the carrier are on the supply side.
On the demand side the different policies are represented by the
list of characteristics making up the level of service vector. If a level
of service vector can be specified which captures the principal deter-
minants of travel demand on the part of the shipper, then a demand model
employing this level of service vector can be used to handle a large
range of policy alternatives. Each of the policy alternatives to be
studied must be transformed by the supply models into the variables in
this vector. As a consequence, one well-specified demand model can handle
a variety of supply side policy investigations. Of course, specifying
the variables to be incorporated in the level of service vector will be
an imports t undertaking.
. nApproved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
28
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Role of Each Element in the Equilibrium Process
Each of the elements of the system has a different function. The
demand model attempts to replicate the decision-making behavior of an
individual shipper of a commodity faced with a number of alternatives.
This behavior varies with the attributes of the product being shipped,
the shipper, and the market. The supply models-must transform the policy
variables under study into a level of service vector for each alternative
which might be chosen. Both are involved in the equilibrium process
since the level of service inputs into the demand element may not elicit
travel demand volumes which are compatible with the original level of
service estimates produced by the supply element. The volume must be
consistent with the level of service at equilibrium.
Our focus here is on the demand element since it is the least
developed and the most crucial in some sense. That is, once a policy
to be investigated has been transformed into the level of service attri-
butes using even the roughest of supply models the volume levels can be
obtained. Even if the transformation of the policy into level of service
attributes is done on a completely intuitive basis, the model can be
useful.- Therefore, it is necessary to consider the possible range of
policy-questions which might be addressed.
Possible Users of Such a Model
A fundamental understanding of the way in which freight flows are
determined would be of use to government planners, policy-makers and
regulators, as well as carriers and shippers. An analysis of freight
flows is required for the evaluation of government policy options in the
following areas of interest:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. Freight transportation facility planning in various
corridors
2. Economic analysis of projects involving the modification
or expansion of the freight transportation systems
3. Impact of changes in regulation which affect the price
and level of service of freight carriage
4. Overall modal policy planning
Carriers would also benefit from a better understanding of how
their supply policies interact with the shippers' demand for the trans-
port of goods. The major areas of concern to the carrier include the
following:
1. Analysis of potential markets
2. Impacts of changes in the level of service
3. Impacts of changes in pricing policy
4. Evaluation of changes in regulations
5. Determining the feasibility of new services
Shippers would like to coordinate their short-run policies with
their long-run plans. However, the complex interaction of the shippers'
behavior with the policies of the other parties makes the task of
coordination very difficult. A better knowledge of this interaction
process would help shippers deal with the following issues:
1. Evaluation of location options
2. Determination of capacity allocation and production plans
Possible Policy Issues to be Investigated with the Model
Perhaps the most important set of issues facing transportation
policy-makers today are those dealing with the future of the railroads.
How can they be made more competitive? Are there ways to rationalize
the network with line abandonments, mergers, or government ownership of
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
the right-of-ways? Would new, advanced, multimodal services be success-
ful? If so, how should the services be priced? How should tariffs be
established? What impact would new services have on conventional service?
For any of these to be handled in more than a superficial manner, a
demand model of the type described here is necessary.
Another important set of policy questions is concerned with dereg-
ulation. The analysis of deregulation policies can be undertaken on a
market-by-market basis with the techniques outlined here. To address
the questions at a more macroscopic level it will be necessary to place
the demand model into a more aggregate framework and consider the system-
wide impacts. This model, however, represents the necessary first step
in such a process.
Policies involving fuel conservation and air quality have recently
received attention. The demand element is especially important in this
area since the ability to predict the impact of various policies on the
overall level of demand may be crucial. The extent of the impact on
choice of mode or shipment size of various transportation control strat-
egies may also be of concern.
Many other questions such as truck size-weight issues, expansion of
the interstate system, rate absorption in ocean shipping, the imposition
of tolls on inland waterways, the development of the domestic air freight
system, etc., can be addressed. The key is the translation of the policy
questions into level of service vectors for each of the alternatives
under investigation, using either a formally developed set of supply
models or good intuitive judgement as previously suggested. The demand
model is therefore a crucial element in any quantitative analysis whether
the policies to be investigated are those of a private sector entrepreneur,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
a public sector decision maker, or a legislative body seeking the impact
of proposed changes in legislation.
The Need for Data on Shipper Behavior
At the present time, forecasting freight movements is not widely
done. A good bit of freight movement data exists, but it is hard to
coordinate and use the data. Their volume alone makes it difficult to
deal with. This would not be such a problem if the data did not lack
several crucial elements needed for accurate forecasting. The data
currently available are largely historical flows, by commodity, over
given corridors and specific facilities. The response to government or
carrier policy is by the shipper. Thus, the shipper's views are extreme-
ly important. To be useful, the data must capture the behavior of the
shipper. In making his choice, the shipper views the alternatives that
are available to him and makes his choice between them based on the
conditions of transport and the relative costs. Data concerning the
choices made by the shipper rarely exist.
It would be difficult to use the data on shipper behavior if it
were available only in tabular form. It is necessary to capture the
values of the shipper; for example, how he weighs transit time as opposed
to reliability or cost. This can be most easily done by the use of a
carefully structured econometric model which captures and makes available
the "valued" choices of the shipper. This is not to say that these are
not useful in their own right. Quite the contrary, there are a wide
variety of uses for the data, including the calibration of other models
for other uses. At the moment, however, data on the choice behavior of
shippers is unavailable in.a coordinated and useful form.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Why Policy-Sensitive Freight Demand Models Don't Currently Exist
Part of the reason the data does not exist and why models which
incorporate it also do not exist is because the models of the past
were not behavioral models. Instead they were aggregate models which
captured correlations of aggregate quantities rather than individual
shipper behavior. The development of disaggregate behavioral models
using modern methods of econometrics is fairly recent. The development
of data sets which can be used to calibrate or test them has not yet
been accomplished.
New methods for predicting passenger travel demand have advanced
very rapidly over the past few years. The profession has come a long
way since the invention of the "gravity model." Disaggregate behavioral
econometric models employing less data collection and more efficient
data utilization have been developed (CRA, 1972; Ben Akiva, 1973).
The disaggregate approach appears to be equally attractive for
freight demand models. In fact, preliminary investigations have shown
that disaggreagte freight demand models are both workable and encour-
aging. The impact of this new modelling approach is likely to have
considerable more import than the various aggregate econometric models
developed during the past ten years. The pioneer efforts were illustra-
tive, but not very practical for prediction purposes.
The Role of Data Collection
The design of such a model must proceed in conjunction with the
design of a data scheme and methods for data collection. Disaggregate
models require much less data for calibration than aggregate models and,
therefore, it is nowt entirely practical to gather the data needed for a
freight pp Approved Re ease 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Objective of Current Work
The overall objective of this research effort is to build upon the
existing work that has been done in freight demand modeling by employing
the recent developments in disaggregate passenger travel demand models.
The first steps in this direction are to review existing sources of data
(including disaggregate data sets if they are available) and to survey
the literature on freight demand models with particular emphasis on
previous research done on disaggregate models. The major elements of
this research effort are the specification of a system of policy-sensitive
disaggregate freight demand models based on a realistic theory of shipper's
choice of origin, destination, frequency, shipment size, and mode of
transport, and the specification of data requirements and their collection
methods. Ultimately, this methodological phase should be followed by
actual data collection, model estimation, sensitivity tests and validation,
and, finally, the use of these models as policy analysis tools.
2. FRAMEWORK FOR FREIGHT DEMAND MODELLING
In this section the framework for freight demand modelling is des-
cribed in terms of the specific shippers' choices and the variety of
factors that determine the pattern of commodity flows.
Shipper Behavior `
The pattern of freight flows results from the actions of numerous
actors. As noted previously, the basic decision-making unit in determin-
ing the demand for freight transport is the individual shipper. Depending
upon the ownership of the freight, the decision-maker is usually located
at either the origin or the destination of a shipment. The behavior
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
that leads to a demand for freight transport can best be described from
the point of view of this individual decision-maker.
A strategic decision (and the most long-range action of any shipper)
is a choice of location. For intercity freight the location of produc-
tion plants and warehouses substantially determines the requirements for,
and the characteristics of, available freight transport services.1
For a given set of production and consumption points, the pattern
of commodity flows is determined by the individual shipper's selection
of markets and suppliers. The quantity of freight of commodity type k
that the producer at point i is shipping to his warehouse at destination
j, for example, is determined by the shipper's choice to market his
product in a specific volume in the market area served by the warehouse
at point J.
Given the quantity of commodity k shipped from origin i to destin-
ation j over a certain time period, an individual shipment (as represented
by a waybill) is determined by the choices of mode m and shipment size
q. There are a variety of feasible combinations of modes and shipment
sizes that can be used to transport a given annual volume between two
locations. Note that the annual volume and the shipment size also deter-
mine the frequency at which shipments are made.
The Choice Hierarchy
Thus, the basic unit of commodity flow--the individual shipment--
is determined by a complex hierarchy of choices made by an individual
shipper. This hierarchy is depicted in figure 2.1. The sequence that
is assumed in this hierarchy represents the different time lags in a
shipper's response to achange in transport policy. For some shippers
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
FIGURE-2.1
HIERARCHY OF SHIPPER CHOICES
Choice of location
and level of activity
Choice of markets
and/or suppliers
locations and
volumes
Choice of transpor
mode and shipment I
size
Commodity
. Flows
V~
Y?3
.Vk
ijmq
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
and commodity types, the transport mode and shipment size may be altered
on very short notice and at a very small cost. However, a location
change almost always requires significant time and capital expenditures.
It is important to note that this choice hierarchy does not imply
a long-run, sequential, causal relationship. In a long-range decision,
a lower-level choice such as the choice of the transport mode, for example,
influences the choice of location, and vice versa.
This means that a lower-level choice is determined in two ways:
strategic (long run) and tactical (short run). A change in a higher-
level choice is likely to result in a reconsideration of all lower-level
choices. Thus, in a strategic decision all the choices are jointly
determined, and in a tactical decision some high-level choices are fixed
and only lower-level choices are adjusted.
Another important implication of this hierarchy is that at any given
point in time we are likely to observe a shipper at a disequilibrium
point. There is a threshold level below which level of service changes
do not influence shippers' choice in the short run. For this reason,
we can observe an inefficient pattern of commodity flows which are due
to the cost involved in adjusting the choices high on the choice hier-
archy.
Factors Affecting Shipper Behavior
The freight transportation situation at the level of an individual
shipper can be described in terms of four groups of characteristics.
First, there are shipper attributes which include those characteristics
relating to the location of the plant, the location of supplies, and
long-run operating policies. Second, there are market attributes which
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
are related to both the long-run choice of markets and the short-term
equilibrium demand for the commodity being shipped. Third, there are
commodity attributes which include the physical characteristics of the
good as well as information concerning its handling and use. The fourth
group consists of transport level of service attributes which depend not
only on the mode, but also the commodity being shipped and the shipment
size being considered.
In the long run, the choice of location, the level of activity, the
shipment size, and the choice of mode are variable at the discretion of
the shipper. The commodity being shipped is assumed fixed, and a unique
commodity type is associated with each basic decision-making unit.
Commodity Attributes
The list of commodity attributes that influence the demand for freight
transport might include the following:
1. Value per pound--basic commodity value at
the origin
2. Shipping density--both of the basic commodity and of the
commodity when packed for shipping
3. Shelf life-number of days to spoilage or obsolesence
4. Product use--manufacturing, processing, final consumption
5. Method of inventory control--single-item, multi-item
6. Stockout consequence--no cost, contribution loss, proba-
bility increase with days, shutdown
A shipper would also consider the temporary storage facilities associated
with the commodity being shipped, although this is variable over the long
run.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Shipper Attributes
Some shipper attributes which are fixed in the short run are
variable in the long run and are determined by the strategic decisions
at the top of the choice hierarchy. The choice of plant location and
the choice of suppliers and markets fixes the distances over which each
commodity must be transported. The choice of plant size (level of
activity) brackets the range in the volume of each commodity which must
be moved. These considerations are summarized by the following variables:
1. Annual production volume--amount produced
2. Number of establishments--number of production plants
3. Location of establishment--on rail siding, water access,
etc.
4. Ownership--whether producer owns the freight
5. Price at origin--whether price is established at factory
6. Decision-maker--point where inventory decisions are made
A shipper would also consider the use of private carriage. This would
effectively replace the short-run mode choice decision with along-run
investment decision.
Market Attributes
Market attributes reflect the market demand for the commodity being
shipped. For the most part, the market equilibrium is variable over the
short run but it can be assumed to be external to the freight demand
decision. Therefore, market attributes influence the intermediate
decisions relating to the level or volume of shipment. The market
variables that influence shippers include:
1. Sales in consuming industry--amount consumed
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
2. Size of consuming industry--number of firms consuming
3. Characteristics of consuming population--socioeconomic
characteristics
4. Size of consuming population--number
5. Location of establishment--on rail siding, etc.
6. Ownership--whether supplier owns the freight
7. Price at destination--whether price is established C.I.F.
destination
8. Decision-maker--point where inventory decisions are made
Mode/Size Choices
In the short run the shipper is left with only the choices of mode
and shipment size. The mode shipment size options considered by a shipper
include all or some of the following:
1. Freight forwarder
2. Airfreight parcels
3. Air container
4. Private truck
5. Contract carrier
6. Common carrier truck LTL
7. Common carrier truck FTL
8. TOFC Plans I-V
9. Rail FCL
10. Rail unit train
11. Inland barge
12. Barge container
13. Sea container
14. Pipeline
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
In some cases, these tactical options will be precluded by higher order
strategic decisions. For example,, the choice of plant location, supplier
location, and market location will determine whether rail, barge, air,
and pipeline are available options in the mode choice decision.
Level of Service Attributes
Once the list of possible mode/shipment size alternatives has been
edited down to the available short-run options, then the final decision
will be based on the modal level of service attributes. For each avail-
able option this group of characteristics
1. Waiting time
2. Travel time
3. Time reliability
4. Loss and damage
5. Transport tariff
6. Other costs
7. Minimum shipment size
8. Special services
The factors influencing shippers' behavior are summarized in figure
2.2. Here an overall view can be gained of the interrelationships between
the variables which are important in the process. We now turn to an
examination of the existing literature to determine the extent to which
models employing these variables in the appropriate way already exist
in useful form.
3. LITERATURE REVIEW
A number of studies of freight demand have appeared in the transpor-
tation and economics literature during the past ten years. The demand
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
C
0
C7
LU
a
Ra..
LL-
C)
4JJ3
L
I1 L 4j
ro 4J
ze Q
-r- 1
4) C C
a) rn
?O .C 4- C L
SO.. 4- r O 0
Cl a) O r? C r- r 4.)
rE L r0 O ro N to
ru r--
O ]
O 1 N (0 N Cl) U
C > E aJ V (1) gCg ?r
c[ -0j . .1~ CL
N
U
Cl. ?r 4.3
C U. 4~ 0) C
??- C N C Q1 C
4) E 4) ?SS. E o 4- .C 0
N r ?r C a) = r O N 4?)
C L U1 L O +) N 4.3 ?r m ro
O 4- C 4.1 ?r U C to C r- ?r +?/ C
N 0 U1 4?) to O r- 0-0
.C ro ?r
O 00 L () O .r r0 N 4-) CL 4-) 4J 1- Ln
a) C N C O t 0) O CO N (1) 0) N
r-?r N?r 0.U N 0. U Cl) gC ?r 'v
N I/) 00. t/) _0.1 O 0.
V!
w
V
Iw
-
N
4) 4-,
?r =
o +r
11 I~
O
r
go
V
?r
C
W
UU<
?
0
}
a)
a)
4.) U 0))
W
?r
i?1
?r
E
L?r4J
4-)
o>a
C1L.
C:)
C
r
N N?r
?r
Q)
C 4/Y S-
il 4-
?r
ro
ro
L
Oct
3
F-
a) N
a) N 0
4- Q. O > Cl
r N m 4-3 U 0
C
4- L O C ?r
fU 3C O > ?r
C S- C:
V) O CL H CL
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
models developed in these studies generally fall into two categories
and representative models from both groups are discussed in this section.
The first group consists of very simple, macroscopic models which pre-
dict only the total freight flow between all origins and destinations.
The data required to calibrate this type of model is relatively easy to
find and therefore most of these models have been empirically estimated.
The principal use of these models is for the prediction of the impact of
system-wide changes in prices or the level of service.
The second group of models consists of microscopic demand models
which address the problem of predicting the freight flow between origin-
destination pairs. These models include variables relating to specific
operations. Therefore, these models can be used to analyze policy
operations dealing with operations on particular links in the freight
transportation network. However, very few of the microscopic demand
models have been thoroughly tested, although some preliminary empirical
work has been done.
The analysis of different freight transportation policy issues
requires models with different levels of detail. Therefore, a freight
demand model may not necessarily be applicable to all policy evaluation
tasks. Both micro and macro models are useful tools.
Sloss (1971)
Sloss has presented a model for predicting the total volume of inter-
city freight traffic carried by trucks. The dependent variable in his
model is the annual tonnage of goods transported by for-hire trucks on
intercity hauls (denoted Vt). The independent variables are: the average
truck revenue per ton (Ct), the average rail revenue per ton (Cr), and
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
a variable representing the general level of economic activity (E).
The economic activity variable was defined as the unweighted sum of
farm cash income, the value of building permits issued and the value
of shipments of manufactured goods. These variables were combined in
a product form model which was calibrated using regression analysis.
The general form of the model is:
Vt a a0 C t a 1 Crag Ea3
The data for the model were obtained from annual reports published
by the Dominion Bureau of Statistics in Ottawa, Canada. Sloss's empir-
ical results imply that the volume of truck traffic is directly related
to the price of rail service and inversely related to the price of truck
service. Sloss also concluded that the quantity of truck traffic is
positively correlated with the general level of economic activity. These
results seem intuitively reasonable.
Sloss's model could be used for a macro-level evaluation of alter-
native pricing strategies. However, it could not be used to evaluate
other system-wide options dealing with changes in operating speeds,
improvements in reliability, etc. Nor could this model be used to pre-
dict the impact of a policy dealing with changes in the price of
carriage of a particular commodity between a particular origin and
destination.
A. D. Little (1974)
As part of an analysis of domestic waterborne shipping, A. D. Little,
Inc. has developed a model which predicts the percentage of intercity
freight traffic moved by water carriers. Unlike most other models which
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
have appeared in the literature, this model emphasizes the strategic
(long-run) aspects of the mode choice decision. In this model, the
level of service variables such as time and cost have been replaced by
variables which indicate the compatibility between the shipper's
operations and the operations of waterborne carriers. For example,
the following variables have been used to characterize shippers in a
BEA region:2
1. Total flow--used because ships and barges are best
suited for carrying high volumes
2. Value per ton--used because water modes typically
carry low-value goods
3. Bulk commodity indicator--used because ships and
barges are well suited for the loading and unloading
of bulk goods
4. Seasonality--used because the high fixed cost of dock
facilities makes it desirable to maintain a high
volume of usage throughout the year
It was found that these four variables are highly correlated with the
fraction of plants located on bodies of water in BEA regions.
No direct measures of level of service were included in this model.
However, two proxy variables were included to represent the differences
between the barge and truck (and rail) levels of service as they might
be perceived in the long run. These variables are the highway distance
(approximately the same as the rail distance) and the ratio of the water
route distance to the highway distance.
The variables described above were used in a nonlinear equation to
predict the fraction of traffic moved by water carriers. The data used
in this study were records of flows of fifteen commodity groups between
BEA regions.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
In terms of policy analysis, strategic models of this kind can be
quite useful. First, they can help identify key variables to use as
long-run policy instruments. Second, they point out how the inherent
advantages of a mode interact with shipper attributes and market attri-
butes in the long-run decision-making process. However, it is obvious
that these models cannot be used for the evaluation of specific pricing
and operating policies which affect long-run, as well as short-run,
shipper behavior.
Perle (1964)
Perle conducted one of the first major studies of freight demand.
Perle used product form models with prices as the only indication of the
level of service on each of the models. He calibrated one model for the
total volume of truck freight and a second for the total volume of rail
freight. The independent variables in both models are the average rail
rate (Cr) and the average truck rate (Ct). Perle used dummy variables
to represent the influence of certain commodity groups and certain regional
characteristics. Since he was using time series data, he also included
a dummy variable to represent the year. But he did not use any market
attributes or variables related to the size and state of the economy.
The general forms of the models are:
brr brt ( r Ri
Vr - arCr Ct i eri
Tj Zk
j drj k frk
V btr btt In Ri
t - atCr ~t i eti
n Tj a Zk)
j dtj k ftk
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Ri - 1 if observation is in ith region
Tj - 1 if observation is in jth year
- 0 otherwise
Zk - 1 if observation is in kth commodity group
- 0 otherwise
Perle used ICC reports to find the total tons carried by each mode during
several years. This data was then used to calibrate several variations
of the basic model shown above.
On the whole, Perle's models did not fit well. Several of the esti-
mated coefficients had the wrong sign. However, the dummy variables
representing commodity specific effects and regional effects significantly
improved the fit of the models. The dummy variable representing the
year was less useful. These results indicate that further effort should
be made to explicitly include commodity and market attributes in freight
demand models.
In terms of policy analysis, the uses and shortcomings of Perle's
models are similar to those of the Sloss model. His models can be used
for the evaluation of pricing options which would affect an entire
system. But Perle's models cannot relate price changes on particular
hauls to impacts on traffic in particular market segments. Moreover,
they neglect many other policy variables such as travel time and reli-
ability. Nevertheless, Perle's work is an important contribution to the
field of freight demand analysis. He used a careful, methodological
approach and his results demonstrate the importance of commodity and
market attributes.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Mathematica (1967)
Mathematics did some theoretical-work in freight demand analysis as
part of the Northeast Corridor Transportation Project. One of the models
which they proposed is an adaptation of the Quandt/Baumol abstract mode
model used in passenger demand modeling . The general form of the model
V . a0 P al P a2 Y a3 Y a4 M as M a6 N a7 (Tb )bo(Tr )bi(Cb )do(Cr ) dl
ijm i j i j i j ij ij ij ij ijm
where:
Vijm - volume of freight flow from i to j by mode m
Pi, Pj a population of the origin and destination
Yi, Yj ? gross regional product of the origin and destination
Mi, Mj ? industrial character indices such as the percent of
the labor force employed in mining and manufacturing
Tij = least shipping time from i to j
Tijm . travel time by mode m divided by the least
from i to j
Cij . least cost of shipping from i to j
Cijm - cost of mode in divided by the least cost from i to j
Nij . number of modes serving i and j
Unlike the other models discussed above, this model deals with the
flow of goods between particular origin-destination pairs. Furthermore,
this model contains a greater range of market attributes and level of'
service attributes, although it does neglect a significant number of
policy variables, such as time reliability.
A second flaw in the Quandt/Baumol model involves intermodal effects.
In this model, changes in the cost and time on an inferior mode will have
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
no effect on the volume carried by the "best" model. This is certainly
counter-intuitive. It is not readily apparent how the model could be
modified to avoid this problem.
Mathematics (1967)
Mathematics has also proposed a freight demand model completely
unlike those discussed above. This model was developed through the
application of classical microeconomics and inventory theory. Without
going through the derivation of the model, the general line of develop-
ment is:
1. Express total annual variable transportation costs of
the industry in city j which is receiving commodity k
from city i as the sum of the direct shipping costs, total
in-transit carrying costs,and safety inventory costs
2. Differentiate total cost to get marginal cost and apply
the optimality rule that marginal cost equals marginal
revenue
Under some assumptions about the cost functions and some approximations
to make the algebra work, the following demand function was derived:
ijm - a0 + a1 (VP1j) + a2 Cijm + a3 (S1 Tijm) + a4 (S1 + Tijm)l/2
Vijm - volume of commodity k moving from i to j by mode m
VPij - the difference in value of commodity k between city
i and city j
" time between shipments of commodity k to j
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Tijm ' average travel time of commodity k from i to 9
on mode m
This kind of microeconomic, inventory-based model might be a good
policy evaluation tool for a shipper doing an analysis of his location
options and general operating policy. It may not be obvious from the
general form of the model, but the following important factors are
implicitly represented by the variables and parameters:
1. interest rate on capital
2. value of the commodity
3. price elasticity of demand for the commodity
4. warehousing costs
5. rate of sales of the commodity
6. cost of stocking-out of the commodity
7. ordering costs
The derivation of the model from a microeconomic point of view allows
the inclusion of these important factors. However, the transition from
a model of an individual firm to a model of the aggregate behavior of
all firms in the origin city was not made explicit in the Mathematica
report. It is not entirely clear how the aggregation issue should be
handled for disaggregate models of this type.
Antle and Haynes (1971)
Antle and Haynes have suggested still another approach to freight
demand analysis. They used a linear discriminant function to predict
the choice of mode, given some basic information about the shipment.
The general form of their model is:
~p1oe834 Refie-s'% z0~47 -1/ 19a:5t.:64-R 9-(t0r i9r000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
where:
X1 = annual tonnage shipped
X2 = distance by principal mode
X3 = average travel time by principal mode
X4 = average shipment size
X5 = rate on principal mode
X6 = rate on competing mode
X7 = handling cost on principal mode
These seven independent variables are used to characterize a single
shipping decision. To use the model, the values for a principal mode
and a competing mode are substituted into the model and a value of Z
is calculated. This value is compared with a test value Z to predict
whether the principal or the competing mode will be chosen. If the
distribution of each of the independent variables is known for a given
commodity, for a given origin-destination city pair, then the volume
Vim can be estimated. So, even though this model is based on discrim-
inant theory, it can be used to produce the same kind of demand infor-
mation as the other models discussed in this section.
The Antle and Haynes model was calibrated on a small amount of data
that was collected on the shipment of coal, coke, petroleum, and
chemicals by barge and by rail in the Upper Ohio River Valley. Using
just the section of the data on coal shipments, they estimated a model
with only four significant independent variables: travel time and rate
on the principal mode, shipment size, and handling cost. Although this
model fitted the data reasonably well, it should be regarded with some
suspicion because it lacks so many important variables. The failure of
the rate on the competing mode to appear in the calibrated model was
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
particularly disappointing.
This model could be used to study the volume of coal carried by
each mode on a particular link in the freight transportation network.
However, the model contains no commodity attributes, which means that a
separate model would have to be calibrated for each commodity being
studied if the model is to be used to evaluate policies related to
particular goods. Despite its shortcomings, the Antle and Haynes study
is important because it helped pave the way for the application of
stochastic, disaggregate qualitative choice models to freight demand.
Rullman (1973)
Rullman used a binary logit model with aggregate data to predict
the truck-rail modal split as a function of level of service attributes,
commodity attributes, and market attributes. The general form of his
model is:
(Rail Share Commodity k
` Truck Share i to j
f (distance, annual volume, value
per ton, difference in rail and
truck freight rates, difference
in rail and truck transit time,
difference in rail and truck transit
time reliability)
As a policy tool, Kullman's model has several desirable characteristics.
It includes a number of level of service variables, including reliability.
It also includes annual volume consumed, which is a market attribute.
Furthermore, Rullman's model is to some extent commodity abstract because
it includes commodity value. (See p. 38, last paragraph.) Although
this model does neglect some important factors, it is a relatively
sophisticated example of a microlevel model.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19: CIA-RDP79-00798A000200020005-9
As a tool for microlevel analysis Kullman's model has three short-
comings. In the first place, it should include a wider range of
commodity and market attributes as well as some shipper attributes.
Secondly, Kullman's empirical work has shown that several important
levels of service variables (such as travel time reliability) cannot be
represented by aggregate data. Aggregate reliability data contains very
little variation among observations and, therefore, the effect of reli-
ability on mode choice cannot be determined. Third, since it is only a
modal split model, in order to predict Vk
ijm' an external estimate of the
total volume of commodity k moving from i to j must be provided. The
major contribution of Kullman's model is the inclusion of a wider range
of variables than most other models. It also demonstrates the feas-
ibility of using the logit model in freight demand analysis.
Hartwig and Linton (1974)
The work of Hartwig and Linton is the most recent effort in the area
of disaggregate freight demand modeling. They used the logit, probit,
and discriminant functions to model the individual shipper's mode choice
between full-load truck and full-load rail.
The data base used in this study consisted of 1213 freight waybills
for full-load truck and rail shipments of consumer durables. From the
bills they determined:
1. The origin and destination of the shipment from which they
determined mileage
2. The date shipped and the date received from which they
determined transit time
3. The freight rate and total charge
4. Shipment weight
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
5. The type of commodity being shipped from which they found
the value per unit of the shipment
In their models, Hartwig and Linton used the difference in truck
and rail transit times, the difference in truck and rail freight charges,
the difference in truck and rail reliability (as measured by the differ-
ence in the standard deviation of the transit time distributions) and
the value of the commodity as the independent variables. The dependent
variable as in the Antel and Haynes and Kullman models is the mode
choice probability. Using these variables, they experimented with logit,
probit, and discriminant models. In general, they found that both logit
and probit performed well, while the discriminant approach yielded some-
what poorer results.
The Hartwig and Linton study indicates that the disaggregate logit
model can be used very successfully in freight demand modelling. However,
their model needs to be expanded if it is to be used as a microlevel
policy tool. Market attributes and shipper attributes should be added
along with some additional level of service variables. Furthermore,
the addition of commodity attributes would make the model useful to
analyze a wider range of commodity types and differential process policies.
In summary, a number of models have been developed and tested which
can be used in macrolevel policy analysis. To date, these models have
not been used extensively. More detailed models capable of predicting
flows of particular commodities between particular cities have developed
more slowly. The principal reason for the slow progress in this area is
the scarcity of data at the required level of detail. Nevertheless, this
second type of model is a potentially powerful method for performing a
detailed analysis of intercity freight operations. It is important to
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
note that, given a microlevel demand model, an appropriate model for
macrolevel analysis can in principle be developed by aggregating the
micro model over the distribution of its independent variables. This
aggregation of microlevel demand models is discussed in the following
section.
4. DISAGGREGATE MODELLING OF SHIPPER BEHAVIOR
This section presents a brief description of the methodology
suggested for modelling shipper behavior. The approach outlined here
avoids many of the problems presented in the work which has been done
to date as described in the previous section. The recommended approach
involves the use of a disaggregate behavioral model.
Why T)isaggregate Models?
A freight demand model is most likely to be used in a study of
freight facilities and policies or for a market analysis of freight
movements of a certain type. The forecasts required are not for the
individual shipper's freight demand but rather for some aggregate patterns
of freight movements. Since the basic decision-making unit for freight
demand is an individual shipper, it follows that the aggregate demand is
simply a sum of individual shipper's demands.
The starting point for a freight demand model formulation is a theory
of the behavior of an individual shipper. However, for the aggregate
forecasting of freight demand, the model of individual shipper's behavior
must be aggregated. The aggregation is based on groupings of shippers
by their geographical locations, industry type, etc. If data is avail-
able at the level of the individual shipper, then this aggregation can
be performed either before or after model estimation.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
An aggregate model is estimated, using the means or totals of the
variables included in the model for an aggregate group of shippers.
Sometimes parameters of the distributions of the independent variables
within the aggregate groups are also included. A disaggregate model
is estimated, using the values of the variables for individual shippers.
The aggregation for forecasting is then performed by integrating the
disaggregate demand model over the distribution of the independent
variables' values for the aggregate group of shippers.
The advantages of estimating a disaggregate model are numerous. One
of the most important points is their efficient use of data. Since the
data is not aggregated prior to model estimation, less original data is
required to obtain reliable estimates of the model's coefficients.
Aggregate data is characterized by a significant loss in the variability
of important variables. Therefore, estimation of an aggregate model
often fails to produce reliable coefficients for variables such as
reliability, waiting time, etc., which do not vary substantially between
large geographic zones but have considerable variations within these
zones.
Because a disaggregate model is not estimated using averages for
specific aggregate groupings, it is not tied to a specific area or a
specific aggregation scheme. Once estimated, a disaggregate model can
be used for a wide range of applications at different areas and different
levels of detail without a need for model reestimation.
Disaggregate freight mode choice models have already been estimated
on a trial basis by Hartwig and Linton (1974) and by Antle and Haynes
(1971). These studies have proven that a disaggregate modelling approach
is workable and attractive for freight demand modelling.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Modelling Qualitative Choice4
At the level of the individual shipper (or the single shipment),
the decisions with respect to transport mode and shipment size, for
example, are characterized as a selection from a set of qualitative
alternatives. In qualitative choice, the alternatives can be indexed
but cannot be rank ordered. Locational, modal, and shipment-size
choices are characterized as qualitative choice problems.
Consider an individual shipper making a selection from a set of
alternatives At where i and j denote alternatives in the set. For
example, the set of alternatives for a model of mode choice and shipment
size, includes all the feasible combinations of available modes and
shipment sizes. For a model that includes a destination choice as well,
the set of alternatives includes all combinations of modes, sizes, and
destinations. Each alternative in the choice set is associated with
some measure of attractiveness (cost) which the shipper is attempting
to maximize (minimize). Given that one and only one alternative is
selected from the choice set then alternative i will be selected if
and only if:
Uit > Ujt, V j c At
where Uit denotes the attractiveness of alternative i to shipper t. In
a mode choice problem, for example, this implies that mode i will be
selected if and only if its attractiveness is greater than or equal to
the attractiveness of any other mode that is available.
The attractiveness of alternative i is a function of the. attributes
of the alternative (e.g., transit time, rate, loss and damage, consump-
tion rate at destination, etc.). Two types of attributes directly
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
influence short-run freight shipment decisions: transport modal attri-
butes (or level of service attributes) and market attributes. However,
the same transport alternative will be evaluated differently by different
shippers of different commodities. Thus, the attractiveness of a given
freight transport alternative is expressed as a function of shipper and
commodity attributes as well as the attributes of the alternatives.
Given data on the observed behavior of shippers and the values of the
important attributes, the attractiveness functions must be estimated if
we are to be able to forecast freight demand under varying conditions.
Due to measurement errors, unobservable information, and other
deficiencies in the available data, we are, in general, unable to estimate
these attractiveness functions with certainty. Therefore, the deter-
ministic choice criterion will not always correctly predict the actual
choices. We can express the attractiveness of an alternative as consist-
ing of two parts, observable and random, as follows:
Uit uit + #it
(1)
where uit is the observable part of the attractiveness function and
eit is an observable random element. If the attractiveness measures of
the alternatives include random components, then only the probability of
choosing each option can be predicted, rather than a deterministic choice.
The probability that alternative I will be selected by shipper t
equals the probability that the attractiveness (cost) of alternative I
is greater or smaller than, or equal to, the attractiveness (cost) of all
other alternatives that are available. Formally, this statement can be
expressed as follows:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
P(i:At) - Prob [Uit >-U jt9 I e At]
(2)
where P(i:At) is the probability of shipper t selecting alternative
I from his choice set At. Substituting equation (1) in (2) we get:
P(i:At) - Prob [cjt - 'it < uit uit, i c At]
(3)
This expression implies that the joint probability distribution of the
random components determines the form of the model that relates the
systematic attractiveness functions to the choice probabilities.
One specific assumption about the random elements joint distribu-
tion leads to the multinomial logit model which is the only probabil-
istic choice model that has been extensively applied to multiple choice
problems. The random elements are assumed to be independently and
identically distributed as follows:
-w
P(E- z
W
N M
O
I[)
u')
N
S3 AOW 30 %
a
0
In
U)
N
S3A0W J0 %
0
co
W -t
0 ? uI
: z
0O a W
a
0 M
U) N
S3AO 4 J0 %
n
5
0
Qf
10
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
O
? o
L I I I
o 0 0 0
0 0 0 0
J.NIOd 83OUO38
Approved For Release 2001/11/19 : CIA-RDP79-00798AO00200020005-9
1 -L----J-
0 0 0
o o 0
a
0 w
0
1-
o z
ar
C;
z
U-
0
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
w
5i
to
I
1
L
iO
O
O.
o
0
ON
o
0
O
0
0
0
(0
to
cr
N
.I.NIOd U3C18OJ8
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
I I I I I
0 0 0 0 0
0 rn OD fl.- to
%-AVO -2
1
0
to
. 88
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
high mean transit times tend to be the most unreliable. Perhaps more
meaningful from a rail operations perspective, it appears that the number
of terminals at which a car is handled is the major determinant of perfor-
mance. For example, consider figure 6. aLK5 is a measure of variation in
transit time. Those moves with zero intermediate yardings exhibit slight
variation and hence tend to be reliable. Those with two intermediate yards
are substantially less reliable. On the basis that moves with more yardings
are the long distance moves in a system, one might argue that distance is
the cause of the unreliability. Some evidence that these reliability
differences are caused by the number of intermediate yards rather than
the distance is shown in figure 7.
That this relationship between terminals and origin-destination
service exists is not surprising when one considers rail operations. It
is the nature of railroad operations that a car encounters numerous oppor-
tunities for delay as it moves from its origin to its final destination.
At each yard, cars moving to common, intermediate, or final destinations
are consolidated into "blocks," placed in a train consisting of one or
more blocks, and handled together to another yard which may be twenty or
more than a thousand miles distant. Whenever a car is set off from a
train or the train reaches its destination, the car is reswitched and
consolidated with other traffic into a new block and a new train. This
procedure is repeated until the car reaches its final destination.
This process of switching and consolidation necessarily results in
longer transit times than would be required for direct movement (such as
by unit train). Equally as important, this process is unreliable. That
is, each time a car is switched, the potential for a missed connection at
that yard exists.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
N N
I 0
co D
~A =
w
..+N m
I-- --~b
D
S`lnVH 3NIl*'1b101 30%
0 0
S-1nVH 3NI1 1VIOI 30 %
0 0
0~0 VQ' N
S1nVH 3NI-I -1V.101. 30 %
H
'Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/1.1/19 : CIA-RDP79-00798A000200020005-9
N
N
W
W
J
O
~
Q
0
O
}
0
}
a
N
N
A
V
NO
N
I
0 0
SWVd 0-0 dO %
r-I
0 0
V N
SWVd 0-0 d0 %
0
0
`- SUM 0-0 dO %
U)
cc
N :
0
O
H
0 0
SJIVd 0-0 d0 %
0
N L) =
N I- ~
LL-
O
b U
I
qT 0
S?J IVd 0-0 d0 %
0
0
SdIVd 0-0 d0 %
Approved For Release 2001/11/19: CIA-RDP79-00798A000200020005-9
W W
F- Q
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Missed connections are critical in that they often lead to car delays
in the order of twelve-twenty-four hours (the time until the next appro-
priate outbound train), large variations in transit time and, hence,
unreliable performance. Table 1 shows the magnitude of transit time delay
and transit time variance as a function of the probability of a missed
connection at a yard. These probabilities are quite realistic in the light
of the various analyses of railroad operating data.6
AVERAGE DELAY TIME AND STANDARD DEVIATION OF DELAY TIME AS A FUNCTION
OF THE PROBABILITY OF A MISSED CONNECTION
Probability of Average Standard Deviation of Transit
Missing Connection Delay (Hours) Times (Hours
Among the causes for missed connections are outbound train cancel-
lations, train length/weight constraints, RIPS, NO-BILLS, and late arrival
of an inbound car. This last cause has been shown to be an important one.
Specifically, if the car arrives later than some "threshold" time, its
connection with an outbound is often missed. Of course, the outbound train
could be held for the car, allowing the connection to be made despite the
lateness of the arrival. However, this may well lead to further problems.
Specifically, Belovarac and Rneafsey7 have shown that the primary cause of
late arrivals at a yard is late departure from the preceding yard. Hence,
holding trains to allow particular connections to be made may well lead to
inbound lateness at succeeding yards and the possibilities of other missed
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
connections. Indeed, Folk8 had examined "no-hold" versus "hold" policies
with respect to overall network effects on car performance.
All in all, the question of yard performance and missed connections
is a complex one, with the various components and operating policies of
the rail network heavily interacting to affect performance.
These research results indicated that potential for improvement in
service existed through operating strategies that would either avoid inter-
mediate yardings of cars (i.e., run-through trains) or improve the proba-
bility of making connections when yardings were necessary (e.g., by opti-
mizing inbound/outbound train connections or by limiting cancellations).
A case study was performed in cooperation with the Southern Railway9 and
various experimental alternatives designed to improve service were imple-
mented. These implementations did lead to service improvements in the
Southern's system and have been integrated into the Southern's continuing
operating plan.
IV RESEARCH NEEDS
With the previous as background, this paper goes on to discuss
research needs of the rail industry in the area of service and service
reliability.
The research needs that exist are subdivided into two major areas
as follows :
1. The shipper's perspective
2. The supply of transportation service
IV-A THE SHIPPER'S PERSPECTIVE
The area can be further subdivided into the shipper's demand for
transportation service and the shipper's operating response to rail
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
service.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
IV-A-1 DEMAND
From the viewpoint of the author, the major gap in rail service
research (and, in fact, in freight service in general) is on the demand
aide of the question. Specifically, we have only the most preliminary
and intuitive ideas about the relationships between level of service
provided by the industry and the volume and traffic mix attracted by the
industry. There is a great shortage of knowledge about what service the
shipper community actually wants and needs and, more importantly, how the
shipper will make his model choice decision as a function of the service
provided. Making major investment decisions or operating decisions in
the absence of demand models (i.e., models for understanding the volume
complications of service decisions) seems quite difficult. Yet, this
seems to be what the industry does (although specific instances of demand
sensitivity or insensitivity have been documented).
Thus, a major research priority should be the development of a level-
of-service and commodity-sensitive demand model for use as a planning tool
by the rail industry in making investment decisions. Of potential major
benefit in this endeavor may be newly developed, disaggregate, demand-
modelling methodology. This methodology has the potential for overcoming
what has long been the major roadblock in the development of such models;
namely, the lack of an adequate data base.10
We should note that there is by no means complete agreement in the
industry on what constitutes service and how service relates to demand.
For example, in informal discussions, various rail officials have indi-
cated that the shipper does not see himself as "buying" reliability from
the industry and that modest changes in volume (if any) would occur were
service reliability to be dramatically improved. By the same token, other
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
rail officials have indicated that service improvements can lead to sub-
stantial attraction of new business. In fact, neither contention has been
supported by a comprehensive set of data, models, and analysis.
In addition to the obvious needs for data, other preconditions exist
for success in this effort. Perhaps most important is the market's per-
ception of what rail service is. In order to develop consistent and usable
demand models, performance measures that are useful from the shipper's
perception must be developed. Of course, these measures must exist (or be
capturable) in the data base used in model calibration.
In summary, to quote from existing work:
If a railroad does not know how service levels affect
traffic volumes, then that railroad does not know what
product it should produce.11
Such a modelling effort will allow the industry to better understand
this issue.
IV-A-2 SHIPPER'S OPERATING RESPONSE TO RAIL SERVICE
The previous section focussed on the attraction (or lack thereof) of
traffic as a function of service. A second shipper-related area is
composed of the operating behavior of a shipper who is, in fact, using
rail. and how this behavior may be related to the quality of rail service.
Shipper behavior and procedures are an important component in the
utilization of the rail industry's resources. For example, consider car
detention. A not-insignificant portion of the car cycle is time-spent
with the car in shipper control.12 An empty car may wait to be loaded
for some time and a loaded car may be used as a rolling or stationary
warehouse by the consignee. Often demurrage is not an adequate incentive
for the shipper to load or unload the cars.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
96
The working hypothesis here is that this shipper behavior is closely
related to the service provided by the industry. For example, one
reason advanced by the shipper community for "excess" car detention on
loading is that poor car-availability leads to very conservative car-
ordering policy. This, in turn, may lead to empty care waiting for long
time periods at the shipper's siding if they are,in fact, delivered in a
timely fashion. Unrealiable delivery of empties or spotting of loads may
also lead to "erratic" shipper behavior because the shipper has difficulty
in scheduling his crews around the undependable service provided by the
industry.
The above suggests that improvements in service (and service relia-
bility in particular) may pay extra dividends in terms of an improvement
in the car detention situation. That is, if the rail industry provides
more reliable service, equipment utilization can be improved because of
better shipper performance which can, in turn, lead (through better avail-
ability of equipment) to still better rail services.
Another related area is the regularization of shipper demands on the
rail system. The irregularity of shipper demands causes poor allocation
of resources as the rail system attempts to respond to peak demands for
service (difficult to predict). This poor distribution of resources
leads in turn to degradation of service on the network taken as a whole.
Regularization of shipper demands can potentially lead to better service
for all shippers. Again, a feedback phenomenon exists. Possibly, if
service reliability can be improved, shippers can be enticed to regularize
their inputs to the system, which will allow the rail industry to improve
service still further.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The above are but two examples of the interaction of the shipper
with the rail system. In all, there appears to be a great deal we do
not know about shipper behavior patterns and how these relate to rail
service. A research effort in this area can yield very meaningful results
from the viewpoint of improved service and shipper satisfaction with rail
service.
IV-B THE SUPPLY OF TRANSPORTATION SERVICE.
The improvement of the transportation service provided by the rail
industry can take place on three fronts. These are:
1. Rail operations
2. Capital Expenditures on Rail Facilities
3. Institutional Changes
To put these needs in perspective, it's important to note that the
previously described M.I.T. research, virtually all of which is supply-
oriented, has merely scratched the surface of the field. This research has
1. Described some service measures
2. Isolated some of the root causes of rail service unreliability
3. Developed various models that can be useful in improving rail
service
4. Demonstrated the usefulness of these models in one particular
operating theatre
This research has had major limitations. It has focussed on the
operating theatre only and, further, (in its implementation) has focussed
on local and individual changes to operations and not on global changes
to network operations.
In short, there is a great deal to do in research in this area. The
remainder of this section goes on to describe these research needs on the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
supply side.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
IV-B-1 RAIL OPERATIONS
It is clear that changes to rail operations can cause major impacts
on rail service. What is less clear is how best to proceed and how much
these changes are likely to cost (or save). Three areas are discussed,
these being:
1. Impacts of Network Wide Operating Changes
2. Data Systems
3. Cost Models
Impacts of Network Operating Changes
The previous research has shown that local changes in rail operations
(e.g., changing a single inbound/outbound train connection) can lead to
the desired service effect. What needs to be better understood is the
benefits (from a service perspective) of wholesale, network-wide changes
to rail operations. Models for the prediction of the benefits of imple-
mentation of:
1. Blocking policy changes
2. Scheduling policy changes
3. Train length policy
4. Through trains
carried out on a network basis are badly needed. It should be noted that
a wide variety of models that begin to address these issues are available.
Network simulation models have been developed by many railroads and by the
industry through the A.A.R.13 These models allow an analyst to simulate
rail operations at very fine levels of detail and can, in theory, be very
useful in the examination of such issues as those noted above.
However, it is the author's impression that these models have not
been partiA t dwP&X R 449gbOW1 A0&e014MD OONS? OOf 5-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
questions. Why this is true (if indeed it is) is not clear. While these
models have sometimes been criticized for being too expensive or difficult
to use, it seems obvious that the costs of running the models will be
trivial compared with the costs of implementing some operating change in
the rail system. Given this situation, it is recommended-that the micro-
simulation approach to rail network modelling be reexamined afresh to
ascertain why efforts along these lines have not been particularly fruitful
to date in the rail industry (unlike the experience in other industries
such as air transportation, manufacturing systems, etc.). If, in fact,
conceptual problems with the microscopic approach can be isolated, further
research on correcting these deficiencies is indicated.
In parallel with this reexamination of the traditional approach to
rail systems modelling, some new approaches are appropriate as well. For
example, less detailed network modelling approaches have proven fruitful
overseas. In particular, ROUTESTRAT,14 developed by British Railways, has
been very useful in scheduling and blocking the BNR system. This model,
which simulates the operation of the systems at a rather macroscopic level
works toward optimizing (rather than simply simulating) the rail system.
Granting at the outset the major differences between the U.S. system and
the British network (e.g., size, one-owner), it would still appear fruitful
to consider the usefulness of such a modelling approach in rationalizing
rail operations in this country.
Data Systems
This discussion of research needs in rail operation should go no
further without alluding to the question of data systems. This is the topic
of several other presentations so we need not dwell on it here. However,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
as a precondition to improving service, the industry needs to measure what
it is doing in rail operations and how well it is performing. Data systems
that will allow the industry to perform such measurements are essential.
There has been a good deal of progress in this area over the past several
years. Among the major efforts are the work in terminal management systems
performed under the auspices of the Labor-Management Committee1s as well as
Southern's TPA,16 Southern Pacific TOPS,17 and other network control
programs. While certainly the industry is in a much better position to
measure its service than it was even a short time ago, further refinement
of rail data systems is needed. In particular, for our service perspective,
the need to measure performance on interline moves is clear. Also of major
importance is capturing the portion of the move from shipper's dock to
originating yard and from terminating yard to consignee's dock. These
local moves are missing from many data systems and yet appear to have a
major impact on service reliability.
The above systems reflect the ability of computer and information
system'technology to measure performance with an eye to isolating problem
areas which can then be improved. This is but the first step in the use
of computers, however. Research into utilizing the computer to allow the
industry-to operate in ways inconceivable without the information processing
and real time capabilities of the machine is needed. The author believes
that the potential of this technology has barely been touched and that
service and operating improvements of major impact are within the industry's
grasp using this tool.
Cost Models
A final subject of proposed research in the rail operations area is
twop the relatiApproved he- For elease 2s001/1 T/19 : CIA-RDP79 00798A00D200020005 9e
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
industry needs better ways of understanding the cost implications of
various operating alternatives and levels of service. The relationships
here are not at all well understood.18
One fascinating aspect of this cost question is the notion that im-
proved levels of service (and, in particular, improvements in'mean transit
time and service reliability) can yield better redistribution of empty
cars and deadheading power, less inventorying of empties, better shipper
car-detention performance, a shorter car cycle, and therefore, better equip-
ment utilization. Hence, under certain circumstances, one may actually be
able to lower costs while providing better service. The relationships
described above are shown pictorially in figure B.
The development of useful cost models is an industry-wide problem.
What is emphasized here is that such models be designed in such a manner
that the implications of operating changes (and service improvements) can
be readily computed.
These cost models, taken together with the service-sensitive demand
models described earlier, would give the railroad analyst the ability to
evaluate fully the implications of service quality modification from the
profitability viewpoint. The relation among the various models is described
pictorially in figure 9.
IV-B-2 CAPITAL EXPENDITURES ON RAIL FACILITIES
The potential changes in rail operations described in the previous
section are characterized by being implementable within a short period of
time. The implicit assumption is that these changes are being implemented
within the constraints of existing resources. The area under discussion
in this section relates to somewhat longer-term changes. These are changes
in the physical facilities available to the rail-operating department,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Strategies for Im.Oroving
Reliability
/
Direct
` Indirect
Impacts
Impac
t
/
Reduced
Trip
Times
Improved
Distribution
Of Empties
Increased
Traffic
Levels
.1 1
Shorter
Load/Unload
Times
Improved Freight
Car Utilization
Increased
Operating
Efficiency
Reduced Need Ability to
For New Cars Move More Traffic
Capital Costs Ope 1 rating
Revenues
and Costs
PROFITABILITY
FIGURE 8 A MODELLING FRAMEWORK
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 103
Operating
Changes
Network / Cost
Models
Service
Changes
Demand
Models
Volume and Traffic
Mix Implications
Shipper Operating
Behavior Models
Cost
Implications
Rail Industry
Profitability
FIGURE 9 IMPLICATIONS OF SERVICE QUALITY MODIFICATION
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
104
obtained through a capital investment program. The research area is the
development of models for effective investment planning and for providing
a useful framework for rail investment planning.
The need is for models that will allow the analyst to predict the
impact of particular investment alternatives on service quality and operating
costs and hence to he able to make an informed choice among investment
alternatives. However, these individual impact prediction models are but a
first step. The inherent difficulty in developing such models is compounded
by the environment in which investment decisions among alternatives is being
made.
1. The investment alternatives are not independent--there are many
different investments under consideration at any Point in time. Yard con-
struction and upgrading, rolling stock of different types, power, electrifi-
cation, new track, and structure are but several of the investment possi-
bilities available. A new level of difficulty is introduced when one recog-
nises that the impact of various alternatives on service and cost are not
independent. For example, the impact of the purchase of new power units is
related to whether or not a particular terminal is upgraded. How these
impacts interact is a difficult modelling issue, as in the complexity intro-
duced by the fact that the combinations of investments can grow to be
extremely large.
2. There is a great deal of uncertainty in the system. The railroad
investment planner faces the classic case of decision-making under un-
certainty. Among the major stochastic elements in the process are:
A. Traffic levels
B. Levels of investment in the future
C. Competitive and regulatory elements
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
D. Uncertainty in the prediction abilities of the impact models themselves.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
In all, the railroad investment problem is a topic worthy of sub-
stantial research. In addition'to the development of a variety of impact
models (the most notable needs being in the area of yard design and capacity
and placement of terminals in the network), a framework for rail investment
planning under uncertainty is needed. This framework could usefully build
upon the work of Pecknold19 which deals with multistage investment planning
in highway systems.
In summary, investment decisions should be made with service quality
in mind. A basic research need is the development of a framework for the
understanding of the impacts on service quality of various investment
decisions taken in various combinations under conditions of uncertainty
with respect to future traffic and investment levels.
IV-B-3 INSTITUTIONAL CHANGES
Operating changes are implementable in short order. Changes derived
from capital expenditures take somewhat longer. Still longer term in nature
are various institutional changes. These are changes in the basic frame-
work within which the industry operates. By their very nature, insti-
tutions were constructed in such a way as to strike a proper balance between
various interest groups and any attempt to change (at a particular point in
time) "the system" is likely to appear to some group as an attempt to take
away some hard-won gains.
Yet, while implementation difficulties are inherent in the-process,
research into the impacts of proposed changes is appropriate and useful.
If the impacts on all actors (industry, labor, shippers, government, etc.)
can be understood, trade-offs can be established and compromises reached.
There are any number of institutional issues one could address. For
the purpose here, only those with quite direct impact on service and
Apps Fr*11Q@WJ4dti0' 3iM ."1? $ DP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
106
Work Rules
The relation between operating policy and service quality is clear.
Likewise, the relation between operating policies and current operating
work rules is strong. A better understanding of how work rules affect
operating policies (particularly train length20) and how relaxation of
particular rules can lead to service improvements is needed.
Network Shape Issues
How the rail network is physically configured and how resources are
allocated to this network are closely related to service and service
quality. At the same time, network shape is institutionally bound up
with the regulatory climate, specifically the I.C.C. Network changes
occur very slowly (e.g., through abandonment and mergers), mostly as a
result of regulation. A more precise understanding of how service
quality relates to network configuration would be useful in illuminating
the trade-offs inherent in any network change alternative.
Fractional Per Diem
The rail industry operates on a 24-hour per diem system, with a
midnight cut-off. Of current research interest is the possibility of
implementing a fractional per diem system. Both 8-hour and hourly
systems have been suggested. An understanding of the service implications
of these alternatives is required.
Organization Structure in Railroad Companies
The typical railroad company in the U.S. has not changed its organi-
zation form for decades. By no means is it clear that the railroad company
is optimally structured to address the service needs of its shippers. The
organization form of some companies seems more suited to the basically
noncompe
titly
1~~I~tp9-~07a.1'82~3at
MPPI VV
clearly no longer exists.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 107
A fresh look at how a rail freight transportation company should be
structured to be able to address 'today's issues and the issues of the future
is needed. Such questions as:
1. How are investment decisions reached?
2. What are the basic information flows in the company and are they
the right ones?
3. What is the proper training for a railroad executive?
4. Is the traditional departmental structure of the industry
operating, transportation, traffic, etc.) the optimal one?
5. How are level-of-service versus cost trade-offs made? Who
makes them?
Much work has been done on organizational structure.21 Some attempt
to utilize these efforts in the rail industry is appropriate.
V SUMMARY
Research needs in the area of service reliability have been described.
Two major areas have been discussed. These are the shipper's perspective
and the supply of transportation service. Within the first of.these is the
area of highest priority in the author's view, namely, the development of
a better and more quantitative understanding of the relation between service
quality and the demand for rail freight service. Also noted here as a
lower priority was the issue of the shipper operating responses to rail
service and how these are related to equipment utilization and, hence, to
rail costs and service.
The supply of transportation service was subdivided into the areas
of rail operations, capital expenditures, and institutional changes.
setting of priorities here is difficult in that the areas are interrelated
and research costs are likely to be very different in size. However, the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
author is inclined to set priorities as follows:
Impacts of Network-Wide Operating Changes on Service
(IV-B-1 p. 98)
Impacts of Various Capital Investment Alternatives on Service
(IV-B-2 p. 101)
Data Systems
(IV-B-1 p. 99)
Overall Investment Framework
(IV-B-2 p. 101)
Cost Models
(IV-B-1 p. 100)
Various Institutional Changes
(IV-B-3 n. 105)
Hourly Per Diem
Work Rules
Network Shape
Organization Structure
A final comment. Service and service reliability is without question
an important issue in the rail industry. It is also a researchable question.
However, research is not enough. It can only highlight the needs and
suggest industry strategies. A continuing and disciplined program of imple-
mentation, experimentation, and innovation within the rail industry is re-
quired if this research is not to produce a set of dust catchers. By the
same token, the research community should discipline itself to produce
intellectually sound but pragmatic, implementable, research results.
Based on recent experiences with the interaction of the rail industry
and the research community, the author has confidence that a great deal can
be accomplished in the future in this area.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
BIBLIOGRAPHY
109
America's Sound Transportation Review Organization. The American Railroad
Industry Today: A Prospectus. Washington, D. C., June 1970.
Association of American Railroads. Yearbook of Railroad Facts. Washington,
D.C. 1971.
Belovarac, K. and J. S. Kneafsey. Determinants of Line Haul Reliability.
Studies in Railroad Operations and Economics, vol. 3. Research Report
R72-38, 1972.
Ben-Akiva, Moshe. "Structure of Passenger Travel Demand Models." Ph. D.
dissertation, M.I.T., June, 1973.
Marc Terziev and Paul Roberts. "Freight Demand Modelling A
Policy Sensitive Approach." Center for Transportation Studies, M.I.T.,
presented to the 47th National Operations Research Society of America
meeting, Chicago, Illinois, April, 1975.
Chandler, A. D., Jr. Strategy and Structure. M.I.T. Press, 1962.
Charles River Associates. The Role of Factor Prices in Transportation
Planning for Developing Countries. Study for the Office of Inter-
national Affairs, U. S. Department of Transportation.
Folk, J. F. "Models for Investigating the Unreliability of Freight Shipments
by Rail." Ph.D. dissertation, M.I.T. Operations Research Center, 1972
(unpublished).
Models for Investigating Rail Trip Time Reliability. Studies
in Railroad Operations and Research, vol. 5, Research Report R72-40, 1972.
Some Analyses of Railroad Data. Studies in Railroad Operations
and Economics, vol. 6. M.I.T. Research Report R72-41, 1972.
A Brief Review of Various Network Models. Studies in Railroad
Operations and Economics, vol. 7. M.T.T. Research Report R72-42.
Hartwig, James and William Linton. Dis aggregate Mode Choice Models of
Intercity Freight Movement. M.S. Thesis, Northwestern University, Depart-
ment of Civil Engineering, June, 1974.
Jennings, A. A. "The Effect of Train Length on the Reliability of Operation
of Freight Yards." S.M. thesis. M.I.T. Department of Civil Engineering,
1972 (unpublished).
Kneafsey, James T. Costing in Railroad Operations: A Proposed Methodology.
Studies in Railroad Operations and Economics, vol. 13. M.I.T. Research
Report R75-15.
Kolsen, H. M. The Economics and Control of Road Rail Competition. Sydney,
Australia: Sydney University Press, 1968.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Kullman, Brian C. "Choice of Mode Between Rail and Truck in the Intercity
Freight Market." Ph.D. dissertation, M.I.T. Department of Civil Engi-
neering, 1972 (unpublished).
Lang, A. S. and R. M. Reid. Railroad Car Movement Reliability: A Prelimi-
nary Study of Line Haul Operations. Studies in Railroad Operations and
Economics, vol. 1. M.I.T. Research Report R70-74.
and C. D. Martland. Reliability in Railroad Operations and
Economics. Studies in Railroad Operations and Economics, vol. 8. M.I.T.
Research Report R72-74, 1972.
Martland, Carl D. Rail Trip Time Reliability: Evaluation of Performance
Measures and Anaylsis of Trip Time Data. Studies in Railroad Operations
and Economics, vol. 2. M.I.T. Research Report R72-37, 1972.
Improving Railroad Reliability: A Case Study of the Southern
Railway, Studies in Railroad Operations and Economics, vol. 10. M.I.T.
Research Report R74-28.
Procedures for Improving Railroad Reliability, Studies in
Railroad Operations and Economics, vol. 12. M.I.T. Research Report
R74-30.
"Origin to Destination Unreliability in Rail Freight Trans-
portation," Engineer's thesis. X.I.T. Department of Civil Engineering,
1972 (unpublished).
Midwest Research Institute. The AAR Network Simulation System, program
documentation for the network simulation model developed for the Associ-
ation of American Railroads, 1971.
Moody's Investors Corporation. Moody's Transportation Manual. New York: Moody
Publishing Co., 1971.
O'Doherty, J. D. "Classification Yard Effects in Rail Freight Movement
Reliability." Civil Engineer's thesis, M.I.T.,1972 (unpublished).
Pecknoid, Wayne M. "Evolution of Transport Systems: An Analysis of Time
Staged Investment Strategies Under Uncertainty." Ph.D. thesis, M.I.T., 1970.
Project Team of Labor/Management Task Force on Terminals. A Program of
Experiments Involving Changes in Terminal Operations on the St. Louis
Terminal Division Missouri Pacific Railroad 1974 Progress Report.
Railroad Systems and Management Association. The Measure of RAilroad
Freight Service. Chicago, Illinois, 1972.
Reebie Associates. "Toward an Effective Demurrage System." Final Report
to the Federal Railroad Administration, July, 1972.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Reid, R. M., J. D. O'Doherty, J. M. Sussman, and A. S. Lang. The Impact of
Classification Yard Performance on Rail Trip Time Reliability. Studies
in Railroad Operations and Economics, vol. 4. M.I.T. Research Report
R72-39, 1972.
Roberts, P. 0., Jr. "The Logistics Management Process as a Model of Freight.
Traffic Demand." Harvard Business School Working Paper, HBS71-11,
April, 1971
, and D. T. Kresge. Techniques of Transport Planning, vol. It.
The Brookings Institution, 1972.
ROUTESTRAT. British Railways, 1973.
Sussman, Joseph M. "A Systems Model of the U.S. RAilroad Industry." pre-
sented at the ORSA/TIMS meeting, Chicago, April, 1975.
and Carl D. Martland, Reliability in Railroad Operations: Execu-
tive Summary. Studies in Railroad Operations and Economics, vol. 9. M.I.T.
Research Report R73-4, 1972.
Improving Railroad Reliability: A Case
Study of the Southern Railway, Executive Summary. Studies in Railroad
Operations and Economics, col. 11. M.I.T. Research Report R-74-29.
Southern Pacific Railway. Total Operations Processing System (TOPS).
Southern Railway. Time Performance Analyzer Manual, 1971.
Transportation Association of America. Transportation Facts and Trends.
Washington, D.C., 1971.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
112
1. J. Sussman, "A Systems Model of the U. S. Railroad Industry" presented
at the ORSA/TIMS National meeting, Chicago, May 1975.
2. Joseph Sussman, Carl D. Martland, and A. Schaffer Lang, Reliability in
Railroad Operations: Executive Summary, Studies in Railroad Operations
and Economics, vol. 9, M.I.T. Report R73-4, 1972.
3. Carl D. Martland, Rail Trip Time Reliability: Evaluation of Performance
Measures and Analysis of Trip Time Data, Studies in Railroad Operations
and Economics, vol. 2, M.I.T. Report R72-37, 1972.
4. A. Scheffer Lang and Carl D. Martland, Reliability in Railroad Operations,
Studies in Railroad Operations and Economics, vol. 8, M.I.T. Report
R72-74, 1972.
2
5. a LH is the variance of transit time.
6. Robert M. Reid, John D. O'Doherty, Joseph M. Sussman, and A. Scheffer
Lang, The Impact of Classification Yard Performance on Rail Trip Time
Reliability, Studies in Railroad Operations and Economics, vol. 4, M.I.T.
Report No. R72-39, June, 1972.
7. K. Belovarac and J. T. Kneafsey, Determinants of Line Haul Reliability,
Studies in Railroad Operations and Economics, vol. 3, M.I.T. Report No.
R-72-38, June, 1972.
8. J. F. Folk, Models of Investigation Rail Trip Time Reliability, Studies
in Railroad Operations and Economics, vol. 5, M.I.T. Report No.-R72-40,
June, 1972.
9. Joseph M. Sussman and Carl D. Martland, Improving Railroad Reliability-
A Case Study of the Southern Railway, Studies in Railroad Operations
and Economics, M.I.T. Report 74-29, 1974.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
10. Moshe Ben-Akiva, Marc Terziev and Paul Roberts, "Freight Demand Model-
ling: A Policy Sensitive Approach," Center for Transportation Studies,
M.I.T. presented to the 47th National Operations Research Society of
America Meeting, Chicago, Illinois, April, 1975; James Hartwig and
William Linton, "Disaggregate Mode Choice Models of Intercity Freight
Movement," M.S. thesis, Northwestern University, June, 1974; and Moshe
Ben-Akiva, "Structure of Passenger Travel Demand Models," Ph.D. disser-
tation, M.I.T., June, 1973.
11. Carl D. Martland, Procedures for Improving Railroad Reliability, Studies
in Railroad Operations and Economics, vol. 12, M.I.T. Report R74-30.
12. Reebie Associates, Toward an Effective Demurrage System, Report
FRA-OE-73-1, National Technical Information Service PB 212-069.
13. Midwest Research Institute. The AAR Network Simulation System, program
documentation for the network simulation model developed for the Associ-
ation of American Railroads, 1971.
14. ROUTESTRAT. British Railways, 1973.
15. Project Team of Labor/Management Task Force on Terminals, A _Program of
Experiments Involving Changes in Terminal Operations on the St. Louis
Terminal Division Missouri Pacific Railroad, 1974 Progress Report.
16. Time Performance Analyzer Manual, Southern Railway, 1971.
17. Total Operations Processing System (TOPS), Southern Pacific Transpor-
tation Co.
18. J. T. Kneafsey, Costing in Railroad Operations: A Proposed Methodology,
Studies in Railroad Operations and Economics, vol. 13, Research Report
R75-15, 1975.
19. Wayne M. Pecknold, "Evolution of Transport Systems: An Analysis of
Time Staged Investment Strategies Under Uncertainty," Ph.D. disser-
to +vet?IIoTReIWg?2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
20. A. Scheffer Lang and Carl D. Martland, op. cit.
21. Alfred D. Chandler, Jr., Strategy and Structure, M.I.T. Press,
1962.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
CURRENT APPROACHES TO TRAVEL DEMAND FORECASTING*
**
Marvin L. Manheim
1. Introduction
One purpose of this paper is to present a brief overview of the
major methods now in use in the U.S. for forecasting future passen-
ger travel in urban areas. Many of these methods have been applied
also to forecasting intercity and national passenger travel and
to predicting freight movements as well.
In the next section, we summarize some of the issues currently
being addressed in urban transportation planning studies in the U.S.
Then we review briefly the basic theory of transportation systems
analysis underlying travel forecasting methods. In the fourth sec-
tion, we summarize the major alternative approaches available today
and show their interrelations in section 5. Examples of two major
approaches are given in section 6. The last section presents our
conclusions.
2. The Role of Travel Forecasting Models in the Analysis of Tranpor-
tation Plans and Policies
Now, more than ever before, a multiplicity of urban and regional
objectives is leading to the proposal of a wide range of policies to
control and shape travel patterns. The number, range, and complexity
of these proposed policies is unprecedented. In order to appraise the
*
This paper draws upon numerous articles and reports to which
many other individuals have contributed. Where specific acknowledge-
ment is practical, appropriate references have been cited. The author
particularly wishes to acknowledge the contributions of several indi-
viduals whose individual and collaborative work, and words, are infused
throughout this paper, Moshe Ben-Akiva, William Jessiman, Steven Lerman,
Wayne Pecknold, and Earl Ruiter, who could almost be listed as co-authors.
**
Appro F6? ea1J 01/11/19: CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 116
advantages and disadvantages of these proposals, it is essential to
be able to predict how travel patterns will be affected.
Among the issues facing urban transportation planners are
these:
1. air quality
2. energy
3. land use policy
4. urban congestion
5. transportation impacts on special groups, such as: elderly,
handicapped, low-income, etc.
6. public transportation financing and new technology develop-
ment
In response to these issues, a wide variety of transportation
,options are being considered. Some of these provide new auto or
transit services, while others are disincentives as, for example,
steps to reduce auto use:
1. new highway facilities
2. new transit service offerings such as rapid transit ex-
tensions, feeder transit service coordination, express bus,
subscription bus, jitney, dial-a-ride, taxi, dual-mode,
light rail transit, etc.
3. taxes on gasoline
4. gasoline rationing
5. gasoline allocation schemes
6. parking restrictions, bans, or surcharges
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
7. vehicle exclusion policies--areas, days of the week, time
of the day, registration sticker schemes
8. time bans on driving on particular days
9. car pooling programs and incentives
10. restrictions on auto ownership
11. taxes on vehicle ownership
12. congestion tolls
13. staggered work hours
14. parking changes
15. changes in transit fares and subsidies
16. ramp metering and other improvements to highway
facilities
17. vehicle emission controls
18. land use controls
In developing a regional or subregional transportation plan
for a particular urban area, policy components such as these are
combined into overall programs.
The challenge of travel forecasting methods is to be able to
predict the responses of potential travellers to such elements.
For example, consider these questions on the impacts of various
policies designed to reduce air pollution from transportation: How
will trip-makers respond to a forced reduction in employee parking
at factories and offices? How will they respond to an early
morning on-street parking ban, such as the one scheduled for Boston
in the spring of next year? Will the result be a decrease in
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 118
trip-making? A change of destination? A shift to public transpor-
tation, to car pools, or to late arrival (i.e., staggered work
hours for employees)? What would the imposition of selective
parking surcharges on existing fees do to demand by the various
modes? What is the impact of the imposition of tolls on certain.
facilities? Would prohibitions on entry by private autos into
certain areas have a similar effect? Do these policies have un-
desirable differential impacts on different segments of society?
What are the short-term and long-term shifts in residence, loca-
tion, choice of work place, or household automobile ownership level
that will occur as a result of various such air pollution control
strategies?
Questions such as these must be answered before a rational
policy for reducing air pollution can be selected. Understanding
travel demand behavior plays a critical role in arriving at these
answers.
3. The Theoretical Basis of Travel Forecasting
Modern travel forecasting methods are based on the theory of
transportation systems analysis. This theory has emerged from
several-sources. I In outline, the prdoblem of predicting the flows
in a transportation system is a simple application of economic
theory: the flows which will result from a particular transporta-
tion system (T) and pattern of socioeconomic activities (A) can be
determined by finding the resulting equilibrium in the transportation
market. If:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 119
V - volume of flow
L - level of service experienced by that volume
F - (V,L) - flow pattern
Then, we find equilibrium by establishing a supply function (S)
and demand function (D), and solving for the equilibrium flows
(F ) consistent with both relations:2
(1.1) L - S(V,T)
V - D(L,A)
Fo - (V0,L0)
(See figure 1, Basic Theory)
While simple in outline, the application of the theory becomes
complex in practice for several reasons:
1. The consumer considers many service attributes of the
transport system when making a choice (e.g., line-haul
travel time, transfer time, walk distance, out-of-pocket
cost, privacy, etc.), and thus, L must be a vector with
many components
2. Determining the demand functions (as well as other elements)
to use is difficult
3. The equilibrium occurs in a network, where flows from
many origins to many different destinations interact, com-
peting for the capacity of the network; and the form of*
these interactions is affected by the topology of the net-
work
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A0002000f?005-9
VO
FIGURE 1,. BASIC THEORY
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
VOLUNE
V
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Thus, fairly elaborate computational schemes are required to ac-
tually determine the equilibrium flows F 0 for a particular (T,A).
In the case of a multimodal network, the symbol V represents an
array of volumes:
(1.2) V - {Vkdmr}, for every k,d,m and r,
where Vkdmr is the volume flowing from origin zone k to destination
zone d via mode m and path r of that mode, and the brackets { }
indicate a set of elements Vkdmr' Ideally, once we have established
our demand and supply functions, we would then like to be able to
turn directly to an equilibrium-calculating procedure to "solve" the e
two sets of relationships to find the equilibrium flow pattern. The
result of this computation would-be the two arrays comprising that
flow pattern:
(1.3) Fo . (V0,Lp), where
V0 {Vkdmr} for every k,d,m, and r; and
Lo {Lkdmr} for every k,d,m, and r.
In words: we should get out of our equilibrium procedure the
volumes, and the levels of service experienced by those volumes,
from k to d by mode m and path r.
Unfortunately, at this state of the science of transportation
modelling, while several systems of transportation models exist,
there is not even one operational model which solves for these
equilibrium flows exactly and directly.3 There are available a
number of different systems of models, each of which represents a
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
different operational approach to. computing equilibrium in trans-
port. networks. The differences in these approaches are reflected
both in the computational algorithms and in the structure of the
demand models which are used.
One particular computational scheme is that which has been used
historically in urban transportation planning studies. In this ap-
proach the equilibrium flows are estimated in a sequence of steps,
commonly called trip generation, modal split, and traffic assign-
ment.4 Correspondingly, the demand function, D, is represented. as
a sequence of functions: trip generation (and attraction) equa-
tions, trip distribution procedures, modal split equations, and the
minimum-path rules of the traffic assignment procedures. We will
refer to this approach as the UTMS--the Urban Transportation Model
System.
More recently, other alternative approaches have been developed.
In the following sections of this paper, we will summarize the major
alternative approaches which now exist.
4. The Major Alternative Approaches
In the preceding sections, the basic concept of a demand func-
tion was introduced as a way of representing consumer behavior.
A wide variety of different types of demand models can be developed.
The objective of this section is to present some of the most impor-
tant distinctions among the types of demand models used in travel
forecasting in order to illustrate the variety of modelling choices
available.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved* F,oi Release 2001/11 /t.9 CI'A-RDP79?-00798A000200020005-9*
FIGURE
The Demand Function
VOLUME
V
LEVEL OF SERVICE TIME, t
Approved For Release 2001111/19 CIA-RDP79-00798AO.00200020005'9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
In this discussion, we will tend to emphasize primarily "struc-
tural" or "abstract" features of the-various choices. However, we
must never forget that the basis of everything we do in demand
modelling is based on trying to represent the behavior of the con-
sumer. To emphasize this, we will relate the discussion of modelling
choices to the way in which the choices open to the individual con-
sumer are assumed to be perceived and operated upon.
In general, in urban transportation, there is a wide range of
choices open to a traveller. These choices are among combinations
(f,d,m,r,h,ao, req, emp), where:
f - frequency of trips
d destination
m - mode
r - route
h - time of trip
so - number of autos owned
res - residential location
emp - employment location
For most of urban travel forecasting to date, it has been
assumed that (res) and (emp) are fixed for each traveller (or alter-
natively, an urban growth and land use model is used to predict these
long-run effects separately from the short-run travel decisions),
and so travel forecasts deal primarily with choices of (f,d,m,r,h,ao).
Further, in the past it has also been usual to separate choices
of time of trip and of auto ownership from the other choices, re-
ducing the primary focus of travel demand models to (f,d,m,r).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Looking at the behavior of groups of consumers, we are thus
concerned with predicting the aggregate total volume choosing a
particular combination, in this form:
Vkdmr - the volume of trips from origin zone k to destination d
by mode m and path r
Note that we have specified the origin of the trips (k), and
that frequency has been replaced by the total number of trips made
(i.e., frequencies of one, two, or more trips per individual per
time period).
There are a wide variety of different ways in which specific
demand models might be constructed for use in transportation
analyses. In this section, we will discuss some of the major differ-
ences in types of demand models: variables included, functional forms,
and three structural features which will be our primary concern:
aggregate versus disaggregate, probabilistic versus deterministic,
and simulteneous versus sequestial.
4.1 Variables Included
Recall that there are two basic sets of variables, activity
system variables (A) and level-of-service variables (L) in the demand
function (1.1). One important set of choices concern what activity
system and level-of-service variables to use. The activity system
may be described in terms of such variables as population, employ-
ment, income, household size, stage in family-life-cycle, etc. The
level of service of the transportation system could be described in
terms of travel time, separated into in vehicle and excess time,
time reliability, service schedule, out-of-pocket cost, perceived
security, tolls, etc.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
4.2 Functional Forms of the Models
A wide variety of alternative functional forms are possible.
The most common are the product, linear, exponential, and logistic
or logit forms.
4.3 Aggregate and Disaggregate
Recall that we are typically dealing with the consumer's choice
of the set of travel options (f,d,m,r). If we consider an individual
traveller (or household) i at location k, we can ask, which combina-
tion (f,d,m,r) does this individual pick?
In this case, our demand function is dlsag re ate; it predicts
the choice of a single decision-making unit (either individual or
household). We can write the demand function as
Xfdalr - gi (A,L)
(4.3-1)
where: Sfdmr - 1 if I chooses (f,d,m,r);
0 if I chooses some other combination
Let us now consider a group of individuals, e, which we will
define as a "market segment." (Generally, we try to group indi-
viduals or households into market segments such that the travel
behaviors of all the individuals in a single market segment are
relatively similar.)
In this case, we write:
(4.3-2) Vkdmr - f5 (A,L)
This aggregate demand function gives the number of individuals in
market segment e at location k who will choose (d,m,r).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved'ror Release 2001/11/19: CIA-RDP79-00798AA00200020005-9
FIGURE 4.2-1
ALTERNATIVE FORMS OF DEMAND FUNCTIONS
LEVEL OF.SERVICE (e.g..,cost, time)
Approved For Release 2001/11/19: CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A00020002000-9
4.4 Probabilistic and Deterministic
A probabilistic demand model gives an estimate of the relative
likelihood of different choices being made. A deterministic demand
model, on the other hand, predicts that one and only one specific
choice will be made.
Usually, disaggregate travel demand models are also formulated
as probabilistic models. If we consider an individual traveller or
household i, we can ask, what is the probability pi that this indi-
vidual at location k will choose a particular combination of
(f,d,m,r)? In general, we know this will be a function of some
of the social, economic, and other characteristics of that indi-
vidual (Ai), the characteristics of the activities which could
be undertaken at various destinations d, (Ad), and the level-of-
service characteristics Lkdmr for each path r in a mode m from
location k to destination d:
(4.4-1) Pi(f,d,m,r) - gi({Ai,Ad, Lkdmr} for all d,m,r]
We write this as a probability in part because there will
likely always be some inherent randomness in each individual's
decision process.
Usually, aggregate models have been assumed to be determinis-
tic, although a probabilistic form would also be reasonable:
(4.4-2) Pke(Vkdmr) - fe (A,L)
4.5 Simultaneous and Sequential
In analyzing the travel behavior of any individual or group
or individuals, a number of alternative assumptions about the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
pattern of behavior can be made. One assumption of particular im-
portance is whether individuals make simultaneous or sequential
choices.
At the disaggregate level, we can write the two cases as follows:5
(4.5-1) Sequential: Pi(f) ? pi(dif)?pi(mlf,d)?pi(rIf,d,m)
(4.5-2) Simultaneous: Pi(f,d,m,r)
These two equations are mathematically equivalent only if, in
(4.5-1) pi(f), the probability of a given level of trip frequency
(f) is indeed independent of destination (d), mode (m), and route
W. Similar conditions must hold for pi(dif), etc. If these con-
ditions are met, then the following behavioral interpretation can
be given to (4.5-1): the potential traveller chooses first how many
trips to make, then a destination, then (conditional on choice of
destination and frequency) a mode, and then (conditional on having
made all the other choices), a route. If these conditions (or an-
alogous conditions for an alternative sequence of decisions), are
not met, then (4.5.2) can be chosen. The corresponding behavioral
interpretation is that the individual chooses his trip frequency,
destination, mode, and route simultaneously.
"Alternative estimation (model calibration) approaches are
available for each case. Estimation of a simultaneous model, such
as (4.5-2), is appropriate regardless of any behavioral choice se-
quence which may exist. Once such a simultaneous model has been
estimated, any desired marginal and conditional probability functions
can be derived using the basic formulas of probability theory--for
example, a .model split probability function, conditional on choice
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
of destination and trip frequency as in (4.5-1): p(mlf,d). How-
ever, since simultaneous models are harder to estimate, information
on the actual sequence of choices, if available either from prior
empirical studies or from behavioral theories, can be used to sim-
plify estimation by using sequential models. If the proper se-
quence is assumed, the results should be identical, since the two
cases are mathematically equivalent and, thus, either can be con-
verted to the other, using the mathematics of probability theory.
If the sequence assumed as a basis for estimation is not correct,
this will become evident in one or both of the following ways:
1. If both types of models are estimated, they will not obey
the relationships required by probability theory. Ben-
Akiva (1973) has demonstrated this case.
2. If only a simultaneous model is estimated, it will not
be possible to derive sequential submodels with the neces-
sary independence properties.
The most general assumption is that the decision process of
the individual traveller is a simultaneous one and, therefore, we
should estimate a simultaneous model. Then, any desired sequence
of models can be derived.
Although from a behavioral point of view it would seem best to
begin with an assumption of a simultaneous choice model (where there
are no behavioral arguments for any particular sequential model),
there may be practical problems. Because of the larger number of
possible combinations of choices, the number of explanatory variables
and allowable interactions between variables may provide some difficulty,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
and the model may become very complex. However, these problems
could be resolved in any of several ways other than by assuming
a sequential model, such as reducing the number of attributes a
traveller is assumed to consider or the number of potential desti-
nations.
A similar distinction can be made at the aggregate level:
(4.5-3) Sequential: Vk - f 1(Ak,Dd'`Lkdmr)
Vkd f2(Ak,Dd'Lkdmr'Vk)
Vkdm f3(Ak'DdLkdmr'Vkd)
Vkdmr - f4(Ak,Dd'Lkdmr'Vkdm)
(4.5-4) Simultaneous: Vkdmr - f(Ak,Dd'Lkdmr)
Again, we note that the sequence in (4.5-3) is one of many possible
sequences.
It is important to note that the sequence in (4.5-3) is the
same general form as in the traditional four-step demand models
used in urban transportation and introduced in section 3--The Urban
Transportation Model System (UTMS). Recent theoretical results,
including the development of a "General Share Model" [12], show
that any desired sequence, such as (4.5-3), can be derived from a
given simultaneous form (4.5-4); and that, under certain conditions,6
an equivalent simultaneous form (4.5-5) can be derived from the
sequential form (4.5-3). (The logic of these results is very
similar to the basic mathematics of probability theory which apply
at the disaggregate levels,) Again, however, in each case it is
likely that equivalent forms (the two simultaneous forms or the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 132
two sequential forms) obtained in the two different ways (direct
estimation or mathematical derivation of one from the other) will.
be different.
Therefore, the most general approach at the aggregate level
is also to assume a simultaneous choice form, estimate that form,
and then derive analytically the corresponding desired sequential
forms, just as was suggested for the disaggregate models above.
In. this way, some of the objections to the conventional four-
step models (cf. section 6.1.5 below) can be overcome; since level-
of-service variables will appear in each step, there can be explicit
behavioral structure, and valid estimation procedures can be used.
4.6 Other Modelling Choices
In addition to those explicitly described in this section,
other key modelling choices include: choice of market segments;
commodity-dependent versus commodity-independent attributes (e.g.,
"abstract mode"); incorporation of activity variables explicitly
or via stratification. These are, in general, choices which must
be made by examination of the results of statistical testing of
hypotheses reasoning.
5. Relationships of Major Model Development Alternatives
The preceding discussion suggests a bewildering variety of com-
binations of modelling choices. In this discussion, we will focus
on the three sets of structural relationships: aggregate versus
disaggregate; probabilistic versus deterministic; simultaneous ver-
sus sequential. There are at present four basic alternatives for
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. Aggregate, deterministic, simultaneous models--Charles
River Associates San Francisco direct demand model
(CRA,1967), for example
2. Aggregate, deterministic, sequential models--the traditional
UTMS models, introduced in section 3, for example
3. Disaggregate, probabilistic, simultaneous models--as esti-
mated by Ben-Akiva (1973), for example
4. Disaggregate, probabilistic, sequential models--the bulk
of the behavioral disaggregate modal split models [Stopher,
1969; Stopher and Lisco; Quarmby; Reichman and Stopher;
Charles River Associates (1972); McFadden]
In addition to these four alternative forms, there are also four
basic ways to use these estimated models in forecasting. Because we
must forecast total trips made (by various modes to various destina-
tions, etc.) rather than probabilities of trips made, disaggregate
models must be aggregated in the forecasting process.
Prediction of equilibrium flows requires an equilibration method
used with an aggregate demand model (either estimated in aggregate
form or constructed from an estimated disaggregate model by aggre-
gation). If the resulting aggregate model is simultaneous, then
equilibrium can be computed in a single step with a "direct" approach.
If the demand model is sequential, i.e., a sequence of equations,
then equilibrium is usually computed in a sequence of steps, in an
"indirect" approach. These two choices for forecasting are indi-
cated as E and F in figure 6-1. The four basic alternative ways
to estimate and use models for forecasting as shown in figure 6.1 are:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 134
1. Estimation of an aggregate, deterministic, simultaneous
model for all aspects of trip choice; and using that same,
model for forecasting in a direct approach. (This is re-
presented by line CE in the figure.)
2. Estimation of a disaggregate, probabilistic, simultaneous
model of all aspects of trip choice, and then using an
aggregation process to obtain an aggregate simulteneous
model for forecasting in a direct approach. (This is
represented by line AE in the figure).
3. Estimation of an aggregate, deterministic, sequential
model of trip choice, by estimating each component in
the sequence, and then using each of the aggregate model
components separately for forecasting in an indirect ap-
proach. (Represented by line DF in the figure.)
4. Estimation of a disaggregate, probabilistic, sequential
model, also estimating each component separately, then
using an aggregation process to separately aggregate each
component up to a sequential aggregate model, and then
using the aggregate model components for forecasting, in
an indirect approach. (Represented by line BF in the
figure.)
In the past, these four approaches have seemed to be essen-
tially different and competitive. However, recent research results-
show that these approaches are all interrelated; they are different
views of the same travel behavior. Once this is understood, rela-
tions among the different views can be developed and, thus, addi-
tional approaches become available.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19: CIA-RDP79-00798A000200020005-9.
135
m
0
IL
Approved For Release 2001/11/19': CIA-RDP79-007984000200'020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The first step in laying out these interrelations is to note
the connection between simultaneous and sequential forms at the
disaggregate level, that of the individual consumer. (The basic
theory was developed by Ben-Akiva (1973). It is possible to go
from simultaneous disaggregate to sequential disaggregate models,
using the basic mathematics of probability theory; and under
specific conditions to go from sequential to simultaneous dis-
aggregate models.
The second step is to note the connection between simultaneous
and sequential forms at the aggregate level, that of a "market seg-
ment" composed of a group of individual consumers. The basic theory
here was developed in the concept of the General Share Model (Manheim,
1973). This development shows how it is possible to go from simul-
taneous aggregate ("explicit") to sequential aggregate models, using
relations very analogous to those applying at the disaggregate level;
and under special conditions to go from sequential aggregate to
simultaneous aggregate models.
The third step is to note the connection between aggregate and
disaggregate forms. A general theory of aggregation is not yet
available; however, assumptions can be made which: lead to directly
useable practical procedures (Koppelman, 1975).
Once these relationships are recognized, additional demand
modelling approaches become available, as shown in figure 6-2. For
example, an aggregate simultaneous model can be estimated directly
or by aggregating a disaggregate simultaneous model. From this can
be derived an aggregate sequential form of use in an indirect fore-
casting approach, such as the four steps used in the traditional UTMS
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
and embodied in numerous software packages. This is represented by
lines CF and GF respectively in Figure 6-2. Similarly, for the se-
quential model, we can derive a simultaneous form of that model for
forecasting in a direct approach. In Figure 6-2, this is repre-
sented by lines BHE and DE respectively. So, for example, in recent
work for CALTRANS (Nestle and Young) the present California genera-
tion, distribution, and mode split relations were combined into a
single simultaneous model.
Thus, we now have a great number of ways to estimate and use
demand models (eight ways, structurally, as shown in Figure 6-2).
The benefit of this wide range of alternative combinations of
estimation and forecasting techniques is that we can exploit the
data efficiency of disaggregate models, and produce models similar
to those conventionally used today, but which are behaviorally
more valid for use in policy-oriented studies.
Moreover, because of these relationships which exist between
alternative estimation methods and alternative forecasting methods,
we can base the choice of forecasting methods on grounds solely of
computational efficiency. We are thus free-to chgose behavioral
assumptions and estimation methods without concern for which form
will be used for forecasting:
1. the "software" for forecasting can be designed to operate
either with the conventional 4-step models or with simul-
taneous models;
2. disaggregate methods can be used for efficiency in model
development;
3. aggregation of disaggregate models can be done to produce
both simultaneous and sequential forms, to allow the user
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19.:'CIA-RDP79-00798A00020002.0005-9 ? 138
Approved For Release 2001/11%19 .' CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
of the models to choose which ones are most useful in a
specific forecasting task.
6. Examples of Major Demand Modelling Choices?
In the preceding section, the set of combinations of major de-
mand modelling choices which have been chosen so far for demand model
development was identified below:
Group I - aggregate, sequential, and deterministic
Group II - aggregate, simultaneous, and deterministic
Group III - disaggregate, sequential, and probabilistic
Group IV - disaggregate, simultaneous, and probabilistic
In the following sections, we will give examples of two of
these groups: I and IV.
6.1 The Traditional Four-Step Approach: Aggregate Sequential Models
To review: The basic choice open to prospective travellers is
among combinations (f,d,m,r,h;ao,res,emp)
where: f - frequency of trips8
d = destination
m=mode
r - route
h - time of trip
ao = number of autos owned
res - residential location
emp = employment location
For most urban travel forecasting, it has been conventional to assume
that residential location (res) and employment (emp) are fixed for
each traveller (or alternatively, an urban growth and land use model
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 140
has been used to predict these "long-run". effects separately from
the "short-run" travel decisions). Thus, models have dealt primarily
with choices of (f,d,m,r,h;ao). Furthermore, typically the choices
of time of travel and of automobile ownership are usually dealt
with separately, leaving (f,d,m,r) as the primary concern of travel
models.
Usually, we are interested in the behavior of groups of indi-
viduals. A market segment is defined as a group of individuals
(or households) with similar travel behavior. Then, if we consider
all of the individuals in market segment e who reside in zone k of
an urbanized area, the demand function can be described as:
(6.1-1) Ve - f(A,L)
where A,L are a variety of relevant activity system and servica
variables describing the choices available; and Vee dmrh;ao is the
total number of trips made by members of market segment e in origin
zone k owning "ao" automobiles, to destination d by mode m and
route r at time h.
In the typical approach, where it is assumed that choices of
time of travel and of auto ownership are made separately from the
others,-this becomes:
(6.1-2) Vkd - f(A,L)
Historically, this demand model has been broken into four com-.
ponents:
(6.1-3) Vkdmr - f 1(A,L) ? f2(A,L) . f3(A,L) ? f4(A,L)
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
where f 1
f2
- trip generation submodel
- distribution submodel
= modal split submodel
- route choice ("network assignment") submodel
By breaking the demand function into four submodels, it was
possible to break the forecasting of future travel into a sequence
of four steps:
trip generation:
Vk = total volume of trips generated by zone k in market
segment e
(6.1-4) Vk = f1(A,L)
trip distribution:
Vkd = total volume of trips originating in zone k and
travelling to destination zone d, in market seg-
ment e
(6.1-5) Vkd - f2 (A,L) .
model split:
(6.1-6)
(6.1-7)
Vkdm = total volume of trips going by mode m from origin
zone k to destination zone d in market segment e
Vkdm - f 3 (A,L) Vkd
network assignment:
e
Vkdmr
Vkdmv
total volume of trips going by route r of mode- m
from origin zone k to destination zone d in mar-
ket segment e
f4 (A,L) ? Vkdm
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
In equations (6.1-5, 6.1-7), the functions f2, f3, and f4 re-
present "share" functions--they serve to split a larger trip total
into a number of components: In equation 6.1-6, for example, f3
is applied to the. total trips from k to d by segment e to obtain
the share of these trips which will go by each of the available
modes, m.
It is instructive to examine each of these submodels in detail.
6.1.1 Trip Generation
Trip generation is the first sequential step, involving the
prediction of total trips from an origin or to a destination by
trip purpose (7). The functional form is usually linear. Symbolic-
ally,
e e e
Vk - a + ibiAki + iCifi(Lk.)
where: Vk - trips of purpose e generated in origin k
ae, bi,ci empirical parameters
Aki - activity system variable i for origin ki
fi(Lk.) - function of level of service variables from k
to all destinations
Typically, activity system variables used are average annual
income, average number of autos owned, number of workers per house-
hold, percentage of households having an income greater than a
specified value; zonal population, acres of land in various land-
use categories and zonal employment. A typical level-of-service
function is accessibility, defined as a function of travel times
or distance; but such functions are seldom used in trip generation.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Example
The two trip-generation equations which follow are typical of
equations used in the traditional four-step approach:
Total Home-
Based Trips
- 0.69 + 1.94 (cars/D.U.)
Dwelling Unit
Home-Based
+'1.39 (residents > 5 yrs./D.U.)
- 1.148 (workers)
Work Trips
+ 0.569 (households)
+ 0.019 (residents > 5 years)
- 0.144 (residents x distance from CBD)
+ 0.488 (residents)
Note that neither equation includes any level-of-service
variables: zonal trips will therefore be predicted to remain
constant, no matter what level of transportation service is
provided.
6.1.2 Trip Distribution
The second sequential step in the conventional four-step ap-
proach is trip distribution, the prediction of trips from origin
to destination. The independent variables are the "trip ends" re-
sulting from the previous step, plus level-of-service variables.
Symbolically,
(6.1-8) Ykd - fn(VkVd' Lkd)
where Vkd . trips of purpose e from origin k to
destination d
- results of the trip-generation step
Vk, Vd e
Lkd
level-of-service variables between k and d
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The two most common functional forms are the gravity model
and the opportunity model. Typical versions of each of these
follows:
1. Gravity Model
(6.1-9) Vkd
efe
Vd kd
e e
dVdfkd
where fkde is an arbitrary function of travel time.
Example
The following table shows typical values of fed for a range of
travel time, and for three trip purposes: home-based work, home-
based nonwork and nonhome-based.
Note that the units of fed need not be defined, since the
quantities appear in both the numerator and denominator of the
gravity model equation. For the same reason, the values can all
be scaled up or down by the same constant without affecting the
model's results.
2. Opportunity Model
(6.1-10.) V e - V exp(-LeSd) I1-exp -LeVd]
where:
Sd - E Vi - "subtended volume"
j - all
destinations for which tkj < tkd
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For'Reiedse 2001/11/19: CIA-RDP79 00798A000200020005-9
Typical Gravity Model
.5.
(Source: Federal Highway Administration)
Table 5.1.1
Travel'Time Factors
e..
fkd
:..Work Non-Work ..Non-Nome-Based
.79
67 .
61
57
50
14 .48.
15 45.
16 10,
17
18
220
210
160
120
130
100
90
80
' 85
70
70.
60
60
55
50
44
39' ,
?
38
35
32
27
30
25
?.
26
21
23
?16
14
Approved For Release; 2001/1.1/19 : CIA-RDP79.p 007.98AO00200020005-9
5
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Example:
In the Chicago Area Transportation Study, three market seg-
ments were used:
e - 1: trips from home to work and all trips to the central
business district (long residential)
e - 2: trips from work to home and all trips from the central
business district (long residential)
e - 3: all other trips (short)
Long residential trips have only nonresidential trip ends as
their subtended volumes, and vice versa. Short trips have only
short trip ends as their subtended volumes. The model parameter
for all trips (e - 1 and 2) is 2.5 X 10-6, that for short trips
(e - 3) is 20 X 10-6.
The model can therefore be stated mathematically as follows:
Vkd - Vk exp [-2.5 X 10-6Sa] {1 - exp [-2.5 X 10-6Va]}
Vkd - Vk exp [-2.5 X?10-6Sa] {1 - exp [-2.5 X 10-6Va]}
Vkd Vk exp [-20 X 10-6 Sa] {1 - exp [-20 X 10-6Vd3]}
Each of these distribution models are "share" models; they
divide the total trips from k,Ve, among all distributions using
a fraction which, when summed over all destinations, equals one.
Travel time by a single mode, usually highway, is typically the
only level-of-service variable used, although in some applications,
a "generalized cost" has been used which is a linear combination
of travel time, distance, and out-of-pocket costs. The level-of-
service variable enters the Opportunity Model in an indirect way
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
only. It affects the ranking.of destinations from each origin,
which in turn affects the subtended volumes which enter the model
directly.
In some applications of both the gravity and opportunity
models, adjustments of the Va's are made after initial application
of Equation 6.1-9 or 6.1-10 in an attempt to force the total trips
to each destination (Ve - E Ve ) to equal the original V a's. This
i id
constraint is not guaranteed by the functional form of either dis-
tribution model. Following adjustments of the original VV' s., the
equations are applied again. Iteration through application of the
equations and adjustment of the original values continues until a
desired level of correspondence between each Ve and Va is reached.
6.1.3 Modal Split
The third sequential step in the conventional approach is
modal split, the prediction of trips by mode from origin to desti-
nation. The independent variables are the trip interchanges re-
sulting from the previous step, plus modal level-of-service
variables. Symbolically,
(6.1-11)
e f e
Vkdm em(Vkd'Lkdq'SkAd)
where Vkdm - trips of purpose e from origin k to desti-
nation d by mode m
Ved - results of the trip distribution step
Lkdq - level-of-service variables for all modes q
between k and d
Sk - socialeconomic variables of travellers in k
Ad - activity system variables in d
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Many approaches have been used to develop functional forms,
fam, for modal split models. The most commonly used prior to the,
last three or four years were regression or table look-up models
based on the relative levels of service offered by each mode (e.g.,
reference [4]). Typically, origin zones have been classified by
income level and auto ownership, and for each subgroup linear equa-
tions or tables are developed which relate fraction of trips by
auto and transit to time and cost ratios or differences.
Example
Figure 6.1-2 shows a set of graphs of modal split relationships
for work trips based on relative levels of service and on
cross-classifications based on the economic status of the
traveller and the relative levels of excess travel time. The
specific variables used in the figure are:
Vertical axis - trips by transit
trips by all modes
TTR - total travel time by transit
total travel time by auto
L - excess travel time bytransit
excess travel time by auto
Ll - 0 to 1.5
L2 1.5 to 3.5
L3 3.5 to 5.5
L4 - 5.5 and over
CR - out-of-pocket cost by transit
out-of-pocket cost by auto
CR1 0.0 to 0.5
CR2 - 0.5 to 1.0
CR3 - 1.0 to 1.5
CR4 - 1.5 and over
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
EC - economic status (median worker income, 1963 dollars)
EC1 . $0 to $3,100 per-year
EC2 - $3,100 to $4,700 per year
EC3 - $4,700 to $6,200 per year
EC4 - $6,200 to $7,500 per year
EC5 - $7,500 per year and over
More recently, the following functional form (termed the binary
choice logit form) has been used for fem, applied to modal split
between auto (a) and transit (t).
e
e 1
(6.1-12) P kdm
1 + exp [hm(Lkdg)]
and hm C - Cm + Eat" (tz - ti ) + Eb-' (cz - ci )
I m kdt kda i m kdt kda
Again, times and costs have been divided into various variables.
The constant, Cm, as well as parameters aM and bm, allow the relative
characteristics of modes not measured by times and costs (such as
comfort and convenience and modal "image") to be represented in the
model. The function hm can be interpreted as a difference in consumer
utility between travel by transit and travel by auto.
6.1,4 Route Choice
The final step in the conventional approach is traffic assign-
ment, the prediction of trips by route and thus by link. Symbolically,
(6.5-13) Vma m f(Ve kdm , ma)
where V - volume of traffic of mode m on link a
ma
L - level of service on link a for-.mode in
ma
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
This equation is deceptively simple; actually, bridging the gap
between predictions of 0-D volumes.(Vkdm) to link volumes (Vms)
represents a significant problem. This is especially true if an
attempt is made to take into account the variation of link levels
of service (usually simply travel time) with link volumes. When
this is not done, the assignment is termed an all-or-nothing
assignment without capacity restraint. Various methods of adjusting
travel times (alternatively, applying capacity restraint) have been
developed. Each one involves either adjusting travel times after a
portion of trips are assigned, and then continuing assignment; or
averaging a number of complete assignments, each one based on the
travel times corresponding to the previous assignment. In present
assignment methods, one of the following assumptions is made about
how travellers chose a path from origin to destination:
1. Each traveller chooses the minimum path. This assumption
leads to a situation where each path used between an 0-D
pair has equal travel time and all paths not used have a
higher travel time.
2. Travellers distribute themselves in such a way that the
time and volume on each path between an 0-D pair are re-
lated as follows:
1
(6.1-14) Vkdr - !kdr ? Vkd
Ef kdr)
r
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Vkdr - volume of flow for 0-D pair k-d on route r
tkdr " travel time on route r
f(tkdr) a decreasing function of route travel time.
6.1.5 Critical Appraisal of the Traditional Four-Step Models
The approach described above is the most widely used transpor-
tation systems analysis approach--the Urban Transportation Model
System (UTMS). It has been applied in over 200 cities in the
United States and in many other cities around the world. The de-
velopment and institutionalization of this approach over the last
fifteen years is a major accomplishment; it is the first large-scale
application of modern systems analysis techniques to problems of
the civil sector.
It is useful to examine this approach critically, from the per-
spective of the equilibrium theory presented in section 3 and the
challenge of today's urban transportation problems.
As we have seen in this traditional approach, the travel demand
models are structured into a sequence of four submodels called:
trip generation, distribution, modal. split, and assignment. Essen-
tially, this amounts to estimating Vkdmr in a series of "successive
approximations": first, Vk, then Vkd, then Vkdm, and finally Vkdmr'
It seems obvious that the following conditions should be met
by any such set of demand models and submodels and the corresponding
equilibrium calculating procedure:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. Level of service, L, should enter into every step, in-
cluding trip generation (unless an analysis of the data
indicates in a specific situation that trip generation is,
in fact, independent of level of service for all market
segments over the full range of levels of service to be
studied).
2. The level-of-service attributes used should be as complete
as necessary to predict adequately traveller behavior. For
example, time reliability, number of transfers, privacy,
etc., should be included if empirical evidence indicates
these are important.
3. The same attributes of service level should influence each
step (unless the data indicates otherwise). For example,
transit fares, auto parking charges, walking distances,
and service frequencies should influence not only modal
split but also assignment, generation, and distribution.
4. The process should calculate a valid "equilibrium" of
supply and demand; the same values of each of the level-
of-service variables should influence each step. For ex-
ample, the travel times that are used as inputs for modal
split, distribution, and even generation, should be the
same as those which are output as results from assignment.
If necessary, iteration from assignment back to generation,
distribution, etc., should be done to get this equilibrium.
5. The levels of service of every mode should influence demand.
Congestion on highway or transit networks, limited capacity
(e.g., parking lots), fares, etc., of each mode should (in
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
general) affect not only its own demand but also the demand
for other modes, at all steps (generation, distribution,
modal split, and assignment). That is, there should be
provision for explicit cross-elasticities.
6. The estimation procedures should be statistically valid and
reproducible.
Careful examination of the traditional approach indicates it
violates each of these conditions. As a consequence, serious ques-
tions can be raised about the biases and limitations of the flow pre-
dictions resulting from use of the models in their traditional forms.
Nevertheless, this set of models--the UTMS--is still in widespread
"production" use in almost all urban transportation planning activi-
ties in the United States.
6.2 Group IV: Disaggregate Simultaneous Models
Modelling at the aggregate level uses data for entire zones
whose sizes range from fractions of square miles for urban applica-
tions to entire metropolitan areas for intercity applications.
Modelling at either of these levels of aggregation smoothes out
most of the variations of the behavior of the individuals who actually
make'the travel decisions being modelled. For this reason, much recent
demand modelling effort has devloped disaggregate models for predict-
ing travel decisions of individual travellers. Initially, these
studies were concerned only with the mode choice decision. The models
developed were individual traveller applications of the form shown in
section 4.3. When applied to individuals, the dependent variable can
only take on the values 0 or 1, requiring a different set of estimation
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
procedures to be used than with aggregate models. In the initial
models of this type, only two modes were included, leading to a
binary choice situation. The binary choice logit functional form
has been used for these models. More recently, multiple choice
models have been developed, using the multiple-choice logit func-
tional form (Ben-Akiva, 1973; McFadden), or multinomial logit,
as it is sometimes called.
One feature of these models which is much discussed in the
literature is their "independence of irrelevant alternatives" pro-
perty. This property has two consequences of concern to transpor-
tation analysts:
1. If a new alternative (mode, destination, etc.) is added,
the percentage decreases in the usage of all existing al-
ternatives will be constant.
2. If one existing alternative is improved, the percentage de-
crease in the usage of the remaining alternatives will be
constant.
The multinomial logit model (MNL) is of this form:
p(i:S) -
Ui
e
E e U j
:ES
where:
x - level of service attributes
0 - Parameters
Ui - EOkXik
k
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
In the case of travel choices:
i - (f,d,m,r) and
S - FDMR, the full set of all possible travel choices.
Thus, the simultaneous choice MNL becomes:
Uf dmr
p(f,d,m,r: FDMR) - e
to Ufdmr
F,D,M,R
We can relate this to the sequential disaggregate form by basic
probability theory:
p(f,d,m,r) - p(f) p(djf) p(mld,f) p(rlm,d,f)
- pf(X,O) ? pd(X'p) . pm(X'O) . pr(X,g)
the specific choice functions
(pf' Pd' Pm' Pr) can be shown to be also multinomial logic
in form [Ben-Akiva, 1973]
p(f:F) -
p(d:D) -
p (m: M) ~ -
p(r:R) -
Uf
e
Ee Uf
feF
Udf
e
E
deb Udf
mdf
e
U mdf
meM
U rmdf
e
rle U rmdf
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The first estimation of a simultaneous disaggregate model was
by Ben-Akiva (1973). The functional form of this multiple-choice
logit model is as follows:
(6.8-1) Pkdm - exp[Ea (A - A , ) + Eb (M1 - M1, )
Pkd'm' q . dq d q q q mq m q
+ EcYk(M2 - M2,q) + Ehq(L 11 dmqL1 ' ) + E 2 2
q g q Lid m q q Y k (L kdmq - L kd'm,q))
P , P , ' - fraction of total trips from household k going to
kdm kd m destinations d and d' by modes m and m'. (Either
d abd d' or m and m' may be the same, but not both.)
Adq, Ad,q
Mmq, m q
i . i
Lkdmq' Lkd' m' q
Yk
- activity system variables
- modal variables
- level of service variables
- household income code 9
As estimated by Ben Akiva, the following variables were used:
1. Activity system variables (Adq):
Adl - number of jobs in wholesale and retail establishments in
the zone of destination d
Ad2 indicator for DBD destinations. (1 if d - CBD, 0 otherwise)
2. Modal variable in separate term (M1 ):
Mm1 1 - indicator for auto usage. (1 if in - auto, 0 otherwise)
3. Modal variable in interaction term with income (Mmq):
M221 - indicator for auto usage. (same as M' )
M ml
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
4. Level of service variables. in separate terms (Lk1
dmq
Lkdml = out-of-vehicle travel time
Lkdm2 ? in-vehicle travel time.
5. Level of service variable in interaction term (Lkdmq
2
Lkdml - out-of-pocket cost.
This model was estimated for auto and transit trips for the
shopping purpose only, and does not deal with trip-making frequen-
cies or time of day choices. It therefore represents a model which
can be used to divide total shopping trips from a household among
the variable modes and destinations. The parameters obtained are
the following:
Associated Variable
Parameter Lavel
Parameter Value
(Adl - Ad,l)
a1
.000171
(Ad2 - Ad,2)
a2
.316
1
1
-
ml Mm'l
bl
-1.36
2 -
Yk(Mml M 2
21)
Cl
.114
1 1
~Lkdml - Lkd'm'1)
h1
-.0633
2 2
Lkdm2 - Lkd'm'2)
h2
-.0164
2 2
Lkdml - Lkd'm'1)/Y
fl
-.0757
In additional work since that time, Ben-Akiva and his associates
have developed additional simultaneous choice disaggregate models.
Some of these models will be summarized briefly here and are described
in detail in the appendix.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 158
The three models described below are:
1. A work mode choice model with an explicit carpool mode
2. A simultaneous choice model for shopping trips including
the choices of travel frequency, trip destination,'and
mode of travel
3. A household auto ownership model
All these models were estimated using the multinomial logit model
with data sets derived from the 1968 Washington, D.C. Home Interview
Travel Survey, conducted by the Metropolitan Washington Council of
Governments. In addition, in recent work, a model of simultaneous
choice of residential location, auto ownership, and mode to work
has been developed (Lerman).
1. Work Choice Model 10
A work mode choice model has been produced as a multinomial-
logit three-mode (auto driver alone, carpool, transit) model, esti-
mated using the Washington, D.C. data base.
The model specification recommended is shown in Appendix A.
This model contains all the normally expected variables--in-vehicle
travel time, out-of-vehicle travel time, out-of-pocket costs, income,
and auto availability--plus some special variables to pick up some
of the factors influencing carpooling. Further description of this
model and a discussion of the variables can be found in Appendix A.
It should be noted that, since the models produce choice prob-
abilities; they do not assume that a traveller will use the same
mode every day. This is particularly important for carpool users
in view of the fact that many existing carpools have been observed
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
not to be an everyday arrangement and carpools are frequently
organized for less than five trips per week. Thus, if a traveller
carpools twice a week, for example, his carpool choice probability
on any given day is 0.4.
2. Non-Work Trip Frequency, Destination, and Mode Choice Models
Two nonwork (shopping) simultaneous disaggregate travel demand
models have been developed on the 1968 Washington, D.C. data base.
The first was the estimation of a joint destination and mode choice
shopping model by Ben-Akiva (1973), described above. The second
was the subsequent estimation of a joint frequency, destination,
and mode choice nonwork model by Adler and Ben-Akiva (1974). Both
are logit models calibrated by maximum likelihood estimation and
both forecast the frequency, destination, and mode choice of a
household's shopping trip as a function of automobiles available,
among other factors. The second model was estimated on a larger
number of observations than the first, has reasonable coefficients
in terms of both sign and relative magnitude, and all the important
coefficients are statistically significant.
The variables used in this second model and the estimation re-
sults are shown in tables B.l.and'?B.2 in Appendix B. The model con-
tains transportation level-of-service variables, socioeconomic char-
acteristics and shopping attraction variables.
Particularly noteworthy in the specification of this model are
its relationships with work trips. The auto availability variable
is formulated as the auto available to the household minus those
used for work trips by household member. This specification assumes
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
that the choice of mode with respect to work trips is of higher im-
portance to the household than the shopping travel choices. This
specification allows for an explicit evaluation of a policy that
the availability of auto for off-peak nonwork trips. This model would
be used after the work mode split model when auto availability was
determined so that it could be input to this shopping model. Further
description of this model and a discussion of the variables can be
found in Appendix B.
3. Auto Ownership Model
Recently, a simultaneous disaggregate auto ownership and work
mode choice model has been developed. This model addresses the
classic "chicken and egg" dilemma of the close interdependency
between auto ownership and mode-to-work choice by developing and
estimating a disaggregate choice. model which assumes these two de-
cisions are made simultaneously. Each combination of auto ownership
level and usual mode to work is a distinct "bundle" or alternative,
one and only one of which is assumed to be selected by each house-
hold.
The model describes the joint probability of a household se-
lecting a given auto ownership level and a given mode to work for
the breadwinner. (The worker with the highest occupation status
code has termed the "breadwinner"; it was felt that his transpor-
tation to work would dominate in terms of demand for the family
auto.) Hence, the choice set for a household consists of the cross-
product of the entire set of modes and the entire set of possible
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
auto ownership levels. However, this can be reasonably simplified
without any major loss in the model's usefulness by assuming that
a maximum of eight choices forms the set of alternatives which is
assumed to be available to any household. These are as follows:
1. Own zero autos and carpool to work
2. Own zero autos and use transit to work
3. Own one auto and drive alone to work
4. Own one auto and carpool to work
S. Own one auto and use transit to work
6. Own two or more autos and drive alone to work
7. Own two or more autos and use transit to work
Note, however, that there is no requirement in the logit model
that every household have all eight alternatives. For some work
trips, transit service is simply not likely to be perceived as
available; these households choose among only five alternatives.
Other households have an income which is simply too low for them
to consider realistically multiple car ownership as an alternative;
these households have only the first five alternatives. A variety
of other combinations is also feasible. For example, in the model
calibration it was assumed that a household will not own more cars
than the number of licensed drivers who might use those cars.
These rules for determining which alternatives pertain to each
household were verified by performing extensive cross-tabulations
on the data set and examining whether any substantial portion of
the sample made a choice assumed to be unavailable.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 162
A broad range of variables was considered for providing possible
candidates in the model specification. An attempt was made to re-
strict the model to variables which are behaviorally related to
the household decision process rather than simply correlated with
actual causal variables.
The model specification recommended is discussed in Appendix C.
The important variables contained in the model include the transport
level-of-service characteristics (in-vehicle travel time, out-of-
vehicle travel time, and travel cost) for both peak and off-peak
travel. In addition, the model includes other factors that in-
fluence carpooling, and variables that determine auto ownership
and mode choice such as housing and socioeconomic characteristics
of the household.
One of the most important aspects of the auto ownership model
development was the determination of significantly, yet logically
different, behavioral responses from different market segments
(groups with homogeneous socioeconomic characteristics). The above
model was estimated for a pooled sample of households but, in addi-
tion, separate models were estimated for each of several different
market segments. This same stratification by market segment could
be applied to all three models (work mode split, shopping direct
demand, and auto ownership) to yield more specific information on
response to the various policies.
Further description of the joint auto ownership/work mode
choice models and a discussion of the variables can be found in
Appendix C.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Note that, while the model is formulated here as a joint (auto
ownership and work mode choice simultaneously) model, separate
marginal (auto ownership alone or work model choice alone) or condi-
tional (auto ownership, given work mode choice; or work mode choice,
given auto ownership) models are directly and easily derivable.
Moreover, the work mode choice portion of this joint model refers
only to the mode choice decision of the breadwinner, not of all
household workers. This is in contrast to the work mode choice
model for all household workers. This gives rise to two alternative
ways of utilizing these three available models for immediate planning
analyses:
1. Use the marginal auto ownership form of the joint model to
yield auto ownership; then use the section l.?.wojrk mode
split model for all workers, yielding complete information
on work trip mode splits and auto availability for other
trips; then run the nonwork demand model to capture changes
in nonwork travel due to increased auto availability.
2. Use the joint auto ownership/breadwinner work mode choice
model for breadwinner work mode split; use the section 1.
work mode split model for all other workers (since there
is a breadwinner variable in that model, there is no bias
in using it for nonbreadwinners only): then, taking the
composite results of those models in terms of remaining
auto availability, run the nonwork demand model.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
4. Features of These Models
It is useful to summarize the major advantages of these models
over other travel demand models that have been developed:
1. Car Pooling
The work mode choice model includes a car pool mode
in addition to drivers driving alone and transit. Other
existing work mode choice models consider auto drivers
and auto passengers as two separate modes. Therefore,
one cannot isolate persons in multiple occupancy vehicles
from those with a driver only. Analysis of special in-
centives to multiple occupancy vehicles can only be per-
formed using a model with an explicit "car pool" mode.
2. Auto Ownership
These models include an auto ownership model sensi-
tive to the important transportation policy variables,
while auto ownership is assumed as an exogenous (independent)
variable in other travel demand models. The importance of
this model for the analysis of today's policy thrusts is
obvious: auto ownership can be influenced by transportation
policies, and these models include this effect.
3. Sample Size
In a recent project to estimate disaggregate travel de-
mand models in the Netherlands, an analysis of the effect
of sample size on the reliability of the estimated coeffi-
cients was performed (Ben-Akiva and Richards, 1975). It was
found that, for a multiple logit mode choice model with ten
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
coefficients, samples of less than 250 observations result
in very unstable coefficients. The standard errors of the
estimated coefficients decreased sharply by increasing
the sample size gradually from 100-150 to 250-300 observa-
tions. Beyond 300, the incremental reductions in the
standard errors with increasing sample size were much
smaller and gradually diminishing. Disaggregate models are
clearly very attractive from the perspective of data collec-
tion economics.
4. Explanatory Power
These models are specified using a greater variety of
explanatory behavioral variables than other available models.
The improved specifications result in increased explanatory
power and reduced uncertainty for forecasting.
5. Theoretical Validity
The assumption of a joint (or simultaneous) decision
process, made in the shopping and auto ownership models, is
more realistic than the sequential decision process implied
by the structure of other available models, both aggregate
and disaggregate.
6. Market Segments
The auto ownership model was estimated using nine mar-
ket segments. The other models also use a richer set of
socioeconomic variables than other existing models in the
field. This allows for a better understanding of how
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
various population groups will respond differently to a
given policy option.
7. Transferability
Because the coefficients of a disaggregate model are
free from aggregation bias, a disaggregate model can po-
tentially be applied to situations (i.e., geographical
areas) which are different from those on which it was
calibrated. However, even a disaggregate model can be
transferred only if it is not seriously misspecified.
These disaggregate travel demand models are based on the
most complete specifications among all similar models
that are known at this time. Therefore, we would expect
these models to produce reliable forecasts for other urban
areas in addition to Washington, D.C. for which the models
were estimated. Some tests that were carried out with
these models on data from New Bedford, Mass., Milwaukee,
Portland, and Los Angeles show that these models can be
used satisfactorily as estimated, or with minor adjustments. 11
8. Trip Frequency
For the shopping trips, the frequency of travel is not
assumed constant as in the UTMS trip generation models.
7. Conclusions
This paper has presented an overview of the major choices for
development of travel forecasting methods:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. In the past, the basic approach used for urban travel fore-
casting has been that of sequential, aggregate, determinis-
tic demand models with an indirect approach to calculating
equilibrium flows. This approach now is seen to have
serious deficiencies, especially when confronted with
today's emphasis on a broad range of transportation options.
2. Recent research indicates that there are at least eight
alternative ways to develop demand models and use them for
equilibrium prediction.
3. From the point of view of equilibration, a simple-step
direct approach is likely to be most valid and most effi-
cient in all except special situations.
4. From the point of view of demand model development, estima-
tion of models using disaggregate methods is much more
economical in terms of data requirements than aggregate
methods, and provides the opportunity for formulation
models soundly based on behavioral theory and estimated
using valid statistical approaches.
5. Numerous practical approaches to aggregation of disaggre-
gate models to produce an aggregated model for forecasting
are available.
6. It is feasible to estimate simultaneous choice disaggre-
gate models, and the results are different from the re-
sults of separately estimated sequential models. The assump-
tion of simultaneous choice is the soundest behavioral assump-
tion except when there are specific behavioral reasons for
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
hypothesizing a particular sequence of choices. For ex-
ample, in current work, one hypothesis being used is of
a simultaneous-block-sequential form:
A. simultaneous choice of residential location, em-
ployment location, and housing type
B. simultaneous choice of automobile ownership level
and mode to work, sequential to (conditional on)
the choices in (A)
C. simultaneous choice of non-work travel, including
frequency, destination, mode, time of day, and
route of nonwork trips, sequential to (conditional
on) the prior choices in (A) and (B).
Alternative behavioral hypotheses might be appropriate in dif-
ferent contexts.
As a consequence of the above, the author's conclusion is:
The preferred approach, in general is to:
1. Formulate a simultaneous or simultaneous-block-sequen-
tial form based upon behavioral reasoning in a parti-
cular context
2. Estimate the corresponding models in disaggregate si-
multaneous (and sequential, where appropriate) forms
3. Construct models for prediction by aggregating ex-
plicitly the estimated disaggregated models
4. Predict flow impacts of alternative transportation
options by equilibrating the resulting aggregate si-
multaneous models in a single-step direct approach.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Where specific situations suggest that short cuts may be accept-
able, use appropriate approximation methods. For example, proposed
changes in transit service frequency in a particular corridor may
be hypothesized as unlikely to affect choices of destination or
auto ownership level; therefore, a conditional mode choice model
could be derived from the simultaneous models and used in a sequen-
tial approach to equilibrium with trip generation and distribution
held fixed. However, for other changes, this hypothesis might be
considered unacceptable, and so the more general approach should
be followed.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Work Mode Split Model
Tables A.1 and A.2 contain the model specification (variables in-
cluded) and estimation results.
Interesting results were obtained for the implied impact of
changing auto availability and also for the importance of the "govern-
ment employee" variable. Both show considerable influence on the
modal split forecast by the model, and both have strong (significant)
coefficients. GW is a dummy variable in the car pool utility function
which is unity if the traveller is a civilian employee of the federal
government. Its positive sign indicates that, all else constant,
federal employees are more likely to choose carpooling than nonfederal
employees, which may be due to more potential for matching (i.e., more
opportunities to choose from and less search time), more active en-
couragement of carpooling by the employer, or more consistent start-
work and end-work times. An important implication of the strength
of this variable is that concerted efforts by large employers to en-
courage and facilitate carpooling can have a significant impact on
the proportion of work trips made by that mode. This variable, in
effect,'is a proxy reflecting the existence of some employer-based
carpool incentive programs. Therefore, in forecasting, this variable
can be used as a policy variable to indicate those employers who in-
stituted such programs by setting its value equal to one (or, if
sufficient empirical data exists, to some other value reflecting
the level of the employer's incentive program relative to the federal
government's).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Variable Code
3.. OPTC/INC
6. ` AALD
Approved For Rel 2S0AP1'KM9 IA 9-00798A000200020005-9
DEFINITION OF VARIABLES
Definition
time (in minutes)/one way'distance
?or shared rids..
1, if -worker is the breadwinner, for. drive 'alone
0, otherwise
1,'if work place is in the CAD, for drive alone
Al.. DCITYa
!2.
DINCc o
13 a INOM8
E4. )TECA8
Alternative:
1 1, for shared ride
0,. -otherwise.
cost (in cents)/household annual
round trip out-of-pocket travel
. income (in dollars)
round trip in-vehicle travel time (in minutes)
round trip out-of-vehicle travel
(in miles)
Q of autos/licensed drivers,`
0, otherwise
u of autos/ licennaed drivers,
0,Xtherwise
0, otherwise
1,' if work place
0, otherwise:
is in' the CBD, for
shared ride
household annual income - 800 A 0 of pcrsona in
(in $), for drive alone and nhared ride
00 .otherwise
0 of 'workers
0, otherwise
in the household, for shared rido
the household
employment density'at the work. zone'(cmployeca per commercial acre)
* one way distance (in miles), for shared ride
0, otherwise . . .
drive alone
shared ride (car pool)
transit
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2bd14/1JQ 2 CIA-RDP79-00798A000200020005-9
WORK MODE CHOICE MODEL
THE MODEL COEFFICIENTS
Drive alone constant
Shared ride constant
?.3. Out-of--pocket travel cost
divided by income
4. In-vehicle travel time
Out-of-vehicle travel time
.divided by distance.
'.Auto.availability
(drive alone only)
Auto availability
(shared r.id1n only)
OPTC/INC
1VTT
AALDa
17Z
Coefficient t-statistic
..3.24
Breadwinner (drive alone only) B '.890
11. CBD work place
(shared ride only) ?
12. Disposable income
(drive alone. and shared ride
only)
13. Number of Workera
(shared ride only)
14. lniployment density
(shared ride only)
of observations a 1114
0 of alternatives w 2924
GWa'.?
DCITYC
DCITY
s
DINC
cos
NWORY.B
DTECAo
-6.86
-5.60
10.08
-.404
-1.36
.0000706
3.46
.0983
1.03'
.000653
1.34.
Approved For Release, 2001/11/19 : CIA-RDP79-00798A000200020005=9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The auto availability variable occurs twice (with different co-
efficients) because of its mode specific character, once in the drive-
alone utility function, and once in the carpool utility function.
Both coefficients have positive signs and, therefore, indicate that
an increase in the ratio of autos to drivers in a household will in-
crease the likelihood of both driving alone and in a carpool, rela-
tive to using public transit. However, because the difference between
the coefficients for driving alone and for carpooling is also positive,
an increase in auto availability (all else equal) will raise the
probability of driving alone relative to that of participating in a
carpool.
The level of service variables: out-of-pocket travel cost (OPTC),
in-vehicle travel time (IVTT), and out-of-vehicle travel time (OVTT)
have significant coefficients with the expected negative signs and
relative magnitudes. For a typical five-mile commuting trip, the
ratio of the out-of-vehicle travel time coefficient to that of in-
vehicle time is approximately 2. This implies that a minute of out-
of-vehicle travel time is twice as onerous as a minute of in-vehicle
time. This result is similar to results obtained in previous studies.
The implied value of time for a five-mile commuting trip by an indi-
vidual whose annual household income is $8000 (in 1968 dollars) is
approximately $2.50/hr. for in vehicle time (and, of course, twice
that for out-of-vehicle time). Again, these figures correspond to
logic and the results of previous mode choice models.
The breadwinning variable is included for drive alone with a
positive coefficient, which represents the higher priority of the
primary worker for the use of any car that is available.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The CBD variable captures the added disutility of driving to
the CAD which is quite likely to differ between the two auto modes,
but nevertheless captures the disutility of driving in the CBD.
As expected, the coefficient of the variable for drive alone is
more negative than that for carpool.
Disposable income is included with one coefficient because when
separate coefficients were tried they were very similar. The number
of workers has been included to capture the effects of intrafamily
carpools. The final variable is an interaction variable that combines
the added convenience of finding a carpool partner if one works in
dense employment areas with the added incentive to carpool over longer
distances. This variable could be improved by collecting more data
on employer characteristics which were not available in the existing
data set.
The two modal constants, for drive alone (Dc) and for shared ride
(Ds) are negative. The larger absolute value of the drive alone pure
(dis)preference indicates it is relatively less desirable than car-
pooling, given that all other variables are constant across all modes.
Both drive alone and carpool are less desirable than transit (implied
zero constant) as far as the constant term alone goes, but the auto
availability variables pick up enough positive value when an auto is
available to reverse this relative effect for drive alone and render
carpool only slightly less desirable than transit.
174
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
APPENDIX B
Shopping Trip Frequency, Destination, and Mode Choice Model
The model specification and the estimation results are shown
in Tables B.1 and B.2.
As can be seen, all the important variables are significant at
the 99% confidence interval (t-statistic greater than 2.3). The co-
efficients of the level-of-service variables (time, cost) have the
expected negative signs and result in reasonable values of time.
For a shopping trip of 2.5 miles and a total round trip travel time
of 40 minutes, a minute of out-of-vehicle time is somewhat less
than twice as onerous as a minute of in-vehicle time.
The attraction variable, In (REMP), has the correct sign (the
higher the retail employment at a given shopping center, the more
attractive it is). The car constant, DC, has a negative coefficient,
reflecting that when the auto is used for a shopping trip it is not
available for other purposes. The CBD constant is positive, indi-
cating that the CBD is a more attractive shopping area generally.
The AAC variable coefficient is positive, meaning that the greater
the number of autos available, the more likely the traveller is to
make?a shopping trip and make it by auto. The positive sign on the
zero-frequency income variable INCF, is consistent with other results
and states that higher income households make fewer shopping trips
(i.e., they are able to maintain larger stocks of goods).
Of the other frequency variables, household size (HHSF) enters
with a negative sign indicating (as expected) that, for a larger
household, the probability of not making a shopping trip on a given
day is less. The density variable DENF is an attempt to account for
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001ABh19 :P3i -RDP79-00798A000200020005-9
SHOPPING JOINT TRAVEL DEMAND MODEL DEFINITION OF VARIABLES
Variable Code Definition
1. DC
176
2. OVTT/DIST round trip out-?of.Vehicle travel time (in minutes)/one way
distance (in wile-.)
3. 1VTT+OVTT round trip. ia-vehicle travel time -h round trip out-of-vehicle
travel time (in minutes) ?? -
4. OPTC/INC round trip out--of-pocket travel cost (in cents)/household
annual income kin code*)
0 of autos available to household - 0 of autos used for work
1, for car
0, otherwise
trips by workers in the household, for car
0, otherwise ? .
1/one way distance (in miles).
7. RE MP retail emn1nvment- at chnnn4nr. rinn4'--lead-4,.r. id.. /E ..o
8. DCBD 1, for CBD shopping destination
0, otherwise'
4 for zero frequency
0, otherwise
10. I1IISF
11. DENT
12. XNCF
f
0 of persons in the household,.for zero frequency
0, otherwise
retail employment density in residence zone (employees per
acre), for zero frequency
0, otherwise
household annual income (in c(?de*), for zero frequency
0, otherwise
41 ternala
No trip (zero frequency), variables (].) - (8) are equal to zero. Trip to
shopping destination d and by node m, for all relevant shopping :sestina-
Lions (including the CBD), and for car and transit modes.
Xncomc code (in 1968 $)
x 0 ?- 2,999
6 10,000
-- x1,999
2 - 3,000 - 3,999
7 a 12,000
- 14,999
3 4,000A: r
pp lip`s
~
or Release 2001/11/19 :8CFA-IkbF 0O079
00020005-9
4 6,000 V,,
9 - 20,000
- 24,959
5 8,000 -- 99999
. 10 - 25,000
+
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Table 'B.2
S1lbPI'r.NG JOINT THAVET. DEMAND 11OI)1:T.
- 177
THE MODEL COEFFICIENTS
Variable
.Coefficient
t-statistic
1.
Car constant
DC
- .555
- 2.13
2.
Out--of-vehicle. travel timel' ?'"'
divided by distance
OVTT/DIST _
- .100
- 3.38
3.
Total travel time
in (IVTT-l?OVTT)
-2.24
--11.85
4.
Out-of-pocket travel cost
divided by income
OPTC/ I nC
- .0242
- 4.20
5.
Autos available for non
work trips (car only)
AAC
.557
5.61
6.
Inverse of distance to
.destination
1/DIST
6.86
]..66
7.
Retail employe:: nt at
? dcstina t i on
In (REM?)
.161
3.29
8.
CBD destination constant
DCBD
.562
2.07
-3.78
- ?4.51
9.
Frequency zero constant
10.
Household size (frequency
zero only)
- .186
- 4.57.
11.
Retail. employment density at
the origin (frequency zero
only)
.000598
1.38
12.
Household income (frequency
.0414
1.18
zero only)
of observations n 1313
{- of alternative? ?~ 44718
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
walk trips to shopping, which are not recorded in the Home Interview
Survey. This variable measures the density of retail employment in
the home zone, and thus is a proxy for the availability of suitable
shopping destinations within walking distance of the home. The posi-
tive sign of the DENF coefficient means that a household living in
a zone with dense retail employment is less likely to embark on a
vehicular shop trip (i.e., is more likely to choose a walk shop
trip instead).
178
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
APPENDIX C
Auto Ownership Model
Table C.1 lists the variables included in the model. The sub-
scripts for each variable indicate the alternatives for which the
variable enters into the utility function. Variables with no sub-
scripts enter into every utility. Seven constants (one for each
alternative except the zero autos-transit to work alternative)
measure a constant relative effect which the remaining variables
do not pick up. These variables represent what might be termed a
'pure alternative effect."
The variables AALDc and AALDs measure the number of autos per
licensed driver, thereby modeling the auto availability in the mode
choice decision. The more autos per driver a household has, the
greater the probability the primary worker will use a car to go to
work; hence, a positive sign would be anticipated.
The variable denoted as Z requires some explanation. It repre-
sents the discretionary income left to a household after transporta-
tion and fixed household expenditures. The rationale underlying the
use of this variable is quite simple. The total utility a household
derives from any auto ownership-mode choice selection depends on the
attributes of the alternative, one of which is the resources remaining
to the household for other uses, if that alternative is selected. The
use of the Z variable is based on the theory that the marginal utility
of money at a given point on the income curve is the same for all cate-
gories of household expenditures. While it is theoretically possible
to incorporate a fuller set of alternatives and thereby eliminate
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
iariable Code
'L Dot
AALDc
'0..
AALDa
11T2
.IVTT
.3. OVTT/DIST
.4 . "LD
.5. fil
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Table C.1
AUTO OWNERSHIP MODEL
DEFINITION OF VARIABLES
Definition
1, for zero autos; shared ride
0,' otherwise
It for one auto; drive alone;
0,.otherwise
1, for one auto; shared ride''
0, otherwise
it
.0, otherwise
1, for two autos;
0, otherwise
( 1, for two autos; shared ride
L 0, cthcr::ioc
1,' for two autos;. transit
.0, otherwise
R # of autos/licensed drivers, for drive alone
0, otherwise
# of auto's/licensed drivers, for shared ride
0,' otherwise
household annual income - 800 * # of peroons.in the household
1000 * # of autos - 250 * daily round trip travel cost (in '$.)
5 1, if household liven in a c'inglo family house, for two autos
0, otherwise
daily round trip in-vehicle travel time (in mir&uLes)
daily round trip out--ofwehitlc travel time '(in minutes)/one way
uiuGance kin m].J.cs)
#- of autos/licensed drivers.'
car f;cneralizcd, chopping travel cot-t/transit generalized chopping
travel cost, for 'one auto
0, othcrwise
Approved For Release 2001/11.11'9.:'CIA RRDP79-00798A00020.0020005-9
for one'auto; transit
Variable Code
.7. DCITYo
.8. DCITYa
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Table C.l(continued)
1, if worker is a civilian employee of the federal government
car generalized chopping travel cost/transit gencralized chopping'
travel cost, for two autos
0, otherwise
1, if work place is in the CDD, for. drive alone
1 0, otherwise
1, if work plat-'is in the CBD, for shared ride
0, otherwise
250 * daily round trip out-of--pocket travel 'cost.
or shared ride.'
for shared ride
0, otherwise
D of workers in the household,
0. otherwise
employment density at the work zone -(employees per commercial
Alternative
acre) * one way distance (in tiles), for shared ride
fl_ nthprwiap
zero autos; shared ride to work
zero autos; transit to.work
one auto; drive alone to work
one auto; shared ride to work
one auto; transit to work
two + autos; drive alone to work
two h autos; shared ride to work
two + autos; transit to work
Approved For Release 200.1/11/19: CIA-RDP79?-00798A000200.020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
possible expenditures on commodities not explicitly modelled, it is
undesirable to do so in a practical model.12 Hence, some relatively
simple means of collapsing the entire set of ignored consumption
choices is required.
This is precisely what the Z variable does. It incorporates
other choices into a term which represents the income remaining to
a ,household after certain fixed household and transportation expendi-
tures. It thus provides a way of including transportation costs
versus capital and operating--directly in the model.
An alternative with zero auto ownership results in a high value
of Z, reflecting the availability of income the household would have
to allocate to the purchase, maintenance, and operation of a car if
it had chosen to do so. Thus, the coefficient of the Z variable in
the utility function should have a positive sign.
The value of the Z variable was also used to evaluate whether or
not a household had sufficient income to preceive an alternative as
being available. Any alternative for which Z was less than zero was.
excluded from the household's choice set.
HT2 is a 0,1 dummy variable representing single family residence.
It was hypothesized that single family residence reduces parking diffi-
culties and increases the need for multiple automobile ownership for
social and recreational trips. The dummy variable only applied for
multiple car alternatives.
In-vehicle travel time, out-of-vehicle travel time, and out-of-
pocket travel cost are self-explanatory. These variables enter the
model with the same coefficients for the two modes. This is based on
the results obtained by ORA (1972) and Ben-Akiva and Richards (1975).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The variable AALD is the number of autos divided by the number
of licensed drivers in the household and measures the need for addi-
tional autos in the household. As the number of autos per driver
decreases, the need for multiple car ownership should increase. The
difference between AALD and the mode specific AALDc and AALDs variables
is that they appear in different equations as measures of different
effects, the former an incentive for owning more cars and the latter
an auto availability effect for mode choice. One would anticipate
that these variables are highly colinear in the data set and that the
estimates of their coefficients would have high standard errors.
ever, this did not turn out to be the case. The reason for this is
that the variables are defined to be zero for different alternatives.
AALD is set at zero for alternatives (1) and (2) only. The modal
c
choice effect AALDc and AALD8 indicated increased car mode preference
when autos per licensed driver increased, while the auto ownership
effect (AALD) indicated increased auto ownership preference when
number of licensed drivers increased. The estimates of the coeffi-
cients of AALD and AALDC were highly significant.
The next two variables, R1 and R2, incorporate the effects of
shopping trip accessibility and thus represent nonwork travel oppor-
tunities. These variables were developed by forecasting the attri-
butes of expected (or average) shopping trip for each household by
both car and transit. This was done using a previously estimated
simultaneous shopping destination and mode choice model estimated
by Ben-Akiva (1973). The attributes of these expected trips (travel
time, cost, etc.) were then weighted using the coefficients of the
previously estimated model. Hence, the ratio of car cost to transit
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
cost was taken in order to measure the relative accessibility of
car shopping travel to that of transit. Note that when transit
is not available to a household for shopping travel, the ratio is
zero (i.e., infinite transit cost).
Expressed mathematically, the expected value of attribute i
for a given mode in, Xim, is as follows:
E[Xim] - E P(dlm) Xidm'
deD
where D is the set of shopping destinations which can be reached
by mode m and P(dlm) is the probability of making a shopping trip
to destination d given that mode m is used. The generalized price
by mode m, GPm, is therefore the following:
GPm - al E[IVTTgI + a2 E[OVTTa] + a3 E[OPTCmI
where E[IVTT m ] is the expected shopping in-vehicle travel time by
mode m;
E[OVTTm] is the expected shopping out-of-vehicle travel time
by mode m;
E[OPTCm] is the expected shopping out-of-pocket travel cost
by mode m; and
al,a2,a3 are the parameters of IVTT4, OVTTm and OPTC., respec-
tively, from a previously estimated shopping travel
demand model.
Finally, the ratio, R, is simply defined as
GP
car
GP transit
The concept of generalized prices was used by both CRA (1972)
and Ben-Akiva (1973).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The.CBD dummy variables are used to reflect the disutility
typically associated with downtown car usage which is not entirely.
measured by travel times and cost. Effects such as high travel
time variance, the frustration drivers experience in congestion,
and the unreliability of on-street parking are all incorporated
into these variables.
. Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 .
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
186
1. M. Beckman, C. B. McGuire, and C. B. Winston, Studies in the
Economics of Transportation, New Haven: Yale University Press
(1956); G. Kraft and M. Wohl, "New Direction for Passenger
Demand Analysis and Forecasting," Transportation Research
Vol. 1 (1967), pp. 205-230; M. Wohl and B. V. Martin, Traffic
Systems Analysis, New York: McGraw-Hill (1967), Chapter 5;
M. L. Manheim, "Search and Choice in Transport Systems Planning,"
Highway Research Record 293, Washington, D.C.: Highway Research
Board (1969).
2. Manheim, op. cit.
3. We deal here with only one of "Wardrop's Principles"; we will
not discuss the other, concerned with global optimization of
the flow pattern, as this is inapplicable to urban transporta-
tion flow prediction. Cf. J. G. Wardrop, "Some Theoretical
Aspects of Road Traffic Research," Institute of Civil Engineers,
Road Paper No. 36, London (1952); M. Beckmann, "On the Theory
of Traffic Flow in Networks," Traffic Quarterly (January, 1967);
W. S. Jewell, "Models for Traffic Assignment," Transportation
Research Vol. 1, No. 1 (May, 1967).
4. B. V. Martin, F. W. Memmott and A. J. Bone, Principles and
Techniques of Predicting Future Urban Area Transportation,
Cambridge, Mass.: M.I.T. Press (1965); various publications
of the U.S. Federal Highway Administration on Urban Transpor-
tation Planning.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Notes (continued)
5. In the equations which follow, standard conventions of prob-
ability theory are used. The relevant concepts, explained
briefly here, are described in any standard text on probability.
P(A) is the probability that a random event, A, occurs.
The joint probability P(A,B) is the probability that two
random events, A and B, both occur.
The conditional probability P(AIB) is the probability of event A
occuring when it is known that event B occurs.
6. The conditions are that f 1,f2,f 3, and f4 in (4.5-3) be "internally
consistent" as defined by Manheim (1973).
7. Adapted from Earl R. Ruiter (1973), HRB Special Report 143;
Cambridge Systematics, UMTA Travel Forecasting Manual; and
Marvin L. Manheim, Fundamentals of Transportation Systems Analysis.
8. Usually per day.
following code was used:
Household Income
(in $1,000/year)
Code
0 - 3 1
3 - 4 2
4 - 6 3
6 - 8 4
8 - 10 5
10 - 12 6
12 - 15 7
15-20 8
20 - 25 9
greater than 25 10
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Notes (continued)
10. Binary work mode choice models have been developed by many re-
searchers; this model is described here because of its relation-
ship to the simultaneous models described following, and be-
cause it is the first disaggregate model to treat carpools
as an explicit mode.
11. A variety of procedures to adjust a model to a different area
are developed and evaluated in Atherton (1974).
12. A more formal case for the use of the Z variable can be made by
utilizing the concept of a utility "tree," in which a household's
utility function is considered to be additively separable. See
Strotz (1957) and Cambridge Systematics (1974).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
REFERENCES
Adler, T. and M. Ben-Akiva. 1974. "A Joint Frequency, Destination
and Mode Choice Model for Shopping Trips," paper submitted to the
Transportation Research Board.
Aldana, E., Richard deNeufville, and Joseph Stafford. 1973. "Micro-
analysis of Urban Transportation Demand." Highway Research Board
1973.
Atherton, Terry. 1974. "A Bayesian Approach to Transferring Stochastic
Disaggregate Travel Demand models," unpublished S.M. thesis, M.I.T.
Beckmann, M.J., C. B. McGuire, C. B. Winston. 1956. Studies in the
Economics of Transportation. Yale University Press.
Beckmann, M. J. 1967. "On the Theory of Traffic Flow in Networks."
Traffic Quarterll. January.
Ben-Akiva, Moshe E. 1973. "Structure of Passenger Travel Demand
Models." Ph.D. dissertation, M.I.T.
Ben-Akiva, Moshe E. and M. G. Richard. 1975. "A Disaggregate Multi-
Modal Model for Work Trips in The Netherlands." Transportation Research
Board.
.Buro Coudappel en Coffeng and Cambridge Systematics, Inc. 1974.
Disaggregate and Simultaneous Travel Demand Models: A Dutch Case Stu y.
Cambridge, Mass.
Cambridge Systematics, Inc. 1974. Introduction to Urban Travel Demand
Forecasting, a tutorial manual prepared for UMTA by Cambridge Systema-
tics, Inc., March.
Cambridge Systematics, Inc. 1974. Multinomial Logit Estimation Package:
Program Documentation. Version 2, Mod 1. Cambridge, Mass.
Cambridge Systematics, Inc. 1974. Series of auto ownership technical
memoranda.
Charles River Associates, Inc. (CRA). 1972. A Disaggregated Behavioral
Model of Urban Travel Demand. Federal Highway Administration, U.S.
Department of Transportation, Washington, D.C.
Charles River Associates, Inc. 1967. A Model of.Urban Passenger Travel
Demand in the San Francisco Metropolitan Area. Report Number 117-1.
Prepared for Peat, Marwick, Livington & Co., December.
Dafermos, S. C. 1972. "The Traffic Assignment Problem for Multi-class
User Transportation Networks," Transportation Science. vol. 6, No. 1,
F!pproved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
190
. 1971. "An Extended Traffic Assignment Model with Appli-
cation to Two-Way Traffic.." Transportation Science. vol. 5, No. 4.
November.
? 1969. "Network Models for Traffic Assignment." Cornell
University Department of Operations Research, Technical Report No. 90.
Ithaca, New York, October
Dafermos, S. C. and F. T. Sparrow. 1969. "The Traffic Assignment.
Problem for a General Network." Journal of Research National Bureau
of Standards-B, vol. 73B. No:_2.
Dial, R. B. 1971. "A Probabilistic Multipath Traffic Assignment
Model which Obviates Path Enumeration." Transportation Research,
vol. 5.
Domencick, T. A., et. al. 1968. "Estimation of Urban Passenger
Travel Behavior: An Economic Demand Hodel." Highway Research Record
238. Washington, D.C. Highway Research Board.
Dunphy, Robert T. and George V. Wickstrom. 1972. "Accessibility:
A Tool for Spatial Analysis." Presented at the 1972 Annual Conference
of the Urban and Regional Information Systems Association. San Francisco.
Ellis, Raymond H. 1966. A Behavioral Residential Location Model (1966)
MS thesis, Northwestern University, June.
"Functional and Detailed Specifications for a Highway Network Equilibrium
Program." Prepared for UNTA Planning Methodology and Technical Support
Division by Cambridge Systematics, Inc., Cambridge, Massachusetts
(forthcoming).
Hudson, James and Steven Lerman. 1974. "Sensitivity Analysis of
Equilibrium Flows in Congested Transportation Networks. Paper pre-
sented at O.R.S.A/T.I.M.S. meeting, Boston
Jewell, W. S. 1967. "Models for Traffic Assignment." Transportation
Research vol. 1, No. 1, May.
Koppelman, Frank. 1974. "Prediction with Disaggregate Models: The
Aggregation Issue." Paper in Highway Research Record.
1975. Ph.D. dissertation, M.I.T.
Kraft, G. 1963. "Demand for Intercity Passenger Travel" in the
Washington-Boston Corridor. Systems Analysis and Research Corporation.
Kraft, G. and M. Wohl. 1967. "New Direction for Passenger Demand
Analysis and Forecasting." Transportation Research vol. 1.
Landau, Uzi. 1975. "Strategic Analysis Models in Transportation
Systems Planning." Ph.D. dissertation in progress, M.I.T.
Lerman, Steven and Nigel Wilson. "An Analytic Model for Predicting
Dial-A-~oJSe~' 'F~ a~ M1/1~ ~4( 20( 9 ation
Researc oar as ngton, D.C.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Lisco, Thomas E. 1967. "The Value of Commuters' Travel Time: A Study
in Urban Transportation." Ph.D. dissertation, University of Chicago. ,
Manheim, M. L. 1975. Fundamentals of Transportation Systems Analysis.
Preliminary Edition, Third Revision.
1973. "Practical Implications of Some Fundamental Properties
of Travel Demand Models." Highway Research Record 422. Washington, D.C.
"Search and Choice in Transport Systems Planning." Highw
Research Record 293, Washington, D.C.
Manheim, M. L. and E. R. Ruiter. 1970. "Dodotrans I: A Decision-
Oriented Computer Language for Analysis of Multimode Transportation
Systems." Highway Research Record 314.
Martin, B. V. and M. L. Manheim. 1965. "A Research Program for
Comparison of Traffic Assignment Techniques." Highway Research Record 88.
Martin, B. V., F. W. Memmott, and A. J. Bone. Principles and
Techniques of Predicting Future Demand for Urban Area Transportation.
M.I.T. Report No. 3.
McFadden, Daniel. 1974. "The Measurement of Urban Travel Demand."
Journal of Public Economics, No. 3.
McGillvray, Robert G. 1972. "Binary Choice of Urban Transportation
Mode in San Francisco." Econometrics. vol. 40.
McLynn, J. M. and T. Woronka. 1969. "Passenger Demand and Modal
Split Models." Northeast Corridor Transportation Project, Report
230, U.S. Department of Transportation.
Murchland, J. D. 1969. "Road Network Traffic Distribution in
Equilibrium." Paper for the Tagungueber "Mathematische Methoden
in den Wirtschaftswissenschaften." Mathematisches Forschungsinstitut.
Oberwolfach, October .
Nestle and Young. 1974. "Short-term Modifications to the Existing
CALTRANS Model System to Accept Supply-Sensitive Assignment Techniques
and Policy-Sensitives/Demand Models." Working discussion paper, M.I.T.,
January.
Putnam, Stephen H. 1973. The Interrelationships of Transportation
Development and Land Development. Volumes I and II, Report prepared
for the U.S. Department of Transportation, June.
Quandt, R. E. and W. J. Baumol. 1968. "The Demand for Abstract
Transport Modes: Theory and Measurement." Journal of Regional Sciences.
vol. 6, No. 2.
Quarmby, D. A. 1967. "Choice of Travel Mode for the Journey to Work:
Some Findings." Journal of Transport Economics and Policy. vol. 1,
No. 3, pp. 1-42
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Reichman, S. and P. R. Stopher. 1973. "Disaggregate Stochastic
Models of Travel Mode Choice." Highway Research Record 369.
Washington, D.C.
Ruiter, E. R. 1973. "Analytical Structures" in Highway Research
Board Special Report 143 . Washington, D.C.
. 1973. "The Prediction of Network Equilibrium: The State
of the Art," in Proceedings of the International Conference on Trans-
portation Research. Held in Bruges, Belgium: Transportation Research
Forum, June.
Ruiter, E. R. and J. Murchland. 1974. Report to UMTA on equilibration
efficiencies, in progress. Cambridge Systematics, Cambridge, Mass.
Stopher, Peter. 1969. "A Probability Model of Travel Mode Choice for
the Work Journey." Highway Research Record No. 283. Washington, D.C.
Stopher and T. Lisco. 1970. "Modelling Travel Demand: A Disaggregate
Behavioral Approach, Issues and Applications. Transportation Research
Form Proceedings.
Strotz, R. H. "The Empirical Implications of a Utility Tree."
Econometrica, vol. 25, No. 2, April.
Systems Analysis and Research Corporation. 1963. "Demand for Intercity
Passenger Travel in the Washington-Boston Corridor," Report to U.S.
Department of Commerce.
Talvitie, Antti and Tom Leung. 1974. "A Parametric Access Network
Model." Paper in draft, Oklahoma.
Turnquist, Mark. 1972. A Cost Model for Incremental Traffic Assignment.
M. S. thesis, M.I.T.
UMTA Transportation Planning System Reference Manual. 1973. Urban Mass
Transportation Administration.
UMTA Transportation Planning System (UTPS) Reference Manual. 1974.
UMTA Planning Methodology and Technical Support Division. Washington,
D.C., August.
Wardrop, J. G. 1952. "Some Theoretical Aspects of Road Traffic
Research." Proc. Institute Civil Engineers. Part II, Road Paper #36.
London, pp. 325-378.
Wilson, Nigel, Wayne Pecknold, and Brian Kullman. 1972. "Service
Modification Procedures for Local Bus Operations of the Massachusetts
Bay Transportation Authority." The Boston Urban Observatory.
Wohl, M. and B. V. Martin. 1967. Traffic Systems Analysis. New York:
McGraw-Hill.
Yamamoto, Warren. 1973. A Comparative Analysis of Transit Alternative
for Non-Work Tri s. M. S. thesis. M.I.T.
prov or Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
EXAMPLES OF COMPUTER APPLICATIONS
IN THE TRANSPORTATION FIELD
Faye C. Johnson
The transportation sector of our economy has integrated the
computer into their operations so profoundly that today we find
this sector to be one of the most sophisticated and advanced users
of computers in the United States. For years American businesses
have used the computer in their accounting and administrative func-
tions but only in the last two decades have they put the computer
to use in the very heart of their business, i.e., operations. We
find in airlines that they now keep all of their passenger reser-
vations on a computer with remote terminals at airports and ticket
agencies throughout the United States. They also make extensive use
of computers in their maintenance, crew scheduling, and airline
ticketing. Railroads, we will see, control their entire car fleets
with the use of computers as well as performing very sophisticated
analytical simulations to run their operations more efficiently.
In motor freight we find their entire rolling stock being tracked
by computers as well as their on-line billing and rating.
This paper will address some of the more advanced computer
applications in only two modes of our transportation sector and,
more specifically, to the hauling of freight in these two modes.
These modes are railroads and motor freight companies. The inter-
dependence of our nation's economy on the transportation sector is
apparent to everyone associated with transportation. Overall,
transportation expenditures, including material inputs to the
production of transportation and services, represent about 18%
*Corp.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
of the United States gross national product and we find these two
sectors hauling over 60% of the total freight ton miles by our
intercity freight carriers. These figures include both for hire
or common carriers and private traffic. Even more significant,
we find that railroads and motor freight generate over 94% of all
the gross operating revenues from the transportation of goods among
the regulated freight carriers.
The fastest growing of these two segments is the motor freight
common carrier. With the invention of the internal combustion engine
at the dawn of this century, the truck transportation industry has
grown in the last thirty years from a very minor entity, in terms
of revenue, to the largest single sector of all regulated freight
carriers. Several factors can be cited as having contributed to
the rapid growth of the trucking industry. These are:
1. The suburbanization of many of the nation's population
.and employment centers
2. Construction of the interstate highway system
3. The inherent flexibility and convenience of trucks
as opposed to other transportation modes
When one looks at the application of computers in these two
industries one needs only to look at how these companies spend
their available money in the operations of their businesses.
Listed below are the percent of expenditures on operating expense
of all the motor freight companies in the United States as of 1973.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
o Transportation--46%
o Terminal operations-22%
o Equipment maintenance--9%
o Traffic--3%
o other expenses--10%
o Insurance and safety--4%
o Administration and general--6%
In railroads as of the end of 1973 we find the following breakdown:
o Transportation--50%
o Maintenance of equipment--21%
o Maintenance of way--18%
o General expense--7%
o Traffic--3%
o Miscellaneous--l%
There is no wonder that both the railroads and motor freight
use the computer more extensively to control their transportation
expenses than in any other function of their business.
The major difference between the two charts is that while rail-
roads include their yard operations as a part of their total trans-
portation expense, motor freight separates their terminal expense
from their transportation expense.
I would now like to address the use of computers in some leading
edge applications for the railroad industry.
As mentioned previously, the railroad industry has attained in
the last twenty years the status of being one of the most sophisti-
cated users of complex data processing equipment for operational
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
real-time applications. This position has evolved over several
decades whereby they, like many other industries, mechanized many
of their batch-related applications through the use of data processing.
Examples of these batch applications presently installed by most of
the major U.S. railroads are:
o Waybilling
o Revenue accounting
o Centralized payroll
o Car accounting
o Disbursements accounting
o Freight claims
o Material management
o Stockholders records
o Reports to shippers on car status and location
These early batch applications, at the time of their development
and implementation, were pressing the state of the art in computer
technology and capabilities. A gating factor was computer capability
and not railroad requirements. Early data processing, utilizing
unit record equipment, had no teleprocessing capabilities and some
of the very early computers also lacked this functional capability.
When teleprocessing equipment became available, the railroads were
among its very first users exchanging information among railroad
yards, agencies, and their home office. It wasn't until the third
generation of computers became available in the 1960s that the rail-
roads were able to apply these management tools to the very heart
of their business, i.e., operations.
Approved For Release 2001/11/19 : CIA-RDP79-00798A0002000.20005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
We have seen the railroads increase their operating efficiencies
over the last thirty years by many means. Some of the more important'
are:
o Dieselization
o Installation of automatic classification yards
o Installation of microwave communication systems
o Extensive applications of computers
It is considered by many that the latter may have more impact
on greater efficiencies in operations than any of the other develop-
ments in the last three decades. It is apparent, with over 1,700,000
freight cars in the U.S., travelling over many thousands of miles
of track, the utilization of this valuable asset becomes a major
challenge of railroad management.
Today we find that a loaded car is moving only 11% of the time
towards its destination. The rest of the time it is empty or
sitting in a railroad yard, a repair shop, or at a customer siding
being loaded or unloaded. Obviously, improving this 11% payload
utilization can mean significant returns to the industry. To be
more specific, assuming an average railroad car costs $25,000 each
and with 1,700,000 cars in our national inventory, for every 1%
increase in the utilization, this industry can have capital expen-
diture avoidance of $425,000,000. Another way of expressing it--
it is equivalent to adding 17,000 additional freight cars to its
inventory.
Improving this utilization is the central objective of.the
use of computers in the railroad industry today. These computer
mechanization activities include such applications as:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. Car Movement Reporting--Keeping the exact status and loca-
tion of all railroad cars and power on its system and re-
porting any change of this status to the appropriate
points within the railroad that has need of this.infor-
mation
2. Scheduling of--Cars--The allocation of available empty cars
to shippers, the disposition of empties, and the determina-
tion of an appropriate sequence of trains for transporting
each loaded and unloaded car from origin to destination
3. Scheduling of Locomotives--The allocation of power in
accordance with train schedules, tonnage, and maintenance
requirements
4. Scheduling of Trains--Determination of departure and arrival
times for each scheduled train; determination of when to
operate "extra" trains and to cancel scheduled trains
5. Emergency Responder -Response to systems outages, e.g.,
derailments, locomotive breakdowns, electrical power and
signal failures, bridge washouts, etc.
6. Scheduling of Track (Dispatching),--Control of train movement
on all main and branch lines
7. Scheduling of Roadway Maintenance--Allocation of personnel
and equipment for inspection, maintenance, and repair of
system track and related facilities
8. Scheduling of Car Maintenance--Determination of time and
place for inspection, maintenance and repair of freight cars
9. Scheduling of Crews--Selection of trained operating crews
in accordance with work contracts, availability, train
schedules, and assignment
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
While this is a very ambitious. application list for individual rail-
roads, we find that many of them today are well on their way in
their development and implementation. To begin this scenario of
applications, a railroad must start with the very complex real-
time application--car movement reporting. The objectives of this
application are:
1. Reduce clerical burden in yards, agencies and, in some
cases, the general office
2. Provide management at all operating levels with complete,
accurate, and timely information for improved control
and utilization of cars, locomotives, and terminal
facilities
3. Establish an historic data base for the eight future real-
time applications noted previously
The first phase of car movement reporting can be categorized as
the data collection phase. Through the use of lease lines or micro-
wave with data'processing terminals installed at several hundred
strategic locations throughout the railroad network, many events
are reported to a central processing unit as they occur. Some ex-
amples of these events are:
. Cars interchanged
. Cars to and from industry tracks
Cars assigned to arriving and departing trains
Cars loaded and emptied
. Cars bad-ordered
Locomotive assignments and status
. Crew assignments
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
After a network is established tto gather this data, then reports
are transmitted to appropriate terminals. Most of these reports
serve the operating department. Reports generally give status of
a car, a group of cars, a train, a yard, etc., or performance of
a yard, group of trains, etc. Examples of these reports are as
follows:
. Detailed and blocked train consist
? Cars which have high, wide, or heavy loads
Car inquiries for location and status for manifest
information and for advanced car tracing
Speed-restricted cars
? Car listing by location
List of trains by route
. Power and caboose reports--on hand and arriving
in yards
Locomotive maintenance due
Cars delayed in a yard
Terminal performance reports
. Yesterday's train performance
Current situation of yard's train and power
After these types of reports are generated, management can
issue the following instructions.
Phase I
Train schedules
Policy governing blocking rules
Movement of power from one location to another
Instructions to hold a car or divert a car
Special and standing instructions
ApproV?{- eReW*ME06fl/tkjt9; (eU P#9e0019S @ P006,46005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
After the data collection. phase is established, other phases
can take place such as:
Phase II
Traffic functions
Reports for shippers
Advance car tracing
Phase III
. Pool inquiries
. Monitor train blocking and train reports
Phase IV
Dispatcher inputs
Initial car distribution functions
Yard summary reports
. Revenue data input
. Additional train reports
Phase VI
. Power, caboose, and crew functions
Car orders and complete car distribution functions
and reports
Phase VII
Phase VIII
Other off-line applications
Car, locomotive, and caboose statistics
Demurrage accounting
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 202
The car reporting application establishes the data base which
is necessary for the railroads to commence the development and im-
plementation of some very exciting additional applications which
will further increase the efficiency of their rolling stock. I
am categorizing these additional applications under the generic
term; railroad operations control. There is widespread agreement
that there are two solutions associated with railway freight ship-
ments that can go a long way towards increasing the efficiency of
their operations.
1. The ability for the railroads to provide shippers with
empty cars for loading of the proper type and grade when
they are required and in sufficient quantity.
2. The establishment of dependable and consistent shipper
to consignee transit time.
It is clear that the optimum utilization of freight cars requires the
solution of extremely complex mathematical scheduling problems. It
is equally clear that concepts of mathematical optimization need
to be adopted in a systematic manner by the railroad industry in
its attempts to solve the varied and interrelated functions of
freight operations.
Until quite recently the technology required to formulate,
solve, and implement the solutions to the highly complex problems
associated with freight operations has not been available.
The necessary ingredients of this technology include telecommuni-
cations, on-line high speed data processors having suitable large
memory capacity, remote terminals for entry and receipt of control
information, as well as efficient algorithms for solving mathematical
models in a quasi real-time environment. Much of this technology is
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 203
now available. Accordingly, railroads are now taking a fresh look
at the railroad operation control system in an effort to achieve
maximum benefit for both the railroad industry and its users.
I previously mentioned eight scheduling functions associated
with railway operation control and, to repeat, these are:
? Scheduling of cars
. Scheduling of yard operations
? Scheduling of locomotives
? Scheduling of trains
. Scheduling of track
? Scheduling of roadway maintenance
? Scheduling of car maintenance
? Scheduling of crews
While all aspects of freight operations contribute to the
quality of service railroads may provide, significant relief
of the primary chronic problems may be expected to result in the
application of modern technology initially to the scheduling func-
tion associated with cars and yard operations. Car scheduling,
of course, embraces the movement of both loaded and empty cars.
Clearly, these two subproblems are highly interrelated in that
typically train traffic is a combination of loads and empties.
To facilitate understanding of the special problems of handling
of empty cars, it will be helpful to describe briefly the essential
scheme for moving cars across a railroad way network.
Freight cars must be moved from their known origin to a
prescribed or computed destination (it may be assumed herein that
origin and destination are contained within the same railway network).
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
To accommodate the multiplicity of diverse origin/destination pairs,
an operating plan has evolved over the years based upon a hierarchy
of trains and strategically situated classifying yards. Cars are
switched in the classifying yards into blocks having a common inter-
mediate or final destination in accordance with a railroad blocking
policy. These blocks are placed on scheduled trains and transported
toward their ultimate destination. The classification policy of a
railroad specifies which blocks of cars may be carried on each
scheduled train, the order of the blocks, and the handling of the
blocks en route. Frequently, the railroad is forced to deviate
from its operating plan because of surges in traffic, mechanical
failure of equipment, or temporary loss of supporting facilities,
i.e., main line, bridges, electrical power, etc. Normal operation
is further confounded by system limitations on basic resources such
as power, train length and tonnage, crew availability and work rules,
yard switching capacity and mainline availability. Deviations from
normal operation often trigger cascading effects of significant
magnitude.
The input to the car scheduling subsystem would be empty and
loaded cars released by users within a given railway network or
received at interchange. Output from the system would be car move
orders which, in essence, represent trip plans for the movement of
each car through the railway network. This system, as I mentioned
previously, is supported by their comprehensive car movement re-
porting and monitoring systems which created the required data base
for car scheduling.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The car scheduling subsystem would contain three major func-
tions.
1. Empty Car Disposition--Determines whether an on-line empty
car may be considered a candidate for loading. In many
instances, regulations prohibit or limit loading of empties
and, in effect, prescribe the destination to which a re-
stricted empty must be moved.
2. Empty Car Allocation--Determine the optimum destination
for each loadable empty according to the overall system
supply, demand, and selected decision criteria.
3. Traffic Scheduler--Determines a suitable trip plan for all
necessary car movements in generating appropriate car move
orders.
Certain of the above types of problems cannot be resolved within
an empty car allocation model or subsystem. However, these problems
must be addressed within an overall freight car management system
in order to provide necessary support for empty car allocation.
At present, railroads allocate empty cars in a manner similar
to the following:
A shipper phones his order to a freight agent. If the order
can be filled locally with unassigned empties, this is done.
Otherwise, the order is transmitted to a division car dis-
tributor. These individuals maintain, to a limited extent,
a record of the required number and type of empties needed
for their division, along with anticipated releases. According-
ly, the division car distributor knows generally what empties
are available for allocation in the shipper's vicinity. He
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
calls an appropriate yard and places a suitable order
for the necessary empties. If a division car distri-
butor is unable to fill an order, he passes it on to the
office of the system manager of car utilization. Cer-
tain individuals in this office are responsible for allo-
cating specific types of cars, e.g., open top cars (plain
flats, coal hoppers, gondolas and chip hoppers). Each
week, the system car distributor's office issues a docu-
ment which contains instructions for disposition of all
system and foreign equipment. Daily modifications to
this document may be transmitted by telephone.
In a similar fashion, each yard creates a daily disposition sheet
which is distributed to car agents, car inspectors, and yard masters.
Generally, empties have a lower priority than loads in that they
will be removed first from a ,train that is too long or heavy. This
practice contributes to local yard congestion and may cause shortages
at other yards.
In order to provide an effective way to utilize a computer-based
technology for the assignment and control of empty railroad freight
cars, it is necessary to model the problem analytically. This can
be done by applying the resource allocation methodology of operations
research.
Factors to be Considered
There are a number of factors which should be considered in the
design and development of an optimum-seeking allocation model. These
include criteria relating to the accuracy and relevance of the model
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
within its functional context as well as within its operational con-
Various railroad policies should be accommodated. Among these
Shortage sharing
Car service rules
. Regulatory constraints
. Per diem minimization
A given customer must ultimately receive a specific car. While
one can devise models to handle the assignment of individual cars,
it is unclear that this is the most effective approach. It may be
more appropriate to group cars into specific "classes" based on car
type, condition, special equipment, etc., and assign them to yards
based on this classification. The specific assignment of individual
cars to nearby customers may be resolved at the yard level, either
automatically (by use of a more detailed optimization model), or
manually (perhaps with the aid of visual displays), or a combina-
tion of both.
Overall control of cars would still be maintained on a systemwide
basis by use of the allocation model on a "global" scale, with the
specific cars assigned to customers selected from those cars globally
allocated to a yard in the customer's vicinity. It would appear
that such a two-level approach (aggregating at the higher level and
"exploding" the aggregated solution at the lower level) is more
reasonable.
This last point relates as well to the operational use of the
model. To be effective, the model must be used regularly on a semi-
reaAMM"After R ns~eE2 1kflUl tcf ilpei~[ d~~-(}~ ,-gin part,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
upon the dynamic character of the problem data. This may be several
times a day or as infrequently as once a day.
While empty cars must be allocated throughout a railway system,
there is a natural division of this problem into two parts. One
deals with the global aspects of empty car allocation--the systemwide
distribution of cars between major yards. The other part of the
problem deals with the local aspects, i.e., ensuring that individual
shipper's requests for empties are satisfied, as well as determining
where released empties must be sent.
In terms of actual railroad practice, the differentiation between
these two problems (global and local) is important. Consideration
of the global problem leads to simplifications since one need not
consider all the detailed differences between car types and can aggre-
gate similar types for routing between major yards. Nevertheless,
at the local level it may become necessary to distinguish between cars.
However, local problems are separable, i.e., are independent of
each other; thus, the local car distributor's knowledge can be utilized
to facilitate solution of the problem. It will often be possible to
take advantage of the distributor's familiarity with the.shippers
in his area regarding such questions as the condition of a released
car, or whether substitution of one type of car for another is accept-
able.
To formalize the distinction between the global and the local
problem, the concept of a car service area will be introduced. A
car service area is that "territory" serviced by a single major yard;
while there may be secondary supply points within the car service
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
area from which cars may be"supplied to customers in the area, any
car which comes from outside the car service area must enter the
area through the major yard.
Major yards are linked by through trains, while individual
shippers are served by local trains or yard engines which typically
run between two supply points, or out and back from a single supply
point.
Car service areas may be considered as a mechanism whereby
local information. can be aggregated for systemwide use. That is,
the net supply or demand in the service area can be determined and
associated with that area's major yard. Thus, in terms of the
global model, systemwide empty car allocation may be based on the
net supply or demand at each of the major yards. Having solved the
allocation problem at the global level, solution of the associated
local problems can be obtained by "exploding" the global solutions.
In the event that there are alternative ways of supplying or receiving
cars within a car service area (for example, when there are secondary
supply points), an additional optimization will be required. Because
of the way in which car service areas are defined, treating the
problem in this two-level fashion (global-local) and optimizing
individually, one can achieve the same overall optimal solution as
if all the problem aspects (both global and local) were combined into
one large problem.
In many cases, the local problem need not be solved by an optimi-
zation technique for there may be no alternatives to choose between.
For example, this occurs when a customer on one branch in the car
service area must receive his cars from a single supply point and,
when cars are released, must return to the supply point. In these
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 210
instances, it is sufficient to'keep a detailed list at the supply
point of requirements and releases; these cars will be moved on
the first available train consistent with the release/requirement
time. If there are insufficient cars available to satisfy require-
ments within a car service area, the global problem solution will
provide for empties to move to the appropriate major yard. Similarly,
if surplus cars are available, the global solution may specify they
be sent to other major yards.
The Allocation Model
The allocation model utilizes a two-level scheme, with one level
treating the global or systemwide aspects of the problem, and the
other dealing with its local aspects. For the global problem, cars
in a class would be aggregated into groups by type, location, and
time of availability such that it would not be necessary to distin-
guish between cars within a group. For the local problems, specific
cars in a class may be considered directly in the model. Thus, a
specific car, if not needed locally, would be grouped and allocated
in a global problem, and then assigned specifically in a subsequent
local problem run.
The determination of the final destinations for empty cars is
possible through the use of simulation techniques. Dynamic scheduling
of cars and trains would be the end result whereby railroads could
meet their overall objectives of railroad operating control systems.
The most important aspect of a viable operating control system
is to provide better controls over the railroads' assets. It will
provide:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. Better control of trains because advanced information would
be available on the exact volume of traffic to be moved
2. Better control of empty and loaded cars--a determination
will be made as to which cars will be moved in what se-
quence along what route and on what train
3. Better control of car distribution--disposition instruc-
tions are part of the computer programs and would direct
the movement of empty cars. The computer would signal
any failure of an empty to move as planned
The development of an intraline operating system is required
before a nationwide interline system can be developed. Railroads
are moving soundly towards their achievement of the first phase.
The Association of American Railroads with their TRAIN II system
is providing the basic framework for a national car interline control
system.
When this event occurs, it is the writer's feeling that the
occurrence will be recognized as the single most significant happening
in the railroad industry during its entire history.
Motor Freight Industry
Now, let's turn our attention to the motor freight industry. This
sector is the lifeline of goods for our country. It's a business that
never closes its doors without affecting every industry and every con-
sumer--a business that's built on giving a mobile society the freedom
it needs to prosper.
The ever-rising cost of labor and equipment, competition, and
unexpected events like our recent energy crisis, present constant
and changing challenges.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 212
For carriers, external, uncontrollable factors such as fuel
prices, labor contracts, financing, and interest rates create prob-
lems in many of their operational areas. But carriers do prosper
in spite of these external factors when management improves its
handling of its own internal operations.
Probably the most important areas that carrier management is
addressing are its freight terminal and linehaul operations. By im=
proving its impact on these areas, management is accomplishing three
basic objectives: service, productivity and control.
Service is the carrier's competitive edge. Providing better
shipment status information and insuring accurate billing and in-
voicing on a timely, consistent basis are the keys to improved cus-
tomer satisfaction.
Productivity for motor carriers is coming from the increased
productivity of people and equipment. The areas which are being
addressed include linehaul equipment availability and utilization,
dock worker productivity, reduction of empty movements, and improved
load factors.
Control of operations means control of equipment in the areas
of balance, interchange, maintenance, and licensing. It means con-
trol of revenue by eliminating lost bills and transcription errors,
and providing positive delivery verification. It means control of
freight to reduce over, short, and damage; reduce inactive freight;
and facilitate tracing.
In view of the demands placed on management today, new tools
and techniques have been developed to sustain the profit and growth
rates of the 1960s. One of the tools that many carriers have looked
to is the computer.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
It has been used in the motor freight industry for some time.
Today it is possible to interface a computer to a communications
network and capture information as events occur. As a result,
information is being retrieved on demand, information that's timely
and accurate, that allows decisions on service and productivity to
be made on a real-time basis.
Implementation of computer-based teleprocessing systems has
typically been an expensive and time-consuming proposition when
undertaken by an individual carrier. In working with the motor
freight industry, IBM has developed a system of application programs
to answer the operational needs of service and productivity, while
avoiding the duplication of substantial dollar expenditures and
effort.
Other similar systems are available; however, because of my
familiarity with the system we developed, I will confine my remarks
to our system as illustrative of what carriers are doing in these
advanced application areas.
We call our system FERST/VS--Freight and Equipment Reporting
System for Transportation/Virtual Storage. FERST/VS is a tele-
processing system designed to provide the management of a motor
freight company with timely and accurate information on the move-
ment of freight and linehaul equipment throughout their system.
FERST/VS offers four application programs to assist operations
management: message switching, equipment control (including a ship-
ment tracing function), freight billing, and rating rate audit. These
programs help improve customer service, increase the productivity of
equipment and personnel, and give motor carriers better control of
operations.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Let's now discuss each of-these areas in detail. Morning re-
ports, terminal performance and statistical reports, hot shipment
advice, or over, short, and damage reports are entered and transmitted
effectively between terminal locations and the central office. This
is called the message switching application.
The motor freight industry has recognized that the key to success
lies in the control of terminal and linehaul operations. Efficient
scheduling of linehaul and city equipment and planning of terminal
resources are critical if the use of manpower and equipment resources
is to be maximized.
If the dispatcher wants a list of all tractors or trailers at
a given location, he can enter an equipment inventory inquiry. He
is presented with a list and status of that equipment.
Changes to equipment status are reflected in the report generated
from the equipment inventory request. This information is as current
as the entries made through the communications devices.
One of the high points of equipment control is the information
it supplies to terminal management on schedules which are currently
en route to their location. Terminal personnel retrieve information
on schedules destined for them within a specified number of hours
(e.g., next sixteen hours) or on total schedules moving in the
system with freight destined for their terminal.
The report generated as a result of this inquiry gives the ter-
minal personnel estimated time of arrival by schedule, weight/cube/
number of bills, and any specific load information entered at dis-
patch, trailer close, or billing time.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The schedules en route report has many potential uses in-
cluding:
Preplanning dock and city equipment requirements
Preplanning trailer spotting at the dock based on
large mark information
. Anticipating equipment availability at a given location
Providing accurate information on inbound tonnage and
shipments
Projecting dock crew workloads
Once a shipment has been recorded in the system's data base,
the current status and location of that shipment is retrieved by
entering an inquiry. The operator receives an immediate response
from the computer showing the location and status of the shipment.
Revenue, shipments, and equipment must all be considered in the
control system for the operation of a motor carrier. Efficient uti-
lization of equipment control system which is integrated into the
freight billing system.
Let's now look at the freight billing application and see how
it interfaces to the equipment control facility in order to provide
a full-function motor freight operations control system.
Equipment control is a computerized system designed to assist
dispatching, keep real-time equipment inventories, and provide ship-
ment tracing and advance schedule and load information.
Equipment control application functions are broken into the follow-
ing three areas:
1. Equipment movement reporting, which consists of simple
input messages used to report changes in status and loca-
tion of linehaul equipment.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 216
2. Inquiry and tracing, which allows the carrier to inquire
as to the current location of equipment, equipment inven-
tories at any terminal location, schedules en route
between locations, and the last reported status of a
particular shipment.
3. Shipment control which gives the carrier the ability
to report changes in status and final disposition of
individual shipments.
Now, let's take a more detailed look at some of the features of
equipment movement reporting and how they are helping the motor
carrier's operation. For example ...
When a schedule (that is, tractor, trailer combination) is ready
to move,.a dispatch message is entered. The message reports the
movement of that schedule to the computer and updates the equipment
files to show those pieces of equipment in transit between terminals.
In order to report trailer interchange between carriers, the
program provides a trailer rental and interchange message. Once
the information regarding the interchange is entered into the com-
puter, the carrier keeps track of the time when the trailer was
taken on-line or put off-line and the return point. This informa-
tion facilitates tracking interchange activity and calculating
accurate per diem charges.
The equipment movement reporting facility also allows the
carrier to report on changes in equipment status. Examples would
be a tractor placed in the shop for maintenance or a trailer spotted
at a customer location for loading or unloading.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Let's now look at the inquiry and tracing facility. Here are
some of the facilities it provides that are helping carriers improve
their linehaul and terminal operations.
Freight billing is divided into these functional areas:
Bill entry
Manifesting
? Delivery reporting
? Management reports
Many edits are performed on the bill before it enters a system.
For instance, pro numbers are checked for validity, destination codes
are checked to be sure they are valid, extensions and totals are
checked for accuracy. Conversational editing of the bills at this
stage helps insure accuracy because the person creating the bill can
make corrections while the original bill of lading is at hand. This
type of bill entry eliminates the need to keypunch bills for entry
into a revenue account system, avoiding a step where transcription
errors can occur.
Since all of the information from the freight bill is available
from the data base, it is possible to do manifesting and to produce
more detailed reports to be used as advanced planning by terminal
management.
The two primary documents produced by the manifesting section
of the freight billing are the tracing manifests and the trailer
summary manifests.
Tracing manifests are available to terminal locations on re-
quest for both inbound and outbound shipments. Inbound manifests
are sorted into alphabetical sequence by consignee; outbound
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
manifests are sorted into alphabetical sequence by shipper. Terminal
personnel uses the tracing manifests in conjunction with the trace
inquiry to answer all shipment inquiries submitted by either shipper
or consignee on a particular pro number.
The trailer summary manifest is an extension of the schedules
en route report which is produced in the equipment control program.
As a result of the freight billing operation, weight, destination,
and additional comments are available. The trailer summary mani-
fest is of particular significance to break bulk operations because
it summarizes shipments, pieces, and weight by ultimate destination.
Terminal management uses the trailer summary manifest in planning
its city pickup and delivery requirements.
Prior to the arrival of freight at destination, full freight
bill delivery sets are printed on request at the terminal. The
delivery sets are used for the delivery of the freight. They may
also be the documents used to control stripping of the linehaul
trailer and pickup and delivery loading.
The freight billing system allows the carrier to report the
final disposition of a shipment via a delivery message. If the
carrier chooses, this message will trigger automatic invoicing.
It may also be used to record over and short information.
Once a system has been implemented to do overhead freight bill
transmission, the centralization of the rating and rate audit func-
tions is a logical next step.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The rating rate audit application package provides the three
functions required of such a system:
1. Rating--separated from bill creation and supported at
one or more central facilities.
2. Rate audit--where the billing system selects those bills
to be audited automatically based on user-defined para-
meters.
3. Rate quote--which allows terminal personnel at remote
locations to submit requests for rate quotations through
their communications devices.
With rating rate audit installed, the billing clerks enter
skeleton bills at the origin terminal. Eating personnel at a central
facility are presented sequentially with the unrated bills grouped by
tariff bureaus. The rate clerk then enters the appropriate rate.
The rating program will then perform the extensions and total the
charges.
The rate audit automatically selects those bills which are to
be audited. The invoicing of bills queued for auditing is prevented
until the audit procedure is completed. Many carriers select the
weight and charge parameters that determine which bills are selected
for auditing.
The rate quote facility allows terminal personnel to enter re-
quests for rate quotations on their communications devices. The re-
quests are presented to rating personnel at the central facility.
They enter the appropriate response, and the request is sent back
to the origin r
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Among the potential benefits to be realized through the use of
the rating/rate audit programs are:
Reducing the wait time for shipments
Alleviating pressure on rate clerks--reducing their
errors
Allowing a greater degree of rate specialization
through centralization of personnel
. Reducing the number of rate clerks
In addition to the on-line capabilities we have discussed, these
programs generate a number of summary and statistical reports that
help a carrier measure performance and more effectively control his
operation.
Inactive freight reports help terminal management control over,
short, and damage, and improve customer service.
To summarize the potential benefits of these on-line applications,
let's recap the areas of service, productivity, and control.
Under service, the shipment tracing inquiries and reports help
to answer customer requests more quickly. The equipment control por-
tion helps to balance equipment, keep track of special equipment,
and make it available where it is needed, when it is needed. Freight
bill transmission helps move the freight faster, reduces over-and-
without bill situations and the on-line editing of bills reduces
biller errors. Delivery times recorded provide a record for manage-
ment of how schedules are being met. The inactive freight reporting
helps terminal management control over and shorts.
Under productivity terminal managers can anticipate manpower
and city equipment needs with the schedules enroute and trailer summary
reports. Central dispatch and terminal managers can improve load fac-
tors, increase utilization, and reduce inactive equipment with the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
accurate, up-to-date equipment inventory. Biller production can be
improved. Rating/rate audit further improves biller production, be-
cause billers no longer have to enter rates and charges on each bill.
Centralized rating also helps clerks become more productive.
In the area of control these applications offer better control
by providing a data base of information on the carrier's operations
when and where needed. That can mean better decisions made, poten-
tial problems anticipated, and action taken before potential problems
become real ones.
Information is provided to:
. Terminal management with--
Schedules en route reports
Empty movement reports
Inactive freight reports
Biller production reports
. To linehaul operations with--
Equipment inventories
Tractor cycle and relay reports
Inactive equipment reports
. To accounting with--
Master bill file verify and balance reports
Delivery verification
. To data processing management with--
Terminal usage reports
Communication circuit analysis
System error incidences
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
And it also provides information to executive management who
can review the same information to measure relative terminal per-
formance, fleet utilization, biller and dock productivity, and many
other facets of his business.
I hope the preceding computer application examples have given
you some feeling for the considerable extent to which our transpor-
tation industry has applied the computer in their day-to-day opera-
tions.
While most of the uses have been in the inteamodel area only,
it will not be too many years away before intermodel exchange of
information and control becomes commonplace. Perhaps some time
in the future we can have a worldwide transportation control system.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
COMPUTER APPLICATION IN THE ALLOCATION
OF AIRLINE RESOURCES
Morton Ehrlich *
Airlines today, at least in the United States, operate in an
environment characterized by:
- Limited resources
- Rising costs
- Restrictive labor contracts
- Government regulatory policies
- Increased competition
- Economic sluggishness, depressing traffic growth
Within this environment, the challenge is not necessarily one of
viability, but one of survival. Classical solution systems are no
longer applicable to today's problems. New and advanced methodologies
and techniques must be developed to respond to current competitive
and market pressures.
This paper addresses the major problem area of the practical
allocation of aircraft resources as a function of market demand and
fleet mix, and discusses the development of new systems designed to
respond to the problem.
There are three key phases which dictate, to a large extent,
the allocation of airline resources:
- market planning, which defines market strategies and
revenue generation
- schedule planning, which converts market strategies to
specific flights, and coordinates implementation
- flight crew pairing and allocation, which attempts to
optimize the use of the human resource
Appf elease iced For R President of 2001/11/19 ng,CFas Dtrn79AggJJ?J200 P020005-9
V
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
All these phases converge into schedule development, the master
control of the consolidated allocation process.
Within each phase, computer systems provide support:
- A marketing information system using data base management
software supports the market planning function. The data
base contains several years of proprietary eastern, compe-
titive, and industry traffic and operational statistics.
- A frequency planning model is used to conduct preliminary
evaluations of future schedules. Results recommend the
optimal sizing and frequency of utilization of aircraft, by
fleet type.
- An online flight crew pairing system provides crew manage-
ment with the capability to interact with computer-stored
files, and update the crew schedule with evolving changes
to the general flight schedule.
- An online schedules development system provides the cap-
ability for online updates to the master flight schedule.
This system provides overnight turnaround of schedule
listings reflecting the latest resource allocation judge-
ments.
Before discussing the resource allocation function, let me
give you an overview of the facilities and hardware used to perform
the allocation process.
EASTERN'S COMPUTER COMPLEXES
Eastern has three major computer centers:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
At Charlotte, North Carolina, the approximate geographical
center of the domestic route structure, are three UNIVAC
494 computers with supportive peripheral hardware and soft-
ware. This center monitors and supports the day-to-day
operation of the airline, through an extensive, high-speed
communication network operating in a real time, online en-
vironment. Statistics collected throughout the day are trans-
mitted between 2:00 A.M. and 4:00 A.M. over a high speed data
link to the administrative center in Miami for report produc-
tion within a batch environment.
In Miami, there are two other major complexes: one, supporting
the systemwide reservations network; the other, providing re-
quisite processing for the administrative, financial, and
engineering and maintenance requirements.
Two IBM 370-195s and two IBM 360-65s configure the reservations
complex. In addition, they
- Support the automated passenger processing systems at our
major airports and ticket offices. This system makes avail-
able to the counter agent within a real time framework:
? Reservations status for all flights
? The capability to offer seat selection within an automated
environment
. The capability to generate automatically tickets from the
reservations record
? Flight operational statistics
? Provides total processing support for several regional
airlines' reservations requirements
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The system also communicates with the administrative computer
complex via a high-speed data link.
The administrative complex is comprised of four IBM 360-65s,
one IBM 360-30, and several smaller computers used either as front
ends for high-speed communication networks or support for interactive
systems oriented to a single function (Exhibit I).. Systems discussed
in support of the resource allocation process all reside within the
administrative complex. Now let me start developing the application
framework.
MARKET PLANNING
The market planning function has the primary responsibility for
revenue generation. Activity groups comprising the function are
responsible for developing market strategies, monitoring and analyzing
market traffic performance, and coordinating product-related input
to Sales, and Advertising programs.
Market Strategy Development
On a monthly basis, market strategy development monitors
(Exhibit II):
- Industry traffic and operational statistics
- Competitive actions
- Eastern's performance for each market on the eastern
system
From three to six months prior to the schedule-planning process,
the top 400 markets are subjected to analysis. At this point in the
process a demand forecast of industry traffic is made for each market.
The results are then used in a forecast algorithm to allocate industry
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Ap+ot jF
_LJ
71
~
1:r P'? ~i r.J ~ ~L?;I 1 ~ ( ~M Ir ~ I _i jl ) ~
~ ,
\' t` // ( ,a c; 1 ~+~' ({ I III :1
-71
.~; r .. 4-- - ? -
V) Z2)
T1
Vii
App
I
I ,I, = 1 i; t
I 1:
-
0
-RDP79-00798A00031 00,
IFel,sg. 9.6CIA-RDP 9:,~1~$i4000200 12ES~0~=9 . ~?f ~J
~l m
2'27
Approved For Release 2001/11/19 : CIA-RDP7,9-09798A000200020005-9
I'
:r. r.
f? N
v
N'
to
N
'
.c
c,
v.
ro
~,
V W.
N U.
a?A?
A A U
4
?
+
+
f
0..4
'
v w
.??
N
?r
-!
N
W
N
w?
W
W
WS
4
N .L'
(p
h
r
.
' ?4
OC
0
r .
1. L)
ltd. N
N
N
N
N
h
h
M
h'
r
?
P,
IL LC
W ,- '
fr ?.
?r
t
+
t
t
a
tv
+
.
+
i
+
u. V,
I- I.)
0) oC
N
0
af'
'C
G
4
0'
;~
V?
'j
co
0
kA
Lp
to
."
f-
p.
'. I-
h N
h
h
h
N
h
'v
M
4. 4
1.4
M'
I.
a
a U
0
0
C)
a
a
0
0
1
0'
0'
0
0
x r
C) O
O
0
0
O
O
0
c)
t7
0
0
0
U)
di:
x a
~+
O
N
r-
N
N
IA
0
0
o
0
0
Vt
W o
0
V'
0
u
C)
a
l
3
ILI
)v+
? a
?
1 ?
?
a
?
?
In
M
.t
N
IA
to
1n
4t
an
Y\
li
?
Z.; I.-
N
N
N
LIN
v
Ai
'C
'T
0
W
0
w -- d
U n :.l
? a
N N
r
+
?
-+
'O
M
.r
.?1
I
-?'
L[ '- N
of
N
.?? a
.?1
W
t .+?
+
t
+
+
1
4
.? i
r~. 'r
a.1-.
a, v
v
v
o
0
o
a
o
v
o
C)
N
G ?R
.
;
I
U w
.
t7
Y 'n
R) S
W
-
LIN
to
in
W
ap
.D
W
f0
:.1
N H
.r ID
'..
N
W.
W
N
w
tti
w
w
,u
W
at W
W N
L? :L
W Q?
01
rI ,
0
Q
"1
?0
h
.r
:r.
-+
vt
?G
IA
a
a
?
?
a
?
;d
4
V
L 'J
W C)?
NI
.r
f.'
-,
W
4t
x
U .J
J
R
.) Z
I1'
W Ut
+
t
w
t
J
+
A
t
U.
t
ut
+
?tl
+
4)
ILI
+
C W
C
4
Y
U: N
I-. U
.C O
4'
r l
.1
C'
cr
N
tV Sll
.,;
.' U
l (y
Cl n:
Ut
O
L
.I?
4
0
J~
V) _j
C'
r.. S
W 4
? ?
a
I
?
?
?
?
IA
N
-?I
u)
M
)
N
h
+
-4
+
t
+
I-?Z
1-
?t CL ,.J
W
W L
t
W U
X N
C
?LL
U
CL LJ a'
C) 2
U z
:U 4
-r N
N
-?I
N
In - -?.
rl
r?
-
N
P
J
,
N
N
.11
f?1
N
N
I'i
N
ft7
C7
CJ
0'
?a
Y1
I
.. w
*
I-
u S
N
M
C
)L
?i
% ~:
L V
^) .0
r4
'C
co
LIN
?T)
N
0)
W
N
0
.a'
.
IF
-' u!
r, N
.A
C)
f?t '
n''
?T
?-
M
M
v
T
0,
.
N
VJ
O
N
M
0)
?U
o
N'
?-f
N
t'
.. N
O .t
N
in
N
0
M
?t
4
.-I
N
'r
.c?
U S
xu
M
0 0
a
0
a
0
0
0
O
O
tJ
~q
O
P4
0
to
+. ~.
NL:I
0 0
o
O
O
0
0
0
o
o
o
0
0
I)
4a
1 u
+ l)
'v
14
W
Ir
.?1
.t
V.
?N
w'
.1
Y)
i
V1 1
~l W
-4 M
?r
N
N
r)
in
I.1
I'')
rt)
'I
.t
?A
6 I
W aJ
-. M
H
N
??d
-l
-4
-?1
N
-Y
H
?.r
H
--J_tJ
m n
.A
1A
N
r)
?
O
!. cr. W
IA
N
-4
N
N
11
W U.) v
v1--
I
I'I u1
I?
???
r
h
0G
W
0)
f?)
U.)
It
fU
OF. LA
?t.
+
?''
t
+
+
to v
t
+
I... u
LIN ?u
V
0
U
0
0
tJ
O
O
U
a
v
ar
tr N
U
U
U
If
C)
C.)
C.)
0
I.)
a
c.)
IN ?'
I
?
I
?
?
a
a
?
?
a
v1-?L,
LIP
v U'
a1' V'
u
'3
U
U
v
0
v
V
u
U
v
C?
V
U
L)
V
U
+
-.
.''' ?..
-'
.-4
+
p.
+
+
+
+
Y
+
+
+
1?- W
.r J
-4
(Ij
rv
r-
U
U
V%
r
N
-4
V.
III
PI
1
IV
cu
CV)
r i
fn
r-
i -i
? ?
?
?
a
N
h
P.
9
tit
?i to
N U'
V.
N
N
N
s
V.
1.+
V'
V'
V'
V'
+. +
+
+
+
+
+
+
+
+
+
+
+
C.) >-
0 O
O
C)
0
0
V
0
W L4
}. .?t
?.i 10t
.!.
0
W
W
-%
f+)
0
W
I')
Ut
C' W
?-4 M
Ir
p1
N
^)
M
J?
I'
?
4
'7
.! C)
.?I -.
..'
M
-?
-.
..y
.?
0'
0
0.
.?'
vt
CA
J
N
A
111
tr
C"
- r w
r i0
41
.1
.
11
V
?1'
H
Il
U'
N
-+
O ,
V ' I r
? ?
?
?
?
?
?
?
N
0
.
N
+
..',
H
N
+
N?
+
fl'
+
+
+ + + + +
Approved 'For Release 2001/11/19
CIA-RDP79-00798A000200020005-9
It.
I).
A.
1
.1.
?
'V
-?
?,/
f'?
I
r
r
t?
1
t
.+
CJ
7 ?
'J
V)
M
G
Q
C,
U
I,t
1
?a
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
projections among Eastern and competitors. Several iterations of the
statistical forecast are required to test multiple competitive assump-
tions.
When finalized, each of the markets will have been assigned a
priority based upon its revenue-generating potential. The priority
is later used to determine alternative tactical plans when recommended
priorities cannot be accommodated because of resource or utilization
constraints.
Traffic Performance Analysis
In depth analysis of Eastern's top 400 markets include enplane-
ments, onboard traffic, 0&D traffic, seats, departures, load factor,
seat share, and trip share. Two reports serve as the basis for the
analysis:
- A market traffic performance analysis (Exhibit III) reports
traffic at the non-directional market level with summaries
of sixty economic and market group entities. This report
is oriented to sales and advertising functions, and provides
the option of adjusting strategies and programs to conform
with market performance.
A directional traffic performance analysis (Exhibit IV) re-
ports the same data elements as the market level report;
however, it reflects traffic performance at the directional
station pair level, with summaries at the station and regional
levels. This report is used during the schedule development
cycle to fine-tune strategic and tactical plans scheduled for
future implementation.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
.,
j1
F
.4
..1 I``r
N
< I
I.
IL
v
?r
)f ' I
.!1
J 'K
? n
J ?~
M1
.T
-M
Approved For Release 2001/11/19 CIA-RDP79-00798A00020002000p-ig3O'
. 1 I? 1 1 1 1 1? 1 t I I t ') { 1 1 1 1 1 1 1 1, I t' 1 1 1 1, { I i
. [ : 1f1 ^J .,J rv ,T..~ rr < .?~ ../ '1 r f.) h i S! ??+ '?\ c, < .J 'A u , N 41 1 ?:J rJ .? 'V .?l ao 'v w. r? .t zJ -+ :J a '? ? ^ ? u, rt < ,) ..\ ?) r a ~..., n :~ r . ? .~
Ul u. , .f,? [A N F .J v1 ~. r M J .) Pl :V ?1 7 y ti ~. ?, ,r, ??? Q? N -' `J ! V\ ~! .Q ~.
?1 ? ?"'?.fV +. M .J N 4' M ??\ ??? .J fl J iJ ,
1,1 y IV f.) ?n 1V N .'1.?? ?J ;?\ 1+1 .4
~ 1 ,
N 'a 1 1 1 I{ ( 1 1 1 1 1 1 1 1. 1 1 1 1 f L. 1 1 1 ( 1~ 1 1 1 t[~
11 1 ..
~... rl .?\ ~r, 7- r?..?~ ?.: J..v. .+ ?r ... ,. ;, ~?\ .A rJ t ~. v ~. ...7 .,) .t ..?.
4Y } 7) D -+ J J. _ .? ?'J w\ R ,? ... M1 I;. w +\ ?- ,. ^1 J -J ?G J h ~?_ "V " ..:I '. -? .V
V I sv Q' ) < r, . J rJ J, .f?1 C? < J'. -? h -~ < s1 w ?') ?J
. 1 .4 N N h ??? < < rJ N M .-1 .f1 ??I ?.f\ ??+ < r N H N ?V V `+,^! "'?. ???;. ,
^\ < . r- , YI r..-, 6- r, 4 ?: ?7 -'+ .a ?Y t ?" I.. ?? . t.' .- ;+ ~ti ') a? 6- ?1 q ~?? 'I):
C r .^ sJ :? :7 ~? ?:. y T) rl ?- ~) J ?? J 1'\ J ?? ?:\ ~T D ? ry T :V d ?- 'J\ J? ;. .} .-? n.
J M ?41 4, h .: J -.. < .r 9 N h e1 .,~ .. ,., < S .'?' .~ rJ ..\ .? t ?'. t @? 7 n t- V 1.1
N. j Z> j
? ? ? ? . . ? ? ? ? ? ? ? ' ?
J
1 S
1 I _
1 1 .
1. y' . ''
1 t ?.)
1 JJ
N n
1 N V~ ?t ?'5 1-? , 1
1 1 1 1 I. 1 1 1 1 1 ~7 J
I , .t
t:1 N
I i 1:
.r .?. . : V I
h! V ?? .n ?S?I .f)
} I
11
.r ^ V -J f1
J .-\ . u . -1 : V ? ' J
?1 .Il 4? ?\
l s V .t r U h
.J N ??
-.? .\ ? V J% 4! Q. r-
.?1 r1 :) J ?V
? I
? ? ? ? ? ? ?
r?Ir a
'1?t, iV
1 1
of ??, N ? ??1 r V' .-
_ r1 J<
. ?'
;. rl ?Y
? N
.J +
11 I I 1 1 1 1
;V J M. wS .J
14' V ??. ? V ^1
IJ
1 1 1 1 1 1 1 1 Ir~~ ?Y y J [ '? t t) .'V J -'\ 1
? 1
1 I
s r? r1 ' 4 ???I , v ? ? ^? " x1 V '. J N J ) N J J
t I
M1 M ?,. r'
`1 .? 'J 1J P1 'J
', J . ? J h ^'1 :
,1J w ~r 6- ') J' _I r
?+1 ..I .-? ??I .~: ? V ?!\ ,'1.
J.
i1'J ~? Y1'7
.V ?J "3 -)
1 I
-4
r )
)S CL'
N'J rr
-, J
? I
C., 4% J, IIJ
i ?
wl .- T. ??1
1 1: I
? . ? ? . ? ? ?? ?
?
?'. .7 ?\h?J?. "I r. .i\ J?'\
.-?.l -f??.4 N??1 N , rJ
1' 1 1 1 I I 1
fI ..1 ?V J? J ?-? ?'J r: `\ 4 ??J ?, J ?- r J h-.,
1)~
.~
.) 11 ..) ?r J\ 1 ?J to .?\ h 'lJ ^'\ N ?lJ M1 .^ 'y 4 ?V
M1 '1) .T rJ {.~ 1 'n n .y .\ .J. I'1 M < 1.1 M1
.?1 iI 1V 'r .11 IV < N .r J J .??? Jr J\.
} y)
? I ?
. 1 I
r- JJ 1.? r^ .3' v J l? lV .J
1 ?1 4 r1 J .1 ,V ??? .V '.. IV .V ? ?V .?1 ..? ?V 'r) .w ... r V I?. -? .r N
1 ?? .
.J r 1 I , I
Approved For Release 2001/11/19: CIA-RDP79-00798A000200020005-9
`1 I- 1.. .1 A I. 1-'?? '\ '1 1- -- t I: . ) 1- ?.) .1 r 1 ~I ?1 1 .J , ?~ J.
-. 1.. 1? -- .. ?I
X
. ? .S .J . J . . r . , ' 1 .1 r _J _r .J' J V A. ?1. .'I . .l. J?' w ? -?. ? K K {
Approved For Release 2001/11/19 : CIA_RDP79_00798AO00200020005-9
I .... I ?; I" 1
1.1 of I, 1 4 -1 17? 4 + ;V
It it I.
f1 I ?;
..R S t', 0
ff ri' !t S'
I ? ? ! ? ?
y
? ? ?? (~ I . ?O C.1 ; ??t
t'. U 1r' I N 0' fn
'4 I
0, tut
,
w ,
Y .
C,)c)
ciC'r
C
O
It
`
1 ?' u 1
.
I T'. ,l
t). Ltt'.l
u r
u
IL1
N
!
N a' Clb'J 470 00'^rJOi1tO CJn n'rn C' - n 0 00r)171 0f)0 0mC'nn ?, C)r, C.Or_'ry O0 e- r,,n p,") ?,
.?' '1.'I 4 ! OC:OVtr' . ? , . .nO . .L. . . . . 7U , act ?:1000)G (,,, Jl J C)n C?0:)OCJ 37U OG ID OU V tl C`
.? ! ?
.-?.: ! W 0 rtC)0ell !hh00.?t)0P-0.0 "^or)C"ooo"nr-e.) C) 0 Cr?r-r7r3 nlr,ryC)N.pCMCr'J.trl:~
?-' I %A L0 I I O?-tOOUIOONO.IO 'r c' co CJ L'?rJ OCC: O.-1 C1 ?'+C. G-?G 1?, 00%r .:l r)rf.C?,I V?GG t7GG
tl u a:, 1, ? I 1 I.? r .?? W 4 N. .?? .r' ?+ . 1 . 1 ! 1
V
U. -4 z t-of I
4J ~) I { ( ! i r
CCX r- r-: I
1 I
CILt01' O.' ( GO0 C)ti0 0 m0 'CM^r1h00 m0 mm0 C)OCmrCrlnCC?rl ltipOmhrJr)r.nnti
0 fO '+) G
.LL ' w Ga:K 0 r1 GSOO V 5_I I I ? ? .. SU'' C?G P? Q,. CIO!) CPO7r?tOL IC'O!l t 1: O! 7r^ 5 4("0
~ W 7` t 4J t>L. I d M .?? , N N .. m rl .t r f+ .-..+ N .-?
L .o. p. , O , ( i 1 ,I { 1 i
?^? lJ Nit Cu ?N M 0'! t` d co ?... O' ..A .~ A If w 01 . 1n tl i
rn ti
? ?L J /^ ?-1 v) '.? v J .3 ! N' N .T c. J r C' ?-I Y a 0. %n.10 .a d m
:C ?l, .. , 4 ! O .? 1 11 1 1 1 ..1 1 j .-? 1 .-?' N .+ .'. .+ 1 1
.C4O'I I.?.C?.t . ? 1 ! ; I I
.J-I C.? N U i I 1 I
0.:.C ?2 I1 I
h 6I ( N N1 ?b to 'D JM'C''J. 01 -4 .4..1 J)J? J M U'O d.+ Al
F I , V A'6' N -t ,'t .J; N. M.:' 0.O C d . C1'I 0 6, o NJl l>
?r. '.-t ..r rJ ?.. r J' N N I N j 1 N 'r -A.-. N ..r
1- O J !
St r?, N 4? t?r r { , I ,
4 IU >? ~ ! ?
r T I O I I { I ' ? ,
1 ~, 1
?
K ! +T ?J) 0 4 .-, M s O ?? ..r ? ... M h 0 O ? N n' of
r? N # ? J K !Ul '7'#
.4 us 5 ??O { N: O ,r ? ..r r.J :J Ina CO. ?.G fI
U dl' .J N, ?V1 I N, M J F; N i) . I O]
?- : 1 ? 1 ? 1 ! n:
.4 t) i 1 II I 1 1; !
I I
4.
.
I 1 1 tf # .-? ?+ M4# Ut , ('I C' J m
t''?
C It b I t 121 m' t>6 61%
C' nl U. I*, #
Lj u M tl
1 Ua
I
'
I ~- N +?? ..? i,? ? 4. 1~ ~". ? ? ti.
J 1 t'. .j r.a ... d , ? ., . ? -, J .? ?L v' .- t? Y . '~ L:. )..1 .J - L) N V,
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
pproved For Release 2001/11/19 : CIA-RDP79-00798A000200020005; 9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
- Competitor actual operational statistics and onboard traffic
for the period January, 1972-December, 1973. These data be-
come available from twelve to twenty-three months after the
fact, and are used to develop passenger departure time pre-
ferences.
- Competitor and industry traffic statistics from January, 1968
to December, 1974. These data are the true and online 0&D
traffic in each market. Market share for each competitor
is available for online traffic. True origin and destination
traffic is available for all domestic markets, and U.S. flag-
carriers' traffic to international destinations. These data
are added to the data base quarterly, but are six to nine
months after the fact.
All data are maintained at the source level (flight or city pair).
Additionally, a consolidated record containing frequently assessed
statistics is maintained at the directional station pair level. Data
base management software establishes linkages among related data at
each level and across multiple levels.
A general purpose report generator provides the capability for
generation:
- In variable formats
- With multiple data sets
- Up to five levels of summarization
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
' Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Structured data files containing unique traffic statistics, in-
cluding 0&D passengers, onboard passengers, and enplanements are
maintained as a subset to the data base. Interactive software com-
ponents of the data base management system permits a library of
statistical subroutines, including
Regression analysis
General purpose simulation system
Time series analysis
Trend analysis
Variance analysis
These operate on this minidata base.
FORECASTING SYSTEMS
Two fundamental techniques are used to forecast traffic. One
method uses:
- Stepwise regression with
- The Durbin Watson statistics to eliminate auto-correllated
variables to develop the forecast algorithm
The other method uses:
- Time series analysis with
- Fourier analysis to overlay seasonality components
Economic indices used in the regression equation to forecast industry
traffic include civilian employment, disposable personal income,
government expenditure, corporate profits, and average air fare.
Forecasted industry data from each method are loaded into files,
and an analysis of variance performed. Resolution of variances intro-
duces the human factor into the forecast algorithm, which, functioning
within arhintb My_e
RM1600M 1fRT. G tR f~ d $ 2 '2( 1p3
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
files), can manually override the statistical projection with a judg-
mental forecast. Once the industry traffic has been forecast, re-
gression analysis allocates the industry among all competitors by
seat share, trip share, and market share. After the regression has
been completed, a test for reasonableness is run to measure forecasted
growth rates (Exhibit VII).
SIMULATION SYSTEMS
Although the simulation model was designed for use at the market
and market group levels, it may also be used at the system level.
Its primary function is to permit the analyst:
- To vary competitive assumptions and reforecast Eastern's
position in the market
- To vary industry forecasts and recalculate Eastern's
O&D traffic
- To vary both competitive assumptions and industry fore-
casts, and determine the impact on Eastern's O&D traffic
The analyst may elect to use either of two subsystems comprising
the model. The first subsystem permits altering the independent
variables in the algorithm by ?1, forecast error value. The second
subsystem requires that the analyst specify a numeric range for the
variables. The model then uses a random number generator to vary the
dependent variable over the specified range.
Another simulation tool is a general purpose simulator. The
system was designed by IBM and is in general use by many airlines.
This model is used primarily to compute a "risk factor" for each
to
1 ?
C
Approved For Release 2001/11/19 CIA-RDP79-00798A000200020005; 9 238
.. ~~
h to ./ ?-I .' . ?\ 4 ??? ? ?' 4 v 4 + m
T f.f .ti N ? ? ? r ? ? ? r ? - ? ? ? ? ? ?
f+'?J V? ? / 1 4 _ 1 , t w ? 1 - 4 ' 1 f 1 1
C3J
?C?4
C) '?J
J .
w
N
Y 4'
?t ? N N h '1 '-f ( N N M M W ?) 1
tr ? .4 N 4 ry n . ? "4 ;r .? r 1 ep ?y f- ' .1
:) 41 M J ?t ?~ '1 4 VI V1 .l Yt ) ?' ~~
? r
I"
?
b
C .-
1:r ?>
J? '2
.J
r ') La
?J 7 a
... W
.l J t?1 '?1 ? 5 -"
w 4 C) .? '1 Yr
w :C !7C l'` w
4 . .1 \1v.J? 4 r
. v ... ?.) i ?h ?.1 :f
\ J ? . e.
IY A
.n y
I:- r
.1? i5tl
4.. .)
?., a
..e ,y .1 43 A
.. ? N )-
J)
t'
'7 " ?l
u) 1
t,Apitroveb
.J? X Iv
s .n
-?t:
f?? ?f '.f ?? ?i I.1 N Y' .?/ .tf ?.?
N ry Ir n vl .f :1 .4 u. W = r.
% .4 + t ' ? ?. ? r ? 1 ' ?
? 19 ? .!1 ? ??) ? vt ? ?r? ? .1 ? .Ae ? .7 ? 1.7 ? 'e1 ? a1
V.4 .?.: '. 4 .??~ 3?t ..: .y.l ' 4 ; 4 44 .1
'.' .1 .7 ' L' .1 V ." 44 /1 745.7 ??1 .1 ..r V1 A ?/) ?r? A ....n
?a? N L- .1. ?U s' e.:.' r? .n tl .re ?? ?/1 r.) ?q ' V .) 4) .' .r q
.?? ?). i. .t V. ? \L IL B. .\. + 1:. ? 't
ILT
.r 4 1'1' .r ?? ?! )? ??? M Ir .4 '.
? t ?.4. r'1 ?C ? T ?r ?4' ?"'1 ?R ? ) ??' ?T
nr -? j J J 'v S ^ ?T 'J ??~ ~?' : J 3 7 f -1 r^ 1 ' l '1
4%a Ala r a r-? f11 t-tl t-t in a of 0? I'D It C)4 'r I
.r 1' r7 -t -? ?7? ?11 ;) -/?rl .V ?n r- Z .' ^t J'S r' p .:? ?.1 U
~l J1 n .r ''~ 'l Itl ,. C' J1 a .1 - N f. .S r 1' 7' J .r
? U' 4 'I T .? .. d' 1.7 -a . S
;? .4
C? .. ?'t
5-
d T M N ?} ..? h '?1 4- .y 1?f w ..t ??? tr 41 ..e .0 N
'C tl? C J .) l: v7 r- O If C)
:J .t r't '1 .?1 ?ry N.. .'t 4' ?' 't
? ? ? a ? ? . ? 4 ?' a ' ? of ? i t ^ f ? ? t ? tt + It + a
t1 .A 1.1 ?.l .1 Irl A ?A N .1 \I) .A -.A
S i?' G7i r t SJC Ma. ^L V .Y i .1R 4)# .7 3' S.4 at A.
I- ?~ ?7.4 r-?7 C1 -1
R
1 1-.
1+
r.
-.
,
-r
i
1 Z
H
O
1
?
{I j
C)'
I
N1
O?
r1
M1
In
1 U7
N
V)
N
co
w
c)
d
h
J
IA
-??
N
N
N
.CI
C !?
'.1
L' P
CJ
~?-
C
Y.
N-
h
J
-
N
J
ru
0
G
O
C7
0
0
V%
M1
-+
-?r
J
-+
G
u1
N
1-
1
fY
a
rl
a r1
U.
th
C'
I Y.
sl
an b
VP
m
K.
n.
I
N
N
611
C. P.-
.) ^
J
IA n
'-
f
r)
a1
N
N
N .r
N
N
J
I-
1
. :u
1
C)
1
!-
I
I 1:1
O
C. N
P_+
O
rA
U.,
~?
M C)
M
-+
Y.
4
CI
U:
\
1
w
!I
1 h
1 r
w
-+
u
1 N
? ~.
1
\
1 U
rn
O? rl
u;
VIa
ll.
.0
N LL
W
r
J Vt
N
:l h
r7
'a '0
.t
?
N O
d'
N O
N
h .{
N
?
h .}
P W O 1
C.N?+ 1
>?rr
-1 .1 ti
O C. a
1 .~
A
-1Jr1JJ
F J J. J J J
L' N U U C. 11
?. 1 1 1 1 1
t N n O?
1 "t 7 'C .C C
NI U. W U.
Approved For FReI c X2001/11/19;: CIA-RbP79-00798A0b02Q00i00g5-9
Y V'1 ll. . ~ . 1 i
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9. 1:)+111111'1' X
tlAi n..t I;t''I/.ICI 11
t~~nl,r LT SC14C~+1L~ Ntn:~lM1N( X
FALL. 75 - VEr? S /18 - M:0N
****/' D1.3 ***v: '
ft:;T rt. T F.?I1 I C. 111F C' CI1r11!
MAI'ICI:T DIPCTN SVA Tl7IP1t STA NI-R STS Fkt 0. 04T1 pATL ARVL
A7(.()411 4TI.I)i R A T 1 8014 .IAX (s;'' 1)?.(t (1/. 111 IUt, M1611 S /' 0'10 1 751;'11 f1+1t)+)
A I 1 I 1111 I .'111 IIAI% 0.' /'~ i t/, 01X111 ', S /',0')111 1 /'~ I 'tiI +11111111
A I I I t',r, I'.()1 IiAIt U'. 11 r' I 1/1 t?I 1111 t /". 1 .'101 UP C)
Ail. 16', U 11I.'> UAIt U! /r '/ 0)I') NIVi 1fSS 1!;0'11)33 751:.'01 0 ()()f)
A T L 2010 2113 DAB 0131 105 MTt,TFSS 750903 751201 0000
ATL 2235 JAX 0024 C:,6 0578 089 NTWTFSS 750903 751201 0000
DA CA TL 11,8 0725 0837 tTL 0632 089 MI b1TFSS 750803 751201 0000
11 AI: 100At 1!23 AI 1. 0?8/t I?4 "1 WTI SS 750')():1 751031 0000
DA t; 13Ct'? 1A~D0~2~t1~608r~hlems.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
ACI:N0T,LFDGFMLNTS
The design and development of UTPS is the product of individuals
too numerous to list. Among the institutions they represent are:
Barton-Aschman Associates; Cambridge Systematics; Coronet Corporation;
Consad Research Corporation; Creighton-Famburg, Incorporated: Deleuw,
Cattier and Company; DTM Incorporated; First Data Corporation; Mass-
achusetts Institute of Technology; Metropolitan Uashington Council
of Governments; National Bureau of Standards; the MITF;F. Corporation:
Peat, Marwick, Mitchell and Company: Planninc Research Corporation:
II. Pratt Associates: Wilbur Smith and Associates: TRW Systems
Group; Alan M. Voorhees and Associates: and the Urban Mass Transpor-
tation Administration, which Provided financial support for the opera-
tion of the above institutions as iYell as the preparation of this
paper and disclaims any Federal policy which might be read into it.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
AN EVALUATION OF THE AIR nUALITY TPTPACTS OF
TRANSPORTATION CONTROL POLICIES IN U.S. U?IRAN AREAS
Gregory K. Ingram
1. INTRODUCTION
Over the past two decades in the United States concern about the
quality of the environment increased dramatically. Concurrently, the
air quality in many urban areas deteriorated, and scientists demon-
strated direct linkages between pollutants such as photochemical
oxidants and the exhaust emissions of motor vehicles. These scientific
findings and the worsening urban air quality led to the 1970 amendments
to the Clean Air Act which set ambient air quality standards and em-
powered the U.S. federal government to require new motor vehicles to
reduce their emissions of certain pollutants by approximately ninety
percent. Although the phased reduction of motor vehicle emission
rates was the major policy proposed to meet the ambient air quality
standards, the 1970 legislation also allowed urban areas to limit
the extent of pollutant generating activit~y if the reduction in
emission rates proved insufficient. In many urban areas of the U.S.,
it appears that transport activity will have to be curtailed sonewhat
if the ambient standards are to be achieved on schedule. A variety
of tr.ansnortation control policies including improvements in public
transit, better traffic controls, and restrictions on automobile use
have been proposed to curtail transport activity in U.S. urban areas.
This paper analyzes a number of these transportation control
policies in two U. S. cities using a computer-based simulation model.
The objective of the analysis is to clarify the potential role of
transportation control policies in improving ai.r quality and to rank.
*Warvard University & National Bureau of Economic Research
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
alternative policies in terms of their cost and effectiveness. The
computer model that is used also could be applied to other urban
transportation policy problems, and the structiire of the model. Is
briefly summarized before the transnortation control policies are
presented.
2. TIT TASSTt! MODEL
The analysis of transportation control policies Is based on the
Transportation and Air Shed Simulation Model (TASSIM) developed with
funds from the U.G. National Science Foundation and the U.!. Depnrtment
of Transportation.1 This model forecasts levels of travel. activity,
pollutant emissions, and pollutant concentrations for distinct zones
within an urbanized region. The overall model, as illustrated in
figure 1, consists of three major components: a transportation submodel.
an emission submodel, and an air diffusion submodel.. Each of these sub-
models incorporates existing modelin' techniques. The transportation
submodel is based upon the urban transportation planning, model used
extensively in transnortation studies. The emissions submodel emnlovs
emission factors published by the U.S. Environmental Protection Agency
to calculate mobile emissions. The diffusion submodel was developed
by meteorologists.
The transnortation component of the TASSTM model is an adaptation
of the standard urban transportation planning (UTP) approach. 2
pproach. The
transportation submodel is composed of the usual steps of trip genera-
tion, trip distribution, modal split, and network assignment. Tn the
trip generation step, the total number of daily person trips which
originate in, or are destined to, each zone is forecast as a function
of the household, land use, and employment characteristics of the zone.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-2, 290
.
FIGURE J.
MAJOR COMPONEU'TS OF
THE TJiANSPORT AND AIR STIED S114UL LTIOrl MODEL
TRANSPORTATION SUBMOD:L
Forecast distributions of transportation activity
(trips and speeds) in each zone by mode and purpose
Lof trip.
EMISSION SUBMODEL
Derive area source emissions from automobile travel
using speed/emission factors. Input stationary
area source emissions by zone and large point
source emissions by location.
DIMSION SUBMODEL
Generate pollution concentrations in each zone
based on emissions and meteorological factors.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
These trip forecasts can be envisioned as the row and column totals
from a matrix that displays trips by origin-destination pair. In
the trip distribution step, a matrix of trips by origin and distribu-
tion is created that is consistent with the trip generation totals.
The trip matrix is produced by employing a distribution rule such as
a gravity model or an intervening opportunities model. In the mode
split step the trip matrix is subdivided into two or more trip matrices
indexed by mode (automobile, transit, etc.). Mode split predictions
are typically based on the socioeconomic characteristics of the popula-
tion and the relative cost and performance characteristics of the modes.
Finally, in the network assignment step, the trips from each origin
zone to each destination zone are loaded on a spatially detailed repre-
sentation of the transportation system. The network assignment typically
assigns trips to the shortest path in the network.
Three major changes distinguish TASSTN's transportation submodel
from the standard UTP anproach. First, the transport submodel has been
calibrated at a level of detail which is more aggregate than is usual
for such models. The Roston version of TASSI;', for example, considers
122 zones and 582 interzonal transport links for both auto and transit.
The TA.SSUN links comprise a spider network whose links represent either
all of the streets and highways or all of the transit routes that
connect zone nodes. Thus, the model produces link loadings on aggre-
gates of highways and arterials rather than flows on individual trans-
portation facilities. The more aggregate nature of the TASSIM model
produces enormous computational savings. Moreover, the spatial, de-
tail retained is sufficient to represent the distribution of transport
activity among zones and the changes in that distribution caused by
transporApprove or Hell ease '>t/~f~/19 : CIA-RDP79-00798A000200020005-9
Approve ~'
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The second change in the transportation submodel involves the
prediction of transit ridership which is shared by the modal split
and network assignment steps of the model. In the standard UTP model,
auto trips and transit trips are assigned separately to their respec-
tive networks. TASSIM employs a more flexible representation of net-
work assignment and mode choice. The two c.ateeori.es of trips repre-
sented are transit trips that are made completely on the transit mode,
and other trips that begin as auto trips. These latter trips can
either switch to the transit mode at some point or'else continue to
to their destination on the auto mode. The use of a composite network
(containing. highway links, transit links, and interinodal links) for
trip assignment enables the TASSTMM model. to simulate the effect of
parking charges and other vehicle restraint schemes on auto trips in.
a more realistic manner than wonnl.d be possible with the standard UT?'
model. The auto networks used in standard models are often so large
(containing 10,000 to 30,000 links) that it would be prohibitively
expensive to expand them to composite networks, whereas the composite
network for the TASSIM model is still of a reasonable size.
Finally, the third change incorporated in the transport submodel
is its use of a diversional routing or multipath assignment procedure
in the network assignment step instead of the all-or-nothing or capa-
city-constrained assignment procedures. The probabilistic multipath
assignment procedure used in the network assignment step selects a
set of feasible paths from each origin node to each destination node
and distributes the trips to these paths by means of a weighting
scheme.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The spatial traffic patterns forecast by the transport submodel
permit the calculation of auto vehicle miles travelled per zone as
well as auto vehicle cold starts per zone.3 Average vehicle emission
rates for a given fleet are multiplied by the estimates of vehicle
miles travelled and cold starts to yield emissions by zone from mobile
sources. Emissions from small stationary sources, such as home heating
units, are added to mobile emissions to give average area-wide emissions
for each zone. An area source diffusion model then predicts the zonal
concentrations of each pollutant which result from the emission patterns
and the meterological characteristics of the urban area.4 Emissions
from large stationary sources, such as electric generating plants, are
treated as individual point sources, and a second diffusion model is
used to forecast the pollutant concentrations in each zone that stem
from these large individual emitters.5
These diffusion models both assume that the relation between
emissions and concentrations is linear and additive, and that the
pollutants represented are nonreactive. That is, the models assume
that emissions can be related to concentrations by an equation of the
form:
C(I) - A(I,J) * E(J),
where C(I) is the n-dimensional row vector of the concentrations in
zone I(1 < I < n), E(.T) is the n-dimensional column vector of emissions
from zone J(1 < J < n) and A(I,J) is the n x n transformation matrix.
Two automotive pollutants, HC (hydrocarbons) and NOX (oxides of
nitrogen) combine in the presence of sunlight to form photochemical
oxidants, which are important secondary pollutants in many metropolitan
areas. The reaction processes that produce photochemical oxidants are
not well known, and at this time there are no simple diffusion models
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
that can predict spatial concentrations of oxidants with an acceptable
degree of accuracy. Therefore, photo-chemical oxidant concentrations
are not forecast in this analysis, and results are obtained only for
the oxidant precursors, 11C and t10X. The diffusion models used in
this analysis are elementary, even primative by some standards; but,
unlike proportional rollback models, they represent the effects on
air quality of redistributing emissions over space. If transport
controls are applied to selected portions of an urban area such as the
central business district, existing patterns of transportation activity
and mobile emissions will be altered. Thus the evaluation of transpor-
tation controls requires a diffusion model.
A major reason for using a model to analyze air quality is the
consistent framework it provides for systematically evaluating poli-
cies that reduce or redistribute emissions in urban areas. 1?hen
the TASSTM model is calibrated to a particular city, a wide range of
transport control strategies can be considered within the same land
use and transportation environment. The choice of the study site
is of crucial importance: analyzing, an atypical city may result in
unwarranted conclusions about policy impacts.
rather than searching for the "average" urban area, this study
has selected two cities, Boston and Los Angeles, which represent
quite different types of urban development in the United States.
Table 1 presents several socioeconomic and transport statistics that
summarize the characteristics of these two cities. Figures 2 and 3
display the analysis zones used in each city. Los Angeles has three
times as many people, but four times as much area and four times as
many automobiles as Boston. ?Moreover, in a typical weekday, residents
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005
Table 1
SUMMARY STATISTICS FOR BOSTON AND LOS ANGELES FOR 1970
Los Angeles
Population
3,023,000
9,008,400
Automobiles Owned
1,096,000
3,930,200
Area (sq. mile)
1,400
5,285
Daily Person Round Trips
3,125,038
12,243,975
Trips to Central Business
District
384,774
416,281
Transit Originating
Round Trips
348,252
.190,'052
Transit Originating
Share of Trips
11.1%
1.6%
Daily Vehicle Miles
Travelled
21,960,336
165,936,096
Average Auto Round Trip
Length (Miles)
Average Transit Round
Trip Length (Miles)
Average Auto Speed (mph)
Hours of Travel
[h.verage Daily Hours of
Travel per-Person
9.91
8.54
2,040,852
19.13
14.17
6,450,413
Emissions by Pollutant
in Grams/Second
19,090
2,839
1,057
64,215
17,601
9,541
Source of transportation and emission statistics: TASSIM simulations
and I..AR'TS Base Year Report.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 296
Figure 2
BOSTON ZONAL SYSTEM!
"r PO-LE I_r
p121
~~- -4 \>---
10. Cell tr.11i::c A)Sprb Wff61 Release 2001111/19: CIA-RDP7 167 0200020005-9"">">
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
in future years will also depend upon the nature of metropolitan nrowth
which actually occurs.
These policy simulations both in Los Angeles and in Boston indicate
that the predicted impact on air quality of reducing automotive emis-
sions Is dramatic. Aggregate mobile emission rates and pollutant con-
centrations of CO, TIC, and NOX in the air quality control regions fall
substantially. For the 1976 fleet aggregate mobile emissions fall by
fifty-two percent for CO, by sixty--two percent for IT, and by twenty-
two percent for ;[OX; for the 1980 fleet aggregate mobile omissions are
down eighty-one percent for CO, eighty-five percent for TIC, and sixty
percent for ITT?X.
Carbon monoxide is emitted primarily by motor vehicles, so the
forecast reduction in zonal CO concentrations closely parallels the
forecast reduction in automotive emissions both in Los k.ns*.eles and in
.)oston. For CO the decline in concentrations is not exactly propor-
tionate to the decline in aggregate emissions because the standards are
projected to have a greater effect on running than on cold start emis-
sions. Since emissions of hydrocarbons and oxides of nitrogen are
produced by both stationary and mobile sources, the forecast reduction
in concentrations for these pollutants shows considerably more varia-
tion.
Because of differences in the composition of stationary source
emissions in the two cities, reducing mobile emissions of hydrocarbons
,tenerally produced a greater reduction in zonal concentrations of hy-
drocarbons in Boston than in Los Angeles. A comparison of NOX concen-
trations in the two cities revealed that the opposite is true for NOX
emission reductions. Nevertheless, in Los Angeles, the zones with the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
highest concentrations of stationary source hydrocarbons had their
hydrocarbon concentrations reduced thirty-five percent by the 1980
fleet emission rates. In the Boston zones with high concentrations
of NOX from stationary sources, a twenty-five percent reduction in
14Oy concentrations was achieved by the 1980 fleet emission rates.
h. Discouraainh Automobile Trips to the Central. Area
Three types of local vehicle restraint schemes were simulated:
parking charges, local licenses, and automobile prohibition. The
restraint areas were defined as zone 1 in Los Angeles and zones 1, 2
3, and 6 in Boston. These zones constitute a generous definition of
the central. business districts (CBDs) in the two cities.
The parking charge simulation raised the average cost of parking
by fifty percent. The area license scheme placed a 254 toll. on all
trips which originate outside and enter the restraint area. Trips
which originate within the area are exempt from the local license toll.
Outright prohibition is similar to licensing except entry cannot he
purchased at any price.
Although traffic flow changes do not endogenously change speeds
on the transport network in TAgSIM, speed changes can be introduced
exogenously if there is reason to believe such changes will actually
occur. For these simulations it was assumed that the reduced traffic
increased auto speeds by ten percent in the restraint area for all
policies except the parking charge, which does not reduce through
trips. In addition, a second prohibition simulation slowed auto
speeds by 10 percent in zones adjacent to the restraint area to repre-
sent the congestion caused by traffic diverted from the restraint area.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The results of the simulations were generally similar for the two
cities. On an aggregate level, all policies increased transit use
and reduced auto vehicle miles traveled. All. simulations except the
Los Angeles prohibition-with-congestion run reduced aggregate mobile
emissions. This exception occurred because the congested adjacent
area modeled in Los Angeles was large, and lower speeds in that area
caused an increase in CO large enough to offset the savines from the
reduction in vehicle miles traveled.
The aggregate improvements from the restraint policies are not
large, however; the really substantial concentration decreases occur
only in the zones directly affected by the policies. The licensing
and prohibition schemes induce auto tripmab.ers destined for the CBT)
to switch to transit outside the restraint area and to complete their
trips on the public mode. In addition, they divert through trips
into adjacent zones to transit, but they do not divert through trips
from the restraint area. `3 As a result, while all three traffic re-
straint policies significantly improve air quality in the restraint
area, licensing, and prohibition produce greater improvements than
parking charges.
It should he noted, however, that the auto trips redirected to
transit, the diverted through trips, and, if it should arise, the in-
creased congestion in neighboring zones caused by traffic-restraint
policies, all can have detrimental effects on the air quality in
these adjoining zones. Through these effects were often not large,
they are a reminder that central area restraints can have certain
undesirable air quality impacts.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
c. Ilakinj Transit Alternatives btore. Attractive
A variety of transit improvements are often proposed as techniques,
for improving air quality in central sections of metropolitan areas.
It is argued that making the transit system more attractive will divert
drivers from their vehicles, reduce vehicle miles travelled (WIT), de-
crease auto emissions, and improve air quality.
The TASSIM model has simulated in Boston and in Los Angeles the
eFfects on air quality of transit fare reductions and performance
increases. Fare reductions of ten percent in Boston-and twenty percent
in Los Angeles were tested. In addition, twenty-percent reductions in
total time required for transit trips were examined. A central area
bus lane scheme was tested in Los Angeles. Finally, the impact of ex-
tending the rapid rail system in Boston was considered.
The results conformed to results of other studies of transit rider-
ship. Responses to fare reductions were less than responses to propor-
tionate travel time reductions. Price elasticities for transit are
low--about -.4 for Boston and -.2 for Los Angeles. In other words, a
twenty-percent fare reduction increases transit ridership by about
four percent in Los Angeles and eight percent in Boston. Transit level
time elasticities are typically larger--about -1.0.9
Because the percent of trips on transit is not large, the fare re-
duction and performance improvement produced barely perceptible reduc-
tions In the total number of auto trips made in the two analysis areas.
The rapid rail extension in Boston did increase the share of transit
trips, but was unsuccessful at improving regional air quality. The
Los Angeles bus lanes improved air quality in the CBI), but resulted
in imperceptible changes for the rest of the region. Since the transit
systems in both cities provide more complete service in the central
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
areas, fare reductions and improved transit travel. times die generate
some noticeable improvements in central area air quality. For example,
in some zones concentrations of. CO were reduced as much as thirty per-
cent by the overall transit travel. time reductions.
Although there is no persuasive evidence that aggregate emissions
are reduced by transit enhancement policies, substantial air quality
improvements can be generated in specified areas. Reductions in .
transit travel times have somewhat greater effects than fare decreases
or extensions of transit service in producing these local improvements.
d. Tiegulating Traffic Flow
The emission rate of automobiles net vehicle mile is a function of
vehicle speed: emissions of CO and 11C decrease sharply with speed,
whereas emissions of VOX increase slightly as vehicle speed increases.
Thus, policies that increase vehicle speeds have often been proposed
to improve air quality. Although increasing vehicle speeds could be
envisioned as a region-wide policy, it is most frequently advocated
for centrall.v located, highly congested zones where average speeds
are quite low, ranging from ten to fifteen miles an hour. TncreasinF,
speeds from such low levels can significantly reduce emissions of
f'O and 11C per vehicle mile. Conversely, policies that may reduce
auto speeds could he expected to degrade air quality.
Four simulations that investigated alterations in vehicle speeds
were carried out for both Boston and Los Angeles. The first policy
simulation increased automobile networl: speeds by ten percent in a
small. central area, and the second policy simulation applied a speed
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
decrease in the same central. area. The third and fourth policy siruula-
ti ons respectively increased and decreased automobile network speeds
in the entire re-ion.
The aggregate statistics from the region-wide speed chances are
consistent with expectations. Faster auto speeds reduce total emissions
of CO and ITC and increase slightly those of NOX.1< Slower regional
auto speeds have the opposite effect. however, changes in the spatial
pattern of pollutant concentrations in Boston differ from those in
Los Angeles. "hen auto speeds fall in Poston, the transit system
attracts enough triprinkers to reduce concentrations of all pollutants
in the central zones. Conversely, faster auto specOs lure away enough
transit riders to increase concentrations of all pollutants in the
central. Zones. in Los Angeles, the transit svatem carries proportionately
fever riders. Therefore, very little slii.ftinr occurs, and changes in
central area concentrations are similar to regional changes.
The small area speed changes reveal another difference between
Boston and Los Angeles. In both cities, local auto speed improvements
discourage transfers from automobiles to transit. Auto speed reduc-
tions have the reverse effect. However, central area speeds are sub-
stantially slower in Boston than In Los Angeles. liven the relationship
between speeds and emissions, a ten percent speed change in Boston
produces a smaller percentage change in individual. vehicular emissions
than a ten percent speed change in Los Angeles. Therefore, changes in
through trips and switching behavior offset changes in individual ve-
hicular emissions in Boston but not in Los Angeles. The small area
speed change simulations suggest that in Boston speed improvement stra-
tegies that are not implemented with vehicle restraint systems that re-
strict auto access to th area will have a hi ph robability of failure.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 309
Tn Los Angeles, they will generate slight changes in the expected direc-
tion, i.e. speed increases will improve air quality. However, it is
clear that in Los Angeles, as well as in Boston, adding vehicle re-
straints when auto speeds are increased would generate greater gains.
e. Increasing Au-to- r,obile Occupancy
It is well known that the average auto vehicle occupancy rate is
quite low relative to vehicle capacity. This knowledge generates
enthusiasm for policies to promote more car pooling for. work. trips.
However, because data on accepted and rejected car pooling opportunities
are virtually nonexistent, few effective policies have been devised to
promote higher vehicle occupancies. Although the TASSIM model yields
few insights about the determinants of car pool.ing, it can be used to
examine the air quality impacts of achieving specified higher levels
of auto occupancy. !,?hen this policy is simulated by the model in the
two cities, substantial and proportionate reductions in auto emissions
and improvements in air duality are obtained.
Of course, car pooling can have effects on transportation demand
that ro beyond simply increasing the average number of persons per
vehicle. If car pooling occurs mainly on work trips, some workers
will leave their automobiles at home several days during the week.
These automobiles will then be available to other household members
who may use them for shopping, school, or personal business trips.
This greater availability of automobiles caused by car pooling might
increase the total number of trips made, thereby offsetting some of
the pains from car pooling;. A partial test of this phenomenon, using
the Boston version of TASSIM, indicated about a third of the expected
air quality improvements are eliminated if car pooling increases auto
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
availability and more trips are made. The number of trips may remain
the same, however, if the primary effect of car pools is to reduce
the number of second and third cars owned. If the increases in auto-
mobile occupancy are achieved primarily through car pooling for work
tripe in central zones, and if the additional trips generated are
primarily suburban shopping trips, then significant central area air
quality improvements may be generated which are greater than the reduc-
tions in aggregate emissions. These considerations suggest that in-
creased car pooling will not reduce automobile emissions uniformly
over an air quality control region.
F. Contra l l in-Urban Devel onnent Patterns
Some analysts, recognizing the interrelrntionship among land use
patterns, travel activity, and stationary emission source locAtions,'
have proposed that land use regulations and controls be used as policy
variables to improve urban air quality. Implementing land use plans
to improve air quality is difficult, however, because little systematic
knowledge is available about the exact relations between land use and
air quality. Therefore, there is little basis in terms of air quality
effects for accepting or rejecting alternative land use plans.
The TASSIM model was used to investigate one particular hypothesis
about the impact of land use on air quality in Boston. This hypothesis
states that higher density development of central portions of metropo-
litan areas improves air quality by reducing the demand for transporta-
tion and particularly automobile travel and, conversely, that existing
patterns of low density, dispersed development degrade metropolitan
air quality by encouraging the use of autos and requiring longer trips.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Two scenarios of altered land use patterns were simulated with
the Boston model. The first increases centralization in the air
auality control region by moving twenty percent of the residences and
employment located in the outlying ring to the central core. The
second scenario increased the extent of decentralization by moving
twenty percent of the core's population and employment to the ring.
Most of the transport statistics produced by these simulated land
use changes conform to the expectations which suggested the policies.
Centralization reduces the aggregate number of trips, trip lengths,
and the aggregate levels of emissions of auto vehicles. Decentraliza-
tion has the opposite effect on all of these variables.
The simulations predict certain other changes which have opposite
air quality implications, however. Centralization causes average
auto speed to decrease. In addition, auto trips to and through the
downtown area increase. The net impact of these forecast changes
from centralization is to increase concentrations of auto pollutants
in the central zones by as much as 10 percent, and to decrease concen-
trations in the ring zones by as much as seven percent. Decentraliza-
tion causes opposite, but more dramatic concentration changes, since
the movement of jobs and people is absolutely greater.
Reducing low density peripheral development and increasing central
densities may not be an effective policy for solving urban air quality
problems unless it is combined with other policies that fundamentally
change patterns of transportation usage. If altering patterns of travel
usage is difficult, then the process of dispersal and decentralization
underway in most metropolitan areas may actually help to reduce the
concentrations of primary pollutants in the centers of those areas.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Since decentralization increases total emissions of TIC and NUX,
however, it might increase concentrations of secondary pollutants
such as photochemical oxidants in some cities.
r;. Summary of Polite Tff.ectiveness
The preceding sections have provider.] a descriptive overview of
the predicted effects that several control policies have on the trans-
portation system and on air quality in two cities. This section
summarizes the results of these policies with a region-wide and local
'' effectiveness' measure to quantify the impact of the policies on air
quality.
The first measure of effectiveness, the regional effectiveness
index described in the first section, measures region-wide violations
of the ambient air quality standards in terms of person-times of ex-
posure to pollutant concentrations that exceed the standards.11 The
second measure of effectiveness focuses on the local or target area
improvement in air quality; it is the percent change in the concentra-
tion of carbon monoxide in the central. business districts of Los
Angeles and Boston. Many of the transportation control policies
that were simulated were applied only to the two cities' central areas,
so this latter measure indicates how much locally applied policies can
improve local air quality. The regional effectiveness index, on the
other hand, better summarizes the impact of policies on the entire
air quality region.
The local regional effectiveness of the policies are compared in
figure 4 for Boston and in figure 5 for Los Angeles. The origins of
these figures represent the benchmark run. Not easily read from the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
313
FIGURE 4
LOCAL AND REGIONAL EFF1 CTIVETNESS OF POLICIES FOR BOSTON
? '6
56
118
44
36
28
211
Transit Improvemcl t L
16
Transit Vxtension
-Boo -6o0 -400 -200
Increase Auto Speeds 0 -1+
Increase Central 0
Auto Speeds
UCentra.l.i ze
-8
1980 Emissions 0
Q)1976 Emissions
0Prohibition & Congestion
OProhibition
QLocal License
Raise
Parking Charge
r?e rni Decentralize
O Decrease Central Auto Speeds
C )Decrease Auto Speeds
OBaisc Auto Occupancy
IR diic=T ansi.t ]tr rc ~__~
200 1oo 600 800
Regional Effectiveness
(Reduction of Peron-Times of
Exposure in Thousands)
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
FIGURE 5 314
LOCAL AIZD REGIONAL EITECTIVEN1:,~Sa OV POLICIES FOR LOS ANGELES
80.1-
1980 Emissionsi
55+
154-
0
Increase Central -~ - _ - `
Auto Speeds lp0 yTransi Rma 'sociuot0ccupancy
Reduce Transit Fares
5r.~
ve.L Times
xeaucc Transittrra
{
-800 -400 h Jo 8ob 1200 1600
_5 : Reduce Central Auto Speeds
Regional Effectiveness
(Reduction of Person-Times of
Exposure in Thousands)
0 Prohibition
50 1976 Emissions
Prohibition & Congestion
0 h5 0 Local license
40 (Raise Auto Speeds
35
75
70
65 - ;
60
30.t-
25-F-
0 Reduce -zU 7
Appru" Fa A*ase 2001/11/19: CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
diagrams themselves is the observation that local rather than regionally
directed policies have the greater relative impact., with the exception
of the emission rate reductions. These diagrams highlight the differ-
ences between the policies in Los Angeles and Boston. Although many
of the policies differ in their relative effectiveness, the only
policies that act in opposite directions in the two cities are the
speed increase and the speed decrease policies, as described under
heading d. above. The only other policy that degrades air quality
is the centralization policy simulated for Boston. Extending the
transit system produces a negligible local improvement, but some
regional air quality degradation in Boston. All other policies to
discourage auto use and to increase transit ridership have positive
regional and local effectiveness. As expected, policies applied
only to the central business districts have greater local than regional
effectiveness.
Finally, figures 4 and 5 illustrate how the effectiveness of'
emission reduction compares with the transportation control policies.
The 1976 fleet emission factors are more effective locally and region-
wide than any of the transportation policies simulated with 1970 fleet
emission factors except for the regional improvement from decentrali-
zation simulated for Boston. The 1980 fleet emission factors produce
substantial improvement in air quality relative to 1976. Because
fleet emission factors will decline so markedly during the late seven-
ties and early eighties, it is probable that vehicle control strategies
will be required to meet air quality standards in most metropolitan
areas for an interim period of only a few years. Of course, in the
longer run, increases in automotive usage could mean that localized
control ~ ~~o~~Cell @ 260M 1'15 Jkr- 79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
4. COST EFFECTIVENESS OF POLICIES IN BOSTON AND LOS ANGELES
Transportation control policies that have similar impacts may
have widely different costs. It is necessary to determine costs
for each policy to make judgments about their relative efficiency.
Many simplifying assumptions must be made to evaluate each policy,
but even crude estimates of the cost and effectiveness of various
policies allow choices to be made among them. In this section re-
source costs for several of the policies simulated for Boston and
Los Angeles are roughly estimated and compared with the two effective-
ness measures used in the previous section.
Many components of the costs of the simulated policies can be
derived from the model itself. For example, when transportation poli-
cies'are applied, the model estimates changes in the use of automobiles,
in transit ridership, and in the amount of time devoted to travel,
which then have to be valued. The model provides no insights about
the capital or operating costs required to implement a policy. These
figures must be estimated separately, but estimating these cities
costs for some policies is difficult because there are few examples
of their use. 12
For each policy the model forecasts total vehicle miles travelled
(VMT) the number of , round trips that originate on transit, the number
of passenger miles travelled (PMT) on transit by persons who transfer
from auto to transit for a portion of their trip, and the number of
hours per weekday that people spend travelling. Table 3 displays
these four quantities from the benchmark runs for Boston and Los
Angeles. Since policy costs are derived from changes in each of
these four measures, the absolute levels displayed in table 3 provide
a bas&pooevedaHdpget setb#ttflqhaCg&,IRDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
317
BENCHMARK TRAVEL MEASURES
FOR BOSTON AR D LOS ANGELES
Transit
Originating
Round Trips
Transit
l'NP From
Auto
Hours of
Travel
21,960,36 165,936,000
348,252
2,040,852 -6,450,413
Source: TASSIM simulations
Boston Los Angeles
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
446,367 ? 326,321
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The changes relative to the benchmark run in each of the four
transportation measures for each policy simulation are displayed in
tables 4 and 5 for Boston and Los Angeles respectively. For example,
the first entry in the VMT column in table 4 indicates that vehicle
miles travelled relative to the benchmark declined by 71,000 in
the parking-charge simulation for Boston. For the same policy relative
to the benchmark run, transit originating trips increased by 9,008,
transit passenger miles travelled (PTIT) from auto originating trips
increased by 379,167, and total person hours spent travelling rose by
8,835. The pattern of changes in the travel measures displayed in
tables 4 and 5 is similar. If a policy increased a particular travel
measure in Boston, it usually increased the travel measure in Los
Angeles. The absolute changes in the travel measures are typically
larger in Los Angeles than in Boston but, when scaled by the size
differences of the two cities, they are generally similar. The most
significant differences between the two cities involve the changes
in transit originating trips. These differences are attributable to
structural differences between the alternative modes of travel in
the two cities: Boston has a much more highly developed transit system
than Los Angeles.
The travel measure changes displayed in tables 4 and 5 were multi-
plied by costs or by prices to value them in dollar terms. VMT are
valued at six cents a mile, an amount that approximates incremental
operating costs for autos. Transit trips are valued at sixty cents
each, which approximates the average cost per passenger trip in the
Boston and the Los Angeles areas. Finally, hours of travel time have
been valued at $1.75, which is one-third of the median wage rate in
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 319
Table 4
CHANCES IN TRAVEL MT:ASU1 ES RELATIVE TO 13ENC1D1AILK
'OR SEVERAL POLICIES APPLIED TO BOSTON
VMT
Transit
Originating
Trips (PMT)
Transit
PMT from
Auto
Hours
of
Travel
Parking Charge
-71,000
9,008
379,167
8,835
Local License
-220,000
-1,297
330,685
17,362
Prohibition
-265,000
592
375,968
21,177
Prohibition &
-444,000
62
403,721
15,795
Congestion
Fare Reduction
-247,000
9,035
72
19,018
.Improved
403,000
21,632
683,345
25,046
Transit
Performance
Transit
22,000
40,905
44,203
17,513
Extension
Raise CED Auto
166,000
1,309
-88,342
-3,247
Speeds
Lower CBD Auto
-166,000
-1,276
105,348
1,003
Speeds
Raise All Auto
659,000
--5,060
-96,110
-110,434
Speeds ?
Lower All Auto
-613,000
4,307
134,601
106,211
Speeds
Increased Auto
-1,975,000
0
0
Occupancy
-1,111,000
36,628
40,598
-13,420
1,.1020,000
-57,606
-83,330
-15,575
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Source: TASSTl1 -,i 1.atic-nr
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
CHARGES XN Ti{AV`,.::L 1?M"ZURES REL!!TIVE TO I3E:ICI"J RK
FOR SF,'VER:'J, POLICIES APPLIED TO LOS A::GEL:r.'S
Transit
Originating
Trips (PMT)
Transit
F?.2 frc~
Auto
Hours
of
Travel
Par}in6 Charge
-2,198,000
4,380
3,].27,652
100,417
Local License
-1,595,000
436
3,017,695
60,070
Prohibition
-2,108,000
? -864
3,901,766
72,350
Prohibition &
-2,080,000
2,679
4,118,656
201,292
Congestion
Fare
-258,000
17,974
--1,541
3,106
Reduction
Central. Bus
-2,546,000
21,000
2,600,000
-2,332
Lanes
improved
-3,302,000
58,703
..3,49(i,156
-16,007
..Transit
Performance
Raise CBD
260,000
-491
-326,231
-23,792
Auto Speeds
Lower CBD
-211,000
210
364,386
16,808
Auto Speeds
Raise All
468,000
-6,640
326,231
-1,172,193
Auto Speeds
Lower All
-478,000
3,876.
509,599
629,898
Auto Speeds
Increased
15,184,000
0
0
Auto
Occupancy
Source: "RTediFR17,RtWse 2001/11/19: CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
the two cities. 13 Si.nce the travel measures are for weekdays, the
weekday valuations have been multiplied by 300 to transform them
into annual values. 14 The results of these calculations are dis-
played in tables 6 and 7. For exariple, the first nonzero entry
in the W7 column in table 6 is -,",1,27,),000. This is obtai.ned by
multiplyinc; -71,000, the charge in V'"IT from parking charges in
Boston, by $.06 and then by 300.
Tables 6 and 7 also display estimates of the capital and adminis-
trative costs of implementing most of the policies. The costs shown
for emission reduction assumes that the aver;w a i.ncremental cost per
car (relative to 1970) of reducing emissions in Boston is $40 in 1976
and $65 in 1930, while in Los Angeles the figures are fi50 in 1976
and $85 in 1980. Capital and administrative costs are not displayed
for some of the nolicies because of uncertainties about policy imple-
mentation. For example, increasing the averace occupancy of automo-
biles by ten percent above their current levels would reduce V"T pro-
portionately, but estimating the cost of obtaining such an increase
is difficult because the determinants of car pooling are not well
understood. Similarly, it would be very diffi.cult to estimate the
costs of centralization for Boston or for Los Angeles.
The costs in tables 6 and 7 are plotted against the effectiveness
of the various policies in figures 6 through 9. For most of the poli-
cies the local effectiveness diagram resembles the regional effective-
ness diagram, and the Boston diagrams resemble the Los Angeles di.a-
grams. For example, in both cities all effectiveness measures increase
steadily from fare reductions to parking charges, to improvements in
transi.t performance, to local licensing, and finally to prohibition.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19.: CIA-RDP79-00798AO00200020005-9
322.
sJ
cccc u
.1 ?ri 43
ri Si to
a. C ?rt
to 14
U a.?
o~
a
to
N
0
to
i+
14'
O O 0
O O
M O
N N
C.
~. N C
r
O
rt
t'.
N
tr1
ap
C-
,
o'
O
N
'
co
r4
0
r4
r?i
w
. f-
w
u.
1
aw
w
q%
1
tit
in
O
O
cl
co
0
to
0
N
Ln
%01
co
N
M
S
tl
%D
w
r1
N1 w
co
w
0
w
r'?t
O
;
1
1
N
r-I
M
O
v
O
N
co
0
LM
to
0
O
Co
w
1
r~
-I-I
1
i
O
(J%
co
OD
Co
r*
r1
h
'T
0
O
0
M
ON
C%
Co
0
0
0
N
N
!V
~,
1
w
.-l
w
N
N
N
O
0
O
M
C-
M
'N
O
0
O
O
O
O
rt
Approved For Release 2001/11/19 : CIA-RDP79-00798A00'0200020005-9
O 0 N at . 9-1 . F 0% a
CQ O C) O ?.7 t1 rIj
C7 O F? Oh r4 fN Co r' H r1 %D 0 co f'1
.y n ra H N ri '9
-( H
r
O r q f-1 N On r-1
w w w 4 w w
rl
cr
O 0 P% N Co
w w A w
cj%
t-i.?1
r r, l r-t
.4 t7 r-i
N ; N
N
?? 0 co O O N tD t
f`. %0 h. 0% -r In
N On F'- Or - tV
rd M ??7 N -' I'-
1 1 1 1 '1
0 0 O O O
O 0 t9 r) M
M rj N
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
323'
o
0
CMr)
O
co
^ -
r1
0
C3
%0
Co
.` )
c
w
o
w
c.
w
O
w
n
w
r??
w
-1
w
r-{
w
C.
w
CO
w
0
L?+
n
-
M
tD
M
Co
N
h
M
00
N
M
.4
-4
CO
;
O
O
r-1
N.
^
It
0
co
M
co
e
N
a
O
n
tr)
CT
d
w
w
w
%0
w
%D
w
N
w
.7
w
.7
td?
O
r" 1
CO
w
E-4 L-c
N
M
M
C?
1
H
y
1
H
V
b0
H
O
O
O.
1\
r-I
N
O
0
Co
0)
ra
1J ?r{
0)
M
00
Co
.7
N -
O.
r-
0
4J c rj to
f:
Cl)
w
Ln
w
N
w
co
w
1
CO
w
1C0
w
M
w
O ?rl c3
b
1?
C)
.
c~+
< ?rt1 H
CO
n
.
.O
O.
C9
co
I
0
to
11 ?r{ A)
?
O
O
n
1`
r-I
.7
r-1
O
M
1-
V) cJ
U)
CC r co
U')
a
H
r-1
%
a)
^
w
co
N'
Co
H
r.
i
0
sa ra P
H?
f-c
%0
OP
H
L-:
,
~
1
-t
W
to
0
&
O
O
--
O
-
C-
IT
CO
%n
0
r
I
N
.D
r-I
IT
.7
ti
N
M
G7
O
O
it
w
O.
w
.~'
w
.7
w
OD
w
IT
w
.D
w
to
H
to
O
0
O
v
U
W
CA
ci
co
1
f~
1
f-
I ce)
i
I
W
&5 4
1
U
4.1
O
O
0
0
b
O
C)
0
0
O
C-
a
CD
CO
v')
tr)
to
,n
O
0
o
O
f"-
n
N-
N-
C)
0
C)
O
?r{ u
w
w
w
w
w
tr.
1-3
Cn
O
?r{ J U)
C ~ ?
1-.
(7%
~'
cn
N
n
N
w
N
w
0
-T
CV
r2
1~
V C- ?r4
a -a
r-1
cn
0
7 00 i.,
tr n 0
M Co H
w' w w
O. 00 ' M
I ra
.D 0 to
f`- O. 0.
I M M
; i r.. j
to .7 - N
O. C11 C> r-1
f~ -7 '.D Cr)
w w w ,?
h 00 co
01
N
0 ? c-- C-
U) to 0 Cl) cJ
10 10 .0 10
N C4 E Q. O
0 0
iJ JJ
A
C]
r-1
H b
V
Co
C
6 )
N
a
$ 1
?~{
o
rr~?
o C)
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release, 2001/11/19 : CIA-RDP79-00798A000200020005=9 324
FIGURE 6
COST J WD LOCAL EFFECTIVEUESS FOR } OSTON
y Transit Extension
9O-
0
A
1
1980 Emissions 0
Transit Improvement
1976 Emissions Q
$4 2 0 Prohibition
51
Local License
Parking C, )Ch arge O Prohibition & Congestion
cn. 14 ,.~ ODecentralize
V 4 Reduce Transit Fares
~QDecrease Central A to Speed::
'0 2b 50
Local Effectiveness
? (Percent CO Reduction in. CTBD)
Approved For Release '2001/11/19: CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
FICU1tE 7
COST AND REGIOIIAL EF17ECTIVENES S FOR BOSTON
loot
.Transit Extenyion
Transit Improvement
0 1976 Emissions
325
c'rohibition
0. Local License
C) Prohibition & Congestion
0 Raise Pa1?kj.nG Charge
Increase Central UReduce Transit () Decentralize
-1160 -"-00 0~ 660 8i 0
Auto Speed.,,
Decrease Central Auto Speeds
Regional Effectiveness
(1eduction of Person-Times of
Approved For Release 2001/11/19 };Q QIF 7a-ROTOFAOQQ C)020005-9
Approved For Release 2001/11/19,: CIA-RDP79-00798A000200020005-9
FIGURE 8
COST AND LOCAL EFFECTIVENESS FOR LOS ANGELES
390
360
330 -r
0 Txansit Improvement
0 Raise Parking Charge
1980 Emissions
1976 Emissions
0
0 Prohibition E Congestion
0 Prohibition
0 Local License
30-
Decrease Central Auto Speeds
( (ILICC '_1't_T1S~l.t F Ires
6 6 o 18 24 30 36 4~ .4 54 60 66 7 7~8
?.30 T Inci'ease Local Effectiveness
Central (Percent CO Reduction in CUD)
Auto
Speeds
Approved For Release 2001/11/19 : CIA-RDP79-00798A00.0200020005-9
Approved For Release 2001/11/19: CIA-IRDP79-00798A0002000200050
PICURE 9
COST AND REGIONAL EFFECTIVENESS FOR LOS ANGELES
1980 Emissions
?1976 Emissions
-200
. 0 Prohibition
o Local License
O Transit Improvement
Q Raise Parking Charge
Decrease Central Auto Speeds
Reduce Trhnsit 17ares
Increase Regional Effectiveness
Central (Reduction of Person-Times of
Auto
Speeds Exposure in Thousands)
c) 200 400 60b 800 1600 12260 1
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Prohibition and congestion are similar to prohibition in. Boston,
but are less effective than prohibition in Los Angeles. The major
differences between Boston and Los Angeles occur in the central area
speed changes, as explained earlier.
Reducing auto emissions to the legislated 1976 and 1980 levels
both in Los Angeles and in Boston is cost effective relative to the
transportation controls, although more clearly so in Los Angeles than
in Boston. Several of the transportation controls buy small impr.ove-
rients in the effectiveness measures with expenditures that are also
small, but large improvements in air quality are obtained only from
the significant reduction in auto emissions over time.
The transit improvement policies other than reducing transit
fares and implementing reserved bus lanes are not very cost effective
in the two cities. In Los Angeles, similar improvements in effective-
ness can be obtained at less cost from parking charges, whereas, in
Boston, improving transit performance is dominated by local licensing,
or Prohibition. The central bus lanes simulated in Los Angeles con-
stitute one of the more cost effective transit policies. The transit
extension simulated for Boston is not cost effective and actually de-
creases air quality as measured by the regional effectiveness index.
One final policy in the Boston simulation that deserves mention
is decentralization. The implementation costs of this policy are
assumed to he zero since the suhurbanization of residences and employ-
ment will undoubtedly continue in the future. Decentralization per-
forms very well on the measure of regional effectiveness because it
reduces concentrations of primary pollutants. The model does not
predict concentrations of photochemical oxidents, however, and
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
decentralization may increase levels of this pollutant by increasing
aggre-ate emissions of. NOX and 11C.
5. CONCLUSION
This analysis of transportation control policies in Los Angeles
and Poston has evaluated the air quality impacts of several policies
by combining transportation and air quality forecasts produced by a
computer simulation model with economic calculations of the costs of
the policies. A comparison of the costs with the air nuality impacts
of the policies yields an assessment of the relative cost-effectiveness
of the policies. This assessment can then be used as one means of
choosing among alternative policies.
Although this paper has focused on air quality impacts, It is
obvious that the transportation portion of the TASSIM model could be
used to carry out preliminary analyses of other issues. For example,
the effect on the transportation system of various land development
policies could be approximated with the model- or preliminary esti-.
mates of ridership levels on extensions to transit systems could be
provided. Of course, at the project level, more detailed analyses
would be required, but the TASSTM model could help analysts identify
those projects which merited further analysis. The model is inexpen-
sive to operate and a large number of alternative policies can be com-
pared fairly quickly at the regional level. Only those policies or
programs which are particularly effective need be subject to project
level evaluation. The use of TASSIM or other so-called "sketch
planning" tools should help give analysts an overview of a wide
range of policy alternative and help them to focus their project
analyses more effectively.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
A_PPM1DTX: COMPUTATIONAL SUMMARY OF TITE MODEL
The TASSIM model is typically executed on the computer as three
separate computer programs named TASAC-D, TA`?SIM, and TASMAP. The
TASAQD program of the TASSIM model supplies data on air pollutant con-
centrations due to large point sources to the TASSTM program. TASAOD
is a simplified version of the Martin-Tikvart Air 'ua].ity Display
?odel. It could be integrated into the same physical code as the
TASSTN program. However, its isolation lowers the costs of TASSTM
forecasts, since the majority of policy simulations look at strategies
which have no effect on point source emissions. Therefore, the TASAnD
program, under integrated operation, would simply reproduce its output
over and over again. For Boston, the TASAOT) program requires slightly
less than two minutes of CPU (Central Processing Unit) time and about
2.3 minutes of total job run time per simulation of the 370/165.
About 115 I:-bytes of core storage are required to compile and execute
the program at a cost of about twenty dollars on the TTarvard-.''.I.T.
370/165.
The TASSIP" program performs virtually all of the operations de-
scribed in the introductory discussion of the conceptual model. It
requires .95 minutes of CPU time and about 2.7 minutes of total job
run time per simulation on the 370/165. The Boston version of the
TASSIN program requires about 156 P.-bytes of core at a cost of about
twenty-five dollars per run. Of course, these run times and storage
requirements will vary if the dimensions of the model change. Faster
run times could be obtained with IT compilation and the use of object
decks, and lower costs could he achieved by storing the input data.
on a user disk.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
TA^NAP is a supplementary program that graphically displays pol.lu-
rant concentrations on maps for the entire study area and for the cen-
tral core area. The naps are visual displays of information printed
out elsewhere in the simulation run. The TASINAP routines could also
he intec'rated into the TASSIP4 program, but using TA!'%Ir to print out
several naps at once is only sli'htl.v more expensive than printing
out a sinc'le man. Therefore, it Is efficient to complete several
simulation runs and then to create maps for all of the runs at once.
Thus, TAVA.P, like TASAOD, is physicall,t separate from the TASST'T
program. Unlike TASAOT) however, TAc*tAP is not essential to the
TA^ST*'' model, since it merely provides graphic displays of outnut
produced by the TASST'T program. A typical map produced for Poston
is displayed in figure 10. The Poston TASNAP program requires .9
minutes of CPU time, 1.4 minutes of total job run time, and 11.8 k.-bytes
core to produce twenty maps.
To adapt the TASSIM model to another city requires tables of travel
times and trips, inventories of stationary emissions, meteorological
parameters, vehicular emission rates, and socioeconomic characteristics
of the population for calibration. Our experience with the Los Angeles
version. of TASSI?' suggests that transf.erring* the model to another city
requires from six to eight man-months and about four hours of computer
time on an T_R*t 370/1.65.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005392
nI N
P P N N P N
N N N N N N
N N
N N
N N N
N N N
N N M
N N U
^ N 1
N N N
N N N
N N Y
N N N P
N N N N N 1
N N P N N N N
N N N N N N N
N N^ N N N N^
N N N N N N N N
N N Y N N 1 1 1
P N N N N N N N N
P N N N N N N U^
^ N N? N N N ry^
N N N^ N N N r
N^ N N N N N 4^ .?.
N N N r N N N Y N n .?.
N N P N N N N 11 N N ??
? N N N N N N N N N n
N N N N N 1 N N N
P N N P N 1 1 N N ...
N N N P N N N U N
II...
N N 11 N 1 I' 11
?
b N N N II I 4 ^^
N N r N t' II
N N P N 1 1
^ 1 1 1 1
P P
1111
1 1 1 1
N .? .??
1 1 Y
1 1 N
1 I N
I, N
1 1 1 N 11 II 11 ?? ~? ?? N, V
1 1 1 II ? fl II . N_ Nn
I I X I II ^ r. ^ t S 11.' I
t ~..-.... N .. ~. .^.. C. 4l l 1
I I N N u .x>[xxx
1 I N N It .. +'N X X K K
1 1 1 { ^^ >CKr; 11
1 11. ?'. 1 1 1 1 1 1 1
N
Ii
VN U
10
FN aT
??11
C> C)
1 'l
N
N
1
N
Y
N
I
1
1
1
1
I
1 I
1 1
1
^
1
1
I
1
.+
1 1
1
1 1
1^
1
1
1
1
1 1
1
1^
.?.
N N N
. I I K
>S K
11 I' , I
^^^ x K X
X
1 II
?
x 74
>< Y. X K
11 U N
-. >
C K K
11 11 f1
^
.
?+ Y,
x x x
c
N N I
P N N N
P 1 N N N N N
N N 1 Y M N N N N
N N N N U N N U II !;
N. N N N 11 N N N N N 1.
N Y It N N 11 N N N N N
^ N N N N N N 11 N
N N N N 11 N 11 N II N 11 N
M N N Y N N N N N N N Y
N 11- N N M M 11 U N N N
It N N N n 11 N 'N N n
~+ 11 Y 4 N N 4 N N ^ N N
..?. N N N tI 11 N N N N N N
~. N N N it N N N N N P N
11 N N II N N N N N N
?`. .'. n N li N V. U N N N r k
.?. N N N N N 1' N N 11 Y N N
7N N N N N N II N N N N /. ^
' . N N ?1 11 U n N N I1 N N n N
.~ 11 M N N N M P 11 11 11 N y N N
+?. N N N N N N N N 11 N N N^ N
++ 11 N N N Y N 11 II N N 1; N N N N
)11 ? ...., y 11 N N N N N N N N N N
NC ^ r.. 11 ti 11 11 I. 11 N N U II M N
K N -II N II 11 P ti N It 1' N N II P N N 1
K X Y. 11 1 11 91111 N 1; N 11 1; N N N 1 1
11 1< 1( >4N U N 11 I it N N N N 11 N 11 11 1 1 1
r >( >( P N 11 y 11 N I I N N N 'I N Y. , I I
? 1 1 1 1 74 4D 11 Y 11 1 1
.. S.H 11
1 1 N N 11 4; 4+ U`^ 11 YD U' 1 ? 1 1 1 1' K Ni 1
1 1 1
1 1 1 1
I I I ,
I I I
1
1 N
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Gregory K. Ingram, Gary R. Fauth, and Eugene Kroch, TASSIM:
A Transportation and Air Shed Simulation Model, Vols. 1 and TI,
Final report to U.S. Department of. Transportation under contract
DOT-OS-30099, "lay 1974.
2. These procedures are described in Urban Transportation Plannin --
Central Information, U.S. Federal highway Administration,
Washington, D.C., March 1972.
3. A "cold start" occurs when an engine that has cooled to the average
air temperature is started. Cold start emissions differ signifi-
cantly from the emissions of an engine running, at its usual opera-
ting temperature.
4. The air source model is described in F. A. Gifford and S. R. Hanna,
"Urban Air Pollution Modeling," presented at Second International
Clean Air Congress, T'ashington, D.C., Deceriber. 1970.
5. The point source model is described in D. 0. Martin and J. A.
Tikvart, "A General Atmospheric Diffusion Model for Estimating
the Effects of Air Quality of One or More Sources," APCA paper
No. 68-148, June 1968.
6. The diffusion models predict annual average concentrations of pol-
lutants. A "Larsen transformation" is used to relate the predicted
averages to the ambient standards. See ralph J. Larsen, "A Mathema-
tical Model for Relating Air Quality Measurements to Air Ouality
stag vtd I'8eP A$ 2&m /g1P14'.'PC1i~=Rb 60'98AUDU1~~51~71.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
NOTES (continued)
7. The emission reductions described reflect legislation in force
as of January 1974.
8. The TASSIM model's price elasticity for parking is about -.3.
Parking price elasticities typically range from -.3 to -.4.
See Damian Kulash, "Parking Taxes for Congestion Relief: A
Survey of Related Experience," Urban Institute working raper
121.2-1, Nay 1973, processed.
9. These elasticities are similar to those estimated in other
cities. see ierald Kraft, "Free Transit Revisited," Public
Policy, Winter, 1973.
10. Automotive emissions reductions from speed increases may be
only temporary since an improved transportation system may
attract more trips.
11. Recall that the simulations do not predict concentrations of
photochemical oxidants.
12. Determining the costs of regional policies such as changes in
regional auto speeds, increases in automobile occupancy rates,
and alterations in land use patterns is beyond the scope of
this study. For this reason we do not include them fully in
our cost effectiveness analysis.
13. Empirical studies of travel behavior suggest that people value
time at approximately one-third of their wage. See Michael E.
Beesley, "The Value of Time Spent in Travelling: Some New
Appro.Hond lkiaTte /44L19'*!CIA9RbPT9a-00tbtA'dbb-200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
NOTES (concluded)
14. Veekday travel exceeds weekend travel, and the use of 300 rather
than 365 provides crude compensation for this.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 336
THE INTRODUCTION OF MATHEMATICAL-ECONOMIC METHODS
AND COMPUTER TECHNOLOGY IN PLANNING AND MANAGING
SOVIET TRANSPORTATION
R. S. Kozin
The present period is characterized by an increasingly complicated over-
all transport system, rising intensity in the use of transportation plant
and equipment, and closer connections between transportation and other
branches of the economy. Successful functioning and development of the trans-
portation system under these conditions depends to a large degree on the
quality of management.
Planning is the foundation of management. Use of mathematical programming
methods and computers have permitted the formulation and solution of a whole
series of technical economic and planning problems with practical importance.
This circumstance has assisted the appearance of theoretical research, in-
cluding such problems as the formulation of a national economic plan considered
as a type of extremal problem in mathematical programming;. Theoretical re-
search has shown the possibility in principle of constructing an optimal
national economic plan. Under socialist conditions the possibility of,working
out an optimal plan becomes practicable. The following conditions assist this:
Presence of a unified institution influencing the planning activities of
all branches and enterprises:
Noncontradictory goals for the functioning of different branches and
groupings ;
The possibility in principle of collecting and preserving all the neces-
sary information on the status and functioning of the branches of the economy.
Introduction of economic-mathematical methods and computer technology in
planning at the present time is directed toward:
*?)ire _
_
_
~~~I~as~f2~$119~5~'7~10(05-9
3~
a
Techn
.c ences
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
137
Obtaining optimal plan solutions for the basic problems of operating
and developing the transportation system;
Achieving consistency among all decisions taken by plan organizations;
Carrying out by automatized means all the formal functions, most of
which at the present time are carried out by planners (acquisition, compi-
lation, and systematization of information, preparation of plan decisions,
definition of their possible consequences, and so on). Automatization simply
of the formal procedures met within the tasks of statistical and accounting
record-keeping, material-technical supply, several planning tasks, and so
forth, will, to some extent, permit a reduction in the management apparatus.
Computers have already been used to solve this kind of problem in transpor-
tation for about twenty years. However a fundamental effect from using
economic-mathematical methods and computers can be obtained in the productive
sphere only through optimization of processes, more rational use of fixed
and circulating capital assets. Besides, theoretical research and practical
experience show that the effect of introducing economic mathematical methods
and computers will be considerably greater if a system of tasks is solved,
covering as a whole the defined functions of production, and not separate
problems relating to certain partial aspects of these functions.
In the USSR three levels of automatized management of the transportation
system are being worked out. On the first level Gospian USSR is working out
an automatized system for plan calculations in transportation (ASPRT) as one
of the functional groupings for a subsystem in planning the national economy
as a whole.
The basic tasks to be solved in the system ASPRT are:
Definition of the transport factor in the development of this branch of
the economy (direct input coefficients in the interindustry matrix);
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Definition of the final demands for transportation on the part of the
population and the economy and the breakdown of the shipments among the
various carriers:
Distribution of investment according to the direction of its utilization
in order to attain a minimum outlay of national economic expenditure on the
required shipments.
Each of these enumerated problems is in itself a large problem, their
solution in a mutually linked manner will permit the working out of scien-
tifically substantiated plans for the functioning and development of the
transportation system of the country.
Each functional subsystem of ASPT, made up in accordance with general
principles, and this includes the complex subsystem for transportation, is
made up of several blocks. In the transportation subsystem, the following
blocks are being worked out:
The general block
In the general block, problems are solved connected with defining the
total demand for freight and passenger shipments and its subdivision among
the various carriers.
The production block
In this block are worked out the integral indicators for the work of
each means of transportation and, also, the volume of work assigned to each
element for all means of transportation.
The science and technology block
In the science and technology block are worked out the possibilities
for making use of the achievements of general scientific-technical progress
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020 3 5-9
for the technical reequipment of transport; in this block are specified also
the principal scientific problems and ways of financing their study.
The capital construction block
Transportation is one of the relatively capital-intensive branches of
the economy. Transportation plant and equipment make up a significant part
of the economy's fixed assets. Moreover, in the future it is intended that
rather large capital investments will be assigned to the development of the
transportation system. The effectiveness of new capital investment in
transportation depends on how it is allocated. Solution of this problem is
basic for the capital construction block.
The material technical supply block
In this block, the input requirements for rolling stock and equipment
for each carrier are worked out, along with the necessary supplies of fuel
(electric energy) and raw materials.
The block for labor and staff
In the labor and staff block we estimate the labor force required to
operate the transportation system and, also, make up plans for training
workers with the necessary qualifications.
The block for costs, profits, and profitability
In this block we calculate a summary measure of the activities of
transportation as a branch of the system.
It is necessary to underline the fact that solutions for the problems
of each block are linked together. Therefore, in order to obtain consistent
answers, plan calculations cannot be carried through just once.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The plan solutions obtained through ASPRT for transportation by these
methods both depend on, and have an influence on, the solutions that are
reached in the other functional subsystems of ASPR. The mutual consistency
of solutions is assured by the unified structure of subsystems, the general
methodological foundation, and the unified technical means for processing
information. The introduction of A9PR at all levels is connected with an
improved methodological basis for solving plan problems through utilization
of optimizing economic-mathematical models. In each of the blocks of ASPRT
listed above, economic-mathematical models are worked out. Two of them--the
planning of rail freight shipments and the planning of capital investments
to increase rail line capacity will be described at the symposium.
The second level of automated systems for management, OASU (branch
automated systems for management), is put together at the ministry level. At
the present time, managing the work of transportation is handled by three
union ministries (railroad, maritime, and aviation) and by republic organs
of management for river and automobile transportation. Roth in the union
ministries and in republic GOSPLAN bodies, corresponding forms of ASU are
being worked out. The planning block in'OASU is one of the main ones. In
this block the planning assignments worked out at a higher level are made
more detailed. This disaggregation is carried out both with numerical indi-
cators and through specifying "addresses" more precisely. Along with the
tasks of transport planning, in OASU we solve in the first instance the
problems connected with operating management of the shipment process, the
planning of freight traffic, operational management of operating work, pre-
paration of documents for performance standards, material-technical supplies,
and various kinds of records. Each type of transportation OASU has its own
specific Approved For Release 2001/11/19 : CIA-RDP7900798A0002000200 5 9 functioning
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
and management of that kind of transportation. It must be noted that trans-
port ASU differs from ASU for industrial branches of the economy. This is
explained mainly by the fact that transportation is characterized by a more
active role (compared with other branches) for the upper level in managing
the productive processes. Such tasks as operational assignment of the
freight car fleet throughout the railroad network, loadings at technical
stations and alone line sections, since they have first-class significance
for railroad operations, can only be decided at the ministry level. Natters
are the same with other means of transportation.
With the introduction of ASU for managing transportation carriers, the
structural table of organization for management will also change and improve.
Thus, while at the present time the railroads have a four-level system of
management: ministry, railroad, division, line organization, with the intro-
duction of ASU only three levels will remain: ministry-railroad-line enter-
prise. With the changed structure of management, there will be a corres-
ponding redistribution of management functions. A three-level system of
management will come into being on the other means of transportation also
with introduction of ASU.
Within the limits of OASU, automated systems for reserving seats on
passenger trains and aircraft are being created and developed. Special talks
at the symposium will clarify these questions.
A third level of automated systems of management consists of ASU for
technological processes and operations. Operating management of the work
of freight classification yards, seaports, airports, etc., is connected with
solving numerous and reciprocally interrelated problems in ASU of technical
processes--this as a rule involves systems working in real time. Along with
information processing equipment, a very important question for this kind
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
of ASU involves the collection and transmission of information. In con-
nection with this, technology for collection (calculating equipment, pre-
paration of documents, etc.) and for transmission (communication lines,
commutators, etc.) must receive the necessary development. The main goal
for operating planning consists of carrying out plans made up for a more
extended period. In general, the principle for tying together all types
of plans, beginning with operating plans straight through to long-term
plans, consists of having the plans made at a low horizon of planning
"concretize" the plans made at a higher horizon of planning.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9343
OPTIMIZING MODELS FOR PLANNING THE OPERATION
AND DEVELOPMENT OF A TRANSPORT NETWORK
I. T. Koslov
The transport network is construed as a common graph whose elements are
constituted by nodes and sections.
Fach element of the transport network is characterized by a range of
technical and operating parameters which define, with an acceptable degree
of accuracy, its carrying capacity and the cost of moving cargo (passenger)
within the boundaries of the element.
The performance of the transport network is determined by the load
(quantity of cargo tonnage or number of passengers) on its component elements.
Such a concept of the transport network is at variance with the notion of the
transportation requirements of the national economy and the population. De-
termination of the transportation requirements of the national economy and
the population is an independent problem in its own right, which lies beyond
the scope of the present paper. But if one is to analyze the degree of utili-
zation of the transport network carrying capacity, the loading of its principal
elements comes in as a very handy tool. Thus, in the problem under consider-
ation, planning of the transport network operation boils down to determining
the loads on its component elements.
To outline the pattern of development of the transport network is
to determine the kinds of reconstructive measures envisaged and the points in
time when the network elements acquire additional capability.
Problem definition
The problem of determining the volume of operation and the ways of
development of the transport network can be solved on the following premises:
*Divisio~~.,,~~g ~n i ute of Complex Transport Problems and
Doctor "art Tec nico~ g~9Q01/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. The transportation requirements of the national economy that the
transport network is to meat are expressed as follows:
The pattern of relationships of the various kinds of cargo between
the network nodes
The number of passenger trains running over each network section
The volume of the local cargo flows over the network sections
Each form of transportation requirements is liable to very in time. In
fact, the transportation requirement for the future is an indefinite quantity
which only lends itself to probabilistic treatment. However, for the purposes
of the present paper, it is assumed that this factor is a determinate quantity.
Thus, the problem solution cannot but he treated as tentative. One can assume
that the indeterminacy cannot detract from the usefulness of the problem to
be considered.
2. The and goal of the transport system consists in meeting transpor-
tation requirements at the least cost. The transportation costs
depend on numerous factors, some of which are relevant to the problem
under discussion:
The choice of cargo routes within the network
The technical state of the main network elements
The transportation pattern
3. Cargo may be carried from one nods to another by quite a few different
routes. Of course, the choice of the route for each individual batch
of cargo never conduces to an optimal solution of the overall problem
for the simple reason that the total transportation costs by and
large depend on the loading of the elements encompassed by a given
route. Furthermore, transport network elements with a limited
carrying capacity often have to be excluded from the haulage routes
for many cargo varieties.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 345
The starting data for a model designed to determine the transportation
costs consist of the loading of the network elements (nodes and, sections).
For this reason, it is the loading of the network elements, and not the cargo
routes, that will form the unknown variables. Obviously, the set of the
chosen routes uniquely defines the loading of the network elements.
4. The network elements are technically developed through reconstructive
undertakings of many kinds. It is possible to limit the number of
such undertakings, but only on the basis of special research. A
research program of this type has been conducted, e.g., for railway
network sections. The results of the study suggest that, of all the
possible measures, only the following are practically significant:
Extension of departure and reception station tracks as conducive to
higher weight ratings of freight trains
Substitution of automatic block signalling for the electric staff
system or semiautomatic block signalling
Construction of additional main tracks on open lines
Substitution of electric locomotives for diesel locomotives
5. The type of traffic pattern significantly affects the economics and
the degree of utilization of the available carrying capacity. In its
turn, the traffic pattern by and large depends on the technological
state of the network elements. However, some variables also exist;
such as, for instance, the type of locomotives used. Since the
locomotive stock is generally limited, it is impossible to use the
most effective units over each section of the network. Thus, a
problem arises as to the most advantageous distribution of locomotives
among the elements of the transport network, which, naturally, cannot
be solved unless the state of technology and the pattern of traffic
are taken into account.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Mathematical definition of the problem
Let's introduce the main designations which will be used in the mathe-
matical treatment of the problem.
Designations
s - number of the transport network sections (s - 1, 2, ..., S)
i - number of the transport network nodes (i s 1, 2, ..., I)
t - time (te/to, Tp/)
p - type of cargo (p - l, 2, ...,P)
R+(i), R 7(i) - totality of network sections going into and out of the node i
xsp(t) - flow of cargo of type p over the section s
yi(t) - operation of the node I
-~ s(t) - vector function determining the technological state of the section s
i(t) - vector function determining the technological state of the node 1
c(t) - traffic pattern
A1(t) - pattern of relationships reflecting the needs of the
national economy for the transportation of cargo of type p
Ns(t) - volume of passenger traffic over the section s
Fs(t) - flow of local cargoes over the section s
K(t) - limit of investments into reconstruction
Nn(t) - required carrying capacity of the transport network
NH(t) - actual carrying capacity of the transport network
(t) - passenger and freight transportation costs
A(t) - reconstruction costs
(t) - weight function to account for the time value of expenditures.
The unknown variables in the problem considered are as follows:
section loads x sp - X
Approves For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
section development s . M
node development H
traffic pattern c
The X and Y variables are determined by the choice of cargo routes.
The section loads will be determined with a breakdown by cargo types condu-
cive to more accurate technical and operating parameters of the operation.
In a general case, the node load should be broken down to differentiate
between the components differing in the level of reclassification. Thus,
for instance, for a railway node, not only the overall car flow must be de-
termined, but also the transit car flow requiring no reclassification, the
transit car flow which does require reclassification, and the local car
flow. Such an elaborate characteristic of node loading can no longer be de-
termined merely by cargo routes, for to characterize the transit flow one
should break it down into the reclassified and nonreclassified flows which
calls for the knowledge of the traffic pattern.
In the problem under consideration, the section and node loads are non-
negative functions of time.
The technical state of the sections and nodes may be characterized,
with a sufficient degree of accuracy, by means of several parameters. As the
latter vary with time, they indicate how the sections and nodes develop. The
functions which define the development of the elements with time are charac-
teristically discrete. Most elements have a limited number of states. Thus,
for instance, the technical state of a railway section is characterized by
hardly more than ten parameters, each of which can assume two or three values.
Such a specific nature of the/I s(-f)uVi(t) function may be used to construct
computational algorithms.
The traffic pattern varies widely depending on the particular kind of
transport. Thus, the rail traffic system is by and large determined by the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
distribution of marshalling operations among the stations and by the loco-
motive service.
The unknown variables are to satisfy three groups of constraints.
Group 1 includes constraints which reflect the compulsory nature of all
transportation demands. Mathematically, these constraints may be represented
as follows:
I,+ Xap(t) SEA-iXsp (t) Aijp(t) - 1 Ajip (t) (1)
p
Group 2 is composed of constraints which stipulate that the needed carrying
capacity of the network should not exceed the actual value:
Nn(X,Y,N,F,t) < Nn(M,H,c,t) (2)
Group 3 comprises constraints whereby the needed reconstruction invest-
ments should not exceed the available level:
A(M,H,c,t) < K(t) (3)
The problem solution is evaluated on the basis of minimal reduced economic
expenditures:
jpC (XY,M,H,c,t) + A(M,H,c,t)j (t)dt - min. (4)
t
?Equations (1) to (4), together with the description of the function
classes to which the unknown variables must belong, constitute the enunciation
of the problem of traffic distribution and network development.
Problem analysis
The mathematicaldefinition of the problem suggests that all the variables
being optimized are functions of time. Assuming that time varies discretely,
the unknown functions can be determined by finding the unknown values of the
scalar variables.
By its type, the problem in question falls in the class of mathematical
programmi1%#PjF r. Recta ,s ,1di &flMIe ~A-f~? 0@R88 ~D5n$rrawly
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
with a view to using the well-developed techniques. Indeed, it is only
constraints (1), which represent constraints of a linear programming problem
in a network form, that have a straightforward structure, whereas the proper-
ties of the space assigned by constraints (2) do not lend themselves to
clear-cut analysis. The left-hand and right-hand portions of equation (2)
are calculated by special and fairly complex algorithms, which is also true
for the left-hand portion of constraint (3).
All three groups of constraints (1) to,(3) should be satisfied at each
point in time independently. This requirement is used to construct a compu-
tational algorithm for solving the problem.
The properties of the functional to be minimized likewise elude analyti-
cal analysis.
However, one can assert that the functional to be minimized cannot be
construed as a sum each term of which depends on just one moment of time.
It is hence impossible to represent the general problem of traffic distri-
bution and network development as a series of independent problems for a
series of time sections.
Thus, the problem under discussion belongs to the category of mathe-
matical programming problems of a general kind.
The actual Soviet transport networks have several thousand nodes and
sections. Since the network carries several score aggregated types of
cargo and the time base runs through at least three five-year periods,
hundreds of thousands of variables will have to he determined if the problem
is to be solved. Undoubtedly, an extremely complex computational task.
Problem solution
To obtain a general solution, the problem defined above will be broken
down into the following parts:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Distribution of the required traffic about the network (R)
Determination of the required reconstructive measures for the
network elements (V)
The solution is arrived at iteratively as follows:
R1+V1+S1+R2+ ... --Sk.
While solving the first part of the problem, both the technological level
of all network elements and the traffic pattern are assumed to be known.
As the algorithm R is being realized, equalities (1) are satisfied for
all time instants of the period in question.
The second part of the problem is_concerned with the measures aimed at
raising the technological level of the network elements. The traffic distri-
bution and the traffic pattern are considered to be known, and the algorithm
must be so constructed as to provide for satisfying inequalities (2) and (3)
given in the mathematical definition of the problem.
The third part of the problem deals with the traffic pattern, the traffic
distribution, and the technological level of the network elements being con-
sidered as invariable.
No proof has been offered to the effect that iteration conduces to an
optimal solution. Moreover, the large dimension of the problem does not allow
for too-many iterations. It follows that an approximate solution is the only
one that can be obtained by the proposed method.
The traffic problem algorithm can be employed as the algorithm R for
determining the traffic distribution about the network. In fact, constraints
(1) are exact analogs of the traffic problem constraints. The national
economic expenditures on traffic, which are part of the general criterion (4),
may be used as the yardstick for determining if the solution is indeed an
optimal one. If the methods used to solve the traffic problem are to be
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
applicable here, the traffic expenditures must be represented as a sum of
expenditures related to the network elements, and this kind of represen-
tation involves a certain error.
Numerous authors have demonstrated that the hypothesis which stipulates
that expenditures grow linearly with the loading of the network element, is
too broad, which calls for the use of more complex functional relationships
between expenditures and loading. Analysis indicates that the real functions
of expenditures all show an important property; viz., the curves are convex
downward.
This made it possible to work out a special algorithm for determining
the network distribution of traffic with nonlinear characteristics. It is
a final algorithm conducive to a truly optimal solution.
The traffic distribution problem as stated must be solved many times for
fixed moments of time within the period under consideration, so that the
network element loads are actually obtained as functions of time.
The optimal development of the technology of the network elements is
determined on the basis of the element load values established at the previous
stage. Mathematically, this part of the problem could be reduced to the
task of finding a minimum of the function of many variables which represent
the sum total of all expenditures over the period under consideration. Such
representation of the problem is made possible by the specific functions of
development of the elements as indicated in the mathematical definition of
the problem. When determining the optimal pattern of development, inequality
(2) is taken into account for each element separately, whereas constraint (3)
must be satisfied for all elements lumped together. To satisy this latter
constraint, a heuristic algorithm is suggested which sets the order in which
reconstructive measures will be taken, it being understood that the order
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
thus set :takes effect only if limited investments are available for the develop-
ment of the carrying capacity.
The optimal traffic pattern is determined by the results of the two pre-
vious solutions. Central to any traffic Rystem (at any rate for the rail-
way network) is the deployment of locomotives. Analysis indicates that this
problem may be solved, given some assumptions, by linear programming techni-
ques.
Special methods exist for scheduling and distributing marshalling jobs,
but in this case approximate methods based on statistical data seem to
suffice.
It is thus clear that the general problem of traffic distribution and
technological development is actually a set of interrelated, highly complex,
and computationally labor-consuming problems. Hence, the problem can only be
solved through the use of efficient computer centers.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
METHODS OF FIVE-YEAR PLANNING OF TRANSPORT-ECONOMIC
CONNECTIONS: THEORETICAL DEVELOPMENTS AND
EXPERIENCE WITH PRACTICAL APPLICATIONS
N. I. Mokrousova and Z. I. Mozgrina *
The volume of future traffic is crucial for many aspects of five-year
planning. For this reason one of the first problems to be solved is one of
planning the transport-economic connections, which furnishes the initial
block of data on the future volumes and directions of cargo traffic.
From the economic point of view, the problem in question boils down to
meeting the country's freight traffic demand at minimal cost within a preset
production framework. In order to accomplish this task, optimization models
are employed to plan the traffic volumes of the future.
Analysis indicates that for the purpose of long-term planning of trans-
port-economic connections, all kinds of cargo may be classified in three
broad groups :
1. Cargoes which are planned on the basis of territorial production-con-
sumption balances, with the transport-economic connections of the
suppliers and the consumers liable to change
2. Cargoes which are likewise planned on the basis of territorial pro-
duction-consumption balances, but with the transport-economic con-
nections remaining invariable
3. Cargoes which are planned without recourse to territorial production-
consumption balances
The first-group cargoes may be broken down into two subgroups:
Cargoes assumed to be homogeneous (most types of timber, cereal,
and petroleum products), which are singled out from the rough lists
of bulk cargoes by virtue of their homogeneous processability as
Senior Research Staff, Institute of Complex Transport Problems
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
well as consumer and other properties. For these cargoes, the
transport-economic connections are determined by optimizing sup-
plies by means of linear programming traffic problem for preset
points of cargo departure and destination and for predetermined
cargo volumes. The volumes of departing cargo are determined by
balance calculations for specific products using the oblast (roughly
equivalent to region) as a base territorial unit. For the purpose
of this paper, this kind of model will be referred to as a one-
product model;
Nonhomogeneous, but interchangeable cargoes, such as, for instance,
fuel coal grades varying in calorific value or chemical fertilizers
containing different levels of nutrients. These cargoes are like-
wise covered by oblastwise and stationwise departure and arrival
schedules. The optimization aroblems involved in the planning of
supplies of interchangeable
problem method (A-problem).
The second group includes mining
products are solved by the allocation
This model will be called a A-problem.
and metallurgical products, construction
materials, petroleum, coking coal, grinding products, and many other kinds of
like cargo. Given certain established transport links, their quantitative
characteristics for the long term are determined by traditional methods, and
the traffic plan is evolved using an optimization model whereby cargoes
(relationships) are superimposed on the transport network following the tree of
the shortest (cheapest) routes. The model used for this category of cargoes
will be termed a rigid-link model.
The third group includes several hundred agricultural products, foodstuffs,
consumer goods, some chemicals, products of the woodworking and paper-and-pulp
industries, metal items, machines, equipment, transport vehicles, industrial
materials, 9150r8 cFtb' AIIJIUMPT"bi/' 9"11TP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 355
The items coming under this heading are universally produced and con-
sumed, so that it is impossible at present to compile territorial production-
consumption balances for these lines of goods. Hence, special methods have
to be devised for their planning.
Let us first dwell in some detail on the planning models of transport-
economic links for the first two groups. The one-problem model, which is
built around a linear programming traffic problem in network terms, is defined
as follows.
Given is a transport network containing n vertices and N arcs, the arcs
being designated as S - 1, 2, ..., N; the vertices is - 1, 2, ..., n and
= 1, 2, ..., n (is abd h9 are respectively the outgoing and the incoming
s
vertices of the arc S). The cost of transporting a unit of cargo along the
arc Cs is known.
Given are points of production and consumption of a homogeneous cargo.
The production volume ak is more than zero, the consumption volume ak is less
than zero, and ak is equal to zero at the rest of the network points, or
vertices, where the cargo is neither produced nor consumed (k - 1, 2, ..., n).
It is required to devise a plan of assigning suppliers to consumers and
a corresponding traffic plan (xs, S - 1, 2, ..., N) such that would involve a
minimum of transportation costs
Cs X s -->41""L'
~c l
while satisfying the following conditions-
production-consumption balance
E ak ? 0, k - 1, 2, ..., n; (1)
flow equilibrium at the network vertices
x x - ak, k - 1, 2, ...9 n;
a is k a
(2)
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
nonnegativity of traffic
x8 >,, 0, S - 1, 2, ...9 N. (3)'
The information input for this model is constituted by an aggregated
conventional transport network. At this stage, the railway network contains
550 nodes (vertices) and 850 segments (sections). Aggregation and substanti-
ation of a conventional network are dealt with in numerous special investi-
gations.
Reduced transportation costs and distances in kilometers were used as
estimates of the network sections. The reduced transportation costs include
the costs of operating trains, servicing them en route and maintaining in-
stallations, as well as the investments into the rolling stock and techno-
logical development. The costs and investments are incorporated in the part
of the model which represents the traffic volume. The sectionwise indicators
were differentiated for bulk cargo and oil products. Bulk cargo was defined
on the assumption that all network sections were equally loaded in all
directions and that the rolling stock was utilized to an average degree (80%
utilization of gondola capacity and 90% utilization of boxcar capacity). For
oil products, the assumptions were that the cargo was transported over all
network sections and the tank cars were utilized to an average degree (80%
utilization of tank car capacity).
While planning specific cargoes, the differences in the utilization of
the rolling stock capacity are allowed for through the use of correction
coefficients.
The estimate of each arc must incorporate the correction coefficient.
In the course of computerized calculations, it was found advisable to use
the main starting network variants for bulk cargoes and oil products, with
the corrections being incorporated in the functional.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Experience with experimental and plan problems indicates that the network
form of the traffic problem is the most convenient approach for planning
practice. Its advantage over the matrix format consists in that it permits
obtaining all elements of the traffic plan directly on the transport network
(traffic over the network sections, relationships, and routes), evaluating the
plan in terms of several criteria, and dispensing with the step of preliminary
computation of the shortest and cheapest routes between specific suppliers and
consumers, since problems for various kinds of cargo with any set of suppliers
and consumers can be solved directly on the network.
The network problem makes use of two kinds of initial information, viz.
constant and variable. The former includes data on the transport network
which are prepared in advance and stored in the external memory of the computer.
These data are used throughout the entire computation cycle, being refined and
corrected from one cycle to another.
The variable information, incomparably smaller in volume, includes data
on the cargo itself; i.e., on the arrivals and departures of the cargo at the
vertices of the transport network.
For the one-product problem, there is a set of programs which, along with
the transport problem program, also includes a program for planning the assign-
ment'of suppliers to specific consumers, the routes of cargo movement being
indicated for each supplier-consumer relationship. The set further contains a
program for determining the volume of work done in ton-kilonleter8. The programs
provide for producing intermediate results (punched-card arrays) for use in
crossfoot models.
The next model is one of planning of nonhomogeneous interchangeable cargo
traffic. Here it is possible to use a linear programming allocation problem
(A-problem) which is mathematically defined as follows:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Given are the points of production i and consumption j
(i - 1,2, ..., n; J - 1, 2, ..., m), as well as the volumes of resources
ai and consumption b1 of interchangeable products. Also known are coeffi-
cients ail which relate the value of production ai to that of consumption b1.
It is required to provide a product supply plan conducive to minimized
transportation costs
xi1 cil -- min
< ai, i - 1, 2, ..., n;
xi1 ail - b1, l - 1, 2, ..., m;
xi1 > 0, A i1 > 0,
where xilrepresents the supply of the products manufactured by the i-th
supplier to the 1-th consumer; and cif represents the cost of transporting a
unit of products from the i-th supplier to the 1-th consumer.
The interchangeable cargoes considered in planning practice allow using
a somewhat more complicated traffic problem rather than special allocation
problem algorithms. This is made possible by the fact that the coefficient
A for one of several suppliers has a constant value. In such a case, the
coefficient XI is constant in each line of the matrix, if the traffic problem
is solved by the matrix method. Thus, it is quite sufficient to divide the
coefficients of the matrix cif by Al and solve a common traffic problem in
conventional product units. For obtaining network results, the relationships
thus found in natural units must be superimposed on the transport network.
For the network case, this problem can be solved in two stages. At the
first stage, the transport problem will be considered for a multilayer network
for the entire volume of products in conventional units, whereas the second
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 359
stage will deal with ordinary one-product problems for each individual product
in natural units, the one-product problems being defined by the results of
the general optimal solution. The network problem proved amenable to solution
because a small number of heterogeneous groups of interchangeable products was
considered: three kinds of fuel coal and two varieties of nitrogenous and
phosphoric fertilizers. The number of identical layers for a multilaver network
corresponds to the number of heterogeneous groups of cargo being considered.
The estimates of the identical laverwise sections are set to he inversely pro-
portional to the coefficient lei. A multilaver network with sequentially
numbered layers can be produced automatically by means of a special program.
The starting economic data for heterogeneous interchangeable cargoes are con-
stituted by the departing and arriving cargo volumes in conventional and
natural units at the vertices of the transport network. The set of programs
of this model includes all programs of the one-product model and an additional
accessory program of transformation of the starting network into a multilaver
one by the preset values of ai.
The rigid-link model essentially contains a problem of distribution of
predetermined cargo loads about the transport network. It is defined as follows:
Given is a transport network containing n vertices and N arcs, the arcs
being-designated by S (S - 1, 2, ..., N) and the vertices by i and j
(1, j - 1, 2, n).
The transportation costs over each are Cs are known. The traffic pattern
is formed as a network of internodal relationships a,1j.
It is required to construct a traffic plan for the transport network
(x , S - 1, 2, ..., N) providing for minimized transportation costs:
s
Cx -I. min.
s s
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
This problem is built around a minimum path-length algorithm.
The cargoes planned with the use of this model are characterized by the
need to retain, fully or partially, the transport-economic links evolved over
a long time, while permitting certain changes in the distribution of production
units and in the territorial production-consumption balances for the long
term. Such conditions imposed on some links stem from technological require-
ments, specialization of some production units in terms of special raw materials
or mixtures thereof, and from some other factors of a similar nature. While
readily incorporated in traditional plans, they sometimes prove unmanageable for
economic mathematical models. To ensure automatic planning of transportation
of this group of cargoes, the simplest possible solution was found; viz., to
combine traditional methods of computation with automatic operations. Under
this system, the interoblast and then internodal relationships accounting for
rigid transport-economic links are handled manually by traditional methods,
but the most labor-consuming computation tasks involved in the superposition
of these relationships on the transport network and determination of the
principal indicators of the plan are done automatically.
The set of programs in this model includes programs for planning the
traffic along the preset pattern of relationships between the nodes, as well
as calculating the work in ton-kilometers.
Summation programs were developed to produce the final results of the
traffic operations, whereby resulting volumes can be obtained for any set of
cargoes. Summation is carried out on the basis of the intermediate infor-
mation furnished by the optimization models. The set of summation programs
provides information on the overall traffic density over the transport network
sections, the general and structural tables of internodal relationships, and
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
361
similar tables for interoblast, interrayon (rayon: a territory far larger
than an oblast) and interrepublican cargo exchanges.
Experience with experimental and plan calculations suggests that the
long-term traffic volumes are generally understated, both in terms of integral
indicators and for the sectionwise flow density figures. The understatement
was tentatively estimated at 1 to 10% for various kinds of cargo. The reasons
for such understatement should he sought in the following factors:
Overaggregated long-term economic information worked out for an
enlarged cargo nomenclature and large economic regions
An aggregated transport network including 550 stations instead of
the 8,000 actual stations and covering a smaller area. than in reality
(113,000 km instead of the actual 135,000 km of overall route length
due to the skipping of all dead-end sidings, underloaded sections
and intranodal tracks)
The optimization models themselves which, under conditions of the
above-described aggregation, provide ideal optimal plans free from
all kinds, even rational, of cross-hauls of actually nonhomogeneous
cargoes (in the one-product model)
In order to allow for the possible understatement of the integral and
sectional traffic plan indicators, use can he made of corrections whose magni-
tude can be determined by comparing the actual data with the calculations and
introducing an allowance for the long term.
All the above-described optimization models for determining transport-
economic links are being currently introduced into the medium-term traffic
planning practice.
The transport-economic links of the third group are planned on the basis
of available statistics. The problem may be defined in the following manner.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Given are an interoblast and interrayon tables of relationships for a
certain base year derived from reports, as well as an interrayon table of
relationships and total freight turnover figure for a given group of cargoes
for the period being now considered, the latter two parameters being found
by a certain method. It is required to determine an interoblast table of
relationships amounts to the corresponding interravon relationships, and the
freight turnover for a given group of cargoes is equal to the assumed given
value r.
Mathematically, the problem is defined as follows:
Given: a matrix A each element of which (aij) characterizes cargo
transportation from the i-th to the 1-th oblast in the base year
(i - 1, 2, ..., n; 1 - 1, 2, ..., n).
a matrix L whose elements lii characterize the distances between the
oblasts;
a matrix C each element of which (Crs) represents the traffic from the
r-th to the s-th rayon in the base year, and
d -
rs i 5
where Jr is the set of all is which belong to the S-th rayon
(r - 1, 2, ..., Z; S - 1, 2, ..., Z) ;
a matrix D each element of which (drs) represents the traffic from the
r-th to the S-th rayon in the year being planned.
A matrix b is to be found such that
max
i,j
bi
drs
ij Crs
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
on condition that
bit aij , Vii
Jd-7 s
- drs, V, S
191 big lij - r,
where bijis the traffic from the i-th to the j-th oblast in the year being
planned.
The above-defined problem is solved in the following way.
First of all, it is necessary to determine coefficients characterizing
the increase in the volume of relationships between the rayons in the year
being planned as against the base year. The coefficients thus obtained are
used to determine a proportional table of interoblast relationships for the
planned period, simultaneously checking whether or not the condition that a
predetermined freight turnover is satisfied. If this condition is satisfied,
the problem is all but solved. If not, then, maintaining the same general
picture of departing and arriving cargo volumes, the relationships between
the oblasts must be so changed as to arrive at the predetermined turover
figure, taking pains to keep deviation from the proportional pattern of
relationships within a minimum, because of the specific nature of this group
of cargoes. With a constant departing volume, the turnover can be changed
by varying the average length of haul. To this end, the relationships must
be changed in the following way.
An additional table of relationships is found which actually amounts to
the difference between the proportion and base tables. As the significance
of relationships within each economic rayon diminishes, they are adjusted
and finally broken down into pairs. The possible variation of the ton-kilo-
meter index is assessed for each pair of adding the longest or the shortest
distance depending on whether the turnover as a whole increases or diminishes.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
364
The variation in the ton-kilometer index is calculated for each pair by multi-
plying the cargo quantity being transported by the distance differential.
Integrating these variations for all pairs of cargo relationships for all
economic rayons, we will determine the maximum value by which the turnover
within a given table of relationships may be changed (increased or decreased).
In order to achieve the predetermined turnover value, it remains to de-
termine the share of the possible maximum variation. This value is equal to
the proportion of turnover deviation of the maximum value by which the turn-
over within a given table of relationships can be varied. The additional and
then the starting tables of relationships may he varied in proportion to this
share.
The cargo traffic density over the sections of the network under con-
sideration cannot be determined merely on the basis of interoblast links.
Investigation of the interoblast exchange table of relationships for the group
of "other" cargoes for the year 1970 indicates that the intraoblast traffic
accounts for thirty seven percent of all shipped goods, but in terms of turn-
over this kind of traffic is insignificant due to the short distances involved.
If the traffic density over the network sections is to be determined, a
more detailed table of relationships is required, for, alongside oblast cities
there exist other big consumers (or suppliers) of "other" cargoes.
Analysis of the points of production and consumption of the group of
cargoes under consideration allowed for determining several aggregated nodes
for each oblast, finding their relative shares of the total turnover and,
thereby, breaking up the relationships of the interoblast table into smaller
units.
Knowing the production and consumption points and using the minimum
path-length algorithm, the table of relationships can be superimposed on the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
network being considered to find the traffic density for the "other" cargoes
by the sections of the network.
The above-mentioned algorithm was used to perform experimental calcu-
lations on the basis of the 1470 railway reports. Analysis suggests that
the calculations are entirely satisfactory for practical purposes.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
PROBLEMS OF OPTIMAL PLANNING AND MANAGEMENT OF
AUTOBUS TRANSPORT IN THE GEORGIAN SSR
G . G . Tsomaja
1. Goal Description
The present level of motor transport technology, the ever' increasing
traffic volumes and the increasingly sophisticated technology and patterns
of the transportation process, all emphasize the urgent need to improve the
methods of planning and controlling the operation of the transport systems.
The motor transport of the Georgian Republic is a highly complex system
which will be progressively more and more difficult to plan and control un-
less modern control methods and technology are used. To meet this challenge,
the "Avtotranstekhnika" /Automotive Transport Technology/ Agency of the
Ministry of Motor Transport of the Georgian SSR has enlisted the cooperation
of the Laboratory for Problems in Automatic Control and Computer Technology
of the V. I. Lenin Georgian Polytechnic Institute to develop an automatic
system for planning and controlling the operation of the passenger motor
transport service. The program has set the following objectives:
Developing an effective network of routes
Selecting optimal types of the rolling stock
Compiling bus schedules
Approportioning routes and buses to motor pools
Planning the technical and economic parameters of the bus
Service for all levels of the control hierarchy
Controlling traffic on a day-to-day basis.
2. Object Description
The Georgian Republic can boast a highly developed bus service: the
annual volume of passenger traffic amounts to 511 million passengers and
* f ~A~ F`1R3'T ~ D f 98 ~~ Pc~bO6 O%d . Dept.
of'Motor Transportation
Approved For Release 2001/11/19 : CIA-RDP79-00798AO00200020005-9
4,533 million passenger-kilometers; the republic operates 565 interurban,
1,001 suburban, and 257 urban bus lines, the length of all routes adding
up to 108,700 km.
In view of the many random factors affecting the transportation demand,
the transport process should be regarded as a probabilistic one. It has been
found nevertheless that population movement, for all its probabilistic nature,
follows certain laws which shape passenger flows in space and time--and fairly
strictly at that. Hence, a possibility arises to work out a network of routes
with a regular bus service on the basis of a fixed schedule, thereby converting
the probabilistic system into the next best thing to a deterministic one. On
the one hand, this approach brings a measure of orderliness into transportation
demand as the passengers are enabled to plan in advance the routes and times
of travel; on the other hand, however, the passengers' freedom of choice is
somewhat curbed. Reconciling these two contradictions is the problem con-
fronting any effort to plan the operation of a passenger transport service as
a technical system.
Several problems also arise if the passenger motor transport service is
viewed as an economic system. Thus, transportation demand can only be met
on the basis of some resources, but these must be administered in the most
effective way. Since the available resources are never unlimited and, de-
pending on the particular conditions of use, vary widely in terms of effective-
ness of utilization and economics, one finds oneself on the horns of a typical
dilemma: either to try to find the most effective use for the available
resources, or to plan resources to meet a prescribed level of effectiveness.
Furthermore, serious problems must be dealt with while considering the
passenger transport service as a social system. At the present level of
civilization, movement turns into one of the most important problems which
Approved For Release 2001/11/19 : CIA-RDP79-00798AO00200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 368
may presage the success or failure of any effort to develop productive forces,
raise the cultural level and living standards, and meet other social needs
of society. Viewed in this light, the passenger transport system is called
upon to satisfy transportation requirements on the basis of social needs.
Such an approach, however, is fraught with conflict between the demands the
transport system must meet as a social system, on the one hand, and as a
technico-economic entity, on the other.
Thus, one can conclude that the object under consideration is a large,
complex, dynamic and probabilistic system whose planning and control present
considerable difficulties.
3. Requirements for an Automatic Stem for Planning and_
Controlling the Bus Service
The problems one encounters while working out and adopting decisions to
set up, develop, and operate a passenger transport system, are by and large
of a poorly structured type. For this reason, no control strategy can rely
on a single utility function which would describe the system in all its
dimensions: technical, economic, and social. Furthermore, social objectives
as often as not defy attempts at quantitative analysis and can only be repre-
sented qualitatively. Most technical and economic problems do lend themselves
to quantitative description, though some of them cannot he formalized. With
this in mind, the requirements to an automatic control system may be stated
as follows:
The system must be a man-machine combination, with the man being
responsible for the final decision adopted on the basis of a high-order
set of objectives which can neither he formalized nor described in
quantitative terms
It must conduce to compatible technico-economic indicators for each
ApptWM Ff IteMar6UT/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 369
It must be adaptable to various modifications of invariable and con-
ditionally invariable input information
It must be able to cope with all problems of technico-economic planning
and day-to-day control both continuously and discretely, providing
final and intermediate results
It must provide all data needed for the purposes of day-to-day control
by all hierarchical levels of the ministry's control system, with the
information verified as to reliability, completeness, and timeliness,
and taking into account the terms of reference of the personnel partici-
patine, In the planning and controlling of the passenger transport
operation;
It must be able to cope with new problems related to the transport
system without requiring substantial modification
The above leads to the conclusion that the man-machine system for
planning and controlling the operation of the passenger motor transport net-
work must be able to solve problems of two classes: those related to an
improved traffic pattern; and those related to the automatic processing of
technico-economic and control data.
4. Solution Experience
As an example of the way to solve first-class problems, let us discuss
how the interurban bus service of the republic is planned and organized. The
traffic plan is based on the population transport demand. Using statistical
methods, one can process the data of passenger flow surveys or line documents
in order to form a three-dimensional matrix of passenger relationships which
would reflect the pattern of population movement in space and time.
A survey of the republican interurban bus service indicates that the
passenger flows fluctuate widely both during the day and from one day of the
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
week to another. Seasonal fluctuations have also been noted. (The survey,
conducted in 1970-71, used special questionnaires for each running bus in
the interurban bus network filled in every day during one demonstration
week every quarter.) After statistical reduction (with a ten-percent
deviation from the meanvalues), the data were used to construct a three-
dimensional matrix for week days of the same type (working days and rest
days) and for different seasons (summer and fall-winter periods). Thus,
four combinations of a three-dimensional matrix of passenger flows were
obtained, which formed one part of the planning data base.
Another component of this data base is represented by the road network.
The peculiar geography of the Georgian Republic shapes its road network which
carries interurban buses: almost all towns and population centers are con-
nected by solitary highways and no dense road network is in evidence. So,
in actuality, the passenger who wants to travel from one town to another
in most cases is left without options. The structure of the highways con-
forms to a single pattern: four routes of the same configuration sprouting
from the republic's capital city. Analysis of the passenger flows suggests
that the movement among the population centers lying on different routes
is negligibly small and can be ignored to the first approximation. For
this reason, a model of the transportation process was constructed on
the basis of one route or direction (western), which can be represented
as a nondirected graph whose nodes are constituted by the towns and the
arcs by the roads interconnecting these towns.
The passenger flows on interurban routes may be carried by rolling
stock of various kinds. So, accordingly, the third component of the data base
comprises rolling stock characteristics, such as passenger capacity, class
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
of comfort, and speed of travel over roads of different categories.
The data base is used to plan a regular bus service. The planning pro-
cedure consists of three steps, viz. (a) development of a network of routesi
(b) assignment of the type of rolling stock; and (c) scheduling of the
service. Step (a) implies listing all routes and all buses on each route as
well as the time each bus starts to run. Each route is defined by the list
of network nodes that the bus must pass in a prescribed sequence in both
directions. Assignment of the type of rolling stock implies selecting
vehicles with the necessary characteristics corresponding to the principles
of the service and permitting to meet them either in the most effective way
or at minimum cost. And, finally, scheduling is a procedure of setting the
time when the buses must pass all nodes along the route, with the names of
the stops and the duration of each stop being spelled out.
The above-described components of the planning process are closely inter-
related and thus necessitate simultaneous consideration.
Attempts at presenting all possible alternatives of bus service plans
as a system of inequalities and equations generally involve large-dimension
problems which often prove too difficult to solve even for computers. So in
practice, optimal solutions remain an impossible dream and one has to be
content with just rational solutions, at least in some cases.
Therefore, the problem has been solved by heuristic programming tech-
niques. A generalized model of the bus service operation has been constructed
on the basis of the following assumptions and constraints which reflect the
chief parameters of the passenger movement process:
The policy of the passenger service from the viewpoint of a set schedule
does not affect the distribution of the paired relationships in space
and tine; i.e., the passengers may be classed only as those catered to
Ai ep )S%4 F#-tRM 3Opa/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The possibility that one and the same vehicle may run on several routes
is disregarded
While assigning a bus, the first to be considered are short routes, and
then long ones. This sequence ensures a maximum number of through
routes and a minimum of transit passengers
Through passengers enjoy priority when buses are assigned to routes,
and only then the sections between the various route nodes which are
not covered by the shorter routes, come in for consideration
Buses are assigned to routes depending on their capacity, larger-capacity
vehicles enjoying priority
A bus run is considered to be acceptable if the dynamic coefficient of
bus capacity utilization lies within admissible limits, the lower limit
being chosen on the grounds of efficiency while the upper on the grounds
of reliability
Permissible limits are set to the time of bus stay at the intermediate
stops of the route as well as to the bus driver's working hours
For some route nodes, the time of arrival of the bus at the destination
is set at the latest possible hour
The problem can be solved if the output of the system includes the list
of routes and the numbers of all buses running on a given route, the time
each bus starts its trip, the type of rolling stock, and the values of the
dynamic coefficient of capacity utilization. Simultaneously, information is
furnished on the bus schedule.
If the above-listed assumptions and constraints are taken into account,
the choice of the route network, the assignment of the buses to their respec-
tive routes and the selection of the suitable type of rolling stock hermit
rationalizing the network and optimalizing the rolling stock. For practical
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
purposes, however, higher-order goals and a number of other constraints intro-
duce certain changes at the expense of system properties.
The above considerations formed the basis of an algorithm which, upon
changes in the solution of the problem, forms a new network of routes, selects
a suitable type of rolling stock adapted to the new conditions, and calculates
the dynamic coefficients of bus capacity utilization. Should the results prove
to be unsatisfactory, this algorithm makes it possible to change the initial
conditions and perform the next step of iteration until acceptable results are
obtained.
Another class of problems aimed at obtaining optimal solutions arises
when one considers the possibility that one and the same bus makes several
trips. In this case, the number of vehicles for a given network of routes
may be minimized. One and the same bus may make two trips provided the
following conditions are satisfied:
The end point of one trip serves as the starting point of the other
The time of completion of the first trip coincides with the starting
time of the second one (allowing
the stop)
for the duration of the bus stay at
The overall duration of the two-step operation does not exceed the
maximum working hours of the driver (crew)
These conditions lead to mathematical minimization problems which can he
solved by known methods.
The bus traffic plan also includes a problem of assigning routes to the
various motor pools, the advisability of having each motor pool cater to a
specified route depending on the pretrip and posttrip expenses as well as on
the technological capability of a given motor pool. The expenses on a zero
run and on the maintenance of a bus at the starting point of the trip outside
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
the motor pool territory (overnight trip) may serve as a criterion for de-
ciding whether or not to assign a given route to a given motor pool. All
the other conditions constitute constraints. The solution to this problem
yields a list of fixed routes for each motor pool and a matching rolling
stock structure.
Thus, the problem of organizing interurban bus services is solved by
heuristic modeling techniques involving a certain hierarchy of subproblems,
whereby the solution for each subproblem is obtained by descending from
optimization to suboptimization or approximate solutions.
After the bus service plan has been compiled, a need arises to determine
the economics of the bus operation. Suitable technical-economic indicators
characterizing the bus operation may be obtained for each route and for each
motor pool on the basis of the bus schedules, the plans of rolling stock
assignments to routes, and routes to motor pools.
Aggregation of the route plan indicators by the hierarchical levels of
the Ministry of Motor Transport allows for planning passenger traffic for
the whole industry in technical and economic terms and objectively enough.
It is further envisaged that branch plans will he coordinated with the
national economic plans through automatic plan information exchange between
the Ministry of Motor Transport and the publican State Planning committee.
The passenger motor transport system falls in the category of open-loop
systems. To make it adaptable to the changing environment, a mechanism is
required for disclosing deviations from the plan as well as suggesting and
adopting decisions with a view to restoring the balance: i.e., meeting the
transportation demand within the set plan targets. Thus, day-to-day traffic
control problems must be dealt with. This system must perform the following
functions:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Determine the actual state of the system (accounting);
Compare the actual state against the plan and pinpoint all deviations
(monitoring)
Analyze the causes behind the deviations (analysis)
Prepare decisions to adapt the system (prediction)
Of the problems listed, the first three readily lend themselves to
formalization and can be described by quantitative models, so that auto-
mation presents no difficulties here. As for problem four, it belongs to
the group of problems which can hardly be formalized. On this basis, when
developing the first portion of the day-to-dav control system, the primary
emphasis was on automating the accounting, monitoring, and analysis functions.
At the present stage, accounting in the bus industry is done by use of
three kinds of special documents, viz. driver's trip ticket, bus ticket
record, and hooking-office record. These documents contain complete infor-
mation describing the state of the object during one working day, so automatic
processing of these data is essential if the accounting problem is to be
solved successfully. Currently, the Ministry of Motor Transport is in the
process of developing an integrated system for processing the data of trans-
port documents, which, when completed, will register and record the indi-
cators characterizing the operation of all kinds of motor transport. This
system will actually mark the emergence of a data bank to cover the entire
sphere of motor transport operation. The same system will also permit
monitoring the state of any object; i.e., checking the actual data against
the planned figures and revealing deviations.
Rearing in mind the multilevel structure of the passenger transport
control hierarchy in the Ministry of Motor Transport, effective control
requires that each level be supplied with enough information and within a
time 1~~R I~ a b1i i e -fFtb~f~- 6 0 21 of the
control system.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
310;
To meet this requirement, the system is broken down into hierarchical
levels, each having its own list of indicators, sensitivity threshold (the
least deviation to which a given level should respond) and schedule of data
delivery. Each higher level differs from the one immediately below by an
extended list of indicators, a lower sensitivity threshold, and longer in-
tervals of data delivery.
Analysis indicates that quantification of the above components is a
formidable task, what with the heuristic nature of the process and the
utter impossibility of formalizing the decision formulation and adoption
procedures. Hence, the quantitative values of the system components which
represent its state were arrived at by means of expert assessment. However,
to satisfy the demands of a great number of consumers and make the system
quickly adaptable to the individual requirements of managers, the system
has a built-in capability of furnishing additional information on demand
(a list of questions worked out in advance). Eventually the system will
permit a real-time man-machine dialog.
Alongside data on the state of the object, the day-to-day control system
must furnish information on the causes of deviation from the planned target
figures and also provide initial data for subsequent planning needs. To meet
these requirements, special data analysis programs are being developed on
the basis of a list of standard causative factors which disrupt the functioning
of the system.
With all the elements of the system operating on the basis of the same
data, the same software and hardware, we are sure we will be able to arovide
effective planning and control of the passenger motor transport operation.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 CIA-RDP79-00798A000200020005-9
AUTOMATION OF BOOKING AND RESERVATION OPERATIONS
ON SOVIET RAILROADS
B. E. Marchuk
It is common practice nowadays to book train tickets in advance through
reservation bureaus, transportation and travel agencies, etc.
Acceptance and registration of reservation orders are rather complicated
and important procedures. Aside from the sheer number of such orders
(hundreds of thousands every day), they come following a highly random pattern,
while would-be passengers present extremely diverse demands.
The advance booking system, which leans on a train seat file, does indeed
facilitate the booking operation; yet, it falls short of the main objective,
viz. high-speed, high-level service.
A typical seat file is found in a special bureau, or in a so-called reser-
vation center which receives telephone requests from personnel manning hooking
offices and direct calls from passengers.
Under such a reservation system, one dispatcher of the reservation
bureau is able to cater to some five permanently operating hooking offices.
But with a large number of booking offices, this system of dispatcher-con-
trolled operation becomes overcomplicated and all but unwieldy: the number
of dispatchers grows proportionately to the number of booking offices; the
time of operation gets longer as dispatchers have to queue up if one and the
same card is required by several people; finally, longer time is needed for
the booking clerk to get through to the dispatcher.
A variety of steps have been taken to improve the dispatcher-control
system and thus streamline the train reservation service, but all of them
have fallen short of their objective and failed to justify the expenses. The
Laboratory Chief of All-Union Scientific Research Instutute of Dept.
4pqQA*Eg# It 2QQt1I1&(19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
trouble is that the gain in processing time is disproportionately small if
the approach consists in simply beefing up the reservation Personnel or in-
creasing the amount of equipment.
There are certain limits to such growth which are rooted in technological,
economic, and biological factors, so that at a certain stage of dispatcher
system growth the quality of service starts deteriorating: the seat-occupancy
rate drops; the newly emergent drawbacks of the system detract from the pas-
senger satisfaction level; the response of the system to ticket cancellation
lengthens. This trend is particularly pronounced where the total number of
hooking workers reaches the 310 to 500 mark.
Hence, a need arose to develop automatic booking systems free from the
above-listed disadvantages and capable of effecting a major improvement in the
booking procedure.
Past experience suggests that at an annual passenger turnover of up to
four million the manual dispatcher booking system is sufficiently effective,
but passenger volumes above 10 or 12 million per annum absolutely warrant a
switchover to automatic booking.
Automatic booking systems using computers, data transmission facilities,
and booking office terminals, provide for automatic collection and Processing
of all information inputs from travel agencies and hooking centers.
Such svstems may be of two types: for simple hooking, and for the
entire scope of booking operations, including automatic sale of all forms of
travel documents. Systems of the latter kind, though 5 to 10 percent more
expensive, are far more economical, since they operate on the basis of much
more extensive passenger traffic information and thus can handle all aspects
of the booking operation. The advantages of such a system more than outweigh
the extra cost involved.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
At present, automatic reservation systems are employed in the railway
networks of Japan, Germany, Spain, Italy, France, Denmark and Canada. Of
these, the most comprehensive system is Japan's MARS-ins with a capacity of
up to a million tickets a day and the capability of catering to some 1,500
booking offices.
Since 1972 the Soviet railways have also had the benefit of an automatic
booking system called the "Express" which is superior to the volume of
operation to any of its European counterparts. However, automation of hooking
operations is a far more formidable challenge in the Soviet Union than in
Europe considering the large volume of passenger traffic and the much greater
railway mileage involved.
The automatic control system "Express" is intended to automate the whole
range of booking operations associated with the processes of reserving seats
on trains, keeping a check on the seats on long-distance trains, and selling
various kinds of travel documents. "Express" is a large computerized control
system operating in real time and designed to offer service to a great number
of passengers.
Analysis shows that in terms of labor consumption the procedures in-
volved in the bookine operation are ranked as follows: ticket registration
and documentation of sold tickets with a breakdown by individual bookinQ-
clerks, 65%; registration of vacant seats and apportioning them among the
booking offices, 20%; and accounting, 15%. It follows that primary emphasis
should be on facilitating the booking clerks' job.
Thus, the "Express" automatic control system executes the following
functions: (a) checking the vacant seats and furnishing them in response
to booking clerks' requests; (b) informing passengers and hooking clerks
alike as to the availability of vacant seats on trains; (c) determining
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
the size of the fare; (d) compiling and printing out the various travel and
auxiliary documents; (e) computing the amount of money from the sale of
tickets and other documents for the whole network of booking offices and for
each booking office taken separately; and (f) carrying out all forms of
statistical and financial accounting relating to passenger transportation.
Structurally, the system is composed of three elements: (a) a computer
center, which is an information storage and processing facility; (h) peripheral
(terminal) equipment comprising booking devices and information displays, which
are installed at the booking offices; and (c) switchgear and data-transmission
equipment linking the computer center with the data terminal equipment.
The computer center is a multiprocessor system made up of three inter-
changeable computers which control all phases of the booking operation at the
various booking points (terminal offices, agencies, railway stations within
the city, central reservation bureaus, etc.). Resides, the same computers
check and perform statistical reduction of the data on the sale, cancellation,
and reservation of train tickets. The computers further compile statistical
and running accounts on the utilization of the rolling stock, on the volume
of ticket sale for each booking office as a whole and for each clerk in the
office, etc. Normally, two of the three computers function in a synchronous
duplex mode, carrying out real time processing of the inquiries arriving from
the booking-clerks and inquiry devices, while the third computer is in hot
reserve, executing statistical accounting functions.
The computer center is so structured as to ensure that the system will
continue to function normally even if any two of the three computers fail.
The computer center capacity is 200,000 seat orders in 300 trains per day
(calculated on the aroung-the-clock basis), with a reservation period of ten
days for one-way tickets and forty-five days for return tickets. For people
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
who wish to reserve tickets long periods (up to two months) in advance,
there is an option of acquiring special "Passeord" coupons which guarantee
a seat on the train, but the actual tickets (without seat indication) are
issued any time before the train's departure.
The booking equipment is designed as an aid to booking clerks and,as
such, is installed directly in the booking offices. This set of equipment
comprises what is known as booking clerk manipulators (BCM), each consisting
of a clerk's console, a ticket printer, and a control device. One BCM can be
used by two booking clerks simultaneously.
The clerk's console is built around a full-size keyboard which permits
cutting down the time required to set an order and minimizing the error rate.
The console is also easy to master and can he readily switched from one
function to another depending on where it is installed.
The information on an order being executed may be sequenced as the
booking clerk sees fit. This information set includes the following data;
point of departure (terminal or station); destination (route); the train
number and the type of car which the passenger prefers; the kinds of docu-
ments ordered and their quantity; the passenger's privileges, if any; the
passenger's requirements to the seat layout; and the type of order (order,
cancellation, group order, report, response to an inquiry, invalidation of
a wrong document, etc.). If necessary, it is possible to fulfill an order
for a specific seat in a specific car.
Booking equipment, depending on the type of job, may be installed both
in regular booking offices and in specialized booking offices which do not
provide face-to-face service to passengers, such as telephone- and mail-order
booking offices; offices invalidating the fares connected with wrong docu-
ments issued by hooking offices; offices reserving seats on requests from
other citis d passing information on vacancies down the train route.
Approved or elease 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Regular hooking offices, which sell and cancel tickets, may be of any kind,
from all purpose to narrowly specialized ones. In principle, each booking
office could sell and cancel tickets for any train and for any date. The
total time it takes the booking clerk to issue two travel documents is not
greater than one minute. The maximum daily capacity of a single booking
office (with due regard for shift rotation of the personnel) may be as high
as 1,500 to 2,n00 travel documents.
The ticket printers automatically prepare travel documents on the basis
of a unitized letterpress printed form, all travel document forms being
letterpress numbered and rigorously accounted for. As a form is fed into
the ticket printer, the latter fills in the requisite information obtained
from the computer center of the system. The first line of the document pro-
duced by the system always contains its name: ticket, pass, children's
ticket, excess fare, excess fare to the children's ticket, excess fare to
the pass, ticket paid for by written order (check), or servicemen's ticket.
If the form is used as an auxiliary document, the first line also carries
its name: report, hoarding ticket, disembarkation information, etc.
The second and third lines are filled, in with the names of the points
of departure and destination, respectively. If the travel document is a
return-ticket, the station names change places.
The fourth line carries digital information (in decimal characters):
the number of the hooking office that has issued the travel document, the
number of the tariff zone of the destination, the time of train departure,
the train number, the train category, the car number, the seat number, the
ticket price, the document code indicating how the document has been acquired
and showing that the advance booking surcharge has been collected.
For passengers planning a chance of train en route, the upper right-
hand comb i L ~Q~1~~+~1~aR-t~e~ gt7$t4~i ~&4@fd0@AZOa@5 nger is
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
supposed to pass on the way. The fare is determined by the system auto-
matically on the basis of the points of departure and destination entered
into the system by the booking clerk at his console.
Information-inquiry devices exist in several versions: seat availa-
bility display boards and information devices for passengers and passenger
service personnel. The display board consists of a booking hall board (BHB)
and a booking clerk display board (BCDB). The BfBs are to be installed in
a prominent position in the large halls of hooking offices to inform the
passengers as to seat availability. The left-hand portion of each BHB
carries permanent information representing the train schedule, while the
right-hand side is variable and automatically shows the seat availability
picture for each train on the first, second, third, fourth, and fifth day
prior to departure. The seat status information for each day is broken down
into six categories. The display board is linked with the computer center
automatically, so that the information is constantly updated as tickets are
bought. Every new day, the information is automatically shifted to the left
by the space of one day. A single board is capable of providing information
on sixty-three trains.
Booking clerk display boards (BCDB) are installed in the booking clerk
booths, enabling the clerk to scan all trains at once for available seats
in one of the six categories. The BCDB displays the numbers of all trains
which have vacant seats of the category required.
The information-inquiry device is built around a typical ticket printer,
differing from the latter only by a simplified manipulator console design
and a more versatile printer.
The data-transmission equipment (DTE) may use half-duplex two- or
four-wire switching or nonswitching telephone channels, transmission being
c p a a seed f 1,200/600 baud. DTE provides communication links
c fuv eu r ortRelease 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
between the computer center and booking offices as well as between computer
centers over distances of up to 5,000 km. The system features thirty-two
intercommunication channels and ninety-six channels for communication with
the peripheral equipment. The maximum carrying capacity of a single
channel is two orders ner second.
when the "Express" systems intercommunicate, they share their terminal
equipment, for the accessing booking clerk to indicate the number of the
particular system he wants. Thus, the hookinc clerk has access to all the
systems and is accordingly capable of procuring seats at all railways. In
this case the computer centers which handle the booking clerk's demand
operate as information switchboards, storing the data transmitted.
Special switchgear is used to switch demands at the terminal locations.
The experience with the above-described system in the Moscow Junction
suggests the following conclusions:
1. The booking clerks' productivity can he boosted twofold or even
threefold.
2. The booking clerk's duties are largely simplified as the automatic
system disrenses with the need to calculate the fare and fill in
the travel document forms; nor is it necessary to compile a lengthy
- report on the travel documents sold, since it is printed out auto-
matically. Under the new conditions, the booking clerk actually
becomes an operator who hears a passenger's demand, inputs it into
the machine and receives the travel documents printed out by the
system.
3. Booking clerks can be trained to work with the system within a short
time (one week). The average error rate for booking clerks thus
trained is three to four percent.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 385
4. The system marks a dramatic improvement in the quality of service
offered to the passengers who are now able to book reservations at
any hooking office (irrespective of its location), for any direction
of travel, and not only personally but also by mail and by telephone.
Furthermore, passengers are kept constantly informed about seat
availability through information devices.
5. The system raises the cost effectiveness of passenger service through
(a) a higher occupancy rate; (h) a more effective utilization of
seats as trains move along their routes; and (c) prompt action on the
cancellation or use of additional cars or trains judging by the
.occupancy information (supplied as the tickets are being sold).
6. The system permits improving the quality of plannine and evaluating
passenger flows on the basis of relevant statistical reports.
According to foreign railways and air carriers, automatic reservation
systems conduce to a five to ten percent improvement in the occupancy rate,
while the investments into such systems are recouped within three to four
years through the cost reduction made possible by higher occupancy rates and
boosted profits from advance reservation.
Under current plans, Soviet railways are going; to be equipped at the
first stage with separate automatic systems to cater to a specified track
mileage. As these systems progressively come on line, they will be linked by
automatic communication lines, so that eventually all the systems will evolve
as a single multicomputer entity with enough scone to cover the entire rail-
way network of the Soviet Union.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
THE SIREN SYSTEM: A NATIONWIDE AUTOMATIC CONTROL SYSTEM
FOR BOOKING AND RESERVING SEATS ON DOMESTIC AIRLINES
V.A. 7hozhik.ashvili et al.
Airline booking and reservation is a typical mass service process of a
high degree of complexity steeminp from the high frequency of flights and
the extremely variable conditions, including the highly fickle weather situ-
ation.
That is the reason why air carriers were among the first to employ
real-time computerized teleprocessing systems and to have achieved the
greatest successes in this field.
Experience of U.S. and European air carriers with systems of this sort
suggests that they are instrumental in atemmin? the tide of personnel ex-
pansion, which would otherwise swamp all companies, and in achieving a
higher seat occupancy rate on the planes.
Aeroflot differs from most of its foreign counterparts in that it carries
out domestic operations on an enormous scale. Thus, the annual passenger
turnover is currently approaching the 100 million mark and shows no sign of
pace slackening.
The booking procedure for domestic flights radically differs from the
international booking operation. Domestic flights have long been a common-
place occurrence, so that it is paramount to relieve the passengers of ex-
cessive formalities and make the booking procedure a fast and simple
operation. The passenger must be able to book a seat on the plane at any
time prior to the takeoff. To achieve this goal, the booking operation must
turn into a simple and fast procedure. Besides, domestic flights are far
cheaper than international flights. The Aeroflot domestic rates are below
Chief Constructor of the Automated System, Main Computing Center,
Ueflppteeed Rdr FMddsdk1Wt*W19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 387
those used by most other carriers. Hence, the automatic system should not
be overly expensive, calling for a thorough analysis of the various services
in economic terms and for a no-frills plan.
These considerations have been taken into account in full measure while
developing the Siren system, a unique system in its own right.
The agent-system interaction procedure is extremely simple and fast.
In most cases the agent has to access the system only once or twice to cater
to each passenger. Tickets are by and large printed out automatically. If
need he, the system may automatically print out several tickets, requiring
no additional accessing.
The response of the system lies in the range from 1.5 to 3 seconds,
while the ticket printing operation takes 5 seconds. Therefore, to take care
of a passenger, all the agent needs to do is to receive a demand and the
payment, which rarely takes more than one or two minutes.
At this rate, around 20,000 passengers can buy tickets in one hour.
These principles built into the Siren system are going to he extensively
used in the nationwide system, too.
The latter system is designed for a capacity of 200 million passengers
per annum. No other system in the world can boast such a capacity.
The-nationwide system is made up of several zonal subsystems covering
the entire territory of the county. The zonal subsystems include the agencies
and airports which territorially fall within each zone. In terms of function,
the zonal subsystems are identical and have the same type of configuration.
The data processing centers (1PC) are interconnected and linked with their
component units by means of a communication network, so that together they
form a single system.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The Aeroflot nationwide system was conceived as a single system in the
full sense of the word. The centers and zonal subsystems are not going to
compete for passengers, but will rather strive for a common goal for the good
of a single carrier. While mapping a complicated route involving changes or
while searching for an optimal growth strategy, the subsystem centers will
view one another as members of the same family pursuing common objectives,
rather than as competing or even cooperating companies. This will certainly
be reflected in the ways user programs will he executed.
Furthermore, there will be an additional task of optimizing the system
as a whole, rendering the Aeroflot system different in essence from the
totality of systems run by different carriers.
In technical terms, the system will widely use minicomputers for small
centers, alongside huge machines that will be installed in major centers.
The first system, the groundwork of the forthcoming nationwide system,
has been in operation since April 1972. It is the above-mentioned Siren
which controls over 500 flights from the Moscow airports daily and affords
fast and equal access to the seats on these flights for 250 agents in Moscow
and another forty-two cities of the country (figure 1).
Designed for a daily capacity of 50,000 seats, the Siren is one of the
largest systems in the world in terms of capacity.
In 1972, the Siren gave service to 2 million passengers; in 1973, 4
million; in 1974, 6 million; and the expected figure for 1975 was in the
vicinity of 8 to 9 million.
Thanks to this system, all operations are now executed much faster and
the productivity of the agents has been raised severalfold, largely adding
to the passenger satisfaction index and thereby boosting the traffic volume.
Upon introduction of the Siren, the seat occupancy rate started to
grow Ap~~o~# ~' - I efebs%'Hb1r11A',usCt TRW P~9 00798A000200of the 02 005 9oportion of
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200016805-9
c Jo~
Approved For . e1e 1 /1 CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
booked seats on the system-controlled flights, while curve 2 presents a
similar picture for the uncontrolled flights. The figure clearly shows
that the controlled flights definitely have an edge on the flights operating
outside the system. The relative magnitude of this edge is shown in figure
2-b. It is expected that in 1975 it will reach approximately 89%. (The dash
lines in the diagrams represent predictions; Ikb means first quarter).
The Siren is built around an integrated system of technical facilities,
all made in the USSR., which include a highly automated data processing
center, data transmission equipment and a large number of display terminals.
The system uses a special language for man-machine dialogue and an operating
system controlling real-time multiprogram operations. The Siren is a means
of automating a number of services, booking operations, and monitoring the
activities of the passenger service personnel.
The need for such an automatic system stems from the fact that the old
methods of booking and reservation as well as the techniques for controlling
these processes run increasingly counter to the mass traffic requirements
in a situation characterized by fast growth of the airline network, larger
passenger aircraft, higher speeds, and increasingly hectic pace in the
activities of airports and agencies. Thus, prior to the introduction of the
Siren, the files of the Main Air Traffic Agency run by the Ministry of Civil
Aviation contained up to half a million seat cards which were accessible to
over 250 booking clerks, dispatchers of the transit services of the Moscow
airports, and personnel of the reservation bureaus and agencies located
outside Moscow. Every day up to 8 to 10 thousand seats were booked by
cable. In spite of the regularly increasing staff handling the booking
demands, the booking clerks had to wait for an average of thirty seconds to
one minute, and as often as not up to ten or fifteen minutes, which caused
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 CIA-RDP79-00798A0002000200e%19
,4 b"
040 Ais 1-94.9. 1970 1911-1FT2 1913 1'9T~r / i "
~oo~sio3
-
-
7 ~K
~,(~~! c7wya, Figure 2a
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
queueing and complaints. The cable demands from other locations for seats
for transit passengers were dealt with only the next day.
Curing one shift in Moscow, several tens of thousands of telephone demands
would be handled. The hectic pace, the often poor audibility, the inevitable
blunting of attention under conditions of incessant monotonous calls, the
swarms of people manually writing down short (five to six digits) notes, all
added up to errors detracting from the quality of the service and causing much
damage. Naturally, execution of orders and monitoring of the personnel's work
were all but impossible.
The Siren system automated to a maximum possible degree the most typical
operations, furnished the booking clerks and dispatchers with diverse and
accurate information and thus made it possible to more than double the booking
speed. As a result, the quality of the service was raised considerably and
sizeable saving of social time was achieved.
The heart of the Siren is a data processing center (DPC) (figure 3). It
is composed of two computer centers with accessory devices whereby the two
computers can operate jointly, and special devices interfacing the computer
facilities with communication channels. Each computer is equipped with a
processor, memory units, input-output devices, and standby facilities.
.The DPC memory (both computers) stores all information necessary for
the system to perform its technical functions: data on schedules, rates,
tariff distances, aircraft seat mock-ups and the sequence in which they are
booked, and restrictions connected with traffic rules. The system functions
are backed up by a software support system made up of a great number of
programs. A specially developed real-time operating system provides for the
operation of the DPC in a variety of modes, communication with the many data
transmission channels, and response to emergency situations.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-3
L___. -
Figure 3: 1 - Mainframe memor (1 reserve)
2 - Processor
3 - Interprocessor link via interface 1R
4 - Input-output channels (1-reserve)
5 - Input/output device
6 - Control devices
7 - Distribution-conversion complex
8 - Magnetic-tape memory unit (2-reserve)
9 - Magnetic-drum memory unit 8 mechanisms (4-reserve)
10 - communication lines
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
ti
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The chief operating mode of the DPC is duplex operation whereby each
request is processed by both computers, the results are compared and issued
to the caller in case of complete coincidence of the two sets of results.
Should there be a failure of any device, the computer center in which the
failure has occurred is automatically switched off, and the system switches
into a simplex mode while the DPC operator is notified about the cause of
the failure. After the fault has been eliminated, the system reverts to
the duplex mode, the switchovers from one mode to the other in no way
affecting the processing operation.
The particular configuration of the DPC which involves two independent
computer centers capable of functioning both jointly and separately, is an
important factor contributing to the realiability of the system as a whole.
To the same end, all principal DPC devices have built-in reserve capability
and there are special programs for speedy faultfinding. Thanks to all these
facilities, the DPC has a high level of reliability.
Special measures have been taken to protect information. Apart from
protection programs, the system uses magnetic drums which continuously
record the information status in real time, as well as any change caused by
booking, cancellation, or any other operation. This information can be
used.to restore the system.
The information stored on magnetic drums is updated by the results of
each individual access.
Operating experience indicates that at the present level of reliability,
the duplex mode is redundant and one of the two computers may be employed to
solve concomitant problems in a batch mode.
The main technological unit of the Siren system which provides for
man-machine interaction, is a console. The operator's dialog with the
system is effected by use of a keyboard dd c~tT ai l hi 1, shows the
Approved For Release 2001/11/19: CIA-RDP79-60798AO ah8Z22OD5=-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
text of the operator's call and then the information furnished by the DPC
in response to the call. The console has a built-in printer for documenting
the most important data obtained by the operator. The booking offices
covered by the system are equipped with printers for making out tickets.
The Siren consoles are installed in the subdivisions of the Ministry of
Civil Aviation Main Agency in Moscow, at the airports of the Moscow area, in
some communities in the Moscow region, as well as in the air travel agencies
of another forty-two cities of the country. The consoles may perform different
functions depending on the particular task with which given subdivisions are
charged.
In terms of functions, the Siren system can be subdivided into the
following technological units:
Department of short-term planning of commercial flight loads (for flights
taking off from Moscow), which includes a day-to-day system control center.
The airports of Moscow are equipped with consoles; using these the control
center dispatchers execute similar functions, and the system with information
about the newly vacant seats immediately prior to flights, and obtain data
on the booking operation for transmission to the computer center.
Most of the available consoles are installed in the booking offices
which cater to passengers who wish to buy tickets in person. A reservation
office has been set up to take care of passengers who make telephone bookings.
This office interacts with the system via its consoles and delivers the
tickets to the passengers by messenger or else issues them in a special
booking outlet at the passengers' convenience.
Certain booking clerks, transit dispatchers at the Moscow airports,
reservation agents, and some other workers cannot interact with the Siren
system since consoles are not available to everybody. To enable the group
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
to have access to the system, the personnel of the commercial flight loads
division receive telephone calls from them, pass the calls to the system,
and transmit the answers to the callers by telephone.
Special mention should be made of a large group of consoles installed
in over forty agencies located outside Moscow. Passengers in these communi-
ties are able to book seats on flights involving a change in Moscow.
It would be erroneous to assume that the range of the Siren is unlimited
by the locations of its consoles. The out-of-Moscow agencies connected to
the Siren form their own networks of agencies of sorts in their vicinity,
which book the seats on the system-controlled flights through the medium of
said agencies with which they keep contact by cable or telephone. Thus, for
instance, the consoles of the Pyatigorsk agency cater to all the booking
points of the Mineral'nye Vody resort area and the other booking offices in
its zone. Similarly, the Alma-Ata agency uses the Siren to hook passengers
from all cities of Kazakhstan and even some cities outside that republic on
flights which involve a change in Moscow. The number of cities thus covered
by the system directly or indirectly is well over 400, their network encom-
passing practically the entire territory of the Soviet Union.
The system contains mock-ups of aircraft with fifty different seat lay-
outs, The seats are sold in two different sequences depending on the type
of aircraft, from the front row to the last one or the other way around.
However, the system can also accommodate passengers claiming specific seats.
With the information completely centralized and the system operating
very fast, the system operators, wherever they may be, enjoy practically
simultaneous and unlimited access to all data. Thus, a seat cancelled, say,
in Magadan in the far east of the country, may be sold within seconds in
Leningrad. It takes from 1.5 to 3 seconds for the answer to come through a
telephone channel and 20 to 25 seconds by cable.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
397
The Siren includes an information display showing the situation with
seat availability for the next seven days. The display incorporates an
automatic calendar and a system-controlled vacant seat indicator. This dis-
play is installed on the airport premises in plain view of the passengers.
An important place in the Siren system is occupied by the communication
network which links the numerous users with the data processing center.
Information is transmitted over standard unswitched telephone and telegraph
channels. In order to improve the reliability of the data communication
network, use is made of accessory devices, viz, data transmission facilities
whose functions include encoding, transmitting, decoding and checking the
information being transmitted for correctness.
Console switches, employed in those cases where more than one console
is installed, contribute to the effective use of the communication channels.
One console switch using one or two communication channels can cater to
eight to sixteen consoles.
The system may perform many different functions, such as issuing or
cancelling seats and selling tickets; furnishing, information about the
schedule, seat availability, number of vacant seats on a particular flight,
changes in the schedules, etc.
A number of functions of the system are executed requiring no operator's
request: calculation of the total rate if a group of passengers requires
service; offering the next flight which has the required number of seats;
making up records; compiling statistics of, say, unsatisfied demand.
The bulk of requests are concerned with booking.
If the operation is ticket selling, the information fed from the Data
Processing Center to the console has the format of a ticket and in the course
of printing all the blanks of the form are precisely filled in with the
names a*ftwe*tFxrdkleamd2f 'ttMdt.9ai(3 7sl07 Ai 11t011AObwC105* date and
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
time of departure, and the rate of payment. The tickets may be sold at a
full rate or at a discount (school children, university students, etc.).
In the latter case, the system checks whether the particular grade of
privilege corresponds to the rate of discount, reminds about the time for
booking; if several seats are requested, the system furnishes the overall
rate with due regard to the size of discount for each passenger.
The seat issuing operation is executed in those cases where the
passenger holds a paid transit or return ticket without a fixed data.
While executing both operations, if the required flight does not offer
the needed number of seats, the system automatically retrieves the two
closest flights (prior to, and after, the flight in question) which have
the number of seats as requested. The system indicates the number of seats
available on the flight requested and furnishes all information relating to
the two alternatives offered. Thus, the operator is able to help the
passenger in choosing the most convenient flight.
The cancelled seats may be returned to the system.
The operator can also use the seat inquiry request to obtain infor-
mation on seat availability on a particular flight as well as the schedule
and rate data. If the seats are unavailable, the system again offers two
closest flights. If the passenger has not named the desired flight, but
only the desired time of departure, a schedule inquiry is in order., In
response to this request the system gives information on several flights in
the same direction, the data being sequenced in a chronological order, as
well as data on the flight numbers, the airports of departure, the times of
departure, and the number of vacant seats on each flight. The package
further contains the data on the rate, the day of the week which falls on
the date given, and the time of inquiry. In case there are many suitable
flights, the inquiry may be narrowed to a certain time interval.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
On the operator's demand, the system sums up his activity for the shift
or for any time of day, indicating the number of booking operations, seats
issued and cancelled, and the receipts.
The dispatchers of the control center run the system through control
requests, such as temporary change of schedule for one or several days or
permanent schedule change, introduction of a new flight, cancellation of a
flight, temporary ban or removal of ban on ticket sales, increase or decrease
in the number of seats, substitution of one type of aircraft for another, and
the like.
Besides, the dispatcher may request information on the passenger load
on any flight, the number of flights in any one direction for which all the
seats were sold out, the results of work of any operator or all operators,
any subdivision or the whole system, the latter document indicating the
number of seats sold as of the time of inquiry, the number of documents
issued by the system, the number of operators employed in the system, etc.
All these operations are executed in real time.
The system operators' work is monitored by use of the daily records
issued by the system. This document outlines all operations carried out
during the day, each operation being entered in the record under the number
corresponding to that on the document printed out by the operator and with
the time of execution indicated. The records contain information on the
type of each operation, the flight number, the date, the seat number, the
rate, the kind of discount, and the commission. If the operation was
executed outside Moscow, the records will show the date of flight number of
the passenger's arrival in Moscow.
Taking into account shift rotation, the Siren system employs over 700
operators, many of whom are located thousands of miles away from Moscow.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
With this number, a constant flux of employees is inevitable, which places
particular emphasis on the ease of trainine, on the one hand, and on the
foolnroofing of the system, on the other. These considerations were allowed
for by the system designers.
Information exchance between the operator and the system proceeds in a
comprehensible language; the names of cities and many communications are
printed out in full; the abbreviations used need no special conversion tables
to he understood. The system contains information on a special instruction
flight which can be used to train operators on the job.
If the operator does make a mistake, requests information on a non-
existent flight, misses some element, or uses a wrong sequence of elements
in the set, the input vetting program will block such a request. The screen
will display the text of the erroneous request and indicate the nature of the
error. Thus, not only is erroneous information barred from the system, but
the operator is assisted in locating and easily correcting his mistake. The
system, in effect, trains the onerators.
Over the period of pilot-scale operation, the system has been enriched
with almost forty new programs and changes in the old programs aimed at im-
proving the quality of the service, raising the efficiency and reliability of
the system, and facilitating the operators' job. All these changes were
effected without shutting down the system.
In 1974, daily control centers were set up at the Moscow airports,
marking a new stage in the system development plan. These centers permit
tickets to be sold practically until the very takeoff, the seats vacated by
passengers switching to earlier flights to he entered into the system, and
additional seats to be entered into the system if a flight is executed by
a plane with a larger number of seats than envisaged by the schedule. Thanks
to these centers, the occupancy rate is bound to rise still higher.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The above-listed advantages of the system can be enjoyed not only by
the hooking offices of Moscow. The benefits of the system are even more
tellin' in the eyes of the passengers chancing flights in Moscow. It must
be admitted that the traditional reservation system for transit passengers
falls far short of meeting the Passengers' requirements. Transmission of
cabled requests, processinj of cables at the post centers, the booking pro-
cedure proper--all take a lot of time, so much so that passengers are able
to acquire a transit hooking with a confirmed reservation the next day at the
earliest, and not infrequently on the third or fourth day. But the Siren
revolutionized the entire transit booking business. Now an efficient service,
equivalent to that offered to passengers booking tickets in the booking
offices of Moscow, is available to thousands of nassengers in many cities
who fly via Moscow. The operator sitting at a console connected with the
data processing center by an ordinary cable link receives information from
Moscow within twenty to twenty-five seconds; the information for non-Mus-
covites is as convenient as that provided to the capital inhabitants. Of
course, this modern form of service is in a different class from the
traditional cable hooking operation. Initially, the plans envisaged that
the Siren would cater to twenty-eight cities. However, already today their
number has passed the forty mark, and the process of expansion of the system
will continue.
It is not in the least accidental that the passenger community has
hailed the Siren with enthusiasm: in 1973 the number of bookings from out-
of-Moscow agencies increased from 65,000 to 593,000 as against 1972: and in
1974 the volume of bookings skyrocketed by almost three times.
It has been mentioned that the Siren is a man-machine system in which
the human operator plays a very important role. The Siren or any other
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
comparable system calls for an intensive training program (740 operators
were trained in Moscow and in other cities), but also for a drastic recon-
struction of the organizational pattern, new technology in all segments of
the service, new instructions, and new manuals.
In one year, i.e., from May 1972 to April 1973, the Siren gave an
economic effect amounting to 3,120,000 rubles, primarily through a higher
occupancy rate. The savings of social time constituted about 3 million
hours within the same year.
The system of moral and material incentives for the personnel has been
revised, and new targets planned for the main a'ency.
The automatic system also performs a major social function. The
operators' condition has been dramatically improved. The booking clerks
are now spared nervous breakdowns which earlier accompanied their attempts
to place a call to the flight loading people, particularly when audibility
was bad; they work much more calmly and smoothly. They can communicate
calmly with the passengers and obtain fast and full answers to many questions
that would otherwise cause difficulties! they no longer have to consult
reference books or seek clarification by telephone. A dramatic cut has been
achieved in the number of conflicts because of lack of seats on flights,
since.the system immediately offers acceptable alternatives, and such an
answer obviously meets with a far less adverse reaction on the part of the
passenger. The passengers also feel much better because now they have to
wait in line half the time they would spend under a manual hooking system,
even during periods of peak demand.
The situation has changed even more dramatically for the better in out-
of-Moscow agencies equipped with consoles. Thus, while earlier passengers
had to come at least twice to procure a seat on a flight passing through
Mosco' now they ceive their tickets immediately. This affected in a major
i4 proved or Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 403
way the structure of transit passenger flows: the number of passengers
coming to Moscow on an "open" ticket, i.e., one containing no date of the
next flight, has been cut down considerably. The vastly facilitated reser-
vation procedure has boosted the number of neople who prefer to fly via
Moscow, having previously reserved a seat on a specified outbound flight.
As transit passengers reserve seats with a change in Moscow, they are
immediately told the seat number, which relieves them from the duty of
havine the ticket registered by the transit dispatcher and thus snares them
the formerly unavoidable exhausting waits to obtain a seat on the next leg
of their travel. This has also benefited the transit dispatchers of the
Moscow airports.
For all these reasons, the volume of passenger traffic has risen by a
substantial margin. Thus, during the first ten months of 1974, the seat
occupancy rate for the Moscow air junction rose by 3.1% as compared to the
comparable period of the previous year, and amounted to three times the
average figure for the rest of the Aeroflot units.
The experience with the development and introduction of an automatic
system for booking and reservation operations has proved beyond all doubt
that systems of this type can find wide application as a means of automating
many processes in the civil aviation field which are characterized by high
speeds of execution and call for fast decisions within a limited time.
Examples of such operations are as follows: daily control of airports,
short-term crew planning, redistribution and mapping: of optimal routes for
supplying spare parts to grounded planes, inquiry and information service,
information flow control on a nationwide basis, and many others.
Centralized storage of vast quantities of diverse and rapidly changing
data, availability of this information at any point in time to any user who
may be th a et?po"i 4fa a '0Sf~9/ e C~I~u S~PfS-007983A dQU20005 ,,)rage
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
and allocation in real time of limited resources (plane tickets, hotel rooms,
hospital beds, tourist passes and the like) to hundreds of users, high reli-
ability of information, high degree of automation of operations--all these
factors testify to the enormous potential of teleautomatic mass service systems,
such as the Siren, and not only in civil aviation, but in many other branches
of the national economy as well.
The large body of designers working with the Ministry of Instruments,
Means of Automation and Control Systems, together with the experts of the
Ministry of Civil Aviation, who designed and built the Siren, can be justly
proud of a major achievement in developing an automatic mass-service real-time
system with a unique information network, which links the remotest cities
of the country with the central computer in Moscow.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9405
INFORMATION MODELING FOR MANAGING SOVIET
MARITIMF. TRANSPORTATION
*
V. S. Bondarenko
Goals and Methods of Modeling
It is common knowledge that the control process is composed of the
following elements: acquisition of initial data characterizing the status
and behaviour of the object of control and of the environment; transmission
of these data to a control facility: accumulation, storage, and analysis of
the information thus received by the control facility; formulation of possible
alternatives of control decisions: selection of the best decision for a given
set of circumstances; and transmission of the latter decision to the object
of control.
Formulation and selection of control decisions are procedures which
lend themselves to varying degrees of formalization, may or may not include
optimization algorithms, and may or may not be computerized. However, what-
ever the case may be, no control decision can be developed without initial
information which is liable to vary widely in terms of composition, contents,
acquisition intervals, and other parameters, the variation being due to both
objective causes (different objects of control, different functional goals
of the objects of control, different quality criteria and the like) and many
subjective factors, including individual traits of the control facility
managers and rank-and-file personnel.
Since the real objects, their functional processes and control algorithms,
are highly complex, a huge number of control alternatives and for practical
purposes, an infinite number of information models can be suggested. Each
information model features its own depth and degree of description of the
object of control, its own mode of inputing data into the control system, and
* Ada o ec For F eileaseen~e1r11 19 artmeR a7 07 ~-00 020 O 05-9
Chief 6 a n ompu ng p ea ans on
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
its own quantitative and qualitative parameters characterizinc* the control
system's response to changes in the object of control and in the environment.
In order to provide the most rational type of control systems combining
high efficiency with favorable economics, it is paramount to analyze thoroughly
and prove the all-round advantages of a suitable information input. .
The comparative approach to the evaluation of various input systems and
the fact that some of the possible alternatives simply cannot be realized,
suggest that it may he helpful to try and simulate information processes.
Information modeling in this context is viewed as an oranic component
of the comprehensive control system model, on a par with economico-mathe-
matical modeling of objects of control where problems of optimization, pre-
diction, and analysis are involved.
However, as distinct from economico-mathematical modeling. employed in
highly developed control systems which are called upon to formalize the pro-
cedures of optimal decision making and quantitative analysis of trends,
cycles and future programs, information modeling is feasible in all kinds of
control systems, including the most trivial computerized and noncomputerized
data processin- systems.
Information models can he broadly classified into two types: particular
models characterizing some local or narrowly functional elements of the
information input of real control systems, and generalized models which may
be regarded as analogs of full inputs of real control systems.
The matrix techniques of analyzing and synthesizing the information
inputs of control systems, which are widely used in modern design institutions,
can be successfully employed for constructing particular and generalized
information models. Comparative technico-economic analysis of these
models--particular and generalized alike--is a method for subsequently
selecting decisii s most co ffornins to he goals and objectives of control.
Approved For Release 001/11/'i9 : CrA-RIiP79-00798A0O0200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
The importance of information models rises as control systems develop,
i.e., with the degree of their sophistication, resolving power and, naturally,
cost.
Objects of Simulation
The information models of control systems should reflect, within certain
limits of accuracy, the state and behaviour of the objects of control and of
the environment, on the one hand; and the response of the control systems to
the changes occurring in the objects of control and in the environment, on
the other. The groups of indicators or information modules which define the
state of the object of control and of the environment, present an important
characteristic of the information input of any control system and, hence,
must be among the primary parameters coming up for simulation study. The
most efficient set of indicators is apparently one that is minimal but
sufficient for describing the object of control and its environment in terms
of formal and informal decision-making algorithms used in a given control
system. But the algorithm structure, in its turn, depends on the available
data base as well as on the scope of numerical methods and the technology.
So, modeling of the set of indicators is closely associated with the
economico-mathematical modeling of the object of control and decision-making
processes. If all these models are viewed together, it is possible to select
effectively the important characteristics of the object of control for regular
measurement, registration, and inputting into the control system.
The need and advisability of inputting into the system of a group of
indicators characterizing the state of the object of control and of the
environment depends on a number of other factors:
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
1. Forecast of changed requirements to the data base of the control
system brought on by the evolution of the object of control and
by the dynamics of its functions and control objectives
2. The technological capability of the available facilities for data
acquisition, transmission, and processing as well as the prospects
of their development
3. Economic considerations and, among other things, evaluation of the
costs of increasing the amount of information inputted into the
system against the effect due to the possible improvement of the
quality of control
In multilevel control systems, the number of hierarchical levels and
the way information-computation tasks are allotted to them may affect the
technical characteristics and the economics of the system. Consequently,
information modeling may be set a goal of determining the most efficient
indicators of information filtration and condensation as the data move from
one hierarchical level to another. Since such indicators depend on the
ability of each level to process information and make decisions, analysis of
the various patterns of allotment of informational and computational tasks
in the course of information modeling may lead to an efficient vertical
structure of control facilities and the optimal information volumes to be
processed on each level.
The modes of inputting and processing of the starting data are important
functions of real control systems, which by and large determine the infor-
mation capacity and sensitivity of the systems. The frequency at which the
on-line numerical values of the'parameters of the object of control and of
the environment are registered and introduced into the system, is either
determined by the control resolution (critical value of displacement) or
else rey~~d~lasfiet~60lJa1@I~4-F2~Pf3~W?9E90909(J-xl~on
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9409
modeling, the mode of information registration and inputting is varied and
each alternative evaluated in technical and economic terms in order to make
a justified choice of an optimal or near-optimal "pulse" of the control
system.
To combat irregularities in the arrival and processing of information
in the course of information modeling, the data are smoothed through varying
the information density of the various groups of indicators during the week
or the day, taking into account the effect of information delays on the per-
formance of the control system.
Modeling of data arrays and banks gives an idea about the structure and
configuration of the data base which is most advantageous for a given control
system. Bearing in mind that the data base is among the most crucial elements
of modern control systems, the various configurations of the arrays making
up the data bank as well as the various relationships between them are
analyzed by comparing the costs of setting up and operating data banks
against the current and predicted requirements of the control systems with
a view to avoiding any reconstruction costs in the future and minimizing
the direct and indirect expenditures.
Any general decision offers some advantages as well as some disad-
vantages-as against a totality of particular decisions. Hence the need to
compare generalized data banks with specialized arrays before deciding which
alternative to adopt. The technical parameters and the economics of a
control system are known to depend on how successfully one evaluated the
comparative advantages of slow and fast external memory units, determines
the efficient balance between the volumes of initial data and the inter-
mediate information package resulting from statistical generalization and
smoothing (mean values, indices, characteristics of dynamic series and the
like), and evaluates the various data configurations and relationships between
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
arrays.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Summing up, one can assert that the goal of information modeling is to
provide answers to the following questions:
1. What kind of information is needed for effective control?
2. What is the economic time frame of data acquisition, transmission
and processing?
3. What are the requirements to the quality of information: complete-
ness, reliability and timeliness?
4. How should the information input develop to match the evolution
of the object of control and of the control system?
An optimal data base for a control system may be put together by aggre-
gating the partially optimized decisions on the composition of the set of
information indicators, the allotment of informational and computational
tasks to the various levels of the control hierarchy, the frequency with
which information modules are entered into the system, the intervals at
which control decisions are made, the success in smoothing irregularities in
the system operation, as well as the composition and configuration of the
data arrays and banks.
Modeling Experience
Information modeling is employed for developing an automatic system
for controlling sea transport operations. In this system, the chief infor-
mation elements are shipment orders and data characterizing the operation
of ships, the situation on routes and in ports, as well as the international
cargo and freight setup. An important role is also played by a constant
flow of information on the cargo shipment capability, technical and operating
characteristics of ships and ports, distances between world ports, freight
and tariff rates, as well as other information of a technical and economic
nature stored in the system.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9411
while designing the automatic control system, the indicators included
in shipment orders, ship and port records, freightage documents, as well as
all the other indicators contained by permanent and dynamic arrays, were
selected by the results of an all-round analysis of appropriate information
models. The models were obtained on the basis of currently available algorithms
for optimizing control decisions, with due regard for the requirements and
wishes of the managerial and rank-and-file personnel of the control bodies
as well as for the recommendations of the designers of the functional sub-
systems incorporated in the automatic control system under development.
The structure of the functional subsystems was defined by the results of
analysis of the activities carried out by the sea transport control bodies.
It included:
- a freight market analysis subsystem, which uses standard freightage
reports as the input, and the results of analysis and forecasting of the
freight market in its various parts by the kinds of cargo and by season as
well as other freightage conditions as the output;
1. A subsystem for day-to-day accounting and analysis of the merchant
marine and port operations. In this subsystem, the input is composed
of daily reports of ships and ports, plans, schedules and timetables
as well as various operating standards; whereas the output consists
of reference and analytical material characterizing the state and
dynamics of shipment operations and the data describing how the
plans, schedules, and timetables are being fulfilled in terms of
fleet operation and port work, and how the operating standards are
being compiled with.
2. A statistics and bookkeeping subsystem, in which the input is built
around statistical and bookkeeping reports of the shipping lines
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
and ports, while the output comprises analytical material charac-
terizing the financial status of the merchant marine and its sub-
divisions with a breakdown into periods and operations.
3. A subsystem of short-term planning of fleet and port operation.
Its input includes the cargo owners' orders, operating standards,
information on the freight market setup and on the times and places
of the ships' fulfilling their obligations. The output of the sub-
system consists of optimal draft plans of fleet utilization as well
as the schedules and timetables of ship operation.
4. A subsystem of long-term planning of sea transport development.
The input of this subsystem comprises the starting material for
forecasting shipping operations and for producing economico-mathe-
matical models of the optimal strategy of development and allocation
of the sea transport resources. Its output consists of the recom-
mendations on plan and design decisions.
Alongside these principal subsystems, the sea transport control system
being developed includes three back-up subsystems: technology monitoring
subsystem, logistics subsystem, and personnel subsystem.
Analysis of variants was found to be the simplest and most effective
means of studying information models and selecting the best design decisions.
The variants differed in the comprehensiveness of the information
picture and in the degree to which they met the demands and wishes of the
control personnel. A list of comvulsory indicators was compiled, which
included all factors indispensable for making forecasts and plans of ship-
ment, ship schedules, and other regular control decisions, as well as for
systems analysis of the most important aspects of sea transport operations.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9 413
Those indicators which are only needed from time to time and only in
response to special requirements or to a disruption of the normal course of
shipment, are considered as claimants to be included among the compulsory
indicators in case statistical evaluation shows them to be needed frequently
enough and economic assessment indicates that the benefits derived from their
routine acquisition outweigh the added costs.
The final selection of the claimant indicators was made on the basis of
several criteria, the most important of which were control problems for which
they were considered necessary, the frequency of their active use, and the
costs involved in their acquisition, transmission, and storage.
To determine the relative contributions of the indicators into the
control process, use was made of questionnaires distributed among the designers
of the functional subsystems and the control personnel. Also used were matrix
and statistical methods of analysis. The results of this work went into a
set of indicators made up of several different groups which were to be
routinely inputted into the system.
The frequency of inputting the various groups of indicators was worked
out on the basis of economic considerations, taking into account the impact
of the inputting intervals on the control system performance. Since direct
relationships were impossible to obtain, the method used was expert assess-
ment subsequently subjected to statistical evaluation. Where input data
serve to obtain mean values, such as characteristics of a dynamic series,
standards and the like, an additional factor was considered, viz. the
stability of the mean values and the stability of the results of control
problems derived on the basis of said mean values.
The three-level structure of the control hierarchy (shipping line-
agency-ministry) mandated a study into the quantity of information needed
by all 1 vels to cope with their respective functions. Here situation
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
simulation and analysis of alternatives were used. As a result, decisions
were made whereby the second and third control levels jointly served by the
Main Computer Center were to be furnished with substantially smaller
quantities of information than previously envisaged, and the information
reaching these levels was to undergo special screening.
All main information arrays were designed as generalized entities to
cater simultaneously to many subsystems and many control problems. The
standardized structure of the information arrays allows of a gradual
switchover to a systematically organized data bank.
Apart from a considerable economy of inputting and updating costs, such
a data bank makes it possible to automate the programmed printing-out and
displaying operations. Thus, the new demands are to be presented in a
specially developed high-level language whose vocabulary is unambiguously
interpreted by the standard program modules. The language vocabulary was
composed of the names of standard arrays and their elements, and the trans-
lation and editing procedures were tied in with information elements and
formally described together with the latter, dramatically cutting down the
programming time and improving the performance of the entire control system.
Modeling Effect
Information models are an effective means of evaluating different con-
figurations of the control system data base and choosing a good decision
before the stage of actual design, thus saving a lot of time and money when
designing new, and reconstructing installed, control systems and-which is
particularly important--affording a practicable way of optimizing control
systems.
Information alternatives are evaluated on the basis of the following
factors: the amount of money spent on designing and introducing a given
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
system or its elements featuring given information characteristics; average
annual costs of maintaining the system or its elements; and average annual
costs of progressive modernization of the system or its elements.
Unfortunately, the information models can be compared only on the basis
of scant and unreliable initial data. Here the basic principles were as
follows:
1. As the accuracy of the initial data decreases, the methods of com-
parison and evaluation become rougher.
2. Particular decisions require more detailed calculation and more
precise techniques of comparison.
3. If reliable initial data are lacking, it is mportant to obtain as
many expert evaluations as possible.
The results of evaluation of the effect of information modeling indicate
that modernization costs are substantially lower than the overall costs of
control system design operations, and modeling effects such savings that the
modeling costs are recouped many times over within a few years of the system's
operation. It is also clear that if the objects of control are to function
efficiently, the inner structure of the control system must be optimized.
Information modeling is not to be considered as an independent stage of
design work. It must be carried out in the course of work on the draft in all
cases where the control system being developed is complex enough to warrant
a search for a priori rational design decisions.
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9
Approved For Release 2001/11/19 : CIA-RDP79-00798A000200020005-9