“cairn

admin

 

PE? 33 3k.

 

3.3%
«a. «1%
4. aw

.1
,
z 5.41.... i
i .91..“ 0 1
.q‘fmmuwm‘mfi
giants

Jim...» I . .w

e. . Lil), V a

{F.rfiq Lrlf'r. A:

S.- .llrfiqm .
$5..

i .

ma .9... z
uUMmunafi. .3.h
g. .. KW. L Mum»

.\1 {3.31. 0
drunk” 1143,

1 1.1.34. .tén

é}?

 

 

THESYS

(;"\:’Afl O\\

 

 

”BRA.“ will\iiiiflifiifiii‘jfllfijwui
Micmgan State
University

 

 

 

This is to certify that the

dissertation entitled

ANALYSIS OF ADAPTIVE SIGNAL CONTROL STRATEGIES IN
URBAN CORRIDORS

presented by

Ahmed Shawky Abdel-Rahim

has been accepted towards fulfillment
of the requirements for

 

Ph .D 0 degree in CiVil Engineering

 

Ali/(a 771 C: , «inf/’1

Major professor

Date 12/10/98

 

MSU is an Affirmatiw Action/Equal Opportunity Institution 0. 12771

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINE return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE
ID, 1 92003
(Emazom g Y 0 r

5.211 :3

 

d

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/” WWW“

 

ANALYSIS OF ADAPTIVE SIGNAL CONTROL STRATEGIES IN
URBAN CORRIDORS

By

Ahmed Shawky Abdel-Rahim

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Department of Civil and Environmental Engineering

1998

ABSTRACT

ANALYSIS OF ADAPTIVE SIGNAL CONTROL STRATEGIES IN
URBAN CORRIDORS

By

Ahmed Shawky Abdel-Rahim

The objective of this study was to determine the merits of alternative signal
control strategies and to quantify the magnitude of the incremental benefits that might be
achieved through changing the control of traffic signals in urban conidors from fixed-
time or actuated control to more advanced computer-based adaptive control systems. Two
adaptive signal control strategies, with two different techniques for traffic profile
predictions, were examined in the study. The analysis was conducted using simulation
modeling of a hypothetical corridor with traffic profiles representing moderate and high
peak-hour conditions. A real-time adaptive control algorithm was incorporated with the
CORSIM simulation model. The algorithm provided real-time adaptive control for the
traffic signals along the corridor based on detector data exchanged with the Simulation

model.

The effectiveness of different signal control strategies was examined using the
same platform. The output of the simulation model showed that adaptive control
strategies resulted in a reduction in average intersection delay and corridor travel time

over optimized fixed-time signals. The reduction was higher during the non-peak periods

and decreased as traffic demand reached its peak with average saving in total travel time
ranging from 1.89% to 5.90%. When compared with the coordinated actuated signals,
neither of the two adaptive control strategies showed a significant difference in corridor
travel time or intersection delay parameters. The saving in corridor travel time under
adaptive signal control system is more sensitive to the PHF of the minor-street traffic
than any other traffic parameter. There was no significant difference in the performance
of the two adaptive control systems that used different traffic-profile prediction

techniques.

A field study to examine the effectiveness of deploying the SCATS adaptive
control system in the Orchard Lake Road corridor showed that the reduction in corridor
travel time achieved under SCATS control was within the benefit limits predicted by the
simulation model. The results also showed that, similar to what the simulation model
output demonstrated, coordinated actuated signals might have achieved a similar

reduction in travel time.

This research has contributed to the understanding of the potential benefits of
adaptive signal control strategies in urban corridors. It also identifies several areas where
there is a need for further research that can, ultimately, lead to a set of rules and

guidelines for choosing among different signal control strategies within an ITS context.

This work is dedicated to my parents, my wife, and my two children

Who all dream to witness my accomplishment.

iv

ACKNOWLEDGMENTS

I am indebted to my advisor, Dr. William C. Taylor, for his invaluable advice and
guidance. Dr. Taylor is a true example of a great educator. His continues support, from

my first days in the program, has helped me achieve this accomplishment.

I would like to express my gratitude to Dr. Richard Lyles, Dr. Thomas Maleck,
and Dr. Joseph Gardiner for their professional suggestions and for helping me develop a

deep insight into the topics.

I would like to express my deep thanks to my wife for her continuos support and
encouragement throughout the years. My deep appreciation to my parents who has always
given me perpetual and unconditional support. Their continuous encouragement helped

me achieve my academic goal.

I would like to extend my gratitude to all my colleagues at Michigan State
University for their unreserved friendships and support. My work experience with them
is very useful for my future. I would like also to thank the Department of Civil and
Environmental Engineering and Michigan State University for providing me with the

opportunity to conduct this research work.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................. viii

LIST OF FIGURES ........................................................................... x

CHAPTER 1

INTRODUCTION ............................................................................. 1
1.1 Description of the Problem ....................................................... 1
1.2 Evolution of adaptive signal control strategies ................................. 2
1.3 Current practices of signal control strategies ................................... 6
1.4 Statement of the problem ......................................................... 9
1.5 Objectives and scope of the research ........................................... 10
1.6 Research approach ................................................................ 11
1.7 Structure of the dissertation ...................................................... 13

CHAPTER 2

LITERATURE REVIEW ..................................................................... 14
2.1 Optimization of traffic signal control systems ................................. 14
2.2 Comparative analysis of traffic control systems .............................. 18
2.3 Current practice of adaptive control systems .................................. 23
2.4 Literature Review Conclusions .................................................. 29

CHAPTER 3

DEVELOPMENT OF THE SIMULATION MODEL .................................... 31
3.1 Simulation Modeling .............................................................. 31

3.1.1 Selection of the Simulation Model Suitable for ITS application. 33
3.1.2 Description of CORSIM (Structure, Principles, and Features) ......35

3.1.3 Modeling an External Controller within CORSIM .................... 37

3.2 Signal Control strategies ......................................................... 40

3.2.1 Coordinated Fixed-Time Signals ......................................... 40

3.2.2 Coordinated Actuated Signal Systems ................................... 44

3.2.3 Real-time adaptive signal control ........................................ 50
CHAPTER 4

DESIGN OF THE STUDY ................................................................... 56

4.1 Study Plan .......................................................................... 56

4.2 Hypothetical Corridor Network ................................................. 58

4.3 Traffic Demand Levels ........................................................... 62

4.4 Measures of Effectiveness ....................................................... 62

4.4.1 Average corridor travel time .............................................. 64

4.4.2 Total corridor travel time .................................................. 64

4.4.3 Average delay per approach ............................................... 64

4.4.4 Average intersection delay ................................................ 65

4.5 Analyzing CORSIM Output/ simulation iteration ............................ 66

vi

4.6 Number of simulation iterations ................................................. 68

CHAPTER 5
ANALYSIS AND RESULTS ................................................................. 71
5.1 Comparative analysis of signal control strategies ............................. 72
5.1.1 Corridor travel time ......................................................... 72
5.1.2 Total travel time ............................................................ 76
5.1.3 Intersection Delay .......................................................... 81
5.2 Signal timing analysis ............................................................ 95
5.3 Sensitivity analysis ............................................................... 106
5.3.1 Sensitivity to minor-street demand level ................................ 107
5.3.2 Sensitivity to maj or-street demand level ................................. 109
5.3.3 Sensitivity to signal phase plan ........................................... 113
5.3.4 Sensitivity to changes in traffic demand over time ..................... 115
5.4 Comparative analysis of adaptive control prediction logic .................. 118
CHAPTER 6
FIELD STUDY ................................................................................. 123
6.1 Field study Objectives and Approach .......................................... 124
6.2 Study Location ..................................................................... 124
6.3 Characteristics of SCATS adaptive control .................................... 126
6.4 Before/after green time and offset Analysis ................................... 127
6.5 The simulation study ............................................................. 132
6.6 Results from the simulation study .............................................. 134
6.7 Comparative cost analysis ....................................................... 139
6.7.1 Components of cost for alternative control systems ................... 140
6.7.2 Components of Benefits for alternative control systems .............. 141
CHAPTER 7
CONCLUSIONS .............................................................................. 145
APPENDICES
Appendix A Data structures in CORSIM .................................................... 151
Appendix B Source Code for the adaptive control logic interface with CORSIM 153
Appendix C Source file for CORSIM ........................................................ 158
BIBLIOGRAPHY .............................................................................. 161

vii

LIST OF TABLES

Table 1.1 Key features for the UTCS three generations of control strategies ............... 3
Table 1.2 Characteristics of the UTCS three control strategies ..................................... 5
Table 2.1 Comparison of travel time by route under different control strategies ........ 21

Table 1.2 Comparison of total travel time under alternative control strategies

with first generation control ........................................................................ 22
Table 5.1 Average corridor travel time (moderate peak- hour case) ........................... 75
Table 5.2 Savings in corridor travel time (moderate peak- hour case) ........................ 75
Table 5.3 Average corridor travel time (high peak- hour case) ................................... 78
Table 5.4 Savings in corridor travel time (high peak- hour case) ................................ 78
Table 5.5 Total corridor travel time (moderate peak- hour case) ................................ 82
Table 5.6 Total corridor travel time (high peak- hour case) ........................................ 82

Table 5.7 Average approach delay for maj or-street traffic (moderate peak- hour
case) ............................................................................................................. 86

Table 5.8 Average approach delay for major-street traffic (high peak- hour case). ....86

Table 5.9 Average approach delay for minor-street traffic (moderate peak- hour

case) ............................................................................................................. 90
Table 5.10 Average approach delay for minor-street traffic (high peak- hour

case). ............................................................................................................ 90
Table 5.11 Average intersection delay (moderate peak- hour case). ........................... 94
Table 5.12 Average intersection delay (high peak- hour case). ...................... - ............ 94

Table 5.13 Percent reduction of corridor travel time under adaptive control
system (two-year operation) ...................................................................... 1 15

Table 6.1 Effective green time and average degree of saturation for Walnut Lake
Road intersection ....................................................................................... 129

Table 6.2 Average corridor travel time (southbound traffic-moming peak) . ........... 138

viii

Table 6.3 Average corridor travel time (northbound traffic-aftemoon peak) . .......... 138

Table 6.4 Percent reduction in corridor travel time (southbound traffic-morning

peak) . ........................................................................................................ 139
Table 6.5 Percent reduction in corridor travel time (northbound trafiic—aftemoon

peak) . ........................................................................................................ 139
Table 6.6 Estimated cost of different generations of UTCS control strategies . ....... 140
Table 6.7 Cost components for alternative control strategies . .................................. 141
Table 6.8 Benefit components for alternative control strategies . ............................. 142
Table 7.1 Percent reduction in total travel time under adaptive control . ................. 148

ix

LIST OF FIGURES

Figure 3.1 TSIS-CORSIM RTE interface structure ..................................................... 39
Figure 3.2 Operation of an actuated phase ................................................................... 47
Figure 3.3 Typical timing dial diagram for coordinated actuated signals in

CORSIM. .. ................................................................................................ 49
Figure 3.4 Network flow diagram for real-time adaptive control ................................ 51
Figure 3.5 Predication sequence for adaptive control strategies .................................. 51
Figure 3.6 Detector location for the two adaptive control strategies ........................... 53
Figure 3.7 Use of downstream detectors to predict vehicle arrivals ............................ 54
Figure 3.8 Optimization sequence for adaptive control strategies ............................... 54
Figure 4.1 Study design ............................................................................................... 59
Figure 4.2 Geometric configuration of the corridor ..................................................... 60
Figure 4.3 Link-node diagram for the corridor ............................................................ 61
Figure 4.4 Demand levels ............................................................................................ 63
Figure 4.5 Average corridor travel time for 40 simulation iterations .......................... 70
Figure 4.6 Number of simulation iterations ................................................................. 70
Figure 5.1 Average corridor travel time (moderate peak- hour case) .......................... 74
Figure 5.2 Average corridor travel time (high peak- hour case) .................................. 77
Figure 5.3 Total corridor travel time (moderate peak- hour case) ............................... 79
Figure 5.4 Total corridor travel time (high peak- hour case) ....................................... 80

Figure 5.5 Average approach delay for major-street traffic (moderate peak- hour
case). ............................................................................................................ 84

Figure 5.6 Average approach delay for major-street traffic (high peak- hour case)....85

Figure 5.7 Average approach delay for minor-street traffic (moderate peak- hour
case). ............................................................................................................ 88

Figure 5.8 Average approach delay for minor-street traffic (high peak- hour case). ..89

Figure 5.9 Average intersection delay (moderate peak— hour case) ............................. 92
Figure 5.10 Average intersection delay (high peak- hour case). ................................ 93
Figure 5.11 Possible delay reduction by adaptive control system. .............................. 97
Figure 5.12 Effective green time for major-street traffic (moderate peak- hour

case). ............................................................................................................ 98
Figure 5.13 Effective green time for minor-street traffic (moderate peak- hour

case). ............................................................................................................ 99
Figure 5.14 Effective green time or major—street traffic (high peak- hour case). ...... 100
Figure 5.15 Effective green time or minor-street traffic (high peak- hour case). ...... 101
Figure 5.16 Degree of saturation for major-street traffic (moderate peak- hour

case). .......................................................................................................... 102
Figure 5.17 Degree of saturation for minor-street traffic (moderate peak- hour

case). .......................................................................................................... 103
Figure 5.18 Degree of saturation or major-street traffic (high peak- hour case). ...... 104
Figure 5.19 Degree of saturation or minor-street traffic (high peak- hour case). ...... 105
Figure 5.20 Benefit sensitivity to minor street PHF. ................................................. 108
Figure 5.21 Benefit sensitivity to minor-street demand level. ................................... 110

Figure 5.22 Potential corridor travel time savings under different demand levels
and PHFs. .................................................................................................. 1 1 1

Figure 5.23 Benefit sensitivity to major-street demand level. ................................... 112

Figure 5.24 Average corridor travel time under fixed-time and adaptive control

(4-phase signal operation) ......................................................................... 114
Figure 5.25 Benefit of adaptive signal control afier one year period. ....................... 116
Figure 5.26 Benefit of adaptive signal control after two-year period. ....................... 117

xi

Figure 5.26 Benefit of adaptive signal control after two-year period. ....................... 117

Figure 5.27 Actual vs. predicted number of vehicles for each cycle (downstream
detector technique). ................................................................................... 120

Figure 5.28 Actual vs. predicted number of vehicles for 5-minte time interval
(upstream detector technique). .................................................................. 121

Figure 5.29 Actual vs. predicted number of vehicles for each cycle (upstream

detector technique). ................................................................................... 122
Figure 6.1 Orchard Lake Road corridor. .................................................................... 125
Figure 6.2 SCATS cycle-by-cycle degree of saturation for through traffic . ............ 130

Figure 6.3 Average corridor travel time under different control strategies
(southbound traffic-morning peak) ........................................................... 136

Figure 6.4 Average corridor travel time under different control strategies
(northbound traffic-aftemoon peak) . ........................................................ 137

xii

Chapter 1

INTRODUCTION

1.1 Description of the Problem

In 1991 the Federal Highway Administration (FHWA) issued a solicitation for
the development and evaluation of a Real-Time TRaffic Adaptive Control System
(RT-TRACS) suitable for use in an Intelligent Transportation Systems (ITS)
environment. The FHWA ITS program consisted of research and operational tests
designed to combat traffic congestion. The thrust of the program was to develop and
implement the technology necessary to mitigate the effects of congestion by providing
more efficient transportation facilities. Some of the elements included in ITS are
transportation management, in-vehicle route guidance systems, integration of multi-
modal transportation, integration of surface streets and freeway networks, and incident
management. An adaptive traffic control system, together with a surveillance system
that provides real time data, are fundamental to integrate different ITS elements. This
adaptive control needs to include forecasting capabilities such that proactive, not

reactive, traffic control is provided.

With the anticipated rapid development of ITS, considerable potential exists to

change the control of many of the nation’s traffic signals, primarily controlled at

present in fixed-time or actuated modes, to more efficient on-line advanced adaptive
control (Davies, 1991). Thus, there is a need to determine the merits of alternative
signal control strategies and to precisely quantify the magnitude of the incremental
benefits that might be achieved through changing to adaptive control strategies. While
the potential benefits of these adaptive control strategies are thought to be significant,
few research studies have compared the effect of adaptive control against alternative
control strategies. The objective of this research is to examine the benefits of
deploying an adaptive signal control system on an arterial street in an urban traffic

network.

1.2 Evolution of adaptive signal control strategies

Afier the introduction of computer-based traffic signal control systems in the
1960s, numerous experiments were conducted to develop more responsive control
strategies. One of the most comprehensive studies was the Urban Traffic Control
System (UTCS) experiment in the 19705 by the FHWA. The UTCS project was
directed toward developing and testing a variety of advanced network control concepts
and strategies and lasted for almost a decade, (Henry, Ferlis, and Kay, 1976). Research
and testing in the UTCS project were divided into three generations. Table 1.1

presents the key features for the three generations of control.

Table 1.1 Key features for the UTCS three generations of control strategies

 

 

Feature First Generation Second Generation Third Generation
Optimization Off-line On-line On-line
Frequency of update 15 minutes 5 minutes 3-6 minutes

No. of timing

patterns up to 40 Unlimited Unlimited
Traffic predictions No Yes Yes
Critical Intersection Adjust Split Adjust splits and Adjust Split, offset,
Control (CIC) offsets and cycle
Hierarchies of Pattern Selection Computation Computation
control
Fixed cycle length Yes Within group of No fixed cycle
intersections length

 

First-Generation Control (I-GC) uses pre-stored signal timing plans that are

 

calculated off-line and based on historical traffic data. The plan controlling the traffic
system can be selected on the basis of time of day, by direct operator selection, or by
matching from the existing library a plan best suited to traffic conditions (volumes and
occupancies). This is called the traffic responsive (TRSP) mode of plan selection.
Plans in l-GC can be determined by any off-line signal optimization method. The
operator determines the mode of plan selection with a frequency of update in the
traffic-responsive mode of 15 minutes. The I-GC also included logic to enable a

smooth transition between different signal timing plans and a critical intersection

control (CIC) feature, that enables vehicle—actuated adjustment of green splits at

selected signals.

The Second-Generation Control (2-GC) is an on—line strategy that computes in

 

real-time and implements signal-timing plans based on surveillance data and predicted
values. The optimization process can be repeated at 5-minute intervals; however, to
avoid transition disturbances, new timing plans cannot be implemented more often
than once every 10 minutes. The 2-GC software contains an optimization algorithm, a
traffic prediction model, sub-network configuration models, critical intersection

control, and a transition model to minimize transition time between two plans.

Finally, the Third-Generation Control (3-GC) strategy was designed to

 

implement and evaluate a fully responsive, on-line traffic control system. Similar to 2-
GC, it computed control plans to minimize a network-wide objective using predicted
traffic conditions for input. The differences with 3-GC were that the time period for
revising timing plans was shortened to 3 to 5 minutes and the cycle length was
allowed to vary among the signals as well as the same signal during the control period.
The switching points within each control interval were also determined by the on-line
optimization procedure. It is noteworthy that 3-GC is similar in concept to 2-GC.
Because of the inherent inaccuracies in the measurement-prediction cycle, neither the

2-GC nor the 3-GC strategies could respond adequately to rapid changes in traffic

flows. Table 1.2 presents the characteristics of different generations of signal control

strategies.

Table 1.2 Characteristics of the UTCS control strategies

 

Generation of Control

 

Benefits Fixed-Time First Second Third

 

+ Signal coordination

+ Change timing easily

+ Monitor performance

and equipment status

+ Data collection for future
policies

+ Responsive to major I
variation

+Responsive to all
variation V

+ Detector needs None Low Moderate High

 

 

 

 

 

 

[Source: The Urban Traffic Control System in Washington, DC, US. Department of
Transportation, Federal Highway Administration, 1974]

A significant advance toward more effective on-line adaptive control was
achieved during the 1980s with the introduction of Split, Cycle and Offset
Optimization Technique (SCOOT) in the United Kingdom and Sydney Coordinated
Adaptive Traffic System (SCATS) in Australia. SCATS is an example of first
generation UTCS control strategy (Lowrie, 1982). It updates signal-timing parameters
based on changes on traffic demand from a library of pre-defined plans. SCOOT,
however, uses an on-line optimization technique and may be considered a 2-GC

strategy (Bretherton, 1990).

1.3 Current practices of signal control strategies

There are several levels of sophistication for signal control systems. These
range from fixed time plans selected by time-of-day through dynamic selection of
fixed time plans to full dynamic selection of cycle length, phase splits, and offsets.
The various signal control systems for urban arterial street networks currently in
operation fall into three broad categories; fixed-time control with plans selected by
time-of-day, semi-actuated or full-actuated control, and on-line dynamic adaptive

control.

Fixed-time control is by far the simplest type of control, as it involves the pre-
calculation of signal timing plans for the entire control period. Typically, fixed-time
plans are optimized off-line based on average historical arrival rates, and then
implemented in the signal controller. They, therefore, have no ability to react to
changes in traffic patterns. However, as traffic volumes, except at capacity, are seldom
constant for an entire control period, this type of control is optimal for traffic
conditions that, while average, may still only occur for brief periods during each
control period, if at all. At all other times, the fixed-time signal plan may not be
optimal. This limitation can be overcome by implementing time-of-day control where
the timing plan is changed automatically at certain times of the day based on expected
temporal traffic peaking trends. This control strategy, other than during transition

periods, maintains the simplicity of fixed-time control.

The main drawback to this type of control is its inability to react to unexpected
deviations from historical trends, such as diversions arising from incidents, or simple
day-to-day random variations of the timing and the severity of the peaks. In addition,
even for predicted traffic conditions, there are only a finite number of time-of-day
plans that can be handled by current controllers. Therefore, during periods of build up
or decay of the peak, the selected time-of-day plan may still not be optimal. These two
limitations spurred the development of adaptive control algorithms. These algorithms,

at least in theory, should provide optimal solutions for all conditions.

Actuated signals allocate the green time for each phase on a cycle-by-cycle
basis. The basic characteristic of actuated control is that the cycle length and phase
splits, and even phase sequence, may vary from cycle to cycle in response to detector
actuation. Detectors, placed upstream from the intersection, provide the actuated
controller with information concerning current demand. Signal actuation, accordingly,
permits phases to start early or to end late to meet the variation in demand. This

flexibility should make it possible to improve the performance of the control strategy.

In urban corridors, where intersections are relatively closely spaced, signals,
operating in fixed-time or actuated modes, can be coordinated to provide progression
for vehicles traveling through the corridor. In this case, all signals along the corridor

should operate in a common cycle-length to maintain offsets and progression patterns.

While coordinating actuated signals would deprive the controller of most of its ability
to vary cycle length to meet demand, they can improve the signal Operation by
reallocating green time among approaches to meet the cycle-by-cycle variation in
demand.

It should be noted that none of the above control strategies has the ability to
dynamically change the phase structure (number, type, and order of phases). Currently,
this decision must still be optimized off-line in most controllers, and is then held fixed
for a specified duration. In actuated signals, however, some phases might be skipped if
there is no demand to activate the actuated controller. Current practice is to use non—
actuated signal control plans if the arrival pattern is predictable, such as when
networks are heavily congested and intersections are closely spaced. Actuated control
is used instead when arrival patterns are less predictable, such as during light traffic
conditions. Current practice also recommends that signals spaced half a mile or less

should be coordinated (Skabardonis, 1997)

Several traffic-responsive signal control strategies were developed for
intersections fumished with traffic detectors. The British SCOOT system is simply an
on-line version of the TRAfic Network StuY Tool (TRANSYT). The control strategy
is to minimize a function defined by vehicle delay and stops at all intersection in the
network in all ranges of demand. Signal splits change incrementally based on current

demand obtained from detectors every four seconds. Offsets and cycle length are

adjusted every few minutes (Robenson, 1991). SCATS of Australia has different
control strategies for various demand levels. Signal plans are selected from embedded
plans developed off-line, while the cycle length is calculated every cycle. The system

also makes use of some local controls to accommodate local traffic conditions at each

intersection (Rathi, 1992).

In the 19905, several adaptive systems were developed and implemented in the
United States. An example of these systems is the Optimized Policies for Adaptive
Control (OPAC), developed by Gartner (1983 and 1991). OPAC uses a mathematical
optimization of multistage decision process to dynamically optimize the signal setting.
The latest version of OPAC (version 3.0) incorporated major enhancements that are
important for efficient signal operation. The enhancements included phase skipping,
optimization of all phases, platoon identification, and modeling algorithm that
provides a coordination mechanism for coordinated signals. OPAC was implemented

in New Jersey in 1996 (Andrews, et. al., 1997).

1.4 Statement of the problem

The benefits of computer-based adaptive signal control systems are not yet
fully understood. The deployment of several adaptive signal control strategies have led
to a mixed perception of their benefits. Field evaluations of the deployed systems

reported different, sometimes contradictory results. Thus, there is a need for research

that quantify the incremental benefits achieved from changing the control of an
existing signal system to adaptive control. There is also a need to examine the
different factors that might impact the benefits of these systems. The research
conducted in this study is original because it tests the benefits of deploying adaptive
signal control strategies in an urban corridor. The performance of alternative signal
control strategies was compared under the same environment using simulation

modeling.

1.5 Objectives and scope of the research
The research addresses the potential benefits of adaptive signal control
strategies for urban corridors. The benefits of adaptive signal control strategies were
tested against fixed-time and actuated signals under different demand levels. The
scope of the research includes:
1. identification of measures of effectiveness (MOEs),
2. development of adaptive control strategies and tests of their effectiveness,
3. determination of the limits of effectiveness of these control strategies with respect to
demand and peak-hour factors,
4. identify which characteristics of the adaptive control logic contributed to the
changes in traffic parameters, and
5. evaluate the effectiveness of different prediction techniques used in the adaptive

control strategies.

10

1.6 Research approach

This research was based on traffic simulation. Simulation models permit the
comparison of different signal control strategies under the same road network and
traffic conditions. CORSIM (CORridor SIMulation), a microscopic interval—based
simulation model, was selected for this study. It has run-time extension capability that
allows external programs to interact with CORSIM and exchange data with it while it
is executing. This feature allows the user to replicate adaptive signal control logic
within the model. CORSIM version 1.04, which was used in this study, includes many
advanced features such as run-time extension capability, traffic surveillance system,
and new enhanced actuated control logic. CORSIM can provide data on the MOEs

suitable for the analysis (Federal Highway Administration, 1998).

The research was based on a hypothetical corridor with demand characteristics
representative of traffic conditions on the Orchard Lake corridor, an urban corridor in
Oakland County, Michigan. The effectiveness of adaptive control strategies was
compared with two conventional corridor signal control strategies namely: coordinated
fixed-time and coordinated actuated signals. Two adaptive control strategies were
developed and examined in the study. Both adaptive systems attempt to optimize the
signal setting by equalizing the degree of saturation in competing approaches. The
difference between the two lies in the way traffic profiles are predicted. In the first

adaptive logic, traffic predictions are made based on detectors placed downstream

ll

from the intersection. Predictions for the traffic profile in any given cycle were based
on the volumes of previous cycles. For the second adaptive control strategy,
predictions were made based on detectors placed upstream from the intersection by
using a “scan ahead” technique that predicts the arrival of vehicles to the intersection
based on estimated time of travel of the platoon from the detector location to the

intersection.

The study is organized into five parts. In the first part, several MOEs were used
to compare the effectiveness of the four signal control strategies under two different
demand levels representing moderate and high peak-hour conditions, respectively.
Having examined the potential benefits of the adaptive control logic, the second part
of the study was a signal timing analysis to examine which characteristics of the
adaptive control logic contributed to the changes in traffic parameters. The third part
of the study was a sensitivity analysis to examine the benefits of adaptive control
strategies under different demand levels and peak-hour factors. The fourth part of the
study aimed to compare the effectiveness of the two prediction techniques used in the
adaptive control strategies. Finally, a field study aimed to validate the output of the
simulation model was performed. A comparison of the cost and benefits of different

signal control strategies was also illustrated in the field study.

12

1.7 Structure of the dissertation

The dissertation is organized into seven chapters. A brief review of the existing
literature (Chapter 2) follows the introduction. The literature review aims at
familiarizing the reader with current research efforts in the area of signal control
strategies. Chapter 3 contains a detailed description of the simulation model
development. It also includes details of different signal control strategies examined in

the study.

Chapter 4 contains a detailed description of the design of the study, including a
discussion of different MOEs used in the analysis, processing the simulation model
output, conidor geometry, and traffic demand levels. The output of the simulation
analysis is presented in chapter 5. Chapter 6 includes the details of the field study
aimed to validate the output of the simulation model. Finally chapter 7 summarizes the

conclusions and suggestions for future research topics.

13

Chapter 2

LITERATURE REVIEW

A literature search was performed to accomplish four primary goals. First, the
current limits of knowledge in the field of traffic engineering relative to signal control
strategies were explored. Second, an understanding of the research gaps in this area,
including the absence of studies performed specifically on the comparative benefits
gained from adaptive signal control were identified. The literature review also
established a base of knowledge from which to launch the proposed study and
demonstrated certain techniques that have been successful in past research. Finally, the
review of past published literature gave insights into the way in which current theories,
technology, and implementation of adaptive traffic signal systems have developed

over the years.

2.1 Optimization of traffic signal control systems

The first generation of signal control strategies was based on off-line
calculations for a fixed-time signal control system. Webster and Cobbe (195 8) suggest
that for an isolated signal, a signal split strategy should equalize the degree of
saturation (D8) of all critical approaches to approximately yield the minimum overall

intersection delay. This technique became general practice for most fixed-time control.

14

For arterial and network considerations, where throughput is more important
than overall delay, Little (1966) introduced an off-line mathematical technique to
maximize the through bandwidth. The principle was to maximize the number of
vehicles able to successfully encounter green signals when traveling along a street.
Over the years, several variations of this approach were developed. NCHRP Report 73
(1969) evaluated several of these offset strategies. Off-line control techniques
investigated in this report were Yardeni's time-space design, Little's maximal
bandwidth, and delay/difference-of-offset. Three responsive control strategies, namely
basic queue control, cycle and offset selection, and mixed cycle mode, were also
evaluated in this report. The results indicate that the cycle and offset selection method
and delay/difi‘erence-of-offsets techniques rank the best for off-peak periods, whereas

the mixed cycle mode and basic queue control were the best for peak periods.

Perhaps the most comprehensive and most widely applied control strategy for
fixed-time control setting on an arterial is based on a computer optimization method.
The TRANSYT method of delay minimization is a popular method to determine
signal timings (Woods, 1993). TRANSYT (Robertson, 1968) is an off-line program
utilizing a modified Webster's method to calculate green splits, and a hill-climbing
optimization technique to determine the offset and cycle length which minimizes a
performance index. The logic for the offset calculation is similar to the

delay/difference-of-offset method evaluated in NCHRP Report 73. Similar programs

15

to TRANSYT are SIGOP (Signal Optimization program) and PASSER II (Progression
Analysis and Signal System Evaluation Routine), which have different optimization
procedures. The control strategies obtained from off-line calculations are effective for
average traffic conditions, but are not responsive to changing traffic patterns. When a

traffic pattern changes, the solutions from these programs are no longer optimal.

Several traffic-responsive signal control strategies were developed for an
individual intersection furnished with traffic detectors. Gazis and Potts (1964)
developed a technique for time-dependent signal setting under varying demand. They
used queue length as an input to minimize total aggregate intersection delay. The
technique is also called "bang-bang" because the green time is set at a predetermined
maximum value for the queued approach, and at a minimum value in other directions.
When a queue in the first approach is cleared, the setting is reversed. This signal
setting does not, in general, minimize the period during which one approach is
congested. D'Ans and Gazis (1976) furthered this control method by means of linear
programming. Church and Revelle ( 1978) formulated similar control strategies with
consideration of maximum waiting time and queue length. When the maximum queue
length was used as a control objective, they found that the solution tended to balance
the queue lengths on the most saturated approaches of each signal phase, and the
signal frequently switched between phases. Michalopoulos and Stephanopolous (1977)

reported that the queue constraint was effective when the demand increases to the

16

limiting value and that the optimal control strategy at saturation is simply the balance

of input-output to maintain constant queue length.

The results of testing four different intersection control strategies, namely basic
queue control, queue-length/arrival rate control, modified space-presence control, and
delay-equalization control are presented in the NCHRP Report 32 (1967). The results
showed that the modified space-presence control strategy yielded the lowest delay
under low to medium intersection demand (up to 2000 vehicles per hour for 4-lane, 4-
leg junction). When the demand was greater than 2000 vehicles per hour, the basic

queue control strategy was better than the others.

Many control strategies have been developed for oversaturated traffic
conditions at an isolated intersection. Gordon (1969) suggested that the control
objective should be to maintain a constant ratio among the respective storage spaces.
Longley (1968) attempted to balance queue lengths on all approaches. NCHRP Report
194 (1976) showed that, although the Longley control logic yielded lower delay than
the off-line calculation, the queue-actuated control resulted in lower delay when the
degree of saturation was above 0.5. The report stated that the objective of signal
control should be to avoid spillback and to provide equitable service. The report also

gave some tactical control strategies to ease queue blockage at an intersection.

l7

Dell’Olrno and Mirchandani (1995), used a simulation study to examine an
approach of real-time coordination of traffic flow in an urban network. The approach
considers available real time data for computing signal timing. It first identifies
platoons and predicts their movement in the network (i.e., their arrival times at
intersection, their sizes and their speeds) by fusing and filtering the traffic data
obtained in the last few minutes. A traffic model is used to propagate the predicted
platoons through the network for a given time horizon. The study showed that this
real-time coordination examined performed as well as or better than off-line
coordination methods such as PASSER II, or TRANSYT. They also concluded that
this approach is suitable for light-to-moderate traffic conditions, but not oversaturated

conditions.

2.2 Comparative analysis of traffic control systems

A step toward today's advanced traffic management systems was made in 1967
with the development of a computer program and methodology to gather and process
data from loop detectors placed near intersections. The Urban Traffic Control System
(UTCS) in the early 1970’s resulted in the development of three generations of traffic
control software (Henry, et al 1976). The first generation uses a set of predefined
timing plans, which were developed off-line. These plans may be called into operation,
manually, by time of day, or by a traffic responsive mechanism. Special features

include dynamic control of critical intersections (CIC). The Second generation uses

18

real time software that computes timing plans based on surveillance data and predicted
changes. The plans are, however, constrained to fixed cycle lengths, but may include
critical intersection control. The third generation is a real-time software system that
computes timing plans based on surveillance data and predicted changes
(approximately every three minutes). The third generation, however, controls
intersections independently with a fully adaptive offset and cycle length

determination.

Although the three generations of control reflected increasing level of
comprehensiveness and complexity, they were not simple evolution of the same basic
system. That is, each generation represented a unique development in control
philosophy. For this reason, the second or third generations may not be “better’ in any
or all applications than the first generation, (Henry, et al 1976 and Kay, Allen, and

Brugerman 1975).

The UTCS, which was installed in 114 intersections in Washington, DC,
operated standard traffic control signals directly from a central computer. Loop
detectors were used as the basic source of traffic flow data. Data from the loop
detectors is transmitted to the computer site using leased telephone lines. The
computer translates the detector signals into traffic flow measures which are used to

select the most appropriate signal timing plan in the case of first generation software,

19

or to serve as the basis of computing the optimal signal setting for the second and third
generations. A two-phase research program to evaluate the UTCS control strategies
was conducted at these intersections. The first phase of the research consisted of an
evaluation of four alternatives of first generation control strategies. Phase II of the
research evaluated the second and the third generation control strategies as well as the
most effective first generation control strategy. Results of the study are presented in

Tables 2.1 and 2.2 (Henry, et al 1975 and 1976).

The first generation was found to be operationally effective and the
least expensive to apply. Second generation control systems also proved to be effective
especially in arterial streets. The third generation however, did not prove effective.
Bell (1990) listed three reasons that the UTCS demand responsive systems failed when
they were introduced in the late sixties. The first was the difficulty of predicting short
term changes in traffic conditions (based on a time scale 5 to 10 minutes). The second
was the disruption to traffic caused when changing from one signal plan to the next.

The third was the limitation of the technologies available.

Much research has been performed to try to quantify the expected benefits of
adaptive control systems. Several studies on isolated intersections claim to show
substantial travel time benefits. Elahi and Goul (1991) stated that the application of

expert systems to traffic control can provide significant improvements over

20

Table 2.1 Comparison of travel time by route under different control strategies

 

 

 

 

 

 

 

 

 

Route Percent Difference
Comparison NO. A. M Noon RM
11 -8.2* -2.1 15.3*
13 -1.3 -7.4 13.1*
Traffic Responsive 22 1 3.3" 33.8* 10.2*
vs. 24 5.1* 6.7 6.2
DC. fixed-time Plans 30 5.9 22.4* 9.4
40 4.1 17.4* -O.7
1 1 -0.2 -1 .0 -1 .2
13 3.3 1 .9 4.9
Traffic Responsive 22 4.7 10.6 * 2.2
vs. 24 4.6 -4.5 2.1*
Time-of-day plan 30 -9.6 9.7* 1.1
40 -10.2 -7.7* 4.7
l 1 4.2 0.3 4.2
Traffic Responsive 13 7.1* 4.1 -1.1
vs. 22 -4.1 1 .4 -3.5
Critical Intersection 24 -9.7 4.3 -29.7
Control 30 -11.2* 7.1* -6.7
(CIC) 40 -0.5 10.5* -4.4

 

 

* Statistically significant at the 95% level.

Positive values indicate lower travel time with the traffic responsive alternative
relative to the other alternatives.

21

Snag—etc.— fi cease—nae“. 8.385
a»: 95:8.— < .9955 he 8338.292?» .555: 3.3 one: 2: 8 2530.. 358323 c.3333 a 823%.: name. guano: <
.35. $3 2: «a «585:»: b.3333... a. 3:20.51 2: 82:5: s.

 

 

 

8.3- 9m- .36- 3d- .88.
a2- ha. *3 a- .92- 8:8 sea =8§e<
1.2- .56. m.m- *odT moron 933 :82
.92- we- 4.”. he- 8:8 9.382
Mm- v6+ o. ~+ ON- cocoa 38m mew—.82
8% as»: Eveoé 09m 628 caucuses EEG o3
.U.Q .GOHMGEmaB 35:00 Remap—0:00 95:. 5:5 35:00 catwaocomv vacuum
3.280 eon—2806 vacuum totem 0:5

 

 

 

 

 

Doggobma «GDOHOAH

 

 

 

35:3 55.225» «Eu .5? 83333 .9555 95:58:. .52:— 2—5 .93.: :38 .«e :emtaaEeU N.N 83:.

22

conventional control systems. Pierce and Webb (1990) reported a reduction in delay
of between 3-24% as a result of using MOVA, a self-optimizing adaptive control
system designed to reduce delay and stops at isolated intersections, to control 20
isolated traffic signals in England. Barcelo et al (1991) stated that a demand-
responsive traffic control system would improve the network performance between

lO-25%.

Stewart and Van Aerde (1997) presented the findings from a systematic
simulation study that attempts to show what types of adaptive control strategies have
the greatest potential of producing incremental travel time benefits for different types
of traffic conditions. The study compared two types of fixed-time control with two
types of adaptive control under different traffic volumes and peak hour factors. The
analysis was performed using the INTEGRATION simulation model. The results
showed that, at low traffic volumes and at a low level of peaking there is little
difference in the performance of any control strategy. In contrast, the study showed
that that when traffic is highly peaked, well-optimized tirne-of-day plans produce
better results than fixed-time plans and adaptive control strategies. The study showed
that adaptive control strategies yield lower travel times than all other types of control

when traffic demand levels evolve slowly.

23

2.3 Current practice of adaptive control systems

Several types of advanced control systems are currently in operation, serving in
experimental traffic control management roles. Two such systems, SCATS and
SCOOT, have developed into the most prevalent real time adaptive traffic signal
control systems. The SCOOT system was developed in Great Britain in the mid-
1970's. The main idea of the SCOOT system was to take an "off-line" model, like the
fixed time signal optimization program TRANSYT, and have it operate "on-line.
Specifically, the system uses traffic data measurements collected from the existing

stream and makes short and long term decisions regarding the traffic signal settings.

Many studies have been completed which claim to have evaluated adaptive
traffic signal systems. Among the earliest reported benefits, a 1966 project in Wichita
Falls, Texas, reported a 16% reduction in stops, a 31% reduction in vehicle delay, a
8.5% reduction in accidents, and an increase in speeds of over 50%. This analysis
compared the computerized system to the single-dial system it replaced. The Fuel
Efficient Traffic Signal Management (FETSIM) and Automated Traffic Surveillance
and Control (ATSAC) programs in California (1976) showed benefit/cost ratios of
58:1 and 9.8:] respectively. ATSAC, which includes computerized signal control,
reported a 13% reduction in travel time, a 35% reduction in vehicle stops, a 14%
increase in average speed, a 20% decrease in intersection delay, a 12.5% decrease in

fuel consumption, a 10% decrease in HC, and a 10% decrease in CO levels.

24

Several of the more recent studies have involved the two most widely used
real-time traffic responsive signal systems, SCATS and SCOOT. The creators of
SCATS, the Australian Road Research Board (ARRB) and the Road and Traffic
Authority of New South Wales carried out many studies to compare the system against
various less sophisticated forms of signal control. One study measured the
performance of SCATS against the control characteristics afforded by systems with
isolated fixed time signal phasing and TRANSYT optimized fixed time control with

and without local vehicle actuation.

The ARRB study conducted their comparison using the floating car travel time
estimation technique to record the "j ourney" or travel time on each link, the number of
stops in each link, the stopped time in each link, and the amount of fuel used in each
trip. The recorded stopped times were later found to be unreliable, so they could not be
used in the analysis. The study was able to compare the different signal systems in
terms of travel time, number of stops, and a derived "Performance Index." The
Performance Index was a weighted-measure of travel time incorporating the number of
stops during the trip. The study found that on one arterial highway, SCATS resulted in
a 23% reduction in travel time and a 46% reduction in stops over isolated fixed time
signals. In the central business district (CBD) study area, the travel time was not

effected and the reduction in stops was 8%.

25

When compared to Linked Vehicle Actuated (LVA) control, SCATS showed
some benefits and some degradation in the recorded performance measures on the
arterial and in the CBD areas. The comparison of SCATS and TRANSYT optimized
fixed times concluded that SCATS can improve travel time and number of stops from
3% to 18%. The actual improvement depends upon the type of road system and traffic

patterns for the area under study (Well 1987).

ARRB also conducted a comparative study of SCATS versus SCOOT. It
detailed the similarities and differences in the data requirements, hardware, and
operation of the two systems. However, a direct field comparison of the operational
differences between SCATS and SCOOT was not possible. Direct comparisons using
simulated or actual data are very difficult, due to the different locations where traffic
flow data is gathered. In SCATS, traffic information is collected at the approach stop
lines. The required traffic flow information for SCOOT is collected upstream of the
stop lines. The two systems also differ in their operating requirements for computer
processing. The paper stated that SCATS is better in some applications because it has
the capacity to estimate congestion better than SCOOT. By contrast, SCOOT can be
more effective in certain heavy flow situations because it incorporates an automatic

double—cycling mechanism, which SCATS did not have.

26

Kelman (1991) presented a performance report on the Metropolitan Toronta
SCOOT system. The study showed that time-of-day plans, in certain circumstances,
were better than adaptive strategies. This is further supported by work done by Gartner
et a1 (1994, 1995). They indicated that, while the largest possible benefits, (+10%
improvement) could be achieved using the most traffic responsive type of control, the
highest dis-benefits, (-10% deterioration) were also associated with the most traffic
responsive controls. Finally, it was suggested by Ort (1995), and Richeson and
Underwood (1996), that the majority of ITS benefits are related to increased safety,
and that these benefits were not directly attributable to the implementation of adaptive

versus fixed-time control strategies.

Abdel-Rahim and Taylor (1998) reported the results of a study conducted to
determine the change in travel time following the implementation of the SCATS in
Oakland County, Michigan. A before/after comparison was used to examine the
change in travel time on a specific corridor. The results showed that the corridor
travel-time improved for both directions for both the peak and the non-peak periods.
The reduction in corridor travel time ranged from 6.6% to 31.8%, with savings in
travel time being higher during non-peak periods. Before/after intersection delay
studies showed that the approach delay for the main street traffic decreased at the
intersections as a result of SCATS implementation. SCATS extended the green time

for the through traffic, reducing the average degree of saturation from 1.02 to 0.87

27

during peak periods and fiom 0.73 to 0.56 during non-peak periods. SCATS reduced
the green time for other approaches increasing the degree of saturation on the minor
approach from 0.52 to 0.63 during peak periods and from 0.22 to 0.31 during non-peak
periods. A before/after offset study showed that the through bandwidth increased
during all time periods for both directions, mainly as a result of extending the green

time for the main street traffic.

Andrews et a1 (1997) evaluated the New Jersey Optimized Policies for
Adaptive Control (OPAC) system. The evaluation compared the performance of
OPAC against a well-designed time-of-day fixed-time signal. The evaluation was
performed under various traffic demand conditions and included both isolated
intersections and arterial sections. The analysis indicated a significant improvement
with OPAC control. OPAC performed its best during oversaturated conditions. It
reduced the travel time and number of stops by about 26 percent and 55 percent,
respectively, for the entire arterial section. OPAC also improved traffic performance
during changing demand conditions. It significantly improved the performance of an
isolated intersection during unsaturated conditions. OPAC reduced stopped delay on
the major-street approach by 40 percent without affecting the minor street

performance.

28

2.4 Literature Review Conclusions

Many research studies have been performed to quantify the benefits of adaptive
signal control systems. Several studies on currently deployed adaptive control systems
claim to show significant travel time and intersection delay benefits. However, the
nature of the benchmark against which these improvements were measured is not
always clearly identified in these studies. For example, it is not known whether these
studies used as their benchmarks optimized time-of-the- day plans or simply used the
plans that were already in place. Therefore, it is possible that any improvements
gained through the implementation of adaptive signal control might also have been
achieved by simply updating the existing time-of-day fixed-time plans to more

optimized plans.

While simulation is an effective tool to compare the effectiveness of different
signal control strategies under the same environment, the literature search found few
research studies using simulation to compare alternative control strategies. This may
be due to the fact that none of the existing simulation models can fully replicate the

adaptive control logic.
The literature review helped identify the research gaps in assessing the benefits

of adaptive signal control strategies. First, none of the past studies compared different

adaptive control strategies for urban corridors against optimized coordinated fixed

29

time or actuated signals. Second, none of past studies used simulation models that can
fully replicate adaptive control logic. Third, none of the past studies examined
different traffic prediction techniques employed in the adaptive control logic, Finally,
none of these studies examined the sensitivity of the adaptive control benefits to

changes in demand levels and peak hour factors.

The review of published literature has .demonstrated the interest in and
importance of comparative studies of the various forms of traffic signal control. As a
result, studies that can assist current and future users of real-time adaptive signal
control systems to determine their expected benefits are extremely valuable. The
research conducted in this study is original in covering some of the areas that have not
yet been addressed. The research used the new version of CORSIM, which has the
ability to model adaptive signal control logic, to examine the benefits of adaptive
control logic against conventional coordinated fixed-time and full-actuated signals in
an urban corridor. The sensitivity of the benefits to changes in demand level and peak-
hour factor was also examined. The effectiveness of different prediction techniques

used for adaptive control logic was also examined.

30

Chapter 3

DEVELOPMENT OF THE SIMULATION MODEL

3.1 Simulation Modeling

Simulation can be used with some degree of confidence to examine the
microscopic as well as macroscopic aspects of travel in traffic networks. A number of
widely disseminated traffic simulation packages are available to support the analysis,
design, and evaluation of a traffic control system’s operation. Simulation models
incorporate analytical models, such as traffic flow models, car-following models,
shock-wave analysis and queuing theory, into a framework for simulating complex

components or systems of interactive components (Waugh, Clark, and Kanan, 1994).

Traffic simulation models can be classified based on the level of simulation
detail to three categories: microscopic, macroscopic, or mesoscopic. Microscopic
simulation models describe the detailed, time-varying trajectories of individual
vehicles in the traffic stream. With this level of details in the simulation, the
microscopic models can be used to study the effects of detailed control strategies on
different traffic parameters (Kosonen, 1990). Examples of the microscopic simulation
models are NETSIM for surface streets and FRESIM for freeways from the integrated
traffic family of simulation models (TRAF). The two models were recently integrated

into one microscopic model for both the freeway and surface street traffic (CORSIM).

31

Macroscopic models represent the traffic stream in some aggregate form (e. g.,
employing a fluid flow analogy or a statistical representation). Less detailed strategies,
involving changes in circulation patterns, for example, may be studied with
macroscopic models. These models may also be used to gauge the impacts of very
detailed strategies outside the boundaries of the area in which these strategies are
implemented (Rakha, 1989 and Rathi, 1990). Examples of macroscopic simulation
models are NETFLO 1, NETFLO 2 for surface streets and FREFLO for freeways from
the TRAF family of simulation models. The three models were also integrated into one

macroscopic model for both freeway and surface streets (CORFLO).

Another example of a macroscopic traffic simulation model is TRANSYT. It is
among the most realistic of those available in the family of computerized macroscopic
traffic models. The traffic model utilizes analytical flow-density-speed relationships
and platoon dispersion algorithm to simulate the normal dispersion of platoons as they

travel throughout the traffic network, (Wallace et. al., 1998).

Recently, several “mesoscopic” models have been developed. These models
track the movements of individual vehicles, as microscopic models do, but model their
movements using macroscopic flow equations. An example of mesoscopic simulation
models is the INTEGRATION model, which was conceived during mid 1980s as an

integrated simulation and traffic assignment tool (Stewart and Van Aerde, 1998).

32

The selection of the level of simulation detail and the model that should be
applied is based on the physical environment to be studied, the level of detailed
analysis desired, the availability of data, and the objectives of the study (Waugh,
Clark, and Kanaan, 1994). As the scope of this study involved studying microscopic
signal control strategies in an urban traffic network, a higher level of analysis is
required, and a decision was made to consider a microscopic or mesoscopic simulation

model for the analysis.

3.1.1 Selection of the Simulation Model Suitable for ITS applications

While traffic simulation covers a wide spectrum of traffic theory and
engineering topics, the current interest in ITS technologies raises significant
challenges for traffic simulation models. Since ITS technologies encompass both local
and area-wide traffic control strategies, the scope of the traffic simulation model must
incorporate microscopic signal control, a vehicle rerouting mechanism, traffic
surveillance, and a dynamic optimization algorithm that is capable of optimizing

traffic signal control on an on-line basis, (Mahmassani, 1992).

A growing variety of simulation models are being developed for ITS
applications. Three simulation models, that have the potential to simulate adaptive
signal control strategies, were examined in the early stages of this study. The three
models are the INTEGRATION simulation model, the Traffic and Highway Objects

for Research, Analysis and Understanding (THOREAU) simulation model, and

33

CORSIM form the F HWA’s TRAF family of simulation models.

The INTEGRATION simulation model, developed in the University of
Queens-Canada, has the ability to model a responsive signal control system through an
automatic signal re-timing algorithm included in the model. When this automatic
signal re-timing is utilized, only the offsets and the lost time specified in the original
signal-timing plan are kept constant. The other timing plan parameters (cycle length
and phase splits) are optimized each time interval based on the volume/saturation flow
ratios in the competing approaches, (Stewart and Van Aerde, 1998). While the model
provides a fiamework for the simulation of ITS implementation alternatives, it does
not allow the users to incorporate their own signal optimization algorithm or adaptive
control logic within the model, which limits the number of control strategies that can
be tested using this model. The model also does not have the capabilities to test

dynamic offset plan alternatives.

THOREAU is an object-oriented, trip specific, microscopic and macroscopic
traffic simulation model designed for comparative ITS analysis. The model adopts
obj ect-oriented, discrete event simulation for vehicle movements along arterial streets,
through intersections and on the freeways. The model incorporates adaptive signal
control strategies similar to the UTCS strategies, which, allows an analyst to study
different adaptive signal control strategies, (Hsin and Wang, 1992). A major

disadvantage of the model is that it requires a dynamic input-output matrix to generate

34

vehicle trips, something that is not generally available. Similar to INTEGRATION, the
model does not allow users to use their own control logic, or to modify the control

strategies included in the model.

The current version of CORSIM has the capability of modeling the operation
of real-time traffic adaptive control signals. The model allows the transfer of detector
surveillance data to an external real-time signal control algorithm, which updates the
signal control parameters on an on-line basis. The user defines the frequency and type
of information to be sent between the model and the external real-time control
algorithm. Based on the review of the available simulation models, a decision was
made to use CORSIM as the simulation tool for this study. The latest version of

CORSIM (1.05 Alpha/I SIS v4.5), issued in March 1998, was used in this study.

3.1.2 Description of CORSIM (Structure, Principles, and Features)

The first version of NETSIM was developed for the Federal Highway
Administration (F HWA) nearly two decades ago. The program, formerly called
UTCS-1, was later integrated into the TRAF simulation system. Over the years,
NETSIM and FRESIM have gained acceptance in the traffic engineering community
as presenting a reasonable representation and MOEs of traffic movement along surface
street and freeway systems, though not in a combined network. In 1997, CORSIM was
introduced as the product of combining FRESIM and NETSIM into a singularly

defined and integrated traffic network, (F HWA, 1997).

35

The CORSIM model is based on the application of interval-based, discrete
event simulation to describe the dynamics of traffic operations. The vehicles are
represented individually and their operational performance is determined uniquely
every second. Furthermore, each vehicle is identified by category, type, and driver-
behavioral characteristic. There are 16 different vehicle types with various operating
and performance characteristics. Drivers are classified into three different driver
behavioral categories (aggressive, normal, and passive). The driver-vehicle
characteristics are assigned stochastically, as are other performance and behavioral
attributes including turning movements, free-flow speed, queue discharge headway,
and acceptable gaps. Consequently, each vehicle’s behavior may be simulated in a
manner reflecting real world processes. In general, most operational conditions
experienced in an urban street network environment are realistically described (Rathi

and Santiago, 1990).

Each time a vehicle is moved, its position (both lateral and longitudinal) on the
network, its relationship to other vehicles, and its kinematics properties (speed,
acceleration and status) are recalculated and updated. Vehicles are moved according to
car-following logic while responding to traffic control devices, pedestrian activity,
transit operations, performance of neighboring vehicles, and other conditions that
influence driver behavior. Actuated control and interaction between passenger cars and

buses are explicitly modeled.

36

CORSIM has the capability to simulate the effects of all types of intersection
control. Signal controllers can be fixed-time, multi-dial, semi-actuated, or full-actuated
signals. Signals can also be set to operate as real-time adaptive signals using the

optional run-time extension capabilities.

The simulation procedure in CORSIM consists of a warm-up (initialization)
period and an actual simulation period during which statistical data are accumulated.
During the warm-up time, traffic generators feed vehicles into the empty simulated
network until equilibrium conditions are reached, that is, the rate at which vehicles
enter and discharge from the network is equal. Having reached equilibrium the model

starts to accumulate statistics on different traffic parameters in the network.

3.1.3 Modeling an External Controller within CORSIM

CORSIM’s fimctionality has been extended through the development of the
new Traffic Software Integrated System (TSIS). TSIS is a Microsoft Windows
(95/NT)-based environment for CORSIM. It provides an intuitive user-interface to
CORSIM (as well as other FHWA models in the future). TSIS has also provided a
fundamentally new approach to using the model to research ITS technologies by
allowing other application programs to interact with CORSIM while the simulation is
executing, and further, these other programs can have an effect on internal CORSIM

operations. This type of application is known as a Run Time Extension (RTE).

37

RTE can be used to test adaptive signal control algorithms for effectiveness,
before they are implemented in the field. With the rapid evolution of technology,
these types of experiments are necessary, not only to assess the benefits of advanced
ITS applications, but also to verify their operational capability before field
deployment. This new capability is a critically important step in extending the

usefulness and effectiveness of simulation-based research and analysis.

To understand how real-time extension programs function within CORSIM,
the structure and the component of TSIS/CORSIM should be understood. The TSIS is
composed of several major components including the “shell” application TSIS.EXE
and several Windows Dynamic Link Libraries (DLL). The four major components of
the TSIS/CORSIM are:

1. TSIS.EXE, the executable MS-Windows program, the new host for operating
CORSIM.

2. CORSIM.DLL, which is built from the configuration-controlled FORTRAN

source code.

3. TSISINTFDLL, the TSIS component that contains the shared memory and the
communications interface, all of which are utilized by CORSIM and CORSIM
RTE.

4. Optional RTE in DLL format, can be designed and implemented to interact with
CORSIM at run-time

38

Figure 3.1 illustrates the overall functional diagram of TSIS and its components

 

 

 

 

 

 

 

 

 

 

(F HWA, 1997)
TSIS.EXE
TSISINTF.DLL
ADAPTIVE.DLL CORSIM.DLL

 

 

 

 

 

 

Figure 3.1 TSIS-CORSIM RTE interface structure

39

3.2 Signal Control strategies

The study was formulated to test the effectiveness of adaptive signal control
strategies against two conventional, widely applied conidor signal control strategies
namely: coordinated fixed-time and actuated signals. Two adaptive control systems
that equalize the degree of saturation on critical approaches were examined. The
difference between the two adaptive strategies lies in the way traffic flow predictions
were made. In the first adaptive control strategy, traffic predictions were made based
on detectors placed downstream fiom the intersection, whereas, in the second
adaptive control strategy the predictions were based on detectors placed upstream of

the intersection.

3.2.1 Coordinated Fixed-Time Signals
3.2.1.1 Definition

In a coordinated fixed time signal system, the cycle length is constant for all
signals during any given control period. Cycle length, green split, and offsets are
optimized off-line based on historical traffic profile data. The values of all signal
timing parameters are kept constant during any given control period, but the signals
may have different settings based on time-of-day. In this study, the signal settings

were kept constant during the one-hour analysis period.

40

3.2.1.2 Optimizing Fixed Time Signal Plans

The design of corridor traffic signal timings has significant impact on traffic
flow parameters. As the major objective of this study is to compare the effectiveness
of different control strategies, it was important to develop optimal settings for the
fixed-time signal control under different traffic conditions. TRANSYT and PASSER
are the most widely used tools for optimizing signal-timing plans for corridors and

traffic networks.

There are five essential elements that constitute pretimed signal timing in a
coordinated signal system: cycle length, number of phases, phase sequence, phase
lengths, and offsets. TRANSYT explicitly optimizes phase lengths and offsets for a
given cycle length. The program is applied at the network and/or corridor level
wherein a consistent set of traffic conditions is known and the system hardware can be

integrated and coordinated with respect to a fixed cycle length and coordinated offsets.

Optimization is performed in TRANSYT through the simulation of vehicle
responses to various signal settings. The traffic simulation model in TRANSYT is
among the most realistic of those available in the family of computerized macroscopic
traffic models. The traffic model utilizes a platoon dispersion algorithm that simulates
the normal dispersion of platoons as they travel downstream. To determine the best
cycle length, an evaluation of a user-specified range of cycle lengths may be made.

TRANSYT does not select the phase sequence--these are required inputs. To examine

41

alternative phase sequences, multiple computer runs are required. When optimizing,
TRANSYT minimizes (or maximizes, depending on the selection) an objective
function called the Performance Index (PI) which is either a linear combination of
delay and stops, fuel consumption and excessive maximum back of queue; or excess
operating cost. This PI is often referred to as “performance optimization”, and is to be

minimized, (Wallace, 1998 and Hadi, 1992)

Another performance index alternative is the forward progression opportunities
“PROS”. It can be used either alone or in combination with the delay and stops (or
fuel consumption) disutility function. This PI is maximized to increase the quality of
perceived progression. The components of the PI are:

1) Standard TRANSYT disutility index (DI), the “standard” delay and stops is a linear
combination of these measures:
DI=[delay on a link * link-delay weighting factor] + [stop penalty * stop * link-
specific stops weighting factors]
2) Excess operating cost,
3) Progression Opportunities (PROS), and
4) Optimization objective function which may be defined by the user in a number of

ways:

42

Standard delay and stopsl
D1 = Excess fuel consumption + [double count progressive links] + [queue penalty]

Excess operation cost

 

 

PASSER is another tool for corridor signal timing optimization. The model
consists of three optimization software programs that optimize traffic timings on
single roadways or entire networks of roadways. PASSER-II can be effectively used
for arterial progression optimization (bandwidth maximization), existing system
evaluation (simulation), and intersection capacity evaluation. PASSER helps traffic
engineers and planners develop timing strategies that optimize the flow of traffic on a
single road or through the entire network by maximizing the progression band. Cycle
and offsets optimization in PASSER is based on bandwidth efficiency optimization.
By determining the maximum band in both directions of a roadway, the program can
calculate timings that allows the greatest number of cars pass through the system

without stopping, (Abu-Lebdeh and Benekohal, 1997).

A key factor in the comparative study conducted in this research is to compare
the adaptive signal control strategies against well-optimized fixed-time and actuated
signals. This will ensure that any benefits reported in the study would be solely the
result of the adaptive control logic and could not be achieved through any other fixed-

time or actuated signal settings. For this reason, different settings for the fixed-time

43

signal control were examined to determine the optimal signal parameters. The
Highway Capacity Manual (HCM) defines arterial/corridor performance in terms of
the average speed of through vehicles, (Anderson, 1996). An optimal setting, as
defined in this study, is the setting that would achieve the highest average

speed/lowest average corridor travel time among all alternatives.

TRAN SYT-7F and PASSER II-90 were used to obtain alternative “optimal”
signal settings (offset, cycle length and green split). The CORSIM simulation model
was used to compare the performance of these alternative settings. Average corridor
speed and total travel time were used as measures of effectiveness to determine the
optimal signal setting among these alternatives. The results showed that offsets and
green split, obtained from PASSER II-90, provided the highest average speed and

lowest total travel time.

3.2.2 Coordinated Actuated Signal Systems
3.2.2.1 Actuated control features and operation

Actuated signals allocate the green time for each phase on a cycle-by-cycle
basis in response to detector actuation. Detectors, placed upstream to the intersection,
provide the actuated controller with information concerning current demand. Signal
actuation, accordingly, permits phases to start early or to end late to meet the variation
in demand. This flexibility should make it possible to improve the performance of the

control strategy.

44

Each actuated phase has the following features that must be set on the

controller:

Minimum Green Time: Each phase has a minimum green time. Older controllers

 

divided this time into two portions: an initial green interval and a unit extension. The
initial green interval was intended to provide sufficient time for all vehicles potentially
stored between the detector and the stop line to enter the intersection, the unit

extension allowed another vehicle to travel from the detector to the stop line.

For all actuated phases, the minimum green time must be established. This is
generally equal to an initial interval that allows all vehicles potentially stored between
the detector and the stop line to enter the intersection. A start-up time of 4 seconds is
incorporated in addition to 2 seconds for each vehicle (the latter based on a saturation

flow rate of 1800 vph).

PassaggTime Interval: The passage time interval allows a vehicle to travel from the

 

detector to the stop line, and is analogous to the “unit extension" of older controllers.
The passage time setting, however, also defines the maximum gap between vehicles

arriving at the detector to retain a given green phase.

Maximum Green Time: Each phase has a maximum green time. If demand is

 

sufficient to retain a given green phase to this limit, the green will arbitrarily

45

terminate. Maximum green times are generally set by working out an optimal cycle

length and phase splits as if the controller were pretimed.

Figure 3.2 illustrates the operation of an actuated phase based on these three
settings. When a green indication is initiated, it will be retained for at least the
specified minimum green period. Additional detector actuation may occur during this
minimum green period. If they occur, and there is at least one passage time left before
the minimum green terminates, no green time is added. If a vehicle arrives during the
minimum green and there is less than one passage time lefi before its termination, an
amount of green time equal to the passage time is added. The controller now enters the
extension portion of the phase. If a subsequent actuation occurs within one passage
time interval, another passage time interval is added to the green. The green is
terminated by one of two mechanisms:

a) a passage time elapses without an additional actuation, or

b) the maximum green time is reached and there is a call on another phase.

Coordinated actuated signal systems operate on a common background cycle

length. Synchronization is provided through the yield point, which is a fixed point in

the cycle length, normally during or at the end of the non-actuated phase.

46

Total Green Time
‘7;

Minimum Green Time Extension Green Period
H an

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. fl '1

 

 

 

 

 

 

 

Actuation V
‘ Maximum Green Period
’ Time
Legend
Detector actuation on Detector actuation on unexpired portion
phase with right-of-way a conflicting phase vehicle extension

Figure 3.2 Operation of an actuated phase

47

The non-actuated phase has some minimum green time, plus all slack time not
used by actuated phases. The green times on all actuated phases vary on each cycle
between a minimum green and a maximum green, depending on the arrival rate of
vehicles and the value of the extension intervals. Fixed force-off points in the
background of the cycle are used to terminate the duration of the actuated phases. If an
actuated phase terminates early, then any green time not used by the actuated phase is
transferred to the non-actuated phase: For this study, the green time for the major
corridor was set as the non-actuated phase. Figure 3.3 illustrates the coordinated

actuated signal controller.

3.2.2.2 Optimization of Actuated Signal Control Settings

Optimization of coordinated actuated signals include two main objectives:

1) optimal design of cycle length and offsets

2) optimal design of detection scheme (detector length and location) and gap-time

settings.

TRAN SYT-7F has the capability of optimizing actuated signal time settings.
However, the program does the optimization through calculating the average values of
green time and cycle length. For the purpose of this study, the optimal cycle length
and offsets obtained form the fixed time signal settings were used as the background

cycle length and offsets for the actuated signals.

48

Phases4 & 8 Clm ...

Phases 1& 5
Begin Pgmissive4pgigd 3 / End Permissive Period 3
/ Phase 1&5 Clearance
mm! wk StartofCoordinstionPhase
Phases 2&6
minimum green
Phases 3 & 7 Clearance
Poms-05M”? YieldPoint&loealT-0
BengermrssrvePenodZ BeginPermissivePeriodl
End Permissive Period 1 I E“ °f mm“ 91"”
Phases 3 & 7 ‘
mrmmurn green
Phases 2&6 Clearance

Figure 3.3 Typical timing dial diagram for coordinated actuated signals in

CORSIM

49

A considerable number of studies have examined the optimal detector length
and location for fully actuated signal systems. Lin (1992), recommended that a 20-foot
detector located approximately 120-200 feet upstream from the intersection would be
optimal for most urban area actuated controllers. Many studies (Lin, 1992), and (Rathi
1994) recommended a gap-out of 3 to 5 seconds to provide an optimal control for
urban area traffic. For this study, traffic simulation was used to determine the optimal
detector location and gap-out among alternative settings. Detectors placed 180 feet
upstream to the intersection with a gap-out time equal to 3 seconds provided the

optimal setting.

3.2.3 Real-time adaptive signal control
Figure 3.4 shows the components of real-time adaptive control. The control consists of

a detection method, traffic prediction model, and signal optimization model.

3.2.3.1 Prediction models for adaptive Signal Control

The effectiveness of adaptive systems is dependent on the accuracy of the real-
time traffic predictions. Traffic predictions are made for some time in the future.
During a computational process for optimization of a specific cycle, only the
predictions from the previous cycle/cycles data are available. Plans optimized using
data collected during the cycle (N) would be implemented in the cycle (N+2) (Figure

3.5).

50

 

 

Signal Timing
Optimization Model X I Real system

1 1

Traffic Profile
Prediction Model 4— Detector data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.4 Network flow model for real-time adaptive control

 

I

 

 

 

Set up
Put new plan into effect
Compute for this period
Plan here '4 +|
Control I 1 I '
Period l N J (N+1) l (N +2) I
Collect data to here

Figure 3.5 Prediction sequence for adaptive control strategies

51

Prediction of the traffic profile can be made in two ways, using detectors
placed upstream or downstream from the intersection. Detectors placed downstream
from the intersection predict a traffic profile for a specific cycle based on the
volume(s) of previous cycles. In essence, they predict changes in traffic demand over
time rather than cycle-by-cycle variation in demand. Detectors placed upstream to the
intersection, however, use “scan-ahea ” and predict the arrival of vehicles during a
specific cycle based on estimated time of arrival of the platoons to junctions further

downstream.

For this study, the effectiveness of both prediction techniques was examined.
In the first adaptive control strategy, predictions of the traffic profile at the intersection
was obtained from detectors placed downstream from the stop-bar at the intersection.
The expected volume for the new cycle (n+1), was estimated, similar to SCATS

adaptive control, based on volumes departed during cycles (11, n-1, n-2) as follows:

V, +1 = 0.5 V, + 0.3 V,,_,+ 0.2 V,,_2

In the second adaptive control strategy, predictions were made based on
detectors placed upstream from the intersection. Detector locations are presented in
Figure 3.6. The use of upstream detector to predict the arrival of vehicle platoons at
the intersection is demonstrated in Figure 3.7. The optimization sequence of the

adaptive control strategies is presented in Figure 3.8.

52

Detector locations
(Minor-street)

 

 

 

 

 

/ -
Detector locations

(Major-street)
fl Downstream detectors

 

 

 

1600 feet

 

 

 

 

 

 

 

< 4
- 7
Detector locations

(Minor-street)

3500 feet

Detector locations
(Major-street)

N q Upstream detectors

Figure 3.6 Detector locations for the two adaptive control strategies

 

 

 

53

Update signal setting

    
  

Intersection location

Average speed 30 mph

\

Detector location T1 T2

 

-I

 

Detector data H
(predicted volmne

Figure 3.7 Use upstream detectors to predict vehicle arrivals

I‘ Cycle N I Cycle N+1 I Cycle N+2

Yellow
intervals All-red

\\ I— interval
:H ' ' flEH:

I‘Collect detector data >I
I Run optimization program

Update signal settings

 

 

 

 

 

 

 

Figure 3.8 Optimization sequence for adaptive control strategies

54

3.2.3.2 Optimization models for Adaptive Signal control

Once predictions were made for the expected number of vehicles in a
subsequent cycle, these data were used to calculate the optimal cycle length and green
split. Cycle length was updated every five minutes. The cycle length was
increased/decreased to obtain an optimal degree of saturation of not higher than 0.9 or
lower than 0.8 on the corridor link that had the highest expected traffic volume. The
green splits, however, were optimized on a cycle-by-cycle basis for each intersection.
Optimal green split was obtained by maintaining an optimal degree of saturation of
0.92 for the minor-street traffic and assigning the rest of the green time for the major-
street traffic. This method would maximize the throughput of the corridor and
maintain adequate delay levels for the minor-street traffic. Offsets were chosen based
on the value of the cycle length from a library of pre-optirnized offset/cycle length
values. This library was developed from TRANSYT-7F optimization of the conidor

under different demand levels.

Before using run-time extensions for testing adaptive signal control algorithms,
the input file must be edited to allow non-source nodes of the network to be placed
under the control of a signal control algorithm separate from and external to CORSIM.
The signal state for these nodes is controlled by the external control algorithm. The
run-time extension program was written in FORTRAN 90 and was compiled using
DIGITAL Visual Fortran V5.0 to obtain the DLL file used by CORSIM. Details of the

optimization program are provided in Appendixes

55

CHAPTER 4

DESIGN OF THE STUDY

There are a number of unanswered issues that should be addressed before
large-scale deployment of adaptive signal control technologies. The magnitude and
consistency of the benefits of these control strategies under various traffic conditions
should be fully evaluated. While field demonstrations are the only way to test new
strategies in real-world conditions, traffic simulation, as an alternative, provides the
mechanism for testing theories, modeling concepts, and control strategies. Simulation
models are particularly valuable in identifying key operational and performance issues
for different control strategies under a range of scenarios. Simulation studies have
many advantages over field demonstrations; they are less costly, the results are
obtained quickly, many variables can be held constant, and the data generated through

simulation include several MOEs that may not be easily obtained through field studies.

4.1 Study Plan

For this study, simulation provided an opportunity to compare different signal
control strategies for an arterial corridor using the same network configuration and
traffic pattern. The study is designed to test different signal control strategies under a

controlled environment using the CORSIM traffic simulation model. The control

56

strategies considered in the study are:

l) coordinated fixed-time signals,

2) coordinated actuated signals,

3) adaptive signal control that equalizes the degree of saturation on critical approaches
with traffic flow prediction based on detectors placed downstream fiom the
intersection (adaptive_l ); and

4) adaptive signal control that equalizes the degree of saturation on critical approaches

with traffic flow predictions based on detectors placed upstream of the intersection

(adaptive_2).

The study was organized in five parts, each designed to achieve one of the
objectives of the research. The first part was a comparative analysis of the
effectiveness of the four signal control strategies under two different demand levels:
moderate peak-hour demand and high peak-hour demand. The second part was a
signal timing analysis designed to test the changes in traffic parameters reported in the
comparative analysis. The third part of the study was a sensitivity analysis to examine
the benefits of the adaptive control systems over fixed time signals under four different
demand levels and peak-hour factors for both maj or-street and minor-street traffic.
Changes in the adaptive control system benefits resulting from adding two left turn
phases to the signal operation were also examined in this section. The fourth part of

the study was a comparison of the accuracy of the prediction techniques used in the

57

adaptive control systems in this study. The final part was a field study to validate the
output of the simulation model. A comparison of the cost and the benefits of different
signal control strategies is also presented in the final part of the study. The study plan

is shown in figure 4.1

4.2 Hypothetical Corridor Network

There is a general resistance among transport modelers to use hypothetical
networks for simulation work because of the potential for introducing unrealistic
features (Van Vuren and Leonard, 1994). Nonetheless, a theoretical network is
commonly used when the main objective is to test a serious of scenarios that have not
been applied in an existing network. Using a hypothetical network allows for the
generalization of the output of the study beyond any site-specific configuration of an

existing network.

A postulated surface street corridor was constructed for this study. The corridor
has a total length of 3.6 mile and includes five intersections. The geometric
configuration of the conidor is presented in Figure 4.2. The link-node diagram of the

tested corridor is also presented on Figure 4.3

58

 

 

Comparative analysis to test the effectiveness of four signal control strategies

 

 

I1

[1

 

 

Moderate peak-hour demand case

Moderate peak-hour demand case

 

 

 

 

 

 

Explain changes in traffic parameters

 

 

I1

 

 

Signal timing and degree of saturation
analysis

 

 

 

 

Sensitivity analysis

 

 

J1

fl LL

 

 

Travel demand and PHF

 

 

 

Phase plan Changes in demand over
time saturation

 

 

 

 

 

Compare the accuracy of the traffic prediction techniques used in the study

 

 

 

Field study to validate the output of the simulation model

 

Figure 4.1 Study Plan

59

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0

Intersection l h E

—. . 6

i

Intersection 2 L g
1

i

Intersection 3 xe,

__ " Q
1

g i
., E
Intersection 4 L

1

i

Intersection 5 L '7
T

i

V:

O

 

 

 

 

 

 

 

 

Figure 4.2 Geometric Configuration of the Corridor

60

i
o—o—o—o—o

®—0—?—9—0
0—0— 3 "6—®

V A

@--0-- i —0—@

 

 

e—o—clr—o—e

8035

Figure 4.2 Link Node Diagram for the Corridor

61

4.3 Traffic Demand Levels

Demand levels were selected to represent different peak-period traffic
conditions in urban areas. Two demand levels with different peaking factors,
representing moderate and high peak-periods were selected for both the major corridor
and minor streets. The traffic volumes and peak-hour factors were based on an analysis
of the volumes on Orchard Lake Road, an urban corridor in Oakland County,
Michigan. The distribution of the traffic over an hour period for different demand
levels is presented in Figure 4.4. As in most corridors in urban areas, during any
specific time period, the demand varies between the two directions of travel along the
corridor (peak and non-peak directions). The demand for the non-peak direction of

travel was set at 75% of the peak-direction demand for both major and minor streets.

4.4 Measures of Effectiveness

The improvement brought about by changes in traffic control can be assessed
using different measures. To compare the effectiveness of alternative control
strategies, several measures of effectiveness that are appropriate for characterizing
traffic flow conditions were studied. The MOEs selected for this study are:
l-Average corridor travel time
2-Total corridor travel time
3-Average stopped delay per vehicle per approach

4-Average stopped delay per vehicle for the intersections.

62

Major Street

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Minor-Street
Demand (vph)
Demand (vph)
1600 16“”
1200
1200 / \
soo 8““
400 4“" f \\
Time (minutes) Time (minutes)
Moderate-Peak Hour Demand
Demand h
1600 (VP ) 1 6133mm (vph)
1200 1200
80° son
15 30 35 60 15 30 35 60
Time (“mm“) Time (minutes)
High Peak-hour Demand

Figure 4.4 Deamnd Level For Major Corridor and Minor Streets

4.4.1 Average corridor travel time

The efficiency of any traffic signal control system can be measured in terms of the
difference in total time taken by a vehicle to complete a designated trip from a specific
origin to a specific destination. Travel time studies are commonly used to measure
arterial levels of service. The travel time characteristics within a corridor are very
useful as an indicator of overall system performance. Travel time is an indication of
delay on a broader scale than intersection delay, which only measures the delay at a
single intersection or specific approach. Average link travel time, as reported by
CORSIM, is the average travel time of vehicles that have completed their trip across

the link.

4.4.2 Total corridor travel time

Total travel time is the sum of the travel time for all individual vehicles completing
their trips. This measure considers the number of vehicles, distance of travel, speed
and delay associated with travel. Total travel-time may be the most appropriate
indicator for a system operator, who seeks to minimize the overall system-wide travel

time.

4.4.3 Average delay per approach
Delay reflects the traffic operations as perceived by system users. The research study

assessed the difference in average approach and stopped delay. The term approach and

stopped delay, as used in this study, are defined as the following:

Average Approach Delay: is the length of time that vehicles approaching the

 

intersection from a particular direction are delayed due to a red signal and/or stopped
queue which prohibits their free flow travel through the intersection. Free flow travel
time represents the time the vehicle takes to travel through the intersection approach
section using the average speed of vehicles on the approach. Average approach delay
is the sum of the approach delay for all vehicles divided by the total number of

vehicles.

Averag£Stopped Delay: is the amount of time that a vehicle spends in a queue plus

 

the time it takes to travel from its position in the queue to the approach stop line. This
is the measure used in the Highway Capacity Manual to assign a Level of Service
(LOS) rating to signal controlled intersections. Average stopped delay is the sum of

stopped delay divided by the total number of vehicles.

4.4.4 Average intersection delay
The weighted-average of the intersection delay for all approaches is a good indicator
of how well the control strategy served the overall intersection rather than for a single

approach.

65

4.5 Analyzing CORSIM Output/ simulation iteration

CORSIM uses the ratio of sample means of observations to calculate the
average per vehicle parameters (average travel time per vehicle, average number of
stops per vehicle, average speed, and average delay time per vehicle). Each of these
MOEs is the ratio of the means of two random variables X and Y. For example, the
average delay per vehicle is obtained by dividing the total delay time accumulated on
each link by the total number of vehicles discharged from the link. To make valid
statistical statements about these parameters, one must apply statistical techniques to
the model output that are based on ratio estimates. The most common methods to
analyze the output of the stochastic simulation models are the batch mean and the

replication methods.

The batch mean method is performed by running the simulation model for one
long run, then dividing it into smaller time intervals (batches). For each batch,
statistics are collected and variability among batches is used to build a confidence
interval on the simulation output parameters. The advantage of the batch means
method is that it requires only one initialization period, which reduces the simulation
execution time significantly. However the length of the intervals should be set long

enough to reduce the correlation among batches.

66

The replication method is performed by running the simulation for a number of
independent runs, each run has its unique initialization time (warm-up period) until the
system reaches an equilibrium condition. After the warm-up period, statistics on
system performance are collected. Each run should have a unique random number seed

to generate random variation among the runs.

Gafarian and Halati (1986), developed a statistically valid method for using the
value of the ratio X/Y as a point estimate for the ratio x/y. The method provides a
measure of the accuracy of the estimates in term of confidence intervals. The
development of the confidence interval for the estimate x/y is based on the
observations {(Xi/Yi), i=1, 2, ,n)} of all independent runs of the simulation model.
The validity of this method depends on how well the assumptions made in the
derivation are met by the CORSIM model. These assumptions are: system in steady

state, independent runs, and normality of X and Y.

Steady state in the CORSIM is achieved by the warm-up procedure. To obtain
independent runs, each run is started with a different seed number. The only
assumption that is only approximately met is that of normality, and it has been shown

that the method is not sensitive to this requirement, (Gartner and Hon 1994)

67

According to the Gafarian and Halati study, (Gafarianl986), when the number
of simulation runs exceed 40, then the coverage probability for the 95% confidence
intervals of the estimator will approach 0.95. In a comparative study for evaluating
alternative control strategies, Gartner and Hon (Gartner, 1992) stated that a sample
size of four simulation runs would provide 0.80 coverage probability of the 95%
confidence intervals. In this research, a pilot study was used to determine the optimal
number of iterations to be applied throughout the course of the simulation model and

to study the variability associated with the randomness of the traffic.

4.6 Number of simulation iterations

The CORSIM model requires a random seed number for the random number
generators that govern the flow of vehicles into the network. CORSIM generates
different traffic patterns of traffic flow and responses to traffic choices. The user
defines two random number seeds. The first one is used for generating vehicles for
traffic streams. The second one is used for all time dependent stochastic processes. By
keeping all other modeling elements constant and varying the seed number for traffic
flow the variance of performance measures due to variation in traffic patterns can be

illustrated.

The preliminary pilot study consisted of 40 identical CORSIM input data sets

in terms of geometry, traffic control, traffic volumes, and turning percentages.

68

However, different random number seeds were used in each run. The simulation runs
produced 40 simulation outputs. Average corridor travel time for the 40 runs is
presented in Figure 4.5. The output shows the variability of the travel time values due
to the randomness of the traffic patterns. The graph indicates that the randomness of
the traffic pattern accounts for variations of up to i1.8% of the mean for corridor

travel time (Figure 4.5).

The data from the 40 simulation runs were firrther analyzed to determine the
number of iterations required for each simulation run. Based on the cumulative
averages from iteration 1 to 40, the average values were plotted against the number of
iterations included, Figure 4.6. The graph show that the cumulative average improves
significantly after 4 to 5 iterations. Accordingly, it was decided to conduct at least five

iterations with different seed numbers for each case.

69

Average corridor travel time

(seconds)

Average corridor travel time

400.0

 

395.0 .
390.0 -
385.0 -
380.0 .
375.0

370.0 -
365.0 _
360.0 .
355.0 I

 

350.0

 

 

10

20 30

run number

Figure 4.5 Average Corridor Travel time for 40 simulation
Iterations

40

 

(a)
GQCOCO
601001

\1
O

wwwwwww
O) \l
01 0|

0)
O

355 .
350

 

O

3

-0

9 O

.-:-’-.-.-.-.-.-.-{}

 

 

0

1

‘I

j

2 3 4 5 6 7 8 9101112131415
number of simulation iterations

Figure 4.6 Number of simulation iterations

70

Chapter 5

ANALYSIS AND RESULTS

The output of the simulation study is analyzed and presented in this chapter.
The chapter is divided into four main sections. The first section includes the results of
a comparison of the effectiveness of different control strategies under different
demand levels. A common set of traffic parameters, generated from the simulation
runs, was used to measure and compare the effectiveness of these control strategies. A
comparison of the signal timing characteristics under different control strategies is
presented in the second section. The signal timing analysis was conducted to examine
how the dynamic green time allocation in both actuated and adaptive signal control

systems contributed to the changes in the traffic parameters.

The benefits of the adaptive control systems over fixed time signals were
examined under four different demand levels and peak-hour factors for both major-
street and minor-street traffic. The results of a sensitivity analysis are presented in the
third section of this chapter. Changes in the adaptive control system benefits resulting
from adding two left-turn phases to the signal operation was also examined. Finally, a
comparison of the accuracy of the two prediction techniques used in the adaptive

control systems in this study is presented in the fourth section of this chapter.

71

5.1 Comparative analysis of signal control strategies

The CORSIM model was used to simulate traffic conditions under four
different control strategies; fixed-time signals, actuated control signals, and two
different adaptive control systems. The effectiveness of these control strategies was
examined for under two different demand levels, representing moderate and high peak-
hour cases in an urban corridor. Traffic parameters, generated from the simulation
runs, were used as measures of effectiveness in the comparative analysis. The
parameters included were; average corridor travel time, total corridor travel time, and

average delay per approach and for the intersection.

5.1.1 Corridor travel time

The efficiency of any traffic signal control system can be measured in terms of
the total time taken for all vehicles to complete their trip throughout the corridor. The
travel time characteristics within a traffic network are very useful as an indicator of the
overall system performance. Travel time studies are commonly used to measure
corridor levels of service and to compare the effectiveness of alternative signal control

strategies.
Figure 5.1 shows the average corridor travel time under four different control

strategies for the moderate peak-hour demand case. The results are aggregated over 5-

minutes intervals. The results are also presented in Table 5.1. All three control

72

strategies showed an improvement in average corridor travel time over fixed-time
signal control. The average corridor travel time was 399.02 seconds/vehicle under
fixed time control. This average decreased to 385.92 seconds/vehicle, 382.58
seconds/vehicle, and 383.57 seconds/vehicle under actuated, adaptive_l and

adaptive_2 controls, respectively.

The overall savings in corridor travel time were 3.28% under actuated-signal
control, 4.12% under adaptive_l control, and 3.87% under adaptive_2 control. For the
three control strategies, the savings in travel time decreased as the traffic demand
approached its 15-minute peak period, then increased again as traffic demand
gradually decreased. The savings in travel time was as high as 8.45% in periods of low
demand and as low as 0.0% during the peak 15-minute period, Table 5.2. The average
savings in travel time during the first 15 minutes of the peak-period were 6.05%,
7.13%, and 6.77% for the three control strategies, respectively. These values were
0.59%, 0.96%, and 0.73%, respectively, during the peak lS-minute period in the peak-

hour.

73

.83 32.52... 829.325
8.93.5. .9559 .3»? 395...... 3...:— oE.» .96.: 3...»..3 nausea ..m 95”....

 

 

 

 

 

 

 

2a....
8 _ m3 8 o a. a 85 mm c on e 2 e 85 25 S e 2 e 85

. n. A M 8...

........................................................................................................................................................................................................ “NM
w
52 n
B
a
m
.3... u
m.
0
J
:9. u
A
n
3.8 m
u. . o
N geesisi ” . U . _ . H H H m
......... _Io>ee€<+ . as. m

< v . . _ _ t i a g . . .
......... 08:83:01 23.3 c 8muu$§8§3ou§ ms.

 

 

 

 

83$... ea. 8: u 9E8. 833%....

 

 

8m

74

Table 5.1 Average corridor travel time* under different control strategies
(moderate peak-hour demand)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Fixed-Time Actuated Adaptive_l Adaptive_2

Time
Average S.D. Average S.D. Average S.D. Average S.D.
0:05 392 3.29 364 3.11 359 3.13 362 2.60
0:10 389 3.37 368 3.62 364 3.43 363 3.12
0:15 396 2.88 376 3.59 370 3.25 372 2.63
0:20 399 2.37 388 2.91 384 2.81 387 2.70
0:25 406 3.54 397 3.85 393 3.56 395 3.40
0:30 409 3.74 406 3.43 406 3.12 408 3.51
0:35 41 1 3.63 407 3.05 404 2.97 407 3.09
0:40 405 3.41 405 3.24 402 2.62 400 3.40
0:45 398 2.95 390 3.65 393 3.52 389 3.83
0:50 397 3.12 387 3.62 379 3.14 382 3.30
0:55 395 4.06 379 2.57 373 3.00 376 3.52
1:00 392 3.05 364 3.84 362 2.99 360 2.83

 

 

 

 

 

 

 

 

 

* Average corridor travel time in seconds/vehicle

Table 5.2 Savings" in corridor travel time under different control strategies

(moderate peak-hour demand)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Actuated Adaptive_l Adaptive_2

Time
Savings Percent Savings Percent Savings Percent
0:05 28.15" 7.17 33.15" 8.45 30.26“ 7.71
0:10 20.61 ** 5.30 24.50" 6.30 25.48" 6.55
0:15 20.40" 5.15 26.41 ** 6.66 24.07" 6.07
0:20 10.77” 2.70 15.46" 3.87 11.79" 2.95
0:25 9.24“ 2.27 13.00" 3.20 10.99” 2.71
0:30 3.20“ 0.78 2.78 0.68 0.69 0.17
0:35 4.11" 1.00 6.90“ 1.68 4.00” 0.97
0:40 0.00 0.00 2.20 0.54 4.33" 1.07
0:45 7.39M 1.86 4.26M 1.07 8.09" 2.04
0:50 9.54" 2.41 17.31 ** 4.37 14.43" 3.64
0:55 16.55“ 4.19 21.99" 5.57 19.42" 4.92
1:00 27.33“ 6.98 29.36M 7.50 31.90“ 8.15
Average 13.11" 3.28 16.44M 4.12 15.45" 3.87

 

 

 

 

 

 

 

 

* savings over fixed-time signal control (seconds)

** the difference between the two means is statistically significant at a = 0.05

75

 

Figure 5.2 and Tables 5.3 and 5.4 present the average corridor travel time under four
different control strategies for the high peak—hour demand case. The overall savings in
corridor travel time were 1.11% under actuated-signal control, 1.23% under
adaptive_l control, and 1.43% under adaptive_2 control. Similar to the moderate
peak-hour demand, the savings in travel time were higher during the low demand
period in the first and last 15-minute periods, and decreased as traffic reached its peak
lS-minute period. The average savings in travel time during the first non-peak periods
were 1.11%, 1.63%, and 2.2% for the three control strategies, respectively. These
values were 0.59%, 0.37%, and 0.64%, respectively, during the peak 15-minute
period. The differences in average corridor travel time among the three control

strategies were not significant during either the moderate or high demand cases.

5.1.2 Total travel time

Total travel time is the sum of the travel time of all individual vehicles
completing their trips throughout the corridor. This measure considers the number of
motorists, distance of travel, speed, and delay associated with the trips. Figures 5.3 and
5.4 show the total travel time under different control strategies for the moderate and
high demand cases, respectively. Similar to the observations made from the average
travel time graphs, all three control strategies showed a decrease in total travel time
over fixed-time control. The decrease, however, was not significant in the high peak-

hour demand case.

76

Abs-#5.: amt: 83895.». 35:8 in”? «555—. .52.: 0:5 595.: .52.:3 99:95. N..». 95»:—

2:;

mo; co; $5 85 35 9.5 mmno 090 35 cmno who 25 mono oono
. . _ . . . . . . . . 8m

 

...... ......... ........ .............. ............ ............. .............. m ........ ........ .............. a.
1 ......... ............. ............ .......... ...... ............. ............ ............. ......... .......... m ............ .omm

,,,,, ., .1 : ,, . ,. .. . E

 

 

. ~|o>u3u<lxl
_-o>_a%<lal

 

 

.............. .. M . . is.

 

3333‘ III
25.. 35..— [9|

 

 

............ ; madnutménsgnnofigzgmcoss.in$
. V u v n 36qu so: 82 n oats 8.53%.:

 

 

 

 

 

con

77

(spuooas) our!) lama; Jopguoo afinJaAv

Table 5.3 Average corridor travel time" under different control strategies

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(high peak-hour demand)
Fixed-Time Actuated Adaptive_l Adaptive_2

Time

Average S.D. Average S.D. Average S.D. Average S.D.
0:05 412 2.53 404 2.38 402 2.39 398 1.96
0:10 415 2.59 412 2.80 408 2.63 408 2.38
0:15 416 2.19 414 2.77 413 2.49 410 1.98
0:20 419 1.77 415 2.21 410 2.13 412 2.05
0:25 423 2.72 418 2.99 420 2.74 413 2.61
0:30 426 2.89 426 2.64 423 2.38 426 2.70
0:35 428 2.80 425 2.33 428 2.27 424 2.36
0:40 421 2.62 414 2.48 419 1.97 420 2.61
0:45 414 2.25 409 2.82 409 2.72 411 2.96
0:50 413 2.39 404 2.79 407 2.40 406 2.53
0:55 411 3.16 407 1.94 408 2.29 408 2.71
1 :00 407 2.33 399 2.97 396 2.28 398 2.15

 

 

 

 

 

 

 

 

 

* Average corridor travel time in seconds/vehicle

Table 5.4 Savings" in corridor travel time under different control strategies

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(high peak-hour demand )
Actuated Adaptive_l Adaptive_2
Time
Savings Percent Savings Percent Savings Percent
0:05 8.45" 2.05 10.21" 2.48 13.80“ 3.35
0:10 3.00“ 0.72 6.77" 1.63 7.44" 1.79
0:15 2.16 0.52 3.34 0.80 6.34" 1.52
0:20 4.10 0.98 9.47" 2.26 7.32“ 1.74
0:25 4.24 1.00 2.20 0.52 9.47” 2.24
0:30 -0.91 -0.21 2.89 0.68 -0.23 -0.05
0:35 2.99 0.70 -0.33 -0.08 4.16 0.97
0:40 6.71" 1.59 2.29 0.54 1.39 0.33
0:45 4.43 1.07 4.43M 1.07 2.60 0.63
0:50 8.60“ 2.09 5.41” 1.31 6.58M 1.59
0:55 3.74" 0.91 3.07" 0.75 3.07” 0.75
1 :00 7.98" 1.96 11.85M 2.91 9.85" 2.42
Average 4.63“ 1.11 5.13" 1.23 5.98" 1.43

 

 

 

 

 

 

 

*Savings over fixed-time signal control (seconds)
** The difference between the two means is statistically significant at a = 0.05

78

 

3".

A83 32.....2: 8.20.3.5
3389.: .9555 flaw.» 39:5... 5...... on... .93... .55. m.m charm

 

93:.
cc”. mné 35 9.5 2.5 mmuc 35 35 an... w. Ho 2 Ho none .55
A1 r . . . . w. r _ \r can
,oow
.ooa

 

 

 

~.o>_.§<|xl
123.31.?
8382+
as. use... IT

 

 

 

littzloulvuu..¢-nn..rvu.lrrununt.Iuialiuufnllllunlllllole llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll

l

 

8.3:... £9. 8.. u 0E8 3393......
. 8.3.... .5... 8: n “Ems .8310}...

 

 

 

. coc—

. co:

- oom—

- cem.

-. . cov—

 

 

8w.

(sainnim'qu) amp [amen [3101,

79

€3.75... AH...
8.32....» .9589 1...»... 2.9.0.5.. 3...... 2.... .92.... .86... v.m 9.5....

2......

no”. 2.”. mm... onno 3.5 3.5 mmno cm... mm... 85 m . He 3 Ho none 85

LP r p -

 

............................................................................................................................................................................................

............................................................................................................................................................................................

......................................................................................................................................................................................

.........................................

 

 

 

 

. - «13:3... le
.1352... LT
8332 + _ _ _ H _ n U u n "
25. BE 16.. 3.3:“. 53. 8... u 6:5. 8338.5

 

 

 

 

 

............ .............. .............. ............. 8.15....) Seuoggaéws

 

c2.

. mm»

. cma

 

-- . one.

 

 

(scannim'w) my PM”) mo;

80

Tables 5.5 and 5.6 show the total travel time under different control strategies
for the moderate and high demand cases, respectively. The tables list the average total
travel time for the full one-hour period and for the peak and non-peak 15-minute
periods. For the moderate peak-hour demand case, the overall decrease in total travel
time was 3.57% under actuated control, 4.46% under adaptive_l control, and 4.05%
under adaptive_2 control. These values were 0.91%, 1.30%, and 1.17%, respectively
during the peak 15-minute period. For the high peak-hour demand case, the decrease
in total travel time averaged between 0.38% to 2.18% under the three control strategies
with no significant difference between peak and non-peak 15-minute periods. Again,
the difference in total travel time among the three control strategies was not significant

for either the high or the moderate demand cases.

5.1.3 Intersection Delay

Corridor travel time is a measure of the overall system performance for the
main traffic along the corridor. It does not, however, address the impacts of the control
strategies on the minor street traffic at the intersections along the corridor. An
intersection delay analysis was necessary to examine the changes in approach delay
under different control strategies. The term approach delay, as used in this study, is
defined as the length of the time that vehicles approaching the intersection from a
particular direction are delayed due to a red signal and/or stopped queue which

prohibits their free flow travel through the intersection. Intersection delay parameters

81

.88.. 82.0.8 0... w......... 0.....0... .8 .5880 .000. 0... 5.3 .88.. 02.0.8-2 0... t
.0580 Raw... 082-00.... .08 0:... .0>0... 82.80 .08. ... mm...>0m...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

w..m Ram .63. 8.. .ms. mmNco. 0.... cm... Ede... $.03. 1.90098... 08...... m.
NV... 3% 0.32 mm... 2.... 2.03. on... 8.0 madam. vodwm. 0.00... 008...... m .
an. 5.6. 8.3.. o... N... wmdt. 0... SM. Wm... .00.... 98.. 0.8.
2.020.. u.....w...>0m 0w0..0>< 8080.. *mw...>0m 0w0..0>< 2.020.. .0830” 0w0.0><
0.... ..1 0w0..0><
.1303... .1305... 02.33.. 02....
9.880.. ...8.....00.. ..w.... 0.0.8.0.... 8.280 2.0.0.2. 8...... 80.50.810.030... 08.. .02.... 8......80 .02.... 9m 030...
.822. 82.108 0... w......... 050.. .8 .5880 .000. 0... 5.3 8.8.. 02.0.8-2 0... 2..
.2280 Raw... 002.100.... .08 0:... .0>0.. 82.80 .08. a. 35.0%..
2.6 vwdm mwfiom mw.m mmdm mmdoa 3... No.3 $.30 wcdg 10.02.18... 08...... m.
h... an... 3.03. cm. wad. mmdwm. .0... mm... Edvm. 3.3. 0.00... 008...... m.
3..V bade 00mg. 9.... No.3. 3.6... mm... 3.0m can... 3.». .. 9.8.. 0.8.
2.080.. *mw...>0m 0w0..0>< 2.080.. *mw...>0m 0w0.0>< 2.080.. *mw...>0m 0m0.0><
0...... 0m0..0><
.1303... .1305... 8030.. we...

 

 

 

 

 

 

92880.. ...8..1..00.. 0.0.0.8.... 8302...... 8.280 2.0.0.2... 8...... 802.580.02.05 0:... .02.... 8......80 .02.... Wm 02......

82

for an intermediate intersection along the corridor (intersection 3) are presented in this
section. The average delay per approach output are the delay for the northbound traffic
(peak direction for the major-street traffic) and for the westbound traffic (peak
direction for the minor-street traffic). Average intersection delay, however, is the

weighted-average of the delay in all four approaches.

5.1.3.1 Average approach delay for major-street traffic

Figures 5.5 and 5.6 show the average approach delay for the maj or-street traffic
under different control strategies for both the moderate and high demand cases,
respectively. Similar to average and total corridor travel time, average approach delay
for the major-street traffic improved under the three control strategies. The reduction
in the maj or-street approach delay explains the reduction reported in both average and
total corridor travel time. Again, the reduction in intersection delay is higher in the

moderate peak-hour demand case than in the high peak-hour demand case.

Tables 5.7 and 5.8 present the average approach delay data for major-street
traffic under different control strategies. The tables present the average delay over the
full one-hour analysis period, the peak lS-minute period, and the non-peak 15-minute
period. In the moderate peak hour demand case, the approach delay for major-street
traffic decreased by 12.78% under actuated signal control, 16.17% under adaptive_l

control and 15.16% under adaptive_2 control. These values were 0.34%, 1.77%, and

83

.0000 ....0...0.. 82.10.02. 0.08.88.00.38...
8......0 80.05.. 8...... 050.. .00..0..0_.0... 8.. .20.. ..00.......0 0w0.0>< Wm 0...»...

0a...

mo“. con. 3... Om... 9.6 0...... mm... omno mm... om... m... c... we... 85

 

. . . w . . .llL l. . IIL I. L

 

 

............ ~|0>0§<le:23:...,m.--...,:.--.m -
—|o>.u§< N H M n n H H H n .
00.0ao<+

 

 

 

 

 

. - . . .
_ . . _ . _ ,
_ . . . _ . _ .
. _ _ . _ _ _
_ . _ . _ . N ,
. . . o _ . _ ,
u _ . . . . —
. . . . _ . - ,
, . . _ _ . _ _ ._
. . . . . _ . , n
_ . . . . . . . .
. . . t u _ . A
_ . . _ _ _ _ _
. . . . . _ . .
. . . _ . _ . _
. _ . . . _ . .
............. ..---..--........-....-.---.u.--..---...----.--., --..--.r.--..-....-..-.-.,.-.. -..,w-..,-.. .-..-r...--...-.. ,...,--=.----.-....--.----..-..------y---.-..-.--...-...,rl
. . u . . . . .
~ . _ . . _ . .
. . _ . . . . i
- - _ . . . - t
I . . . . . . .
. . . , u . ‘I
. . . . . ‘11
l‘ - . . . .
. . . .
/< . _ . . \
. _ _ .
.lll'lllll.~ I, . . . a . _
. I . . . .
. I I . . \ .
</ . . 4 _
l n .
..................................................... 1%,]- uqnv.-nu.ou.n:vn-.-.|u-.-”.-uu. \ tun-o--- nu-..------. o--nv»..|v-..--...l...t........v-...vr..yl
£1

. . . . . . . . _ .
. . , v . . u _ . .
............ on... x...+
1 A _ . . . A . . . _ .
n . _ . _ . . . — - . .

 

o.

om

em

ow

(apgqu/saos) Kalap qanoaddv

84

.0000 ....0...0.. .02....02.
5.... 0,030.0... .95....0 2.0.0.5.. .0...... 050.. 80.3.3.3... .0. ~.0.0.. ..000.....0 0u0.0>< 9m 0...»...

on...

no“. co”. mmno omno 3.5 005 mmuo omno mmno omno 25 ofio mono oono

. L _ . . .. L _ L . _ . .llullfi o

 

 

 

 

(apgqu/soas) Kalap qanmddv

 

 

 

 

............. u . ov
a o>_a€<l.T
_|o>_a%<|al
8033+

.......... Uguvofiml.‘ ('.L.T om

 

 

85

0000.. 800-000 0... w0...... 050.. .0 0800.0 .000. 0... a...» 00.8.. 0.00.8-0. 0... t.
.0.0...0> \ 0000. .8800 .00w.0 08.0.0000 .26 00.0.. 0000.000 0. 0w0.>0m*

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8.0 00.. 00.5 N00 mm. 00.3 006 .0... 00.3 00.NN 10.009000. 0.00.8 m .
m... 0m... 00.0w 00... mm... m0.0~ Rd 00... 00$ mmém 0.00... 0080.8 m .
:0 o... a... 00... 8.. 000. 00.0 8.0 0.... 8.00 08.. 25.
.0080. 0000300 0w80>< 8080.. 00000.00 0w000>< .0080. u..0w...>0.m 0w0..0><
08.... 0w0.0><
0.300%... 33%.... 8033. 020”.
90080.. .00.....00.. 5.... 0080.080 .0....00 .00.0....... .0...... 050.. .00..0..0_.08 .0. 00.0.. ..000.....0 0w0.0>< 0.0 0.00...
0080.. 800-000 0... w0..8.. 0.....00. .0 .8080 .000. 0... 5.3 0080.. 0.00.8-0. 0... t.
.0.0...0> \ 0000. .8800 .0030 080-00.... .0>0 00.0.. 00088.0 0. 0w0.>0m_..
R... 00.0 :0. 00.00 00.0 00... 03.0 .0... 00.... 9...: £380-85 3:50 0.
xx... 0.... 00.3 R. mm... 00.8 0m... no.0 0...~ mm..m 9.00... 00.8.8 0.
0.0. 00.~ 0.0. 2.0. w... .00. and. mm.N 0.... 00.0. .80.. 000.
8080.. n..0w0.>0m 0w0.0>< 8080.. u..0.w...>0m 0w0.0>< 8080.. 0000300 0m0.0><
0.0.... 0w000><
012.0003. .0800... 02023. 80...

 

 

 

 

 

 

$8080.. 88.0.02. 0.0.0008. 00.w0.0..0 .8800 808...... .0...... 0....0... .00..0..0_.08 .0. 00.0.. 0000...... 000.03.. 0.0 0.00...

86

0.88%, respectively during the peak 15-minute period. During the non-peak 15

minutes period, the reduction in intersection delay was as high as 30.30%.

In the high peak-hour demand case, the average reduction in intersection delay
was 3.93% under actuated signal control, 4.39% under adaptive_l control and 5.11%
under adaptive_2 control. The average increase during the non-peak 15-minute period
was as high as 8.03%, and during the peak 15-minute period was as low as 1.30%. The
difference in approach delay for maj or-street traffic among the three control strategies

was not significant.

5.1.3.2 Average approach delay for minor-street traffic

Figures 5.7 and 5.8 show the average approach delay for minor-street traffic
under different control strategies for the moderate and the high demand cases,
respectively. The results are also summarized in Tables 5.9 and 5.10. Average
approach delay for the minor-street traffic increased under the three control strategies.
The average increase in delay was 6.3% under actuated signal control, 7.15% under
adaptive_l control, and 7.79% under adaptive_2 control in the moderate demand case.

These averages were 2.06%, 2.95%, and 4.22%, respectively, in the high demand case.

87

.0000 00080.. .08....00. 80.038.880.080
.0..000 80.00.... .0000 050.. .00..0..00.8 .0. .20.. 0000....0 03.0.5. 0.m 0.0m...

2....

no“. 0.... mm... cm... 0.0... ow... mm... omno 0N... cm... 0.... 0.... 00... co...

 

 

 

 

 

 

 

. _. . . + . __ . + . . 0. o

Nuo>0§<|xl , . _ , , _ _ _ , .

. ....... .no>0§<+ .............. .............. ........... ............. ........... ............ ............. - o.
vounau<+ . , _ _ . , _ _ . .

000800+

......................................................................................................................................................................................... I ON

.................................................................................................................................................................................... T OM

,- 00

0 cm

 

 

 

o0

(alagqu/saas) Kalap qauoaddv

88

.0000 00080.. 80.....00.
00.... 0080.080 .0.800 80.0...0 .0000 050.. .00..0...00.8 .0. 00.0.. 00008.0 000.03. 0.0 0.00....

2....

mo”. co”. mm... cm... 0.0... ow... mm... cm... mm... cm... 0.... 0.5 00... co...

 

 

 

 

 

 

 

.......................................................................................................................................................................................

 

 

 

 

.l T L. i 0 0 0 0 L L. 0 .
0.2.00...le _ _ . n u t H h H ‘
.8381: m m M M W W m H. m M

B.§.<+ ........ ......... .............. ........ , ..... ........... .
000800+ _ .

l

o

O
—t

O
N

o o
0- m
(mama’s/sass) £3199 qanmddv

O
V3

00

89

02.00 500.000 00. 0.0.00 0000.. 00 .00080 .000. 00. 00.3 02.00 30000-2 00. 2..
6.0.00.0 \ 0003 .0008 .0006 0000-005. .26 00.00 0000.000 0. 0000000..

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

and- 8..- 000m 000- 00. .- ~0.Nm cm..- 00.0- 000m 000m 10.000000. 0.0000 m.
~00. 0.0- 00.0w 0.0- m..- 00.mm mmN- 0m..- no.0m .0.Nm 6.000. 00000.0. m.
00.0- 0 . .N- 00.mm 3.0- mm. .- mm.mm 00.0- no. .- 00.0w 0... . m 0.000 000.
000.00 0000.00 000.05.~ 0000.00 0000.00 000.000~ 0000.00 0000.00 000.0><
0.0.... 000.0><
0.20.03 330.00 30300 020.
€000.00 .000-0000 00.0. 8.00.0.0. .0..000 30.0.0.0 .0000 0.000.. 30.3.3000 .0.. 00.00 0000.000 000.0>< 0..m 0.00H
00000 0.00-000 05 0.0.00 0000.0 .0 0.0080 .000. 00. 00.3 00000 0.00.8-2 0.0 3.
00.0.0? \ 0000. .0008 000.0 0.00-0000.. .0>0 00.00 0000.000 0. 000300...
$0- 0:- 3.3 8..- 3.0- 2.2. 8.0- 0.0- .0.:- fié 1.000985 22.0: 2
00.0.- 00..- 00.00 00.0- 80- 00.00 00.0- 00... 00.0w 3.00 0.000. 00.0000 m.
00.0- 00.».- 0m.00 m .6- 0mm- 00.00 3:0- 0.00- $.00 0060 0.000 000.
.000.00 00,000.00 000.0>< 0000.00 0000.00 000.0>< 0000.00 00000.00 000.0><
000... 000.0><
0.300000. 30.02 8032 B0.

 

 

 

 

 

 

00000.00 .000-0000 0.0.0000: 8.00.0.3 .05000 30.0.0.0 .0000 0000.. 000.00.30.00 .0.. 00.00 0000.000 000.0>< 06 0.00...

90

As both adaptive control strategies used a shorter average cycle length (100
seconds) during the non-peak periods, the increase in minor street delay was lower
than that under actuated control, which was forced to use a cycle length of 120
seconds throughout the analysis period. The three control strategies, however,

achieved the same reduction in major street approach delay.

5.1.3.3 Average intersection approach delay

The average approach delay for the intersection is the weighted average of the
delay on all approaches. It takes into consideration the average delay per approach as
well as traffic volumes for each approach. Thus, it is an indictor of the overall benefits
of alternative signal control strategies at an intersection. Figures 5.9 and 5.10 present
the average intersection approach delay under different control strategies. Tables 5.11
and 5.12 present the average over the full one-hour analysis period, the peak 15-
minute period, and the non-peak 15-minute period. Overall, the intersection delay
decreased by 6.4% under actuated signal control, by 8.36% under adaptive_l control,
and by 7.42% under adaptive_2 control in the moderate demand case. The delay
decreased by 2.11%, 2.14% and 2.25% for the three control strategies, respectively, in
the high demand case. Similar to all other parameters, the reduction in delay is higher
during the non-peak 15-minute period, and decreases as traffic volumes reach their 15-

minute peak period.

91

A83 can—:2. .52—
.:u«.— 83938703895» 35:3 25.3.51 .36.:— bzou £283.? 3:89.35 owfio>< ad 2sz
25,—.

mo; co; mmuo cmuo. mvno ovuo mmuo omno mmno omno 25 ofio mono 85

ll V if if _ if .l V F V F r F + o.o
. . . V

l 4 l l

_ . v V —
_ V . _ _ _ , . . a V _
c 4 . ¢ _ _ A — V . V _
. V _ . q _ a a V _ — —
V . V V _ _ V n V _ V V

...... ....L-..'.‘.v-.-l...-...-..-.-xl.....-...‘.-...V.........--.-L.......y...i.L-...-. l..-..V'..1......-..rl.-i...-....-.l..-...1...-.V..--.l..---.-r-..--....-.I.L.--.-..‘..-y.¢l Oco~

. . _ . _ _ _ . . _ . c
. , _ _ _ _ . . . _ . .
. , . _ . _ _ . . _ . .
. . . _ _ . . . . . . ,
. . . _ . _ _ . . . _
. . . . _ . _ , _ . _

odm

 

odm

 

 

 

 

 

 

......... mnozaé<lxl T odv
_no>u%c<lcl m m m m m m m m m n
Email-I _ , . _ . . , . . _

......... 5......1. s.

 

 

 

. . _
. _ _ . . . . . . . .
. _ . . _ . < . V V V
_ A . _ _ - - . _ _ .
. . . _ . _ . . . . .
> b if L . t i » >l > L Once

(alagqu/saas) Knapp qanoaddv

92

mo

V

328 “Ea-=2. .52—

.:3a :35 8389:... 35:3 25.55.. .365. 530—. 5.3.5.? 5:39.85 03.33.. 3% Ping

”_ co;

V

mmno own

25,—.

o memo ovuo mmuo ammo mmno omuo mfio ofio mono oouo

 

_

 

11

L. V

 

nnnnnnnnnnnnnn

llllllllllllllll

 

mnoéaé... le
Egan? Ial

3332 III
2:: 85 ll

 

 

F L F V a r a V V

 

 

 

 

o6

- 93

q
o
N

odv

C.
o
M
(apgqu/soas) Kalap qauoaddv

odm

ode

93

$2.5.» Son-28 05 wage cash mo 226:5 ammo— ofi :23 @023 8558-2 55 2.
A2929, \ 88V 35:8 32me oat-H55.“ 5>o tow—nu 585% E $5.3m...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mcé hm; mmdm 26 $6 8.0m SUN 36 $.om mmSN 138952V 52:28 m _
wmd- wed- wwdm vmd 36 ~53 3d Rd mmdm Eda £33 52.58 m V
mmN mod mmsm EN cod mmSN SN and omsm Eda A25: 5:8
385m *mwfiém owm5>< 552$ kwfiém own5>< 2505A 533% $853..
22C. owm5><
N-o>_a%< WEE“? @252. BE
€52.56 Son-v.3.— AMEV 535225 25225 255E“. 5.2:. .356 .5325: 2.355952: 53.5.5. afim 53:.
@023 Boa-28 55 mega 0552 we 2:528 ammo— 55 :23 v3.5a 52228-2 55 .3
$523 \ moomv 35:8 32me 582686 .5>o >306 58.5% E meme...
3.3 ohm 3.3 ct: ooé mm“: 3.2 QM 5.3 wodm 2.339285 3:58 m _
wmd- m flo- Nvém 3.0- mod- mmdm 3d- Ed- mvdm Rom 0:53 5558 2
N: ow; Nmfid 3.x mod aNNN ovd cm; ENN mmém 98: 28V
:50qu amwarém owa5>< «cocoon *meEm oma5>< 2595A .3333 $853..
58:. owe5><
m-uéaé... alozaéac 2.232... 85

 

 

 

 

 

 

€52.56 uses-5:52. 895—225 8%525» 252.3 255.56 5.2:. .5256 £28.39. 2353.52: owa5>< —~.m 53.2.

94

5.2 Signal timing analysis

There are three possible ways an adaptive signal can contribute to a reduction
in delay for the major street traffic. The first is to allocate more green time to this
approach; the second is to utilize the existing green time more efficiently, (i.e. with
less lost time); and the third is to improve the corridor progression, and thus reduce the
stopped time component of delay, Figure 5.11. The total effective green time allocated
to each approach, aggregated over 5-minute periods, under different control strategies
was compared. The effective green time for each phase represents the combined green
and amber periods minus the lost time at the beginning and end of that phase. The lost

time was assumed to be three and half seconds per cycle.

The results of the green time comparison for both maj or-street and minor-street
traffic are presented in Figures 5.12 and 5. 13 for the moderate demand case, and
Figures 5.14 and 5.15 for the high demand case. The results showed that all three
signal controls, actuated signals, adaptive_l control, and adaptive_2 control, allocated
additional green time to the maj or-street traffic while reducing the green time for the
minor-street traffic. The green time added to the maj or-street traffic was higher during
the 15-minute non-peak periods and decreased as traffic volumes approached their 15-
minute peak. During the 15-minute non-peak period in the moderate demand case, the
green time allocated for the major street traffic increased from 10.88 minutes under

fixed time control to an average of 11.80 minutes under actuated control, 12.16

95

minutes under adaptive_l control, and 11.64 minutes under adaptive_2 control.
During the peak 15-minute period, all three signal control strategies assigned
approximately the same green time allocated to the approach under fixed time signals.
In the high demand case, the green time allocated for major street traffic, except during
the first and last 10 where it was slightly higher, was equal to that assigned to the
approach under fixed-time signal control. The analysis showed no significant
differences in the green time allocated for each approach among the three control
strategies, which explains the similarity in travel time and intersection delay reduction

achieved by the three control strategies.

To measure how effectively the green time allocation was utilized the degree
of saturation, aggregated over 5-minute periods was compared under different control
strategies. The ideal saturation flow was assumed to be 1800 vehicles per hour. The
results showed that, in general, when the three control strategies extended the green
time for the major-street traffic, they reduced the average degree of saturation,
reducing approach delay and hence corridor travel time. Consequently, when they
reduced the green time for the minor approach, the degree of saturation increased,
increasing the delay for this approach. The reduction/increase in intersection delay for
each approach was a function of the amount of increase/decrease in the green time
allocated for the approach. The results are presented in Figures 5.16, 5.17, 5.18, and

5.19.

96

TIME
-22:

 

 

 

 

 

 

 

DISTANCE
- I I
E I I
I I
I I
' CASEII
MS I I /
Mag-mm - . nmmoonassron
' '-11'-
/ L—
I
DISTANCE

 

a I CASE!!!

{m- M

 

 

Figure 5.11 Possible delay reduction by an adaptive control system

97

A33 6533. sacs-:3.— 89.0025
95:... gonna-no.3:— 58 2:: 50..» ofifieflfl «Hm 95w:—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

25...
35 one 35 £5 £5 as :5 85 85
«Sensual
Wages-T
maniacal-I fl
guest-T
o 9 [Ill-1A i 6 6
% Iilj 1 V

 

 

 

 

 

 

 

 

 

 

 

 

md

m.—

98

Wm

Wm

(samurai) [BMQJlly ainugm g 13:! amp [133.13 aApaayC-I

mé

339 2:25.. 33.23.. 3.2325
95.5 «cabana—:5:— ..£ 2:: :99.» 253mm mfim 0.5»:—

 

 

 

 

 

 

 

2::
3; 35 c2 95 on”: ”Na 85 35 85 85
, L L P . . o
-3
, _
I 2
............................................................................................................................................................................... r N
a ................................................................................................................................................................. ...-% W.N
1 ..................................................................................... I m
-ppuhuhpupni -uu-:- ......................................... ‘ ............................................... . ................................... ‘ .................... r mom
N uémauwlxl _ _ M U . _ _
...... w§§+ V
33:31... N H H w H H ,
...... wfiuuwuwhlél , m6

 

 

 

(seminal) [mum]! amuym 9 43d am}; {133.13 aAyaayg

99

A33 ens—5.. SEE—«on :35
353.5 .3.5..8?E ..8 2:: 38..» 253.5— vfim 9:.me

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2:;
3; 35 one $5 8o ”3 85 :5 85 85
.1 o
~u§m§wlxl
1 Nuufimuawnf no
1 ~§§BV+ _
3:33.331?
2
_ N
3
m
3
Willi 34 I I II‘MHfl v
3

 

 

 

 

 

 

 

 

 

 

 

(83311111111) [3.1133111 31n111111 9 .13d 3111;; 1133.13 3111133333

100

$28 6:25.. SSE—a2. gum—c
353.: Bahamas—:5 ..8 0:5 :09.» 95253 mfim 3.53..—

 

 

 

 

 

 

25-.-
3; 35 one $5 35 «Na 35 35 86 85
- _ L _ _ o
- no
u - _
................................................................................ . m4
............................................................................. . N
.................................................................................................... 11.---..- 1 m.N
. .......................................................................................................................................................................... r M
f ...... N'Néguuw ................................................................................................................................................. v W.M
3 3:35.31?
..................................................................... .--------- .-..----- 1 V
nuguuVIIl . u U _ U H .
...... 2:33.331? 3.

 

 

 

(s31n111111) [3.1133111 3111111111 5 .131] 3111;; 1133.18 3.1933ny

101

A323 6525—. 32—3—33 83.8.5.5
can... Submasmafi be .3323: a: 352— 3% «Sum...—

2::-

vo: $6 35 9.5 omuo ”N5 35 3 He 85 85

 

 

33333:... n H 1 a H n M
Wuémuuwlqlsw ...... w ............. .................... m ......... , .......... ............... .................... ...................

 

 

 

...... 1:333...-
/

................... .:,- /I.

........................................................ -./.

1111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111

 

 

 

vd

ed

5.0

#4

«4

1109mm“ go 33.1330

102

A83 .5382. 33.3—32— oufioecfiv
can... Humbug—35:. .8.“ 25353. no 8..qu :6 9:.me
25,—.

3; $5 omuo 9.5 :5 «Na 35 35 85 85

 

4‘ 4 1 J 4 A]

 

 

 

 

«Jaws—xix:
WESVLI
...... muse-inn:
3:33.79:
‘ l ,l‘n-Y’}, \ .‘lhc‘- -.
........................................................................................ \n V’/.\\\\“ 1\ .

 

. Nd

“I
o

‘0.
o

“2
o

 

 

NA

nopunms JO 33.133q

103

A83 6535—. SSE—won :95
953.: «3.5-3738 ..8 55.253. no 3..qu afim 3.53m

 

 

 

93E.

x: as one 9.5 one ”do :5 :5 B o 85
NISSESgVIXI .................. -------------- .................. ..................... - ................ -
~I§~E§VL¢I H m m m m m

maaauvxlll ----------- - ------- ------------------ ------------------ -------------------------------------- .
aéEhmbm , . 7 _

 

 

 

 

 

 

Yo

We

06

50

NA

uopumns 10 33.1330

104

vou—

Aomau 5:352. 33.3—32— gas
953.: «3.5.35:— ..£ 533.53. mo 8..on afim ouswi

 

 

  

 

25,—.
a o 2 o a. a one an o a o z o 85 85
, o
N33233:; m 7
,,,,, Niuéfiunvxlqlf- - .md
..........+. h H N W
,,,,,,,, Wfiéld - - é
.................................................................................................................... v coo
------ . wd
- #

 

 

N;

uopmnws 10 33.133q

105

During the 15-minute non-peak period in the moderate demand case, the
degree of saturation for the major-street trafiic decreased from an average of 0.81
under fixed time control to an average of 0.77 under the three other control strategies.
This reduction was a result of increasing the green time for the major-street traffic.
Simultaneously, the degree of saturation for the minor-street traffic increased from an
average of 0.77 to an average 0.95 as a result of reducing the green time for this
approach. During the peak lS-minute period the three control strategies as well as the
fixed time control maintained the degree of saturation for the two approaches at

0.99%.

5.3 Sensitivity analysis

This part of the analysis explores the sensitivity of the benefits of the adaptive
control strategies to changes in demand levels and peak hour factors for minor-street
and major-street traffic. The benefit, in this part of the study, is defined as the percent
reduction in average corridor travel time achieved by the adaptive control strategy.
Due to the similarity in the performance of both adaptive control strategies examined
in this study, the sensitivity analysis was carried out only for the adaptive_l control

strategy.

106

5.3.1 Sensitivity of adaptive control benefits to minor-street demand level

The sensitivity of adaptive control benefits was first examined under different
peak-hour factors (PHF) for the minor-street traffic. The traffic volume and the PHF
for the major-street traffic were kept constant. The traffic volume for the minor-street
traffic was also kept constant with four different values of PHF: 0.85, 0.90, 95, and
0.99. Five simulation iterations were performed for each PHF. The results of the
analysis are presented in Figure 5.20. As can be seen, the percent reduction in average
corridor travel time is sensitive to the PHF for the minor street traffic. That is, as the
PHF decreases the benefit increases. The percent reduction in travel time ranged from
13% to 7.8% when the PHF was 0.85, and from -l.2% to 1.44% when the PHF was
0.99.

Lower values of PHF mean that the ratio of the volume during the non-peak
periods to the volume during the 15-minute peak period is lower. Because the green
time allocated to this approach, under fixed-time control is calculated based on the
peak lS-minute traffic, the approach is served with an excess amount of green time
during periods other than the peak 15-minute. The adaptive control system utilized this
extra-green time by reallocating it to the major-street traffic. As the PHF decreases, the
excess green time allocated for the minor-street traffic increases, providing the
adaptive control logic with more potential benefits. Similarly, during the peak 15-
minute period, there is little or no excess green time to be reallocated, eliminating the

potential benefits of the adaptive control system.

107

ELEV ..efiah .53..— ..aom 39......33... 8 5352.8 52.3. aflm 9...»...

3.3% Sam—.33.—

 

mwd L.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m 8.0 8.0 3.0 mad md mad
. . . . _ . _ P F P
C . l '
a ' ' ' C '
l l l l .
mam-.32... 1 o. 1 d J . I -- .
0% Ce 0 o O
E. E: o I ll o
m... E: o I . w
an E: o o
I. E: o . .
:9. 8.. u uEE. 82.3.35...

 

8..." "in .5... 8: u 0E»: 82.933:

 

 

 

or

N?

(11133134) 311111 plum; Jopguog) 111 S3111Ans

108

To examine the sensitivity of the benefits to minor-street traffic demand, the
PHF was kept constant for four different demand levels: 150, 200, 250, and 300
vehicles per hour. Figure 5.21 shows the savings in corridor travel time under these
four demand levels using four different PHFs. For the same PHF, the benefit slightly
decreases as the minor-street traffic demand increases. This indicates that the percent
reduction in travel time is more sensitive to changes in PHF than to changes in traffic
demand, at least at these demand levels. Figure 5.22 shows the potential reduction in
corridor travel time under different demand levels and peak hour factors for minor

street traffic.

5.3.2 Sensitivity of adaptive control benefits to major-street demand level

To examine the sensitivity of the benefit of adaptive control systems to
changes in major-street traffic volumes, four different demand levels were examined.
The minor-street demand level and PHF were kept constant. The PHF for the maj or-
street traffic was also kept constant. Five simulation iterations were performed for
each demand level, the results are presented in figure 5.23. No significant changes
were found among the benefits obtained under different demand levels. This indicates
that changes in major-street traffic demand, within the values examined in this study,

had no significant effect on the percent reduction in corridor travel time.

109

own

3...... 2.9.3.56 .8. 6.3.56 353... .033........... e. 5.3.355 52.3.. .N.m 9...»..m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3...». 6.3.56 35......
com 0mm com om. co.
m. ..A ..n .m A o
? Invll
ill l0 l0 ll- N
T lll ll/l v
fl 6
mad u urn. x
mad u “in. IOI
a... u ..Ea + l. - o.
and u “in. IOI
_ 8...» “in .5: 8: u 058.. 893.3%:

 

.N.

 

(11133134) 311111 [3.11111 101111103 111 83111Ans

llO

can... 59.5.55... 5.. a... .2... .95. 6525..
25.55.. 5...... 2:89... .22.... 3.25.5 .... .9223 2.... .52.... ..o.........U .3252... and 9...”...

58......» 59.5.55:

8.. SN 8N om.
mm... . _ ,

 

 
 

 

 

ell-id 333138-10“!W

 

 

on... _o>a...
3......00
E «9.3..» .x.

111

om»... ..

.255... can... .3...?...._.a... .. ...... .95.... ......53 ... $3.»... .. 3.323% 3% 9...»...

Eat 05¢... $2.935...

 

 

 

 

 

 

 

 

 

 

 

 

 

.8. ...... 8N. ...: 8: .... 8.. .8 8.
. . . r _ _ _ . o
N
x
. . m .
Ih‘JIIII.IIIIIIIIIQIIIIIIIII‘ v
4 I 4
O x
.
mmm5>< .0. n
mfica X
82...: x
mfic... 1 m
«.25. I
..%E: 0
.... u ...... 332...:
....n "in. ...... ... u .52. .83....5:

 

 

0..

(°/.) awn |9Aan Jopguoo u! sBugAes

112

5.3.3 Sensitivity of adaptive control benefits to signal phase plan

The objective of this part of the study is to examine whether the reduction in
corridor travel time reported in the earlier parts of the study is limited to the two-phase
signal operation examined in the study or can be generalized for other signal plans. A
four-phase signal plan was considered in this part of the study by adding a protected
left-tum phase for major-street and minor-street traffic. Corridor travel time was
examined under two different signal control strategies: fixed-time and adaptive_l
control. Five simulation iterations were performed for each case, the average corridor

travel time under the two control strategies are shown in Figure 5.24.

The results showed a similar pattern to that obtained under two-phase signal
operation. Similar to the results obtained earlier, the reduction in travel time was
higher during the low demand period in the first and last 15-minute periods, and
decreased as traffic reached its lS-minute peak period. The average reduction in travel
time during the non-peak periods was 4.14%. The reduction was only 0.36% during

the peak-15 minute period.

113

€039.30 :25.» 02.2. -...
.25.» 3.2—...... ...... 2.... we... 5...... as... .55... 32...... ouu.o>< ca... 2.6.".

2::

mo; 00... mmno on... .36 ofio mmuo 090 mm... omuo ...—.6 o _. Uo mono

 

 

 

 

 

. F . . . . . . . . . . con
i . . . . - .
. o . . . . . . . . — ‘
. u . . . . _ . . . .
. . . . . _ . . _ .
_ . . . . . . . . .
_ ~ _ . v . _ . .
. ............ . ....... ......... ..t. ....... . ..... ........... . ...... . ...... ....... z
. A . . u , A , . .
. - . v . _ . , . v
I- ........... . ........ .1... -l ....... . ......... lyr ....................... . .... ... ,.y....w ........... . ....... v.-. ........................... _ ............. m ........ ..nxnil ovm
_ n . . . . . . .
. . . . .
, . . . . . _ .
. , . . . . . .
............. . ............. . ............. -. ............. . ............ . ........ .......... t .. :. ........ ....... I
. _ . . . A . . v _ .
...... t . r . 7 00m
_ _ . , . . . .
_ _ . . . . . . _
. . V . . _ . . _
. . . . . , . - -
. o . . . . _ . .
. - . . . . V _ . .
..... - - _. 2 I 1 00V
_ n _ _ . _ . . _
u . . . _ o
u u . . . . .
. . . . -
_ . . . — . v
. . . . . .
. . . . . . .
_ . w - — _ .
vavllll zlll .............. Jr XI II. I t'llltJlllllll IIIIII 4' IIIIIIIII I1|1..luv llllllll . iiiiiiiiiiii 1"
. a . . . _
_ c _ . . , .
. . . . . , V
. u _ A . .
. u . . . , .
. - _ . .
Inl1llll !!!!! .QtOOI IIIII III‘I lllllllllllllllllllll . lllllllllllll . iiiiiiii ill. ............ r ............ . ........... vl‘
_ . . _ _ _ 03.
o> a ll: . , . . , , . _ ,
............ .......................................... ......................... ........................... , ........ r
m 5| X_n_ , _ , , , . . GOV
O I . . . . . _ _ , .
. I , . , . . _ , .
....... j ............ ,. ............ . ............. .......... ............ . ....... ............ . ........ ........ ............. . .........
_ g . . . . _ . . . . .
. n . . . . _ . . _ . .

 

 

 

com

114

(spuoaos) own plum Jopguoa OBBJOAV

5.3.4 Sensitivity of adaptive control benefits to changes in traffic demand over

time

The objective of this part is to examine the long-term benefits of adaptive
control strategies. For this purpose, it was assumed that the major-street traffic would
grow at an average of 6% annually. The growth rate for the minor-street traffic was set
at 2% annually. The benefits of adaptive signal control were examined afier one and
two years under two assumptions. The first one assumed that the signal plan will
remain with the same setting optimized using the base year traffic demands. The
second case assumed that the signal settings would re-optimized using the new traffic
demands. The results are shown in Figures 5.25 and 5.26 and summarized in Table

5.12.

Table 5.13 Percent reduction of corridor travel time under adaptive control
system (two-year operation)

 

 

 

 

 

 

 

Base-year First-year Second year
Adaptive vs. non-optimized fixed-time 4.16 9.06 26.33
Adaptive vs. optimized fixed time 4.16 3.04 1.57

 

When compared with the fixed-time settings, an optimized setting based on the
base year traffic, the benefit of the adaptive control system increased significantly with
time. The savings in travel time increased from an average of 4.16 % in the base year

to an average of 26.33% after two years of operations. This represents a significant

115

 

...».o... .5... ......0. .....w.» 022...... .... 2.25.. mmfi 9...»...—

 

 

 

08F
no”. co; mm... on... 35 9.6 mm”. on”. muno omuo mfio or. no... oouo
. .y p P p . P , p F p F o
- 06:69.... 3.8.30.5: Lal
0.5.69... um~_E_..oIII co.
o>=amuml¢l
com
com
cow
com
com

 

 

 

116

(puoaoa) cum warm afiuaav

no;

...5.» ...... .52.. .82.... .....w... 032...... .0 2.5.5.. 3% ......w...

0E;

00; mm... on... 9.5 CEO and cone mmuo omuo n F H0 020 no...

 

 

— P p p h P b . > .— p

 

- 06:60.... um~_E_.a0.:0c +

«5.-.... 8.5.810: ‘
9.3m... IOI
ill . I, l . I||I|I.. III-ill , m m

.

 

 

oono

o

co.

com

com

oov

com

com

(puooos) cum ”an; efiuoav

117

benefit of the adaptive control system as it is optimized on a day-to-day basis. When
compared with a fixed-time signal that is optimized based on the new traffic demands,
the benefits of the adaptive control decreased from an average of 4.16% in the base
year to an average of 1.57 % afier two years of operation. This is a result of the

increase in the degree of saturation resulting from the growth in traffic.

5.4 Comparative analysis of adaptive control prediction logic

In the first adaptive control strategies examined in this study (adaptive_l),
traffic predictions were made based on cycle-by-cycle traffic data collected
downstream from the intersection. That is, the prediction of the number of vehicles
expected for any specific cycle was calculated based on traffic volumes in the previous
three cycles as explained in section 3.2.3.2 of this dissertation. Figure 5.27 shows the
predicted and actual number of vehicles for both maj or-street and minor-street traffic
for one-hour period. The comparison showed that the error in the predicted number of
vehicles per cycle ranged from 0 to 3 vehicles for minor street traffic (average number
of arrivals per cycle = 10 vehicles) and from O to 5 vehicles for major-street traffic
(average number of arrivals =40 vehicles). When the prediction horizon was increased

to 5-minute intervals, the accuracy of the predictions increased (figure 5.28).

In the second prediction technique, upstream detectors were used to predict the

number of vehicles expected based on the estimated travel time between the detector

118

location and the intersection. A comparison of the predicted and actual values is
shown in Figure 5.29. The error in the predicted values ranged fiom O to 2 vehicles for
both major-street and minor-street traffic. The error is basically due to the difference
between the stochastic value of the vehicle speeds and the average value used in the
prediction model. It should be noted that the scan-ahead prediction model introduced
in this study represents an ideal case. No mid-block traffic entered or exited in the
approach between the detector location and the intersection. It should be noted that
while the first prediction technique was not as effective in predicting the cycle-by
cycle variation in demand as the second technique, there was no significant difference
in the performance of the two adaptive control systems that used these two prediction

techniques.

119

2.2.5... 5.005.. 0.00.5330...
0.0.3 ..000 ...... 8.02.0.» .... .0.—.0.... 005.00.... ...» ...-.04. 3% 0...»...

0.2.0 .0.. 020...? .0 .0082. 33.0..

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

on 9. 0.. mm on mm on m. o_. m o
_ ..i \‘w o
_ x
_ \
4 \ m
\
o o
o 1 o.
\
\
I \
111 {I h 11 11. m.
_ \
\
\
TI \ Eili- om
v \ \
. .
A m \ \ mm
‘ r \ _
Tire . 1- .8
_ z \ \ fl “
,V \ . 1. .1. mm
, w. u r
A o c ..\ o . , ,
1: O O “(Ade .m p _ L.. O?
0 H0. 0 0 fl
’ ?
O t 1 # L. mV
\0\ c o .
. .

 

 

 

 

 

 

 

 

 

0.050 10:! sepmen J0 Jeqwnu paiogpead

120

€2.58 .8555“. anhm=3o3
5.285 58: Safififi 5.. 8522' .5 5:55: 55:5...— .m> 5:5< wflm «...—Ma

3523 5 52:5: .83

 

 

 

 

on? car om cm on o
F _ W W ‘1 o
m m \
m m \
M m \
w m \
................................................................................. k on
u \ n
m \ _
m \
. . w \ _
m N m\ m
A ............... l ........... ....... ............. \w ................... ............................ v 00
U m \ m m
h m \ _
w \
W \
m\
.................................................. \f.-C--f;..--..,.-..--.,-..--..-:.-. - T om
\
\
\
W \O M W U
.................... o\A our
\9 m m H
O \
..\ A . .
\ m m
L 03

80“)!qu ;o Jeqwnu peaogpmd

121

62—58 Acacia-3.3 5555.. 8.5.53:
293 .55 .8.— 5529» .5 53:5: 55:53 .9» 553. afim charm

226 5.. 552; 5 525:. 535a

 

 

mv 0v mm on 0N ON or or o
w W + . _. w 1. .. L. o
M.
........... ”V OF
N m m m m 0“ m
m m m m m \ m
4 ., H _ ”\
D m m m \m
m m m m \ n
H H H U \
w m m m\
.....1...... ............................. W-..--------..--A...---...---1--.“ ............. \ ......................................................... ON
m m m m \
u H n n \
w m m m\
m m m X
_, W _. \ u
m H m \
H, w V\
W w W x 8
m u m \\ .
m m Om
m M Q
N H O \H
W 3» _
........... V‘ 1 1....... CV
“OW\ H _
O. m _
V H _
x.
1 L - L o...

 

 

 

9.vo Jed sepgqu .30 Jeqwnu peaogpead

122

Chapter 6

FIELD STUDY

In 1991, the Road Commission for Oakland County (RCOC), Michigan,
embarked on a major demonstration project involving the implementation of
Intelligent Transportation Systems (ITS) technology. This technology included an
adaptive traffic control system, SCATS, implemented on the road network within
Oakland County, Michigan. The signal improvements are part of the FAST-TRAC
(Faster and Safer Travel-Traffic Routing and Advanced Controls) system, a national
demonstration project for Advanced Traffic Management System (ATMS) and

Advanced Traveler Information System (ATIS), (RCOC, 1996).

A before/after study, (Taylor et. a1, 1997 and 1998), to examine the impacts of
SCATS implementation on Orchard Lake Road traffic revealed that the conidor travel
time improved for both directions of travel for both the peak and the non-peak periods.
The reduction in corridor travel time ranged from 6.56% to 31.80% with savings in
travel time being higher during the non-peak periods. Similar results were obtained
from studies carried out by the Australian Road Research Board (ARRP) and the
Department of Main Roads in New South Wales to evaluate the performance of

SCATS. Their comparison of SCATS and TRANSYT optimized fixed time signals

123

showed that SCATS can improve the travel time and number of stops from 3% to

18%, depending on the network and traffic characteristics, (ARRP 1982, and 1988).

6.1 Field study Objectives and Approach

The first part of this study was designed to determine which characteristics of
SCATS adaptive control caused the savings in travel time on Orchard Lake Road.
Before/after green time and offset analyses were carried out to examine how the
dynamic green time allocation (cycle length and phase plans) and the dynamic
progression (offset plans) contributed to the changes in intersection delay and corridor

travel time.

The second part of the study was a simulation analysis to examine whether
similar savings in Orchard Lake Road corridor travel time could have been achieved
through conventional control strategies namely; coordinated actuated and optimized
fixed-time signals. A comparison of the cost and benefits of different control systems

are presented at the end of this field study.

6.2 Study Location
The selected study location is a corridor (Orchard Lake Rd) in an urban district
in Oakland County, Michigan. The geometric configuration of the corridor is

presented in Figure 6.1.

124

€058 beam 08% 523 .8ch3 53:8 98% SEA 2380 fio oEwE

 

 

 

 

 

 

 

 

0:2 —.m
h p h p p L p p p
FoSLeSa 7334 «8: a «8? a «a: a a in d :5 a
m
«Eon anagram 3392 23333 13820 gawk—Em
Box 33 2280
/. M w k7 m Jr _._._ b 8 m
«1.2.33:— aafia _ — 354.233:— 353322 __ __ __ 28:55;
.32 33583

Al IIIW
go a OH. 5.32

6.3 Characteristics of SCATS adaptive control

Input data for the SCATS system installed in Oakland County are collected via
a system of video image devices mounted overhead on the signal strain poles or
attached to mast arms. The traffic information collected in the field includes the
discharge characteristics (i.e. flow and occupancy during the green phase) on each
intersection approach. This data is transmitted to a regional control center where the
SCATS control program attempts to maintain the highest degree of saturation at the
intersection. This is done by choosing optimal signal timing parameters from a library
of pre-defined signal timing and offset plans that were optimized off-line under

different traffic conditions (Lowrie, 1982 and 1990).

For dynamic signal coordination, SCATS chooses from four different pre-
defined offset plans. Two different offset values “PPl” and “PP2” are defined for each
plan. PPl is to be used if the cycle length is equal to or less than a pre-defined value
“C1” and PP2 is to be used if the cycle length is higher than or equal to a pre-defined
value “C2”. If the cycle length is between “C1” and “C2”, the system chooses the
offset linearly between PPl and PP2. Unlike the fixed time mode where all the signals
are coordinated to one intersection, SCATS coordinates the leading phase in one
intersection with a designated phase in another intersection, typically the leading phase

of the preceding intersection.

126

SCATS uses detectors placed downstream from the intersection to collect
traffic data of vehicles departing from the intersection during the green interval for
each approach. The system predicts the traffic profile for the next cycle based on the
volumes of previous cycles. This method allows SCATS to predict changes in traffic
demand over time rather than predict cycle-by-cycle variation in demand. This method
is similar to the prediction technique used in the second adaptive control strategy

examined in the first part of this research.

6.4 Before/after green time and offset analysis

The green time allocation and offset data for the after period were obtained
from SCATS system monitoring files. The file reports the cycle-by-cycle green time
allocated for each approach, the actual number of vehicles that departed during the
green phase per lane per approach, and the link and offset plans employed during this

cycle.

Before SCATS was employed, the signals were operating as fixed time signals
with the plan pre-determined according to the time of the day and the day of the week.
For the weekdays, three different plans were used, morning rush (6:00 - 9:00 AM),
evening rush (3:00 - 7:00 PM) and non-rush periods (9:00 AM - 3:00 PM) and (7:00
PM -6:00 AM). The cycle length, green split and offset remained constant during each

of these plans. As in any fixed-time signal system, the signal plans were optimized off-

127

line based on average historical data. The plans were optimized using the PASSER II-

90 program to maximize the through bandwidth for Orchard Lake traffic.

Because the cycle length was different for the before and afier periods, and was
variable in the after period, a comparison of the total effective green time per hour was
used in the analysis. The total effective green time per hour allocated to each
movement during the before and after periods were compared for the intersections
along the corridor, the results are presented in Table 6.1. The before/after green time
study revealed that SCATS, in general, extended the effective green time for the major
road traffic. This increase in green time led to a corresponding reduction in delay and
stops. Extending the green time increases the percentage of non-stopped vehicles and
reduces the stopping time for stopped vehicles. SCATS also used longer cycle lengths,

with less lost time per hour, thus utilizing the green time more efficiently.

A cycle-by-cycle degree of saturation comparison was used to test how
effectively the green phase was used for each approach during a cycle. When the total
entering traffic volume approaches the intersection capacity, a degree of saturation
equal to or near an optimal value of 0.90, would mean that the system effectively
assigned the right amount of green for that approach. Figure 6.2 presents the
before/after cycle-by-cycle degree of saturation for northbound and westbound traffic,

respectively (Orchard Lake Road and Walnut Lake Road intersection).

128

Table 8.1.a Green Time and Volumes (Walnut Lk Road Intersection)

 

_

Orchard Lk. Rd.

 

Walnut Lk. fid.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Through Volume Through Volume
Time Fixed Time SCATS SB NB Fixed Time SCATS EB we
6:00 AM 26 33 '75—7 295 18 13 351 65
7:00 AM 26 30 1678 690 18 16 620 160
8:00 AM 26 30 1703 700 18 17 589 215
9:00 AM 26 35 1204 620 16 14 353 195
10:00 AM 26 34 986 739 16 14 298 184
11:00 AM 26 34 969 948 16 14 318 198
12:00 PM 26 34 985 905 16 14 311 241
1:00 PM 26 33 931 885 16 14 271 228
2:00 PM 26 34 1076 1057 16 14 324 228
3:00 PM 24 28 946 1164 19 16 375 398
4:00 PM 24 27 890 1397 19 16 378 474
5:00 PM 24 24 918 1626 19 18 384 613
6:00 PM 24 27 968 1476 19 17 348 534
7:00 PM 26 33 885 1134 16 15 303 240
8:00 PM 26 34 880 1021 16 13 278 201
9:00 PM 26 32 887 884 16 14 273 196
10:00 PM 26 32 527 551 16 15 247 183
11:00 PM 26 30 299 388 16 14 150 131
Table 6.1.!) Average Degree of Saturation (Walnut Lk Road Intersection)
o'_'rchard'L—k. Rd. Walnut Lfid.
SB NB E3 wT3
Time Fixed Time SCATS Fixed Time SCATS Fixed Time SCATS Fixed Time SCATS
6:00 AM 0.51 0.40 0.20 0.16 0.69 0.96 0.13 0.18
7:00 AM 1.14 0.97 0.47 0.40 1.09 1.20 0.31 0.35
8:00 AM 1.16 1.00 0.48 0.41 1.15 1.22 0.42 0.44
9:00 AM 0.81 0.61 0.42 0.32 0.79 0.90 0.44 0.50
10:00 AM 0.66 0.51 0.50 0.38 0.67 0.78 0.41 0.48
11:00 AM 0.65 0.51 0.64 0.50 0.71 0.82 0.44 0.51
12:00 PM 0.66 0.52 0.61 0.48 0.70 0.78 0.54 0.60
1:00 PM 0.63 0.49 0.59 0.47 0.61 0.68 0.51 0.57
2:00 PM 0.72 0.56 0.71 0.55 0.73 0.84 0.51 0.59
3:00 PM 0.70 0.59 0.86 0.72 0.72 0.85 0.76 0.90
4:00 PM 0.65 0.59 1.03 0.93 0.72 0.83 0.90 1.04
5:00 PM 0.68 0.67 1.13 1.06 0.73 0.75 1.05 1.07
6:00 PM 0.71 0.64 1.09 0.98 0.66 0.74 1.02 1.13
7:00 PM 0.59 0.48 0.76 0.61 0.68 0.72 0.54 0.57
8:00 PM 0.59 0.46 0.69 0.54 0.62 0.74 0.45 0.53
9:00 PM 0.60 0.48 0.59 0.48 0.61 0.69 0.44 0.50
10:00 PM 0.35 0.29 0.37 0.31 0.55 0.58 0.41 0.43
11:00 PM 0.20 0.18 0.26 0.23 0.34 0.37 0.29 0.32

 

 

 

 

 

 

 

 

 

 

129

 

 

 

 

19:00

‘| b ‘ ‘2 h: . . . .
Mf'““-... I $n .
00.1! 4'“
fi‘ .- I
'5"

«v.31: . - . .. “g
-.s I ::: {.‘.'_':fi' ' '

o. ‘3“

 

P

_ ,. ..‘b. I.
«:1‘. l
a n . ‘

...x----é':

“Moon‘s"

 

 

l
17:00

1 o - D o o I D
...: C I . -
s1}..-
0 all a: a ::.1_'_"
.. a n r a 4- ::.::.._ ..
..P‘o") Pfig': . a .-.
new»! :." .° .—
c'.-.-.'.1“
“'1,

 

Pf. _ -
1...: -O"'

u ‘N. .... ...-..-
. . b-.:!==§b
_ -. 1

 

15:00

1' - .
. ."-:;:“
~“‘~~.-.‘%‘% e. ...-I.“ I .-

ie'H‘e‘h') .

 

.‘.'.‘J 'l ‘0

 

‘15:: t ' ’

““.-.o. 0'. - g - - .
upolsfifi$$$1$$'-'.: :1}}?-::“‘1~

 

 

..‘r.: o-..
Jill’l'l]; L-:.....::I:$-
1. 4...;{i fat...“
-3.
...“.O
""3.:‘;b|.:::.:;- 8 0
‘:--.-._’=W a g
"193:3 ." ‘-
..::-.'J|e
...I;.I.: I‘e‘|I.
”'43....‘F I o n

-
00...]. "
II.
"2' r:-.:-:-:-::
Mfi'.‘:‘.',

I--.=J

4‘?
I

”Oil-'5'. -. -.‘ .- .1

 

SCATS

 

es§5\'-'—'-q -‘ n no..—

1100

4

O. .3";

w. ' r I-
W .0. ::-'.’ ’~

col... : : :0...-
O

 

' ' ’ ' 55w .
‘D.... ... C
“I. .2“

 

 

“'-
«hue: mun-neat:

 

:5
‘5:
i;
i
Figure 6.2 SCATS/Fixed Time Cycle-By-Cycle Degree of Saturation for Northbound Tra ic

---Fixed‘l”ime|
l
t
i..
4::
900

 

 

 

 

 

 

 

1.60

 

 

 

 

 

1.40
2
6
O 40 .
0.20
0 00
7.

uonumes go 001er

130

The result shows, in general that SCATS provided a consistent degree of
saturation. However, the average difference in the degree of saturation between any
two consecutive cycles under SCATS control was 0.14 for northbound traffic and 0.19
for westbound traffic). This is larger than might be expected from an
adaptive/responsive system. This might be a result of the SCATS logic, which only
allows the phase plan or cycle length to change after receiving two votes in any three
consecutive cycles. This limits the ability of the system to respond to the cycle-by-
cycle changes in demand. A second factor is that each intersection is forced, under the
marriage mode, to have a common cycle length even if it is longer than the
intersection optimal cycle length. A third factor is the prediction techniques used in
SCATS which predicts the short-term variation in demand rather than cycle-to cycle

variation.

Improving progression for the traffic along the corridor is another potential
method of achieving a reduction in delay. Improving the progression would minimize
the number of vehicles arriving during the red phase, thus reducing the delay and
stops. A before/after progression analysis was conducted to investigate how SCATS
dynamic progression and offset plan contributed to the savings in delay and travel
time. Progression diagrams were used to compare the effectiveness of the before and

after progression.

I31

The total width of the through bandwidth per hour was chosen to be the
measure of effectiveness (MOE) in this analysis. This measure takes into consideration
the difference in cycle length between the before and after periods as well as the
variation in cycle length during the after period. The results showed that SCATS, in
general, provided a wider through bandwidth for the major street traffic during all time
periods. The results also showed that both fixed time and SCATS controls achieved an
optimal through bandwidth for the peak direction traffic during both the morning and
the afternoon peak periods. The through bandwidth under each control had an optimal
value equal to the total green time per hour for the intersection that had the lowest
green time for the northbound/ southbound traffic. This increase in the through
bandwidth under SCATS control is attributable mainly to the increase in the green

time allocated for the northbound/southbound traffic under SCATS control.

6.5 The simulation study

The objectives of this simulation analysis were to determine if conventional
signal control strategies can provide savings in travel time for Orchard Lake Road
traffic similar to those achieved by SCATS. The simulation analysis was carried out
for three different signal control strategies:
1) the base case, in which signal timings were set to represent the signal timing

conditions during the before period. The signals were operating under fixed time

132

signals with a cycle length of 120 seconds during the peak-periods. The results from
these simulation runs were compared against those collected in the field during the

before period to calibrate the simulation model.

2) re-optimized fixed-time signals, in which the cycle length was increased to 140
seconds during the peak-periods. This cycle length value represents the average
cycle length employed by SCATS during the peak period. The green split for
different approaches was set to the average values employed by SCATS. The offset

plans were obtained through the PASSER II-9O optimization of the network.

3) coordinated actuated signals, in which the common background cycle length and the
offsets were set similar to the second control strategy. The major conidor traffic
was set as the non-actuated approach and the gap-out was set to 3 seconds for all

other approaches.

The analysis was canied out for two time periods; the morning peak period
(7:00 AM to 9:00 AM) and the afternoon peak-period (4:00 PM to 6:00 PM). The
southbound traffic was the peak direction during the morning peak-period, whereas,
the northbound traffic was the peak direction during the afiemoon peak period. The
simulation was done using the CORSIM simulation model. One of the recent

improvements to CORSIM that made the simulation output more reliable, is the ability

133

 

to change traffic demand and turning movements within each time period. Traffic
volumes were obtained from the SCATS monitoring file, which reports traffic
volumes for each lane on a cycle-by-cycle basis. Simulations were run five times using
five different random number seeds for each case. The results from the five runs were

averaged to account for the variability associated with the randomness of the traffic.

6.6 Results from the simulation study

Figures 6.3 and 6.4 present average corridor travel time under different control
strategies for southbound traffic (morning peak-period), and northbound traffic
(afternoon-peak period), respectively. The corridor travel time obtained from
simulation with the signal settings similar to that during the before SCATS period
showed a similar pattern to the travel time data collected in the field. The error
between the simulated and actual travel time values ranged from 1.01% to 3.07% for

southbound traffic and 0.46% to 3.86% for northbound traffic.

The difference between the actual and simulated travel time is attributed to the
stochastic nature of the traffic. While the simulated data is the average of all vehicles
for five different runs using different random numbers, the field data is the average of
112 observations in a one-day period. While mid-block volumes were included in the
simulation runs, the volume and distribution of such traffic might be different than

what is actually in the field. This could be another possible source of the difference.

134

Nevertheless, the travel time pattern from the simulation results was similar to that

observed in the field, which validates the output of the simulation analysis.

Tables 6.2 and 6.3 present the average values of corridor travel time under
different control strategies during the peak 30-minute period, the non-peak 30-minute
period and the overall average. Tables 6.4 and 6.5 present the percent reduction in
conidor travel time under SCATS control compared to actual fixed-time data (before
SCATS data), optimized fixed-time (simulated data) and coordinated actuated

controllers (simulated data).

When compared to the optimized fixed-time signal, SCATS reduced arterial
travel time by an average of 5.61% and 3.00% during the non-peak 30-minute periods
for the southbound and northbound traffic, respectively. During the peak 30-minute
period, the reduction in travel time averaged 0.26% and 0.49%, respectively. These
values are similar to those obtained from the simulation model in the first part of this
study. The PHF for the minor— street traffic during these time periods ranged from 0.87
to 0.93, with an average volume of 243 vehicles per lane per hour. The expected
benefits from an adaptive system with minor street-traffic demand similar to these
values would be in the range of 3% to 6% (Figure 5.22). The values obtained through

this simulation study were within this range.

135

 

In.“ I'll ..-_._ ._ v.
I'

 

30:233.. 5552: - 058“— 952.530».

3583“» .0.—Eon 2295—. 35:: 2:5 .39: 32:3 oumee>< «6 2:9“.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

25...
85 9a one 26 85 me; one. are 8c.
. 8m
8m
\ 1w one
J _ Ill. 8e
#
A88 o: uov 33:53 333+ W one
:88 ozuo 83:53 eeeeeéilxli ,
e83 830 83:53 8.2.3 a
fies mp<ow+
fies 85 Bell
H %

 

 

com

(spuooes) owl; lanai; eBmeAv

136

 

Averaging cooEocToF—ub canon—535
«05383 .223 EoBEu Leon: 2:: _o>m.a 33:3 ouflo>< to 959...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

oEF
oon nan onun nan oonn n3. one 26
. 4 on~
m
A83 o! «.9 Rings 333+
:63 oiuo 83:55 eeooeelxl oon
:88 ofiuo 85:85 eeooexolol
ins mpgmlll
is: 2.: oselol . _
t onn
ooe
. 2 one
_
_
oon
onn

 

 

 

 

 

 

137

(epuooes) our!) ienu; eBeJeAv

Travel time values under actuated signal control obtained through simulation

were quite similar to those obtained in the field under SCATS control. The difference

between the average travel time under these two types of control ranged between

0.02% to 2.76%. This validates the output of the simulation model which

demonstrated that actuated signal control can achieve a reduction in travel time similar

to that obtained through adaptive signal control.

Table 6.2 Average corridor travel time* (southbound traffic-morning peak-

 

 

 

 

 

period)
Optimized
SCATS Before SCATS fixed-time Actuated
(field data) (field data) (simulated data) Simulated data
overall average 344.2 374.1 357.6 350.7
Average (non-peak)” 303.0 341.3 321.0 31 1.6
Average (peak)*** 381.5 406.0 382.5 380.5

 

 

 

 

 

* travel time in seconds

** average values during the first 30-minute period
*** average values during the peak 30-minute period

Table 6.3 Average corridor travel time* (northbound traffic-afternoon peak-

 

 

 

 

 

period)
Optimized
SCATS Before SCATS fixed-time Actuated
(field data) (field data) (simulated data) Simulated data
overall average 409.6 440.1 421.5 414.6
Average (non-peak)M 381.5 420.2 393.3 392.1
Average (peak)*** 429.4 434.0 431.5 429.5

 

 

 

 

 

* travel time in seconds

*"‘ average values during the first 30-minute period
*** average values during the peak 30-minute period

138

 

 

 

Table 6.4 Percent reduction* in corridor travel time under SCATS control
compared to different control strategies (southbound traffic-morning peak-

 

 

 

 

 

period)
Before SCATS Optimized fixed-time Actuated
Field data simulated data Simulated data
overall average 7.99 3.75 1.85
Average (non-peak) 11.2 5.61 2.76
Average (peak) 6.03 0.26 -0.29

 

 

 

 

*Percent reduction = (travel time (difi'erent control strategies) - SCA TS travel time) / travel time

Table 6.5 Percent reduction in corridor travel time under SCATS control
compared to different control strategies (northbound traffic-afternoon

 

 

 

 

 

peak-period)
Before SCATS Optimized fixed-time Actuated
Field data simulated data Simulated data
overall average 6.93 2.82 1.21
Average (non-peak) 9.21 3.00 2.70
Average (peak) 1.06 0.49 .02

 

 

 

 

*Pereent reduction = (travel time (different control strategies) - SCA TS travel time) / travel time

6.7 Comparative cost analysis of alternative signal control systems

Transportation engineers and planners use cost/benefit comparisons to decide

whether the benefit of any signal control system would justify the cost of installing,

operating and maintaining the system. Perhaps the most comprehensive comparison of

the cost of alternative control strategies in the literature is that included in the UTCS

study (Henry, et. a1, 1976).

139

 

 

 

The average cost per intersection is highly dependent on the size of the
network. For the first generation of control strategies, the average cost per intersection
(in 1975 dollar) was $17,353 and $13,988 for networks of 100 and 500 intersections,
respectively. Anderson (1984) presented the subsystem cost of an adaptive signal
control system as a percent of the total cost over a ten-year period (1970-1980). The
results suggest that the detector, the communication, and the central computer cost, as
a percentage of the total cost, decrease over the 10-year period. He suggested that the

cost of these subsystems would continue to decline over the years.

6.7.1 Cost components of alternative signal control systems

The cost of implementing a signal control system depends on many factors.
The geographic location, the area type, and the size of the system are among the
factors that affect the overall cost of the system. Data regarding the cost of fixed-time
and actuated systems were obtained through different sources and represent average
values. Data regarding the cost of SCATS adaptive system were obtained from OCRC
files, and it represents the cost of installation with 1994 dollars. Table 6.7 shows the

components of the cost of different systems.

140

 

Table 6. 7 Cost components of different signal control strategies

 

 

 

Fixed-Time Actuated Adaptive
Signal Controller Low High High
Detectors None High High
Communication None None High
Central Computers None None High
System maintenance Average Average High
Data maintenance High High None

 

 

 

 

The average cost of replacing a fixed-time signal system with a coordinated
actuated system would be in the range of $50,000 to $70,000 per intersection (Carrier
and Gable, 1998). Based on OCRC cost data, the cost of implementing SCATS in the
Oakland County traffic network ranged from $92,000 to $114,00 per intersection. This
cost included the detection system (four Autoscope@ cameras per intersection), the
local controller (one controller per intersection), and the average cost of the central

computer and the communication system per intersection.

6.7.2 Benefit components of alternative signal control systems

After determining the expected cost of alternative control strategies, the next
step is to assess the potential benefits of these alternatives. All relevant factors should
be evaluated and quantified before any rational decision can be reached. Table 6.8

presents the components of the benefits for alternative signal control systems.

141

 

 

Table 6. 8 Benefit components of different signal control strategies

 

Fixed-Time Actuated Adaptive

 

 

Improvement in traffic operation: ‘

-optimal control during peak-periods
-optimal control all time periods
-respond to incident and special event

-Safety improvements

-data maintenance and signal setting update

 

- data for ATIS applications

 

6.7.2.1 Benefit to traffic parameters (normal traffic conditions)

Campbell (1988) and Witkowski (1992) suggested that the change in the value
of the estimated peak-hour person-hour of stopped delay or travel time during
weekdays can be used as the measure of effectiveness when comparing different signal
control strategies in urban areas. The value of the benefit can be assessed by
multiplying the estimated saving in total travel time per hour by the assumed average
hourly value of an individual travel time (average $4/hr in urban areas, Witkowski

(1992)).

Assuming that the Orchard Lake Road corridor has a moderate peak-hour

demand that lasts for two hours in the morning and two hours in the afternoon peak.

142

An adaptive control system would reduce the total travel time from an average of
223.76 vehicle-hours/hour to 213.78 vehicle-hours/hour, with savings 9.98 vehicle-
hours/hour, (Table 5.5). Assuming an average occupancy 1.2 per vehicle, four peak-
hours per day, and an average of 253 weekdays per years, the total benefits over one-
year period would be:

9. 98 x 1.2 x 4 x 253x $4/hr = $48,488/j’ear

6.7.2.2 Benefit to traffic parameters (incident and special-event conditions)

Many studies concluded that adaptive control systems have the potential of
reducing total delay resulting from incident situations. It is difficult; however, to
quantify such benefit due to the randomness associated with incident occurrence,
duration and severity. An average value, based of historical data of the non-recurring
congestion resulting from incidents and special events can be used as guide to quantify

such benefits

6.7.2.3 Changes of the adaptive control system benefit over time span

Another significant benefit of an adaptive system is the ability to
optimize the system on a real-time basis, something that is very costly and labor-
intensive for non-adaptive systems. As illustrated in section 5.3.4 of this study, the
reduction in conidor travel time changes over time. When the fixed-time signal plans

were kept unchanged, the average reduction in travel time increased from 4.16% in the

143

 

'E ...'. .T

base year, to 9.06% in the first year, to 26.33% in the second year. When compared
with signal plans that optimized annually, the reduction in travel time decreased from
4.16% in the base year to 3.04% in the first year, then to 1.57% in the second year.
The benefit of the adaptive control, thus, would be tied to the frequency of updating

the signal plans in the network.

Signal plan updating requires extensive data collection and many man-hours of
data management and network optimization. It would be justifiable thus, when
estimating the benefit of the adaptive system over a span of time, to include the
savings in data management and signal plan updating as a relevant benefit of the

system. The value of this benefit would vary among agencies.

6.7.2.4 Other benefits of adaptive control systems

The benefits of adaptive signal control systems extend beyond the reduction in
delay and travel time. Adaptive systems are the core of any Advanced Traveler
Information Systems (ATIS) application as they provide the system with real-time
traffic data required for route choice algorithms. While significant, it might be difficult
to explicitly quantify its dollar value outside the ATIS benefit context. Many studies
reported some safety benefits associated with adaptive signal control. The benefits
ranged from marginal (Hanbali and Fornal, 1997) to significant as reported by

Richeson and Underwood (1996).

144

 

Chapter 7

CONCLUSIONS

This research examined the potential benefits of adaptive signal control
strategies for arterial corridors in urban networks. Two adaptive signal control
strategies were examined against two conventional control methods, namely
coordinated fixed-time and coordinated actuated signals. The difference between the
two signal control strategies was in the way they predict traffic. In the first adaptive
control strategy, a prediction of the traffic in any given cycle was made based in the
traffic on the previous three cycles using traffic data collected via detectors placed
downstream from the intersection. In the second adaptive control strategy, a prediction
was made using detectors placed upstream to the intersection that scan the traffic
expected to arrive at the intersection. The research was conducted using the CORSIM
simulation model. Using the same platform, these four signal control strategies were
tested. The measures of effectiveness used to identify the performance of each control

strategy were average travel time, total travel time, and intersection delay parameters.

The effectiveness of the four signal control strategies was examined under two
different demand levels. The demand levels were chosen to replicate moderate and
high peak-hour conditions in an urban corridor. The output of the simulation study

showed that both adaptive control strategies showed a reduction in corridor travel time

145

 

over well-optimized fixed-time signals. The percent reduction in total travel time was
higher in the moderate peak-hour demand case. Within each demand level, the savings
in total travel time was higher during the non-peak periods and decreased as traffic

demand reached its peaking (Table 7.1).

Table 7.1 Percent reduction in total travel time under adaptive control systems

 

lS-minute non-peak periods 15-minute peak period

 

High peak-hour demand 1.89 0.40

 

 

Moderate peak-hour demand 5.90 1.20

 

 

 

 

When compared with the coordinated actuated signals, neither of the two
adaptive control strategies examined in this study showed a significant difference in
corridor travel time or intersection delay parameters. This was true in both moderate
and high peak-hour demand levels. Thus, it can be concluded that, within the demand
levels examined in this study, a coordinated actuated signal system can provide

savings in corridor travel time similar to those achieved by an adaptive control system.

The output of a signal timing analysis study showed that the reduction in
conidor travel time was achieved mainly through reallocating the extra green time,
provided to the minor-street traffic under fixed time control, to the major-street traffic.
This happened when the minor-street traffic demand was lower than its peak design

values during the non-peak periods. Similarly, during peak periods, when minor-street

146

 

F‘—

traffic demand reaches its peak design value, the value of the extra green time that can
be reallocated decreases, decreasing the potential corridor travel time savings by

adaptive signal control systems.

The sensitivity of the adaptive signal control benefit to various traffic
parameters was also examined. These parameters included minor-street traffic demand
and PHF, major-street traffic demand, the phase plan, and a change of demand over
time. The output showed that the saving in corridor travel time under adaptive signal
control system is more sensitive to the PHF of the minor-street traffic than other traffic
demand parameters. As the PHF decreases, the excess green time allocated for the
minor-street traffic increases, providing the adaptive control system with more
potential benefits. No significant relationship was found between the percentage
reduction in corridor travel time and changes in minor-street or maj or-street demand

levels.

The corridor travel-time savings were not limited to a specific signal plan.
Similar conidor travel time saving was obtained when the signals operated in two-

phase and four-phase plans.

The reduction in corridor travel time changes over time. When the fixed-time

signal plans were kept unchanged, the average reduction in travel time under adaptive

control increased from 4.16% in the base year, to 9.06% in the first year, and to

147

 

26.33% in the second year. When compared with signal plans that optimized annually,
the reduction in travel time decreased from 4.16% in the base year to 3.04% in the first

year, then to 1.57% in the second year.

The accuracy of the two prediction techniques used in the adaptive control
systems was examined. The output showed that detectors placed upstream to the
intersection predicted the expected number of vehicles for each cycle more accurately.
However, the accuracy of predictions made using detectors placed downstream from
the intersection improved when the prediction horizon increased to 5-minute intervals.
There was no significant difference in the performance of the two adaptive control

systems that used these two prediction techniques.

A field study to examine the effectiveness of deploying SCATS adaptive
control system in the Orchard Lake Road corridor showed that the reduction in
corridor travel time achieved under SCATS control was within the benefit limits
estimated by the simulation model. The results also showed that, similar to what the
simulation model output demonstrated, coordinated actuated signals might have

achieved the same reduction in travel time.

This research has contributed to the understanding of the potential benefits of

adaptive signal control strategies in urban corridors. It should be noted that the output

reported in this study is limited to the geometric configurations and demand levels

148

 

examined in the study. The benefits of the adaptive systems might vary if the spacing
between the intersections is significantly different than those examined in the study,
since the signal spacing is a significant factor in determining the benefits of signal

progression.

The study also identifies several areas where there is a need for further research.
This research can, ultimately, lead to a set of rules and guidelines for choosing a signal
control system among different alternatives within an ITS context. The followings are

some proposed areas for further research in the framework of this study:

1. The study can be extended to cover additional demand levels during non-peak
periods and oversaturation conditions.

2. The study can be extended to examine different geometric configurations,
(intersections that are closely spaced or set further apart).

3. The same simulation model can be modified and used to examine the effectiveness

of adaptive control systems for urban networks.

4. Different adaptive control logic can be tested using the same simulation model
introduced in this study.

5. The same study framework can be used to test the effectiveness of adaptive control

strategies in urban networks under different demand levels.

149

 

APPENDICES

150

 

Appendix A

Data structures in CORSIM

 

Variable name

Description

Fortran
common block

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IMXLNK Maximum allowable number of links in a network SNET

IMXNOD Maximum allowable number of nodes in a network SNE'I‘Z

IMXDET Maximum allowable number of detectors in a network SNET3

YlNlT Flag that records if the simulation has reached equilibrium or GLR091
not.

DWNOD[IL] Downstream node number of link 1L, where IL is the CORSIM SIN036
link ID.

UPNOD [IL] Upstream node number of link IL, where IL is the CORSIM SIN062
link ID

NMAP[IN] User-specified node number that corresponds to the internally- SIN075
assigned NETSIM subnetwork node, number, IN

CLOCK Current time since start of the simulation in seconds. SIN104

TI‘LNK Total number of links in subnetwork. SlNl 16

DTLNK[DT] Link Number surveillance detector DT is on. SlN700

DTLEN(DT) Length of Detector DT in tenths of a foot SIN313

DTMOD(D'I) 1-3 Detector type (O=Presence, 1=Passage) SIN314
4-10 Speed of vehicle when passing the passage detector.
1 1-23 Vehicle count since beginning of simulation.

DTNLNK(DT) Detector identification number of the next detector on the same SIN 308
link as detector DT.

DTPOS(DT) Distance between the detector's downstream edge and the SIN312
downstream stop-bar, in tenths-of-a-foot.

DTFLNKUL) Detector identification number of the first detector on the SIN307

 

referenced link, 1L.

 

 

151

 

 

 

Variable name

Description

Fortran
common block

 

DETON(DT)

Bit specific array:

1 (0,1) if detector was (on,off) for first 0.1 second

2 (0,1) if detector was (on,off) for second 0.] second
3 (0,1) if detector was (on,of1) for third 0.1 second

4 (0,1) if detector was (on,off) for forth 0.1 second

5 (0,1) if detector was (on,off) for fifth 0.1 second

6 (0,1) if detector was (on,off) for sixth 0.1 second

7 (0,1) if detector was (on,off) for seventh 0.1 second
8 (0,1) if detector was (on,off) for eighth 0.1 second
9 (0,1) if detector was (on,off) for ninth 0.1 second
10 (0,1) if detector was (on,off) for tenth 0.1 second

SIN070

 

TOTDET

Number of surveillance detectors specified for this
subnetwork. The data for each detector is contained in arrays
DTCTRl, DTCTRZ and DTCTR3.

SIN109

 

AMBSPCUL]

The subscript K is the link identification number and contains
the movements which are in amber.

1 (0,1) if signal for right tumors (is, is not) amber

2 (0,1) if signal for through vehicles (is, is not) amber

3 (0,1) if signal for diagonal tumors (is, is not) amber

4 (0,1) if signal for lefi tumors (is, is not) amber

5 - 11 Time remaining (sec) in amber for right turners

12-18 Time remaining (sec) in amber for through vehicles
19-25 Time remaining (sec) in amber for diagonal tumors

26-32 Time remaining (sec) in amber for left turners

SIN366

 

 

SDCODE[IL]

 

Signal codes of current, active, signal interval controlling
traffic on this link, fetched from XINTl or XTNTZ.
1 Signal code for right turn vehicles
2 Signal code for through vehicles
3 Signal code for diagonal (left or right) vehicles
4-5 Signal code for left turn vehicles
Where Signal Code is defined as follows
0 GO, permitted and protected
1 NO GO, not permitted
2 COND, GO, permitted but unprotected (left tumers

 

SINOSO

 

152

 

 

 

Appendix B

Source Code for the adaptive control logic interface with CORSIM

C Program ADAPTIVE_1.for - PC version
C By: Ahmed Shawky Abdel-Rahim

C Date: 4/27/98

C Last updated: 6/16/98

C "‘ FOR PHD DISSERTATION "*"
C Purpose: INTERFACE on a real-time basis with CORSIM
C

C -—---------- GLOSSARY OF VARIABLE NAMES

 

 

CLOCKR REALTIME CLOCKTIME

CREAL STRING OF FLAGS FOR SECOND-SPECIFIC DATA

CREALD STRING OF FLAGS FOR DEBUG DATA

CRELCF CHARACTER - PATHNAME WHERE THE REAL-TIME SOFTWARE
READS AND WRITES TO THE CONTROL FILES

CRELDF CHARACTER - PATHNAME WHERE THE REAL-TIME SOFTWARE
READS AND WRITES TO THE DATA FILES (DIRECTORY WHERE
NETSIM WAS EXECUTED)

FLREAL ARRAY OF FLAGS SPECIFYING WHICH FILES TO CREATE FOR REAL-
TIME TRAFFIC ADAPTIVE SIGNAL CONTROL SYSTEMS

ISMLEN TIME TO COMPLETE SIMULATION TO END OF CURRENT TP, SECONDS

IT INDEX OVER TIME PERIOD DURATIONS

IW FLAG SET (0,1) IF CONTROL FILE (HASNT,HAS) BEEN UPDATED

ITOTME TOTAL EXECUTION TIME SET FOR NETSIM SIM RUN, SECONDS

LENINT LENGTH OF A TIME INTERVAL, SECONDS

LNTMPR ARRAY OF TIME PERIOD DURATIONS, SECONDS

LU21 PERIPHERAL UNIT NUMBER 21

LU26 LOGICAL FILE (CONTROL FILE FOR SECOND-SPECIFIC DATA)

LU27 LOGICAL FILE (CONTROL FILE FOR DEBUG DATA)

LU33 LOGICAL FILE (FILE FOR DEFINING PATHNAME, ITOTME, LENINT,
AND LENPRD)

WRETV FLAG SET (T,F) IF (IS,ISNT) TIME TO RETRIEVE NETSIM DATA

IMPLICIT INTEGER (A-Q, S-V, X), REAL (R, Z), LOGICAL (W, Y)

O OOOOOOOOOOOOOOOOOOOOOOOO

COMMON /GLR898/ CLOCKR
COMMON /GLR183/ CRELCF
COMMON /GLR184/ CRELDF
COMMON /GLR894/ FLREAL( 5)
COMMON /GLR060/ LENINT
COMMON IGLR014/ LNTMPR( 19)
COMMON /GLR888/ LU21
COMMON IGLR893/ LU26
COMMON /GLR895/ LU27

153

 

COMMON /GLR185/ LU33
C
CHARACTER CREAL‘32, CREALDH 1, CRELCF‘BS, CRELDF*35
C
INQUIRE (FILE='RELNET.I>TH', EXIST=WEX)
IF (.NOT. WEX) THEN
WRITE (*, 7002)
CALL EXIT
ELSE
C
C -—-- CHECK PATHNAME FOR THE CONTROL FILES.
C
CALL FOPEN (33)
READ (LU33, 7003) CRELCF
IC = ICHAR (CRELCF(1 : 1))
IF (IC .LT. 65 .OR. IC .61“. 122 .OR.
1 (IC .GT. 90 .AND. IC .LT. 97)) THEN
WRITE (I, 7004)
CALL EXIT
ENDIF
IF (CRELCF(2:2) .NE. 2') THEN
WRITE (*, 7005)
CALL EXIT
ENDIF
IC=36
5 CONTINUE
IC = IC - 1
IF (CRELCF(IC:IC) .EQ. ' ' .AND. IC .GT. 1) GO TO 5
IF (CRELCF(1C:IC) .EQ. 1') CRELCF(IC:IC) = ' '
C
C ---- CHECK PATHNAME FOR THE DATA FILES.
C
READ (LU33, 7003) CRELDF
IC =1CHAR(CRELDF(1:1))
IF (IC .LT. 65 .OR IC .GT. 122 .OR.
1 (IC .GT. 90 .AND. IC .LT. 97)) THEN
WRITE (t, 7007)
CALL EXIT
ENDIF
IF (CRELDF(2:2) .NE. ':') THEN
WRITE (*, 7008)
CALL EXIT
ENDIF
IC = 36
10 CONTINUE
IC = IC - 1
IF (CRELDF(IC:IC) .EQ. ' ' .AND. 1C .GT. 1) GO TO 10
IF (CRELDF(IC:IC) .EQ. '\') CRELDF(IC:IC) = ' '
WRITE (I, 7011) CRELCF, CRELDF
PAUSE
CLOSE (LU33)
ENDIF
C

154

 

Plane

 

 

C --—- GET TIME INVARIANT DATA.
C
CALL FOPEN (26)
READ (LU26, 5001) ITIME, IRCODE, INCODE, CREAL
IF (ITIME .EQ. 1 .AND. INCODE .EQ. 1) CALL RINv
CLOSE (LU26)
C
C ...—- DETERMINE TOTAL LENGTH OF CORSIM SIMULATION RUN, IN SECONDS.
C
ITOTME = 0
DO 20 IT = 1, 19
ITOTME = ITOTME + LNTMPRaT)
20 CONTINUE
IT = 1
ISMLEN = LNTMPR(IT)
C
C ---- LOOP OVER SECONDS.
C
DO 60 [CLOCK = 1, ITOTME
CLOCKR = [CLOCK
IW = o
WSEC = MOD (ICLOCK, FLREAL(1)) .EQ. 0
WTI = FLREAL(2) .EQ. 1 .AND. MOD (ICLOCK, LENINT) .EQ. 0
WTP = FLREAL(3) .EQ. 1 .AND. MOD (ICLOCK, ISMLEN) .EQ. 0
WDB = FLREAL(4) .EQ. 1
WRETV = WSEC .OR. WTI .OR. WTP .OR. WDB

C ----- UPDATE THE SIMULATION RUN LENGTH IF INTO THE NEXT TIME PERIOD.

1F (ICLOCK .EQ. ISMLEN + 1 .AND. ICLOCK .NE. ITOTME) THEN
IT = IT + 1
ISMLEN = ISMLEN + LNTMPR(IT)

ENDIF

C ---- READ THE SECOND-SPECIFIC CONTROL FILE. "WAIT" BY LOOPING
C ---- IF NECESSARY UNTIL CORSIM PROCESSES ANOTHER SECOND AND UPDATES
C ---- CONTROL FILE (INCODE = 1).

25 CONTINUE
30 CONTINUE
CALL FOPEN (26)
READ (LU26, 2000) ITIME, IRCODE, INCODE
IF (INCODE .NE. 1) THEN
CLOSE (LU26)
GO TO 30
ENDIF
BACKSPACE LU26

C ----- CORSIM HAS PROCESSED ANOTHER SECOND. RESET FLAGS.

IRCODE = l
INCODE = 0

155

 

 

C ---- REALTIME IS READY TO RETRIEVE CORSIM DATA (WRETV = TRUE).
C
IF (WRETV) THEN
C
C ---- THE CORSIM SIMULATION TIME IS THE SAME AS THE REALTIME
C ---- PROCESSING TIME.

C
IF (ITIME .EQ. ICLOCK) THEN
C
C ---- GET DESIRED NETSIM DATA.
C
IF (WSEC) CALL RSEC (CREAL)
IF (WTI) CALL RTI
IF (WTP) CALL RTP
C

C ---- MAKE CHANGES TO SIGNALS AND OUTPUT IF IT IS TIME.
C
IF (WSEC) THEN
IDUM = 0
DO 351=1, 10000
IDUM = IDUM + 0
35 CONTINUE
CALL ROUTS (CREAL)
ENDIF
C
C GET DEBUG DATA.
C
IF (WDB) THEN
40 CONTINUE
CALL FOPEN (27)
IF (ICLOCK .EQ. 1) THEN
READ (LU27, 6000) ITIMD, IRCODD, INCODD, CREALD
ELSE
READ (LU27, 2000) ITIMD, IRCODD, INCODD
ENDIF
IF (ITIMD .EQ. [CLOCK .AND. INCODD .EQ. 1) THEN
CALL RSECD (CREALD)
CLOSE (LU27)
ELSE
CLOSE (LU27)
GO TO 40
ENDIF
ENDIF
C
C ---- REALTIME IS READY TO RETRIEVE BUT MUST WAIT UNTIL CORSIM
C ----- CREATES THE DATA (FOR THIS SECOND). UPDATE CONTROL FILE.
C

ELSE
WRITE (LU26, 5000) ICLOCK, IRCODE, INCODE, CREAL
CLOSE (LU26)

GO TO 25

ENDIF

156

"ml

 

‘i

 

C --—-- REALTIME IS NOT READY TO RETRIEVE. UPDATE CONTROL FILE AND
C ---- RETURN TO PROCESS NEXT SECOND.

C
ELSE
IW = l
WRITE (LU26, 5000) ICLOCK, IRCODE, INCODE, CREAL
CLOSE (LU26)
ENDIF
C
C ---- BE SURE TO UPDATE CONTROL FILE.
C

IF (Iw .EQ. 0) THEN
WRITE (LU26, 5000) ICLOCK, IRCODE, INCODE, CREAL
CLOSE (LU26)
ENDIF
END

157

 

 

Source file for CORSIM

Appendix C

CORSIM main source file - Prepared by Ahmed Shawky Abdel-Rahim
dissertation

300 300 300 300 300 300

8021
21

1

1

11
8011
8022
22

12
8012
8023

23

13
8013
8024

24

14
8014
8025

25

15
8015
8035

w
0101

l—‘NNWWAAU‘

for Ph. D. Thesis
April,

1 0

122000
22000

12

23
32000

232000

132000
32000

13

24
42000

242000

142000
42000

14

25
52000

252000

152000
52000

15

35
52600

352600
43700
53700
33700
43700
23200
33200
13200
23200

1998

-15

6O

NNNl\)NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN

7981

9 24
300000

300 300 300 300 300

7T
TT
TT
7T

7T
TT
TT

7T
TT

7T

7T
TT
TT
7T

7T
TT
TT
7T

TT
7T
TT
7T
7T
7T
7T
7T
7T
7T
7T

11
8021
8011

21

12
8022
8012

22

13
8023
8013

23

14
8024
8014

24

15
8025
8015

25

8035

LA)
(Ail—‘15

158

98

300

31

35

25

24
15
23
14
22
13
21
12

ll

21

12

22

13

23

14

24

15

25

(A)
wal—‘UINUTW A

(A)

30
30
30
3O

30
30
30
30

3O
3O
30
3O

30
30
3O
3O

30
3O
3O
3O

45
45
45
45
45
45
45
45
45
45

7581

11
11
11
11

11
11
11
11

11
11

11

11
11
11
11

ll
11
11
11

11
ll
11

ll
11
11
11
11
22
11

 

 

 

31
8031
8021

21

11
8011
8022

22

12
8012
8023

23

13
8013
8024

24

14
8014
8025

25

15
8015
8035

LA)
U‘

8(3

waHHl—‘HHNNUJUJAAUWU‘

HNNNNNww
Feulblybokdhiuiw

31
1
31
21
l
21
11
1
11
22
2
22
12
2
12
23
3
23
13
3
13
24
4
24
14
4
14
25
5
25
15
5
15
35

LA)
HMI—‘WNAUM-fi-U‘Lfl

bacon on
l-‘Ol—‘H

112
56

3200
3200

100
90
100
100
90
100
100
90
100
100
90
100
100
90
100
100
90
100
100
90
100
100
90
100
100
90
100
100
90
100
100

100

10

10

10

10

10

10

10

10

10

10

10

10

10

10

10

10

10

10
10

10

LfiU'lilsz

ALDBJF‘H

TT
7T

87
87
87
87
87

UiU'lUlflflUi

8031

2

l

23 5
23 5
23 5
23 5
23 5

159

11

00000

00000

2

00000

20
20
20

00000

20
22
22

00000

45
45
0

00000

11
11

00000

00000

11
11
11
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
21
35
35
35
35
35
35
35
35
35
35
35
35
35

 

 

E3021
8022
8023
8024
E3025
E3011
8012
8013
E3014
E3015
8035
8031

28012
38013
48014
58015
2299 2200 9922
2299 2200 9922
2299 2200 9922
2299 2200 9922
2299 2200 9922
5 1
5 1 22480
5 l 2950
21 475
11 475
22 475
12 475
23 475
13 475
24 475
14 475
25 475
15 475
352200
311600
31 l 1 2
35 5 5 4
4000 18400
4000 15200
4000 12000
4000 8300
4000 4600
4000 2000
4000 21600
6000 18400
6000 15200
6000 12000
6000 8300
6000 4600
2000 18400
2000 15200
2000 12000
2000 8300
2000 4600
7000 18400
7000 15200
7000 12000
7000 8300
7000 4600
1000 18400
1000 15200
1000 12000
1000 8300
1000 4600
4000 500
4000 22600
3

0022
0022
0022
0022
0022

0000
0000
0000
0000
0000
1

l

0000
0000
0000
0000
0000

200
200
200

0000
0000
0000
0000
0000

0000
0000
0000
0000
0000

160

0000
0000
0000
0000
0000

0000
0000
0000
0000
0000

35
31

0000
0000
0000
0000
0000

0000
0000
0000
0000
0000

35
35
35
35
36

36
36
36
42
42
42
50
50
50
50
50
50
50
50
50
50
50
50
90
90
170
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
210

 

BIBLIOGRAPHY

161

 

BIBLIOGRAPHY

Abdel-Rahim, Ahmed (1998) Analysis of Corridor Delay under SCATS Control,
Intelligent Transportation Society of America’s Eighth Annual Meeting, Detriot,
USA, D-ROM.

Abdel-Rahim, Ahmed (1998) The SCATS Effect, Traffic Technology International,
Feb/Mar, pp. 57-60.

Abu Lebdeh, G., and R. F. Benekohal (1997) Development of traffic control and queue
management procedure for oversaturated arterial, Transportation Research Record
1603, TRB, National Research Council, Washington, DC, pp 150-155.

Allsop, RE. and J .A. Charlesworth (1977) Traffic in a signal-controlled road network: an
example of different signal timings inducing different routing. T rafi‘ic
Engineering and Control, May, pp 262-267.

Ancker Jr., C]. and V. Gafarian (1967) A simple renewal model of throughput at an
oversaturated signalized intersection. Transportation Research A, Vol 1, pp 57-
65.

Andrews, C. A., S. M. Elahi, and J. E. Clark (1997) Evaluation of New Jersey Route 18
OPAC/MIST Traffic Control System. Transportation Research Record 1603,
TRB, National Research Council, Washington, DC, pp 150-155.

d’Ans, GO and DC. Gazis (1976) Optimal control of oversaturated store-and-forward
transportation networks. Transportation Science, Vol 101, pp 1-19.

Bell, MC. and PT. Martin (1990) Use Of detector data for traffic control strategies. Third
International Conference on Road Trafi‘ic Control, Institute of Electrical
Engineers, Conference Publication NO 320, pp 86-89.

Bretherton, R.D. et. al. (1986) The use of SCOOT for traffic management. Second
International Conference on Road T rafi‘ic Control, Institute of Electrical
Engineers, Conference Publication NO 260, pp 81-84.

Bretherton, RD. and GT. Bowen (1990) Recent enhancement to SCOOT - SCOOT

version 2.4. Third International Conference on Road Traffic Control, Institute of
Electrical Engineers, Conference Publication NO 320, pp 95-98.

162

 

 

Davies, P. et. a]. (1991) Assessment of advanced technologies for relieving urban traffic
congestion. NCHRP Report 340, Transportation Research Board, Washington,
DC.

Dell’OlmO, P., and p. B. Mirchandani (1995) REALBAN: An Approach for Real-Time
Coordination of Traffic Flows on Networks. Transportation Research Record
1494, TRB, National Research Council, Washington, D.C., pp 106-114.

Drew, DR. and C. Pinnell (1965) Some theoretical considerations of peak-hour control of
arterial street systems. T raflic control: Theory and instrumentation, Instrument
Society of America, Plenum Press, NY, pp 91-123.

Easa, SM. and AD. May (1981) Evaluation of traffic management strategies in central
business districts. Transportation Research Record 816, National Academy of
Sciences, Washington, DC, pp 1-10.

Gafarian, A.V., and A. Halati (1986) Statistical Analysis of Output Ratios in Traffic
Simulation. Transportation Research Record 1091, TRB, National Research
Council, Washington, D.C., pp29-36.

 

Gartner, N. H., C. Stamatiadis, and P. J. Tarnoff (1995) Development of Advanced
Traffic Signal Control Strategies for Intelligent Transportation System: multilevel
Design. Transportation Research Record 1494, TRB, National Research Council,
Washington, D.C., pp98-105.

Gartner, NH. and D.L. Hou (1994) Comparative evaluation of alternative traffic control
strategies. Transportation Research Record 1 3 60, Transportation Research
Board, Washington, DC, pp 66-73.

Gartner, N. H., P. J. Tarnoff, and C. M. Andrew (1991) Evaluation of Optimized Policies
for Adaptive Control Strategies. Transportation Research Record 1324, TRB,
National Research Council, Washington, D.C., pp105-114.

Gartner, N. H. (1983) OPAC: A Demand-Responsive Strategy for Traffic Control.
Transportation Research Record 906, TRB, National Research Council,
Washington, DC.

Gartner, N. H. (1982) Prescription for Demand-Responsive Urban Traffic Control.
Transportation Research Record 881, TRB, National Research Council,
Washington, D.C., 73-75 pp.

Gazis, DC. (1964) Optimum control Of a system of oversaturated intersections.
Operations Research, Vol 12, pp 815-831.

163

Gerlough, D.L. (1965) Simulation as a tool in traffic control system evaluation. T raflic

control: Theory and instrumentation, Instrument Society of America, Plenum
Press, NY, pp 71.

Gerlough, D.L. and RA. Wagner (1967) Improved criteria for traffic signals at individual
intersections. NCHRP Report 32, Transportation Research Board, Washington,
DC.

Gordon, R.L. (1969) A technique for control of traffic at critical intersections.
Transportation Science, Vol 3, NO 4, pp 279-288.

Hadi., M. and C. E. Wallace (1992) progression-Based optimization Model in
TRANSYT-7F. Transportation Research Record 1360, TRB, National Research
Council, Washington, D.C., pp 74-81.

Hanbali, R. M., and C. J. Fomal (1997) Methodology for Evaluating Effectiveness of
Traffic-Responsive Systems on Intersection Congestion and Traffic Safety.
Transportation Research Record 1603, TRB, National Research Council,
Washington, D.C., pp137-149.

 

Henry, R., D., R. A. Ferlis, and J. K. kay (1976) Evaluation of UTCS control strategies.
U. S. department of Transportation, Federal Highway Administration,
Washington, DC.

Head, K. L., P. B. Mirchandani, and D. Sheppard (1992) Hierarchical Framework for
Real-Time Traffic Control. Transportation Research Record 1360, TRB, National
Research Council, Washington, D.C., pp 82-88.

Hsin, V. K., and P. T. Wang (1992) Modeling Concepts for Intelligent Vehicle Highway
Systems (IVHS) Applications. Proceedings of the 1992 Winter Simulation
Conference, Arlington, VA December, pp. 1201-1209.

Institute of Transportation Engineers (1989) Toolbox for alleviating traflic congestion.
Washington, DC.

Kagolanu, K. (1994) A comparative study of traffic control system. IT E Compendium of
Technical Papers, Institute of Transportation Engineers, pp 374-3 78.

Kay, J .L., J .C. Allen, and J .M. Bruggeman (1975) Evaluation of first generation

UT CS/BPS control strategy. Report FHWA-RD-75-26, Federal Highway
Administration, March.

164

Kosonen, I. (1990) A Simulation tool for traffic signal control planning. Road Traffic
Control, Third International Conference on Road T raflic Control, Institute of
Electrical Engineers, Conference Publication No 320, pp 72-76.

Kwon, E., and Y. J. Stephanedes (1998) Development Of an Adaptive Control Strategy in
a Live Intersection laboratory. Proceeding of the 78th Transportation Research
Board Annual Meeting, Washington, D.C., CD-ROM.

Lin, F., and M. C. Percy (1985) Vehicle-Detector Interactions and Analysis of Traffic-
Actuated Signal Controls. Transportation Research Record 971 , TRB, National
Research Council, Washington, D.C., pp112-120.

Little, J .D.C. (1966) The synchronization of traffic signals by mixed-integer linear
programming. Operation Research, Vol 14, No 4, pp 568-594.

Longley, D. (1968) A control strategy for a congested computer-controlled traffic
network. Transportation Research, Vol 2, pp 391-408.

Lowrie, PR. (1982) Sydney coordinated adaptive traffic control system-Principle,
methodology, algorithm. International Conference on Road T raflic Signaling,
Institute of Electrical Engineers, Conference Publication No 207, pp 67-70.

Lowrie, P. R. (1991) SCATS, Sydney Coordinated Adaptive Traffic System, A Traffic
Responsive Method of Controlling Urban Traffic. Road and Traffic Authority of New
South Wales, Sydney, Australia.

May, AD. and RC. Montgomery (1986) Avoiding the need for manual intervention in
signal control: A case study of Bangkok. Second International Conference on
Road T raflic Control, Institute of Electrical Engineers, Conference Publication NO
260.

May, AD. and FD. Montgomery (1993) The assessment of fixed-time signal control in
highly-congested conditions. Traflic Engineering and Control, pp 176-179.

McShane, W.R. and RP. Roess (1990) Traffic Engineering. Prentice Hall, Englewood
Cliffs, NJ.

Michalopoulos, PG. and G. Stephanopoulos (1977a) Oversaturated signal systems with

queue length constraints-I: Single intersection. Transportation Research A, Vol
11, pp 413-421.

Michigan Department of Transportation (1992) Metropolitan Detroit Advanced T rafiic

Management System (A TMS). Metropolitan Transportation Center, Traffic &
Safety Division.

165

 

 

Montgomery, F0. and DJ. Quinn (1992) Queue management strategies using fixed time
signal control. IEE Conference 355, April, pp 80-84.

National Cooperative Highway Research Program (1967) Improved criteria for traflic
signals at individual intersections. Report No 32, Highway Research Board,
Washington, DC.

. (1969) Improved criteria for traflic signal systems on urban arterials.
Report No 73, Highway Research Board, Washington, DC.

 

. (1971) Improved criteria for traflic signals in urban networks. Report NO
124, Highway Research Board, Washington, DC.

 

Nip, K.F. (1975) A comparative study of traffic control schemes at an isolated
intersection. Trafi‘ic Engineering and Control, pp 264-269.

OECD (1985) Dynamic traflic management in urban and suburban road systems. Road
Transport Program, Reported by Scientific Expert Group MC3, Paris, France.

 

OECD (1981) T raflic control in saturated conditions. Road Research, Paris, France.

Passer II-90 Program User Manual.(l991). Texas transportation Institute, College Station.

Peytechev, E., and A. Arglie (1998) Integrative Framework for Predictive Evaluation of
Traffic Control Strategies. [Online] Available
http://www.doc.ntu.ac.uk/RTTS/projects/epe/projecthtm, June 28, 1998.

Pignataro, L.J. (1973) Trafi‘ic engineering: Theory and practice. Prentice-Hall,
Englewood Cliffs, NJ.

Rakha, H. et. a1. (1989) Evaluating the benefits and interactions of routes guidance and
traffic control strategies using simulation. VNIS’89, IEEE- VTS- V399, pp 296-
303.

Rathi, AK. and A.J. Santiago (1990) The new NETSIM Simulation program T rafi‘ic
Engineering and Control, pp 317-320.

Road Commission for Oakland County, (1993) F AST-TRAC is on the Cutting Edge Of
Traffic Management Technology. Beverly Hills, Michigan.

Road and Traffic Authority of New South Wales, (1991)SCATS User Manual - V4.7.
Road and Traffic Authority of New South Wales, Sydney, Australial.

166

 

Robertson, DJ. and RD. Breherton (1991) Optimizing networks of traffic signals in real
time - The SCOOT method. IEEE Transactions on Vehicular Technology, Vol 40,
No 1, pp 11-15.

Robertson, GD. (1986) Dual strategy U.T.C. - vehicle actuation under computer control
and SCOOT. Second International Conference on Road Traflic Control, Institute
of Electrical Engineers, Conference Publication NO 260, pp 95-97.

Robertson, GD. (1987) Handling congestion with SCOOT. Trafi‘ic Engineering and
Control, April, pp 228-230.

Ross, D.W., R.C. Sandys, and J .L. Schlaefli (1971) A computer control scheme for
critical-intersection control in a urban network. Transportation Science, Vol 5, No
2, pp 141-160.

Rothenberg, M.J. (1985) Urban congestion in the United States: What does the future
hold? IT E Journal, Vol 55, No 7, pp 22-39.

 

m—AL-
.J
|
1.
\

Saraf, R. (1994) Adaptive Traffic Control using Neural networks”, Ph. D. dissertation,
Department of Civil and Environmental Engineering, Vanderbilt university,
nashville, TN, 1994.

Sarakki, V. and J. Kerr (1994) Effectiveness of IVHS elements on freeway/arterial
capacity-Concept and case studies. IT E Compendium of Technical Papers,
pp 303-307.

Skabardonis, A., R. L. Bertini, and B. R. Gallagher (1997) Development and Application
of Comtrol Strategies for Signilized Intersections in Coordinated Systems.
Proceeding of the 77th Transportation Research Board Annual Meeting,
Washington, D.C., CD-ROM.

Stewart, J. A., K. lepik, and M. Van Aerde (1998) Benefit Sensitivity of Adaptive Traffic
Control Strategies at Isolated Traffic Signals. Proceeding of the 78th
Transportation Research Board Annual Meeting, Washington, D.C., CD-ROM.

Stewart, J. A., and M. Van Aerde (1998) An Assessment of Adptive Co-ordination of
traffic Signal Offset Within Integration. Proceeding of the 78th Transportation
Research Board Annual Meeting, Washington, D.C., CD-ROM.

Taylor, W. C. and Abdel-Rahim, Ahmed (1997) Analysi of Corridor Delay under SCATS
Control- Orchard Lake Raod, Final Report, Submitted to Oakland County Road
Commission.

167

Tranoff, P. J ., and P. S. Parsonson (1981) Selecting Traffic Signal Control at Individual
Intersectiosn. NCHRP Report 233, TRB, National Research Council, Washington,
DC.

Transportation Research Board (1984) Development in traflic signal system.
Transportation Research Circular 282, National Research Council, Washington,
DC.

Van Aerde, M.W. and S. Yagar (1990) Combining traffic management and driver

information in integrated traffic networks. Third International Conference on
Road T rafi‘ic Control, Institute of Electrical Engineers, Conference Publication NO
320, pp 11-15.

Wallace, C. E., K. G. Courage, and M. A. Hadi (1998) Methodology for Optimizing
Signal Timing: TRAN SYT-7H User Guide. University of Florida Transportation
Center, march.

Waugh, M.R., J .E. Clark, and A. Kanaan (1994) Using simulation models to aid in
determining traffic control strategies for IVHS/ATMS signal systems. IT E
Compendium of Technical Papers, Institute of Transportation Engineers,
Washington, DC, pp 470-475.

Witkowski, J. M. (1992) Cost-Effectiveness Analysis for Signilized Intersection
Improvements. Transportation Research Record 1360, TRB, National Research
Council, Washington, D.C., ppl3-22.

Wood, K. (1993) Urban traflic control, systems review. Project Report 41, Transport
Research Laboratory, Crowthome, UK.

Wright, C. and K.W. Huddart (1989) Strategies for urban traffic control. Discussion
Paper 6, Middlesex Polytechnic, November.

168