" I A?’"I 7 7 ~‘ 0."?

' lA‘<"u'.n\.\
“Aw ‘-’ { 2; Y gm
":‘g'; '4. Alia; _ fl «I . 3‘ “I?

J

n .,
"‘;\’Vt

“ QC‘x‘uK}

A
"
_;'._. 1 v— iii—{4%.
2:1,: AA “ix." ,N
’i ‘ at!"
...

Cm

.
W... 5"“: _. .\
. . :‘wnuaarze (\E
‘ (
:31; M V m: " "1‘ 'fifiai 3...;2353,‘
VD. ‘.

A , V, f . q: 4 . _ . _ '
775‘ W‘h‘ . 1" ‘ ‘ , ' . . ‘ T ‘ '1“ a
Mfr am. A» {it .f .A . Kr; f. ' a” "e?
‘ a " '55:“- ‘€
:1 $5

 

 

 

”25 I Q» (64 '7 ’
ll‘llllulilllllll':HUI“ Willi“ l “l l . ' “MARY"?
3 1293 00585 2565 r in.“
J "Diversity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled
Assessment of the Applicability of
"TRANSYT-7F" Optimization Model to
the Traffic Conditions in the Cities
of Al-Iéhobar and Dammam, Saudi Arabia.

presented by .

Nedal Taisir Ratrout

has been accepted towards fulfillment
of the requirements for

PhLD, degreein CiV1l Engr.

, / _/
16/144:an c7. /‘~’-’v/;’-/i

Major professor/I

Date W

MS U is an Aflirman've Action/Equal Opportunity Institution 0—12771

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/Equal Opportunity lnditutlon

ASSESSMENT OF THE APPLICABILITY OF "TRANSYT-7F"
OPTIMIZATION MODEL TO THE TRAFFIC CONDITIONS IN THE CITIES
OF AL-KHOBAR AND DAMMAM, SAUDI ARABIA.

BY
Nedal Taisir Ratrout

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Civil and Environmental Engineering

1989

L’ V ’1.- (NC) “v

ABSTRACT
ASSESSMENT OF THE APPLICABILITY OF "TRANSYT-7F"

OPTIMIZATION MODEL TO THE TRAFFIC CONDITIONS IN THE CITIES
OF AL-KHOBAR AND DAMMAM, SAUDI ARABIA.

BY

Nedal Taisir Ratrout

Several studies showed that traffic optimization produces
significant benefits in terms of reduced delay, vehicle stops,
and fuel consumption. With the increasing complexity of urban
networks, computer-based traffic algorithms are necessary to
obtain an optimal timing plan.

The main objective of this study was to select a can-
didate algorithm for optimizing the traffic in Saudi Arabia
and to calibrate the algorithm (model) to fit traffic condi-
tions in the cities of Dammam and Al-Khobar, Saudi Arabia.

An exhaustive research of the literature was conducted
to identify all network optimization models. The features of
these models were compared, and it was concluded that the
TRANSYT-7F model is the best candidate for application in
Saudi Arabia.

To use the model accurately, the saturation flow rates,
start-up lost time, extension of effective green, and average

vehicle spacing were evaluated in the study area. Their

values were found to be similar to those used in the United
States.

Calibration of the TRANSYT-7F model consists of deter-
mining that value of the platoon dispersion factor (PDF) which
reduces the discrepancies between the simulated and observed
flow profiles. The average best-fit PDF values were 28 and
40, respectively. The TRANSYT-7F manual suggests a value of
25 for low friction links and 35 for moderate friction links.
Both sets of values were used to conduct a sensitivity
analysis involving 4 parameters and 25 hypothetical networks.

The optimal timing plan using the calibrated PDF values
did not product significantly better results than optimization
runs using the values recommended in the manual. This led to
the conclusion that there is little value in developing a
calibrated set of PDF values for use in the cities of Dammam

and Al-Khobar.

Dedicated to my father and my mother whose blessing,
love, support, and continuous encouragement have been greatly

instrumental in the accomplishment of this feat.

iv

ACKNOWLEDGEMENTS

All praise and thanks are due to Allah, Lord of the
Universe, for His merciful divine direction throughout my
study.

I wish to express my deep appreciation and gratitude to
Professor William Taylor, my advisor and committee chairman,
for his valuable assistance and encouragement in the conduc-
tion and completion of this dissertation, and throughout my
entire doctoral program.

My appreciation and gratitude are also due to other
members of my guidance committee, Drs. Gokmen Ergun, Thomas
Maleck, and Professor V. Mandrekar, for their useful sugges-
tions and constructive comments.

Finally, thanks are due to King Fahd University of
Petroleum and Minerals for supporting this study and to the
Dammam traffic police department for facilitating the data-

collection phase of this study.

TABLE OF CONTENTS

Chapter Page
1. INTRODUCTION TO THE STUDY ........... ......... 1

1.1 Introduction ........ ..... ............ 1

1 O 2 The PrOblem O O O O O O O O ..... O O O O O O O O O O O O O 3

1 O 3 Obj ect ives O O O O O O O O O O O O O O O O O 0 O O O O O O O O O 6

2. LITERATURE EVIEW O0.0.0..........OOOOOOOOOOOO 8

.1 Advantages of traffic optimization ... 8

.2 Network optimization models ......... 10

.3 Reasons for selecting TRANSYT ......... 25

.4 Calibration of the TRANSYT program ... 28

2.4.1 British experience ........... 30
2.4.2 Experience in Canada and

the U.S.A. ......... ...... .... 32

2.4.3 Experience in other countries 38

2.5 Conclusions .......................... 50

3. THE TRANSYT-7FMODEL ......OOOOOOOOOCOOO0.0... 52

3.1 Model structure ............ ......... . 52
3.2 Input requirements ................... 54
3.3 Output characteristics ............... 55
3.4 Tasks performed by the model ......... 56
3.5 Limitation of the model .. ...... ...... 57
3.6 Computer requirements .. ............. . 59
4. DATA COLLECTION METHODOLOGY ........... ....... 61
4.1 Adaptation of the model .............. 61
4.2 Site selection .... ..... . ....... ...... 62
4.3 Data collection ........ ....... ....... 63
4.4 Schedule of activities ... ............ 67

vi

5. DATA ANALYSIS

5.1
5.2

5.3

01m
01¢

vii

Driver performance characteristics ...

Estimating PDF values in the

study area ........

Calibration of the model .............
5.3.1 Procedure .

5.3.2

Calibration results ..........

Validation of the model ..............

Sensitivity analysis

5.5.1 Procedure .

5.5.2
5.5.3
5.5.4

Conclusions

6. SUMMARY AND CONCLUSIONS ...

APPENDICES ......OOOOOOOOOOOOOOOOO.

A.

B.

C.

D.

E.

BIBLIOGRAPHY

NETWORK DIAGRAMS ....

SPEED DATA .........

Networks used in the analysis..
Sensitivity results ..........

LISTING OF THE FORTRAN PROGRAMS USED
IN THIS STUDY .....

GRAPHS OF

VS. PDF VALUES” .....

TIME-SPACE DIAGRAMS

”SUM OF ABSOLUTE DIFFERENCE

69
69
71
75
75
79
83
89
9O
94
96
113
115
122
122

125

127

136
148

152

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

2.1

2.2

2.3

2.4

2.5

4.1

5.1

5.2

5.3

5.4

5.5

5.6

LIST OF TABLES

Parameter values recommended in
NCHRP report #233 and in TRANSYT-7F
manual ......00.000000000000000.0.00

Characteristics of sites studied by
McCoy et a1 ........................

Best-fit parameter values along sites
studied by McCoy et a1 .............

Results of analysis of variance in
Axhausen and Korling study. Effect
of "source" on a ...................
Mean of a for the level of the

design factor in Axhausen and Korling
study ..............................

Summary of TRANSYT-7F data
requirements .............. .........

Average vehicle spacing and
extension of effective green in
the study area ............... ......

Start-up lost time and saturation
flow rate in the study area ........

Geometric features and traffic data
of the studied links .......... .....

Suggested roadway friction and PDF
values ......OOOOOOOOOOOOIOOOOO .....

Calibration results ..... ...... .....

Validation results ....... ..........

viii

34

36

37

47

47

65

70

7O

72

74

82

84

Table

Table

Table

Table

Table

Table

Table

Table

5.7

ix

Comparison between the flow profiles
obtained using the best, calibrated
and recommended PDF values ......... 86

Optimization of results of different
percentages of original length ..... 97

Simulation results of different
percentages of original length ..... 99

Improvement in performance measures
with different percentages of original
length ............................. 101

Improvement in performance measures
with different percentages of original
velume ....I.........OOCOOOOOOOOOOOO 108

Improvement in performance measures
with different levels of friction .. 110

Results of reversing the node order
in optimizing the hypothetical
arterial (original case) ...... ..... 112

Speeds in the study area ........... 125

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure
Figure
Figure
Figure

Figure

LIST OF FIGURES

Page

The structure of the TRANSYT
program

......O..........OO........ 53

A sample output of the program
"CALIB.' ......OOOOOOOOOOOO00.0.0.0. 78

Calibration of the link 9401 ....... 80

Observed and simulated flow
profiles of link 9401 ....... ...... 81

Sketch of the two-dimensional
network

..................... ...... 95

Improvement in total delay vs.
length of links .............. ..... 103

Improvement in uniform stops vs.
length of links ............. ...... 104

Improvement in PI vs. length of
links

................OOOOOOOOOOIOO 105

Sketch of King Abdul Aziz
Street

......O......O........OO.... 123

Sketch of First Street (Dammam) .... 124

Typical phase sequence in the
study area

........................ 125
Calibration of link 9101, Dammam .. 136
Calibration of link 9201, Dammam .. 137
Calibration of link 9301, Dammam .. 138
Calibration of link 9403, Dammam .. 139

Calibration of link 9503, Dammam .. 140

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

xi

Calibration of link 9603, Dammam ..

Calibration of link 9401,

Al-MObar ......0..000..0..........

Calibration of link 9201,

Al-MObar .0.......00........0.....

Calibration of link 9001,

Al-mObar .00....0..0.0...0

Calibration of link 9103,

A1-K11°bar .0........00...00..0....0

Calibration of link 9303,

Al-mObar ......0....0.0.0......000

Calibration of link 9503,

Al-mObar ..............0..........

Optimal time-space diagram for
the linear network ................

Optimal time-space diagram for

the two-dimensional network

141

142

143

144

145

146

147

148

150

CHAPTER 1

INTRODUCTION TO THE STUDY

In contrast to many developing countries, Saudi Arabia
is a rich country. It has been changing very rapidly from
being a pre-industrial society to a modern industrialized
country. This rapid growth exerted pressure on public
utilities in general, and on transportation facilities in
particular.

To cope with this rapid growth, the government has built
a huge (i.e. more than 35,000 KM) and modern network of roads
and highways throughout the country. In fact, at the present
time, Saudi Arabia possesses one of the best transportation
facilities in the Middle East" Following construction of this
road network, the attention is now focused on operating and
maintaining this network in the most efficient way. A large
portion of this modern network is concentrated in the large
cities, such as Riyadh, Jeddah, Dammam, and Al-Khobar.

Dammam city lies in the center of the Eastern province

of Saudi Arabia by the Arabian Gulf. It is the third largest

2

city in the country, and the largest in the Eastern province.
The city of Al-Khobar is less than twenty kilometers from
Dammam, and is considered to be the second.most important city
in the Eastern province. Generally speaking, both cities are
considered to constitute a single metropolitan area. Both
Dammam and Al-Khobar are close to the city of Dhahran, where
the third largest international airport in the country and.the
headquarters of the Arabian American oil Company (ARAMCO) are
located. They are also very close to King Fahed University
of Petroleum and Minerals (KFUPM), one of the finest univer-
sities in Saudi Arabia. The state of Bahrain, which has one
of the important financial and stock markets in the Middle
East, is less than fifty kilometers from either city. This
unique location of Dammam and Al-Khobar made them the major
commercial and economic activity center of the Eastern
province.

There are approximately ten major arterial streets and
fifty signalized intersections in each city. All signals are
fixed timed" .Neither of the two cities possesses.a comprehen-
sive plan for collecting traffic and operational data. Such
data, if it exists, is usually collected as a part of a
specific study or project and not.as>a part of a comprehensive

plan.

Winem-

One of the most important transportation problems that
is typically faced by traffic engineers is the optimization
of traffic flow in an urbanized area. The process of op-
timizing traffic flow in an urban network is achieved by
interconnecting and operating the signalized intersections to
minimize the delay and stops in the system. This minimization
of delay and stops in a network will provide a convenient
driving environment, improve the network capacity, and reduce
excess fuel consumption. As an example, in a comprehensive
optimization study conducted in the United States (8), it was
concluded that for the average intersection in the study,
"each year vehicle delay was reduced by 15,470 vehicle-hours,
455,921 vehicle stops were eliminated, and 10,524 gallons of
fuel were saved".

With the increasing complexity and magnitude of urban
signal networks, traffic optimization is almost an impossible
task to perform manually. Recently, several optimization and

simulation computer models have been introduced:

1. TRANSYT ...... ........... .... network optimization
2. SSTOP .. ..................... network optimization
3. SIGOP III ......... ... ...... . network optimization
4. SIGRID ......... ........... . network optimization
5. COMBINATION ...... ........... network optimization

6. PRIFRE ..... .......... ....... freeway optimization

4
7. PASSER II .................. arterial optimization
8. SOAP ................... intersection optimization
9. PASSER III ...... diamond interchange optimization

10. SUB ...................... arterial bus simulation

11. NETSIM ........................network simulation.

Computer-based models for network optimization all
contain an algorithm for computing signal offsets and splits
that minimize some combination of delay and stops in a
network. These models have made the procedure for optimizing
the signal timing plans in a given urban area relatively easy,
regardless of the complexity and the magnitude of the traffic
conditions in the given area. Moreover, the introduction of
such optimization models in a microcomputer version made these
models available and practical to be used in any country or
city in the world with very little investment and without the
need of highly experienced personnel.

In each of these models, there are a number of constants
that represent the driving habits and traffic conditions in
the country where the model was originally introduced and
calibrated. Typically, these constants include a number of
traffic characteristics, such as saturation flow, vehicle
spacing, queue dispersion, headway distribution, and driver
response to different signal phases.

It is well known that such traffic characteristics (i.e.
constants) can vary considerably from one society to another.

Because of such a possibility, any traffic signal timing

5
optimization model (and any model that deals with traffic
operation) has to be tested for its validity in the place
where it is proposed to be used. Consequently, some, or all,
of the model constants might have to be modified to match the
traffic conditions and driving habits in the country in which
it will be used.

In addition to these constants which might need some
modification, optimization models require considerable data
that describe the physical features of the network and the
traffic characteristics. These include intersection and road
geometry, traffic flow and volume, signal timing plans, speed,
and traffic composition.

In western countries such as the United States, this type
of data is easily and economically obtained using sophisti-
cated equipment and techniques. Moreover, most of the time,
such data is readily available, and often in a computerized
format. In contrast, in Saudi Arabia, the data collection
system is undeveloped, and frequently data of this nature does
not exist, or it is not updated regularly.

Therefore, to assess the applicability of any network
optimization model to the Saudi Arabian traffic conditions,
one first has to collect the required input data for that
particular model. The model must then be analyzed to deter-
mine which constants and internal relationships need to be
modified (if any) in order to calibrate the model for the

local driving habits and traffic environment of Saudi Arabia.

5
optimization model (and any model that deals with traffic
operation) has to be tested for its validity in the place
where it is proposed to be used. Consequently, some, or all,
of the model constants might have to be modified to match the
traffic conditions and driving habits in the country in which
it will be used.

In addition to these constants which might need some
modification, optimization models require considerable data
that describe the physical features of the network and the
traffic characteristics. These include intersection and road
geometry, traffic flow'and volume, signal timing plans, speed,
and traffic composition.

In western countries such as the United States, this type
of data is easily and economically obtained using sophisti-
cated equipment and techniques. Moreover, most of the time,
such data is readily available, and often in a computerized
format. In contrast, in Saudi Arabia, the data collection
system is undeveloped, and frequently data of this nature does
not exist, or it is not updated regularly.

Therefore, to assess the applicability of any network
optimization model to the Saudi Arabian traffic conditions,
one first has to collect the required input data for that
particular model. The model must then be analyzed to deter-
mine which constants and internal relationships need to be
modified (if any) in order to calibrate the model for the

local driving habits and traffic environment of Saudi Arabia.

6

This calibration effort will provide a better understanding
of the traffic characteristics in Saudi Arabia. This, in
turn, will provide some general clues about the applicability
of traffic models introduced in western countries (such as the
U.S.A.) to the traffic environment of Saudi Arabia. Further-
more, the study will result in a better understanding of the
virtues and deficiencies of the selected model. Finally this
study will contribute to improving and optimizing the traffic
environment in Saudi Arabia.

All network optimization models were reviewed (chapter
2) to select the best candidate for optimizing traffic in
Saudi Arabia. It was concluded that the TRANSYT-7F model is
the most appropriate model in this regard. The main reason
for selecting the TRANSYT-7F model is the fact that it is the
only one which was extensively and successfully used in many
countries under various traffic conditions and driving habits.
In addition to this, the TRANSYT-7F model has the ability to
handle many special traffic conditions, such.as more than four
phases in a cycle and sign controlled intersections. This
ability makes the model applicable to almost every network

configuration in Saudi Arabia.

Midi/£-

This study is designed to achieve the following main

objectives:

7
To identify the similarities and differences in
traffic characteristics and.driving habits on urban
networks between the United States and Saudi Arabia.
To adapt TRANSYT-7F for use in optimizing the
traffic in the study area.

To perform a parametric analysis on the TRANSYT-7F
model to determine which constants and internal
relationships (if any) need to be modified to
calibrate the model for the study area traffic
conditions.

To determine if the model calibrated to the local
traffic conditions provides better results than its
original calibration.

To assess the applicability of the TRANSYT-7F model

to the Saudi Arabian traffic conditions.

CHAPTER 2

LITERATURE REVIEW

2. vanta es of t f' o timizat'o .

Traffic optimization. is, by definition, the act of
developing an optimum signal timing plan (optimum signal
offsets and splits), by which the signalized intersections in
a given network are interconnected and operated to minimize
delay and stops. This minimization of delay and stops in the
network will provide a more convenient driving environment,
improve the network capacity, and reduce excess fuel consump-
tion (1, 2, 3, 4, 5, 6).

One of the most comprehensive and important examples of
the potential benefit of traffic optimization is the National
Signal Timing Optimization Project (8). In this project,
eleven cities in the United States optimized a portion of
their street network (ranging from 23 to 81 intersections per
city) using the TRANSYT-7F program. It was concluded (based
on TRANSYT estimates) that, for the average intersection in
the project, "each year 15,470 vehicle-hours of delay were

saved, 455,921 vehicle stops were eliminated, and 10,524

9
gallons of fuel were saved". It was also reported that
driving through urban areas became faster and easier as a.
result of implementing the optimized signal timing plans. It
was estimated that two million gallons of gasoline per day
could be conserved if the signal timing at most of the 240,000
signalized intersections in the United States were optimized.

Similar conclusions were also reported in a case study
of fuel efficient traffic signal operation in the city of
Garden Grove, California (9). Using the TRANSYT program
(version 8), improved traffic signal timing was developed for
a test network (70 intersections) in Garden Grove. The
following was concluded from the study:

The field test indicated that significant im-
provement. in ‘traffic flow* and. fuel consumption
result from the use of timing plans generated by
the TRANSYT optimization model. Changing from.pre-
existing to an optimized timing plan yields a
network wide 5 percent reduction in total travel
time, more than 10 percent reduction in both the
number of stops and stopped delay time, and 6
percent reduction in fuel consumption.

Rach (1) summarized. the results. of' an optimization
project carried out by the British Road Research Laboratory
using the TRANSYT program. He reported that, as a result of
that optimization project, the mean travel time was reduced
by 16 percent and the effective capacity was increased by 25
percent.

An improvement in the air quality is another possible

advantage of traffic optimization. Schlappi (10) found that

there was a relationship between vehicle stops and carbon

10
monoxide concentration. He reported that a ten percent
reduction in the number of vehicles stopping would result in
a five to seven percent reduction in the concentration of

carbon monoxide.

2.; Network optimization models.

Computers were first introduced as a possible tool for
studying and analyzing traffic problems in the 19505.
However, it was not until the late 19605 that computers and
computer programs were practical and widely used in analyzing,
designing and evaluating traffic facilities (11).

Prior to this, the timing (optimization) of a network
was done manually by either the volume priority method, or
the preferential street method (2). The procedure in these
two techniques consists of ranking all links in the network
in order of decreasing link volume or in order of decreasing
preference (i.e. importance) . Link offsets are timed to
insure good individual progression, starting with the link of
the highest rank and continuing in order of decreasing rank,
until reaching those links whose offsets are determined by
other previous settings.

However, because such manual techniques are cumbersome
and time consuming, it was more common to provide preferential
treatment (i.e. good individual progression) for only a small

number of arterial streets in the network. Clearly enough,

11
such manual procedures do not provide the optimal timing plan
for a network. Nevertheless, it was the only practical and
logical procedure for timing a network of signalized intersec-
tions without the help of computers.

One of the earliest models for network optimization is
the COMBINATION program developed in the U.K. and used by the
Greater London Council (1,12). The COMBINATION program is
based on the assumption that delay depends solely on the
offset difference between the signals at each end of the link
and not on any other signal in the network.

Basically, the program calculates the delay/difference-
of-offset relationship for each network link, and then
combines links in series or in parallel to obtain a set of
optimum offsets such that the network delay is minimized.
Consequently, the program does not include delay caused by
random fluctuation in the traffic. Also, it does not calcu-
late the optimum signal splits at the intersections in the
network under consideration. The signal splits at each
intersection have to be known (used as input) in order to use
the model. The program represents the earliest effort in
utilizing computers for network optimization. It was used a
number of times in the 19603, mainly in the U.K. (1). The
literature does not show any current utilization or modifica-
tion in the program.

Robertson (13) used the COMBINATION program for some

time, which in turn stimulated him to write the TRANSYT

12
(TRAffic Network StudY Tool) program in 1967. He wrote in
this regard (11):
TRANSYT grew from my chance to use the COM-

BINATION method and study its virtues and vices,

both of which center on the simplicity of the

platoon structure. I wrote TRANSYT program in 1967

using assembler language, and it was first tested

later that year on the Cromwell Road in London.

Since that time, Robertson and his colleagues (13) made
several major’ improvements on ‘the (original program, and
produced nine versions of the TRANSYT program (in addition to
the original program), with the latest version being
TRANSYT/Q, released in 1987. Based on the seventh version of
TRANSYT (TRANSYT/7), the Federal Highway Administration
produced an Americanized version of TRANSYT, referred to as
TRANSYT-7F. The TRANSYT program is a macroscopic, deter-
ministic optimization model. It is comprised of two main
sections; a traffic model and an optimization procedure.
The traffic model is a macroscopic, deterministic simulation
model. The term "macroscopic" refers to the fact that the
model considers platoons of vehicles (hereafter called
platoons) rather than individual vehicles. The simulation
process is based, primarily, on simulating the dispersion of
platoons as they progress along network links. This is done
by using a platoon dispersion algorithm developed by Robert-
son (13). The algorithm describes (collectively) the desire
of individual drivers to maintain comfortable time headway as

they progress along network links. It was found that the

algorithm (i.e. the comfortable headway) is a function of

13
roadway characteristics, as will be discussed later. Based
primarily on Webster's methodology (14), the traffic model
calculates delay and stops for each network link. Following
that, the weighted sum of the delay and number of stops
suffered by all vehicles in the network is obtained and.called
the "performance index," or "PI" of the network.

The optimization procedure is an iterative, gradient
search (hill-climbing) technique that optimizes signal phase
lengths (i.e. splits) and offsets of a signalized network.
The first step in the optimization process is to determine
the performance index (PI) of the original signal timing'plan.
This is done by the traffic model, discussed previously. For
offset optimization, the offset of the first signal in the
network (as input by the user) is increased by a pre-specified
amount. The traffic model is then called to recalculate the
new PI. If the new PI is less than the previous value, the
TRANSYT program continues to increase the offset by the same
amount as long as the PI continues to decrease. On the other
hand, if the new PI is greater than the previous value, the
program will decrease the offset by the same amount and
continue to decrease the offset (by the same amount) as long
as the PI continues to decrease. The optimum offset of this
signal is achieved when no further improvement can be made
(i.e. PI can not be decreased) by varying its offset. The
same procedure is repeated for every signal in the network

under consideration. The phase length optimization process

14
is similar to the offset optimization process discussed above.
Finally, the optimum signal timing plan (i.e. optimum splits
and offsets for all signals) is reported.

The TRANSYT model can be used as a simulation and as an
optimization tool for arterial roads as well as urban net-
works. A more detailed discussion of the TRANSYT-7F program
is presented in chapter three of this study.

Practically speaking, all researchers involved in traffic
operation agree that TRANSYT has been widely and successfully
applied throughout. Europe, the ‘United. States, and. other
countries (8, 13, 15, 17, 18, 19, 20, 24, 25, 26, 27, 28, 30,
38, 39, 4o, 41, 43, 44, 45, 46, 47, 48, 49). For example,
Rouphail (15) reported that TRANSYT "has been successfully
applied at many intersections in Europe and the United
States." Cohen and Liu (16) also wrote in this regard:

The TRANSYT model is the most widely used
computer program for developing signal-timing plans

for urban signal systems. An Americanized version

of the program, TRANSYT-7F was developed for use in

the United States and has been successful.

Rach (1) described and evaluated the TRANSYT program,
together with other network optimization models. He also
summarized the results of one of the earliest applications of
TRANSYT. Rach wrote: “

Overall,TRANSYT has been demonstrated to be
reliable and effective both as a design and as an
evaluation tool. In a study carried out by the

Road Research Laboratory, it was found that TRANSYT

accurately predicted network delay. Also, the

hill-climbing process in TRANSYT was found to be

very effective in obtaining optimum offsets. When
compared to existing signal timing settings, the

15

TRANSYT settings reduced mean travel time by 16%
and increased effective network capacity by 25%.

As a part of the National Signal Timing Project (8),
TRANSYT was used by eleven cities in the United States
representing' a ‘wide range. of‘ geographical locations and
traffic characteristics. The cities optimized the signal
timing in a portion of their street network (on average 46
intersections per city) and evaluated the effectiveness of
the optimized signal timing plans. It was found that these
plans provided significant reductions in vehicle delay,
vehicle stops, and fuel consumption. Therefore, it was
concluded that, "TRANSYT-7F‘is a very valuable tool for signal
timing optimization projects."

Wallace (17) critically reviewed the TRANSYT program.
He concluded that:

TRANSYT-7F is a major new tool available to
traffic engineers for analysis of traffic signal
system, evaluation of alternative control strate-
gies and design of optimal signal setting.

Currently, TRANSYT-7F is in use by over 400 cities,
states and consultants throughout the United States (18) .
For example, TRANSYT-7F was recently used in North Carolina
as a part of its management program for energy conservation
(19). TRANSYT-7F is the only network optimization model that
is reported in both the Software and Source Book (18) and the
Handbook of Computer Models for Traffic Operation Analysis

(3). In fact, the handbook (3) described TRANSYT-7F as being

"one of the most widely used design models."

16

Although TRANSYT was originally developed as an op-
timization program, its "realistic" traffic simulation model
(20) makes it a valuable candidate for traffic evaluation.
In fact, Yagar and Case (21) recommended that "TRANSYT be
seriously considered for any evaluation purpose to which it
is applicable." McCoy et al (20) described the traffic
simulation sub-model in the TRANSYT program as one of the
"most realistic in the family of macroscopic computerized
traffic simulation models." Dudeck et al (22) reported that
NETSIM (state-of-the-art microscopic model for traffic
simulation) and the TRANSYT-7F model produce ”compatible
estimates of travel although the differences are at times
appreciable."

The TRANSYT model was also used effectively as an
arterial optimization model (3, 16, 18, 23, 26, 29, 31). For
example, Skabardonis and May (23) used MAXBAND, PASSER-II
(state-of-the-art models for arterial optimization), and
TRANSYT-7F in optimizing an 11-signal arterial. It was found
that, in terms of traffic performance, no model was capable
of producing signal timing plans that are superior to those
generated by the TRANSYT-7F model.

Another early model for network optimization is SIGRID
(SIgnal GRID design) developed by Traffic Research Corporation
for the Toronto traffic computer control system in 1964 (1).
The program is a time-volume geometry method of optimizing

offsets in a network. The program does not optimize the

17

individual signal splits and link offsets. The program user
has to predetermine the optimum signal splits and link.offsets
(i.e. optimum offset difference for each successive pair of
intersections) by another program, or simply from experience.
Having’ these ‘values for' each. network link, the program
minimizes the discrepancy between the optimum offsets and the
actual ones. Therefore, the program does not necessarily
minimize the system delay. It should be noticed that the
program uses oversimplified assumptions in calculating average
waiting times, and hence, these times do not necessarily
reflect the actual delay characteristics. The literature does
not show any modification or current application of the
programs. Generally speaking, the SIGRID program is unsophis-
ticated and obsolete.

Another major breakthrough in the field of signal network
optimization was the development of the SIGOP (SIGnal Optimi-
zation Program) model. Originally, SIGOP was developed by
Peat, Marwick, Livingston and Company, for the U.S. Bureau of
Public Roads (1). Basically, the original SIGOP model was
nothing more than an extended version of the above SIGRID
program. However, later versions of SIGOP were modified and
improved substantially. The latest version of the model
(SIGOP-III) was developed by KLD Associates, Inc. for the
office of Research, Federal Highway Administration (32).

Generally speaking, both SIGOP—III and the previously

mentioned TRANSYT-7F are similar optimization programs that

18

«can be used for optimizing arterial roads and grid networks
<3f urban streets. Both programs are macroscopic models that
«can be used for design and evaluation purposes. Each model
(consists of two main parts: a traffic flow algorithm and an
toptimization sub-model. The major difference between SIGOP-
III and TRANSYT-7F is in the structure of the objective
function which is used as an optimization criterion. The
tobjective functions of both models are expressed directly in
terms of vehicle delay and vehicle stops. However, the
objective function of SIGOP-III has a third term reflecting
excess queue length relative to available storage capacity.
Furthermore, unlike TRANSYT, which allows all splits to vary
(subject to a minimum green constraint) in order to achieve
the lowest value of the objective function, SIGOP-III calcu-
lates minimum green requirements using Webster's method (3).

The TRANSYT traffic sub-model provided the basis for much
of the SIGOP-III traffic sub-model. The platoon dispersion
technique used in TRANSYT (Robertson platoon dispersion
algorithm) is used indirectly in.SIGOP-III. Delay and queuing
calculations in both models are based on Webster's methodol-
ogy. Unlike TRANSYT, the SIGOP-III program suffers from two
major limitations. First, the program does not explicitly
deal with minor intersections (i.e. controlled by stop or
yield signs) and, secondly, it can not be used for intersec-
tions having more than four phases in a cycle. Furthermore,

links longer than one mile are not accepted by the program.

19

Generally speaking, the SIGOP-III model seems to receive
little attention from researchers and traffic engineers. The
literature does not indicate any significant application or
validation of SIGOP-III.

In the early 19703, Datta.et al (34, 35) developed TRASOM
(TRAffic Signal Optimization Model) as a part of a "Traffic
Signal Optimization Project" for the Oakland County Road
Commission, Pontiac, Michigan. The model is basically a
linear (road) optimization program that utilizes a sequential
optimization process similar to the preferential street
method, discussed previously. In a given network, TRASOM
determines an optimal linear solution (the best progressive
system) for every roadway in the network separately. Unlike
the preferential street method, TRASOM utilizes the speed-
volume relationship of the road under consideration as a
constraint in determining the optimal progressive speed on
that particular road. After determining the optimal linear
solution for all the roadways constituting the network, the
model fits in the intersecting nodal offsets according to a
pre-specified sequential strategy, in a manner similar to the
one used in the preferential street method. This sequential
strategy is established by rank ordering each road in the
network. on some jpriority system, such as importance of
direction of travel, maximum critical demand volume, and so
on. The assigned priorities do not change the cycle splits

on any road in the network. However, since the optimal

20

offsets on all roads might not be attainable, these assigned
priorities establish. the sequence 'used. in. determining a
feasible network solution. Road offsets are timed to provide
the optimal individual progression, starting with the road of
the highest rank, and continuing in order of decreasing rank,
until the model reaches some road whose offsets (one or more)
are already determined by previous setting. Consequently,
this model does not provide a mathematically guaranteed
optimum network solution. However, Datta et al reported that
the model optimization process provided network solutions
which were feasible, practical, and near optimal. The model
was also efficient in terms of its computer-time requirement.
The model has been used by the authors in designing timing
patterns for 200 traffic signals covering an area of 36 square
miles in southeast Oakland County, Michigan. The TRASOM
program was fully developed using private funds and has been
used only by the authors in designing signals. The model is
obsolete when compared to present programs such as TRANSYT-7F
and SIGOP-III.

A relatively new traffic signal optimization program is
the SSTOP (Signal SysTem Optimization Program) developed in
the late 1970s (34, 35) . The program was originally developed
as an on-line traffic signal optimization program for Metro-
politan Toronto. The program was then converted to an off-
line version (optimization program for fixed-time signal

systems) by the Traffic Research Group at McMaster University

21
under contract to Transport Canada. The program is a macro-
scopic, deterministic network optimization model.

The first operation of SSTOP is to read in, check and
sort the input data for obvious errors. Once this step is
completed, the program starts the optimization process by
determining the practical minimum and optimum cycle lengths
for each individual intersection based on Webster's cycle
length procedure. The practical minimum cycle length is the
best cycle length as determined by Webster's procedure taking
into consideration the minimum green time necessary to provide
adequate capacity for each phase, and satisfying pedestrian
walk time requirements. The practical optimum cycle length,
on the other hand, is the minimum cycle length which allows
for an additional 10 percent capacity to cope more efficiently
with random traffic fluctuations.

The next step in the optimization process is to select
the best cycle length for the system (single cycle length for
all intersections) from pre-specified cycle lengths. For
overall network control, the program accepts up to ten
specified candidate system (network-wide) cycle lengths. For
each candidate cycle length, intersections which can not be
placed under coordinate control in the network are identified.
This is done by comparing the candidate cycle length with the
practical minimum cycle length of each intersection in the
network. Whenever the candidate system cycle length is less

than the required individual minimum cycle length (i.e.

22

practical minimum) at an intersection, that intersection is
placed under isolated control. The signals at the isolated
intersections are allowed to operate at their optimal cycle
length (practical optimum cycle length), thereby avoiding the
possibility of over saturated signals in the network. Once
these intersections are identified and isolated from the
system, the rest of the intersections are placed under
coordinated control.

Following this step, signal splits for both coordinated
and isolated signals are calculated, based on Webster's method
by setting the green times proportional to their respective
volume/saturation ratios. For every link in the coordinated
network, link entry flow patterns are generated using signal
and volume data from the upstream signal. Using Robertson's
platoon dispersion model (used also in TRANSYT), link exit
flow patterns are then formed at the downstream signal. For
every offset difference between the upstream and downstream
signals, a delay (i.e. uniform delay) and stop value is
calculated from.thetarrival and.departurejpatterns at the link
exit signal. This computed value is based on the logic
adapted from the COMBINATION program (previously mentioned),
and hence, it does not include delay caused by random fluc-
tuation in the traffic. Therefore, a random delay component
is added to the above delay value. This delay component, on

a given link, is based on its degree of saturation.

23

A performance index is then calculated for each given
offset difference between the upstream and downstream signal.
This performance index is obtained by the addition of the
uniform delay, random delay, and number of stops resulting
from any particular offset difference. The optimum offset
difference for the pair of signals under consideration is the
one which produces the minimum performance index.

Delay-offset and stop-offset relationships (curves) are
established for every link in the coordinated network. The
optimum offset for some links might not be achievable when
the links are considered, collectively, as a whole network.
This is because at least one of the two intersections, at both
ends of a link, might be a part of one or more links whose
offset have to also be taken into consideration. Therefore,
a technique adapted from the SIGRID program (previously
discussed) is used to calculate the optimum offsets for the
coordinated network. The SIGRID program will determine
network offsets that are as close as possible to the optimal
offsets of individual links. determined. previously' (those
resulting in the minimum performance index of each individual
link).

Delay and stops for these coordinated links whose offsets
are varied from the optimal ones are reobtained from the
delay-offset and stop-offset curves previously calculated.
The performance indices of isolated signals are also deter-

mined (in terms of delay and stops) using the technique

24

developed by Webster. Finally, the overall system performance
index is obtained by adding the performance indices of the
isolated and coordinated signals. The overall performance
index of each candidate system.cycle length is obtained in the
manner described above. The system cycle length which
produces the minimum overall system performance index is
considered the optimum cycle length.

The jprogram ‘was first tested for effectiveness and
validity in 1979. Three field demonstration projects were
conducted in three Canadian cities (34) . The program was
tested again and compared to the TRANSYT/7 program in 1980.
This study was conducted in the downtown area of Galt in the
city of Cambridge, Ontario (34). The signal timings.generated
by SSTOP were similar to those obtained from TRANSYT/7. Lam,
et al (34), commented on the results of these tests as
follows:

Generally, the results of SSTOP vs. TRANSYT
comparison indicate that the signal timings gener-

ated by SSTOP compare favorably with those gener-

ated by TRANSYT. Input preparation for SSTOP is

easier and faster than for TRANSYT. Computer

requirements and running costs are much lower for

SSTOP.

However, they also commented on the features which the
model lacks. Lam, et a1, wrote in this regard the following:

Various theoretical refinements are possible

to increase the flexibility of SSTOP. A.brief list

of possible additional features is:

- To provide for a network-wide lost time
parameter

- To provide the ability to input link speci-
fic stop penalty factor

- To be able to include a double cycle length
in the coordinated network

25
- To provide special treatment of over and
under saturated intersections
- To be able to force congested intersections
to remain in a coordinated network
- To be able to coordinate networks based on
user input signal splits.

The literature does not show any other (i.e. other than
what was mentioned previously) reported study on the applica-
tion or validation of the SSTOPjprogram. It seems that Canada
is the only place where the program was ever tried. Based on
a private communication with the developers, it is understood
that SSTOP has an "error" in the way it handles left-turns,
and that this error will not be corrected because it will be
"expensive" to do so (50).

It is clear from the discussion in this section that
TRANSYT, SIGOP-III, and SSTOP are the state-of-the-art
computer models for network optimization. It is also clear

that TRANSYT is superior and more appropriate in fulfilling

the objectives of this study.

2. easo s o selectin TRANSYT.

One of main reasons for selecting the TRANSYT model for
this particular study is the fact that it has been widely and
successfully used in many countries. In the previous section,
a number of examples were cited. Furthermore, TRANSYT is the
only optimization program which has been tested and calibrated
in several countries. Several validation and calibration

studies.of‘TRANSYT were conducted in the U.K. (where the model

26

was originally developed), Canada, Australia, Germany, Sweden,
and other countries. This is in addition to the major effort
performed by the Federal Highway Administration in developing
the Americanized version of the TRANSYT/7 program
(TRANSYT-7F), and then in testing it in eleven cities as a
part of the jpreviously' discussed. National Signal Timing
Optimization Project (8) . The experiences of all these
countries in the calibration and application of TRANSYT proved
that the model is transferable under various traffic condi-
tions and driving habits. These experiences will aid in
achieving the objectives of this study.

In addition to this, TRANSYT (in contrast to SIGOPIII)
has the ability to handle special traffic conditions, such as
up to seven phases in a cycle and sign-controlled intersec-
tions. This ability makes the TRANSYT program suitable in
Saudi Arabia, where such traffic conditions are not uncommon.
Furthermore, the fact that TRANSYT (in contrast to SSTOP) can
be used as a design and evaluation tool for arterial roads
makes it.a practical model to be transferred to Saudi Arabia.

In summary, the reasons“ for selecting TRANSYT are as
follows:

1. TRANSYT is the only network optimization model which

has been subjected to validation and calibration
studies in several countries under various traffic

conditions and driving habits.

27
TRANSYT is the only model which was extensively and
successfully used in practice.
TRANSYT is an optimization as well as a simulation
model. This feature makes TRANSYT easy to validate
and calibrate in any country.
TRANSYT can be used as a design and evaluation tool
for arterial roads as well as urban networks.
TRANSYT is available in a microcomputer version
which makes it available and practical to be used
in any city or country in the world with very little
investment.
The objective function (i.e. the function that will
be minimized) in TRANSYT is very flexible. This
function is a linear combination of delay and stops,
in which the user can express the importance of
stops relative to delay for each link in the network
according to his objectives and convenience.
TRANSYT has the ability to handle the following
special situations

a. Sign-controlled intersections.

b. Up to seven phases in a cycle.

c. The use of double cycles for major
intersections.

d. The use of half cycles for minor inter-
sections.

e. Grouped nodes.

28
f. Mid-block sources.
9. Multiple links at common stop line.
h. Bus operations.
i. Multiple greens for a movement.
j. 100% green operation.
k. Bottlenecks.

Other network optimization models (i.e. SIGOP-III and
SSTOP) can not be used in most of the above situations.

The Americanized version of TRANSYT (i.e. TRANSYT-7F) is
used in this study because it is available free of charge and
without a licence. Therefore, transferring the model to Saudi
Arabia will not introduce any legal problems. The modifica-
tions introduced in later versions (British versions 8 and 9)
do not affect the objectives of this study. This is because
the traffic model (platoon dispersion algorithm) which was
used in the TRANSYT-7F program is also used in versions 8 and
9. Consequently, the results that will be obtained from
calibrating TRANSYT-7F (proper value of the parameters that
describe drivers performance characteristics and habits) can

be applied to any later version of the TRANSYT program.

2.4 gaiibzatiog of the TRANSYT program.

The reliability and effectiveness of the TRANSYT model
in simulating and optimizing the traffic flow in a given

network depends, primarily, on the ability of its platoon

29
dispersion algorithm to accurately predict the flow pattern
from one signal to another. The platoon dispersion algorithm
used in TRANSYT is considered to be one of the most realistic
algorithms used in macroscopic traffic simulation models (20) .
It is based on the theory that a platoon of vehicles starting
from an upstream intersection will continuously disperse as
it travels downstream along the link. Robertson (13) devel-

oped the following recurrence relationship to simulate this

phenomenon:

Q1(i+Bt) = F*Q(i) + {(l-F)*Q ‘(iwt-ln (1)
where,

<N(i+flt) = number of predicted arrivals in interval i+fit at

a point downstream of a signalized intersection,

number of departures in interval i from the
signalized intersection,

Q(i)

fl = empirical travel time factor expressed as ratio
between average travel time of leading vehicle in
platoon and average travel time of entire platoon,

t = average travel time from the signalized intersec-
tion to the point at which the platoon is being
calculated,

F = empirical smoothing factor, which controls rate at
which platoon disperses, expressed as
F - l/(l + afit),

a = empirical dispersion factor.

It is clear that the amount of dispersion in the traffic
flow pattern, as predicted by the above recurrence equation,
depends on the values of a and B. Generally speaking, a large
dispersion (which is usually caused by a high level of

friction to traffic flow) is represented by high value of a.

30
A value of 1.0 for 3 represents a situation in which platoons
do not disperse (average travel time of leading vehicle in
platoon and average travel time of entire platoon are equal).
On the other hand, a value approaching zero for 3 represents
a large dispersion. Theoretically, the parameters a and )6 can
take any value between zero and one. Consequently, the
successful utilization of the TRANSYT model depends, to a
large extent, on the selection of values for a and.fi that.best
represent and replicate the traffic flow pattern in a given

network.

2,5,1 Bririsp Experience.

Robertson.(13) studied 700 platoons at four sites innwest
London, England, and concluded that values of 0.5 and 0.8 for
a and B, respectively, produce the best agreement between
actual and computed platoon dispersion. Although the sites
selected by Robertson covered a wide range of traffic and
roadway conditions, he indicated that the appropriate values
of a and 3 might be a function of roadway characteristics such
as roadway width, slope, traffic composition, crossing
pedestrians and parking activities. The fact that such a
relationship exists between the parameters (i.e. a and B) and
the roadway' characteristics ‘was reported. by several re-
searchers in the United States, Europe, Canada, Australia, and
other places. McCoy et a1 (20) summarized results of three

major studies on platoon dispersion conducted in England.

31
These studies are discussed (based on McCoy documentation) in
the following three paragraphs.

Collins and Gower (38) studied the dispersion of platoons
of passenger cars in the suburbs of London, England. Data
were collected along three-lane dual carriage-ways (high-type
arterial road). The values of a and 5 that produced the best
agreement between the predicted (by Robertson platoon disper-
sion model) and the observed flow patterns were then found to
be 0.20 and 0.80, respectively.

In Sheffield, England, El-Reedy and Ashworth (39)
examined dispersion of platoons of vehicles along a 33ft wide
single carriage-way. The road was subjected to a 30 mph speed
limit, and had a bus volume of 12 per hour. Data were
collected at three different stations on three different days.
The stations were on a 5 percent downgrade at distances of
1082, 1378, and 1837ft downstream from a signal. At the
second station (i.e. at 1378ft from the signal), the best
agreement between the observed flow patterns and those
predicted by the platoon dispersion model was achieved when
a and B were set to be 0.6 and 0.63, respectively. At the
third station (i.e. at 1837ft from the signal), the best
agreement was observed when values of 0.70 and 0.59 were
selected for the parameters a and 6, respectively.

Similar studies were conducted by Sneddon (40) in
Manchester, England. The data were collected on two different

sites. The first site was a three lane dual carriage-way with

32
10 to 15 percent commercial vehicles in the peak hours, and
reasonable freedom for over taking; 'The other site was.a two-
way road 35ft wide with 2 to 3 percent commercial vehicles,
two narrow lanes in the direction studied, and severely
restricted over taking. With the parameter 6 being fixed at
0.8, it was found that the values of a that provided the best
fit between the observed and predicted platoon dispersion were

0.4 and 0.63 for the first and second site, respectively.

2.5,; Erperieppe in Canada and rpe U,§.A.

In Toronto, Canada, Lam (41) conducted platoon dispersion
studies on Leslie street (a high type four-lane two-way
suburban arterial with no driveways and with left turn bays)
in order to evaluate the applicability of the parameter values
suggested by Robertson (i.e. a = 0.5 and B = 0.8). Using
these suggested values for the parameters, Lam applied the
Robertson platoon dispersion model to six roadway segments on
Leslie street for three times a day. He found that the
average error in the computation of delay was 13.8 percent.
Lam then calibrated the platoon dispersion model using his
observed traffic flow date, and found that an a of 0.24 and
B of 0.8 provided the best fit between observed and predicted
platoon dispersion for the conditions existing on Leslie
street. The calibrated.model (i.e. a = 0.24 and fi = 0.8) was
capable of reducing the error of the delay estimate to 8.2

percent.

33

To validate the accuracy of the platoon dispersion model
for traffic conditions found in the U.S.A., Tarnoff and
Parsonson (24) collected traffic flow data on Route 7 east of
its intersections with. Towlston IRoad in. Fairfax County,
Virginia. Data (i.e. time of arrival of every vehicle) were
collected at 100, 400, and 800ft from the intersection. Since
the data collected closely matched the free flow suburban
arterial case evaluated by Lam, Tarnoff and Parsonson decided
to use the parameter values determined by Lam in their study.
Consequently with a and B being equal to 0.24 and 0.8,
respectively, the Robertson dispersion model was applied to
the data collected at the 100ft station to predict the platoon
flow patterns at the 400 and 800ft stations. At both the 400
and 800ft stations, close agreement was obtained between the
observed and predicted flow patterns. Enrthermore, better
agreement was obtained during peak periods, where heavier
traffic volumes are experienced. This was explained by the
fact that the heavier traffic volumes in peak periods reduced
the variability of the data, and thus the accuracy and
reliability of the predictions were increased. Although it
was acknowledged that additional research should be conducted
to refine the relationship between the parameters a and B and
the roadway conditions, some general recommendation regarding
the relationship between these two parameters and the roadway
environment were deve10ped. These recommendations are shown

in the first two columns of Table 2.1.

34

TABLE 2.1 Parameter values recommended in NCHRP' report #
233 and in TRANSYT-7F manual.

 

 

 

 

 

 

* TRANSYT-7F
NCHRP 233 Manual
Roadway
a B a 8 Characteristics
0.50 0.8 0.50 0.8 Heavy friction]
0.37 0.8 0.35 0.8 Moderate friction2
0.24 0.8 0.25 0.8 Low friction3 I

 

(Source:References 7, and 24)
Tarnoff and Parsonson (Reference 25)

1 Combination of parking, moderate to heavy turns,moderate

to heavy pedestrian traffic, narrow lane width: traffic flow
typical of urban CBD.

2 Light turning traffic, light pedestrian traffic, 11- to
12-ft lanes, possibly divided: typical of well designed CBD
arterial.

3 No parking, divided, turning provisions, 12-ft lane width;
suburban high-type arterial.

Based on the study of Tarnoff and Parsonson (24) and
preliminary work performed by the University of Florida,
parameter values that are almost identical to those recom-
mended by Tarnoff and Parsonson were suggested in the user's
manual for TRANSYT-7F (7). These values are shown in columns
3 and 4 of Table 2.1. However, the manual indicates that
these parameter 'values are only' a general guidance for
TRANSYT-7F users that are based primarily on the Tarnoff and
Parsonson work. Consequently, one should not consider such

agreement between the parameter values of the TRANSYT-7F

manual and those suggested.by Tarnoff and Parsonson (see'Table

35
2.1) as a validation of the relationship shown between the
roadway conditions and the parameter values.

It should be noted that the recommended values for the
parameters a and ,6 in the TRANSYT documentation for the
original program (13) and in ‘versions through, to, and
including TRANSYT/7 is 0.5 and 0.8, respectively. However,
the Transport and Road Research Laboratory (TRRL) has changed
these values in TRANSYT/8 (42) to 0.35 and 0.8 for a and 6,
respectively. It is interesting to note that these values
(i.e. a = 0.35 and fl = 0.8) are also used as a default values
in TRANSYT-7F which is an Americanized version of the British
TRANSYT/7.

It is essential to understand that, in all TRANSYT
versions, the user is only allowed to select the value of a
in accordance with the roadway characteristics. All TRANSYT
versions do not allow the user to change the default value of
the parameter 8 which is kept as a constant value of 0.8.

As a part of the National Signal Timing Project (8),
TRANSYT-7F was calibrated and used by eleven cities in the
United States. The evaluation report of the project (8)
indicated that TRANSYT-7F was calibrated by adjusting some of
its input parameters, such as the platoon dispersion factors.
However, more explicit information about this calibration
process was not provided.

McCoy et al (20) studied the dispersion of 1700 platoons

of passenger cars under low friction traffic conditions in

36
Lincoln, Nebraska. The study was conducted on six arterial
street segments downstream of signalized intersections. All
of the selected sites were level, tangent sections without
parking. Other attributes of the studied sites are shown in

Table 2.2.

TABLE 2.2 Characteristics of sites studied by McCoy et al.

 

 

Lanes Observed Speed Peak
Width Driveway Limit Period
Site No. (ft) Access (mph) Studied
1: 1 13 None 35 p.m.
2“ 1 13 Limited 35 p.m.
3“ 2 12 None 45 a.m.
4” 2 12 None 45 a.m.
5M 2 13 None 45 a.m.
6 2 12 Limited 45 a.m.

 

 

 

(Source:Reference 20)

*

Two-way two-lane arterial street.
Four-lane divided arterial street.

0.

At each study site, data were collected (time of arrival
for every vehicle) at four stations downstream from the
signalized intersection. The first station was located
immediately downstream from the intersection. The other three
stations were located at 300, 600, and 1000ft downstream from
the first station. The Robertson dispersion model was then
applied to the data (i.e. platoon flow patterns) collected at
the first station to predict the platoon flow patterns at the
other three stations. Flow patterns were predicted for

several combinations of a and B. The values of a and B were

37

varied in increments of 0.01 over the range of 0.00-1.00 and
0.5-1.00, respectively. The pair (combination) of a and B
values that produced the best agreement between the observed
(actual) and predicted platoon flow patterns at the three
downstream stations (i.e. at 300, 600, and 1000ft) was
selected as the best value of a and B for the study site. In
a similar manner, the best-fit value of a with B being equal
to 0.8 was determined for each study site. This was done to
provide a basis of comparison with other platoon dispersion
studies and because the default value of B (B = 0.8) in the
TRANSYT program is not under control of the user; The results
of the study are summarized in Table 2.3.

TABLE 2.3 Best-fit parameter values along sites studied by
McCoy et al.

 

 

Best-Fit Value Range of No. of
Parameter Value of a for Platoon Platoons

Site a B B = 0.8 Size
1: 0.22 0.99 0.51 5-15 294
2" 0.20 0.96 0.35 5-20 319
3N 0.16 0.95 0.38 5-38 309
4“ 0.13 0.97 0.35 5-23 303
5“, 0.14 0.99 0.36 5-23 286
6 0.16 0.96 0.38 5-15 180

 

 

 

(Source:Reference 20)

* Two-way two-lane arterial street.
Four-lane divided arterial street.
The Kolmogorov-Smirnov (K-S) test was used to evaluate
the goodness of fit of the actual (observed) flow patterns
with those predicted by the calibrated platoon dispersion

model (i.e. using best-fit values of a and B in the model).

38

Similarly, the test was also applied to flow patterns pre-
dicted by the platoon dispersion model with the best-fit
values of a for B equal to 0.8 (column 4 in Table 2.3). In
both cases, the K-S test showed that the observed flow
patterns fit those predicted by the calibrated platoon
dispersion model at the 10 percent significance level. It
was also found that larger platoons dispersed slightly more
than smaller ones. It was concluded that the appropriate
values (average results of the study) of the parameters a and
B for passenger-car platoons under low friction traffic flow
conditions on urban arterial streets are 0 equal to 0.21 and
B equal to 0.97 on two-way two-lane streets, and a equal to
0.15 and B equal to 0.97 on four-lane divided streets.
Consequently, it was stated that the user of the TRANSYT
program should be able to specify the proper values of B as
well as a.
2 4 e 'e e ' ot nt ' s

In the State of Kuwait (on the northern border of Saudi
Arabia), Castle and Bonniville (43) conducted platoon dis-
persion studies along a number of high standard arterial roads
in the fringe area of the city of Kuwait and on one street in
its central business district (CBD). The distances between
signalized intersections along most of the arterial roads were
in the range of 1000 to 2000m (i.e. 3300 - 6600ft). Data

were collected at three stations on each study segment. The

39

first station was located just beyond the upstream intersec-
tion. The other two stations were located downstream from the
intersection at varying distances from the first station. The
maximum distance between any two stations on any link was 2.2
km. The Robertson dispersion model was then applied to the
data collected at the first station to predict the flow
pattern at the two downstream stations. This was done using
a value of 0.8 for the parameter B and five different values
for the dispersion factor a. The values of the parameter a
investigated were 0.5, 0.4, 0.35, 0.30, and 0.20. The value
of a that produced the best agreement between the predicted
and the actual flow patterns at the two downstream stations
was then determined.

It was found that a value of 0.5 for the parameter a,
provided the best fit between the observed and predicted flow
patterns in the central business district“ On the other'hand,
a value ranging between 0.3 and 0.4 for a was fOund to be
appropriate in predicting platoon dispersion on most of the
high standard arterial roads. On only one of the arterial
roads, it was found that an a value of 0.5 (i.e. as in the CBD
area) was appropriate in predicting platoon dispersion. It
was thought that the extremely high speeds along that par-
ticular arterial road (averaging 56 mph) might contribute in
suppressing the free flowing conditions, which are usually
conducive to a high degree of platoon cohesion (i.e. low value

of a). It should be mentioned that the discussion of Castle

40
and Bonniville implied that, for the same traffic conditions,
the value of a for a particular link is more affected by the
average speed than by the distance between successive inter-
sections (i.e. link length). In fact, it was indicated that
platoons were observed to remain grouped and did not reach
random flow over distances of up to 2000m.

To examine the sensitivity of timing plans to different
values of a, data were prepared by Castle and Bonniville for
a small network in the study area (five intersections in the
fringe area of Kuwait) and three optimized timing plans were
calculated (by TRANSYT) using three values for the parameter
a (0.5, 0.4, and 0.3). The performance indices (total delay
in vehicle hours per hour plus weighted number of vehicle
stops) for these three timing plans (PISO, PI,0 , and P130) were
then determined.

A series of six: TRANSYT simulation runs 'were then
performed in which the program merely simulated the effect in
terms of delay and number of stops which would result from a
specific set of signal splits and offsets. In any single run,
signal timing (splits and offsets) were fixed at the optimized
values found earlier with a specific a value (0.5, 0.4, or
0.3). However, a different value of a was used in the platoon
dispersion model. For example, in the first simulation run,
signal timing was fixed at the optimal values found with a =
0.5 but an a of 0.4 was used in the platoon dispersion model.

The value of the performance index calculated in this

41

simulation run was denoted as PIsomo' The value of the
performance index determined by optimizing the network with
a being equal to 0.4, was denoted as P140. A comparison
between the two indices P150,“ and PI,0 will indicate the extra
delay and stops that would result from implementing a signal
timing plan calculated with an a of 0.5 if, in fact, an a of
0.4 describes more accurately the platoon dispersion on that
particular network.

Similarly, five additional simulation runs were conducted
to determine the values of PISO/SO' lelw, Plan/30' PIso/wv PIso/sov
and their values were compared to the performance indices of
the optimized plans, P130, PI“), and PISO. The experiment was
conducted for two sets of traffic flow along the tested
netwbrk, which represented the a.m. and p.m. peaks. It was
found that the maximum difference in the PI resulting from
assuming an inappropriate (1 for the network never exceeded
1.56 percent of the optimum value. This difference (1.56
percent) represents the extra delay and stops that would
result from implementing a signal timing plan calculated with
an a of 0.3 if, in fact, an a of 0.5 describes more accurately
the platoon dispersion on that particular network. The value
of the error was determined by subtracting PIso from P130,” and
dividing the result by PIso (i.e. {378.4 - 372.6} / 372.6).

Although it was acknowledged that different values of a
produced noticeably different patterns of arrival on any

single link, it was concluded that the TRANSYT optimized

42
signal timing plans (for a network) were not very sensitive
to the value used for the parameter a, in the 0.3 to 0.5
range. Castle and Bonniville wrote in this regard the
following:

In effect, the requirement to minimize delay
throughout the whole network, and not merely on an
individual link, acts as a constraint on the
potential sensitivity of TRANSYT plans to platoon
dispersion rates. The results obtained suggest
that for a wide variety of network configurations
and characteristics, the use of the default platoon
dispersion factor of 35 is quite adequate. The
optimized timings found by TRANSYT appear to be
sufficiently insensitive to the precise value of
this factor that most users can expect satisfactory
results by utilizing the program's default value.

Smelt (44) studied the dispersion of 47 platoons of
traffic along the "Princes Highway East" which is a major
suburban divided arterial road in Melbourne, Australia. Data
were collected (time of arrival for every vehicle) during the
p.m. peak period at five stations along the major road. These
stations were located at 200m, 400m, 600m, 800m, and 1200m
downstream from a signalized intersection on that road. The
observed platoons at each station were grouped into four
categories according to their sizes. The Robertson dispersion
model was then applied to the data (i.e. platoon flow pat-
terns) collected at the first station (at 200m from the
intersection) to predict the platoon flow patterns at the
other four stations.

To calibrate the platoon dispersion model to reproduce

the actual traffic pattern as close as possible, pairs of a

and B were tested together. The values of a and B that

43
produced the best agreement between the observed and predicted
(by platoon dispersion model) flow patterns were then deter-
mined for each category of platoon size at every station.
Consequently, 16 calibrated pairs of a and B were obtained (at
each station, a pair of a and B was found for each size
category).

It was found that all pairs of a and B were similar and
that that platoons remained together and did not disperse
greatly (i.e. did not reach random flow) over the 1000m
between the initial station and the final station. Therefore
it was concluded that "a and.B do not vary greatly either with
increasing platoon size or increasing length of section."
Finally, Smelt reported that the results of his study show
that the average values of a and B are 0.19 and 0.89, respec-
tively.

In Malmo, Sweden, Leden (45) did a validation study of
TRANSYT and Rosim (a simulation program of a road network
controlled by traffic responsive signals) in which he
mentioned that the value of the parameter a depends on the
road characteristics and on the average speed along the link
under consideration. In one of the studied links, it was
shown that the platoon dispersion model was much more sensi-
tive to changes in B than in a. On this particular link,
Leden found that the values of a and B that provided the best
fit between the observed and predicted traffic flow patterns

were 0.3 and 0.8, respectively. However, more explicit

44
information regarding the relationship between the road
characteristics and speed, and the values of a and B, was not
given in the study.

Axhausen and K6rling (46) examined the effect of varia-
tions in the parameter a and link speeds along a real network
of eleven intersections in the cities of Pforzheim. and
Karlsruhe, Federal Republic of Germany. Link speeds varied
between 30 and 60 km/h (18 - 38 mph). The default value for
the parameter a was chosen to be 0.4, however on some of the
links it was lowered according to the recommendation given in
the TRANSYT-7F users manual. The network was then optimized
and the resulting performance index (PI) was determined. This
optimization run was referred to as the base case.

A number of additional optimization runs were conducted
by varying (increasing or decreasing) the speed and the values
of a along all the links by a specific amount. For example,
in one of the runs, the values of a were increased by 25
percent along all links, and at the same time link speeds were
decreased by 25 percent. The optimal signal timing plan
(offsets and splits) and the performance index (PI) for these
runs were then determined. Each of these optimal signal
timing plans was simulated using the original values of a and
link speeds (i.e. the same values of a and link speeds used
in the base case) and the resulting performance indices were
compared with the performance index (PI) of the base case.

It was found that the maximum difference is about 5 percent

45
of the value of the performance index of the base case. On
one of the studied links, it was found that inappropriate
values of the parameter (1 resulted in a large number of
additional stops. It was also mentioned that incorrect values
of a and speed could lead to "grave errors" in the choice of
the appropriate offset.

Axhausen.and.K6rling wrote the following in this regard:

It is also obvious that errors in the speci-
fication of a and speed might lead to grave errors

in the choice of the offsets. These grave errors

do not create problems for the theoretical validity

of the TRANSYT results, but for the application of

TRANSYT results to real networks, as a consulting

engineer will be hard pressed to defend such errors

as the result of an optimization of the signal

setting.

A second study was conducted for the calibration of the
parameters a and B. Axhausen and Kérling reported that out
of the large number of possible factors that might affect the
proper values of a and B, only five seemed to be the most
important in the "European context," namely, number of lanes
available to traffic, slope, crossing pedestrians downstream,
flow'conditions at the stop line (disturbancezby narrow lanes,
crossing pedestrians, turning vehicles blocking the lane), and
parking activities.

To study the effect of these factors on the parameters
a and B, data were collected from eight sites in the cities
of Karlsruhe and Pforzheim during the p.m. peak. The "least-

squares" criterion was used to calibrate the platoon disper-

sion model. It was found that there were a large number of

46
a and B pairs that produced almost the same results in terms
of the "least-squares" criterion. Therefore, it was concluded
that the model should be calibrated for a, while fixing B at
a given value. Axhausen and Kérling wrote in this regard:
A closer look.at the squared deviation surface

showed that there is a wide region of pairs of a

and B for which the height of the surface is nearly

equal.

They continued:

The value of B has to be fixed externally,

i.e. as the ratio of the travel-time of the first

to the mean or median vehicle of the platoon, to

find an unambiguous solution for 0.

Because of this, and since the default value of B (ize.
B = 0.8) in the TRANSYT program can not be changed, it was
decided to calibrate the model for the default value of B.
However, it.was acknowledged.that the model with B being equal
to 0.8 did not generally produce the absolute minimum of the
calibration criterion. The study revealed that the mean value
for a is 0.37, which is close to the default value of 0.35
used in TRANSYT-7F and TRANSYT/8. Analysis of variance was
also introduced to test the relationship between the values
of a and the five factors (mentioned previously) that describe
the roadway conditions. The results of the analysis are
summarized in Tables 2.4 and 2.5.

The results showed clearly that the number of lanes,
slope, and crossing pedestrians were the proper factors in

explaining the variance. Hence, one can conclude that these

three factors are very important (at least, in this study) in

47

TABLE 2.4 Results of analysis of variance in Axhausen and

 

 

K6rling study. Effect of "source" on a.

Source SS df F Significant at
Number of lanes 0.15 1 7.36 0.01
Slope 0.09 2 2.43 0.12
Parking activity 0.00 1 0.00 0.95
Crossing pedestrians 0.49 l 25.56 0.00
Flow at stop line 0.00 1 0.03 0.87
Subtotal : Model 0.74 6 6.37 0.02
Error 0.27 14

Total 1.01 20

 

 

 

(Source:Reference 46)

TABLE 2.5 Mean of a for the level of the design factor in
Axhausen and Kérling study.

 

 

 

 

 

Source Level Mean
Number of lanes One 0.46
TWO 0.29
Slope Slightly upwards 0.48
Level 0.33
Slightly downwards 0.34
Parking activity Parking activity 0.37
No parking activity 0.37
Crossing pedestrians Crossing pedestrians 0.53
No crossing pedestrians 0.23
Flow at stop line Disturbed 0.37
Smooth 0.38

 

(Source:Reference 46)

 

 

48

selecting the proper value for a. Axhausen and Korling
recommend that in any further study of the problem the number
of measurements taken be increased and to include other
factors in the analysis. These factors might include the
available width of the road, the percentage of trucks, and
mean speed on the links It was also suggested to describe the
study sites, as much as possible, with quantitative variables.
For example, it was suggested to use number of parking and un-
parking vehicles instead of "parking activity," saturation
flow instead of "flow at stop line,” and so on.

Axhausen and K6rling (46) summarized the results of four
platoon dispersion studies conducted in the Federal Republic
of Germany, Poland, Florida and Texas. These studies are
discussed (based on Axhausen and Kérling documentation) in
the following four paragraphs.

Dhardi (from ref. 46) examined the dispersion of platoons
along three sites in West Germany. Each site had two lanes.
The traffic flow was disturbed by parking activities at the
first site, and by pedestrian activities at the second site.
The traffic flow at the third site was not disturbed. With
the parameter B being fixed at 0.8, it was found that the
values of a that provided the best fit between observed and
predicted traffic flow patterns were 0.26, 0.32, 0.24 for the
first, second and third site, respectively.

Tracz (47) studied the dispersion of platoons along two

sites in Poland. Each site had two lanes with an average

49
width of 4m per lane. With B being fixed at 0.8, the platoon
dispersion model was calibrated. It was found that the a
values that provided the best fit between the observed and
predicted platoon dispersion were 0.3 for one site and 0.23
for the other site.

Lorick (48) calibrated the platoon dispersion model along
three sites in Gainesville, Florida” Each site had two lanes,
an average lane width of 3.7m, no pedestrian activities, and
all sites were on a level slope. The best fit values of a and
B for the three sites were different. These values ranged
between 0.3 and 0.5 for a and between 0.65 and 0.8 for B.

Denney (49) studied the dispersion of platoons along one
site in Austin, Texas. The site had three lanes, drive ways
that disturbed traffic flow, an average lane width of 4m, and
no parking activities. With B being fixed at 0.8, the platoon
dispersion model was calibrated. He found that the value of
a which provided the best fit between the observed and
calculated traffic flow patterns was 0.25.

Among all the studies mentioned in this section, the
user's manual for TRANSYT-7F is the only source which provides
general guidelines for selecting the values of a. Neverthe-
less, five important conclusions can be drawn from these
studies:

1. The reliability and effectiveness of the TRANSYT

model depends on selecting the proper values of a

and B.

50

2. The proper values of a and B are a function of the
traffic conditions and geometric features of the
network under consideration.

3. For similar traffic conditions and geometric
features, different values of a and B are reported
in different countries. This might be explained by
the fact that driver performance characteristics and
habits differ from one society to another. Conse-
quently, one can hypothesize that the values of a
and B are affected by the driver performance
characteristics and habits.

4. Calibrating the TRANSYT model by determining the
most appropriate (best-fit) value of the parameter
a while keeping the parameter B constant at its
default value of 0.8 produces satisfactory, but not
necessarily the best results.

5. Length of links and size of platoons seem to have
little influence, on a and B.

6. The performance index (PI) of an entire network is
not very sensitive to values of a between 0.3 and

0.5.

The basic conclusions from the above literature review

are the following:

51
Traffic signal optimization is an essential step
for achieving smooth and efficient flow on a street
system.
From all network optimization models, TRANSYT is
the only model which has been extensively and
successfully used in practice. Furthermore, the
model can be constructed to handle the Saudi Arabian
traffic conditions.
The TRANSYT-7F model is appropriate in fulfilling
the objectives of this study.
The TRANSYT-7F model has to be tested and calibrated
for the traffic conditions in Saudi Arabia.
Even though the parameter B is not under the control
of the TRANSYT user, calibration of the model can
be accomplished by determining the appropriate value
of the parameters a.
The appropriate values of a and B are a function of
the traffic conditions, geometric features of links,

and driver performance characteristics.

CHAPTER 3
THE TRANSYT-7F MODEL

3.1 Model grrpgture.

As discussed previously, the TRANSYT-7F program consists
of two principal parts: a traffic model and an optimization
procedure. A pictorial representation of the interaction
between these two parts is shown in Figure 3.1, The optimiza-
tion procedure utilizes the traffic model to evaluate the
effect of changes in the signal phase durations and offsets.
Optimization of the signal settings (signal phase durations
and offsets) is achieved by minimizing an objective function
called the Performance Index (PI). This Performance Index is
a linear combination of delay and stops, in.which the user can
express the importance of stops relative to delay for each
link in the network according to his objective and conveni-
ence. A more detailed discussion about the traffic model and
the optimization procedure was given in chapter 2 of this

study.

52

53

\l

.EmpwoLo Hymz<me one mo mcsuosoom are

 

 

 

 

xopcm
ouantot-m

8.035:
E «noun
+ $200

 

 

 

 

 

 

 

3568
:32»
535330

 

 

 

Esp-coed

cater Ezdo

 

 

- /[

 

 

 

 

 

 

Sap

so?! Ewao

3:38

Eco? 302

.59:
seat

_.m 35w:

 

 

one?!

ace. 3:...

 

 

 

 

 

 

 

..\‘\ ,

 

 

a... so...

33 x3332

 

 

 

54

The input data required to operate the TRANSYT-7F program

can be grouped into six major categories as follows;

1.

Network representation. The entire network must be
represented in terms of nodes (intersections) and
links (streets).

Network data. These data describe the geometric
features of the network. Basically, this category
includes link lengths and number of lanes per link.
Timing data. The entire signal timing plan has to
be input in the program. The plan includes cycle
length, signal splits, an initial set of offsets,
phase duration, and sequences.

Traffic data. This category of data includes
average cruise speed, traffic classification, flow-
from mid-block sources, and volume data (total flow
by link and by movement).

Driver characteristics data. These data describe
the driver performance characteristics and include
saturation flow rates, start-up lost time, green
extension time, average vehicle spacing, and the
platoon dispersion factor a. It should be mentioned
that TRANSYT-7F has default values for these
parameters that can be used if more precise informa-

tion does not exist.

55
6. Control data. These data include all the parameters
by which the user may instruct the program to
perform different tasks. For example, the user can
instruct the program to perform a simulation or an

optimization run.

3.3 Output charagtezistics.

The TRANSYT-7F program produces five major categories of
outputs, namely, input data report, traffic performance table,
signal timing tables, time-space diagrams, and flow profile
plots.

The input data report is a listing of the input data
reported in the same order and format the data were input to
the program. Warning and error messages are also included in
this report. This diagnostic report is used to detect missing
or questionable data resulting in warning and error messages.
Although this report is optional and can be suppressed, the
warning and error messages will always be reported.

The traffic performance table reports several measures
of effectiveness (MOE's) of traffic performance for each link
in the network separately, and for the entire network,
collectively. The most important measures of effectiveness
are the following:

1. Total delay (veh-h/h)

2. Average delay (sec/veh)

56

3. Uniform stops (veh/h, & %)

4. Performance index

5. Total travel time (veh-h/h)

6. Fuel consumption (ga/h).

The signal timing tables provide a complete signal timing
plan. This plan includes interval lengths and offsets for all
the nodes in the network under consideration.

The flow profile plots are used, primarily, to calibrate
the platoon dispersion algorithm used in the TRANSYT-7F
program. These plots are optional and can be requested for
some or all the links in the network under- consideration.
Basically, the flow profile plot is a histogram that shows the
average flow rate at the stop line during the cycle length.

The time-space diagrams are optional and can be requested
for any contiguous series of links. These diagrams are
valuable tools in evaluating the through progression of

traffic along network streets.

W;-

The TRANSYT-7F model can perform any of the following
tasks:

1. Simulation run. Traffic simulation is performed to

evaluate the effectiveness of an existing signal

timing plan. Simulation runs are also used in

calibrating the model (see section 4.4).

57

2. Optimization run. If the network under considera-
tion has a single cycle length (or a multiple of
this common cycle length), the model will optimize
signal splits and offsets to determine the optimal
signal timing plan. On the other hand, if the
network has different cycle lengths, the model will
first select the optimal cycle length (by executing
a cycle length evaluation run), and then optimize
the signal splits and offsets.

3. Cycle length evaluation run. The cycle length
evaluation run is performed to select the optimal
(best) cycle length, for a given network, from a
number of specified cycle lengths. To accomplish
this, the model will optimize the entire network
using each specified cycle length. The cycle length
that produces the lowest value of the performance
index (PI) is the optimal cycle length.

4. Input data scan run. None of the above tasks is
performed in this run. The input data are just

checked for any errors or missing information.

3. ' 't t'o ode .

The TRANSYT-7F has three limitations. These limitations

are summarized as follows:

58
As an evaluation (simulation) tool, the model is
only applicable to systems (networks or roads)
possessing a single cycle length or a multiple of
this common cycle length.
The model does not explicitly optimize phase
sequences. Consequently, to obtain the best
(optimal) phase sequences at a given node, a number
of possible or desirable phase sequences have to be
identified. Following that, an optimization run has
to be performed for each set of phase sequences.
The set of phase sequences that produces the lowest
value of the performance index is the optimal phase
sequences.
The model does not consider volume to capacity ratio
(V/C) and queue capacity as constraints in the
optimization process. It is not uncommon for the
model to provide an optimal signal plan that
produces queues longer than links (spill over) or
a V/C ratio in excess of 95% along some links. In
most of these situations, the problem can be
alleviated with simple adjustments of the signal
splits and/or by increasing the cycle length by a
small percentage.
The TRANSYT-7F program does not optimize signal
offsets and splits simultaneously. The program will

optimize either one in each optimization step. A

59
default list of optimization steps is available in
the program. However, the user can change the size,
sequence, and/or the number of steps in this list.
Although the default list provides an acceptable
signal timing plan, it is not necessarily the
optimal (best) one.

5. The TRANSYT-7F program assumes that vehicles enter
the network (i.e. from the external links) at a
constant uniform rate. (Although this assumption is
not theoretically appealing, it is not unrealistic
over long period of analysis such as one hour.

6. The only platoon dispersion factor under the control
of the TRANSYT user is "a" (i.e. PDF). The param-
eter "B", on the other hand, can not be altered from
the default value of 0.8.

3.6 m ut r l ements.

The microcomputer version of the TRANSYT-7F program will
run on any IBM PC (or PC compatible) microcomputer that has
at least 256K of memory and two double disk drives. The DOS
2.0 operating system (or any later DOS version) is also
required to execute the program.

The execution time in optimization runs is a function of
the number of nodes and computer architecture. For an IBM

microcomputer possessing a math co-processor, the execution

60

time is 5 minutes per node. However, without the math co-
processor, the execution time may be as high.as 30 minutes per
node. On the other hand, the execution time in simulation
runs is only a few seconds per node.

The TRANSYT-7F program can handle up to 50 nodes and 250
links. Larger networks can be analyzed using the overlaying
technique, or by reducing the size of the link and node-

related arrays as discussed in the TRANSYT-7F manual (7).

CHAPTER 4

DATA COLLECTION METHODOLOGY

The methodology followed in this study consisted of
collecting the required input data for a selected network in
the study area, comparing the traffic flow patterns computed
by the model with those observed in the field (along the test
network), calibrating the model to decrease the discrepancy
between calculated and observed flow patterns, validating the
modified model in another test network, and comparing the
results of the calibrated model with that of the same model

using default values for the model parameters.

4. a t'o th 'de .

As discussed in chapter one, the successful utilization
of the TRANSYT-7F model (as well as any optimization or
simulation model) depends on selecting the proper values of
the factors that describe the driver performance charac-
teristics. In the TRANSYT-7F model, there are four such
factors that have to be determined (i.e. measured from the

field) for the study area traffic conditions. These are the

61

62
start-up lost time, extension of the green phase into the
clearance interval, average vehicle spacing, and saturation
flow rate.

In addition to these factors, the platoon dispersion
algorithm used in TRANSYT-7F has to be calibrated for the
traffic conditions existing in the study area. As discussed
previously, this algorithm describes the desire of individual
drivers to maintain. a comfortable time headway’ as they
progress along network links. The platoon dispersion al-
gorithm is affected by the drivers performance characteris-
tics. The algorithm used in TRANSYT-7F was calibrated for
conditions existing in the United States, and is not neces-
sarily valid for the traffic conditions in the study area.

As discussed in previous chapters, the calibration
procedure consists of determining that value of "a" which,
when used in the platoon dispersion algorithm, produces best
agreement between the simulated and observed flow profiles.
A more detailed discussion on this procedure is given in the

next chapter.

4.; Site selection.

The TRANSYT-7F model can not be used to simulate traffic
in a network possessing different cycle lengths. Thus to be
capable of simulating the traffic in a network (and conse-

quently to calibrate the model), the entire network must have

63

only one cycle length or a multiple of this common cycle
length. To select a study section, the cycle length of every
signalized intersection in the cities of Dammam and Al-Khobar
was determined from the field. Unfortunately, only two
arterials in Dammam. and one in .Al-Khobar satisfied the
criterion of a common cycle length. Furthermore, one of the
two arterials in the city of Dammam.was undergoing a substan-
tial pavement rehabilitation, which made it unavailable for
this study. Consequently, only two arterials (one in each
city) were used in this research.

Each arterial consisted of four signalized intersections,
three lanes in each direction, and they were both located in
areas of mixed residential and commercial activities.
Sketches of the two arterials studied and a typical phase
sequence are provided in Appendix A. Further description of

these two arterials is given in the next chapter.

4.; Data collection.

The data collection phase was a major step in this study.
The TRANSYT-7F requires a considerable amount of data that
describe the‘geometric features of the network and the traffic
characteristics within it. These include road geometry
(length and. width), signal timing' plans, control ‘volume
counts, traffic flow by link and by movement, and the average

speed on each link. This is in addition to the previously

64
mentioned data needed to describe the drivers performance
characteristics. A summary of TRANSYT-7F data requirements
is shown in Table 4.1.

Because traffic and operational data does not exist for
the study area, the data needed for this study was collected
from the field. With the exception of the average vehicle
spacing and the platoon dispersion data, the procedure
recommended by the TRANSYT-7F manual was followed in col-
lecting and presenting the required data (7).

The average vehicle spacing is used by the TRANSYT-7F
program in the calculation of queue capacity along each link
in the network under consideration. The program uses only
one value for the average vehicle spacing in this.calculation.
This is logical, given the fact that this value is a charac-
teristic of the driver and is not a function of road features.
The average vehicle spacing in the study area was determined
by randomly choosing a link from the study area and studying
the queue characteristics along this link. A total of 105
queues were observed along the selected link. Vehicle spacing
was calculated for each queue by dividing the length of the
queue by the number of vehicles. The average vehicle spacing
was then determined by averaging the values of the vehicle
spacing obtained from all the studied queues. The average
vehicle spacing was 7 meters, and the variance was 0.27

meters, providing a 95% confidence interval of i 0.1 meter.

65

TABLE 4.1. Summary of TRANSYT-7F data requirements.

 

MAJOR CATEGORY

DATA TYPE

 

 

 

 

 

 

 

Network Data 1. Nodes (i.e. intersections).
2. Links (i.e. streets).
3. Link distances.
4. Parking and turn restrictions.
5. Bus routes.
Timing Data 1. Existing cycle length.
2. Existing offsets.
3. Existing interval duration.
4. Existing phase lengths.
Driver Performance 1. Saturation flow.
Characteristics 2. Start-up lost time.
Data 3. Green extension time.
4. Average vehicle spacing.
5. Platoon dispersion data.*
Speed Data 1. Cruise speed on the links.**
2. Bus dwell time.
Volume Data 1. Control volume counts.
2. Total flow by link and by
movement.
- 3. Flow from mid-block sources.
4: Classification of traffic.
5. Input flows from upstream links

 

(Source:Reference 7)

* For model calibration.

** See Appendix B for speed data.

 

66

A total of 369 observations were made at the eight
intersections studied for determining the start-up lost time.
The mean value was 2.49 seconds, and the variance was 1.42.
For determining the extension of the effective green, a total
of 304 observations were made at the same locations“ The mean
value and the variance were 3.44 and 1.70 seconds, respec-
tively. Because TRANSYT-7F requires single digit values for
the start-up lost time and the extension of the effective
green, the values of 2 and 3 were used, respectively.

The platoon dispersion data, or equivalently, the
arrival flow patterns, were needed to calibrate the platoon
dispersion algorithm used in TRANSYT-7F program for the
traffic conditions existing in the study area. The arrival
flow pattern (profile) was obtained for every link in the
study area. This was done by counting the vehicles arriving
at a point upstream of the investigated intersection at a
distance large enough to insure that the arriving flow'pattern
was not disturbed or influenced by the vehicles queued at the
downstream intersection. This distance ranged from 80 to 200
meters depending on the traffic characteristics of each
investigated link.

Given that the cycle length in the study area was 120
seconds, the cycle was divided into 24 intervals (steps) each
of which was five seconds long. The number of vehicles

arriving in each step was counted separately. This was done

67
over 30 consecutive cycles (TRANSYT recommends at least 10

cycles), which represented a one hour study period.
4.4 edu o 'v't'es.

The data used in this study was collected during the
summer of 1988 in the cities of Dammam (First Street), and
Al-Khobar (King Abdul Aziz Street). During this period there
were no abnormal conditions that might have affected the
traffic characteristics in the study area.

The help of fifteen civil engineering students (studying
at KFUPM) were required to undertake the data collection task.
All of the students had previous experience in collecting
traffic data. In addition, they were all extensively trained
to collect the data needed in this study.

Volume, speed, signal timing, and driver performance
characteristics data were collected near, but not during the
morning peak hour. In the city of Al-Khobar the data was
collected.between 9:00 and 10:00 a.m., and in Dammam from 9:30
to 10:30 a.m. The morning peak hour, which is around noon in
both cities, was found to impose a large burden on the student
collecting volume data, and consequently increases the chance
of human error. In addition, it was uncomfortably hot for the
students to conduct their tasks reliably during the noon hour.
Other physical and geometric data, such as number of lanes,

locations of non-signalized intersections, length of links and

68
turning bays were collected either in the early morning, or
late in the afternoon.

The first ten days were devoted to locating and deter-
mining the cycle length of every signalized intersection in
the cities of Dammam and Al-Khobar. The next two weeks were
spent in selecting and training the participating students.
Another three weeks were spent in collecting the required

data.

CHAPTER 5

DATA ANALYSIS

5.1 Driver performegce characteristies.

As mentioned in previous chapters, the first step in
conducting this study was to collect the required data for
the model. One type of required data is the driver perfor-
mance characteristics, which include the extension of effec-
tive green time, average vehicle spacing, start up lost time,
and saturation flow rates. These driver characteristics, as
observed and evaluated in the study area, are summarized in
Tables 5.1 and 5.2, in which one can easily see that they are
not very different from what is usually encountered in the
United States. More specifically, comparing the value of
these characteristics with those given in the TRANSYT-7F
manual for different categories of drivers seems to indicate
that drivers in the study area can be classified some where

between normal and aggressive.

69

70

Table 5.1 Average vehicle spacing and extension of effective green in the study area.

 

The study area

TRANSYT-7F recouIIIendati ons

 

8Extension of

Average vehicle

bExtension of

cAverage vehicle

 

 

effective green spacing effective green spacing
(sec) (dm) (sec) (dm)
dChange period
3 70 minus 0-2 78
seconds

 

 

 

 

 

Only integer seconds may be used in TRANSYT-7F.
Recommended values for normal and aggressive drivers.

Default value.

Change period in the study area ranged between 3 to 5 seconds.

Table 5.2 Start-up lost time and saturation flow rate in the study area.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

City of Dame-I City of Al-Khobar TRANSYT-7F Recommendations
cSaturation cSaturation cSaturation
aStart-up flow rate aStart-up flow rate aStart-up flow rate
lost time (vphg/lane) lost time (vphgllane) Condition lost time (vphg/lane)
(sec) b (sec) b (sec) b
Thru Turn Thru Turn Thru Turn
1720 1690 2 i 1700 1650 Aggressive 2 1800 1700
i drivers
f 116ml 3 1700 1600
1 drivers

 

a: Only integer seconds may be used in TRANSYT-7F.
b: Protected, unopposed turning movements.
c: Rounded to the nearest 10 vehicles.

 

71

5.2 Eetimating PDF velees in the etugy etea.

The two arterials (one in Al-Khobar and one in Dammam)
used in this study were, generally speaking, very similar.
They were both high type arterials located near the central
business district. Each arterial consisted of four major
signalized intersections with four approaches in each.
Consequently, the flow profiles could be observed for six
links (three for northbound traffic and three for southbound
traffic) connecting the four intersections in each arterial.
The physical, traffic, and geometric features of these two
arterials are described in detail in Table 5.3. By comparing
these features for each link with the characteristics of
different levels of roadway friction as given in the
TRANSYT-7F manual, the roadway friction on each link was
classified as being either low or moderate. Since the manual
recommends a PDF value in correspondence to the roadway
friction level, the recommended PDF value for each link was
found from the manual as shown in Table 5.4.

The twelve links (six in each city) were classified into
eight and four links having low and moderate roadway friction
characteristics, respectively. To divide these two categories
of links (i.e. links with moderate friction and links with low
friction) evenly between the calibration and validation
studies, the links serving northbound traffic in.Al-Khobar'and

Dammam were used in the calibration study, and the southbound

7'2

.3505... 00.. 0.30.... 30.. 0.0.... 058. 05 3 5.000.802. 3.00.509 05 to...

6053:: 00.. 0.30.... ...o.» 0.0.... «ER. 0.... 00 0.5.o>

 

.80. SmuGLOdOUUG

 

 

 

 

 

 

 

3. . 3 2K. 21> 03+. :0 «a p00... 0:0.
noon 0:0: 0:0: 0:0: 3K ...}... ~n 0:: mm m Rm 0:.....0.. 5:030...
0>+ 0.80 0:0 002.6
3.30.. 009.008.
5.20.... 3 3 - t .e.u:0o.00.. 0:00 8.0.. 29 90.. .23.
530... name 0.0.0.8: 0:0 1.9.0.380 3... «Na ...}... .3 0:0 -n n o.n
3.2.00 009.000...
C - 3 33:00.00: 0.50 83
no.3 0.9.0.8: o:- .0.0..0..e.ou so. ..No .23 no 0:0: eon n o.n
3.9.0.... 009.000:
2 - 2 1.2.8.8.. 8.8 ...... 2..., E:
53 30.0.8: 0:0 6.0.0.380 30. .3 .23.. co 0:0 eon n o.n
3.30.. .8.
...-E... 3 C . 3 .3283». :3 8m. 2.... .... .63
5.0.. .80 so. 0:0 3.9.05.3 30> .3 2.3 00 0:0 s~n n o...
.3003; 2030....
3 - ...... 2K. 00:30... 0.5...
33 0:0: 0:0: 0:0: man ...}... 3 0:0: has n 0... 0330.... £5.08:
0). ...... poise
.03: - 03:. 8:: 2.... .0 21>. 2.0....e .... c...
a 5.00.... on... 0..» 0:30 0.30... r.o> 3...... p083 053009.005 50:0. 00:0. 50...
8.000.... x:... 05...... 5.3.0.. .0 5.80000: 2:0... .009 .0>< ..o:.l .0 ‘ ~..:.. 0 0:... ...-.80..

 

 

5.20.. ...... a...

 

 

 

 

 

 

 

 

 

 

in ..0 >50

.92.. no.3: 0... .o 33 0.30..» ...... 00.5.00. 0.30.53 n.m 0......

 

73

300.050 00... 3.3.0 30.. 0.0.... 0:.00 05 0» 53000.03. 100.39. 05 0.0...
.00>.0000 003 0...0.0 30.. 0.0;: .:.00 0:0 00 00:.0>

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3.33 30.
3.0. . 3. 33:00.00. :0. .00. 20> 00. £3. 00.0 0.3.3.0000
n03 30. 0:0 3.0.30 30> 000. .03 ..0 0:0 0..” n ..n .0 095.. 0.000
3.0:00 30.
.320... 3 SN . -. 1:50.02 .3. 80. 20> .... ...}.
00:00 mono 20. 0:0 .0.0.00§00 >.0> 0~0p .2\02 00 0:0 ..n n ..n
>0.0:00 30.
.- . .m. 3.33.00. :0. .00. 20> m... .mm+.
no.0 30. 0:0 3.0.0.0000 30> .00. ...}... 00 0:0 0.... n ..n
3.3 03.30... 20> 00+.
..~ . -. 3.33.00. .00. 20> 0... .2}.
.000 03.30.. 0:0 3.0.0.0000 30. 000. ...}... 00 03 0... n ...n
3.3 30.
000500 0» .- - m~. .0..:00.00. 30. .00.
5.0.. .0~0 30. 0:0 3.0.0.050 30> 03. ...}... n0 0:0: En n ..n
3.0000 :0.
.m~ - 0m. .0.0:00.00. 30. .00. .00> 0n.+.
.000 :0. 0:0 .0.0.00500 >.0> .00. .2x0. 00 0:0 0~n n u.n
.30: - 30:. 00.. 05 .0 20>. 00.306 .... ....
0 5.50... 3.0 05 0:03 0...0.u #033 3...... 00000. 083000.03. 50:0,. 00:0. 50....
5.33.0 ..:.. 0:...00 3.3.00 .0 5.500000 0...0.. 00000 .0>< .050 .0 0 ~30... a 0:0. 3.050..
5.80.. .80 8....

 

 

«(8:32: .... >h .u

.. .0. 0:00. M... 0.00.

Table 5.4 Suggested roadway friction

74

and PDF values.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Used in PDF. Roadway City and 1 Link # I
value characteristics direction (Node-Node)§
calibration 25 low Al-Khobar 9401 ,
friction N.B. (24-23) ;
calibration 25 low Al-Khobar 9201 g
friction N.B. (23-22) g
calibration 35 moderate Al-Khobar 9001 i
friction N.B. (22-21) .
!
validation 25 low Al-Khobar 9103 ;
friction 8.8. (21-22) g
validation 25 low Al-Khobar 9303 E
friction 8.8. (22—23) T
validation 25 low Al-Khobar 9503
calibration 25 low Dammam 9101
friction N.B. (12-6)
1
‘calibration 25 low Dammam 9201
] friction N.B. (6-7)
gcalibration 35 moderate Dammam i 9301
1 friction N.B. { (7-1)
ivalidation 35 moderate Dammam 9403
g friction N.B. (1-7)
ivalidation 35 moderate Dammam 9503
ivalidation 25 low Dammam 9603
L friction N.B. (6-12)

 

 

 

 

 

*: Recommended by the manual of TRANSYT-7F in

level of roadway friction (third column).

accordance to

75
links were used in the validation study. This distribution

of links between the two studies is shown in Table 5.4.

5.3 Calibration of the model.
5.3,; Procedure.

As discussed previously, model calibration consists of
determining the value of the platoon dispersion factor
(referred to as "a" in the Robertson dispersion algorithm,
and as "PDF" in TRANSYT-7F) that when used in the TRANSYT-7F
model, achieves the ibest. agreement. between ‘the observed
traffic flow patterns and those predicted by the model.

Starting with the first link serving northbound traffic
in Al-Khobar (between intersections 24 and 23 as shown in
Table 5.4), the PDF value which produces the best agreement
between the observed and simulated flow patterns was deter-
mined by conducting several simulation runs for the entire
arterial. In each simulation run, a different value of PDF
was used for the first link; The PDF values investigated were
in the range of 0.15 to 0.60. For all other links in the
arterial, the recommended PDF values were used. In each
simulation run, a flow profile plot was requested for the
first link. By comparing the flow profile plots obtained with
the observed flow pattern and using the sum of absolute
difference criterion (which will be discussed shortly) the

best fit PDF value was determined for the first link. Using

'76

the best fit value of PDF for the first link and the recom-
mended PDF values for all other links, the same procedure was
repeated for the second and third links serving northbound
traffic in the city of Al-Khobar. This methodology was then
repeated for the links serving northbound traffic in Dammam.

The determination of the best fit value of PDF for each
link was not a straightforward issue, as one might imagine,
because of two basic difficulties. The first difficulty is
the fact that the flow profiles obtained from TRANSYT-7F
represent the shape of the flow pattern, but do not indicate
the number of vehicles arriving in each five seconds of the
cycle. These profiles are presented as vertical lines drawn
over a horizontal axis that represents the 120 second cycle
length. The length of these vertical lines represents the
relative number of vehicles arriving in each interval of the
cycle. To convert these vertical lengths into the number of
vehicles arriving in each interval over the study period, the
procedure recommended in the TRANSYT-7F manual was followed.
This procedure determines the scale of these vertical lines
by dividing the maximum flow along each link (reported with
each plot) by the maximum number of vertical symbols in any
plot. “By calculating the number of vertical symbols (i.e.
length of vertical lines) in each interval and using the
previous vertical scale together with the proper conversion
factors, the total number of ‘vehicles arriving in each

interval was determined.

77

The second difficulty faced in the comparison effort was
how to select easily and efficiently from the large number of
the simulated flow'profiles the one which is in best agreement
(best-fit) with the observed flow pattern. The criterion used
for this purpose was the minimization of the absolute value
of the total differences between the number of vehicles
simulated and observed in each increment over the study
period. Although this sometimes can be done visually by
comparing the simulated and observed flow histograms, quan-
titative methods are more reliable, especially when the
differences between the two flow patterns is relatively small.

The normalization and matching procedure require a con-
siderable amount of time and effort if done manually.
Therefore, a FORTRAN program was developed to conduct these
tasks. This program converts the vertical lines reported for
each interval in the simulated flow profile into number of
vehicles, compares this number to the actual flow, finds the
absolute difference, and sums these differences for the 24
intervals of the cycle. The program is given in Appendix C
under the name "CALIB". A sample output is given in Figure
5.1 for the first link serving northbound traffic in Al-
Khobar.

78

.«
.0
.MI
.0
.«
.«
.0
.N
.«
.0
.«
.0
.«
.0!
.MI
.00
.«
.0
.nl
.«
.«
.0
.«
.0
NJdum mun

:.mHA<U:

.0
.0
.0
.0
.0
.0
.0
.0
.M
.N
.N
.0
.0
.0
.M«
.0«
.«0
.00
.00
.00
.N«
.0
.N
.N
NJ¢U0 UCQ

Emuwouo may no ooduso wHQEmw < _.m wuswwm

.0
.0
.0
.0
.0
.5
.0
.0
.0
.N
.0
.0
.0
.0
.0«
.fl«
.0”
.00
.00
.00
.fl«
.«
.0
.N
UJ¢U0 50¢

II«0«£I>

.000« I 0060 QWPU«0W¢Q k0 200332
«00« I 0¢¢U DU>¢WUGO l0 200232
«N I 03Q¢¢0 23180 2« tun
.000« I &L«0 300 :30
.00« I lu«0 00¢ :30
0«.flfl «0.0 «N 00.0« N 0«
N0.« «0.« 5« 00.0« N 0«
00.00« 00.0« M« 00.0“ M 5«
”0.« «0.« 5« 00.0« N 0«
00.0 «0.0 0« 00.0« N n«
0«.0N N0.“ 00 00.0“ M 0«
00 .M 00 .« MN 00 .00 n M«
«N.00« «0.0« 00 00.0« N N«
00.00 «0.0 N« 00.5 « ««
00. 00. 5 00.5 « 0«
00.0 «0.0 «« 00.5 « 0
00. 00. n« 00.0« N 0
M«.0fl «0.0 MM 00.fl« N 5
«5.00 00.0 "N 00.«fl 0 0
00.00« 00.N« 00 00.00 5 0
00.000 00.0« «0 00.05 0« 0
50.0« 00.0 Nfl« N0.5N« 0« fl
N”.00« M«.0« 000 00.«0« 00 N
«0.0«« 00.0« 50« 00.50« «N «
50.0 00." NN« 00.0«« 0« 0“
00.00 00.0 00 50.50 0 MN
00 on 00 .« 0 00 .5 « NM
“0.0“ «0.0 M« 00.5 « «N
«0.« «0.« 0 00.5 « ON
“000 k00¢ 0062 Run! 0>Q2 052«

50000000«.0 IaBIIILflIL £U£« 0\« n 0003 NJ¢U0
”000 304k KC!

«0 Inn

«000 0 ¥Z«J

«£002 500 0500

79
5. 2 a ' ra 'on resu ts.

Applying the previous procedure for any link, one can
find that PDF ‘value whose fIOW’ profile satisfies (i.e.
minimizes) the sum of the absolute differences criterion.

Figure 5.2 shows how the sum of absolute differences
changes with different values of PDF for the first link
serving the northbound traffic in Al-Khobar. With the aid of
such a graph, the PDF value that satisfies the sum of the
absolute differences criterion can be readily identified,
which in this case is 31. The general "U" shape of this graph
was found to be a common characteristic in all of the inves-
tigated links. Similar graphical representations for all
other links are given in Appendix D. Figure 5.3 shows the
observed traffic flow pattern (for the same first northbound
link) drawn over the simulated one using the best fit PDF
value of 31. It is clear from this figure that the two flow
profiles are very similar, as one would expect.

Table 5.5 Summarizes the results of the calibration
effort. As can be seen from this table, theme is a clear
tendency for the best PDF value to be greater than the one
suggested by the TRANSYT-7F manual for both low and moderate
friction categories. More specifically, the best PDF values
ranged between 26 and 31 for low friction links, and it was
exactly 40 fbr the links with moderate friction, while the

suggested values were 25 and 35, respectively. The average

80

..oqm xcm« mo :OMomubm0oo

~.m «uswmm.

 

mwfii9> mam

0v @—

:—

 

 

J‘l q

.. m. ... ..
.mk)(k1.;rr»;fl

cotmfito Ejm 333g

1

 

 

 

.gem es=5 lo seapaea.4qu
mqaszg-aa

8.

mm—

on.

mom

onm

mam

SBONBHBAfiIU 31010888 30 HHS

 

 

._Oqa xcMfl wo mmdmwoua 30am cmom~ssmm ocm pw>pmmno m.m whowmm

 

8|

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

@000.» US...
8.9:8—8898898393“.
. . 1 1 4 . . . 1 4 luWUl
,8
low
u .00
u... .8 m
.400— an
mi
T! ..-. ..om— m.
.3
fl 1o: 5
7w .59
.. 032326 ,8.
... Do>ummn=u I4 .. HUN
0mm

 

 

.n use ..aea 52.4 in gs=ueaa 394;
gamszg-a<

 

 

._0qo xcfid mo mmfimwoho soda coom_:smm pom pm>ummbo m.m muowmm

 

81

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

n.000v W5...
8. o: Q: :0 ow as on 8 9 am cm a. a:
..om
. .9.
..-. .00
"i -8 mm
48— AM
mm
..-. 10m— m.
.. ....
% ...... a
T w loo.
.. Uwuo.3E_m 18—
.. Dm>ummDD ii {SN
omeu

.n use ..gea 52.4 is gs=laaa ”94¢
g£m@:¥-a«

 

 

82

.mCOwumccmEEoumu «cocoa mhlawmz<ma Eoymum
.mcofiuoocmssoomu «cocoa mhne>mz<ma eouhua.
.m0«wuoum
30«u omqupmum can om>ummno comaumn wocmuomuwo wu:«omno 0:» nonwEHCME uocu mom ocean”
.An seafloov mam ummn
mnu an cwuoHUmua 0:0 as» can mHHmouQ BOHH cm>uwmno cmmsumn mocmumuuwp mu:«omno no samum
.n.m
mance cw cm>fiu m« mfid«o> pm>uwmno .oo« « A05:«o> pm>umm00\mocmumuuwo mu:«omno mo somVuH

 

 

 

 

 

 

 

 

AH u so .m.z cofiuofiuu
Honm amesno oumumoo: mm cc pad e.v~
As n we .m.z cofiuofiuu
Homm sesame son ma m~ mad H.wH
Aw I NHV .m.z :0H90«Hu
Hoam 3.5560 304 ma on mad «.na
Adm u may .m.z newuofluu
Hoom umnonxufl< ouauwoo: mm o¢ ova . e.na
ANN u mac .m.z cofiuofiuu
Homm unnonxuac 304 mm hm mma a.ma
Anm u «mo .m.z :ofiuofluu
Ho¢m unconsaaa so; mN an and H.nH
cofiuomuwa m:«o> 0:«o> mocmumMMfio ma:«o>
Aoooznmuozc can mofiumwumuomuano mom mom musaomnc :0
* Kawq hufio amzomomm pmummmmst ummmm 00 32mm uouum a.

 

 

 

 

 

 

 

 

.mu«:mmu :o«umunfi«mu m.m GHQMB

83
best PDF values were 28 and 40, for the low and moderate

friction links respectively.

WW.

Using the calibrated PDF values found in the previous
section (28 and 40), the flow along each validation link was
simulated. The resulting flow profiles were then compared
to those observed in the field.

The flow along each of the above links was also simulated
using the PDF values suggested by the TRANSYT-7F manual and
the resulting flow profiles were compared to those observed
in the field. This was done to compare the calibrated PDF
values with those suggested by the manual. The results of
this validation effort is summarized in Table 5.6.

Considering the "sum of absolute differences" as the com-
parison criterion between the results obtained using the
calibrated and recommended PDF values, the following con-
clusions were reached:

1. With the exception of the last link in Dammam

(# 9603), the calibrated PDF values provided
superior results.

2. The differences are not large. The improvement in

the comparison criterion is less than 2% of the

observed volume along any link.

84

.0«flmoua 30«u omquomuQ pan Um>ummno cmmzumn mucoumuuwo mu=«omnm mo 59m r«*
.m.m manna :H cm>«o ma mea«o> .oo« « AxcflH may mco«m we:«0>\mocmumuu«c mu:«omna no Esmv

"a

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.m.m :owuowum
«com smegma sou mm oHH a.mH mm mad p.ma
.m . m COHHOMHW
mama smegma mumumuoz mm mud o.v~ oe had n.-
.m.m cowvowuh
no¢m smegma oumumcoz mm and «.mm ow mad n.a~
.m.m coflHOHHm
nomm umnonxuac sea mm «ma n.sH mm osa c.6H
.m . m COHUOHMm
nomm umnonxnaa aoq mu omH m.¢H mm sea n.¢d
.m.m coHuowum
noam umnonxuac son mm and a.ma mm 5nd a.ma
msam> .uuwo osam> .uufla
cofiuomufia mom .mn< «Houum w mom .de «Houum w
umnasz 6cm mowumfiumuomumzo ««35m ««35m
xcfiq aufio amzumom
mmflusum mnu ca
sawumocmaaoomu mhlawmz4ma=ocsou m:«m> cmumunfl«mo

 

 

.muaammu cowumpflam> m.m 0«nma

85
3. The reason why the last link in Dammam (# 9603) is an
exception is the fact that the best fit PDF value for
this link (see Table 5.7) is equal to the recommended
value (25), where the value used in the validation was

28, the average value for all low friction links. The

difference (in terms of the comparison criterion) between

the results obtained using the calibrated and recommended

PDF values for this link is small.

Because the number of links used in the calibration and
validation processes were relatively small (6 links in each),
it was thought that confidence in the resulting values would
be enhanced if one can demonstrate that the calibrated PDF
values would not change if the calibration and validation
links are reversed, or if all the links are used in the
calibration process. Thus, the links which were originally
used in the validation process were calibrated and their best
fit PDF values were determined. These values are reported in
Table 5.7. For convenience, the table also re-summarizes the
results of the original calibration process which were
previously reported in Table 5.5. Consequently, Table 5.7
contains the results of calibrating and validating all links
serving both northbound and southbound traffic in the study
area. A study of Table 5.7 leads to the following conclu-
sions:

1. If the links serving southbound traffic are used in

the calibration process, then the average best fit

86

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.oco 8.25:8 of :2: 8:5 .2 688.538. u.
.8— . «8303029333 33030 50 .53 u.
828:3 a 5:2. .8- .82.._5:8 a 5:8 3
53:3 55.5525: 2: :5 032, 58 2.9.2302... "n
6032.; >262... 50 an: :93 ..3 53¢ a...» c. “.53 03..) 3.. 3226 2: “N
$339353 3:: son... ..3 «3:95 .8: 33:00.5 .8. Bianca 5253 «3:20:53 3303. of .3 I5 05 «3.3:..- uogu 032, 5a.. of 2.
88 5... 3282. 5:8 an 5: .5. 3 3. .55— 3 3. 55. 82
5855-:
38 5... :3 5:8 5~ an. 55. 8 on, 55. R l a, .5, on:
.85-:
88 5... 3.. 5:8 5~ 5.: 55. 8 S. 5.5 .5 on. :9 $2
...505-.<
88 55 3.. 5:8 55. a: 55. 8 5: 55. 55. a: 5.9 E
. lilo .
88 55 32.8.. 5:; an 8. 5.5 3 t. 55... on t. n.- «N5
in
8.3 55 32.8.. 5:; 55 on. .C.- 3 5. 5.5 S a. 5.8 .8
ED
38 5... 3282. 5:8 55 on. 5.: 8 5: 5:: 8 t. 5:: :5
Ens-Io
82 5... :3 5:8 55. ~n. 5.8 8 8. «5. o~ 5: .5— .8
losing
88 5... :3 5:8 5~ on. ~5. 55. 55. 55. 8 5: 55. 55
. go .
3:; .33 «3.) 5:5 .32, 5:5
8m0005mo cm as .3< 050550 N 52 .00< 050550 fl be .3‘ 050550 N
5 .8. 828.: W3 .55 .55 .55 253
5c: 58 5.83.. 5.533 . . .. «53>
nco_uaucoeaouo¢
55-25.25 $53298 18 :5

 

 

 

 

 

 

 

 

 

 

.503; “35 nVoncfliauog ace £33323 .533 of 8.33 $05300 3:595 33.. 2: 5033 555.398“. 5.5.. 033

87

.08 $39333 of :22 .533 5a.. Egon—Boo...
.oo— . «95.388203; 3303. we .53

 

 

 

 

 

 

 

 

 

 

 

828:.) u 5:: ac. 60:25:: u 5:8 3
53:... 55.5.25 .5 E .32, 3.. 829588 5
£032.: 8330.. we on: so: ..3 >3: 3.: E ‘53 33> 52 39.2,- of “N
$539.05: so: can» .55 50:59:. to: $3950.55 u:- Vuzoono :00:qu «3:20:53 3303- of 5° .55 23 moi-:5... =5 03; 59. of "—
memo .m.m 30.— .U:o> mm 00— M.N— ow 05— 0.0? mm mcr 0.2. moor
$85.2
mono .a.m 30.. .Um.a> mm. 09. 0:2. QN new n.0r ”N 5.: n5: 0ND—
.85.:
nova .m.w 30.. .vza> m~ Fm— a.m.—F QN ND. 0.»: w~ NMF 0.2. 50°F
5.525.:
0130> #5:. 035a) .*w_o Onda> .wwmn
c3333 E mom .2: £0.30 x ”3.. .8: 3.3.3 a no... .2: 35...; x
a can 5.52.: wow: .5m 5m 5m 21:
xc: 3.5 >563: 3.33.75 .. c . Q5.o>
ncoCaEEuom
55:52:: Ncoznszau 18 35

 

 

 

 

 

 

 

 

 

 

4.3.5.33 fin 03:

88
PDF values are 28 and 40 for low and moderate
friction links, respectively. These are the same
values obtained in section 5.3, when the best PDF
values of the links serving northbound traffic were
averaged. The same is also true when links serving
both northbound and southbound traffic are used
collectively.
Using the "sum of absolute differences" as a
comparison criterion between results obtained by
different PDF values, the best results are obtained
for the best fit PDF value for each link.
Except for two links marked by an "*" in the table,
the calibrated PDF values provided superior results
(in terms of the above comparison criterion), for
all the links in the study area. Nevertheless, the
degree of this superiority is not large. The
improvement in the comparison criterion (as defined
above)-was less than 3% of the observed volume for
any link.
The reason the recommended PDF values provided
better results than the calibrated ones for the two
links (marked by * in Table 5.7), is the fact that
the best fit PDF value for each of these link is
closer to the suggested PDF value than it is to the
average calibrated one. However, the difference (in

terms of the above comparison criterion) between the

89
results obtained using the calibrated and recom-
mended PDF values for these links is very small.
This difference never exceeded 0.7% of the observed

volume for any links.
Consequently, one can conclude that, on average, the best
PDF values in the study area are 28 and 40 for low and
moderate friction, respectively. Furthermore, it has been
shown that these best fit PDF values provide some improvement

over the PDF values suggested by the TRANSYT-7F manual.

It has been shown in previous sections that, in terms of
reducing the discrepancies between simulated and observed
traffic flow patterns, the calibrated PDF values provided
slightly better results than those recommended by the manual.
However, this difference is only important if the consequences
of developing and implementing an optimal signal timing plan
differ depending on which set of PDF values are used. More
specifically, it is more important to determine if the
performance measures (i.e. delay, stops, etc.) would be
improved by using the calibrated PDF values over the recom-
mended ones in optimizing a network in the study area. It is
also important to determine if such improvement is large
enough to justify the considerable amount of time, money, and

effort spent in developing the calibrated PDF values.

90

To make these determinations, the performance measures
resulting from using different sets of PDF values in op-
timizing the studied arterials have to be compared and
evaluated. This comparison will indicate the significance of
the calibrated PDF values in improving the traffic perfor-
mance. Nevertheless, this will not be sufficient to draw
conclusions on the importance of the calibrated PDF values in
optimizing traffic along other arterials or networks in the
study area. The two studied arterials do not cover the
variety of network configurations in the study area. To
overcome this dilemma, the two arterials were used in devel-
oping a large number of hypothetical network configurations.
Since any network configuration is described by four principal
parameters, namely, length, volume, friction, and complexity
(arterial, or two-dimensional), different levels of these
parameters were used in developing the hypothetical networks.
These networks ‘were used. collectively in evaluating ‘the

significance of the calibrated PDF values.

5. P oc .
The following specific procedure was followed throughout
the sensitivity analysis:
1. The calibrated PDF values (i.e. 28 and 40 for low
and moderate friction, respectively) were used in
developing an optimal signal timing plan for the

network under consideration. To assess the

91

reasonability of the optimal timing plans provided
by TRANSYT-7F for the two networks (mentioned
above), optimal time-space diagrams are provided in
Appendix E for selected routes in the two networks.
The optimal signal timing plan (signals splits and
offsets) was simulated using the same calibrated
PDF values. Thus, the performance measures (i.e.
delay, stops, PI, etc.) resulting from implementing
the optimal signal plan were determined. For
convenience, these performance measures will be
referred to as P2412340.

Using the recommended PDF values (i.e. 25 and 35
for low and moderate friction, respectively),
another optimal signal timing plan was developed
for the same network. This plan will represent the
result that would be obtained by not using, or
knowing, the appropriate PDF values of the study
area.

To evaluate the non-optimal plan (part 3), the
network was simulated using the appropriate (i.e.
calibrated) PDF values. Consequently, the above
plan was developed using the recommended PDF values
and simulated using the calibrated ones. The
performance measures resulting from simulating this

plan are referred to as PM25&35°

92
5. Comparing the performance measures obtained in.part
four (P112555) to those found in part two (P1423340) ,
will indicate how much improvement can be achieved
by using the calibrated (i.e. the appropriate) PDF
values. IMore specifically, the following criterion

was used to determine the level of improvement;

[H Pnzsass ' PM28:40 ) / PMzauo } * 10° 1

Positive 'values of the above. criterion, means that the
calibrated PDF values provided better results (positive
improvement), and vice versa. This is true for all
performance measures except system speed. Since higher system
speed means better performance, negative values of the above
criterion represents positive improvement, and vice versa.
Three major parameters were studied in the sensitivity
analysis: friction, volume, and length of links. Each
variable was investigated on the maximum possible range that
could be achieved without violating any of the following
criteria;
- The value can not exceed the published TRANSYT-7F
acceptable range.
- The value will not produce a volume to capacity ratio
(V/C) greater than 95% on any link.
- The value will not.produce:a spill over situation (i.e.

queue longer than link) on any link.

93
To include the effect of geometric complexity in the analy-
sis, these factors were investigated over a hypothetical
arterial, and then over a two-dimensional grid network. The
characteristics of the hypothetical arterial and the grid
network are discussed in the next section.

Changing the level of volume and friction was, con-
veniently, accomplished via the usage of cards number 36, and
39, respectively. However, the model does not have such a
facility for changing the link length. Since changing the
length of all links manually is tedious, a FORTRAN program
was developed to accomplish this taSk. This program reads
the original data, changes the length of links by whatever
percentage needed, and then rewrites the whole data deck with
the new link lengths. This program is given in Appendix C
under the name "LONG".

There is an implicit assumption in conducting the
sensitivity analysis that the calibrated PDF values will not
change with different levels of the investigated factors. In
other words, the calibrated (i.e. best fit) PDF values found
for the original arterials will still be the best fit PDF
values if the volume or length of links are altered. This
assumption (which may, or may not be valid) is essential to

the conduct of the sensitivity analysis.

94
5.§.g Networks used in the analysis.

To cover a variety of link characteristics and traffic
patterns, the two arterials used previously in the validation
and calibration process were connected together to form a
single hypothetical arterial consisting of eight signalized.
intersections. Furthermore, to include the effect of network
complexity in the analysis, a two-dimensional grid network was
made up from the same two arterials. In this hypothetical
network, each arterial was used twice in an alternated way,
and they were all oriented parallel in the East-West direc-
tion. All intersections of these four arterials were inter-
connected from the North-South direction with hypothetical
links to form a grid network consisting of 16 signalized
intersections. A sketch of this network is given in Figure
5.4. The hypothetical arterial and the grid network are
referred to as the one and two dimensional network, respec-
tively. It should be mentioned that a minor adjustment was
introduced in the two-dimensional network when used in
analyzing volume sensitivity. This modification was to
increase the volume along a few links to 40 vehicles per hour.
Since the minimum volume accepted by the model is 10 vehicles
per hour, this increase (to 40 vph) was essential to study the

effect of reducing the volume along all links by 75%.

95

.xp03omc amcofimcmEMpIOBS ass «0 coomxm c.m obowwm

 

VP)
11 {New J1 twwllt:
RM?!

*T

 

 

 

 

 

 

 

 

 

 

A Q
1 E....% is
5
4
w
\ I

m
kg -VM/V AT
9
w
w
G \W AT
tQt§§Q/w A
w

UGQQ)? - No.

 

 

 

 

9
DP)
Iql

JL 6L

T‘

>\WHV efixflv sK‘fiV

96
5.5.3 Sensitivity results.

Although the behavior of all performance measures were
investigated in this study, a special emphasis was placed on
the behavior of the performance index (PI) throughout the
study. The reason behind this emphasis is the fact that the
performance index is the only criterion recognized and used
in the optimization process. Fnrthermore, the performance
index is, as discussed in previous chapters, a linear
combination of delay and stops to which all other performance
measures are related. More importantly, the fact that it is
a combination of delay and stops provides the user with an
average or overall assessment of the traffic performance.

' e t ' ks

Values of this investigated factor ranged from 45% to
260% of the original length of each link. The performance
measures resulting from optimizing and simulating the networks
with each investigated value are documented in'Tables 5.8, and
5.9.

Using the procedure and criterion discussed previously
in section 5.5.1, Table 5.10 shows the percentage improvement
in performance measures resulting from using the calibrated
(rather than those suggested by the manual) PDF's. A pic-
torial representation of this improvement is also provided for
delay, stops, and the performance index in Figures 5.5, 5.6,

and 5.7, respectively.

9‘7

.co_zu.n_.ou so ucaoc u:.o> lea

uu_4<u uu~.»»mz<¢h >3 uoucoesouog 03.0) mom

.co_uo~_l_udo
.zuuco. xc_. .ac_u_to oz» no..o_u.:s gu_g: touuow on»

uz<¢p

38

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.cuuco. doc_o_to we monoucouaod ucogoe._u so «b.3mot yo co_uo~_s_udo a.m odou—

a~.go~ «o.m~ oo.o.m. «who. ..s. ~m..n. no ._5<u. as a n~ baa 3.
~c.oo~ co.m~ .~.nom. ammo. .... ou..np no a<¢p mm a m~ “no xmq n.
os.ou~ ~o.nn no.a~o. oo~m. o.~. hn.ame as ..5<u as a o~ “no ~c
...~a~ oo.nn No.3.a. when. a... «a.mn. as =<¢p an a m~ “no go» .—
~o.oo~ .o.mn s..oqo~ om~m. ..n. gk.cm. no ._4<u o. a n~ “no a,
.o.oo~ oo.mn mm..q°~ anon. a.m. o~.°m_ on asap an a m~ “an use. 6
om.6.n m~..¢ so.-s~ «was. ~.~— ....6_ no ._z<u so a o~ sac o
kn.¢.n 6.... no.-¢~ n.-_ n.~. ...~q. no =¢¢b an a m~ “an non. s
oa.oo~ «p.56 oo.¢oa~ --. .... on..n. no ._z<u as a o~ “no 6
868 3.3 on. $3 o-t .... : 8.3.— 3 .3: an a 3 3o x8. a
.o.o.n ~n.o. n~..o~n mono. o.~. ~e.~q. as ..5<u as a o~ ago 6
mo.~on n~.o¢ om.oo~n muse. ..~. aq.~¢. mu z<¢p mm u m~ “ac xo- n
66.3," so..m o¢.no~n soak, ~.~. so.oq. no ._5<u as a o~ ago ~
mo.o~n an.oe ~m.mo~n omen. c.6F n...o. mu z<¢p an a m~ .ao noo~ .
..a. xtgsax. A_.. A.;\>. .>\ou.. ..;\>. luau. to;
xouc. vegan ..cou macaw >._oa >a.»a synced to .3..> .a_.oooa .o_.a.u.as .oz
oucasgowgoe laua>m .03; Eco*_c: .u>< .auoh o.u>u outaomu mom coma sausage sou
....au.< u< ...;

 

9E3

.co_uatn_..u sn acne. u:..> to;

"._5<u a.~.h»mz<¢p sn uoucussouot o:..> no; uz<¢p no
.co_uou_s_uoo "ado "a

.guoco. x:_. ..c_o_to as» .._.a_».as gu_gz fauna. ogp "a

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-¢.ncm oe.o~ mm.o~o~ nnmpn m.n— m¢.on~ no u_4<u co a ow ado o
pn.oem om.o~ op.omn~ -epn «.np no.¢n~ no 2<¢~ an a m~ ado ume s
no.00m so.~m ~n.~mon «mnom s.m— mq.m- on u_4<u oo a n~ ado o
as.oom oo.om ~¢.omon oo~o~ a.m— mo.o- on z<¢p mm a m~ ado aoow m
um.mom «5.0q nn.¢nom ~emnm 0.6. c«.om~ mo ¢_h<u so a o~ ado e
no.n~m ~—.~c «n.noom o~pnn o.ep so.mm~ no a<¢~ mm a m~ ado new. n
on.~om oo.oe op.oooo «mo~n a.m— om.-~ no n~4<u ow a c~ ado . ~
or.~om co.o¢ no.pooo no-m «.m— -.oo~ mo a<¢p mm a m~ ado xn nm~ —
A_ao Aagxlxv A_.o Ach>v A>xuoao aezx>v «come mom
nooc_ pecan .ucou coaum >-.oa >odoo caucus we o:.-> .d_cuaoo go_.a.u.:z .oz
oucuEtOCBom coua>m .oam ecocmc: .u>< .ouoh odu>u outaomu mod cash sausage you
areas»; "a “to;

...u.ucouv o.m odour

S99

..cs:.ou gueo mom pountn_.ou use gu_: ca.a .aco.a .us_uao ocu uc_uo.:s_a >n
...c.. ... to some». ..c_o_ro as. ao..a_u.:a gu_g: rezone use
.ogonssc so; «som>oad mosque:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o—.go~ oo.e~ ~o.—~mp «woo. q..— oo.—np no oq a om oq a o~ «F
om.¢o~ n~.¢~ om.o—mp m~mo. o... o~.nnp «0 cc a «N mm a m~ nae n—
no.pn~ oo.nn mo.o—np been" s.-w on.mnp as ow a u~ no a am ~p
om.no~ mo.nn «a.mpn— ~npmp a.—p no.0np as oe a o~ mm a m~ goo —«
nn.oon n¢.mm no.~mo~ «.nmp p.m. m¢.Fm— no ow a o~ oq a o~ o—
~p.oo~ on.mn oo.oqo~ omen. ~.np us.—m— on oe a om an a m~ noon 0
oo.~pn oo.—e no.~wq~ scoop e.~— om.~e- no ow a o~ cc a u~ a
oo.~pn «o.o¢ oo.p¢e~ who». a.mp uu.ne— an ac a ow an a m~ «on— s
nq.pon oo.~¢ ~s.o~n~ nun». m.p— so.~n— no co a o~ co 4 o~ o
om.—on oo.o< oo.o~o~ oonsp o.—— o~.nnp no ow u o~ mm a mN soup n
——.o—n on.c¢ on.no~n ~o~op ~.~p mn.o¢¢ as oo a u~ cc a o~ c
nm.~—n n~.n¢ «o.oo~n «snow o.~— n¢.o¢— as so - o~ mm a m~ xo- n
oo.opn on.om «o.¢-n nn—op «.Np om.n¢— no oe a o~ cc a cm ~
o¢.«~n o—.ov on.o—~n Fame. ~.e— cm.nop mo co a o~ mm a m~ xooN p
:3 22a: 2: 2.35 932: 72?. 5:. 5:.
xouc- noon» .mcou ocean >o.oo >o.oa goon. donuma .at_udo .uco_o .asmudo
oucoscowtoa ecum>m .03; stom_c: .u>< douo~ sauce. ecu oc.~o.:¢_o as» uc_c_auno to..d_u.:z .oz
o.u>u c_ non: o:.o> c. noun o:.a> guacodn noun
«:3; co_uo_:l_a not» poc_oubo «cannons «oceanoeeoau and no;
.u_toua< u< atom

.zuoco. .ocmomgo co monoucoutod «cueocw_p wo audzmoa co_uo.:E_m o.m o.no~

 

l()0

.Acsa.ou zucu no; pouaeb_.ou oz» gu_3 code doca_u .oamudo ecu oc_uo.:ema >o
..xc_. ... .o guano. ..c_o_.o as. mo_.a_z.:a gu_gx .o.u.. o;—
.atonssc so; ¢:o_>ota account

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

an.nem o¢.o~ o¢.o~n~ pnmpn a.m. «4.on~ no co 4 u~ no a n~ a
umq
ou..¢m o¢.o~ mn.ooo~ «mo—n m.np a~.on~ no o« a o~ mm a m~ s
om.~om oa.on oo.~oom onnom a.m. o~.-~ on co a o~ oe a o~ o
goo"
«..nnm Pm.on on.ooon nmoow ”.0. oo.«o~ on cc a n~ an a m~ m
50.nsm oo.u¢ «o.F—om nmpnn «.6. o~.om~ no so a cw o« a o~ q
non.
pv.osm oo.~¢ ne.noom msown 3.6— os.om~ mo oe a o~ mm a m~ m
oo.~om oo.oe om.oooo nmo~n o.m- qm.-~ no ow a om cc 4 a~ ~
an.nn~
so.~om oo.oe mn.oooo eco~n a.mp oo.-~ no cc a o~ mm 4 mm p
a.m. .gssxc ._.c ..=\>o .>\uo.. .Lg\>c c..a ca.a
xouc_ noon” .mcou cacao >a.uo >o.oc Agony .ocumu .ge_udo .oco_a dos_ado
ancestowaoa soua>m dose seo»_c: .u>< .auop guaco. ogu ocmuo.:I_n ozu oc_c_ouno ao_.o_u.:x .oz
o.o>u c_ to»: 03..) :_ now: and.) guacohn some
«can co_uo.:s_a sate ooc_oubo mogaooos uucaaaoceoau use can
xtoxuoz "a stag

.A.p.ucoov 0.m 0.30»

 

101

pence gauze. o.o>u _u¢_udo ecu cog» ucoto~._o a, uo:.o> can caucus-coo; u~.»>mz<¢p >5 nosey cacao. oduxo .cl_udo
.guoco_ .ac.u_ao c. oococo ac .o._ - anon anon

.>_u>_uuodmoe .ao:.a>
and nouoen_.oo use nousossooot ecu ac_n: >3 uoc.ouao no.o>u .ae_udo an a so “no:.o> not tau-en_.cu ozu u:_u: so

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

no
.ouo_ca0gdda ago: or. u.uoa nouosn_.oo ecu c. coco voucouuouog as» Lo>o to; uoaatn_.oo ocu oc_o: >u no
.xgoauoc .ao.uoguon>g .oco_¢cos_u.ozu ul_u-~ u.a_touta .uo_uoguod>z "I_u.p "b
.axcm. ..a *0 sauce. .oc_u.to ecu ¢o_.d_u.as sums: gonna» oz» no
oa.w+ -.—+ po.o+ ~o..+ o—.n + o~.~ + .oov cc a c~ an a m~
emu-~
Ann a any goo—c
oq.o- oo.o+ mp.o- no.—. o~.o + p~.o + I_u.p. cc a o~ an a m~
cc a u~ mm a m~
Amos
po.o+ hm.o+ mo.o- -.o- po.o + oo.o + a.m-. non.
ow a n~ an a m~
w¢.o- oo.o me.o. as.o- oo.o oo.o .mov ow a o~ mm a m~
l_o-~
Anew Non—
eo.o+ oo.o+ oo.o op.o- 50.0 + «n.o + lwu-p co a o~ mm a m~
3 a 8 m... a a
‘on a who
n~.o+ Fn.c+ no.0- ov.o+ Nm.. + p... + ¢_p-p. xo-
ac a n~ an a m~
co.o ~o.o+ oo.o «o.o- oo.o oo.o + «may cc 1 o~ an a m~ an.nn~
n_p-~
Ame a may
¢¢.—+ n~.n+ -.o- ¢~.o- ~m.qp+ mo.n-+ e_u-p. oq a n~ an a m~ noom
._ao
nonc— uoodm .mcou «dawn >0.oo >o.oo u:.o> oada>
oucasto»eo¢ sauo>m dose Igo~_c: .o>< .ou0p co_uotn_.ao us.h>mz<¢h Lo_.d_u.:x
xtoxuozn success
automoas oucaeLowaoa c_ acoso>oads_ oacucoueomo uo:.o> not
.guaco. .ac_u_to co «monocoutod acutocc_u su_3 metamooe cucaucocaoa c_ ucoeo>oads_ c—.m 0.3a»

102

.>.u>muoodnoe .ao:.a>

can nouacbmdou use poucaeeouoc as» oc_a: so noc_ouao ao.u>u doa.udo no a ow “aoa.a> can now-tn_.oo ogu uc_u: so
use 583 303 .228 of :2: 228:: a. 832, .2 22258.. $23.: 3 use £83 3&6 .828

.zuoco. .ocmomco c. oncogu 0: .o._ - «moo ouaa

 

 

 

 

 

 

 

 

 

 

“a
.3352er 0.8.- 0.3 «19. payment-u 05 3 «0.8 320388.. of ..96 3... 239538 05 9:3 3 no
.xtozuoc .au_uogu0d>g docosacasmn-o:u "smu-~ u.o_touao .ou_uoguon>z ue_u.- “a
.mxc_. ..a .0 guano. .uc_u_to 0:» ao_.d_u.:e go_s3 sauce. 0:» "a
-.o. «o.o. em.o- n«.o- oo.o No.9. “moo cc a o~ mm a m~
Imu-~
.mov nae
oo.o+ eo.o+ op.o- 90.9- m~.p + ~n.— + s_p-p so a o~ an a m~
oe a o~ mm a m~
Ache
no.o+ nn.o+ ~n.o+ om.o+ m¢.o + po.o + a_p-. non
co 4 o~ mm a m~
._av
xooc. panda .acoo ado». >o.oa >o.oa oada> o:.o>
occasLoCLoa saua>m .03; egow_c= .u>< .auo» comuatn_.ou mk-p>mz<¢p to_.dmu.::
£33.... 565....
mouse-o: oucaeLOCLoa c_ acuso>oeda_ oooucoutoau no:.-> mom

 

 

 

 

 

 

 

.a.u.ucouv o—.m o.na»

.mxcwfi mo :uwcmH .m> zmfioo Amuoo om ucwsm>ouosH m.m muswwm

 

Ihmu‘w... urn/Emma n6 N

 

8N9NO~N8N8_8_9_8_8_8 8 9.8

 

l03

 

 

85
;a&..
M
.stmw
1“
3:
Ladmfl
M”
l
LBS.1
UH
.85:w
1
mm
18d.
.. .a_Lmt¢
a. .. {0302 .84.

 

 

 

mas—A LO shezwa .9»
.5ij 423;? 2: :zwaweyagga:

 

 

104

.mxcwfi mo :uwcoa .m> mcoum EuOmwcs ow ucoEm>ouaEH

o.m upsmsm

 

152m..— n-HZHDHS n5 N

 

 

 

 

8N 9am CNN 8m 8. 8_ 9.— am_ 8_ 8 8 av om
. 4 . q 1 4 . . . 4 1 . 8.0—-
m ...oraztc .Aagv
. xto>ymz q.
. mm
.. 8 o- w
.A
.. w
.. 84.. N
.. 1.
Io!- m
.... .. 8.N.. E
. ..... ....
[Rwy/s“ .M \WVN I 8.0 S
4---- , .
8.N

 

 

was. a ”3 2.53m.-

meéhm gage—2: z=.bzwzw>amgz_

Jibr

 

 

.mxcwa mo suwcma .m> Hm cw ucoEo>ouoEH m.m ouswflm

 

IhmZmJIIfimHSnDN
QMQNQNNBN8_:_Q_=N_8_8 8 av om

 

 

105

 

 

 

1 8.0..

.\.QI q .\ 89D 2
\s I
X. 1 8.0 no
\ 0
..d. M.
. 8.. w.
. N
II-
a. . I
1 B _ N
d
I

.. 8 N .

.. _U_meu¢
.. ngme 1 and

 

 

 

muz_s lo steam;
.6» .e 2: szmzwsegmz=

 

 

106

It is clear from Table 5.10 and these figures that there
is no specific pattern or general relationship between the
improvement in any performance measure and the length of
links. Furthermore, there is an inconsistency in the results
from the arterial and the results from the network. It is
also clear from Table 5.10 that with the exception of the case
of "arterial at maximum length" (first row in the table), the
change in any performance measure was less than 4% in all
cases studied.

In the case of the "arterial at maximum length", where
the largest difference in performance measures existed, the
difference between the optimal cycle chosen (by the model)
with the calibrated PDF values and.the suggested ones was also
the largest. This difference was twenty seconds. In all
other cases such differences, if they exist, never exceeded
five seconds. The relatively large difference in the arterial
case at maximum length is mostly attributed to the difference
between the two optimal cycles chosen by the model with
different PDF values. Even in this case, the improvement in
the performance index, which is the criterion used by the
model in the calibration process, never exceeded 1.89%.

All of the above observations tend.to indicate that using
the calibrated PDF values does not consistently provide a

better solution for all levels of length investigated.

107

g: Volume along links

Values of this factor investigated ranged from 25% to
114% of the original volume along each link. Table 5.11 shows
the percentage change in performance measures resulting from
using the calibrated platoon dispersion factors at each
investigated value. As in the previous factor (i.e. length),
Table 5.11 shows no specific pattern or a consistent relation-
ship between improvement in any performance measure and
volume. With the exception of the case of "network at minimum
volume" (2nd row), the maximum change in any performance
measure never exceeded 3.29%. With the same exception, the
difference between the optimal cycle length chosen with the
calibrated PDF values, and that chosen with the suggested
ones, never exceeded five seconds.

For the network at minimum volume case, this difference
was fifteen seconds. This relatively large difference between
the two optimal cycles is the major reason why this particular
case possesses the largest change in the performance measures.
Even in this case, the largest change in performance index was
less than 3.8%.

C: 'c ' ve

The TRANSYT-7F model recognizes only three categories of
friction, namely low, moderate and high. As discussed in
previous sections, links with high friction characteristics
did not exist in the studied arterials. Hence, calibrated

PDF values were developed only for the low and moderate

lC)8

.~.m.m co_uuoo 00¢ "co_uu.g» no. sauce. ocm~>dnco c_ no»: use «son on» >.uuoxo no: a. use: pom: xeozuoc on» "a

.>.o>_uuoamug .mo:.o>

can vouaaa_.au uca noucossouo. ogu uc_¢: so noc_auna on.uxu .os_uao .‘ a a. uuoaso> no; non-tn..ou can uc_o: so

652 583 3%". .128 of 5:: 2233.. 3 832, .2 838.88; $26.3: .5 .83. 583 3&0 .828 u.
.oua_coocdno meal at: a.uoa nouaea_.ou one mm moco noocoesouoe ogu Lo>o no; pouaen_.au ogu comm: >o an

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.xaozuoc .au_uo:uon>s .oco_ucme_p.o:u "l_u.~ u.a.gouea .ou_uoguoe>z "a.m.. «a
ow a o~ mm a m~
“moo
oo.o+ Fo.o+ me.o+ —o.o+ om.- + mo.¢ + s_p-~ New.
oe a ow mm a m~
oe a o~ mm a m~
“moo
oo.o oo.o oo.o oo.o oo.o oo.o l_p-— app—
oq a o~ mm a m~
~..w- mm.p- m~.o- ~—.—+ oo.~ - o~.n . Ana a oova oq a on an a m~
I_p.~
“no a any. goo.
oo.o. oo.o+ mp.o. no.—. on.o + p~.o + e.p.p ow a o~ mm a m~
oo.o- o—.o- mn.o- om.o- oo.o - on.o - Ammo oe a o~ mm a m~
I_p-~
“moo xno
pp.o- no.o+ oo.o- o~.o- oo.o oo.o + l_p-— oc a o~ mm a mm
ss.n- op.m- oo.o- os.n+ pn.~p- Pn.~.- .oo a moo. ow a o~ mm a m~
s_p.~
Aooo xm~
o~.o+ m¢.o+ so.o- ~¢.o- oo.p + so.p + e_p-— oc a o~ an a m~
A_oo
xouc~ noon“ .mcoo ocean >udoo >a.uo o:.o> u:.o>
coca-3t: 239$ .2... Ea. :5 .o>< .33 5395. .8 {-5233 832,
xaoxuoza axc_.
moeauoas oucaetoctod cw ucoso>0gds_ oooucouauan mono.) no; ya u
.oe:.o> .ac_o_eo so mooaucootod newcome_u :u_3 nauseous occascocaod c. aco§a>otna_ —..m «on.»

109

friction categories. However, to increase the number of
values over which the friction level could be studied, a
calibrated PDF ‘value ‘was assumed for* the Ihigh friction
category. Since the calibrated PDF values were higher than
the recommended ones by three and five points for low and
moderate friction, respectively, it was assumed that the
calibrated PDF value for high friction links would be higher
than the recommended one by five points. This is only a
hypothetical assumption, and should not be considered as a
more appropriate PDF value of high friction links in the study
area.

By using combinations of these three categories, five
friction levels were investigated as shown in Table 5.12.
There is no systematic pattern or general relationship between
improvement in performance measures and level of friction.
This conclusion is true whether'we include or exclude the last
two levels (i.e. level 4 and 5), where the calibrated PDF
value for the high friction links was assumed and not deter-
mined. It is also clear from Table 5.12 that the change in
performance index never exceeded 4.27%.

This percentage change in.performance index (i.e. 4.27%)
is the largest value obtained in the sensitivity study. To
determine whether such a value has practical significance, a
simple experiment was conducted. In this experiment, the
arterial case with its original parameters was optimized

twice. The only difference between the two optimization runs

l 10

mod 00.005..00 use u0uc0laou0g 0:0 no.0: >9 00c_0ubo 00.u>0 .05_ueo .0 a 0. “00:.0> mod 00

.Ouflu fluln «ii

.>.0>.uu0d00e .00:.0>

000n_.0u 0:0 oo.o: >n

acne. guano. ...». .....ao 0:. gag. .co.o.._u ._ .oa..> 1°; sauce-sou». .-.»ma<¢. so acne. guns». o.u>o ..a..ao "-

.>.co 0:.0> possum. c< "-

.00-.0aocaa0 0.05 0.0 0.1oa tau-Ln..0u 0g» »_ 00cc 000:0550000 0:0 L0>o mom p0.003..00 0;» oc.»: >o an

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.xaozuoc .0u_uoguon>: .0co.0c0§_u-oxu "l.p- u.a_t0uga .00_u0300d>: "6.0-» "0
m»-n+ no.o+ no.o+ om.»- oo.o»+ oo.o»+ “no a no.
s_u- -zo.: ..<
Ame a mm. mm. on .n .0>0.
»o.n+ »».»+ sn.o+ 00.»- em.o»+ oo.o»+ a.m-».
»n.»- ~».- ~».o- o~.no om.m - om.m - amo a on.
I.u- so.c no. .00:
.ma 0 mm. mm. a a. an a an .. .o>u.
«0.»- ow.»+ on.»- »o.o- e».~ + mo.~ + t_p-»a
oo.o- m».»- s».o- »o.~+ nn.n - ~0-n - “no a mo.
Imp-w. .005 ..<
“no. o0 an .n .0>0.
no.»+ no.o+ »o.o+ »n.»+ 00.» + no- + I.u-»
oo.»+ -.»+ »o.o+ ~o.»+ o».n o os.~ + goo.
I.p- .006 0:0 so.
“no a oova ox a o~ an a m~ .~ .0>0. a.
oq.o- oo.o+ m».o- no.»- o».o + »~.o + s_p-»
»-e+ go.n+ oo.»+ om.o- oo.o + oo.o + .oo a coo.
l_u- so. ..<
Ana. o~ m~ .» .0>0.
~o-o+ »Q.o- oo.o- e».»+ Mm.» - so.» - a.m-»
5.0.
x00:— 00000 .mcou 00000 >0.0o >0.0o 0:.0> 0:.0>
00:0Icooa0a suuu>m .03. Ico»_co .o>< .000» oo.o-tn_.0u 1»-»»mz<¢» .0>0.
80030020 co.»o...
00030006 00casco»eod c. ucoso>00au_ 0000:0050mn 00:.0> moo
.co_uu.g» »o 0.0>0. u:000»».p n».: 0003000: oucoeco»aod c. uC0E0>otde. ~».m 0.90»

111

is the node (i.e. intersection) order used by the model in the
optimization process. In the first run, the nodes were
ordered as they were faced by the northbound traffic. Thus,
the model first optimized the most upstream (external) node
in the northbound direction. Then, moving in the same
northbound direction, the second node was optimized, and so
on. In the second run, the nodes were ordered as they were
faced by southbound traffic. Consequently, the node order of
each run is a mirror image of the other, as can be seen in
Table 5.13. This table shows that just reversing the node
order produced a 3% change in the performance index. Theoret-
ically speaking, using either node-order is equally correct.
Recalling that 4.27% is the maximum change in the performance
index encountered in the sensitivity study, it appears that
the model is relatively insensitive to volume, link length,
and PDF values when compared to the 3% that can result.by just
reversing the node order in the optimization process.

It is interesting to note that the negative changes,
which were very frequent throughout the sensitivity study,
are contradictory to common sense. A negative change means
that if two optimal signal timing plans were developed for a
given network, the first using the wrong PDF values and the
second using the correct ones, the traffic performance
resulting from implementing the first optimal plan will be

better than that resulting from implementing the second one.

1 l2

.000 .h 0:0 ~» 0000: 0003000 0. 0 000: .0.. .0.0.0 0:0 c. 0000: 00 00000 000x0 0:0 0. 0.0»

.0030“ 0:0 00:0 000000 0. 00.000000 003002000: 000 c. 000000.000 00000 00.0 0:»
.03.; 9585302032. 053502. . 3.8538

.000000000 00000>00 0:»

.x0300 0:0 0300030020 0003 00.000000 0:»

6.00!

 

 

 

 

 

 

 

oo.n+ «0.0+ oo.»+ 0m.~+ oo.n+ s0.n+ 300 0:0000 0:0 000.0
a 0:0 00 00030003 00000000000 0003000 000000000o a

00.0o~ 00-mn nm-00o~ nmom» o.n» 0~.om» 0:300 ~» .0 .0 .» .eu .n~ .- .»~ mm a m~
003000

»s.»o~ mm.0n o~.o»o~ anew» 0.~» 0o.m0» 0:300 0~ .- .n~ .¢~ .» .0 .0 .~». mm a m~
502.0

0.0. .5300. A... .0:\>0 a>xo000 a0s\>.

x000. 00000 .0000 00000 >0.0o >0.0o 000000000000 0. 000

00000000000 0000>0 .030 0000.03 .o>< .000» 00.000000 00000 0002

 

 

 

 

 

 

 

 

.00000 .0c_u_00. .0_00000 .00.00:000>: 0:0 00.0.5.000 c. 00000 000: 0:0 0:0000>00 00 00.3000 n».m 0.00»

5.5.4

113

Conclusions

The general conclusion of the sensitivity study can be

summarized as follows:

1.

There is no specific pattern or general relationship
between the change in performance measures and any
of the investigated parameters.

The change in the performance index, which is the
most important performance measure, never exceeded
4.27%. It has been shown that even this maximum
percentage appears to have no practical signifi-
cance.

With.a'very feW'exceptions, the changes in all other
performance measures were less than 4%.

The exceptional cases existed at the extreme values
of the investigated factors. For example, minimum
volume in the two dimensional network and maximum
length in the arterial case. In each of these
situations, the difference between the optimal cycle
chosen (by the model) with the calibrated PDF values
and that chosen with the suggested ones is rela-
tively large. Consequently, it seems that the
change in the performance measures is mostly
attributable to the relatively large difference
between the two optimal cycles chosen by the model

with different PDF values.

114

From the above four points and from the detailed dis-
cussion in the previous section, it appears that the changes
in performance measures are within the accuracy of the model,
and theme is no evidence to consider such changes as sig-
nificant "improvements" resulting from using the calibrated
PDF values over the suggested ones.

In summary, the sensitivity study tends to indicate that
the calibrated PDF values do not provide significant practi-
cal improvement in the overall traffic performance, regardless
of link length, volume, friction or complexity of the networks

investigated.

CHAPTER 6

SUMMARY AND CONCLUSIONS

Traffic optimization is the act of developing a signal
timing plan in which the signalized intersections in a given
network are operated to minimize delay and stops. Several
studies showed that this minimization of delay and stops in
a network provides a convenient driving environment, improves
the network capacity, and reduces excess fuel consumption.

The main objective of this study was to select a candi-
date tool for optimizing the traffic in Saudi Arabia, and then
to test and- assess the applicability of this tool to the
traffic conditions in the cities of Dammam and Al-Khobar,
Saudi Arabia.

With the increasing complexity and magnitude of urban
signal networks, manual traffic optimization is an impossible
task to perform. An exhaustive research of the literature was
conducted to identify signal network optimization models. The
features of these models were compared, and it was concluded
that the TRANSYT-7F model is the best candidate for applica-

tion to the traffic in Saudi Arabia.

115

116

In any optimization or simulation model, there are a
number of parameters (constants) that represent the driver
performance characteristics in the country where the model
was originally introduced and calibrated. It is well known
that such characteristics can vary significantly from one
society to another. Therefore, the successful utilization of
any traffic model depends on selecting the proper values of
the parameters that describe the driver performance charac-
teristics in the area where the model is to be used.

In the TRANSYT-7F model, these parameters are average
vehicle spacing, start-up lost time, extension of the green
phase into the clearance interval, and saturation flow rates.
In addition to these parameters, the platoon dispersion
algorithm used in the TRANSYT-7F model has to be calibrated
for the traffic conditions existing in the area where the
model will be applied. This algorithm portrays the need of
individual drivers to maintain a safe and comfortable headway
as they progress along network links. Hence, the platoon
dispersion algorithm is also affected by the driver perfor-
mance characteristics. Consequently, to assess the applica-
bility of the TRANSYT-7F model in optimizing the traffic flow
in the cites of Dammam and Al-Khobar, Saudi Arabia, the above
four factors had to be measured in the study area and the
platoon dispersion algorithm had to be calibrated for the

local traffic conditions.

117

To select study sites for this analysis, the entire road
network in each city was investigated. One criterion used in
selecting the study sites was that the chosen networks have
only one cycle length or a multiple of this common cycle
length. Since the TRANSYT-7F model can not be used to
simulate traffic (and consequently can not be calibrated) in
a network possessing different cycle lengths, this criterion
was essential to the conduct of this research. Only one
arterial in each city was found which satisfied this require-
ment. Each arterial consisted of four signalized intersec-
tions with four approaches in each, three lanes in each
direction with curb parking, and they were both located in
areas of mixed residential and commercial activities. All
intersections had 120 second cycle lengths divided into four
separate phases.

The required traffic and operational data for both cities
were collected near the morning peak in the summer of 1988.
Other physical and geometric data were collected either in the
early morning, or late in the afternoon during the same
period.

It was found that the values of the extension of the
effective green and the start-up lost time were three and two
seconds, respectively. The average value of the saturation
flow rates of both cities were 1750 vph and 1670 vph. for
through and protected turns traffic, respectively. The

average vehicle spacing was seven meters. Comparing the

118
values of these parameters with those documented in the
TRANSYT-7F manual for different categories of drivers indi-
cates that the drivers in the study area can be classified
somewhere between normal and aggressive. Therefore, the value
of these parameters are not very different from what is
usually encountered in the United States.

The calibration of the TRANSYT-7F model consists of
determining that value of "PDF” (0) which when used in the
platoon dispersion algorithm, produces best agreement between
the simulated and observed flow profiles. To accomplish this,
the observed arrival flow pattern (profile) was obtained for
every link of the studied arterials. Using different values
of "PDF", a large number of simulated flow profiles was
obtained for each link. With the aid of a FORTRAN program,
the value of "PDF" which minimizes the value of the absolute
difference between the observed and simulated flow profiles
(i.e. best-fit) was determined for each link.

The average best-fit PDF values were 28 and 40 for the
low and moderate friction links, respectively. On the other
hand, the TRANSYT-7F manual suggests a value of 25 for low
friction links and 35 for moderate friction links. The flow
along all the links were simulated using both sets of "PDF"
values and the resulting“ profiles ‘were compared. to the
observed ones. It was concluded that the average best-fit
"PDF" values provide some improvement over those suggested by

the TRANSYT-7F manual. More specifically, the improvement in

119
terms of reducing the value of the absolute difference between
the observed and predicted flow profiles was less than 3% of
the observed volume for any link.

To determine if the consequences of developing and
implementing an optimal signal timing plan differ depending
on which set of "PDF" values are used, a sensitivity analysis
was conducted. In this analysis, a large number of hypothe-
tical networks were coded using data from the two studied
arterials. To do this, the two arterials were connected
together to form a single hypothetical arterial consisting of
eight signalized intersections. Furthermore, to include the
effect of network complexity in the analysis, the same two
arterials were used in making a two-dimensional grid network
consisting of sixteen signalized intersections. Following
that, the length, volume, and friction of links, in both the
arterial and the network, were varied systematically (one
factor at a time) to produce a large number of hypothetical
network configurations. For each.hypothetical configuration,
an optimal signal timing plan was developed using the average
best-fit "PDF" values. This plan was then simulated with the
same "PDF" values. The performance measures resulting from
this simulation reflect the consequences of implementing an
optimal signal timing plan developed with the average best-
fit "PDF" values. For the same network, a second optimal plan
was developed using the suggested "PDF" values. This plan was

also simulated using the average best-fit "PDF" values. The

120
performance measures resulting from simulating the second
optimal plan indicates the consequences of implementing an
optimal signal plan which was developed using the suggested
"PDF" values. Comparing the performance measures of the two
simulation runs is an indicator of how much improvement can
be achieved by using the average best-fit "PDF" values.

It was concluded from the sensitivity analysis that the
change in the performance index, which is the most important
performance measure, never exceeded 4.27%. To prove that even
this maximum percentage has very little practical signifi-
cance, an experiment was conducted. In this experiment, the
hypothetical arterial was optimized twice. The only dif-
ference between the Optimization runs is the intersection
order used by the model in the optimization process. It was
found that just reversing the intersection order produced a
3% change in the performance index. Comparing this 3% change
to the maximum change in the performance index encountered in
the sensitivity analysis (4.27%) clearly showS‘that‘the‘change
has little practical significance.

It was also concluded that, with a very few exceptions,
the change in all other performance measures was less than
4.0%. In each of the exceptional cases, the difference
between the optimal cycle chosen (by TRANSYT-7F) with the
average best fit "PDF" values and that chosen with the
suggested ones is relatively large. Consequently, it seems

that the change in the performance measures is mostly

121
attributable to the relatively large difference between the
two optimal cycles chosen by the model with different sets of
"PDF" values.

In summary, this study indicates that there is little
value in developing a calibrated set of "PDF values for use
in the cities of Dammam and Al-Khobar. Since this study was
conducted on a small sample of the road network in these two
cities, this conclusion should not be taken for granted in
other networks in Saudi Arabia. It is suggested that similar
work be done in other major cities such as Riyadh and.Jeddah.

Working with the TRANSYT-7F program in this study
demonstrated that the program is easy to understand and use.
Consequently, its introduction to traffic engineers in Saudi
Arabia should not yield major problems. The program is
flexible and can be applied to almost any network configura-
tion in Saudi Arabia. However, its data requirements are
immense, especially since traffic and operational data are
not readily available for the user. Presentation of the
required data in an acceptable format to the program is
another tedious task. Nevertheless, the TRANSYT-7F program
is still the best network optimization tool and its usage

should be encouraged throughout Saudi Arabia.

APPENDICES

APPENDIX A

NETWORK DIAGRAMS

l22

 

 

.00000m 000< .30n< wcfix mo :000xm _.< 003w00

QkQ V\% W\ W\.\\\v®\mv\0\ anRNV
>\\<Q\Vh- .nvx-VNRKRVK VVR k Qx<
NNQANW SN \Qox

 

I») L-

 

 

sxkhfi. 11.>&¥$. 0\‘GN _

 

 

 

 

 

 

.- r». J v»- u»- _
\\\-¢-§ 11 0.x 90. 11§m‘\ A.“

123

.A50860av 00000m 00000 «c £000xm N.< opswfim

mNVkQVv-ux W§\0\W¢\0\ hWQN-V
\<\4 Q\\h1 .01 NQRK Qx-VK VVR k QR
1V KAN“ Q\ x Q?

 

L1 L1 1P1 L

 

_ {fin-h. _ tfixm-

00...:

 

 

 

 

 

 

 

1.1 thm- 141 {Wm 1.1 tu“ 1.

 

 

 

 

 

1»- —1 L1 L

 

wfinhfl.

0500 11 23101-4 000.00 1.0.300.

EN

.0000 00300 050 :0 00003000 000:0 .0000>H m.< 003w00

 

124

 

 

$0 0.00
0.00 000 .000 1
NM

§~ saw 080 0.000
"mall-1'.

“QM NsNN 8N QxNN. ¥\§hu$(\\

30.40 /
2- S- 0“. on tuznkgq
\ n m. \ 72-00%
M

«00.038000 Mm. kwxxVM-QES

 

 

 

 

 

 

 

 

APPENDIX B

SPEED DATA

125

.030chomu 000 mcHumon am

.0.

 

 

 

 

 

 

 

 

m-HH mm 0H nuflom nomm NH 0 505500
H.NH 0m 0H suflom nomm @ h EMEEMO
H.NH mm 0H Suflom nova h H BMEEMQ
m.HH om 0H nuhoz Honm H h Enfifima
N.nH mw 0H SUMO! Homm h m EMEEMQ
m-HH mm «H SHHOZ HOHm 0 NH EMBEMG
H0>00020 Ammx. 00H0500
0000000300 00000 00000 no umnaoz
0mm 00030 003532 000000000 x000 mcoHuoomumucH >000

 

 

 

 

 

 

 

 

.0000 >0300 0:0 :H 00000m H.m 0Hnma

.000000000 000 00000000 >0 "4

 

 

 

126

 

 

 

 

 

¢.NH 00 NH zudom nomm «N I nN 00non¥IH<
n.NH 00 NH £000m mama MN I NN 00QOSXIH<
o.NH 00 NH nunom nOHm NN I HN 00QO£xIH<
m.NH 00 NH 00002 Hoom HN I NN 00QO£MIH<
m.NH n0 NH 00002 HONm NN I nN 00nonle¢
m.NH 00 NH 00002 Hovm MN I «N 00nonle<
H0>00000 00000 0000000
0000000000 00000 00000 00 000002
«mm 000:« 000002 000000000 0000 0000000000000 >000

 

 

 

 

 

 

 

 

.A.0.0:oov 0.0 00000

APPENDIX C

LISTING OF THE FORTRAN PROGRAMS
USED IN THIS STUDY

127

.0000000 3000
no>stnao .00 00000000 >00§S£CE L0; .0000 .04000 #00
.DILflaun 0002 a 0002 0003000 IU£l000500 0:» .0000
.0002 a 0002 0.03000 lac-£00000 00000000 .0000
.0G>Ll0£0 coal :0 00000:!) 0l>Lllao mo 0.0032 .0002
.0fl>&t0£0 £01. 00 0.0000!) 0.0003000 00 0.00:2 .0002
.00>Ltuc0 £000 00 .0000 0.00000) .00
00 000000 00000500.. .0005». DI000550I 00 000532 .m>02
.Llnssc 01>L00£0 .chzn
.xsluaonm

.0003000 0:0 .hno .N
.00: 0000000 '20 £00 .000 .0

.n-000 030030

.00.:

0O. .000 ulna—000 0000 .UO0L00 xusun 000 LD>O
00>000£0 £00. £0 00000:.) no 00005: DO>LU000 Ink .000 .m

.Lonssc x000 0:. 0:00) 000 00030000 0000 .0IUOE

h>mz0zh I00 >0 0.000000 II 3000 3000 5:00:08 000

Inuxu 000 LO 0I>000£0 :00. so» 200000>I 0100000)
00 c 0 .000 0.0000.) 0:0 90 000000 000003000 Ink .000 .N

.000000 >030! I00 LI>O

Iauxu .00 $0 00>00000 £00. £0 nt>ttnno 00000:.)

no 0002:: 00 0000000 0.0003000 00» 000>£OU

00 00>200. $0000» CO0ILO>COU In» 000 .00000000

0:» on.uao0a >00ascas so» .4000. 00.0: 000300;. .400 ..
.0000» 030:0 DIL03000

*******¥****************¥****

.lllUOta
* 00000000000 .00 £0 0.03 no: 00 .:m~400= $000000 '00 00 00:0
ttttit##0##titttttttttttttttttttttittttttttttttttttttttttttttttttttt
# t
t 00400 t
0 *
0*ttittttttttttttttttttttttittittttttttttttt*ttttttttttttttttttttttt

*
****************************¢*#

128

onczam.

o.ouaz:0

o.on4coma

o.ou40000

0.0I000020

o.onuoom

0.01000020

o.onuam¢

o.ouun

o.onzoaz

on0>az

oncoaz

0:002.

oux

on.

0312.4

onuxct

onuan

o.on4000
.2.*¥¥*#*¥**#******tttttttttttttttt*t$¥ttt*tttttlttfitttttt$*¥t**t2:32:82.

2.202.103.000. .oca.nm4.u.n. zmao

..302.I030¢00. .030.Im4.m.¢. zmao

..a4o.u0:0¢00. .00¢.Im4.u.0. zmao

..040.I0:0¢00. .mca.lm4.u.0. zmao

2.040.103.000. .400.Im4.u... 2000

u>200.40000.4000:.uaom:0.uoom.uonc:0.uamc.uo.4000.¢0nz 0.4000

man 000002.

00thmmk0a cmh00¢010
*ttttttttttttttttttttttitttttttttttttttttttttttttttttttttttttttttttt

it.titttttttttttt.ttttttttttttttt#tttttttttttttttttttttttttttttttttt
.04000 000 00. yo 8:0 .010000 23000 20 000
.0000 00. so 0:0 .0000 300 £30

.lomd 00. $0 53$ .lknfi 000 £30
...—.0. 030 .>00. .00 0..30.0 .0£.£.ps0a .04000 000

..0000L0 3000
us»..:s.n on» ac.»ao.a 0...:cas so» 0.000 .04000 man

*****#
*###**

129

em..nx h on
.n.oz.30¢
..940.Imapcrm. .oma.n04.u.n. zmao
..a00:.nm:pahm.n.00040
090020.0900202000.c.0p.¢3
0:2.»200 mm
0:2..200 .
quma.4¢umc..o..n.0k.¢3
mom.Jcoma.Jcomc.maam.0900.cu¢2.¢oaz.m>az.zhz..nnn.¢.0r.03
acumxmoazu4¢uma
Jammxzu¢2u4¢uma
uncm+0aomzmuuacmzm
0¢*0anuncm
unma+uomazmuuomcxm
.ma.mm¢nuomc
Acumxuouuom
muazu.oo..*¢u¢2.nuo
mocz+¢zam.|¢zam.
moaz+azamuazam
0>zoo\.m>a2*0x¢z.|¢uaz
mo¢z.zhz..ooaa20.nnn.n.o¢0¢
m>nz.oono20.¢¢¢.m.oc0¢
c0..|. . on
#1:!!!*******#*##********#***$.32?!*#*****************#**********$3.393...
.ooo.¢.0p.¢3
Jcom.uxcz.uoa.xz.4.mmm.¢.0h.¢3
p0mhca.oo..¢. 0h.¢3
h0mpco.oo..n.a¢0¢
22.4.mxcz.una.molazm.mmm.m.oa0¢
0:2.»200 mm
u>zoo.4¢om.mmlozu.......ocmm
*##**#*#¥#it.it.tit.****#**¥***¥****t*tttttttttttttttttfitttttttttttit?!
.n.nz.30¢
.m.nz.30¢
...az.30¢
t.t.***.¥********#*******#************************3*8.33.8.3...*****#***¥**.3.

130

t :0.400: 2000000 00 020 t
8.85.8.3...8:3...3****¥****.¥***$*#$3*¢**#3:.fittttitittzltttittt********#******
****#***#$¢#8:!#****#*t§.*ttttttttttttittt.*tt***#**#*$tt##titttttttttt

020

0000

..no.o.zo> ..0.m.0..uac-uosaug 00:. 0x. . 0000 04¢00..x.\.+
0...:040 x02..x.\.0...000..x.\.0... c 22.4..x..002000 000

..04000 0.0..xn.+

.04000 000..xn..04000 000..x0..0000..x0..0000..x0..¢0¢2..x0.+
.0002..xn..0>02..x0..002...x.\v0¢z¢00 000

.o.o.0..n 00.0 000 200..xm\\.+
o.0.0..n 00.0 000 :00..xm\\.0a:000 000
.o.o.0n.0.o.00.0..m.00.n.0.0az¢00 000
.0.00.. I 0000 0000.0000 00 000202..x0.002000 mom
.0... I 0000 00>¢0000 00 000202..xm\\\.002000 .on
.0..00:¢00 00¢
.0.m.0¢zmo0 nnn
.0.n.002000 000
.0.0.00.0¢z¢00 ...
.c.¢.x..0azzo0 oo.
.0... I 010000 23000 2. 0.0..x0\.002000 no.
.0..x..0..002000 00.
.o.o.00.002000 .o.
it.****¥*******t.it#*#******#¥***¥*¥**¥**¥****#***¥#***#**#*txltxltttttxl

002.0200
002.0200 0.
*tttfitttfitfittttttttttttttit$$¥t¥t¥¢tt$¥$¥t$t¥¥t¥t$¥¥tt¥¥t$¥t¥**#*#**

0200.mon.¢.00.¢3

0200...on.¢.00.03

200..no..¢.00.03

NDZMPZOU bk

WDZ~PZOU b
mm¢u+23multflmm
Alanvmmdlmm¢~
QnIGHIKQH

0..0..00n020.00..n.0000

l3!

002.0200
002.0200 0
000.0.....n.00.03
000.0...u020.....0.0000
002.0200 00
004.00n020.000...0000
“‘002.0200
***#****###tttttttttittttttttt****#*******0.0.0:!*ttttttttttttttttttttt
.m.0z.300
...0z.300
o.ou0zJ
.88....8.*************t8:851!!!t.......#85:?!******#******************#****it.1:!!!
..302.|000000. .0000.I00.0.n. 2000
..000.u000000. .0000.|04.0.0. 2000
..000.n000000. .400z.|00.0... 2000
00.020 0:0000
0...n...¢...0...0......+
.o...0..0..0..0..n..¢..n..0.....000 000002.
o0:000.0 000000010
it,*itttttt*itttt.*****¥¥¥**tt$t¥¥t0.3.3.0.?!*******##**¥*#*¥****#¥*0.0.8:?!

# ..uhlh>mz¢¢h £00 at. .000 tic. 00.0000 0:0 .thZ .N t
t .00: E‘LaOta In» Lou .h¢n¢ .. t
t .00..0 000000 0
t t
t .0000 uhlb>m2¢mb .Oc.u.Lo Ugh. .bdoo .m t
t .nxc.. no acacia *
t .acau.Lo 0:» ut.~0.0.08 £0.23 £00000 0:0 .4002 .. t
t .....0 0009. t
t t
* .nx:.. 0o ... 000:0. :
# Ito acaaCIzu L00 won: an: a. .:OZOJ: Edsaosa 0:» 0.20 #
tt##0####0##:##0##:titttttttttttttttttttitttttttttttttttttttttttttit
# t
t 0204 t
t t

#*********.¥******8.8:!************************************************

132

002.0200
0...0...c...0...0......+
.0...0..0..0..0..0..¢..0..0.......1020.000.0.0000
002.0200

0.010.0
0.01n.0
0.01¢.0
0.010.0
0.010.0
0.01..0
0.010.0
0.0100
0.0100
0.0100
0.0100
0.0100
0.01:0
0.0100
0.0100
0.01.0
010..
010..
010..
010..
010..
01...
010..
010.
010.
010.
010.
010.
01¢.
010.
010.
01..

I33

040.0.001020.....0.0000
002.0200
.0.0z.300
..000.1000000. .0000.100.0.0. 2000
..302.1000000. .0002.100.0.¢. 2000
..0000.1000000.0.00000
..0000.1000000.0.00000
..0000.1000000...00000
002.0200
002.0200
0 00 00
002.0200
0.0.0.0.¢.0.0.0.0.0+
...0.0.0.00.00.00.00.00.¢0.00.00..0.c¢¢.0.00.03
002.0200
0..¢0..10.0
n..*0..1n.0
0..*0..1¢.0
0..¢0..10.0
0..¢0..10.0
...:0..1..0
0..¢0..10.0
0.:0..100
0.00..100
0.:0..100
0.:0..100
n.*0..1n0
0.:0..1¢0
m.:0..100
..00..1.0
002.0200
0. 020
0.:0..100
0000
.0.*0...¢000100
2000.001 .00.... .00. 00 .00. ...0.

I34

0.0.0.0.¢.0.0.0.0.0...0+
.0.0.00.00.00.00.00.00.00.00..0.001020.¢¢¢.0.0000
0.010.0
0.010.0
0.010.0
0.010.0
0.010.0
0.01..0
0.010.0
0.0100
0.0100
0.0100
0.0100
0.0100
0.0100
0.0100
0.0100
0.01.0
010..
010..
010..
010..
010..
01...
010..
010.
010.
010.
010.
010.
01¢.
010.
010.
01..
002.0200
000.0.....¢.00.03

00

135

*****#*******t*ttttfit************************t_**************¥*******
* =0200= 2000000 00 020 *
**********$*¥**3..8.#***************************t******************¥***
8:3.t..3.t.##2##******#*****t***¥t**¥t*3.*********#****¥#0....*****#**********#
020
0000
.0.000..00z000 ¢¢¢
.n.0..002000 000
.0.0.0.00:000 000
.000.x..00z000 ...
**********#*********0.1.3....*t***********¥**¥¥t***t***#*****¥#***$****#*
..0000.1000000.¢.00000
..000000.1000000.0.00040
002.0200
002.0200 00
00 00 00
0...n...¢...0...0......+
.0...0..0..0..0..n..¢..0..m.....000.¢.00.03
0.010..
0.010..
¢.01¢..
0.010..
0.010..
..01...
0.010..
0010.
0010.
0010.
0010.
0010.
¢01¢.
0010.
0010.
.01..

APPENDIX D

GRAPHS OF "SUM OF ABSOLUTE DIFFERENCES
VS. PDF VALUES"

.smEEmo ..o.a 0:.. .0 =o..m.n..mo ..0 0.50.0

 

W04"; “an.
00 mm... 8 mm om m. o—
. w 4 . . . 8 _

 

136

 

SBONSUBi-JIU 31010888 :10 HHS

00.00100 030 043.0090

 

mKN

 

 

.20 .0... 00 220000.03
202200

 

 

.EmEEmc ._o~o xcMfi mo newumunw~mu ~.o euswwm

 

 

137

 

Wan-....) “HE
ow mm 8 mm cu m. o.
4 q 4 . 4 1 o:
.I- .8. Wu
H
fl
. _
.k if. 8 w
M s
I\ .9; m
I\\\ a
\ 18. m
M
. L 8— W“
N
«:0
co_._m:..o EDm mfiiomnc . o: S
8—

 

 

 

.9mm ¥g=4 go zeapaga_4qu
zazzaé

 

 

.smesmo ..omm xcfifi no coflumhnfldmu m.a muswfim

 

I38

 

qudg> mam
my“ cc mm Amy mv HUV mm Hz” mm .Hw m. OMH:

 

lJ-quhngo-IH/ . 8.

.\.... /
/

.Qmm

Lomm

SSONQHBdflIU BLHWDSBH £0 HHS

cotmu ...o Sam 04.3.80;

.nmm

 

 

 

.9nm ¥s=4 go agapaga_dqu
zazzaé

 

 

.Emesma .mocm xcfifi mo :ofiumuawfimu q.n «hawflm

 

mqfiik> “Hr;

am we av mm an mm om m—
‘ ‘ . 4 . . + mm—

 

139

 

..\!...-\Aa/)nfi . B.

/ .3
/

mm—

1

.nb_

SSSNSHBdfiIU 31010883 d0 HHS

C0_gma.LU Ejm mw3.omnm
MK—

L

 

 

 

mega ¥s=4 mo ae=paga_4«u
zazgaé

 

 

.EmEEmn .MOmm xcfifi no cofiumunwfimo m.a musmfim

 

140

 

 

wins; ”an.
B 8 me Q on 8 mm ON @—
XJ ‘ q q q 4 ‘ o__
.2...» é. s
H n
”w
\ / LB— 0
4.
..\x it! to: w
ea
nu
..Dm. mm
1
// .8. M
.2. mm
..fi
18. ....Nm
575.1..0 EJw magomnt I 1 8— 8
8m

 

 

 

mama gz=4 mo EQEPaga_4«u
zazzeé

 

 

141

.EQEEmo .moom xcfifi mo coflumunwfimo

0.9 mhswfim

 

Wants. ”Ba.

9 on 8 mm. om @—

o—

 

 

ma

4

. a? 4 q
\ /

co_._m:._o EDm mfigomnum

1

1

 

 

 

990m gs=4 LO ze=hdgfl.4¢u

242245

8—
o:
8.
R:
Q.—
8—
8—
a:
8.
8—
8m

SBONSHBdflIU 31010888 :10 HHS

 

 

.uanosx-d< ..oqm xcflfi mo cofiumhnfifimo “.9 muawfim

 

ow mm 8 Wu cm 9 o.

 

I42

 

x ,/

H

/ 18m

SBONSHEd‘dIO almoseu :IO HHS

52.5.2.6 Ejm 33.0QO
BN

1

 

 

 

.gqm ¥s=4 go ze=pqgn_4qu
m<mg2g-a«

 

 

.hmnonxufi< ..omm xcfid mo coflumunfifimo m.a upswflm

 

wDJI> mam

 

143

 

av mm on mm cm 0. o—

d 4 . 4 ,1 m.—
.\.-./ ,
. I,

/ L8—

SBDNBHBfldIO 31010888 £0 NOS

coCmaCo Ejm 33.3

 

# 9N

 

 

.9mm gs=4 go ze=paga_gqu
meagzg-aa

 

 

M4

.umnonx-fl< ..oom xcwfi mo :owumupwflmo

m.o «hawwm

 

mfifi#&> mam

00 on Am, _uV 0v mm on mm on

m—

o—

 

 

\

-_
In
l\\\\llI-I IJI

\ /.
/

co_Lou_Lo EJo mu:_omnc

L

 

 

 

.gom gs=4 no seepaga_4qu
meaézy-a«

Qu—

. ww—
. an.
1 mm—
..omm
l mvm
.QKN

mam

. can

mVn

losm
imam
.lomv

SBUNSHijIU 31n1oseu so wns

 

 

I45

.umnonxud< .mc_¢ xcflfi «o soflumunfidmu

o_.a muawflm

 

o—

 

 

mqflfiy) mam
av 0v mm on aw om @—
\Iui./ L
\\\\\ Jr!
\. ./ 1
. L
I 1

co.L0u.Lo 53m ou:_omn¢

 

 

 

8.5 “.5... ”a 8253...:
magsmée

Qu—

mn—

Dn—

mm—

Dam

mvm

QKN

SSONBHdeIU ElHWOSBU so HHS

 

 

.umnocx1fi< .moma xcfifi mo cofiumunflfimu __.a muswflm

 

 

146

 

34...; ”an.
cc 8 8 ow cm 9 o.
1 1 1 1 4 1 9:
1|).
. S
.qur. 1.2.m
. .f/////, nu
1:
.. 1 8. w
8
0
...I
0
al.—
1 8m 3
a
...m
...u.
1 BM 8
2..
N
0
..e3
co_gou_go 53m ou3_omo¢ 1nv~

 

 

 

menu gs=4 no seepaga_4qu
$32,813

 

 

147

.pmnonx1~< .momm xcflfl mo cofiumunflfimo

N. .a 83»:

 

 

 

wins ”an.
a. 9 8 8 mm 8 m. o.
N); . 18.
..J.

Ih/IL\R J/r ..om_
1 Bu
.LDnN
.1mmm
1 8m

co_gou_go £30 0u3_omn¢ I
58

 

 

 

88 as: "a 3.25::
$351.2

SBONQHSAHIU BlHTOSBH d0 HHS

 

 

APPENDIX E

TIME-SPACE DIAGRAMS

l48

.Hmwumupm uwnosxufi< wcofim owmmmuu wagonsusom
.xuo3um: ummcwfi mzu how Emuwmflu womamnmefiu dmewuao

23

_.m mugwwm

 

NUSRh§Q

YRV

thQ‘

‘Wx‘

KR \§\\

 

85 5:4.
3 [MEI/V fa? 5.5. fflfi/f/L‘
@522 f0»? .5 5 7369/‘774'

-

(7&1! [IA/67%

 

 

MRVQ>x

 

\N

MN

#0

85
)7”! (Jet)

 

 

 

149

.Hmfiumuum EmEEmo wcon ufiwwmuu vzsonzusom Any
.A.o.uc00v ..m wuswfim

 

MQRRKH

er

:19:

t Vh.‘

§2Q8\

 

85 sec.
: 5,915” to»? .5. a. 73mm:
£227 fax? 5 5 main:

[/(ZZ KIA/67H

 

HQQz

 

 

 

\$

Nx

I70

83
rm: (5m)

 

 

 

150

.Amwumuum unnosxufi< wcofim owmmmuu ucaonnuaow Amy

.xuoaumc Hm:0wmcmafluuozu mcu new smuwmmu mumqmlmsmu amefiuao

~.m mpswfim

 

Ld$$Rh§V

«SQ

sK§9V

:3 ts!

 

dfl?.5£c
i éflé'i/V f0? 5.5. fflflFF/Z
Pia {OR .5 8 726416475

¢CY222714£M4577/

 

 

A‘b

BO
7%ME'CSQCJ

 

.NQQ>\

 

 

 

\\ Nx

 

 

151

. Emesm wco m owmmmuu canonnusom ADV
Hmwumuuw a A .A.c.uc0uV ~.m muswwm

 

.wwkfi 9Q

“(Q

#3:

\‘Nfih

thu\k\

 

3 6/PJZ’A/ K0,? 5.5. fflflff/C
'A’ifl f0? 55 7369/776

(VIZ! KIA/6777’ :80 566.

50
fl»?! (5:4)

 

NQQ?

 

 

 

/60

 

 

BIBLIOGRAPHY

BIBLIOGRAPHY

Rach, L. "Traffic Signals." Transportation and Traffic
Engineering Handbook, Institute of Transportation
Engineers, 1982.

Pignataro, L.J. "Traffic Engineering: Theory and
Practice." Prentice-Hall, Inc., New Jersey, 1973.

U.S. Department of Transportation, and Federal Highway
Administration. "Handbook of Computer Models for Traffic
Operations Analysis." Report FHWA-TS-82-213, December
1982.

Hielm, 0., and Hammarstroen, U. "Measurements of Fuel
Consumption. A Validation Study of TRANSYT. Potential
for Fuel Saving in Coordinated System of Traffic Sig-
nals." National Swedish Road and Traffic Research
Institute, Report 271, Sweden, 1981.

Robertson, D.I., Lucas, C.F., and Baker, R.T. "Coordi-
nating Traffic Signals to Reduce Fuel Consumption. " TRRL
Report LR-934, Transport and Research Laboratory, 1980.

Euler, G.W., and Schoene, G.W. "FHWA-Sponsored Project
Proves Cost Effectiveness of Signal Timing Optimization. "
Transportation Research News 103 , Transportation Research
Board, November 1982.

Wallace, C.E., Courage, K.G., Reaves, D.P5 Schoene, G.W.,
and Euler, G.W. "TRANSYT-7F User's Manual". Transporta-
tion Research Center, University of Florida, Gainesville,
Florida, June 1984.

Federal Highway Administration, and. University of
Florida. "National Signal Timing Optimization Project,
Summary Evaluation Report." May 1982.

Wagner-McGee, and Associates. "Fuel Efficient Traffic
Operation and Evaluation: Garden Grove Demonstration
Project." Report P-400-83-004, California Energy
Commission, February 1983.

152

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

153

Schlappi, M.L. "Signal Coordination As Its Affects
Carbon Monoxide Concentrations." Session 13, Institute
of Transportation Engineers, Compendium of Technical
Papers, 1984.

Robertson, D.I. "Research on the TRANSYT and SCOOT
Methods of Signal Coordination." Institute of Transpor-
tation Engineers Journal, January 1986.

Homburger, W.S. , and Kell, J.H. "Fundamentals of Traffic
Engineering." Institute of Transportation Studies,
University of California, Berkeley, 1986.

Robertson, D.I. "TRANSYT: A Traffic Network Study
Tool." Transport and Road Research Laboratory,
Crowthorne, Berkshire, England, Report LR 253, 1969.

Webster, F.V., "Traffic Signal Setting." Road Research
Technical Report 39, London, 1956.

Rouphail, N.M. "Analysis of TRANSYT Platoon-Dispersion
Algorithm." Transportation Research Record 905, Trans-
portation Research Board 1983.

Cohen, S.L., and Liu, C.C. "The Bandwidth-Constrained
TRANSYT Signal-Optimization Program." Transportation
Research Record 1057, Transportation Research Board,
1986.

Wallace, C.E. "At Last - A TRANSYT Model Designed for
American Traffic Engineers. " Institute of Transportation
Engineers Journal, August 1983.

U.S. Department of Transportation. "Microcomputer in
Transportation: Software and Source Book," Report UMTA-
URT-41-87-1, June 1987.

The North Carolina Department of Transportation, and the
Institute for Transportation Research and Education.
"North Carolina's Traffic Signal Management Program for
Energy Conservation." Institute of Transportation
Engineers Journal, December 1987.

McCoy, P.T., Balderson, E.A., Hsueh, R.T., and Mohaddes,
A.K. “Calibration of TRANSYT Platoon Dispersion Model
for Passenger Car Under Low—Friction Traffic Flow
Conditions." Transportation Research Record 905,
Transportation Research Board, 1983.

Yager, 8., and Case, E.R. "Using TRANSYT for Evalu-
ation." Special Report 194, Transportation Research
Board, 1981.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

154

Dudeck, G.R., Goode, L.R., and Poole, M.R. "TRANSYT-7F
and.NETSIM Comparison of Estimated and Simulated Perfor-
mance Data." Institute of Transportation Engineers
Journal, August 1983.

Skabardonis, A., and.May, A.D. "Comparative Analysis of
Computer Models for Signal Timing." Transportation
Research Record 1021, Transportation Research Board,
1985.

Tarnoff, P.J., and Parsonson, P.S. ”Selecting Traffic
Signal Control at Individual Intersections." NCHRP
Report 233, Transportation Research Board, June 1981.

Coeymans, J.E., and Willumsen, L.G. "Adapting TRANSYT
to Conditions in Developing Countries." Puanning and
Transport Research and Computation, 1986.

Jovanis, P.P., May, A.D.,and Deikman, A. "Further
Analysis and Evaluation of Selected Impacts of Traffic
Management on Surface Streets." Institute of Transpor-
tation Studies, University of California, Berkeley,
October 1977.

Schoene, G.W., and Euler, G.W. "Let TRANSYT speed
Traffic Flow." American City and County, Volume 97,
Number 12, December 1982.

Davis, S.W. "Signal Timing based on TRANSYT-7F."
Engineering Bulletin of Purdue University, Number 153,
Purdue University, 1982.

Cohen, S.L., and Mekemson, J.R. "Optimization of Left-
Turn Phase Sequence on signalized Arterials." Transpor-
tation Research Record 1021, Transportation Research
Board, 1985.

Chang, E.CP., and Messer, C.J. "Warrants for Intercon-
nection of Isolated Traffic Signals." Texas Transporta—
tion Institute, Texas A&M University, August 1986.

Jrew, B.K., Parsonson, P.S., and Chang, E.CP. "Investi-
gation of Optimal Time to Change Arterial Traffic Signal-
Timing Plan." Transportation Research Record 1057,
Transportation Research Board, 1986.

Lieberma, E.B., Lai, J., and Ellington, R.E. "SIGOP-III
User's Manual." Transportation Research Center, Univer-
sity of Florida, Gainesville, Florida, July 1983.

Kaplan, J.A. , and Powers, L.D. "Results of SIGOP-TRANSYT
Comparison Studies." Traffic Engineering Journal,
September 1973.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

155

Datta, T.K, Litvin, 0.14., and Taylor, W.C. "Progressive
Signal Signal System in Network of Arterial Streets Using
TRASOM." Institute of Transportation Engineers Journal,
June 1979.

Datta, T.K, Litvin, D.M., and Taylor, W.C. "Progressive
Signal System in Network of Arterial Streets." Transpor-
tation Research Record 597, Transportation Research
Board, 1976.

Lam, J.R., Ugge, A.J., and Allen, B.L. "Development and
Applications of the SSTOP Traffic Signal Optimization
Program." Proceeding of the 7th Annual Conference,
Institute of Transportation Engineering, Toronto, Canada,
1982.

McGill, J. "SSTOP - A Signal System Optimization
Program. " Institute of Transportation Engineers Journal,
March 1984.

Collins, J.F., and Gower, P. "Dispersion of Traffic
Platoon on A4 in Hounslow, U.K." TRRL Report SR 29UC
Transport and Road Research Laboratory, Crowthorne,
Berkshire, England, 1974.

El-Reedy, T.Y., and Ashworth, R. "Platoon Dispersion
Along a Major Road in Sheffield." Traffic Engineering
and Control, April 1978.

Sneddon, P.A. "Another Look at Platoon Dispersion-3:
The Recurrence Relationship." Traffic Engineering and
Control, February 1972.

Lam, J .K. "Studies of a Platoon Dispersion Model and Its
Practical Applications.” Proceeding, Seventh Interna-
tional Symposium on Transportation and Traffic Theory,
Institute of System Science Research, Kyoto, Japan, 1977.

Vincent, R.A., Mitchell, A.I., and Robertson, D.I. "User
Guide to TRANSYT version 8." TRRL Report 888, Transport
and Road Research Laboratory, Crowthorne, Berkshire,
England, 1980.

Castle, D.E., and Bonniville, J.W. "Platoon Dispersion
Over Long Road Links." Transportation Research Record
1021, Transportation Research Board, 1985.

Smelt, J .M. "Platoon Dispersion Data Collection and
Analysis." Proceeding, Annual Meeting of the Australian
Road Research Board, Volume 12, part 5, 1984.

45.

46.

47.

48.

49.

50.

51.

52.

156

Leden, L. "A validation Study of TRANSYT and ROSIM."
University of Lund, Lund Institute of Technology,
Department of Traffic Flaming and Engineering, Lund,
Sweden, 1976.

Axhausen, R.W., and Kdrling, H.G. ”Some Measurements of
Robertson's Platoon Dispersion Factor." Paper submitted
to the 66th Annual Meeting of Transportation Research
Board, Washington, January 1987.

Tracz, M. "The Prediction of Platoon Dispersion based
on Rectangular Distribution of Journey Time." Traffic
Engineering and Control, 1975.

Lorick, H.C. "Analysis and Development of Fuel and
Platoon Models.” M.S. thesis, Department of Civil
Engineering, University of Florida, Gainesville, Florida,
1981.

Denney, R.W. "Traffic Platoon Dispersion Modeling:
Evaluation the Mechanism and Empirical Support." Paper
presented at the 65th Meeting of the Transportation
Research Board, 1986.

Letter from S.W. Erwin dated February 29, 1988.

Kell, J.H., and Fullerton, I.J. "Manual of Traffic
Signal Design." Institute of Transportation Engineers,
1982.

Box, P.C., and Oppenlander, J .C. "Manual of Traffic
Engineering Studies." Institute of Transportation
Engineers, 1976.