Development of a traffic-actuated signal timing prediction model

MISSING IMAGE

Material Information

Title:
Development of a traffic-actuated signal timing prediction model
Physical Description:
xiii, 171 leaves : ill. ; 29 cm.
Language:
English
Creator:
Lin, Pei-Sung, 1964-
Publication Date:

Subjects

Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1995.
Bibliography:
Includes bibliographical references (leaves 164-169).
Statement of Responsibility:
by Pei-Sung Lin.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002070305
notis - AKQ8570
oclc - 34360604
System ID:
AA00004713:00001

Full Text









DEVELOPMENT OF A TRAFFIC-ACTUATED
SIGNAL TIMING PREDICTION MODEL











By

PEI-SUNG LIN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENT FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY






UNIVERSITY OF FLORIDA

1995


UNIVERSITY OF FLORIDA LIBRARIES













ACKNOWLEDGMENTS

This dissertation cannot be accomplished without the assistance of many people. I

wish to express my sincerest thanks to my supervisory committee, dear friends and lovely

family who helped make it possible.

First, I would like to express my extreme gratitude to professor Kenneth G. Courage,

chairman of my supervisory committee, for giving me the opportunity to pursue my graduate

studies under his enthusiastic guidance. He not only provided me the financial assistance but

also gave me many levels of support. Every time when I face any difficulty in research, he has

always inspired me with his ingenious idea. His lofty standard has always been a source of

motivation to me. I will never forget what he has done for me during my studies.

Dr. Charles E. Wallace is the Director of Transportation Research Center. Under his

leadership, I felt very warm in my mind when I was in this big family. Although Dr. Wallace

was very busy, he always made time for me. He has provided his professional guidance to

my research and lofty standard to my dissertation. I deeply believe his comments on my

dissertation will be beneficial for my professional career. For this, I am eternally grateful.

Dr. Joseph A Wattleworth served as one of my committee member. Although he has

been retired, his assistance and personal support are sincerely appreciated.

Dr. Sherman X. Bai has been a source of inspiration and motivation. His technical

support and personal caring were really invaluable to the success of this research. He has








helped me far more than being a member of the committee. He has become a good friend of

me. I want to thank him for his professional support and true friendship.

Dr. Mang Tia served as one of my supervisory committee member after Dr. Joseph

A Wattleworth retired. His assistance in this matter is really appreciated. Without his help,

the requirement for my dissertation cannot be completed.

Dr. Anne Wyatt-Brown served as the outside member. She has been of great help

with her assistance on the improvement of my technical writing. In addition, her patient

instruction, sincere encouragement are highly appreciated.

I am indebted to Dr. Gary Long for his guidance and encouragement during my

studies although he is not on my supervisory committee. Special thanks go to William M.

Sampson, manager of McTrans Center, and Janet D. Degner, Manager of Technology

Transfer Center, for providing me financial assistance and the personal supports.

I would also like to express my gratitude to my colleagues for their assistance. I like

to thank Shiow-Min Lin, Yu-Jeh Cheng, Cheng-Tin Gan, Jer-Wei Wu, Min-Tang Li, Chian-

Chi Jiang and Randy Showers for their encouragements. I also like to extend my thanks to

David Allen, David Hale, Jim Harriott and James Kreminski for their proofreading.

Finally, I want to express my deep appreciation to my family. My parents, Chang-

Lang Lin and Li-Chen Hsu, continually supply me their unwavering love, sincere inspiration

and selfless support throughout my life. My brother, Pei-Yi Lin, and my sister, Li-Ling Lin,

continually give me their encouragements and supports. My wife and best friend, Hui-Min

Wen, is willing to share every good time and bad mood with me. Her patience and love give

me the warmest feeling and the best support to complete my research.
















TABLE OF CONTENTS


ACKNOWLEDGEMENTS


LIST OF TABLES

LIST OF FIGURES ......

ABSTRACT ........

CHAPTER 1. INTRODUCTION .

Problem Statement. .
Objectives ..
Organization .

CHAPTER 2. BACKGROUND .

Introduction .
Literature Review ..
Preliminary Model Development
Simulation Models .
Arterial Considerations .

CHAPTER 3. MODEL DEVELOPMENT


Introduction. ...
Determination of Arrival Rates
Permitted Left Turn Phasing
Compound Left Turn Protection
Applications. ...


. 1

S2
4
5


S6


6
6
S.29
S.54
.57


. .63
. 63
. 66
. 76
. .78









CHAPTER 4. MODEL IMPLEMENTATION


Introduction .
Structure and Logic of the ACT3-48 Program .
Extension of the Development of Coordinated Operations .


CHAPTER 5. MODEL TESTING AND EVALUATION .

Introduction .
Fully-actuated Operation .
Coordinated Actuated Operation .
Further Evaluation of the Analytical Model . .

CHAPTER 6. EXTENDED REFINEMENT OF THE ANALYTICAL MODEL

Introduction .
Refinement of the Analytical Model for Volume-density Operation
Refinement of the Analytical Model to Incorporate "Free Queue" Parameter
Incorporation of the Analytical Model into the HCM Chapter 9 Procedure .

CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS .

Conclusions .
Recommendations .


APPENDIX


UNIFORM DELAY FORMULAS .


BIBLIOGRAPHY


BIOGRAPHICAL SKETCH .


.80
.81
.91


.94


.94
.94
105
108

122

122
123
131
143

146

146
149


170















LIST OF TABLES


Table

2-1 The iteration results and its convergence for the illustrated example.

6-1 Through-car equivalents, EL, for permitted left turns in a shared lane
with one free queue. .......

6-2 Through-car equivalents, EL1, for permitted left turns in a shared lane
with two free queues .. ......


Page

. .52


135


136














LIST OF FIGURES


Figure Page

2-1 The operation of an actuated phase under significant demand .35

2-2 Dual-ring concurrent phasing scheme with assigned movements 38

2-3 The relationship among the components in the phase time .39

2-4 Queue accumulation polygon for a single protected phase .. 42

2-5 The intersection used as an example for circular dependency illustration 50

2-6 Queue accumulation polygon in the first iteration of the illustrated example .51

2-7 Iterative loops in the phase time and cycle time computation procedure 53

2-8 Phase time comparison between EVIPAS and NETSIM 56

2-9 Conceptual relationship between major street g/C and minor street demand .58

2-10 The location of studied intersection .. 59

2-11 Prediction of the major street g/C ratio based on a power model for minor
street traffic volume. ....... .61

2-12 Prediction of the major street g/C ratio based on a logarithmic model for minor
street detector occupancy .. 62

3-1 Arrival rate over a full cycle with coordinated operation 64

3-2 Queue accumulation for a single protected phase 67

3-3 Queue accumulation polygon for a permitted left turn from an exclusive lane
with opposing lane number greater than one 69








3-4 Queue accumulation polygon for a permitted left turn from an exclusive lane
with opposing lane number equal to one .. 70

3-5 Queue accumulation polygon for a permitted left turn from an exclusive lane
with sneakers .72

3-6 Queue accumulation polygon for a permitted left turn from a shared lane
(g g ). .75

3-7 Queue accumulation polygon for a permitted left turn from a shared lane
(g
3-8 Queue Accumulation polygon for protected plus permitted LT phasing with
an exclusive LT Lane ........77

3-9 Queue Accumulation polygon for permitted plus protected LT phasing with
an exclusive LT Lane 78

4-1 Major structure of the ACT3-48 program .. 82

4-2 Case 1: Phase sequence for simple permitted turns .. 83

4-3 Case 2: Phase sequence for leading green .. 84

4-4 Case 3: Phase sequence for lagging green .. 84

4-5 Case 4: Phase sequence for leading and lagging green .. 84

4-6 Case 5: Phase sequence for LT phasing with leading green 85

4-7 Case 6: Phase sequence for leading dual left turns 85

4-8 Case 7: Phase sequence for lagging dual left turns .. 85

4-9 Case 8: Phase sequence for leading and lagging with dual left turns .. .86

5-1 Cycle length comparison for a 1.5-sec allowable gap setting 96

5-2 Cycle length comparison for a 3.0-sec allowable gap setting 97

5-3 Cycle length comparison for a 4.5-sec allowable gap setting 97

5-4 Composite cycle length computations with all gap settings .. 98








5-5 Percent of phase terminated by maximum green time for each gap setting .99

5-6 Phase time comparison between the Appendix H method and NETSIM. .101

5-7 Phase time comparison between the proposed model and NETSIM 101

5-8 Intersection configuration of Museum Road and North-south Drive on the
campus of the University of Florida 102

5-9 Phase time comparison between the analytical model and field data 104

5-10 Phase time comparison between NETSIM and field data 104

5-11 Phase time comparison between arterial street and cross street 106

5-12 Relationship of estimated phase times between NONACT and NETSIM 108

5-13 NETSIM arrival distributions for a single-link of 100-ft length 111

5-14 NETSIM arrival distributions for a single-link of 1000-ft length 111

5-15 NETSIM arrival distributions for a single-link of 2000-ft length 112

5-16 NETSIM arrival distributions for a single-link of 3000-ft length 112

5-17 Comparison of analytical and simulation model arrival distribution for
single-lane, 100-vph flow 114

5-18 Comparison of analytical and simulation model arrival distribution for
single-lane, 300-vph flow 114

5-19 Comparison of analytical and simulation model arrival distribution for
single-lane, 500-vph flow 115

5-20 Comparison of analytical and simulation model arrival distribution for
single-lane, 700-vph flow 115

5-21 Comparison of analytical and simulation model arrival distribution for
single-lane, 900-vph flow 116

5-22 Optimal NETSIM single-lane link length for various phase termination
headway settings 116








5-23 Comparison of analytical and simulation model arrival distributions for
two-lane, 200-vph flow 118

5-24 Comparison of analytical and simulation model arrival distributions for
two-lane, 600-vph flow 118

5-25 Comparison of analytical and simulation model arrival distributions for
two-lane, 1000-vph flow 119

5-26 Comparison of analytical and simulation model arrival distributions for
two-lane, 1800-vph flow 119

6-1 Variable initial feature for volume-density operation 124

6-2 Gap reduction feature for volume-density operation 125

6-3 Phase time comparison for volume-density operation with a zero detector
setback 130

6-4 Phase time comparison for volume-density operation with a 150-ft detector
setback 130

6-5 Phase time comparison for volume-density operation with a 300-ft detector
setback 131

6-6 Phase prediction for single shared lane with free queues 142

A-i Uniform delay formula for ingle protected phase 154

A-2 Uniform delay for permitted left turns from an exclusive lane (n >1) 155

A-3 Uniform delay for permitted left turns from an exclusive lane (n,=1) .156

A-4 Uniform delay for permitted left turns from a shared lane (gq>gf) 157

A-5 Uniform delay for permitted left turns from a shared lane (gg,) 158

A-6 Uniform delay for compound left turn protection: HCM Chapter 9 Case 1 .159

A-7 Uniform delay for compound left turn protection: HCM Chapter 9 Case 2 160

A-8 Uniform delay for compound left turn protection: HCM Chapter 9 Case 3 .161








A-9 Uniform delay for compound left turn protection: HCM Chapter 9 Case 4 .162

A-10 Uniform delay for compound left turn protection: HCM Chapter 9 Case 5 163














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

DEVELOPMENT OF A TRAFFIC-ACTUATED
SIGNAL TIMING PREDICTION MODEL

By

Pei-Sung Lin

December, 1995

Chairman: Kenneth G. Courage
Major Department: Civil Engineering

The Highway Capacity Manual (HCM) provides a methodology in Chapter 9 to

estimate the capacity and level of service at a signalized intersection as a function of the traffic

characteristics and the signal timing. At traffic-actuated intersections, the signal timing

changes from cycle to cycle in response to traffic demand. An accurate prediction of average

phase times and cycle length is required to assess the performance of intersections controlled

by traffic-actuated signals. The current technique in Appendix II of HCM Chapter 9 for this

purpose has not been well accepted.

This dissertation describes a more comprehensive methodology and a more

satisfactory analytical model to predict traffic-actuated signal timing for both isolated and

coordinated modes. The proposed methodology and model have been verified by simulation

augmented by limited field studies. The results are very encouraging with respect to their








general reliability and their compatibility with the current HCM Chapter 9 structure. The

techniques developed in this study would provide an important contribution to the

methodology of traffic engineering for traffic-actuated signal timing prediction and improve

the analytical treatment of traffic-actuated control in the HCM Chapter 9.













CHAPTER 1
INTRODUCTION


The concepts of capacity and level of service (LOS) are central to the analysis of a

signalized intersection. Level of service is expressed as a letter grade from A through F that

describes the quality ofperformance of a signalized intersection from the driver's perspective.

It is evaluated based on the average stopped delay per vehicle for various movements within

the intersection. The 1985 Highway Capacity Manual (HCM) [1] prescribes a methodology

in Chapter 9 (Signalized Intersections) to estimate the LOS as a function of the traffic

characteristics and the signal timing.

Intersection traffic control is characterized as "pretimed" if a predetermined timing

plan is repeated cyclically or "traffic-actuated" if the operation varies from cycle to cycle in

response to information from traffic detectors on the roadway. Pretimed control is usually

appropriate for constant traffic demand, while traffic-actuated control is better suited to

variable traffic demand. Pretimed control is much easier to analyze, but traffic-actuated

control offers more in the way of performance to the motorist.

Whether an isolated actuated controlled intersection or a set of coordinated actuated

intersections, the operational performance largely depends on traffic patterns and the actuated

controller parameters to be discussed in this dissertation. A well-designed actuated control

plan that responds appropriately to traffic demand can significantly reduce delay and fuel








2

consumption. More advanced forms of adaptive traffic control strategies were introduced

recently, but the traffic-actuated control concepts still play a very important role today.

Because of its superiority, traffic-actuated control has become the predominant mode

throughout the U.S.A. in spite of its analytical complexity.

Many traffic-related measures at a signalized intersection, such as intersection

capacity, vehicle delay and queue length, are determined by the phase times and cycle length.

For traffic-actuated control, the phase times and corresponding cycle length vary from cycle

to cycle in response to the traffic demand. Therefore, it becomes desirable to predict the

average phase times and cycle length for traffic-actuated control which are the main inputs

in the procedure contained in the HCM Chapter 9 for the computation of intersection capacity

and vehicle delay.

Improvement of the analytical treatment of traffic-actuated control presented in the

HCM Chapter 9 is the subject of this dissertation. An enhanced analytical model will be

proposed and tested.

Problem Statement

Capacity and delay are two major measures of effectiveness for the analysis of a

signalized intersection. The procedure contained in Chapter 9 of the HCM is used almost as

a standard to estimate the intersection capacity and vehicle delay. For traffic-actuated

operation, accurate estimates of intersection capacity and vehicle delay must rely on accurate

estimates of the signal timing. However, the primary technique, presented in an appendix to

HCM Chapter 9, to predict the signal timing for traffic-actuated operation has been the

subject of much criticism in the literature [2, 3, 4, 5, 6].








3
This technique is based on the simple assumption that a traffic-actuated controller will

maintain a high degree of saturation (95% in the HCM Chapter 9 procedure) on the critical

approach to each phase. It does not consider any controller parameter that influences the

signal timing in field. This has created many questions in the literature regarding the validity

of the assumption and the simplistic nature of the technique. Therefore, the main deficiency

of the technique in the appendix to HCM Chapter 9 comes from improper analytical treatment

of traffic-actuated control

Some analytical work has been done on estimating individual phase lengths for traffic-

actuated operation [2, 6, 7, 8, 9, 10], but a method for treating the entire phase sequence,

given a specified set of traffic volumes, controller parameters and detector placements, does

not exist. Simulation is currently the most reliable method for determining the signal timing.

Simulation is a powerful tool, but its application is best suited to situations that do not lend

themselves to analytical treatment. Furthermore, the signal timing prediction based on

simulation is time consuming.

Therefore, the development of a practical analytical technique to predict traffic-

actuated signal timing is very desirable. Such a technique would provide an important

contribution to the methodology of traffic engineering and improve the analytical treatment

oftraffic-actuated control in the HCM Chapter 9.

In this case, there are three major questions with the development of the traffic-

actuated signal timing prediction model The first question is whether the model can improve

the analytical treatment oftraffic-actuated control in the HCM Chapter 9 procedure on signal

timing prediction. The proposed methodology must be in a form that can be incorporated








4
into the HCM Chapter 9 procedure. The second one is whether the model can treat the entire

phase sequence for a specified set of traffic volumes, actuated controller parameters,

intersection configuration and detector placements. The third question is whether the model

can accurately and quickly predict the signal timing for traffic-actuated operation.

Objectives

The signal timing of a traffic-actuated signal will vary from cycle to cycle in response

to traffic demand. The goal of this research is to develop an analytical model to accurately

predict the average signal timing for traffic-actuated intersection for both isolated and

coordinated modes. Although performance measures (delay, stops, queue length, fuel

consumption, etc.) are sensitive to actuated controller parameters, it is necessary to note that

this study does not include the development of an optimization methodology of the actuated

controller parameters. The specific objectives of the research are stated as follows:

1. Review the literature that deals with the subject of traffic-actuated control,

particularly related to signal timing prediction and vehicle delay estimation.

2. Develop a model to improve the analytical treatment of traffic-actuated control

in the HCM Chapter 9 on signal timing prediction and present the methodology

in a form that may be incorporated into the HCM Chapter 9 procedure.

3. Assure that the developed analytical model can predict the average phase times

and corresponding cycle length for a specified set of traffic volumes, controller

settings, intersection configuration and detector placements.

4. Test and evaluate the analytical model using simulation and limited field data to

assess its accuracy and feasibility of implementation.










Organization

This dissertation includes seven chapters that are devoted to developing an analytical

model to predict the signal timing for traffic-actuated control The first chapter provides a

general introduction to the dissertation topic, problem statement and research objectives.

The next two chapters describe the development of the proposed analytical model

Chapter 2 presents the background knowledge required for the analytical model development

and a preliminary model that deals with protected movements from exclusive lanes only.

Chapter 3 extends the methodology developed in Chapter 2 to include shared lanes, permitted

left turns and compound left turn protections (permitted plus protected phasing and protected

plus permitted phasing).

The implementation of the proposed analytical model and procedure in a computer

program to predict traffic-actuated signal timing is addressed in Chapter 4. The

computational process of this program is also presented.

The comparisons of predicted phase times between the proposed analytical model and

simulation and field data are presented in Chapter 5. An intensive evaluation is also made on

the comparison of vehicular arrivals at the stopline produced by simulation and the proposed

analytical model.

Chapter 6 presents several refinements of the proposed analytical model to achieve

a stronger capability on the phase time prediction. Uniform delay formulations developed in

this study for traffic-actuated control are shown in the appendix. The final conclusions and

recommendations are stated in Chapter 7.













CHAPTER 2
BACKGROUND


Introduction

There has been a substantial amount of research conducted on traffic-actuated control

which provides essential information for traffic-actuated operating characteristics and signal

timing prediction. This chapter first reviews past and current research on the traffic-actuated

control followed by the presentation of a preliminary model structure of phase time prediction

for fully-actuated operation. Each traffic control concept or theory that has contributed to

the model development is addressed separately. Next, two simulation models, TRAF-

NETSIM (NETSIM) and EVIPAS, which have been adopted by this study as evaluation tools

for the proposed analytical model, are introduced. Finally, a preliminary consideration of

signal timing prediction for semi-actuated traffic signal coordination is described. Further

model development and implementation for traffic-actuated signal timing prediction will be

mostly based on the background knowledge presented in this chapter.

Literature Review

Traffic-actuated control has been used since in the early 1930s. Whether an isolated

actuated controlled intersection or a set of coordinated actuated intersections, the operational

performance is largely determined by the traffic arrival patterns and actuated controller

parameters. The arrival patterns refers to the arrival headway distributions. The basic








7
actuated controller parameters include the minimum green time, maximum green time and

allowable gap settings (vehicle interval or unit extension). A well-designed actuated control

plan that responds appropriately to traffic demand can significantly reduce delay and fuel

consumption. Therefore, shortly after actuated signal control was first introduced,

researchers began to study the influence of traffic arrival patterns and departure characteristics

at a signalized intersection with traffic-actuated control Many researchers also focused on

the optimization of controller settings, detector placement, and the relationship among them.

Recently, some researchers began to develop models to predict traffic-actuated signal timing

for the purpose of more accurate capacity computation and delay estimation.

Review of past and current research is an area which definitely merits attention. There

has been a substantial amount of research conducted on traffic-actuated control which will

contribute to this dissertation.

The development of a traffic-actuated signal timing prediction model is the subject of

this study. Thus, the major emphasis ofthis literature review is on the traffic-actuated control

that particularly is related to operation characteristics, vehicle arrival headway distributions

and signal timing prediction models. The procedure contained in the HCM Chapter 9 is used

almost as a standard to analyze signalized intersection capacity and level of service, so the

literature review also covers the methodology for both capacity computation and delay

estimation. One item that needs to be reviewed carefully is a new program called EVIPAS.

The EVIPAS model is an optimization program which is able to analyze and determine the

optimal settings of controller parameters for traffic-actuated control The results of the

testing efforts on EVIPAS are report in this chapter.










The main topics of the literature review include

Traffic-actuated control definitions;

Warrants for traffic-actuated control;

Benefits and operating considerations for traffic-actuated control;

Effects of coordination and phase-skipping for traffic-actuated control;

Late-night, low-volume operation of coordinated actuated systems;

Evaluation of traffic-actuated control by simulation;

Prediction of phase times and cycle length for traffic-actuated control;

Delay models for traffic-actuated control;

Signalized intersection capacity models for traffic-actuated control; and

Overview and evaluation of "Enhancement of the Value Iteration Program for

Actuated Signals" (EVIPAS).

Traffic-actuated Control Definitions

Three basic forms oftraffic control: pretimed, fully-actuated and semi-actuated were

mentioned by Orcutt [11] in 1975. He indicated that pretimed control was used primarily in

the Central Business District (CBD) area, especially where a network of signals must be

coordinated. He defined actuated signals in terms of equipment that responds to actual traffic

demand of one or more movements as registered by detectors. If all movements are detected,

the operation is referred to as "fully-actuated." If detectors are installed for some, but not all,

traffic movements, the term "semi-actuated" is applied. Orcutt suggested fully-actuated

control should normally be used at isolated intersections.








9
Precise definitions of the basic controller types were described by the National

Electrical Manufacturers Association (NEMA) standards [12] in 1976. According to the

NEMA standards, the basic controllers include pretimed, semi-actuated, fuly-actuated

without volume-density features, and fully-actuated with volume-density features. In the

remainder of this dissertation, fully-actuated without volume-density features will be just

called fully-actuated control, and fully-actuated with volume-density features will be called

volume-density control

Warrants for Traffic-actuated Control

Warrants for selecting traffic control modes, which are very useful for practicing

engineers, have been researched since the early 1960s. Studies of delay at actuated signals

have been made for the purpose of evaluating warrants for this type of control on the basis

of the information in the 1961 edition of the Manual on Uniform Traffic Control Devices

(MUTCD). This information was expanded by the Texas Department of Highways and

Public Transportation into a graphical format. The graphical relationships were studied in

1971 by Vodrazka, Lee and Haenel [13], who concluded that they provide good guidelines

for selecting actuated equipment for locations where traffic volumes do not warrant pretimed

signals. The current edition of the MUTCD stops short of numerical warrants for choosing

between pretimed and traffic-actuated control, but it does suggest certain qualitative

conditions under which traffic-actuated control should be implemented.

Benefits and Operating Considerations for Traffic-actuated Control

In 1967, Gerlough and Wagner [14] began to compare pretimed control effectiveness

with volume-density control. They found that traffic-actuated control at higher traffic








10

volumes degraded performance. One of the problems cited for volume-density control was

that the duration of green for each phase was dependent on the estimated queue length at the

beginning of the phase. Difficulties with queue length estimation made this type of control

less effective.

Long-loop presence detection operates by producing a vehicle call for the duration

of time that the vehicle is over the detector. This is as opposed to the mode of small-area

detector operation in which the detector outputs a pulse of less than 0.1 seconds when the

vehicle is first detected. This latter mode of operation is known as passage, pulse or count

detection. The long-loop presence detector with fully-actuated controllers in a mode known

as lane-occupancy control or loop-occupancy control (LOC). LOC operation occurs when

the controller is programmed for an initial green interval of zero. Extensions are set either

to zero or to a very low value. There is no need for a non-zero initial interval or minimum

green time because the long loops continuously register the presence of any vehicles that are

waiting, causing the controller to extend the green until the entire queue is discharged. The

result is a signal operation that responds rapidly to changes in traffic flow.

In 1970, Bang and Nilsson [15] compared LOC operation with small area detector

(pulse detector) operation. They concluded that delay was reduced 10 percent and stops by

6 percent under the same traffic conditions with LOC. In 1975, Cribbins and Meyer [16]

compared pulse and presence detectors. They concluded that the longer the length of the

presence detector on the major approach to the intersection, the longer the delay. They also

concluded that the highest intersection travel time values occurred when either long-loop

presence or pulse detectors were used on both major and minor approaches. The intersection








11

travel time was defined here as the average time it takes a vehicle to pass through an

intersection, whether it is stopped or slowed.

Numerous theoretical studies on traffic signal timing were conducted between 1958

and 1970. The theoretical work on pretimed control by Webster [17] in 1958 and Miller [18]

in 1963 has been applied to the computation of optimum cycle lengths as a function of vehicle

arrival rates. It has also been used for evaluating vehicle delay, intersection capacity,

probability of stops and so on. These results were also well validated through the comparison

of field data. In 1969, Newell [19] and Newell and Osuma [20] expanded the body of theory

by developing relationships for mean vehicle delay with both pretimed and actuated control

at intersections of one-way streets and intersections of two-way streets, respectively. Newell

[19] demonstrated that the average delay per vehicle for an actuated signal is less than that

of a pretimed signal by a factor of about three for intersections of one-way streets. Osuma

[20] considered intersections oftwo-way streets without turning vehicles. For the particular

traffic-actuated policy which holds the green until the queue has been discharged, the traffic-

actuated control will not perform as well as pretimed control under the following two

conditions: 1) flows are nearly equal on both approaches of a given phase and 2) the

intersection is nearly saturated.

In 1976, Staunton [21] summarized the work of numerous signal control researchers.

In his paper, the comparisons of delay produced by pretimed control and actuated control,

as a function of vehicle volumes, were presented. Staunton demonstrated that fully-actuated

control with 2.5-sec extensions will always be better than the best form of pretimed operation,

given optimum settings for all volumes. Longer values for the extensions can easily degrade








12
actuated control performance. His conclusions were based on simulation, but the details of

the detector configuration were not specified. In view of the 2.5-sec extension time, short

or passage detectors were probably used in his study. The performance estimates from

Staunton were supported by Bang [22].

In 1981, Tarnoffand Parsonson [23] compiled an extensive literature review on the

selection of the most appropriate form of traffic control for an individual intersection. Three

complementary approaches were used to evaluate controller effectiveness: 1) field data

collection using observers to manually measure vehicle volumes, stops and delay; 2)

simulation using the NETSIM model developed by the Federal Highway Administration

(FHWA) to evaluate control system performance; and 3) analytical techniques developed by

the research team and other agencies. The general conclusions from their extensive literature

review were as follows:

1. Pretimed controllers operate most effectively when the shortest possible cycle

length is used subject to the constraints ofproviding adequate intersection capacity

and minimum green times for pedestrians and vehicle clearance intervals.

2. The delay produced by fully-actuated controllers is extremely sensitive to the value

of the extension that is used. In general, shorter extensions reduce vehicle delay.

3. For small area detectors (motion or pulse detectors), at low and moderate volumes

when extensions of two or three seconds are employed, the use of the filly-

actuated controllers will reduce delays and stops over those which can be achieved

using pretimed controllers. When high traffic volumes occur both on the main

street and on the side street causing the controller to extend the green time to the








13
maximum on all phases, the fully-actuated controller will perform as a pretimed

controller, producing comparable measures of vehicle flow.

4. The relative effectiveness of the various control alternatives depends on the quality

of the signal timing employed. A poorly timed actuated controller will degrade

traffic performance to as great an extent as a poorly timed pretimed controller.

Through the detailed evaluation of controller performance, the conclusions by Taroff

and Parsonson [23] are described as follows: semi-actuated controllers produce a higher level

of stops and delays for all traffic conditions than either the fully-actuated or pretimed

controllers. However, for side street traffic volumes that are less than 20 percent of main

street volumes, there is an insignificant difference between semi-actuated and fully-actuated

controller effectiveness. Fully-actuated controllers produce significant benefits when used in

an eight-phase and dual-ring configuration over that which would be possible with a four-

phase pretimed controller.

From the simulation results on small area detectors for fully-actuated control, the

location of 150 ft produced a level of performance far superior for a 3-sec vehicle interval to

that of closer detectors for the approach speed of 35 mph. Tarnoff and Parsonson concluded

that it is appropriate to locate the detector such that the travel time is equal to the extension

time. It was also concluded that for volumes in excess of 450 vehicles per hour per lane,

additional improvement can be realized through the use ofthe added initial feature of volume-

density controller for an approach of 35 mph or higher.

They indicated that further modest gains in performance for fully-actuated control

were possible with the use of long loops and short (or zero) initial and extension settings.








14
This application was found to produce a performance similar to a 2-sec extension time with

a short loop. From the simulation results, they concluded that LOC was more effective than

pulse detection over a wide range of traffic volumes. LOC offered the further advantage not

reflected by the simulation results of screening out false calls caused by vehicles approaching

but not traveling through the intersection.

Volume-density controllers provide the greatest benefit at intersections with high

approach speeds where a detector setback in excess of 125 ft from the intersection requires

a variable initial green time. Tarnoff and Parsonson found that the variable initial and gap

reduction options of the volume-density control did not improve the controller's performance

over that of a filly-actuated controller unless the option is used to reduce the vehicle

extension to a value that is less than the one used for the filly-actuated controller. Thus, if

the volume-density controller is timed to provide a 3-sec passage time and a 2-sec minimum

allowable gap, its performance will be superior to that of a filly-actuated controller with a

constant 3-sec allowable gap. Noted that the above simulation results given by Tarnoffand

Parsonson do not properly account for the problem of premature termination of green due to

the variation in queue discharge headways that occur under normal operating conditions.

In 1985, Lin [24] studied the optimal timing settings and detector lengths for filly-

actuated signals operating in presence mode using the RAPID simulation model He

suggested the optimal maximum green for hourly flow patterns with a peaking hour factor of

1.0 was about 10 seconds longer than the corresponding optimal greens with a peaking hour

factor of 0.85, and the optimal maximum green was approximately 80 percent longer than the

corresponding optimal greens. This result was similar to the 1.5 times pretimed split








15

suggested by Kell and Fullerton [25]. Lin indicated that optimal vehicle intervals were a

function of detector length and flow rate. For detectors 30,50 and 80 ft long, the use of 2-

sec, 1-sec and 0-sec vehicle intervals can lead to the best signal performance over a wide

range of operating conditions, respectively. The use of vehicle intervals greater than zero

second for detectors 80 ft or longer is not desirable unless the combined critical flow at an

intersection exceeds 1,400 vph.

In order to improve the VIPAS model, a new optimization algorithm and a new

intersection simulation were designed and programmed. The original VIPAS traffic

characteristics and vehicle generation routines were combined with these new models to

create the enhanced version called EVIPAS. In 1987, Bullen, Hummon, Bryer and Nekmat

[26] developed EVIPAS, a computer model for the optimal design ofa traffic-actuated signal

The EVIPAS model was designed to analyze and optimize a wide range of intersection,

phasing, and controller characteristics of an isolated, fdlly-actuated traffic signal It can

evaluate almost any phasing combination available in a two to eight-phase NEMA type

controller and similar phasing structures for a Type 170 controller. The model has been field

tested and validated.

In 1987, Messer and Chang [27] conducted field studies to evaluate four types of

basic fully-actuated signal control systems operating at three diamond interchanges. Two

signal phasing strategies were tested: a) three-phase and b) four-phase with two overlap.

Two small-loop (point) detection patterns (single- and multi-point) were evaluated for each

type ofphasing. They concluded that 1) single-point detection was the most cost-effective

three-phase design; and 2) multi-point detection was the more delay-effective four-phase








16

configuration. Four-phase control characteristically operates a longer cycle length than the

three-phase for a given traffic volume. This feature may produce higher average delays unless

the cycle increase is controlled to the extent that the internal progression features of four-

phase control can overcome this deficiency.

In 1989, Courage and Luh [28] developed guidelines for determining the traffic-

actuated signal control parameters which would produce the optimal operation identified by

SOAP84. They also evaluated the existing signal control parameters on an individual traffic-

actuated signal The significant conclusions are summarized as follows:

Under low volumes, the maximum green settings have little or no effect on the

performance of actuated signal controllers. Under moderate volumes, shorter maximum

greens increase the average delay considerably. Longer maximum greens, however, have no

significant effect on delay. Under high volumes, the maximum green settings become more

important. There is a setting which minimizes average delay. Other settings with longer or

shorter maximum greens will produce more average delays. The optimal maximum green

setting can be achieved by running SOAP under actuated control with an optimal saturation

level set in the BEGIN card. They indicated that the settings that are optimal at some time

may not be appropriate in other times of day.

The value of 0.95, which is the default value used in SOAP actuated control, was

suggested for multi-phase operation and a slightly higher saturation level may be desirable for

two-phase operation. For approaches with a reasonably even distribution of traffic volume

by lane, settings of 4.0, 2.0 and 1.4 seconds were recommended by the study for one, two and

three lanes as the best values for unit extension, respectively.








17

In the same year, Bullen [29] used the EVIPAS simulation and optimization model

to analyze traffic-actuated traffic signals. The variables studied were detector type, detector

placement, minimum green time and vehicle interval. The evaluation criterion was minimum

average vehicle delay. The study showed that the optimum design of a traffic-actuated signal

was specific for some variables but relatively unaffected by others. The design was critical

only for high traffic volumes. At low volumes, vehicle delay is relatively unaffected by the

design parameters studied in his paper. The most critical variable Bullen found was vehicle

interval, particularly for passage detectors, where it should be at least 4.0 seconds regardless

of detector placement and approach speed. This conclusion somewhat contradicts previous

study results. However, it should be noted that the EVIPAS model used by Bullen considered

variable queue discharge headways.

Detector configuration is essential to the success of actuated control Kell and

Fullerton [30] in their second edition ofthe Manual of Traffic Signal Design in 1991 indicated

the small area detector might ideally be located three or four seconds of travel time back from

the intersection, with the allowable gap set accordingly. Similar principles were proposed in

previous research by Tarnoffand Parsonson. Kell and Fullerton also indicated that, in some

states, the detectors setback were determined on the safe stopping distance. The main

purpose is to avoid the dilemma zone in which a vehicle can neither pass through the

intersection nor stop before the stopline. For long loop detectors, they indicated the concept

of loop occupancy can provide good operation when vehicle platoons are well formed. The

use of several smaller loops instead of one long loop was suggested to solve the problem of

random vehicles causing excessive green.








18
In 1993, Bonneson and McCoy [31] proposed a methodology for evaluating traffic

detector designs. They indicated that the safety and efficiency of a traffic detector design can

be determined by the probability ofmax-out and the amount of time spent waiting for gap-out

and the subsequent phase change. The stopline detector and advance loop detector with

presence and pulse mode were discussed, respectively. The methodology presented by

Bonneson and McCoy determined the optimal combination of design elements in terms of

safety (via infrequent max-out) and operations (via a short waiting time for phase change).

The design elements included detector location, detector length, vehicle speed, passage time

settings, and call extension setting). They concluded a large maximum allowable headway

will have an adverse effect on performance by increasing the max-out probability and the

length of wait for phase change.

Effects of Coordination and Phase Skipping for Traffic-actuated Control

In 1986, Jovanis and Gregor [32] studied the coordination of actuated arterial traffic

signal systems. In the past, all optimization methods required that each actuated signal be

converted to its nearest equivalent pretimed unit. Using bandwidth maximization as a starting

point, a new procedure was developed by Jovanis and Gregor that specifically accounts for

actuated timing flexibility. Yield points and force offs at non-critical signals are adjusted so

they just touch the edges of the through-band while critical signals are unmodified. This

method was applied to a data set describing midday traffic conditions on an urban arterial

system of six signals in central Illinois. Simulation was used to evaluate these signal timings

and compare them with corresponding pretimed alternatives. They were surprised to find out

that pretimed, coordinated control appeared superior in general to actuated coordinated








19
control in this experiment. They also concluded that the level of service of side streets was

much more important for pretimed than actuated strategies.

In 1989, Courage and Wallace [33] developed the guidelines for implementing

computerized timing designs from computer programs such as PASSER I, TRANSYT-7F

and AAP in arterial traffic control systems. The coordination of a group oftraffic-actuated

signals must be provided by some form of supervision which is synchronized to a background

cycle length with splits and offsets superimposed. Both external and internal coordination of

the local controllers were addressed.

This report focused on the external coordination of traffic signal controllers.

Permissive periods were introduced to indicate the time interval following the yield point

during which the controller is allowed to yield to cross street demand. If the computed splits

are longer than the minimum phase times, it might be possible to establish a permissive period

without further sacrifice or compromise on the rest of the sequence. The methodology of

computing permissive periods was introduced. The effect of phase-skipping due to lack of

traffic demand was also presented. The Timing Implementation Method for Actuated

Coordinated Systems (TIMACS) program was developed to perform the computations of

permissive periods.

Since most previous studies were more specific to certain geometric and phasing

combinations, the qualitative and quantitative evaluation methodology for coordinated

actuated control needed to be fully investigated. In 1994, Chang and Koothrappally [34]

designed a field study to demonstrate the operational effectiveness of using coordinated,

actuated control They concluded 1) there was significant improvement, based on both delay








20

and number of stops, between the semi-actuated control, fully-actuated, and pretimed

coordinated timing during the study; 2) there were no significant differences in performance

among all the semi-actuated operations as long as the progression-based signal coordination

timing was developed correctly; and 3) the use of longer background cycle lengths generally

caused fewer arterial stops. However, it would generate much higher overall system delays.

Late-night, Low-volume Operation of Coordinated Actuated Systems

Coordinating the timing of adjacent signals to promote progressive traffic movement

was recognized as one of the most effective means for reducing vehicular stops, delay, fuel

consumption and exhaust emissions. Early efforts on the subject of signal control always

indicated the need to interconnect signals into a single system and to work toward maimizing

progressive movement during peak periods.

In 1990, Luh and Courage [35] evaluated the late-night traffic signal control strategies

for arterial systems. They stated that late-night, low-volume arterial signal control involved

a trade-off between the motorists on the artery and those on the cross street. The

conventional measures of effectiveness such as stops, delay, and fuel consumption were not

appropriate for evaluating this trade-off Luh and Courage proposed a methodology to

choose between coordination and free operation on arterial roadways controlled by semi-

actuated signals when traffic is light. The choice was made on the basis of a disutility function

that was a combination of the number of stops on the artery and the average cross-street

waiting time. The results indicated that this method provides a promising tool for late-night

arterial signal control










Evaluation of Traffic-actuated Control by Simulation

Simulation modeling has become an extremely important approach to analyzing

complex systems. After 1980, more and more simulation modeling was used in traffic

operations. In 1984, Lin and Percy [36] investigated the interactions between queuing

vehicles and detectors for actuated controls, which govern the initiation, extension, and

termination of a green duration. They emphasized that a model used in the simulation analysis

should be calibrated in terms of observed characteristics such as queue discharge headway,

arrival headway, the relationship between the arrival time of a queuing vehicle and the

departure time of its leading vehicle, the number of queuing vehicles in a defined area at the

onset of a green duration, and the dwell time of a vehicle on the detection area. They also

indicated, under a presence control, the chance for premature termination of a green duration

increases when detector lengths are shortened and a detector length of longer than 80 ft can

effectively eliminate the premature termination. Using long detectors, however, results in

longer dwell times and may reduce control efficiency.

Lin and Shen [37] also indicated that the modeling of the vehicle-detector interactions

should take into account the stochastic aspects of queuing in relation to detectors. The use

of average characteristics of departure headway could result in underestimates or

overestimates of the probabilities of premature termination of the green.

Later, Lin [38] evaluated the queue dissipation simulation models for analysis of

presence-mode fully-actuated signal control The queue dissipation models used in the

NETSIM program and the VIPAS program were evaluated. He indicated that both models

were capable of producing realistic departures of queuing vehicles from the detector area.








22

The models were rather weak, however, in representing other aspects of vehicle-detector

interactions. A major weakness of the model in NETSIM was that the simulated movements

of queuing vehicles have little to do with the discharge times generated separately from a

probability distribution. The weakness of VIPAS was that the Pitt car-following model used

in VIPAS did not provide a flexible model structure for calibration. Therefore, the outputs

of the model could not be made to conform easily and simultaneously with observed

departure, arrival and dwell characteristics of queuing vehicles.

In 1988, Chang and Williams [39] investigated the assumption that independent

vehicle arrivals at traffic signals, such as in the Poisson distribution, have been widely used

for modeling delay at urban intersections. The study introduced an effective yet economic

approach to estimate the degree of correlation among arriving vehicles under given conditions

and geometric characteristics. With the proposed technique, traffic professionals can easily

determine if the existing delay formulas and other traffic simulation models based on the

Poisson distribution are applicable.

The presence of high variability in traffic simulation results often leads to concern

about their reliability, and consequently precludes a rigorous evaluation of the target traffic

system's performance under various control strategies. In 1990, Chang and Kanaan [40]

presented the variability assessment for NETSIM. The batch-means method, which allows

the user to assess the variability of parameters, such as the average delay per vehicle, through

a single relative long run, was introduced. This study provided a good contribution to traffic

simulation users, given the large expenditures on computer simulation.










Prediction of Phase Times and Cycle Length for Traffic-actuated Control

Traffic-related phenomena at a signalized intersection, such as lane capacity, delay and

queue length are influenced by the green times (or phase times) and cycle length. For traffic-

actuated control, green split and cycle length fluctuate with respect to the traffic demand.

Consequently, it becomes desirable to predict the average phases times and cycle length. The

phase time is equal to the green time displayed plus intergreen time (clearance interval or

duration of yellow plus all-red).

In 1982, Lin [7] began to develop a model to estimate the average phase duration for

fully-actuated signals. The model was developed primarily on the basis of probabilistic

interactions between traffic flows and the control He assumed the arrival at the upstream

side of an intersection would be at random, so the arrival pattern in each lane was represented

by a Poisson distribution. Later, Lin and Mazdeyasna [8] developed delay models for semi-

actuated and fully-actuated controls that employ motion detectors and sequential phasing.

These models were based on a modified version of Webster's formula. The modifications

included the use of average cycle length, average green time, and two coefficients of

sensitivity reflecting the degree of delay sensitivity to a given combination of traffic and

control conditions. In 1992, Lin [2] proposed an improved method for estimating average

cycle lengths and green intervals for semi-actuated signal operations as mentioned before.

In 1994, Ak9elik [9, 10] proposed an analytical methodology for the estimation of

green times and cycle length for traffic-actuated signals based on the bunched exponential

distribution of arrival headway. The discussions in his papers were limited to the operation

of a basic actuated controller that used passage detectors and a fixed allowable gap setting.








24

Both fully-actuated and semi-actuated control cases were studied. A discussion of arrival

headway distributions was presented since the estimation of arrival headways is fundamental

to the modeling of actuated signal timings. The formulae were derived to estimate the green

times and cycle length based on the bunched exponential distribution of arrival headway.

The bunched arrival model was proposed by Cowan [41] and used extensively by

Troutbeck [42, 43, 44, 45, 46, 47] for estimating capacity and performance of roundabouts

and other unsignalized intersections. The bunched arrival model considers that the bunched

relationship between vehicles increases when the flow arrival rate increases. Since the

bunched arrival model appears to be more representative of real-life arrival patterns in

general, Akgelik used this arrival model for deriving various formulae for the analysis of

traffic-actuated signal operations. The random arrival model which uses negative exponential

or shift negative exponential distribution of arrival headway can be derived as a special case

of the bunched arrival model through simplifying assumptions about bunching characteristics

of the arrival stream. The methods given in his papers provide essential information (average

green times and cycle length) for predicting the performance characteristics (capacity, degree

of saturation, delay, queue length and stopped rate) of intersections.

Delay Models for Traffic-actuated Control

With the increase of computer software, the comparison of different traffic programs

for pretimed and actuated controls became intuitively appealing. In 1974, Nemeth and

Mekemson [48] compared the delay and fuel consumption between deterministic Signal

Operations Analysis Package (SOAP79) and the microscope and stochastic NETSIM

simulation for pretimed and actuated controls. They indicated that in terms of delay








25

prediction, SOAP79 and NETSIM were found to be entirely compatible except for the

difference in delay definitions.

In 1988, Akcelik [49] evaluated the 1985 HCM [1] delay formula for signalized

intersections. He stated that the HCM formula predicted higher delays for oversaturated

conditions. An alternative equation to the HCM formula was proposed. This formula gave

values close to the HCM formula for degrees of saturation less than 1.0, and at the same time,

was similar to the Australian, Canadian and TRANSYT formulas in producing a delay curve

asymptotic to the deterministic delay line for a degree of saturation greater than 1.0.

The signalized intersection methodology presented in the 1985 HCM [1] introduced

a new delay model Lin [3] evaluated the delay estimated by the HCM with field observed

delay in 1989. Some inconsistency existed in the delay estimation between the HCM results

and field observation. He suggested improving the progression adjustment in the HCM

procedure and using a reliable method to estimate average cycle lengths and green durations

for traffic-actuated signal operations.

In 1989, Hagen and Courage [50] compared the HCM [1] delay computations with

those performed by the SOAP84 and TRANSYT-7F Release 5. The paper focused on the

effect of the degree of saturation, the peak-hour factor, the period length on delay

computations and the treatment of left turns opposed by oncoming traffic. They indicated

that all of the models agreed closely at volume level below the saturation point. When

conditions became oversaturated, the models diverged; however, they could be made to agree

by the proper choice of parameters. The computed saturation flow rates for left turns

opposed by oncoming traffic also agreed closely. However, the treatment of protected plus








26

permitted left turns produced substantial differences. It was concluded that neither SOAP nor

HCM treats this case adequately.

A delay model was recommended in the HCM [1] for level-of-service analysis at

signalized intersections. The use of this model for the evaluation of traffic-actuated signal

operations required the knowledge of the average green times and cycle length associated

with the signal operation being analyzed. Since the method suggested in the HCM to estimate

delay of traffic-actuated signal operations was not reliable, Lin [2] proposed an improved

method for estimating average green times and cycle length in 1990. The method was

appropriate for semi-actuated signal operations. Lin stated that the method was sufficiently

simple and reliable. Realistic examples were used to illustrate the application of the method.

In 1993, Li, Rouphail and Akcelik [4] presented an approach for estimating overflow

delays for lane groups under traffic-actuated control using the 1985 HCM [1] delay model

format. The signal timing used in the delay model was from a cycle-by-cycle simulation

model This study was limited, however, to two-phase single-lane conditions. The results

indicated that the signal timings are much related to the controller settings, with longer

extension times producing higher cycle length. It was found that overflow delay increases

with longer extension times. Further, by applying the 1985 HCM delay formula to the

simulated signal settings, the resultant delays were much higher. This implies the need for

calibration of the second delay term to account for the actuated control effects.

Signalized Intersection Capacity Models for Traffic-actuated Control

Intersection capacity analysis is essential for measurements of most traffic control

effectiveness. The first U.S. Highway Capacity Manual (HCM) in 1950 contained a chapter








27

for estimating the capacities of signalized intersections. Numerous studies were undertaken

to evaluate the different aspects of signalized intersections, and many capacity methods were

developed. In 1983, May, Gedizlioglu and Tai [51] began the evaluation of eight available

methods for capacity and traffic-performance analysis at signalized intersections including

pretimed and actuated controls. The eight methods included the U.S. Highway Capacity

Manual method (1965), British method (1966), Swedish method (1977), Transportation

Research Board (TRB) Circular 212 planning method (1980), TRB Circular 212 operations

and design method (1980), Australian method (1981), National Cooperative Highway

Research Program (NCHRP) planning method (1982), and NCHRP operations method

(1982). They concluded that the NCHRP operations method and the Australian method were

found to be the most cost-effective.

In 1991, Prevedouros [5] studied the traffic measurements and capacity analysis for

actuated signal operations. He verified that the methodology in Chapter 9 of the 1985 HCM

[1] was not appropriate to treat the pretimed and actuated controls identically, especially

concerning the estimation of capacity and performance of existing intersections. The main

sources of error and their potential impacts were presented. He developed a comprehensive

data collection and analysis methodology to complement the procedure in the 1985 HCM.

Overview and Evaluation of EVIPAS

EVIPAS [52] is an optimization and simulation model for actuated, isolated

intersections. It is capable of analyzing and determining the optimal settings of controller

parameters for a wide range of geometric configurations, detector layouts, and almost any

phasing pattern available in a single or dual-ring NEMA and Type 170 controllers. It will








28

generate the optimized timing settings for controllers ranging from pretimed to volume-

density actuated controllers. The optimum settings of timing parameters include minimum

green time, maximum green time, unit extension, minimum gap, time before reduction, time

to reduce, variable (added) initial and maximum initial for each phase.

The value of optimal timing is defined as the timing setting which results in the

minimum "total cost." The model allows the user to define "total cost: to include a variety

of measures of effectiveness, such as delay, fuel consumption, depreciation, other vehicle

costs and emissions. The EVIPAS model allows for two modes of operation. In its

optimization mode, the model is used to obtain optimal timing settings by a multivariate

gradient search optimization module and an event-based intersection microscopic simulation.

In the simulation mode, EVIPAS allows the evaluation of a prespecified signal plan just by

microscopic simulation.

For the capacity and level of service of traffic-actuated control, the performance

outputs are primarily concerned, which includes the summary of delays and signal

performance. The summary of delay table provides delay statistics for the intersection and

for each approach and lane. The summary of signal performance table shows the average

phase length and average cycle length. All above delay measures, average phase length, and

average cycle length are based on the microscopic simulation results.

Since both EVIPAS and TRAF-NETSIM are microscopic simulation models, the

phase time comparison between these two models becomes necessary. The phase times

estimated from the EVIPAS simulation model will be compared with those from NETSIM

simulation model later in this chapter.










Preliminary Model Development

The purpose of this study is to accurately predict the average phase times and

corresponding cycle length for actuated operations. The preliminary model developed in this

study is limited to through movements and left turns with "protected only" phasing from an

exclusive lane. However, this preliminary model is very useful for later development of a

complete and comprehensive model.

It is important to note that the proposed preliminary model is mainly based on the

methodology proposed by Courage and Akcelik [6] for evaluating the operation of a traffic-

actuated controller in their working paper NCHRP 3-48-1 for National Cooperative Highway

Research Program (NCHRP) project 3-48, "Capacity Analysis of Traffic-actuated Signals."

The proposed analytical model for predicting average phase times and corresponding

cycle length applies several traffic engineering concepts and theories. They include traffic-

actuated operation logic, dual-ring control concept, average phase time prediction for traffic-

actuated signals, queue accumulation polygon (QAP) concept, vehicle arrival headway

distribution, circular dependency relationship and sequential process. These concepts and

theories are used for both preliminary and comprehensive model development, and their

model implementation.

In the discussion of the preliminary model, the method in the Appendix I to HCM

Chapter 9 will initially be reviewed following the model development issues. Then, each

concept and theory used in the model development will be presented. Finally, the

computational framework proposed by Courage and Akelik [6] for modeling traffic-actuated

controller operations will be summarized.










Review of the Appendix II Method to HCM Chapter 9

In the HCM [1] Chapter 9 Appendix I methodology, an actuated signal is assumed

to be extremely efficient in its use ofthe available green time. Thus, the average cycle length

is estimated using a high critical volume over capacity ratio (v/c) which is approximately equal

to 0.95. In other words, the controller can be effective in its objective of keeping the critical

approach nearly saturated. The formula for the average cycle length may be stated as

C = L/(1-Y/X,) (2-1)

where

C, = the average cycle length;

L = the total lost time per cycle, ie., the sum of the lost times associated with the

starting and stopping of each critical lane group in the phase sequence;

Y = the critical flow ratio, determined as the sum of the flow ratios (v/s) for the

individual lane groups that are critical in each phase. The flow ratio for each

lane group is defined as the ratio of the traffic volume (v) to the saturation

flow rate (s); and

X, = the target degree of saturation (volume/capacity ratio or v/c ratio). A value

of 0.95 is suggested in Appendix II for traffic-actuated control

After the average cycle length has been computed, the average effective green time

(g) for each lane group I can be determined by dividing the average cycle length (C,) among

lane groups in proportion to their individual flow ratios ( (v/s) ) over target degree of

saturation (X,). The formula of average effective green time for each lane group may be

stated as








31

= C, [ (v/s) / X ] (2-2)

The effective green time, rather than the signal displayed green time, is usually used in signal

timing computation which is the signal displayed green time plus the intergreen time (change

interval or yellow plus all-red clearance) minus the lost time in the phase.

As mentioned before, the Appendix II method for estimating the signal timing for

actuated operation has been questioned in the literature. There are three major problems with

the Appendix I methodology:

1. The assumption that a traffic-actuated controller will maintain 95% saturation on

the critical approach to each phase has not well been accepted. Several studies

have indicated that a somewhat lower degree of saturation often results.

2. The effects of actuated design parameters such as minimum green time, maximum

green time, unit extension and detector configuration are not reflected in the

formula for average cycle length, so it is not sensitive to the above parameters.

3. The simplistic nature of this model does not provide for real-world complications

such as minimum or maximum green time setting, shared-lane permitted left turns,

left turns that are allowed to proceed on both permitted and protected phases,

phase skipping due to lack of demand, constraints imposed by coordination, etc.

In their working paper, Courage and Akgelik [6] indicated that the limitations of

Appendix I technique can be overcome, but not without adding considerable complexity to

the computational procedures. The HCM has traditionally dealt with "single pass" analytical

models that may be described in manual worksheets. The updated version (1994) of HCM

Chapter 9 worksheets are analytically much more complicated, however, they have retained,








32
with one minor exception, their "single pass" characteristics. Therefore, Courage and Ak9elik

proposed a model with a sequential process of multiple iterations to improve the model

addressed in Appendix II with "single pass."

Model Development Issues

Since the entire Chapter 9 methodology has reached the limits of single-pass

procedures, the limitations of Appendix II, as mentioned before, cannot be addressed without

resorting to complex iterative procedures. Because the limitation of the Appendix II

technique is a result of the primitive treatment of actuated control, it may only be overcome

by improving the actuated control model Thus, the model to be developed in this study must

be able to perform effective comparisons between the pretimed and traffic-actuated control

modes. The model must also be functionally capable of providing reasonable estimate of

operating characteristics (timing and performance measures) of traffic-actuated controllers

under the normal range of practical design configurations. It must be sensitive to common

variations in design parameters. The design parameters include

Actuated controller settings (minimum green time, maximum green time and

allowable gap);

Conventional actuated vs. volume-density control strategies;

Detector configurations (length and setback);

Pedestrian timings (Walk and Flashing Don't Walk, FDW);

Left turn treatments (permitted, protected, permitted and protected, and not

opposed); and

Left turn phase positions (leading and lagging).








33
Additional input data are needed to improve the accuracy of the analysis methodology.

The information that is already required by the Chapter 9 procedure will naturally be used to

the fullest extent possible to avoid the need for new data. Most of the additional data items

are related to the operation itself The proposed model will be based on the standard eight-

phase dual-ring control concept that is more or less universally applied in the U. S. A. In this

study, a standard assignment of movements to phase is adopted. It can greatly simplify the

development and illustration of all modeling procedures without affecting the generality of

the capacity and level of service results.

It is difficult to analytically deal with the very low volume operation that typically

occurs late at night. Under this condition, the repetitive cyclical operation upon which the

analysis is based no longer applies. The effort required to develop a model for dealing

accurately with delays of a few seconds per vehicle (i.e., level of service A) is difficult to

justify. An approximation of the operating characteristics for very low volumes will generally

be acceptable from a capacity and level of service perspective.

In the literature, many analytical studies on traffic-actuated operation assume that

passage detectors are used, whereas in actual practice, presence detectors are much more

common. Passage detectors transmit a short pulse to the controller upon the arrival of each

vehicle. Presence detectors transmit a continuous signal to the controller as long as the

vehicle remains in the detection zone. For purpose of this study, variable length presence

detectors will be assumed. The operation of using passage detectors to detect vehicles may

be approximated by using short length presence detectors.










Traffic-actuated Operation Logic

Actuated operation is one kind of traffic control which uses the information collected

by detectors to determine the signal timing of an intersection. The detector type can be either

passage or presence. The main advantage of traffic-actuated control is that the traffic signal

can properly display the green times according to traffic demand. There are three types of

actuated controllers. They are semi-actuated, fully-actuated and volume-density.

The operation of the semi-actuated signal is based on the ability of the controller to

vary the length of the different phases to meet the demand on the minor approach. Maximum

and minirmm green times are set only for the minor street. Detectors are also placed only on

the minor street. On the other hand, fully-actuated controllers are suitable for an intersection

at which large fluctuations of traffic volumes exist on all approaches during the day.

Maximum and minimum green times are set for each approach. Detectors are also installed

on each approach. The volume-density control is one kind of actuated control with added

features which 1) can keep track of the number of arrivals, and 2) reduce the allowable gap

according to several rules. It is usually used at intersections with high speed approaches. For

illustration purposes on actuated control logic, a passage detector will first be assumed

because it is simple. Some basic term definitions are addressed as follows:

Initial interval is the first portion of the green phase that an actuated controller has

timed out for vehicles waiting between the detector and stopline during the green

time to go through the intersection.

Vehicle interval, also called "unit extension" or "allowable gap" is the time that the

green time is extended for each detector actuation.










Maximum green time is simply the total green time allowed to the phase.

Minimum green time is the shortest green time that can be displayed.

To avoid vehicles being trapped between the detector and stopline, it is necessary that the

vehicle interval be at least the "passage time" of a vehicle from the detector to the stopline.




Maximum Green Time

Minimum Green Ti Extension Period

trial Int.Ie. It.









Time *
t---------^

Detector actuation on phase with right-of-way
SUnexpired portions of vehicle intervals



Figure 2-1. The operation of an actuated phase under significant demand.



Figure 2-1 shows the operation of an actuated phase under significant demand. Prior

to the beginning of the figure, a "call" for green had been put in by the arrival on the studied

approach. Then, the phase with right-of-way on this approach will first display the initial

interval plus one unit extension for the arrival The sum of initial interval and one unit

extension is usually called minimum green time. During the minimum green time, if an








36
additional vehicle arrives, as shown in Figure 2-1, a new unit extension is begun from the time

of detector actuation. The unexpired portion of the old vehicle interval with the shaded area

shown in the figure is wiped out and superseded. If vehicle actuation continues, the green

time will also be extended with the same process until the maximum green time is reached.

The total extension time after the minimum green time is referred to as the extension period.

If the traffic volume is less intense, the extension period will not reach the maximum

green time. When a vehicle interval expires without an arrival of a new vehicle (indicated by

an asterisk [*]), the green time will be terminated and the signal light will turn to yellow plus

red clearance if there is a vehicle waiting for the next subsequent phase. Since in this

illustrated example significant demand is assumed, the maximum green time is reached.

Dual-ring Control Concept

In a pretimed controller, the controller operates under a single ring sequential timing

process. Each phase is taken as an interval of time in which specified traffic movements are

serviced. Of course, it is possible to have a given movement served on more than one phase

of the sequence, and it is expected that a combination of two non-conflicting movements (two

through, two left turns or a left turn plus through) will be serviced on any one phase.

However, the vast majority of modem traffic control systems use NEMA standard

traffic-actuated controllers which employ a dual-ring concurrent timing process. By keeping

the non-conflicting phases in separate rings, it is capable of displaying them simultaneously

to optimize the combinations of movements which are displayed on each cycle. Since the

standard eight-phase dual-ring operation is more or less universally applied in the U. S. A.,

and in this study a standard assignment oftraffic movements is also based on the dual-ring








37

NEMA phase configuration, it is essential to be familiar with the dual-ring concurrent phasing

scheme with assigned movements. The dual-ring phasing scheme and operation logic will

be presented next.

The standard dual-ring concurrent phasing scheme using NEMA phase definition is

shown in Figure 2-2. In Figure 2-2 phases 1, 2, 3 and 4 are belong to ring 1, while phases 5,

6, 7 and 8 are belong to ring 2. A specific traffic movement is assigned to each NEMA phase

as shown at the corer of each phase box. For example, NEMA phase 2 is an eastbound

through movement and NEMA phase 7 is the southbound left turn. In a standard dual-ring

concurrent phasing scheme, east-west movements are assigned to the left side of barrier

(phases 1, 2, 5, 6), whereas north-south movements are assigned to the right side of barrier

(phases 3, 4, 7, 8). The barrier can be reversed to assign north-south movements to the left

side and east-west movements to the right side if needed.

Traffic movements for phases 1 and 2 conflict with each other. It is also true for

phases 3 & 4, 5 & 6, and 7 & 8. Since the conflicting phases on each ring are sequential, on

the side of the barrier, none of the phases within ring 1 will conflict with any of the phases

within ring 2. In such a way, non-conflicting phases can be displayed simultaneously to

optimize the combinations of movements.

An example of the dual-ring concurrent phasing scheme based on the east-west

movement in Figure 2-2 is presented as follows. The phase sequence begins with the non-

conflicting combination ofNEMA phases 1 and 5. The next phase sequence can be either the

combination of NEMA phases 1 and 6 or NEMA phases 2 and 5 according to the traffic

demand. If the demand for eastbound left turns is heavier than that of westbound left turns,








38
in general, the combination of phases 2 and 5 will display most of the time. Finally, NEMA

phases 2 and 6 will display.

It is a standard convention to assign the odd number to the left turn in any phase pair

(1-2, 3-4, 5-6, 7-8), and the even number to the through movement. This reflects the

popularity of leading left turn protection. When the lagging left turn protection is to be

implemented, the phase assignment may be reversed (ie., even number to the left turn).

Although any phase may be theoretically designated as the coordinated phase in each ring, it

is common to designate the phase with through movement on the left side of the barrier as

the coordinated phase. It is necessary to note that the above phasing assignments conform

to those used by the PASSER II arterial signal timing. It has also been adopted by the

WHICH program for mapping data into NETSIM.


Barrier


Ring 1






Ring 2


Left Side of Barrier
( E-W Movements )


Right Side of Barrier
( N-S Movements)


Barrier


Figure 2-2. Dual-ring concurrent phasing scheme with assigned movements.










Average Phase Time Prediction for Traffic-actuated Signals

The main objective of this study is to accurately predict the average phase times for

traffic-actuated signals. The average phase time includes two major portions of timing. One

is the queue service (clearance) time and the other is the extension time after queue service.

Therefore, accurate phase time predictions are mainly dependent on the accurate predictions

of both queue service times and extension times after queue service.

Before the illustration of the methodology to predict average phase times and cycle

length, some key term definitions need to be addressed first. The phase time is the signal

displayed green time (controller green time) plus the intergreen time. In the signal timing

analysis, the effective green time and the effective red time are frequently used. Therefore,

appropriate conversion of the displayed green time to an effective value is required before the

signal timing analysis.





SR G I



r g
Sgqst e
Phase Time -


Figure 2-3. The relationships among the components in the phase time.








40

The relationship among phase time (PT), displayed green time (G) and effective green

time (g) shown in Figure 2-3 is expressed as follows:

PT = G+I = t, +g = t, +gst+g, (2-3)

subject to PTm < PT < PT,=

In more detail, Equation 2-3 can be expressed as follows:

PT = t, +Go+EI = t= t+gqt+eg+I = t +g + s+ + t (2-4)

subject to PT, < PT < PT,

where

I = intergreen time (yellow plus all-red);

ti = lost time, which is the sum of start-up lost time, t., and end lost time, td;

gq,, G, = the queue service time (saturated portion of green), where gq, = G= ;

e., E. = the green extension time by gap change after queue service, where eg =

EI, and the total extension time, EXT, is defined as (ei+I) or (Eg+I);

ge = the effective extension time by gap change after the queue service period,

where g = e + I td; (2-5)

PTn, = the minimum phase time, PTm, = G. + I, where G. is minimum green

time; and

PT. = the maximum phase time, PT,= G + I, where G. is maximum green

time.

Queue Accumulation Poygon (OAP) Concept

The analysis of queue accumulation polygon (QAP) is an effective way to predict the

queue service time, gq, (= G.). The QAP is a plot of the number of vehicles queued at the








41
stopline over the cycle. For a single protected phase which could be the through phase or the

protected left turn phase, when traffic volume does not exceed its capacity, QAP is just a

single triangle as shown in Figure 2-4. In Figure 2-4, g, stands for actual queue service time

(in this case, gq~ = g), while g, is the effective extension time after queue service. More

complex polygons occur when a movement proceeds on more than one phase.

Based on the vehicle arrival rate q, during effective red time, the accumulated queue

(Qr) before the effective green time can be estimated. The time taken to discharge the

accumulated queue can be computed simply by dividing the accumulated queue of Qr with the

net departure rate (s q) which is equal to the departure rate (s) minus the vehicle arrival rate

(q) during the effective green time. For this simple protected phase, the departure rate s is

equal to the saturation flow rate. The target v/c ratio may be considered in the peak hour

analysis. However, it must be set to 1.0 to determine the actual queue service time. In order

to determine the critical queue service time (gq) of different lane groups within the same

phase, a lane utilization factor is considered in the computation of actual queue service time

(g). In general, g, can be estimated from the following formula:


g = fu q (2-6)
s-(qIXT)


where

fq = a queue length calibration factor [7] proposed by Ak9elik to allow for

variations in queue service time, where

= 1.08- 0.1 (G/ G )2 (2-7)









42

f = a lane utilization factor for unbalance lane usage based on the HCM Table

9-4;

q, qs = q, is red arrival rate and q. (veh/sec) is green arrival rate (veh/sec);

XT = a specified target volume/capacity (v/c) ratio; and

r, s = r is red time (sec) and s is saturation flow (veh/sec).

In multi-lane cases, the saturated portion of green time should represent the time to

clear the queue in the critical lane (i.e. the longest queue for any lane) considering all lanes

of an approach in the signal phase. More complex polygons occur when a movement

proceeds on more than one phase. The computation for queue service time is mainly based

on the QAP concept.










o
0



0
oa
.a/


Time (seconds)


Figure 2-4. Queue accumulation polygon for a single protected phase.










Vehicle Arrival Headway Distributions

Arrival headway distributions play a fundamental role in the estimation of green

extension time, eg (or E), at actuated signals. The bunched exponential distribution of arrival

headways was proposed by Cowan [36], which considers that the bunched relationship

increases among the arriving vehicles when the traffic volume increases. The free

(unbunched) vehicles are those with headways greater than the minimum headway (A), and

the proportion of free vehicles (p) represents the unbunched vehicles with randomly

distributed headways. Thus, the measurement of the proportion of free vehicles (p) depends

on the choice of minimum headway (A). The proportion of bunched vehicles in the arrival

stream is (1-p). In this arrival model, all bunched vehicles are assumed to have the same

intra-bunched headway (A). The cumulative distribution function, F(t), for this bunched

negative exponential distribution of arrival headways, representing the probability of a

headway less than t seconds, is


F(t) = 1 ype- -A) for t >A (2-8)

= 0 fort

where

A = minimum arrival (intra-bunch) headway (seconds);


(p = proportion of free (unbunched) vehicles; and










X = a parameter calculated as


SIAq (2-9)

subject to q < 0.98/A

where

q = total equivalent through arrival flow (vehicles/second) for all

lane groups that actuate the phase under consideration.

A detailed discussion of the application of this model on actuated control and the

results of its calibration using real-life data for single-lane and multi-lane cases are given in

Akelik and Chung [53]. The more commonly used simple negative exponential and shifted

negative exponential models of arrival headways are special cases of the bunched exponential

model Therefore, in this study, the bunched arrival model is used to estimate the extension

time after queue clearance.

The method for estimating the green extension time, e, for an actuated controller that

uses a passage detector and a fixed gap time (unit extension) setting (e.) was described by

Akqelik [34]. In this study, presence detectors are assumed. The headway (h) between two

consecutive vehicles is equal to the sum of the gap time (e) and the detector occupancy time

(to). Therefore, the headway, hk, that corresponds to the allowable gap time setting, e,, is

h= e. + to (2-10)

where to is the detector occupancy time given by

to = (Ld +L )/v (2-11)


where










Ld = effective detector length (ft);

L = vehicle length (ft); and

v = vehicle speed (ft/sec).

There is no need for the estimation of an extension time if the actual queue service

time, g, is less than the minimum effective green time, gn, or g, is greater than the maximum

effective green time, g Ifg, < g< then g willbe set to gan and ifg &> gn then g is equal

to g,. Detection of each additional vehicle between gnn and g in general, extends the

green period by an amount that is, in effect, equal to the headway time, h.. The green period

terminates when the following two conditions are satisfied.

1. the headway between two successive vehicle actuations exceeds the headway that

corresponds to the gap time setting, h > ho (gap change); or

2. the total green extension time after the expiration of minimum green time equals

the maximum extension setting. It is equivalent that g is equal to g..

During a gap change, the green period terminates after the expiration of the gap time.

Assuming that the termination time at gap change is the headway corresponding to the gap

time setting (ho = e. + to), the green extension time, eg, by gap change can be estimated from

the following formula [6, 9, 10] based on the bunched exponential headway distribution.


e 1
eg= pqq (2-12)



Once the green extension time, e, is obtained, the effective extension time, g&, is just equal

to the sum ofeg and the intergreen, I, minus the end lost time, td, as shown in Equation 2-5.








46
As mentioned before, the more commonly used simple negative exponential and

shifted negative exponential models of arrival headways are special cases of the bunched

exponential model. For simple negative exponential model, use


A= 0 and (p =1


(therefore X = q)


and for the shifted negative exponential model (normally used for single-lane traffic only), use

P = 1 (therefore X = q / (1- Aq)) (2-14)

These two models unrealistically assume no bunching ((p =1) for all levels of arrival flows.

The bunched model can be used either with a known (measured) value of (p, or more

generally, with a value of


relationship suggested by Akqelik and Chung [53] can be used for estimating the proportion

ofunbunched vehicles in the traffic stream ((p)


p = e -b (2-15)


The recommended parameter values based on the calibration of the bunched

model by Akqelik and Chung [53] are:


exponential


Single-lane case:


A = 1.5 seconds and b = 0.6


Multi-lane case (number of lanes =2): A = 0.5 seconds and b = 0.5

Multi-lane case (number of lanes > 2): A = 0.5 seconds and b = 0.8


(2-16a)

(2-16b)

(2-16c)


(2-13)










Effect of Phase Skip

The minimum phase time requires more attention when the phase may be skipped due

to low traffic volume. The minimum phase time would only be valid if the controller was set

to recall each phase to the minimum time regardless of demand. On the other hand, the real

significance of the minimum phase time in an actuated controller is that a phase must be

displayed for the minimum time unless it is skipped due to lack of demand. This situation may

be addressed analytically by determining the probability of zero arrivals on the previous red

phase. Assuming a bunched arrival headway distribution, this may be computed by using the

following equation:


POV = p e- A) (2-17)


where

POv = probability of zero arrivals during the previous red phase; and

R = previous red phase time.

So, assuming that the phase will be displayed for the minimum time, except when no vehicles

have arrived on the red, the adjusted vehicle minimum time then becomes

AVM = MnV(1 Po) (2-18)

where

AVM = the adjusted vehicle minimum time; and

MnV = the nominal minimum vehicle time.

The similar concept for adjusted vehicle minimum time may also be applied to compute an

adjusted pedestrian minimum time.








48
If the phase may be skipped due to lack of demand, the adjusted minimum phase time

is the maximum of adjusted vehicle minimum time and adjusted pedestrian minimum time.

It will be used as lower bound of the predicted phase time. When the Pov for a phase is not

zero, the estimated phase time must also be modified by multiplying the (1-Pov) factor to the

original total extension time. Therefore, the predicted phase time becomes

PT = tj +Gqs +(1 Pov) (Eg +I) (2-19)

Circular Dependency Relationship and Sequential Process

The determination of required green time using the Appendix II method is relatively

straight forward when the cycle length is given. However, traffic-actuated controllers do not

work on this principle. Instead, they determine, by a mechanical analogy, the required green

time or phase time given only the length of the previous red interval The green time or phase

time required for each phase is dependent on the green time or phase time required by the

other phases. Thus, a circular dependency relationship exists between actuated phase times.

There are two way to resolve this type of circular dependency. The first one is

simultaneous solutions of multiple equations. The second one is a sequential process

involving repeated iterations that converge toward a unique solution. Either method could

be applied to solve this dependency problem. Since the simultaneous solution will not lend

itself to the complications that must be introduced to solve the more general problem,

Courage and Akgelik [6] proposed and set up an iterative procedure that will apply to the

general problem. This iterative procedure is adopted in this study to predict the average

phase time for general cases, not just limited to the protected phases.








49
An initial set of values for all phase times must be established before the iterative

procedure may begin. With each iteration, the phase time required by each phase, given all

of the other phase times, may be determined. If the minimum phase times turn out to be

adequate for all phases, the cycle length will simply be the sum of the minimum phase times

of the critical phases. If a particular phase demands more than its minimum time, then a

longer red time will be imposed on all the other phases. This, in turn, will increase the phase

time required for the subject phase. Through a series of repeated iterations, the circular

dependency will come to an equilibrium and converge to a unique solution. When the

convergence of cycle length is reached, the final cycle length and phase times are determined.

This convergence may be demonstrated easily by using an simple example. Consider

the intersection shown in Figure 2-5. This is a trivial intersection with four identical single-

lane approaches carrying the through volume of 400 vph. A saturation flow rate of 1900

vphgpl is assumed. Each phase has assigned the following constant parameters:

Detector: 30 feet long, placed at the stopline

Intergreen time (I): 4 seconds

Lost time (t,) 3 seconds per phase

Start-up lost time (td): 2 seconds per phase

Minimum phase time: 15 seconds

Allowable gap: 3 seconds

Maximum phase time: 50 seconds

No pedestrian timing features.

No volume-density features.












N FHE1-NS SB










t t



HE1-NS NB

Figure 2-5. The intersection used as an example for circular dependency illustration.



The QAP for the first iteration in this example is shown in Figure 2-6. Initially, the

"trial time" is the nominal minimum vehicle time (15 seconds in this case) for each phase.

Although this operation has four through phases, it can actually be treated as a two-phase

operation because the northbound through phase is identical to the southbound through phase

and the eastbound through phase is the same as the westbound through phase.

Hence, the trial cycle length is equal to 15 x 2 = 30 seconds. This initial timing would

result in an effective red time of 18 seconds for each phase. The traffic volume for each

approach is 400 vehicles per hour. In other words, the arrival rate is equal to 400/3600 =

0.11 vehicles per second. Therefore, during the effective red time, the accumulated queue

can be computed as the product of arrival rate and effective red time, which is equal to 0.11

x 18 = 2 vehicles.














Required phase time
o 3 ----------i
13+5.16+8.3 16.46 see I

S13 15.16 sec 8.3 see
g Isecl I
.11 x 18 2.00 veh I
2


P 0.11 vps 0153 0.11
o* 1 -- 0'42 vps

z 15 + 3 18 sec
effective red r tl gg ge

0 10 20 30
Time (seconds)


Figure 2-6. Queue accumulation polygon in the first iteration of the illustrated example.




The departure rate is the saturation flow rate (1900 vehicles per hour), so the phase

can discharge (1900/3600)= 0.53 vehicles per second. Therefore, the net service rate is equal

to (0.53 0.11) = 0.42 vehicle per second. Since f is about 1.07 and f, equals 1.0, the actual

queue service time, g, taken to discharge the queue will be (1.07 2) / 0.42 = 5.16 seconds.

The extended green time, e., is 5.3 seconds and intergreen time, I, is 4 seconds. Thus, the

total extension time, EXT (= es + I), is 9.3 seconds (= 5.3 + 4) and the effective extension

time, g equals 8.3 seconds (= 5.3 + 4 1). By using Equation 2-3, the phase time, PT,

equals the sum of the lost time, tz, the queue service time, gq,, (= g, in this example) and the

effective extension time, g& which is about 16.46 seconds (= 3 + 5.16 + 8.3). By using

Equation 2-4, PT is equal to the sum oft,, gq,, (= g in this example) and EXT which is also

about 16.46 seconds (= 2 + 5.16 + 9.3). The new phase times produce a new cycle length








52
of (16.46 x 2) = 32.92 seconds. This will generate another version of Figure 2-6 with

different dimensions. By repeating these calculations with a new cycle length each time, the

computed cycle length will converge to within 0.1 second. Convergence for this example is

especially rapid. The process is very reliable. Table 2-1 shows the iteration result and

convergence for this trivial example.



Table 2-1. The iteration results and its convergence for the illustrated example.

Itera- Cycle Old Acc. Total Total New New Differ-
tion phase queue service time ext. time phase cycle ence
time (= t + g) (= e + I) time
(sec) (sec) (sec) (sec) (sec) (sec) (sec)
1 30.0 15.0 2.00 7.16 9.3 16.5 32.9 2.9
2 32.9 16.5 2.16 7.57 9.3 16.9 33.7 0.8
3 33.7 16.9 2.21 7.68 9.3 17.0 33.9 0.2
4 33.9 17.0 2.22 7.71 9.3 17.0 34.0 0.1


Computational Process

The computation framework with five worksheets proposed by Courage and Ak9elik

[6] for modeling traffic-actuated controller operations for simple through and protected left

turn phases is adopted by this study as the basis for more complicated and general model

development. This computation process will be introduced here. The worksheets play a very

important part in overcoming the "black box" image of a complex model such as the one in

this study. They provide a structure for presenting the results of intermediate computations

in a common form that is compatible with their proposed techniques.









53

Worksheet 1 is "Traffic-actuated Control Data Input". "Lane Group Data" is shown

in Worksheet 2. Worksheet 3 is "Traffic-actuated Timing Computation" and Worksheet 4

is "Required Phase Times". The last worksheet, worksheet 5, is "Extension Times Based on

Allowed Gaps". The worksheet format offered a clear and concise way to document the

information. This format is also consistent with the current HCM. While the worksheets

themselves are quite simple, the overall procedure contains iterative loops. In this research,

the worksheets proposed by Courage and Akgelik will be modified and enhanced for the

computation of more general and complicated scenarios, not just limited to simple protected

scenarios. The complete procedure involving the five worksheets is illustrated in Figure 2-7.

This figure shows the five worksheets, the main information flow path and two iterative

loops indicated as "Loop A" and "Loop B".



1. Data 2. Lane






A Adjustments SYMBOLS
S^..* Main
Information
Flow

** Worksheets
5. Phase W
Extension n -
Times : Iterative Loops

Fu s
Figure 2-7. Iterative loops in the phase time and cycle *..... computation procedure.


Figure 2-7. Iterative loops in the phase time and cycle time computation procedure.








54
Loop A. Required Time Cycle Time Adjustment: This is an external iteration

between Worksheets 3 and 4. It is required to make the phase times converge

to a stable cycle length. Worksheet 4 must also refer to Worksheet 5 if phase

time extensions are required to compute the required phase times.

Loop B. Phase Extension Time: This is an internal iteration within Worksheet 5.

It is only required when gap reduction is employed. When the allowable gap is a

function of the phase time, the phase time cannot be computed without iteration.

Simulation Models

Simulation is one of the most powerful analysis tools available to those responsible

for the design and operation of complex process and systems. Simulation might have more

credibility because its behavior has been compared to that of a real system, or because it has

required fewer simplifying assumptions and thereby has captured more of the true

characteristics of the real system.

NETSIM, a popular and powerful microscopic traffic simulation model, has been

continually developed by the Federal Highway Administration (FHWA) for many years.

NETSIM is able to model an eight-phase, dual-ring controller explicitly, recognizing all of the

phase-specific parameters. EVIPAS is an optimization and simulation model for actuated

controlled, isolated intersections. In the simulation mode, it is also capable of providing

simulated phase times for a wide range of actuated parameter settings. Thus, both the

NETSIM and EVIPAS simulation models could be used as tools to verify the phase time

estimation from the proposed analytical model










Comparison between NETSIM and EVIPAS Simulation Results

Since both NETSIM and EVIPAS are simulation-based models, it becomes necessary

to compare the simulation results between EVIPAS and NETSIM. The current version of

NETSIM (P version 5.0) produces very detailed tables of several performance measures. It

does not, however, provide sufficient information on the operation of the controller itself in

the standard output tables. To obtain this information, it was necessary to develop a

postprocessor to extract the operational data from special files used to support the animated

graphics features ofNETSIM.

The actuated-controller data for each second of operation are recorded and stored in

a text file that is given a file name with extension of ".F45" by NETSIM. The format of

".F45" files is hard to read. After it is properly converted, a readable text file can be produced

with extension of".X45". Then, a postprocessor was developed to read the .X45 file and

produce a summary ofthe operation. Both conversion and postprocessor were combined into

a program called "NETCOP" for "NETSIM Controller Operation Postprocessor." It provides

phase-specific information such as percent skipped, percent gapout, percent maxout, average

cycle length, average phase time, adjusted cycle length and adjusted average phase time. The

"adjusted cycle length" is computed by subtracting the number of seconds of dwell (ie., the

time during which no demand was registered on any phase) from the total number of seconds

simulated before dividing by the number of cycles. The adjusted phase time is computed

according to the adjusted cycle length. Since the adjusted phase time from NETSIM can

represent the effective use of phase time, it is adopted for later phase time comparison.









56

Although the simulation techniques used in EVIPAS and NETSIM may differ in some

degree, theoretically, the phase time estimates for the same traffic conditions, geometric

configurations and actuated timing settings should be close. Thus, an evaluation has been

made by comparing the simulated phase times from both NETSIM and EVIPAS based on 9

hypothetical examples with traffic-actuated operations. These examples cover both two-

phase and multi-phase actuated operations. The comparison result is shown in the Figure 2-8.

In a simple regression analysis between the above two simulated phase times, a 0.96

coefficient of determination, R2, was achieved. As expected, the simulated phase times from

EVIPAS are very close to those from NETSIM simulation, which demonstrates that the

EVIPAS model has the similar effectiveness on phase time estimation as NETSIM.





70


I R2 = 0.96





S50 2



S.......... ...............NETSIM S..... simulated Pha--- -----e Ti --e ( --ec) --
< u a


E-M S----ate---)---- ----------------
0 10 20 30 40 50 60 70
NETSIM Simulated Phase Time (aec)


Figure 2-8. Phase time comparison between EVIPAS and NETSIM.










Arterial Considerations

As urban roadways become more congested, and resources available for building new

facilities become more limited, transportation professionals are exploring all possible

alternatives to improve the existing transportation systems. Fully-actuated traffic signals are

powerful for isolated intersections, but not proper for coordinated intersections. One major

area that holds great potential in reducing urban congestion is the implementation of

coordinated semi-actuated traffic signals on arterial streets. Unlike pretimed signals, semi-

actuated signals are intrinsically more intelligent and complex to implement and they provide

a better coordination than pretimed signals [34].

In semi-actuated operation, detectors are placed only on the minor street and

exclusive left turn lanes on the major street. No detectors are installed for the through

movements on the major street. Under this operation, once there is no detection on the minor

movements, the green will always come back to the through (coordinated) movements on the

major street. Therefore, the major benefit of using semi-actuated control is to assign unused

minor street green to the major street. This concept is fundamental to the signal timing

prediction for coordinated semi-actuated operation.

In pretimed control, the phase time of each movement is fixed, so the effective green

over cycle length ratio (g/C) for each movement is a constant. Unlike pretimed signals, the

phase time of each lane group for actuated signals does not stay constant but fluctuates from

cycle to cycle, so the g/C ratio for each movement fluctuates. Therefore, the major difference

between pretimed and traffic-actuated intersections lies in the g/C. The g/C ratio is important

because it is required for the capacity and delay computation. It is necessary to note that the








58

coordination for actuated signals must be provided by some form of supervision which is

synchronized to a background cycle length. Since coordinated semi-actuated control is

frequently used at intersections along an artery, the g/C ratio of the critical through-lane

group on the major street is especially important.

The true g/C ratio of each movement is difficult to access in current practice directly

from a traffic controller. Therefore, an alternative for obtaining the g/C ratio is needed. It

is found, in general, that the g/C ratio on a major street is high when the demand on minor

streets is low, and vice versa. This observation suggests the potential for using minor street

traffic demand measures to predict the major street g/C ratio. To verify the concept of the

assignment of unused minor street green to the major street, attempts have been made to

quantify the relationship between the major street g/C ratio and minor street traffic demand.






100
S0 Low Medium High

0





'a
0C


Minor Street Demand, x


Figure 2-9. Conceptual relationship between major street g/C and minor street demand.








59
Since traffic volume and detector occupancy are easy to measure with reasonable

accuracy for minor street demand, a conceptual model focusing on these two variables was

developed to predict the major street g/C ratio. As the minor street demand, x, increases

from zero, the major street g/C ratio, y, should start as a sharply decreasing function. The

rate of decrease should attenuate as x increases. The conceptual relationship between y and

x is shown in Figure 2-9.

Data Acquisition

Data from a closed loop signal system were used for the calibration of the candidate

models. The data set includes 98 intersection records generated from four coordinated semi-

actuated intersections during seven time periods along SW 34th St. in Gainesville, Florida.

A layout of the arterial system is shown in Figure 2-10. The SW 34th St. artery is a major

street, and Radio Rd., Hull Rd., SW 20th Ave. and Windmeadows are four minor streets.

Protected left turn signals exist on the major street at all four intersections.


N
1 Radio Rd.

2 Hull Rd.
SW 20th Ave. 3
Archer Rd.
Windmeadows



SW 34th St.


Figure 2-10. The location of four studied intersection.








60

The Traffic Actuated Controller Monitor/Analyzer (TACMAN) computer package

[54] uses information collected by a microcomputer-based control system, the Signalized

Intersection Monitor (SIMON) [54], to produce hourly phase-specific descriptive information

and performance measures. The descriptive information includes traffic volume, detector

occupancy, etc., and performance measures include stops, delay and fuel consumption. The

measures used in the study pertain to the critical movements.

Model Calibration

Statistical Analysis System (SAS) [55] programs were applied to perform correlation

and regression analyses. Regression techniques were used to calibrate the parameters of the

conceptual models. Regression results show that either the minor street volume or detector

occupancy can explain most of the variation in the major street g/C. Furthermore, the F and

t values indicate that both candidate models are overall significant and the coefficients in each

model are also individually statistically significant. The power model (volume model)

produced the highest adjusted R2 using the traffic volume data, while the logarithmic model

(occupancy model) was better for detector occupancy data. These two models can be

expressed as follows:

MAJG = 100(1 + 0.01 VOLUME) -3045 Adj. R2= 0.88 (2-20)

MAJG = 100 12.0954 LN(1 + OCCUPANCY) Adj. R2 = 0.91 (2-21)

where

MAJG = g/C ratio (%) for the major street critical through movement;

VOLUME = hourly volume for the minor street critical movement; and

OCCUPANCY = hourly detector occupancy for the minor street critical movement.








61

The curves for these two minor street demand measures are shown in Figures 2-11

and 2-12, respectively. As expected, the shape of both curves conforms to the described

conceptual model Although the occupancy model might be slightly better than the volume

model, the volume model may frequently be used because minor street volumes are often

known, estimated or forecast when no information is available about detector occupancy.

Based on the concept that the unused minor street green time is assigned to the major

street in coordinated semi-actuated operation, the relationship between the g/C ratio of the

critical through-lane group on a major street (major street g/C ratio) and the traffic demand

of the critical movement on minor streets (minor street demand) was quantified by the above

volume model and occupancy model In the further study, this concept will continue to be

implemented in the phase time estimation for coordinated semi-actuated operation.





S100

o 90
0 80

S70'

S60

50 Ii -- ,
0 100 200 300 400 500
Minor Street Traffic Volume (vph)




Figure 2-11. Prediction of the major street g/C ratio based on a power model for
minor street traffic volume.

























0 5 10 15 20 25 30
Minor Street Detector Occupancy (%)




Figure 2-12. Prediction of the major street g/C ratio based on a logarithmic model for
minor street detector occupancy.













CHAPTER 3
MODEL DEVELOPMENT


Introduction

The scope of the preliminary model development presented in chapter 2 was limited

to basic through movements and protected left turn movements from an exclusive lane. This

chapter continually explores the analytical basis for extensions of the preliminary methodology

to cover permitted left turns in both shared and exclusive lanes and the complicated

compound left turn protection (Le., protected plus permitted phasing or permitted plus

protected phasing).

Determination of Arrival Rates

In the previous analytical work, the arrival and departure rates were constant

parameters determined externally. The arrival rates were determined by the specified traffic

volumes and the departure rates were determined by the saturation flow rates. Neither

depended on the signal timing. In fully-actuated operation which is very often used at an

isolated intersection, the progression effect is not considered, therefore, the arrival type 3

(green arrival rate = red arrival rate) is appropriate. For coordinated semi-actuated operation,

since the progression effect is an important consideration, the use of different arrival types to

represent progression quality is required. The derivation of arrival rate is presented first in

this chapter for later application on coordinated semi-actuated intersections.









64

With arrival type 3, the arrival rate is constant over the whole cycle at q veh/sec. With

other arrival types, two different arrival rates must be computed. q, is the arrival rate on the

green phase, and q, is the arrival rate on the red phase. For a given average arrival rate, the

values of q and q, will depend on the platoon ratio, Rp, associated with the arrival type, and

the green ratio, g/C, which is a part of the timing plan. The HCM defines the platoon ratio

as follows:

Rp = PC/g (3-1)

where

P = the probability of arrival on the green;

g = the green time for the phase; and

C = the cycle length.






Basic Continuity Relationship: qr + qgg = q(r+g)


o

o q = average arrival rate

r g
A. r--------------
(t:___ O____


Time (seconds)


Figure 3-1. Arrival rate over a full cycle with coordinated operation.









65
From Equation 3-1,

P = R g/C

Now, with arrival type 3, R = 1, so P = g/C, and q == q = On the other hand, if Rp 1,

then the arrival rates will be different on the red and green phases, as illustrated in Figure 3-1.

The problem is to determine q and q given q, Rp, g and C.

From Equation 3-1:


P(r+g)
g


But, by definition,


p gqg
q(r+g)


Sgqg(r+g)
q(r+g)g


R q
q


Therefore,


qg = qR


(3-2)


As an extension of this derivation, the red arrival rate may be determined from the continuity

relationship shown on Figure 3-1.











qrr + qg = q(r+g)

From which,


qrr = q(r+g) (qR)g


Therefore,


q(r+g) (qR)g (3-3)
r

Permitted Left Turn Phasing

Before beginning the discussion of permitted left turn (LT) phasing, the concept of

green time determination for a protected phase is briefly reviewed first. This concept is based

on the queue accumulation polygon (QAP) shown in Figure 3-2 which was presented earlier.

It is convenient to be shown here again for the illustration. The QAP shows the number of

vehicles accumulated in a queue on a signalized approach over one cycle of operation. Each

cycle is assumed to repeat the same pattern indefinitely. The number of vehicles accumulated

at any time in the cycle may be determined as the difference between the cumulative arrivals

and departures since the start of the cycle.

The queue accumulation increases throughout the red phase at the rate ofq, veh/sec.

The maximum accumulated queue occurs at the end of the red phase, and is indicated on

Figure 3-2 as Q. During the green phase, the queue decreases at the net departure rate, (s -

q), until it has been fully serviced. The time required to service the actual queue is indicated

on Figure 3-2 is g. The green phase will continue until the occurrence of a gap in the arriving








67

traffic of sufficient length to cause the controller to terminate the phase. The effective

extension time is indicated on Figure 3-2 as g,. Of course, this whole process is subject to

a specified maximum phase time. An analytical model for predicting g,, g& and average phase

time, PT, was presented in the preliminary methodology. In this simple case, the queue

accumulation polygon is just a triangle. In the remainder of the cases to be discussed in this

dissertation, the QAP will assume a more complex shape.













r
!* Qr









Time (seconds)

Figure 3-2. Queue accumulation polygon for a single protected phase.



Permitted Left Turns from Exclusive Lanes

The basic QAP concept may be extended to cover a slightly difficult case in which

a permitted left turn is from an exclusive lane, yielding to oncoming traffic, instead of a

protected movement. This introduces a couple of important changes in the QAP. Since the

number of opposing lanes, np, may influence the net arrival rate during the period when the

opposing queue is being serviced, it will be considered for the permitted left turns form
opposing queue is being serviced, it will be considered for the permitted left turns form








68
exclusive lanes. Based on the number of opposing lanes, the QAPs for condition 1 (nW, > 1)

and condition 2 (np = 1) shown in Figures 3-3 and 3-4 are discussed, respectively.

If the number of opposing lanes is greater than one (see Figure 3-3), the queue

continues to accumulate throughout the first part of the green with the arrival rate of q, while

the opposing queue is being serviced. The time required to service the opposing queue is

indicated on Figure 3-3 as gq. There is no chance for a permitted left turner to make a

maneuver during the period ofgq. The maximum queue, indicated on Figure 3-3 as Qq, can

be computes as follows:

Qq Q +qg *gq (3-4)

If the number of opposing lane is equal to one (see Figure 3-4), the left turns from the

opposing lane do create the chance for the left turns from the exclusive left turn lane to make

maneuvers. According to the HCM Chapter 9, the adjusted saturation flow rate, s,, during

the period ofgq can be computed using the following formula:

sq = / EL2 (3-5)

where

sq = the permitted saturation flow rate during the period of g;

s = the protected saturation flow rate; and

E, = the left turn equivalence as determined from Equation 9-22 in the HCM.

the queue continues to accumulate throughout the period of g with the net green arrival rate

of (q sq) while the opposing queue is being serviced. The maximum queue, indicated on

Figure 3-4 as Qq, can be computes as follows:

Qq = Q + (q s,) *gq (3-6)









69

Thereafter, the LT vehicles will filter through the opposing traffic at a rate determined

by the opposing volume. This is indicated on Figures 3-3 and 3-4 as the permitted saturation

flow rate, Sp. The net departure rate is shown as the difference between the permitted

saturation flow rate and the green arrival rate, which is equal to (s,- q.). Fortunately, the

HCM Chapter 9 worksheets already provide the means to compute these values. The value

of gq is determined explicitly on the supplemental worksheet for permitted left turns. The

value of permitted saturation flow rate, s,, may be determined as

s, = s / EL (3-7)

where

s, = the permitted saturation flow rate;

s = the protected saturation flow rate; and

ELI = the left turn equivalence as determined from Figure 9-7 in the HCM.





0

4 Qq
4)qg

SQ
S.S- q8

r


Time (seconds)

Figure 3-3. Queue accumulation polygon for a permitted left turn from an exclusive
lane with opposing lane number greater than one.
















o q
.B Qr q sq5

I I s-
s 1 i q8
o qr Iq

Sr

Time (seconds)


Figure 3-4. Queue accumulation polygon for a permitted left turn from an exclusive
lane with opposing lane number equal to one.


Green Time Extension for Permitted Movements

The model for estimating the effective green time extension, g, assumes that the

arrivals after the queue has been serviced will be free-flowing as they cross the detector. This

will not be the case for permitted left turns. A complex stochastic model would be required

to treat this situation in detail

It should, however, be possible to use the left turn equivalence, ELI, described above

as an approximation in this situation. In other words, the equivalent through volume, VL EL,

would be used in place of the actual left turn volume, VL The green extension time must be

determined using an equivalent through volume of(VT + VLELI) in place of the actual volume

of (VT + V).










Effect of Sneakers

Sneakers are permitted LT vehicles that exit the intersection at the end of the green

phase, usually during the intergreen interval It is common to assume that a maximum of two

vehicles per cycle may be released from the queue. Sneakers are treated implicitly in the

Chapter 9 worksheets by imposing a lower limit of two vehicles per cycle on the capacity of

each exclusive LT lane with permitted movements. For purposes of this analysis, sneakers

must be recognized explicitly in the QAP. This requires the definition of some new terms:

Maximum Sneakers, Sm: The maximum number of LT vehicles released at the end

of the green phase assuming that the LT queue has not already been serviced.

Permitted phase terminal queue, Qp: The number of vehicles accumulated at the

end of the permitted phase before sneakers have been released.

Adjusted permitted phase terminal queue, Qp: The number of vehicles accumulated

at the end of the permitted phase after sneakers have been released.

Actual Sneakers, Sa: The actual number of sneakers released at the end of the green

phase. This is determined as Min (Sm, Qp).

For the purpose of illustration, the effect of sneakers on the QAP for the condition,

np > 1, is shown in Figure 3-5. This illustrates the case in which the phase is terminated by

the maximum green time before the queue of LT vehicles is completely serviced. If Qp is

greater than zero, then the maximum phase length will be displayed as a pretimed equivalent.

The adjusted permitted phase terminal queue, Qp', is equal to Qp minus Sa. If Qp is greater

than zero, then the v/c ratio for the approach will be exceeded. These parameters will be

involved in a more complex way in the analysis of compound left turn protection.















i Qq

Sqgq
I Qr ... / Sp-qg

Sr Qp

||-- S
Time (seconds)


Figure 3-5. Queue accumulation polygon for a permitted left turn from an exclusive
lane with sneakers.


Multi-Lane Approaches

IfLT vehicles are accommodated in an exclusive lane, it is reasonable to assume that

there will be additional lanes available to handle the through traffic. It is also expected that

the detectors installed in both the through and left turn lanes will activate and extend the same

phase. Under these conditions, the queue service time, g,, for the through and LT lanes will

be different. The required phase time (RPT) is the sum of lost time (t), effective extension

time (g = g + I t, ) and the maximum value of through queue service time, g, (= g), and

left turn queue service time, gq,(= g + g,):

RPT = Max ( Through gq, Left turn g,) + t, + g (3-8)

and eg must be determined using an equivalent through volume of(VT + VLEL) in place of the

actual volume of(VT + VL).










Shared Lane Permitted Left Turns

The shared lane permitted left turn case is only slightly more complicated than the

exclusive lane permitted left turn case. The additional complication may be seen in the QAPs

shown in Figures 3-6 and 3-7. The concept of free green, g,& must be introduced here. In an

exclusive left turn lane, the first vehicle in the queue will always be an LT vehicle. However,

in a shared lane, the first vehicle could be either a through or LT vehicle. The portion of

effective green blocked by the clearance of an opposing queue is designated as gq. During the

time, an LT vehicle may be stopped by the opposing traffic, but a through vehicle will not.

Until the first left-turning vehicle arrives, however, the shared lane is unaffected by left

turners. The free green represents the average green time from the beginning of green that

will be available to move through vehicles in the shared lane. The Chapter 9 supplemental left

turn worksheets provide a method for computing gq and gf As indicated on Figures 3-6 and

3-7, there will be a net discharge rate of(s q) during the free green interval, and the queue

remaining at the end of the interval is represented as Q,

The computation for queue service time, gq, is more complicated for a shared lane

than that for an exclusive left turn lane or through lane. Basically, the gq for a shared lane

can be divided into two parts. One is the green time before the beginning of the actual queue

service time (g&), and the other one is g, itself

In Figure 3-6, since g, is greater or equal to g,, the gq, is just equal to the sum of g

and g, Note that according to the HCM Chapter 9, Q, can be computed based on the number

of the opposing lanes. The detail description for the computation of Qq is in the HCM

Chapter 9. Figure 3-7 represents the scenario that gq is less than gf. The accumulated queue,








74

Q, which belongs to through vehicles, will be served at a net service rate of(s qa). Assume

the time to clean all of these accumulated through vehicles is designated as &. Ifgf is less than

g, then gqg, will equal the sum ofgf and g&. On the other hand, ifgr is greater or equal to go

then g just equals g, because g, is zero for this condition. The time to clear all accumulated

through vehicles, g, may be represented by Equation 2-6. The computation for gq, may be

summarized as the following equations. If g is greater than or equal to g, the queue service

time, gq can be computed as

g& = gq + gs (3-9)

Ifgq is less than g, the formula for the computation of the queue service time, g, becomes

gst = gf + g when g,<&g (3-10)

gqst = gt when gf&g& (3-11)

The remainder of Figures 3-6 and 3-7 follows the same process as the exclusive lane

cases shown in Figure 3-3 and 3-4. The permitted movement saturation flow rate for the

shared lane must, however, be computed somewhat differently. In an exclusive LT lane, the

left turn equivalence, EL, was applied to all of the vehicles in the LT lane. In a shared lane,

it is only appropriate to apply this factor to the LT vehicles. So, s, should be computed as


s
sp 1 + PL(EL 1) (3-12)



where

PL = the proportion of left turns in the shared lane, as computed by the supplemental

worksheet for permitted left turns, and all other terms are as defined previously.









75

eg should be computed in the same manner as described for the exclusive LT lane case and

using an equivalent through volume of (V, + V, EI) instead of the actual volume.




o
0
3




.0 I
o qr Qf
z 1
1 s I P-qg
Z7, ^


Time (seconds)


Figure 3-6. Queue accumulation polygon for a permitted left turn from a shared lane
(gq> g).






Qr
0 0

.9 -s-qg

i 1

qr ff
Ssp-qg


Time (seconds)


Figure 3-7. Queue accumulation polygon for a permitted left turn from a shared lane
(gq <&.










Compound Left Turn Protection

The QAP concept may be extended to cover the case in which an LT movement

proceeds on both permitted and protected phases from an exclusive lane. One important

difference between the simple permitted LT phasing and the compound protected LT phasing

is the assignment of detectors to phases. It is assumed that detectors will be installed in all

LT lanes to ensure that LT vehicles will not face a permanent red signal. The discussion of

simple permitted LT phasing assumes that the detector in the LT lane (either shared or

exclusive) will actuate the same phase as the concurrent through traffic. On the other hand,

it is logical to assume that a protected left turn will have a detector that actuates the protected

left turn phase. This has very important implications for the analysis of compound left turn

protection. It means that LT vehicles will not extend the permitted phase. When they occupy

the detector during the permitted phase, they will simply be placing a call for their own

protected phase. In the analysis of compound left turn protection, it is necessary to make a

strong distinction between protected plus permitted (leading), and permitted plus protected

(lagging) left turn phasings. Each of these cases will be analyzed separately.

Protected Plus Permitted Phasing

The QAP for protected plus permitted phasing is presented in Figure 3-8. In Figure

3-8, it is necessary to note that if the number of opposing lanes, n,, is greater than one, s,

will be equal to zero. Keep in mind that the QAP is used for the purpose of determining the

length of the protected phase only. The length of the permitted phase will be determined by

the simple process of its corresponding through phase because there are no permitted left

turns that actuate the detector. The The most important piece of information provided by









77
the QAP in this case is the size of the queue accumulated at the beginning of the green arrow,

Q0. With protected plus permitted phasing, this is equal to the queue at the end of effective

red time, Q, Given %Q, the determination of green time follows the simple procedure for

protected movements.








r
SQr = Qga


-Pq-r
o4- q
QP S \Is-qg








Figure 3-8. Queue accumulation polygon for protected plus permitted LT phasing with
an exclusive LT lane.


Permitted Plus Protected Phasing

This case is illustrated in Figure 3-9, which is essentially the same as Figure 3-8,

except that the order of the phases has been reversed. It is very important to note that the

protected (green arrow) phase must be presented last in both cases, because this is the phase

whose length we are trying to determine.

Again, it is needed to know the number of vehicles accumulated at the beginning of

the green arrow phase, Q., which is equal to Q'in this case. This will raise an interesting








78
question. Suppose that the value of Q9 is zero. This could happen if the permitted phase was

able to accommodate all ofthe left turns. Theoretically, the protected phase should never be

called under these conditions. However, there is a stochastic element which dictates that all

of the phases will be called occasionally. In this scenario, the adjusted minimum phase time

will be used to estimate the phase time for this protected phase.






Q1
o


|lf Q, g,..
Sp-qr





r

Time (seconds)


Figure 3-9. Queue accumulation polygon for permitted plus protected LT phasing with
an exclusive LT lane.


ATime (selications

The analysis presented in this chapter fills the gaps left in the preliminary

methodology. The complete analytical basis for a practical computational method to predict

traffic actuated signal timing are conducted. This method should be sensitive to a wide range

of actuated controller parameters. The QAP concept is especially attractive because it can

provide a clear picture for estimating the signal timing. Another essential benefit is that it can








79

also provide a direct estimate of the uniform delay that is compatible with the current HCM

Chapter 9 delay model The methodology presented in this chapter will be incorporated into

the computational framework described in the preliminary model structure to develop a

complete model implementation for predicting the signal timing at a traffic-actuated signalized

intersection. The model implementation and model evaluation will be presented in the next

two chapters, respectively.













CHAPTER 4
MODEL IMPLEMENTATION


Introduction

A specific analysis program, ACT3-48, was developed by Courage and Lin in this

study as a tool to implement the developed analytical model and procedure to predict the

traffic-actuated signal timing. The original worksheets have also been modified in accordance

to the analytical modelto cover all possible movements described in Chapter 3. The ACT3-

48 program can produce intermediate outputs in a format identical to the modified

worksheets. The computer program is required because the iterative nature of the procedure

makes it totally impractical for manual implementation. The program is able to evaluate the

proposed analytical models using a variety of data. In this chapter, the computer program

structure and logic are presented.

The analytical model developed in this study is for isolated mode of actuated

operation. An effective method to predict the phase times for coordinated mode of actuated

operation is to apply the analytical model for isolated mode to predict the phase times of

actuated phases and then properly assign the unused phase times to the non-actuated phases

(arterial through movements). By appropriate implementation of the analytical model for

isolated mode, a procedure has been built to predict the phase times for coordinated mode,

which will also be addressed in this chapter.










Structure and Logic of the ACT3-48 Program

The major structure of the ACT3-48 program is shown in Figure 4-1. It is not

difficult to recognize an iterative loop inside the flow chart. The iterative loop is required to

make the cycle time converge to a stable value. The program structure is divided into six

major parts: 1) Data Input, 2) Lane Group Specification, 3) Accumulated Queue

Computation, 4) Extension Time Computation, 5) Required Phase Time Computation, and

6) Cycle Time Adjustments. These six parts will be addressed separately.

Data Input

The data input for the ACT3-48 program is from the WHICH program, which not

only has a user-friendly input scheme, but also provides sufficient information for actuated

operations. After the data is input, the ACT3-48 program can be executed from WHICH to

process these data and compute phase times.

In the "control specification" shown in Figure 4-1, control treatments are determined

first according to the input data. These treatments include left turn types (protected,

permitted or compound protection), phase swaps and overlaps. In addition, a left turn

equivalence, ELi, is computed. Finally, the phase sequence pattern is recorded.

The phase sequence pattern needs to get more attention because in each iteration, it

is required for the accumulated queue computation and phase time prediction. The

computation of most time elements such as g, r, gf and gq in the QAP are also based on the

phase sequence pattern. The initial phase pattern is from the input of WHICH. Due to the

dual-ring control logic, the phase sequence pattern may change during the iterative process.

Therefore, possible phase sequence patterns are required to be considered in the program.























































Figure 4-1. Major structure of the ACT3-48 program.








83

There are eight possible cases of phase sequence patterns in all. For the purpose of

illustration, only the phase sequence patterns in the north-south direction are shown in figures.

Case 1 is a standard case for permitted turns which is shown in Figure 4-2. Case 2 is the

phase sequence for leading green which is shown in Figure 4-3. In contrast to Case 2, Case

3 is the phase sequence for lagging green which is presented in Figure 4-4. Case 4 shown in

Figure 4-5 is the phase sequence for leading and lagging green. Case 5 is a left turn phase

with leading green which is shown in Figure 4-6. Case 6 is leading dual left turns, and Case

7 is lagging dual left turns. Cases 6 and 7 are shown in Figures 4-7 and 4-8, respectively. In

Cases 6 and 7, the phases for dual left turns will terminate simultaneously. Finally, Case 8

shown in Figure 4-9 is leading and lagging with dual left turns. Case 4 and Case 8 are

interchangeable. For example, when the volume of northbound left turn volumes are heavy

and through traffic is light, Case 4 may become Case 8.









3 4




7 8


Figure 4-2. Case 1: Phase sequence for simple permitted turns.


















3 4






7 8




Figure 4-3. Case 2: Phase sequence for leading green.







3 4






7 8




Figure 4-4. Case 3: Phase sequence for lagging green.








3 4






7 8


Figure 4-5. Case 4: Phase sequence for leading and lagging green.


















3 4







7 8




Figure 4-6. Case 5: Phase sequence for LT phase with leading green.






3 4






7 8




Figure 4-7. Case 6: Phase sequence for leading dual left turns.








3 4






7 8


Figure 4-8. Case 7: Phase sequence for lagging dual left turns.















3 4




7 8



Figure 4-9. Case 8: Phase sequence for leading and lagging with dual left turns.



In "actuated parameter specification", the actuated parameters are specified based on

each NEMA phase which was defined in Chapter 2. The actuated parameters consist of

minimum initial, maximum initial, minimum phase time, maximum green time, allowable gap

recall, detector configuration and so on. Based on the control specification and the actuated

parameter specification, Worksheet 1: Traffic-actuated Control Input Data can be produced

by the ACT3-48 program.

Lane Group Specification

In "lane group specification", the ACT3-48 program will determine the phase

movements for each NEMA phase and the lane group movement within each NEMA phase.

For example, ifa NEMA phase includes all left turns, through traffic and right turns, the phase

movement of this NEMA phase will be specified as "LTR". If a lane group within the NEMA

phase is just for through and right turns, it should be presented as "TR". The purpose for this

specification is to associate each of the lane group with its NEMA phase.








87
The lane group specification is convenient and necessary for later computation. For

example, the phase time computation for permitted left turn movement from an exclusive lane

("L") will be different from that for through and right turn movements ("TR"). In addition,

the ACT3-48 program will also determine the traffic volume (veh/hr), arrival rate (veh/sec),

saturation flow (veh/hr) and departure rate (veh/sec) for each lane group based on the input

data from WHICH. Worksheet 2: Lane Group Data is then generated.

Accumulated Queue Computation

In a new iteration, the most important step is to create a new queue accumulation

polygon (QAP). With the phase sequence pattern, phase times and other information from

the last iteration, the new QAP can be produced easily for simple through or protected

phases. It becomes more difficult in computation for permitted left turns from either an

exclusive or a shared LT lane and compound left turn protection because more information

is required before accumulated queues can be computed.

In the analytical model for permitted left turns from an exclusive lane, the opposing

queue service time, g, is needed for QAP. For the permitted left turns from a shared LT lane,

both gq and free green, gf are required. In the compound LT protection, an exclusive LT lane

is assumed, so gq must be obtained. Fortunately, the method for estimating g and gf has been

presented in HCM Chapter 9, which is used in this study to compute g and g. In the ACT3-

48 program, the computation of gq is based on 1994 version of HCM Chapter 9. Since the

computation of& is based on the length of green time in the latest HCM version which may

cause unreliable convergence of cycle length, the method presented in 1985 HCM Chapter

9 is used instead. Based on the number of opposing lanes, there are two scenarios described