$1.57“ 64.1. v - ‘
. nufiufl.

. . 4,, . . . .. ._ ”v . .. ... 1.. .... . ;, ... . . 7.. o. RM“) .le

.
. .n . a
3.. .1w 3...: 1&5
-

 

11917;...
i L .. . .. 3.. .1.Hu......nuh
‘ . .LévVnkvrunUMefi. . . h . . . . .. b. . “ . . .
. :t . r: «lidxfi. . . .

L, k$mw raga, g ; ‘mafizk.
:” 9.. .
.1 . .
._.., .

 

II
A

v
null.

.‘t.

vh‘

firm... .: y
9",»; Jan-.1. r I
.. .- l unratb

9 .33..- v. ‘1‘!

v 20L...

.
1. . .
if?

. S; harm?
. infimvubfi.
Yr? :32}.
E. ell-.1 5.5.4.43.
I
.IU'. it!!!
unbkufixu

: V
43 I... flu... y: ...u.wf«;2..l
. us 4mm. zhnu ¥mfilhfi§hhuu V
x a. s y z . ..
1.1.9 ‘ .. . 5% Eli}...
”rift." t k... 4"

.
A. 53.”?! u .. X e--.)

. i .uur»..a¥..
5.:
.. L... .3.
.. E». .11; 1:...
(.233... Q

. o ’31.: 2.. Ixnltvls. L
2.5.... .3553“: niftflmatiuwtznu
gnu. r169... man“: :34...“ .. , .......u».h4v hufi‘.
.Il JILI". .. l1! glirl‘lpy‘nl {5...}.
.tumoniurfi. “£15,: :flefihifl. an 7 .mfiiihfl.
.. ....... ..‘....... .
is»... utlyt..-) 2.1.. Ji
‘3 0» V}. " ' All I ‘5
u“ an . .95.»...1...‘ 3.1;: . .

 

‘ q. . . , .1 def.
I“. I. . Is . ..
. on 1.1.- UN, ‘. . . .
l .‘f ’Qtn-O i ‘0 n."p‘-ll>
r. n! k “U531... u. Pfiuflu.‘ J..u.»1.m...n.r.h
in: .7..w;.o.l;fl!. L! 5‘. . u... 12.591. . 1..
.bny‘b' o. I}: ‘JIQIs‘vJ. Ob I49.

WW3 .hmeJmWMuflrS . :3 .
. a. u x-|.. ‘ - ‘
. t: SIMHWNFZ HmWMd “1.??er. ._. ....v :3! :. z _
. t! u‘ . - h. 31...”: -5 . I . . ‘ I .
“4......anth .v.fib;....9nw:.lhllm.nN«u.u {who i . V . . ., .u...9...fi.v....rh¢.,.uuuu.
.15: SunflzfiuNWcflJmhnln , ‘ . a .. . . , . . . . . .
.11... mwumx‘uzllumt .532..- . ‘ . . . . ‘11: , .
, .smgrfi (I... «firinmbqtu fzb... . . . . . . a Vouvu‘ ..—D‘..P...
. Rd” He‘u...“ - ruflfivfa [ A ‘ .
. ..p..niu.}mw .. . ‘
.. t 1!! ..

.u‘aufi...n¢

.cn(|ucl.ltit.v~a» (I! . alcri
ll . 3.. 14 .. 114i . .Ir.
.\ufl.-f&~ hrh”. .I‘ T a
nvflufln. nfllflnfl... . ..
lvfllhhsc. gin-Htui fry! 51(1); ‘
‘ .l...tl..:.!...-.:... $13.3. . twin}, 1...: .zéa........,r...1....\..:
. . .n .A..!.X.3s§ri.ai.2 2 .u.v: vfifihfi. . 1..
. {Infinvunlfl .41.} . J.r.nlv..{:-.vix . . 1w

\vl.

0743' Qllvv tic}!

. v. 3 .
O‘H‘Ora .3 ‘5: ‘ .
3.123 no. . My. l1: 9
‘33..“ 311... 3.64... nu.
if... .flhflgférfi. .1 1.»
u. 2.3. If... :1- .
v1.30
1 Pl! .Izl! l.
k§u2.r.«hmu.hu.uulh v.0); . .
10...! affilii 3L. $31.. 30:; :1 (1.1».

, .ch

. .LHI“.

- u ‘u I . I h ' .
.ewfikwwmwflw

)!
“I?

 

1..

 

‘.‘§. ‘ ‘31:! 1!“.ercll H I.

v.1.VIvI... v. I-vlrv , .. . )Ix in .«vtniu .. If: (5|?
:|);..-n|‘l.l 46111») : 1 y . . .. ‘nrhkyI06. ; N..y..hbn.. MV‘....|U.H|L
. n 5 bed ,.L w ill! -. Itoiwv IIN‘ 110.9111. 1 ‘l'vuV :1. ln. ‘ , 1. ......VI . .. . .7
l | IO! ‘ (O... . ‘1“. A .c..«.l . J. .3 : - . .2.‘1I|llf.lhb}.r. ‘IIU .‘0

- I' i '

I.

 

THESIS
/7

A

Ibiiiflfl'lllll'il [initiallflflififliiilliiflfifli'flfill ._.

3 1293 01570 6942

This is to certify that the

dissertation entitled

Development of a Modeling Approach to Analyze
Intersection Traffic Delay Under the Control
of a Real—Time Adaptive Traffic Signal System

presented by

P. Brian Wolshon

has been accepted towards fulfillment '
of the requirements for

 

 

Ph.D. degree in Civil Engineering

WW C. fffl/‘r/é"

Major professor J

Date 3// 3/9 7

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

LIBRARY

Mlchlgan State
University

 

 

 

PLACE IN RETURN BOX to romovo this checkout from your record.
TO AVOID FINES return on or bdoro doto duo.

DATE DUE DATE DUE DATE DUE

 

_ gfieéao
JAN)“. I ‘ him 1 ' 111'

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

:L_J
j! I l

 

MSU lo An Affirmative WM Oppoltunlty Institution
Wane-m

DEVELOPMENT OF A MODELING APPROACH TO AN ALYZE
INTERSECTION TRAFFIC DELAY UNDER THE CONTROL OF A
REAL-TIME ADAPTIVE TRAFFIC SIGNAL SYSTEM

By

P. Brian Wolshon

A DISSERTATION
Submitted to
Michigan State University

in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY

Department of Civil and Environmental Engineering

1997

ABSTRACT

DEVELOPMENT OF A MODELING APPROACH TO ANALYZE
INTERSECTION TRAFFIC DELAY UNDER THE CONTROL OF A
REAL-TIME ADAPTIVE TRAFFIC SIGNAL SYSTEM

By

P. Brian Wolshon

The United States Department of Transportation has recently begun implementation of the
national demonstration project for suburban Advanced Traffic Management (ATMS) and
Advanced Traveler Information Systems (ATIS) in Oakland County, Michigan. As part of
this project, the City of South Lyon has recently converted the control of its signalized traffic
network from optimized fixed-time control to the Sydney Coordinated Adaptive Traffic
System (SCATS). SCATS is an automated, real time, traffic responsive signal control
strategy. Under SCATS, the timing of the signals is governed by a computer-based control
logic. The system has the ability to modify signal timings on a cycle-by-cycle basis using

traffic flow information collected at the intersection approach stop lines.

The expected benefit from such a system comes from its ability to constantly modify signal
timing patterns to most effectively accommodate changing traffic conditions. While the
potential benefits from this control structure may be significant, few research studies have
used computer simulation modeling to document the effect of implementing this method of
signal control. The objectives of this research study are to analyze and document the changes

to certain traffic flow and delay parameters which have occurred in South Lyon as a result

of implementing SCATS signal control.

The study will make use of the “before/after” analysis technique to measure certain aspects
of the traffic delay conditions in periods prior to and following the implementation of
SCATS control. The research will be structured within a framework of several component
segments including data collection, reduction, analysis, simulation model construction, and

statistical testing.

Dedicated to my wife Kelly,

my parents Paul and Susanne, and my brother Jeff

iv

ACKNOWLEDGMENTS

I would like to express my appreciation and gratitude to my advisor and committee chairman
Dr. William C. Taylor, for his invaluable advice and guidance throughout my studies at
Michigan State University and during the completion of this research. His patience, kindness,

and insights will never be forgotten.

My appreciation and gratitude are also extended to Dr. Thomas Maleck, Dr. Richard Lyles,
and Dr. Joseph Gardiner for serving as members of the guidance committee and for their
assistance. My gratitude is also extended to current and former graduate students Dr.
Sorowit Narupiti, Mr. Mohammed Saif, and Mr. Ahmed Abdel-Rachman in the Department
of Civil and Environmental Engineering at Michigan State University. Their assistance,

suggestions, and encouragement were also key to the completion of this research.

Words can not express my most sincere love and gratitude to my wife Kelly. Without her
constant support, and patience I could not have completed this project. I wish to thank my
brother Jeffrey to whose heights of creativity and intelligence I have always aspired. To my
father, I wish convey my gratitude for his love, support, and constant encouragement
throughout my life. Through his example of hard work and a sense of humor, I have learned
that the most important achievements in life are happiness and the knowledge that I have
always tried my best. Finally, I would like to express a debt of gratitude to my mother
Susanne. I know that the completion of my Doctoral Degree means more to her than it could
ever mean to me. It has been because of her continuous love, support, sacrifice, and
unending devotion to her children that my brother and I have the confidence to strive for

greatness.

TABLE OF CONTENTS

 

LIST OF TABLES ........................................................................................ ix
LIST OF FIGURES ...................................................................................... xii
CHAPTER 1: Introduction ..... 1
1.1 FAST-TRAC and SCATS ................................................................ 2

1.2 The South Lyon Traffic Network ....................................................... 4
CHAPTER 2: Literature Review - 9
2.1 Adaptive Traffic Control Systems .............................................................. 10

2.2 Traffic Simulation and Modeling Tools ..................................................... 15

2.3 FAST-TRAC and Related Intelligent Highway Projects ........................ 20

2.4 Comparative Studies of Adaptive Traffic Signal Systems ...................... 22

2.4.1 SCATS ....................................................................................... 22

2.4.2 SCOOT ....................................................................................... 25

2.5 Literature Review Conclusions .......................................................... 26
CHAPTER 3: Research Objectives and Approach . ............. ........... 28
3. 1 Objectives .................................................................................................... 30

3.1.1 Total Intersection Delay .............................................................. 31

3.1.2 Approach Delay ........................................................................... 32

3.2 Experimental Approach ............................................................................... 33

3.2.1 Data Collection ............................................................................ 34

3.2.2 The “After”Traffic Analysis ......................................................... 36

3.2.3 The “Before”Traffic Analysis ....................................................... 37

3.2.4 Comparison of Results and Conclusion ....................................... 38

vi

vii

 

 

CHAPTER 4: South Lyon Traffic Analysis - - - 40

4.1 Data Collection ............................................................................................ 41

4.1.1 “Before” Data ................................................................................... 42

4.1.2 “After” Data ................................................................................ . 47

4.1.3 Related Data ................................................................................. 50

4.2 Development of the Traffic Analysis Models ............................................ 51

4.2.1 The SOAP Program and Its Limitations ....................................... 51

4.2.2 “After”Delay Calculation Model ................................................. 57

4.2.3 “Before”Delay Calculation Model ............................................... 60

4.3 Model Limitations ....................................................................................... 62

4.4 Delay Mode] Output Data ......................................................................... 67

CHAPTER 5: Comparison and Interpretation of Modeled Delay 81

5.1 Total Intersection Delay .............................................................................. 82

5.2 Approach Delay .......................................................................................... 90

5.3 Delay Data Comparison Conclusion ........................................................... 97

CHAPTER 6: Summary and Conclusion - . - - 100

6. 1 Summary of Research Study ........................................................................ 100

6.2 Conclusions ................................................................................................. 102

6.2.1 “Direct” Conclusions ................................................................... 103

6.2.2 “Indirect”Conc1usions ................................................................. 105

6.3 Future Research Needs ................................................................................ 107

6.3.1 Advanced Simulation Systems .................................................... 107

6.3.2 Improved Methods of Analysis .................................................. 108

6.3.3 Cost/Benefit Assessments ........................................................... 108
APPENDICES

A. Delay Analysis Model Source Code ............................................................ 1 10

B. Delay Analysis Model Ouput ...................................................................... 1 19

LIST OF REFERENCES . --------- ------- -- - - - 126

 

Table 4.1
Table 4.2a

Table 4.2b

Table 4.2c

Table 4.2d

Table 4.3a
Table 4.3b
Table 4.3e
Table 4.3d
Table 4.3e
Table 4.3f
Table 4.4a
Table 4.4b
Table 4.4c
Table 4.4d
Table 4.4c
Table 4.4f
Table 4.5a
Table 4.5b

LIST OF TABLES

Field Recorded Traffic Data ........................................................................ 52
Green Phase Split in Optimized Fixed Time Operation - Pontiac Trail and Ten
Mile Road Intersection ................................................................................. 63
Green Phase Split in Optimized Fixed Time Operation - Pontiac Trail and
Nine Mile Road Intersection ....................................................................... 63
Green Phase Split in Optimized Fixed Time Operation - Pontiac Trail and
Eleven Mile Road Intersection ..................................................................... 64

Green Phase Split in Optimized Fixed Time Operation - Reynold Sweet

Parkway and Ten Mile Road Intersection .................................................... 64
Delay Mode] Output - Pontiac Trail and Ten Mile Road, Monday ............ 70
Delay Model Output - Pontiac Trail and Ten Mile Road, Tuesday ............ 70

Delay Model Output - Pontiac Trail and Ten Mile Road, Wednesday ....... 71
Delay Model Output - Pontiac Trail and Ten Mile Road, Thursday .......... 71
Delay Model Output - Pontiac Trail and Ten Mile Road, Friday ................ 72
Delay Model Output - Pontiac Trail and Ten Mile Road, Daily Average 72
Delay Model Output - Pontiac Trail and Nine Mile Road, Monday ........... 73
Delay Model Output - Pontiac Trail and Nine Mile Road, Tuesday .......... 73
Delay Model Output - Pontiac Trail and Nine Mile Road, Wednesday ..... 74
Delay Mode] Output - Pontiac Trail and Nine Mile Road, Thursday ......... 74
Delay Model Output - Pontiac Trail and Nine Mile Road, Friday ............... 75
Delay Mode] Output - Pontiac Trail and Nine Mile Road, Daily Average .. 75
Delay Model Output - Pontiac Trail and Eleven Mile Road, Monday ....... 76
Delay Model Output - Pontiac Trail and Eleven Mile Road, Tuesday ....... 76

ix

Table 4.5e
Table 4.5d

Table 4.5e
Table 4.5f

Table 4.6a

Table 4.6b

Table 4.60

Table 4.6d

Table 4.6e

Table 4.6f

Table 5.1
Table 5.2

Table 5 .3
Table 5.4a

Table 5.4b

Table 5.4c

Table 5.4d

X

Delay Model Output - Pontiac Trail and Eleven Mile Road, Wednesday .. 77

Delay Mode] Output - Pontiac Trail and Eleven Mile Road, Thursday ...... 77
Delay Mode] Output - Pontiac Trail and Eleven Mile Road, Friday ........... 78
Delay Model Output - Pontiac Trail and Eleven Mile Road, Daily

Average ........................................................................................................ 78
Delay Model Output - Ten Mile Road and Reynold Sweet Parkway,
Monday ....................................................................................................... 79
Delay Mode] Output - Ten Mile Road and Reynold Sweet Parkway,
Tuesday ....................................................................................................... 79
Delay Model Output - Ten Mile Road and Reynold Sweet Parkway,
Wednesday .................................................................................................. 79
Delay Model Output - Ten Mile Road and Reynold Sweet Parkway,
Thursday ...................................................................................................... 80
Delay Model Output - Ten Mile Road and Reynold Sweet Parkway,

Friday .......................................................................................................... 80
Delay Model Output - Ten Mile Road and Reynold Sweet Parkway,

Daily Average ............................................................................................ 80
Statistical Comparison of Total Intersection Delay ..................................... 84
Percentage Comparison of Total Intersection Delay ................................... 87
Statistical Comparison of Approach Delay ................................................ 91

Percentage Comparison of Total Intersection Delay - Pontiac Trail and Ten
Mile Road ................................................................................................... 94
Percentage Comparison of Total Intersection Delay - Pontiac Trail and Nine
Mile Road ................................................................................................... 94
Percentage Comparison of Total Intersection Delay - Pontiac Trail and Eleven
Mile Road ................................................................................................... 95
Percentage Comparison of Total Intersection Delay - Reynold Sweet Parkway

and Ten Mile Road ..................................................................................... 95

Figure 1.1
Figure 1.2
Figure 4.1a
Figure 4.1b
Figure 4.10
Figure 4.2
Figure 4.3

LIST OF FIGURES

Area Map ..................................................................................................... 5
South Lyon Vicinity Map ............................................................................ 6
Signal Phasing Diagrams and Lane Use Configurations ............................. 43
Signal Phasing Diagrams and Lane Use Configurations ............................. 44
Signal Phasing Diagrams and Lane Use Configurations ............................. 45
Sample Segment of SCATS Output File .................................................... 48
Sample Segment of Delay Model Output ................................................... 68

xi

Chapter 1.

Introduction

The United States Department of Transportation has recently begun implementation of the
national demonstration project for suburban Advanced Traffic Management (ATMS) and
Advanced Traveler Information Systems (ATIS) in Oakland County, Michigan. As part of
this project, the City of South Lyon has recently converted the control of its signalized traffic
network from optimized fixed-time control to the Sydney Coordinated Adaptive Traffic
System (SCATS). SCATS is an automated, real time, traffic responsive signal control
strategy. Under SCATS, the timing of the signals is governed by a computer—based control
logic. The system has the ability to modify signal timings on a cycle-by-cycle basis using

traffic flow information collected at the intersection approach stop lines.

The expected benefit from such a system comes from its ability to constantly modify signal
timing patterns to most effectively accommodate changing traffic conditions. While the
potential benefits from this control structure may be significant, few research studies have
documented the effect of implementing this method of signal control. The objectives of this

research study are to analyze and document the changes to certain delay parameters which

2

have occurred in South Lyon as a result of implementing SCATS signal control.

This study will make use of the “before/after” analysis technique to measure certain aspects
of the delay conditions in periods prior to and following the implementation of SCATS
control. The research will be structured within a framework of several component segments
including data collection, reduction, and analysis; simulation model construction; and

statistical testing.

1.1 FAST-TRAC and SCATS

The South Lyon signal improvement project is a small segment of the FAST-TRAC (Faster
and Safer Travel - Traffic Routing and Advanced Controls) project“) FAST-TRAC has
involved the conversion of more than 1,000 pretimed and actuated signalized intersections
in Oakland County to SCATS control and has established a county-wide, real-time route
navigation system?” While FAST-TRAC has been managed primarily by the Road
Commission for Oakland County (RCOC), it has been a cooperative effort between many
federal, state, county, and local government agencies; as well as private corporations and

universities.

The SCATS system was developed in the 1970's by the Roads and Traffic Authority of New
South Wales, Australia. The operational aspect of SCATS has been compared to the type
of control “provided by a traffic control officer stationed at the intersection controlling traffic
to insure that congestion is reasonably equal among the various approaches. The primary
difference is that today’s real time adaptive control systems can anticipate the arrival of
vehicles from preceding intersections and respond accordingly.”‘4) It functions by making
constant modifications to traffic signal timings in real-time in response to the variations in
traffic demand and system capacity. It has advantages over the police officer by evaluating
and controlling the signal system on a system wide basis rather than on an isolated

intersection by intersection basis. It operates by using traffic sensors to monitor flow

3

conditions and thus coordinate signal timings in order to minimize stOps and delay time when
the system is at or near capacity. SCATS attempts to maximize the system capacity and
minimize the possibility of traffic jams by controlling the formation of queues. One of the
ways that SCATS can accomplish this is by providing a progression of green signal phases
to reduce stopped vehicle queues, thereby reducing delay and decreasing the network travel

times“)

Input data for the SCATS system is collected via a system of traffic sensors. The sensors
may be inductive loop detectors imbedded in the pavement or, as in the case of the South
Lyon system, video image devices mounted overhead on the signal strain poles or attached
on mast arms. The traffic information collected in the field involves the discharge
characteristics (i.e., flow and occupancy during the green phase) on each intersection
approach. This data is transmitted to a regional control center where the SCATS control
program attempts to most effectively maintain the highest degree of saturation on the

intersection downstream of the collected traffic data.

SCATS divides the network into systems and subsystems. Each subsystem contains a single
“critical“ intersection, usually where two high volume roadways intersect. SCATS control
logic incorporates a dynamic process whereby intersection signal phasing is coordinated.
This system is known as “marriage” and “divorce.” Married intersections coordinate timings
to allow platoons of traffic to pass through. A divorce occurs when two intersections no
longer require coordination to maximize traffic flow through the network. The divorce is
implemented after three consecutive cycles warrant a divorce, insuring additional stability

within the system.

Other differences that separates SCATS from conventional fixed timed systems is its ability
to modify timing strategies to fit various control philosophies and to collect, process, and
maintain a history of traffic statistics for an area. Signal phases can be set to equalize

saturation on all approaches or they can be arranged to give priority to a particular direction

4

of importance. Since the SCATS system requires the use of certain traffic data information
it has the ability to record and store these statistics to monitor the strategic performance of
the system, detect signal faults, and allow manual overrides of the signals under special

operating circumstances.

1.2 The South Lyon Traffic Network

Oakland County is located in the southeastern comer of Michigan, immediately north of the
City of Detroit. Throughout the past 15 to 20 years Oakland County has experienced an
explosion in commercial and residential development. Accompanying the growth in
development was an equally significant increase in traffic congestion”) The City of South
Lyon is located in the southwest corner of Oakland County, approximately 40 miles
northwest of the City of Detroit. Figure 1.] illustrates the location of South Lyon relative to
the Detroit metropolitan area. It is presently a small, semi-rural/suburban community of
approximately 25,000 people. The traffic congestion problems are comparatively minor
compared to many cities in Oakland County. However, like the rest of Oakland County, the
land within the City and surrounding Township is experiencing considerable commercial and
residential development and increases in the amount of traffic congestion are expected to
follow. It was expected that the introduction of the SCATS system of advanced traffic
management will allow the community of South Lyon to accommodate the anticipated

increases in traffic in a cost effective and efficient manner.

The layout of South Lyon road network is arranged in a perpendicular grid system of primary
roadways. The major roads are spaced at approximately one mile intervals. The lone
exception to the grid layout is Reynold Sweet Parkway. The Parkway serves as a bypass
route for through traffic around the central business district. The layout of the South Lyon
street system can be seen in more detail in Figure 1.2 The majority of the South Lyon traffic
load is carried on Pontiac Trail and Ten Mile Road. Pontiac Trail is the main north-south

arterial roadway in South Lyon, providing access to Interstate 96 which is located

5

 

 

Wm mwmrrmnrmmto'm

 

    

coMMEncerwpmr

W

   

   
   
   
     
     

'Lake
''''''' .. . .RD ‘ »~‘
I ... Blanmer Na. 1. 1
'9 ButnParir “'
' 91 ;.‘-_‘
IBAR 111) K r
>31.-. .. 9V ._ ,
o D E ‘Q/ m 9,
I 1' °’ 1 :W
M.“ 1‘, PEARSON ‘._ 3
$3 9‘ i x a;
z m‘
j.

  
 

,/
NMSVED

  
 
      
      
    
 

  
  

    
  
 

     

2
E
3 E12
° :1 MfliEilrgsou
x 1 F >
y I 2:1 flaws in ‘ J .,. §: HTS :
1’ "f g g g 12 MILE no ~
«2k _.\ .i/ w 1 in. >

, H “F: ’ g :12 MILE an E

12 MlLE RD ‘5‘ 1 _ x E
. 3 1RD a § 2 2

MILE

1m
-5 .1
.1; ,
I!
U

GRISWOLD

  

GREENFIELD

FENKELL .

    

 
  

1

           
  
  

 

Enoch/mt: Ito ‘ a“

    
  
    

    
  

s

mxaono

 
   

PLYMOUTH ”1105

=
r11
—1
« z:
2
—1

0110501

  

-5. ;
FIGURE 1.1

      

  

        
          

NUREAFDI-

 

  
 

 

 

   

  

 

     

——~L“ -. ----- , ~ AREA MAP

. g. . J ./
rm 3; 1 .L \. “151:1?”
. ’ - 1 “ “Maw . 0% J.» %1%\ SOUTH LYON Ams EVALUATION STUDY
._ .3 :REELMD “Z I ' 12-..-.. r-n Pnflrwav L, {KS _ ,. m

 

 

”“5141!th
IWMWIUMMWW ‘

 

 

 

 

FIGURE 1.2

 

SOUTH LYON VlClNlTY MAP

 

SOUTH LYON ATMS EVALUATION STUDY

DOM 51”! mn
”All!!!“ 0' GI. NO mm. m

 

 

 

 

7

approximately four miles north of town. The overwhelming majority of commercially
developed land within South Lyon is located directly adjacent to Pontiac Trail. Ten Mile
Road is the prime east-west arterial serving the traffic demand to the commercial centers

located to the east.

The local street network is also arranged in an approximate grid layout. Its density is
considerably greater than the primary road system. The area of greatest density is located
within the vicinity of the central business district, surrounding the intersection of Ten Mile
Road and Pontiac Trail. As a result of the dense commercial development, two minor street
intersections with Pontiac Trail have been signalized. The South Lyon road network under
study also incorporates two at-grade railroad intersections. One of these, to be described in
detail later, forms an awkward triple intersection at a point where two roads and the railroad

tracks all coincide.

There were six signalized intersections in operation within the South Lyon traffic network
during the study period of 1995-1996. The installation, operation, and maintenance of these
signals are under the jurisdiction of the Road Commission for Oakland County. The closest
traffic signal outside of the study network is located approximately five miles away.
Originally, all of these signals operated on a well coordinated, pre-timed basis. No traffic
adaptive control measures, like actuated signal timing, were initially at work within South
Lyon. The Oakland County ATMS project involved the conversion of all six signals to
SCATS control. It was thought that the relative isolation of South Lyon and its small signal
network would make it an ideal “laboratory” for an evaluation study of the SCATS system

of advanced traffic management.

It has been hypothesized that the ATMS proposed for South Lyon will result in
improvements to the traffic flow conditions within the controlled network. With the
resources allocated for the design, implementation, testing, and research of FAST-TRAC and

the potential worldwide implementation of adaptive traffic signal control systems, like

8

SCATS, there is a need to assess its benefits. The following research study has documented
performance measures before and after the installation of the SCATS system, to address such

questions.

Chapter 2.

Literature Review

To establish the current state of the art a literature review was completed. The literature
search was performed to accomplish four primary goals. First, the current limits of
knowledge in the field of traffic engineering research relative to SCATS signal control were
explored. Second, the literature search lead to an understanding of the "gaps" resulting from
past research. The research gaps included the absence of studies performed specifically on
the comparative benefits gained from adaptive signal control. The literature review also
established a base of knowledge from which to launch the proposed study and demonstrated
certain techniques which have been successful in past research. Some of these were applied
to the South Lyon evaluation study to aid in the accuracy and efficiency of the experiment.
Finally, the review of past published literature gave insights into the way in which current
theories, technology, and implementation of adaptive traffic signal systems have developed
over the years. The search has shown how theories have been built from prior experiments
and how and why some ideas of twenty and thirty years ago have now come into practice

while others were unsuccessful in field applications.

10

The following sections of the literature review have been divided into the main areas of
traffic research that relate to this study. They include the historical development of adaptive
traffic signal control, the state of current and future practice of advanced traffic management
systems, the development and use of computer simulation and analysis tools to model traffic
related events, and an investigation into the experiments and theories which are under study
in Intelligent Transportation Systems (ITS). The last two sections of this chapter will
provide documentation of the results of recent comparative studies of adaptive signal control
systems and a demonstration of the need for research into more advanced measurement and

analysis techniques in this field.

2.1 Adaptive Traffic Control Systems

The majority of travel delay experienced on arterial roadways is governed by the
intersections along the route. Reductions in stop delay at intersections can enhance the
efficiency of a road and reduce the amount of time required to travel through a particular
corridor or network. One of the best ways to achieve a reduction in delay is to provide an
efficient progression of signals. A method of providing effective signal progression is to
anticipate the arrival of vehicles and modify signal timings to match the arrival pattern. The
Sydney Coordinated Adaptive Traffic System is one of several forms of advanced traffic
management which are in operation or under development that use anticipatory and adaptive

techniques to increase the efficiency of a road network.

Adaptive traffic control systems are traffic signal systems which have the ability to change
the phasing and timing strategies and coordination of signals within a network to meet the
demands placed upon them. These types of systems may take on many forms and levels of
complexity and automation. Their primary purpose is to increase the efficiency of a system

by maximizing the way traffic is managed.

11

The concept of traffic signal systems which could be modified to more efficiently move
traffic has been studied for many years. The practice of traffic signal optimization began in
the 1920's.“) As signal control technology became more advanced, signal systems began to
use pretimed coordination plans. The original "traffic responsive" signal systems were
merely pretimed signal control systems which could be modified to fit the anticipated traffic
demand pattems.‘7’ Such systems, still prevalent today, are designed to optimize traffic based
on average time-of-day and day-of-week conditions. These types of control work well and
are considerably less expensive to install, operate, and maintain than more complex computer

controlled systems.

Real-time adaptive control is a fairly recent development. Past research publications have
shown ongoing testing and development was underway in the early sixties. For real-time
systems to be possible, accurate detection, rapid data processing, and effective and speedy
transmission of data are necessary. One of the initial studies performed to evaluate the use
of computer and video technology was published in 1967.“) The study was one of the first
attempts to evaluate the feasibility of using video detection and recording equipment to
chronicle traffic activities near roadway intersections, detect the occurrence of accidents, then
record and store this data for further use. The system was not very SOphisticated and required
a technician to watch the video feed and look for incidents. The technology of the day did
not make it possible to automatically search and record events for use on a real-time basis.
However, the study was able to demonstrate the utility of incorporating automated

technology to search for and record certain elements of the traffic stream.

Another advance toward today’s advanced traffic management systems was made in 1967
with the development of a computer program and methodology to gather and process data
from loop detectors placed near intersections to determine certain traffic flow parameters.
This research and development project”) also demonstrated how such data could be used to
determine queue lengths and delay at signalized intersections. By today’s standards this

model was primitive. However, the technology demonstrated in this project has evolved into

12

one of the integral elements which allow real-time adaptive control systems to function

today.

The development and implementation of adaptive control systems have become possible over
the past fiVe to ten years due to a number of very important factors. Two of the most
important are concerns about the increased levels of congestion and advancements in
computer technology and data processing efficiency. The number of annual vehicle miles
traveled is increasing at a steady rate and it is not economically, environmentally, nor
spatially possible to construct new roads or expand existing ones to meet the anticipated
traffic. Therefore, the answer to improving transportation efficiency lies in the effective
management of existing systems. Real-time adaptive traffic management systems allow

traffic professionals to more effectively manage the existing transportation infrastructure.

Several types of advanced control systems are currently in operation, serving in experimental
traffic control management roles. Two such systems, the Sydney Co-Ordinated Adaptive
Traffic System (SCATS) and the Split, Cycle and Offset Optimising Technique (SCOOT),
have developed into the most prevalent real time adaptive traffic signal control systems and

are poised to become the automated traffic management systems for the next century.

The SCOOT system was developed in Great Britain in the mid 1970's.“” The main idea of
the SCOOT system was to take an "off-line" model, like the fixed time signal optimization
program TRANSYT, and have it operate "on-line." Specifically, the system uses traffic data
measurements collected from the existing stream and makes short and long term decisions
regarding the traffic signal settings. The great advantage to a real time, on-line system like
SCOOT is that it can respond to the current traffic conditions with almost no lag time and
it greatly reduces the need for constant human interaction in adjusting the signal timings.
Traffic stream measurements are collected automatically and adjustments to the timing plans
are made automatically. This is in contrast to fixed time, non-automated systems where there

is a need for traffic engineers and field technicians to study the traffic conditions and

13

implement the changes.

SCOOT is currently in operation throughout Great Britain as well as in Thailand, Hong
Kong, China, the Phillippines, and Canada. The first SCOOT implementation project in the
United States occurred in Oxnard, California in 1992.“°’ Studies performed by the
Transportation Road Research Laboratory in London, England found that, under certain
conditions, the SCOOT system provided a 20% improvement over a standard fixed time

signal system.”

While the SCOOT system represented a major advance in the management of signal timing,
it had some shortcomings (especially in its earlier forms) that made the parallel development
and implementation of another real time adaptive traffic control system even more desirable.
The SCATS system was developed in Australia during the same time as the SCOOT system.
SCATS, like SCOOT, is an automated traffic control system whereby traffic signal timings
are adjusted in response to variations in traffic demand. The SCATS system operates on a
real-time basis, using the current traffic flow information to modify and coordinate signal
timing. To minimize stops and delay time when the system is at or near capacity, SCATS
attempts to maximize the system capacity and minimize the possibility of traffic jams by
controlling the formation of queues.‘2'3'”’ SCATS accomplishes this by providing a
progression of green signal phases to reduce stopped vehicle queues, thereby reducing delay

and decreasing the network travel times”)

The SCATS system also offered some distinct advantages over SCOOT. Primary among
these was its ability to omit individual signal phases from the signal cycle if the traffic
demand did not warrant them. For example, a protected left turn phase could be eliminated
from the cycle if no vehicles were present; as detected by the sensors. The other advantage
of the SCATS system was its ability to readjust the phase sequence pattern to best take
advantage of the traffic conditions. This feature requires drivers to be more vigilant because

one cycle may start with a leading protected left turn phase, while the next cycle might

l4

feature a lagging protected left turn phase.

Development of yet another advanced traffic control project is currently underway in Europe.
The research and development associated with the HERMES (High Efficiency Roads with
Rerouting Methods) project has taken place in parallel with the testing, development, and
implementation of SCATS in Australia and North America. The HERMES project relies on
the origin-destination characteristics of the network and automatic incident detection
information to execute the advanced traffic management process.” The HERMES system
utilizes origin-destination algorithms to determine traffic flows within the control system.
The origin—destination information is used to set strategies for arterial signal control and
freeway ramp metering. One of the primary advantages of the HERMES system is that it
seeks to go a step further than SCATS or SCOOT by searching the network for the
occurrence of disruptions or incidents which can restrict the flow of traffic. Then it
implements new routing strategies, based on the location and severity of the incidents, to

most efficiently route vehicles around the effected area.

The automatic incident detection element of the HERMES system is contained in the
MONICA (Monitoring Incidents and Congestion Automatically) algorithmm'”) Once an
incident is detected in the system, a process of rerouting the traffic is enacted. The
difficulties of implementing a fully integrated system like HERMES mean that some aspects
of the project remain in the simulation and deve10pment stage. However, work continues
since the integration of an automatic incident detection and traffic rerouting system within

an advanced traffic signal control system remains a desirable goal.

Another form of advanced traffic control under study and development is the OPAC
(Optimization Policies for Adaptive Control) demand-responsive traffic signal control for
individual intersections.“5"6’ This ATMS structure uSes a dynamic programming approach
to determine an optimal signal control strategy. It then compares the optimized timing to a

more conventional optimized signal timing pattern, based on volumes and past experience.

15

The research study indicated that the OPAC strategy demonstrated a significant improvement
over fixed time control. The conclusions were drawn from simulation trials and analyses
using Highway Capacity Manual‘m criteria. The study did not include field verification of
the simulation trials. The obvious limitation of the OPAC system is that it optimizes
signalized intersections in isolation, in contrast to SCATS and SCOOT, which seek system

optimization.

2.2 Traffic Simulation and Model Tools

As the volume of traffic increased and individual signals were coordinated into signal
networks, the need to measure the effect of signals on the efficient movement of traffic
within networks became more important. Advanced technologies have allowed more
sophisticated analysis techniques for the study of traffic control. The advanced methods to
study traffic control have been researched and developed concurrent with the advanced
controls themselves. Studies to measure the performance of traffic signals have been

underway, in one form or another, since the invention of the signalized traffic control.

Research into the principles of how signals contribute to delays began to take shape in the
1940's and 50's. Many of the early studies of signal control analyzed intersections in
isolation. They focused on the effect of signals on the approach traffic arriving at an
intersection. One of the earliest available studies used mathematical equations to evaluate

starting delay and how it could impact the capacity of an intersection.”

As technology levels increased and the need to study and understand the effects of signalized
traffic control became more apparent, more complex study methods were employed. During
the 1960's, efforts were made to test mathematical models of various timing controls with
the assistance of computers. One of the earliest published studies to investigate the use of
computer simulation, lauded the anticipated results that it could bring to the field of traffic

engineering research.” It was hoped that the use of computer simulation would also help

l6

spur the development of more complex mathematical models constructed to describe the
behavior of traffic flow. This study attempted to describe the physical characteristics of the
intersection, the generation of vehicles, and the acceptance of left turning gaps, in terms of
mathematical models. It was correctly assumed at the time that computer simulation models
could be used to study hours of traffic conditions without effecting actual operating
conditions in the streets. The procedures of these models were coded and simulation trials
executed by a computer to determine the accuracy of the model predictions. The early
research also demonstrated how the state of computer technology at that time did not allow
this study to be completed. The lack of research computers and the frequent breakdown of

existing equipment hampered many early efforts.

A study published in 1963 that continued the initial computer simulation project was more
successful in gaining some output results. The research study was one of the first to use the
results of computer simulation to evaluate various patterns of signalized control“) The
project involved the simulation of some 60,000 hours of traffic behavior to compare some
of the basic relationships between traffic volumes and vehicular delay under various traffic

control and volume conditions.

The lessons learned from early computer simulation studies remain valid today. For
example, the occurance of rare events such as accidents, especially if they were not
recognized, could effect studies involving field observed data. Computer simulation allowed
such events to be included by the use of probability distributions. Experiments that
contained rare events could be tested over and over to focus on different aspects of traffic
behavior. The study pointed out that computer simulation could also prove to be an
extremely valuable tool for the completion of “before” and “after” studies because of the

extreme difficulty in obtaining similar conditions to evaluate a change in control.

Later research studies continued to advance the knowledge base and complexity of computer

simulation applications to traffic engineering. A simulation study completed in England in

17

the 1960's attempted to assess the implementation of certain signal control improvements to
an area before it went into operation?” The study was also one of the first to make use of
simulation to model an area wide network rather than just an isolated intersection. Another
interesting research technique documented in this study involved the use of events which
were observed in the field but not reflected in the original simulation output. The events
were used to reprogram the model to allow it to more realistically simulate the observed

behavior.

The use of computerized traffic simulation and analysis programs have developed from use
as strictly a research device, to very valuable tools in modern traffic engineering practice.
Two programs, SOAP and TRANSYT-7F, were developed independently during the 1970's.
After years of separate use the two programs are often used together for the evaluation and
analysis of traffic signal plans. TRANSYT was developed in England to analyze
coordinated progression of traffic signals on arterial roadways. It helps a users to evaluate
various signal cycle lengths which will promote the movement of traffic through a particular
corridor. The system attempts to promote the formation of traffic platoons and minimize the
number of stops at red lights. The program will also produce space-time diagrams which are
also helpful in the analysis of traffic flow. The Signal Operation and Analysis Package
(SOAP) was developed'for the analysis of single isolated signalized intersections. It was
originally coded at the University of Florida in the 1970's and, like TRANSYT, has
undergone numerous modifications to upgrade it accuracy and usefulness. It can be used to
evaluate traffic signal design alternatives at four-legged intersections with or without
protected left turning phase intervals in the signal sequence, including fixed time, semi-
actuated, and fully actuated control. One of the benefits of the SOAP program is that it
allows the operating characteristics of an individual intersection to be expressed in terms of
specific measures of effectiveness (MOE). SOAP MOE’s include the calculation of various
vehicle approach delays, number of stopped vehicles, fuel consumption, queue lengths, and
other flow characteristics.” Its primary limitation, however, is the lack of a mechanism to

adjust vehicle arrival distributions. The SOAP delay analysis algorithms assume a random

18

arrival pattern in all analyses.

Development of increasingly more detailed and sophisticated traffic simulation continued
through the 1970's and 80's. Simulation has developed into a standard application tool for
the modern traffic engineer. One of the most widely used traffic simulation programs today
is the NETSIM computer model.” NETSIM, as its name suggests, is a microscopic road
network simulation program. It was initially developed by the Federal Highway
Administration in the mid-1970's as a method of modeling traffic events on arterial road
networks. It is extensively used by transportation engineers and planners to analyze various
highway design and traffic planning scenarios. The NETSIM program has continued to
evolve and add more complex features to keep pace with advancements in traffic control
technology. More advanced versions of NETSIM, currently under development, will also
include features to assist in the modeling of adaptive signal control. Unfortunately, these

systems were not available to assist in the completion of this study.

Recently, NETSIM has been incorporated into the TRAF family of traffic simulation
programs. The TRAF modeling family encompasses a variety of independent freeway,
arterial, and rural road simulation programs. The programs within TRAF are ROADSIM,
a two lane roadway simulation model, FREFLO, a macroscopic freeway simulation model,
NETFLO 1 and 2, macroscopic urban network simulation models, and FRESIM, a
microscopic freeway simulation model. The central concept of the TRAF system is a
compatible format of data exchange which allows them to be compatible with one another.
This compatibility allows the various programs to be linked together to form broad-based
multi-level traffic simulation models. These multi-program TRAF models allow the
simulation of large scale traffic networks that may include both freeways and arterial

roadways.

While the TRAF system allows a compatible format for the simulation of separate arterial

roadway and freeway sub-networks, the most advanced simulation tool for the coordinated

l9

modeling of arterial and freeway traffic systems is the INTEGRATION simulation
program?” The INTEGRATION model was developed in Canada in the 1980's and
incorporates many concepts which are very useful for the analysis of urbanized traffic flow.
Primary among these is a strategy to model various forms of traffic assignment. The
INTEGRATION program allows users to assign traffic volumes to a road network based on
various conditions. These could include shortest travel distance, shortest travel time, or
special conditions such as traffic rerouting due to incidents and congestion. Research is also
underway to investigate the addition of dynamic route guidance in the INTEGRATION
model?” A dynamic route guidance simulation model would be a very valuable tool to

advance research in the area of intelligent traveler information systems.

The continued development of more advanced methods of computer simulation has played
a key role in the study and cultivation of many traffic engineering theories. Current
simulation models are able to very accurately predict detailed traffic stream parameters such
as delay at signalized intersections, if the appropriate traffic flow data are known.“5'2°’ A
detailed assessment of current traffic simulation modeling was documented in research
papers by Wang and Hsin‘z'” and Wang and Niedringhausf”) These papers documented
some of the needs and requirements necessary in the next generation of traffic management
simulation programs. The later presented the development process of the THOREAU
(Traffic and Highway Objects for Research, Analysis, and Understanding) program. The
THOREAU model appears to present a promising platform for the construction of complex
traffic simulation models. It makes use of many innovative concepts in the field of computer
programming and data processing, including object oriented programming and parallel
processing. These allow for more detailed simulations of real world events and allow for
faster construction and execution of the simulation models. THOREAU can also be used to

model adaptive traffic signal control”)

The recent interest and funding dedicated to Intelligent Transportation Systems has also

allowed new simulation models to be developed at a more rapid pace. A recent study

20

sponsored by the United States Department of Energy identified some of the best available
traffic simulation models.‘29’ These include freeway and arterial simulation models like
INTEGRATION and the TRAF family of simulation programs. The study also pointed out
the importance of the development of more sophisticated simulation models and some of the
critical requirements that future generations of traffic simulation models should include.
These ranged from the simulation of traffic flow, to the modeling of driver route selection
behavior, to the environmental impacts of traffic activities. The authors indicated that future
models will need to be more “user friendly” and allow users to quickly and accurately model
complex systems which may involve very sophisticated traffic control and guidance
activities. The text stressed the need and importance for such models and the need for further

research into these and related computer simulation programs.

2.3 FAST-TRAC and Related Intelligent Highway Projects

SCATS hardware and computer technology have been implemented on the road network
within Oakland County, Michigan.” 3" The signal improvements are part of the FAST-
TRAC (Faster and Safer Travel - Traffic Routing and Advanced Controls)” system, the
national demonstration project for Advanced Traffic Management Systems
(ATMS)/Advanced Traveler Information Systems (ATIS). The project involves the
conversion of over 1,000 pretimed and actuated signalized intersections to SCATS control.
In 1995, there were about 200 SCATS controlled intersections operating within the Oakland
County system. Traffic flow data are gathered at each intersection and transmitted to a
central control center. This center serves as the heart of the traffic management system. It
houses the SCATS computer program and the various control systems required to execute

the modifications to the signal controllers.

An evaluation study is currently underway of the Troy FAST-TRAC network“) One of the
primary goals is to document the impact of SCATS control on intersection delay. While the

study is not complete, some preliminary data have been compiled. It appears that the SCATS

21

method of ATMS has some tangible benefits over fixed time signals, under certain traffic
conditions. The Troy evaluation study, involved a “before” and “after” technique to evaluate
the system performance. The measure of effectiveness used to compare the “before” and
“after” periods was average stop delay per vehicle at the intersections. Autoscope video
image detectorsm’ were used to record traffic volumes and video taped data were used to
record the stopped vehicle queues on the approach legs. Various recording and simulation
methods were employed to determine the delay. One method involved the use of video
camera data to calculate an average delay per vehicle by timing the stOp delay experienced

by the first, last, and “middle” vehicles in the stopped queues on each of the approaches.

The FAST-TRAC project is only one of several ITS projects which are in operation or under
development in the State of Michigan”) The SCANDI (Surveillance Control And Driver
Information) project is another ATMS project which has been in operation since the 1960's.
It is comprised of a system of loop detectors, closed circuit cameras, and changeable message
signs located at various positions on three major freeways within the City of Detroit. The
system was installed to assist drivers in anticipating congestion and give information on
events such as road construction. An extension of the existing system is currently under
contract. The upgrade will involve the addition of systems for automatic incident detection
and motorist aid. Other Michigan ITS projects include DIRECT (Driver Information Radio
using Experimental Communication Technologies) and MACS (Mainline Automated
Clearance System). These projects involve the implementation of systems for advanced

traveler information and commercial vehicle operations.

Although the primary application of advanced technology for the improvement of highway
transportation has been focused on congested urban settings, the use of emerging ITS
technology has also been proposed for rural areas. One such project is currently under
development in Minnesota. It involves the use of advanced traveler information systems and
advanced transit information services“) The deployment of ATIS systems are of particular

value to tourist traffic. Visiting drivers are in greater need of travel information services

22

because they are often unfamiliar with the local road system. They are also in frequent need
of special interest information like the location of specific points of interest, gas stations, and
road construction zones. Recent studies have also stressed the need for intelligent
transportation systems in rural areas to assist with advanced warning for unexpected road
hazards, speed enforcement, and commercial vehicle navigation and safety.” The 1992
Intelligent Vehicle Highway System (IVHS) Strategic Plan prepared by the United States
Department of Transportation outlined the application of advanced technologies in rural
areas for the identification of hazardous weather conditions on roadways and the

coordination of public transportation services for lower mobility individuals.

2.4 Comparative Studies of Adaptive Traffic Signal Systems

Many studies have been completed which claim to have evaluated adaptive signal traffic
signal systems. Several of the more recent studies have involved the two most widely used
real-time traffic responsive signal systems, SCATS and SCOOT. Past studies have
completed comparisons of these systems against various forms of less sophisticated forms
of signal control like coordinated fixed time actuated control. The depth and level of detail
incorporated into these comparative studies has varied, although none of the previous studies

has used simulation nor stopped delay as a performance measure.

AM
Many studies of adaptive control have been carried out by the creators of SCATS, the

Australian Road Research Board (ARRB) and the Department of Main Roads in New South
Wales. One study evaluated SCATS against various forms of non-adaptive forms of signal
traffic controlesb The study measured the performance of SCATS against the control
characteristics afforded by systems with isolated fixed time signal phasing and TRANSYT
optimized fixed time control with and without local vehicle actuation. The study, however,
documented solely the results of the performance comparison of SCATS and TRANSYT

optimized timings.

23

The ARRB study made their comparison using the floating car travel time estimation
technique to record the “journey” or travel time on each link, the number of stops in each
link, the stopped time in each link, and the amount of fuel used in each trip. The recorded
stopped times were later found to be unreliable, so they could not be used in the analysis.
The accuracy and reliability of the fuel meters also prevented a valid analysis of fuel
consumption data. The study was able to compare the different signal systems in terms of
travel time, number of stops, and a derived “Performance Index.” The Performance Index
was a weighted measure of travel time incorporating the number of stops during the trip. The
study found that on one arterial highway, SCATS resulted in a 23% reduction in travel time
and a 46% reduction in stops over isolated fixed time signals. In the central business district
(CBD) study area, the travel time was not effected and the reduction in stops was 8%. When
compared to Linked Vehicle Actuated (LVA) control, SCATS showed some benefits and
some degradations in the recorded performance measures on the arterial and in the CBD
areas. The comparison of SCATS and TRANSYT optimized fixed times concluded that
SCATS can improve travel time and number of stops from 3% to 18%. The actual
improvement depends upon the type of road system (CBD network, arterial corridor, etc.)

under study.

A similar comparative study carried out by the authors of the previous study evaluated a
precursor to SCATS.” The paper documented a performance study of the Department of
Motor Transport (DMT) signal system against optimized fixed time networks of isolated and
coordinated signals. The study also analyzed bus priority signal preemption under different
forms of control. Using travel time as its measure of effectiveness, the research study
demonstrated that the DMT control strategy showed an improvement of 35% to 39% over
optimized fixed time control during morning and evening peak hours. However, the DMT
system increased the minor cross street delay. The bus priority system under DMT control
showed an improvement of 11% to 26% in bus travel times over the optimized fixed time

operation.

24

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

SCATS does not have.

The study indicated that the two systems are also similar in many respects. They both
implement frequent but small changes in cycle time, phase splits and offset to accommodate
rapid fluctuations in traffic demand. Past field trials have shown that during certain time
periods, fixed time signal control can out perform both of these two systems.“'7"'9’ The study
concluded by stating that neither system is perfect and that both systems could benefit from

additional improvements.

Another comparison of SCATS to other forms of advanced signal control was completed
prior to the installation of the Oakland County FAST-TRAC project."'°’ The study was
conducted to compare four different types of advanced traffic management systems; SCATS,
SCOOT, UTCS, and OPAC. The comparison study was completed to select the type of real-
time adaptive signal control that would be used in the County. The study did not incorporate
field experiments to support the selection decision. Each system were compared on the
merits of certain criteria that was of particular value to the Road Commission for Oakland

County.

25

W

The amount of available published research on comparative studies involving SCOOT
appeared to be greater than that of SCATS. This is the result of SCOOT's greater level of
use in North America. A study was recently completed as part of the Metropolitan Toronto
SCOOT demonstration project. The initial size of the Toronto SCOOT network was 75
signalized intersections. A report of preliminary findings was prepared and presented in
1994.09) The report claimed that the Toronto SCOOT project resulted in a decrease of 6% to
1 1% in travel time along the study corridor during peak periods of traffic demand. The study
also documented travel time improvements that measured as high as 34% on certain arterial
roadways after major sporting events. The preliminary report showed that benefits were also
apparent in vehicular and pedestrian safety. The study did acknowledge that under certain
flow conditions in Toronto, some SCOOT controlled intersections could be out performed
by fixed time control. The study did not, however, document the details of the experimental
design, the analysis techniques, nor all of the data collection methods and performance

measures.

Another recent SCOOT evaluation study in North America was completed in Oxnard,
Califomia."°’ The Oxnard project involved the first installation of SCOOT in the United
States. The report documented an observed improvement in total system delay and number
of stops. Unfortunately, the study was only a cursory report of observed traffic conditions.
The study conclusions were based on the SCOOT data collection statistics for recorded stops
and delay. Therefore, the author stated that the amount of data collected for analysis did not

allow the results to be scientifically corroborated.

Other studies of SCOOT have examined traffic operating conditions under off-peak traffic
flow conditions. One assessed the change in certain performance measures during low flow
conditions.“'” The study used the number of stops that were experienced by vehicles
traveling through both a corridor and grid network of SCOOT governed intersections. The

study was conducted in suburban London, England and used data collected during off-peak

26

times. The study found that in the corridor study area, the number of stops was reduced by
17% compared to fixed time control. In the grid network the study found that SCOOT
control also resulted in a 17% savings when compared to fixed time signal control.
However, the study also showed that during very low flow periods (1:30am to 2:00am)
SCOOT was out performed by vehicle actuated (VA) signal control. This was reasonable
to expect since VA control can eliminate signal phases on the cross-street and SCOOT

cannot.

2.5 Literature Review Conclusions

The review of past published literature into the various areas related to the study has
demonstrated the interest in and importance of advanced traffic control systems, computer-
aided traffic modeling, and comparative studies of the various forms of traffic signal control.
A review of the literature has shown the important role that ATMS will play in more
efficiently utilizing our transportation infrastructure. Today, real-time adaptive signal control
systems like SCATS and SCOOT are available for mainstream use in the United States. As
a result, studies which can assist current and future users of such systems to determine their

expected benefits are extremely valuable.

Unfortunately, the existing published research literature has also indicated the need for
further research into more detailed measures and newer methods to assess the changes which
can be expected from adding SCATS control to a signalized network. Some of the past
studies have conducted comparative studies using route travel time and the stop frequency
field measurements and have shown the benefits of SCATS and SCOOT over optimized
fixed timed systems. However, the study authors have used travel time or the number of
stops as their only performance measures. While these performance indicators are very
helpful, they cannot reveal all of the details of what is occurring in the system. They only
present the “big picture.” The South Lyon evaluation study was proposed to overcome this

existing lack of information.

27

The primary reason for the lack of detailed analysis has been the lack of sophisticated data
collection and analysis methods and tools. Now, traffic flow databases can be compiled
using automatic collection and storage techniques. Traffic flow data can be collected
continuously; at any time and under any flow condition. This is possible because of systems
like Autoscope sensing technology and complex and high-speed computer systems. As a
result of commercially available computer simulation modeling and analysis tools like
SOAP, NETSIM and THOREAU, traffic flow analyses can be completed quickly and

inexpensively under repeatable and controlled conditions.

The South Lyon evaluation study will take advantage of these more sophisticated analysis
tools to complete a comparative study of certain detailed aspects of real-time adaptive signal
control. The research project will use some recent methods of data collection and computer
simulation to model and test the effect of the change to SCATS signal control in South Lyon

using intersection approach delay performance measures.

Chapter 3.
Research Objectives and Approach

It has been hypothesized that the use of advanced traffic management systems, like SCATS,
will result in significant improvements to traffic flow performance within a controlled
network. Prior studies‘4"°""39""’ have demonstrated that SCATS and SCOOT can reduce
route travel times and intersection delay when compared to conventional fixed time systems.
Other studies have shown how adaptive control systems can be used to improve the
operation of transit bus systems or flow conditions during special event situations, like
sporting events. However, the published research literature has also demonstrated a lack of
Significant in-depth research into some of the specific ways that these flow improvements

are attained.

One of the primary reasons for the lack of information about certain flow performance details
is that they are difficult to measure, calculate, and analyze. For example, intersection stop
delay is one of the prime contributors to travel time changes during the day. It is also a
useful measure of system performance because it can indicate the efficiency of signal phase

splits and is used to rate the Level of Service of signalized intersection approaches in the

28

29

Highway Capacity Manual.“7’ Two comparative studies (36'3” that were reviewed in the
preceding chapter documented how the use of SCATS resulted in reduced travel time and
number of stops at red lights. However, neither of these studies used stopped delay as a
measure of effectiveness. The use of travel time and number of stops is valuable in assessing
certain aspects of system performance. However, they do not provide a level of detail
necessary to analyze certain aspects of urban traffic operations. The use of intersection delay
statistics give details about signal efficiency and can better explain the changes which are

brought about by SCATS signal control.

One of the past studies‘36) attempted to use stopped delay as a measure of effectiveness,
however, it was not possible because the recorded stopped times were found to be
“unreliable.” Delay statistics are often difficult to field measure. Timing information for
stops, starts, and turning movements must be collected for all vehicles on an approach.
Reduction of the data is also complicated. Recently developed techniques like video image
sensing and recording have made it possible to collect data in a more simplified fashion.

Unfortunately, even with these advances, data reduction remains a long and tedious process.

For some time it has appeared that the most promising method to overcome some of the
shortcomings of the past research was the use of simulation. Simulation programs like
NETSIM and SOAP provide a very useful means to assess certain traffic parameters under
highly controlled and repeatable experiments. To their disappointment, past researchers have
also found these programs to be limited in their ability to simulate adaptive traffic signal
control. The Federal Highway Administration has recently contracted to upgrade the
NETSIM program to incorporate an adaptive signal control interface mechanism. Within
this new system, traffic flow information can be relayed to an external program which can
affect signal phase timing changes within the NETSIM simulation run. The process involves

a continuous feedback loop between the movement of traffic and modification of signals.

30

Due to the proprietary nature of most adaptive signal control systems and the inherent
difficulties involved in making modifications to the existing NETSIM code, the realization
of a NETSIM based real-time control simulator remains in the developmental stage. Until
this new system is perfected, the simulation of real-time adaptive traffic signals remains a
cumbersome process. However, the simulation of real-time traffic signal control can be
accomplished by using a series of incremental step models which modify signal cycle and

phase splits on a cycle-by-cycle basis.

This study will take advantage of the incremental simulation process to model a network of
real-time adaptive traffic signals. This research will advance the understanding of adaptive
signal control by completing an evaluation of delay time characteristics within the South
Lyon ATMS project area. The results of the study will provide new evidence to evaluate the
claim that improvements to certain traffic delay parameters can be expected when SCATS
is introduced to an area. The study will test the hypothesis that significant and measurable
changes (improvements) to certain delay characteristics are likely after the implementation

of SCATS. To attain this goal, several research objectives will be met.

3.1 Objectives

The prime objectives of the South Lyon ATMS evaluation project are to record, document,
and analyze the changes which occurred to the traffic flow conditions within the South Lyon
signalized road network as a result of the addition of the SCATS signal control management
system. To accomplish this objective, this study will incorporate a detailed process of data
collection, simulation model construction, and statistical analysis. Data collection will be
cam'ed out using various means, including field observation, and video image sensing. The
analysis of the data will be completed with the assistance of the SOAP isolated signalized
intersection simulation and analysis program. Several SOAP models will be constructed for
each intersection during the “before” and “after” analysis periods to calculate certain

measures of operational performance recorded during the separate analysis periods of the

31

study. After the performance measures for each period are calculated, a comparative analysis
of the two intervals will be carried out using suitable statistical testing procedures. The
analysis will indicate whether the measured change in operating conditions after the

implementation of SCATS was statistically significant.

To determine the extent of the change in quantifiable terms and document the results, the

study will address the specific questions shown below.

1) What is the difference in the total intersection delay time in the South Lyon road
network for the existing non-adaptive, pretimed, signal scheme and the SCATS
adaptive control mode? Is the measured difference in the total intersection delay time
statistically significant or is it a result of random changes which occur within the
traffic stream?

2) What is the difference in the average approach delay times at selected intersections
within the South Lyon road network between the existing non-adaptive, pretimed,
signal scheme and the SCATS adaptive control mode? Is the measured difference
in intersection approach delay time statistically significant or is it as a result of

random changes which occur within the traffic stream?

The research questions will establish the measures of performance to compare the two
different signal control strategies during the “before” and “after” periods of the study. The
improvement brought about by changes in traffic control can be assessed using many
different measures (i.e., safety, delay, travel time, etc.). The two measures selected for this
study were based on their relevance to the goal of reducing traffic delays in the community

of South Lyon. Each is described in detail in the following sections.

1W3):
The total delay at an intersection is defined as the difference in travel time experienced by

a vehicle as it is affected and unaffected by the traffic control at an intersection. It includes

32

the “lost” time due to deceleration and stopped delayf‘m It is also the composite of other
delay sources, including queue delay at over-saturated intersections and “unnecessary”
stopped delay. Information acquired from a recently completed study of a SCATS
implementation project in Troy, Michigan has shown that the unnecessary stopped delay
could be of particular significance to the South Lyon study. Unnecessary stopped delay is
defined as “the portion of the stopped delay which occurs when there is no vehicle entering
an approach on the opposing legs of the intersection.”“'” During the SCATS implementation
project in Troy, many intersections received exclusive left turn lane signal control where
none had previously existed. As a result, left turns could be completed only under a
protected left turn phase. Status reports performed after the change demonstrated the
existence of added delay and user dissatisfaction at locations where permissive left turns
were no longer allowed. There are two intersections with protected/permissive left turn
phasing in South Lyon. It is expected that the total delay statistic will measure the

unnecessary delay at these intersections.

Many different analysis techniques for the calculation of total intersection delay have been
developed“) The most widely used employ mathematical models to calculate various
aspects of total delay which are the result of vehicle arrival patterns. The comparison of total
delay during the “before” and “after” periods of this study will be accomplished through the

use of the SOAP simulation model.

LiLAanQasthelax

The research study will also assess the change to the specific portion of intersection delay
known as approach delay. The approach delay is measured separately at each approach. It
is defined as the length of time that vehicles approaching the intersection from a particular
direction are delayed due to a red signal phase and/or a stopped queue in front of them which
prohibits their travel through the intersection.““’ To express approach delay in more useful

terms it is often converted to average stopped delay per vehicle on a given approach during

a specified time interval. This is the measure by which the Highway Capacity W1”)

33

assigns a Level of Service rating to signal controlled intersections.

The average approach delay per vehicle can be measured in various ways, using various
means. One method is to record the delay experienced by vehicles located at three positions
within the st0pped queue. Delay measurements are taken from the first vehicle arriving after
the start of the red phase, the last vehicle to join the stopped queue, and that of a vehicle
lying somewhere in the middle of the queue. The three delay times are averaged and
assigned to each of the total number of vehicles in the queue. To study an hour long period
this process of delay measurement and assignment would need to be completed for each
approach during all signal cycles within the hour to determine the average delay during that
time period. The average approach delay times during the “before" and “after” periods of

this study, will be accomplished automatically using SOAP traffic simulation models.

3.2 Approach

The experimental approach used to determine the extent of the changes after the
implementation of SCATS, will involve a modified “before/after” analysis. A typical
“before/after” analysis technique requires the collection and study of like data elements
during time periods both before and after a change is made to a system. The premise of this
type of study states that when all other factors involved in the system are controlled for, any

differences which are recorded are attributable solely to the changed variable.

For reasons that will be detailed later in this chapter, the research methodology of this study
will require a modification of the traditional “before/after” design. The analysis of the
“after” condition will take place prior to the analysis of the “before” condition. The “after”
study will analyze traffic conditions in South Lyon after the implementation of SCATS
traffic signal control. The “before” analysis will evaluate the same traffic conditions under
optimized fixed time traffic control. The same measures of operational effectiveness; total

intersection delay and average intersection approach delay will be used to assess both

34

”before” and “after” periods.

As is often the case with many research methodologies, the “before/after” technique of
analysis contains some inherent weaknesses. Some of the shortcomings of this technique
stem from the lack of total control over the fixed factors involved in the experiment. The
experimental assumption is that all factors remain constant and all recorded changes would
be the result of the change to SCATS signal control. Unfortunately, traffic volumes, turning
movements, and incident occurrences can vary significantly from hour to hour, day to day,
and week to week. Therefore, it would be impossible to hold all of the factors constant in
the field. To restrict the amount of variability in these elements during the separate periods
of the analysis, computer simulation will be used. Simulation will allow controlled and
repeatable experiments to be performed on the traffic stream. With the use of the SOAP
intersection simulation and analysis program, it is also possible to calculate and present
detailed information about the performance of a signal at an intersection. To accomplish the

same level of detailed analysis in a field setting would be nearly impossible.

The research strategy will incorporate four primary steps of data collection, ”before” and
“after” simulation experimentation, and a comparative statistical analysis. Each of the four
steps will build upon the results gained from its predecessor. The first step will involve the
collection and sorting of all data elements required to construct and execute the comparative
traffic simulation models. The second and third steps will include the assembly of the
models to conduct the “after” and “before” analyses. The final step of the study framework
will complete the statistical comparison of the “before” and “after” measures of
effectiveness. Detailed descriptions of each of these steps is presented in the following

sections.

Win
The data collection phase of the study will involve the collection of both physical elements

and traffic flow parameters of the South Lyon traffic network. The physical elements of the

35

system include features of the South Lyon road network such as the number of road
segments, intersections, approach lanes and use configurations, as well as traffic control
items such as posted speed limits. The original fixed time signal timing information will also
be recorded during this time and subsequently verified against Road Commission signal log
records. All of these data elements will be collected manually, primarily through visual

inspection.

The collection of traffic volumes will not be completed in the field, rather they will be
computed from SCATS data files. These files are created by the SCATS control software
and are stored at the Road Commission for Oakland County traffic operations center in Troy,
Michigan. Traffic flow information is collected at the intersections in South Lyon by the
Autoscope video image detection system. The Autoscope detection architecture allows the
presence of vehicles to be acknowledged at each approach to the SCATS intersections.
Traffic information is collected primarily for the purpose of selecting an appropriate timing

plan for the intersection. However, this information can also be stored for later analysis.

A typical SCATS data file contains several important pieces of data which can be used to
analyze the operation of the signal and traffic conditions. It is made up of a stream of data
records that include a cycle-by-cycle history of signal phase splits, cycle length, and approach
degree of saturation. Using relatively simple computer programs, the files can be sorted and

the pertinent traffic volume and signal timing information extracted for use.

The collection of traffic volume and signal timing data for the “after” analyses will not be
conducted immediately after the change to SCATS signal control in South Lyon. The
collection of the data will begin after an “acquaintance and adjustment” period has taken
place. The adjustment period will be required for several reasons. Most important, it will
allow drivers to adjust to the new signal phasing strategies. SCATS is not only real-time
adaptive, it may also adjust, rearrange, or eliminate certain phases from the signal cycle.

Thus, drivers who were familiar with the fixed time signal operation will be allowed to adjust

36

their driving habits to fit the new SCATS control plans. The adjustment period will also
allow the Road Commission for Oakland County time to correct and “fine tune” the

operation of the SCATS system.

6‘ ,9

The second step of the experimental approach strategy will be to complete the “after”
analysis study. The “after” analyses will be conducted using the stored signal and traffic data
and physical features of the South Lyon collected in the first step. A SOAP simulation
model of the traffic conditions for a one hour interval will be coded with this information.
Separate models will be constructed for each of the six SCATS controlled intersections with

the network.

Unfortunately, the SOAP modeling environment, nor any other currently commercially
available traffic modeling software, allow modifications of traffic signal timings in response

to traffic volume to be made during program execution.

To overcome this lack of sophistication, the “after” analyses will be conducted using a series
of incremental SOAP simulation runs. Each SOAP model series will correspond to a time
interval as it was recorded in South Lyon. Each increment in the series will represent a
single signal cycle time “slice.” The SCATS implemented signal phase split plan, offset, and
cycle length for each one cycle increment recorded in the data file will be used to code the
model. The length of each time slice will depend on the length of signal cycle length which
was selected by the SCATS control program. The output data from this sequence of

simulation runs will be compiled to analyze the full one hour period.

The approach traffic volumes used in the “after” simulation models will be augmented by
turning movement volumes as they occurred in the field. This is useful since the SCATS
control structure implements signal timing plans based on a “saturation equalization”

concept. Some of the intersections have phase splits that are modified based on the degree

37

of saturation for critical movements on the constituent approaches. Additionally, SCATS
implements cycle length and phase splits changes based on the efficiency of the preceding
cycle. If one particular movement was below saturation during a cycle, it will in theory,

receive less green time during the next cycle.

The final phase of the “after” analysis will be to use the “after" traffic model output to
calculate the various measures of effectiveness (MOE). These measures will include the
total intersection delay, average intersection stopped delay, and the average travel time for
selected routes within the network and will serve as the basis for comparison against the

“before” simulation results.

‘6 ’9

Step three will encompass the “before” traffic analysis. The “before” simulation model will
represent the traffic flow conditions in South Lyon prior to the implementation of the SCATS
control system. Instead of using field traffic volume data collected in South Lyon under the
fixed time signal settings, the “before”analysis will use the same traffic volumes as the
“after” study. This arrangement will insure a direct comparison of conditions. The traffic
signal timings will be modified from the “after” models. An optimized version of the RCOC
fixed time signal phase split, cycle, and offset timing plans will be used to process the traffic
demand. Separate “before” models for each intersection will be executed to calculate the
signal MOE’s. However, each run will model a full one hour period instead of the multi-step

approach used in the “after” analysis.

An optimized version of the original RCOC fixed timing plan will be used so that an
objective comparison can be made between SCATS and fixed time control. The original
Road Commission pretimed signal phase split, cycle, and offset timing plans were developed
to efficiently accommodate peak and off-peak traffic demand using a system of historical
traffic volume and flow monitoring. The original RCOC signal timing plans may not

represent the time settings for the greatest reduction in delay for traffic volume recorded

38

during the study time periods. The fixed signal timings will be optimized using the SOAP
phase split optimizer before being used to calculate the output MOE’s. The use of these
optimized timings will help to demonstrate the difference in delay conditions between
SCATS and optimized fixed time control. Any improvement which is gained from SCATS
will represent the result of its utility rather than the result of a comparison to a poorly timed

fixed time system.

Separate simulation models representing each of the six SCATS controlled intersections will
be constructed and executed during the “before” analyses. In contrast to the segmented
“after” models, each of the “before” simulation models will be executed as a single
continuous model. The one hour approach traffic volumes will be the same as “after” study
and turning movement percentages will use the same distribution as the “after” analysis.
Since the signals timing plans are fixed they will not be influenced by the level of traffic.

The sole difference between the two analyses will be the traffic signal timings.

The final phase of the “before ” analysis will be to determine the SOAP “before” traffic
model output measures of effectiveness. As in the “after” analysis, the measures will include
the total intersection delay, average intersection stopped delay, and the average travel time
for selected routes within the network. After these figures are calculated they will be

compared in the final phase of the study.

3 2 | C . [E 1 1 C l .
The final step of the project will be to compare and document the observed differences in the
measures of effectiveness between the “before” and “after” periods. The comparison will
be performed using statistical testing procedures which have been developed specifically for
evaluating the results of experiments identical to that which has been proposed in this study.

The primary statistical testing procedure that will be used to compare the “before" and

“after” simulation output data will be the t-test. The t-test incorporates a procedure which is

39

ideal for making paired comparisons of two data sets, especially “before/after” studies. A
comparison of the mean total delay, stopped delay, and average travel time in the “before”
period will be made against the mean total delay, stopped delay, and average travel time in
the “after” period. The t-test allows a statistical determination of the degree of difference
between the two study times to be calculated. It will also allow the significance of the

difference to be determined to a specified level of confidence.

The t-test will be used to evaluate the hypothesis of the study; that SCATS control will result
in a significant change to the average values of the performance measures in the South Lyon
system. The test procedure will be based on the mean and standard deviation of the delay
and travel times determined for the “before” and “after” analyses. The t-testing method
makes it possible to determine if the change is attributable to SCATS or if it is due to random

variations in operating conditions of the traffic.‘45‘“”

Chapter 4.
South Lyon Traffic Analysis

The analysis of traffic flow conditions before and after the implementation of SCATS traffic
signal control was conducted using mathematical and computerized modeling and analysis
techniques. Currently, there are several computer based analysis packages which can be used
for the evaluation of traffic signal performance. The levels of input and output data
complexity found in these packages vary significantly. One commercially available traffic
signal analysis tool is the Arterial Analysis Package (AAP). The AAP incorporates several
different computer modeling and analysis programs that can evaluate different aspects of
signalized traffic flow. The intent of the AAP system is to coordinate the use of these
component programs to permit the simulation, analysis, and comparative evaluation of
different signal timing strategies for a network of traffic signals. One of the constituent
programs is TRANSYT-7F. TRAN SYT-7F is used for the analysis of arterial traffic flow.
It assists in the selection and evaluation of traffic signal offsets and cycle lengths. Another
component of the AAP is the Signal Operations Analysis Package (SOAP). SOAP focuses

on individual intersections. It allows automated comparisons and evaluations of various

40

41

intersection signal timing strategies.

Currently, neither the AAP, NETSIM, nor any other existing traffic modeling system permit
a detailed analysis of real time adaptive traffic control systems like SCATS. To overcome
the current lack of analysis tools, a new technique was developed to model the “before’ and
“after” delay conditions of the South Lyon traffic network. This new technique uses a
combination of the SOAP signalized intersection traffic analysis package, existing
mathematical models, and a new computational computer program developed specifically

for the processing SCATS output data files.

The goal of this methodology was to use it to construct a set of traffic analysis models
capable of simulating the intersection approach delay conditions of the South Lyon system
before and after the implementation of SCATS. While the SOAP program could not be used
directly for this purpose, the new analysis procedure was developed based on the SOAP
delay computational methodology. A number of steps were required to create and make use
of this analysis procedure. The following sections of this chapter describe the process and

assumptions that were used to create the various models and present the results.

4.1 Data Collection

The first phase of the model development involved the collection and reduction of all
elements necessary to describe the conditions at the South Lyon intersections. The SOAP
program and its constituent equations are fairly detailed in their analysis of traffic conditions
resulting from various signal timing strategies. By contrast, the input data requirements are
not nearly as detailed. As a result, the data required to code and execute the models were
relatively easily obtained. The acquisition of additional data to compare the model results

to actual traffic conditional was also required.

42

The collection of data fell into three primary categories. The first category included the
SCATS traffic signal timing and the traffic volume information. These data were collected
remotely and stored automatically by the FAST-TRAC SCATS system. The second category
of data were collected by more traditional methods of direct observation. This data included
elements such as intersection approach geometries and the fixed time signal plans. The last
set of data collected for use in this study were collected as part of related SCATS evaluation
studies in South Lyon. This information was used primarily for the comparison and

verification of the model output data.

The collection of the "before" traffic volume and traffic signal timing data took place during
the spring, summer and fall of 1995; prior to the implementation of the SCATS signal
control and Autoscope video imaging system. The fixed time signal timing data was
collected during a field visit on April 22, 1995 and verified against Road Commission for
Oakland County signal log records. There were six signalized intersections within the South
Lyon street network. They are the intersections of: Pontiac Trail with Nine Mile Road,
Reynold Sweet Parkway, McHattie Street, Ten Mile Road, and Eleven Mile Road. The only
signalized intersection not in the Pontiac Trail corridor was the Ten Mile Road/Reynold
Sweet Parkway intersection, located approximately 1,000 feet east of Pontiac Trail. As part
of the FAST-TRAC project, the control of all six of these intersections were converted to
SCATS control. A street map showing the locations of each of the study intersection was

presented earlier in Figure 1.2

The “before” signal cycle lengths, phase patterns, and intersection approach lane geometry
are illustrated in Figures 4.1a, 4.1b, and 4.1c. The cycle lengths at all intersections, except
for the Pontiac Trail/Eleven Mile Road signal, were 80 seconds. At this location the traffic
signal was set to a cycle length of 70 seconds. Thus, there was no coordinated progression
on Pontiac Trail between Ten Mile and Eleven Mile Roads. All signals operated under a six

phase operation except for the signals at the Ten Mile Road intersections with Pontiac Trail

 

LANE USE

.111.

.0”

 

ELEVEN MILE ROAD 2

% SIP
j]

 

PONTIAC TRAIL

SIGNAL TIMING

 

MOVEMENT Tm:
”“55 mrsnvu.

N/S E/W

 

 

I GREEN RED 34 SEC.

 

2 AMBER RED 4 SEC.

 

3 RED RED 1 SEC.

 

4 RED GREEN 25 sec.

 

5 RED AMBER 4 SEC.

 

5 RED RED 1 SEC.

 

 

 

 

 

 

CYCLE LENGTH: 70 SEC.

 

 

LANE USE

Elk

Shh

 

 

31>

J
T?

 

f

PONTIAC TRAIL

'0 Ito Ito

 

CYCLE LENGTH: 80 SEC.

FIGURE 4.10

 

SIGNAL PHASING DIAGRAMS AND
LANE USE CONFIGURATIONS

 

 

 

SOUTH LYON ATMS EVALUATION STUDY

now 51"! mm"
”MIMI 0' GI. IN WIN. m

 

 

LANE USE

 

EZ/W

 

TEN MILE ROAD => I

REYNOLD SWEE

mmv ' i fl
W

 

SIGNAL TIMING

 

MOVEMENT TIME
PHASE INTERVAL
scum an E1 /w

 

 

I GREEN RED RED 12 SEC.

 

2 AMBER RED RED 4 SEC.

 

3 RED RED RED 1 SEC.

 

GREEN
4 RED (pgfiugs. GREEN so sec.
INT)

 

GREEN

 

 

 

5 RED (pew.s,AMDER 4 SEC.
LT.)
GREEN
5 RED {pacing RED 4 SEC.
LT.)
. AMBER
7 RED (AMBER RED 4 SEC.
L.T.)
8 RED RED RED 1 SEC.

 

 

 

 

 

 

 

CYCLE LENGTH: BO SEC.

 

 

LANE USE
W

11%

 

PONTIAC TRAIL

T n1r>

REYNOLD SWEET
PARKWAY

SIGNAL TIMING

 

MOVEMENT TIME
PHASE INTERVAL
N/S E/W

 

 

I GREEN RED 54 SEC.

 

2 AMBER RED 4 SEC.

 

3 RED RED 1 SEC.

 

‘ RED GREEN 18 SEC.

 

5 RED AMBER 4 SEC.

 

5 RED RED 1 SEC.

 

 

 

 

 

 

CYCLE LENGTH: 80 SEC.

FIGURE 4.1 b

 

SIGNAL PHASING DIAGRAMS AND
LANE USE CONFIGURATIONS

 

 

SOUTH LYON ATMS EVALUATION STUDY

WSTAIEMIY
”WUOKWWMM

 

 

45

 

 

 

LANE USE SIGNAL TIMING
I
MOVEMENT TIME
”“55 INTERVAL
N/s E/w

 

 

 

£3 1 GREEN RED 45 SEC.

%1 k) f 2 AMBER RED 4 SEC.

NINE mu: ROAD 9 3 RED RED 1 SEC.
__ i] i? 4 RED GREEN 22 SEC.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SIGNAL PHASING DIAGRAMS AND
LANE USE CONFIGURATIONS

=1 5 RED AMDER 4 SEC.
1?
.-
0 6 RED RED ISEC.
S
:2; CYCLE LENGTH: Do SEC.
D.
SIGNAL TIMING
MOVEMENT TIME
”“55 1NTERVAL
N/S E/w
LANE USE "
W ‘ “'5‘" "5° “ “C'
2 AMDER RED 4 SEC.
fig 3 RED RED iSEC.
4%; k) f 4 RED GREEN 2:: SEC.
MCHATTIE STREET 9 5 RED AMBER ‘ 55°-
__ i] i? 6 RED RED 1 SEC.
S; _, CYCLE LENGTH: so SEC.
E
p.
0 FIGURE 4.1::
S
.—
2
o
l

 

SOUTH LYON ATMS EVALUATION STUDY

room STATE mm"
“PM“! 6 Cl NO mm m

 

 

 

46

and Reynold Sweet Parkway. At these two intersections the signal timing configurations
allowed for separate left turn phases on each of the approach legs. The road geometry of the
link segments and intersection approaches was measured at the same time as the traffic signal
timings. All of the approaches to the study intersections featured an exclusive left turn lane.
With the exception of the south approach to the Pontiac Trail/Eleven Mile Road intersection,
all approaches to the study intersections also featured shared through/right turn lanes. None
of these approach geometries were altered during or between the "before" and "after" phases
of the study. Originally, there was no exclusive left turn phasing at the Pontiac Trail
intersection at Nine and Eleven Mile Roads. After the completion of the FAST-TRAC
project the north and southbound left turn movements at each of these locations were given

permissive/protected left turn phasing.

Most of the road segments, or links, between the intersections featured two lane cross
sections. The exceptions to the two lane cross section were segments of Pontiac Trail and
Reynold Sweet Parkway. A continuous center lane for left turns exists on Pontiac Trail
between Ten Mile Road and Reynold Sweet Parkway. The common center left turn lane also
extended north of Ten Mile Road for 500 feet. A continuous center lane for left turns existed
on Pontiac Trail for a distance of approximately one half mile north of Nine Mile Road. The
center turn lane also extended south of Nine Mile Road for a distance of approximately 700
feet. At the intersection approaches, the standard two-lane cross sections increase to more
than two lanes. Wider approaches are able to accommodate exclusive and shared

through/tum lanes.

A schematic diagram of the Reynold Sweet Parkway and Ten Mile Road intersection is
shown in Figure 4.1b. The position of the railroad track at this location required the
construction of a "double signal" at the westbound approach to the intersection. Westbound
traffic is controlled by a multi-phase signal allowing a pre-emptive red for train crossings or

permissive/protective left turn when no trains are expected.

47

The posted speed limits varied in the South Lyon network. In the central business district
vicinity of the Ten Mile Road/Pontiac Trail intersection, speeds were restricted to 25 miles
per hour. Posted speed limits on Pontiac Trail are 35 miles per hour immediately outside the
core commercial district and 45 miles per hour in the vicinity of Nine and Eleven Mile

Roads. The posted speed limit on Reynold Sweet Parkway was 35 miles per hour.

The collection of the "after" traffic volume and signal timing information took place during
the week of May 6th through May 10th, 1996. The "after" data elements of the system were
composed primarily of the traffic volume and traffic signal timing features of the system.
Both of these key statistics were collected automatically by the SCATS data processing
system. In addition to adaptive signal control, the SCATS system has the capability to
collect and store a number of important details relating to the flow of traffic and control of

signals at the intersection.

Each of the signals controlled by SCATS incorporates a system of video imaging cameras
which are positioned to record the presence of vehicles for critical movements at the stop line
locations. A "critical movement" is a left turn, right turn, or through movement which is
allotted green time during the signal cycle. At a minimum, through movements must be
recorded to allot green time to the approaches. At locations where a separate left turn
phasing is used, Autoscope cameras are added to record the left turn traffic volumes. No
right turn traffic volume information is collected in South Lyon. The right turn and through

traffic both use the through lane and are unopposed by conflicting traffic movements.

A typical SCATS output file contains a wealth of information. A segment of the Ten Mile
Road/Pontiac Trail intersection SCATS output data file is shown in Figure 4.2. The SCATS
output file consists of a stream of cycle-by-cycle information containing key performance
statistics for each critical movement on the constituent approaches. The first line of each

record shows basic information such as the date, time, region, and location of the

IfigurelLZ

48

Sample Segment of SCATS Output File

Pontiac Trail/Ten Mile Road Intersection

8-MAY-96 06:59 WLAKE

INT SA/LK PH
35 s 7 ' 2
35 s 8 ' 1
35 s 9* 4
35 s 10* 3
35 s 11 ' 5
35 s 12 ' 6
35 s 13* 7
35 s 14* 8

8-MAY-96 07 01 WLAKE

INT SA/LK PH
35 s 7 - 2
35 s 8 ' 1
35 s 9* 4
35 s 10* 3
35 s 11 ' 5
35 s 12 ' 6
35 s 13* 7
35 s 14* 8

8—MAY-96 07 02 WLAKE

INT SA/LK PH
35 s 7 ' 2
35 s 8 ' 1
35 s 9* 4
35 s 10* 3
35 s 11 ' 5
35 s 12 ' 6
35 s 13* 7
35 s 14* 8

8—MAY-96 07 03 WLAKE

INT SA/LK PH
35 s 7 ' 2
35 s 8 ' 1

‘ 35 s 9* 4
35 s 10* 3
35 s 11 ' 5
35 s 12 ' 6
35 s 13* 7
35 s 14* 8

SS 1

PT! DS
42! 41
42! 39
10! O
10! O
18! 72
18>1l9
11! 0
11! 54
SS 1

PT! DS
41! 38
41! 37
10! 0
10! O
18! 69
18>115
12! O
12! 52
SS 1

PT! DS
39! 22
39! 43
10! 0
10! 0
27! 14
27! 51
10! O
10! 0
SS 1

PT! DS
23! 51
23! 87
11! O
11! 0
27! 48
27! 57
10! O
10! O

PL1.3
VO VK!
8!
9!
O!
O!
7!
10!
0!
3!

I5

<'U
g1.

OOQNOOCDIbOt“WOQNOOQQOL—‘NOKOUIOOChm

8!
8!
O!
O!
6!
10!
O!
3!
4.3
VK!
4!
9!
O!
O!
2!
6!
O!
O!
PL4.3
VO VK!
6!
11!
O!
0!
7!
7!
0!
0!

P
V

OOQWOOkOIb

PV
DS

4.3 CL
vo VK!
-I

—!

—!

-!

-!

3 3!
-!

-!

4.3 CL
vo VK!
-I

-!

-!

-!

-!

3 3!
-s

-!

4.3 CL
vo VK!
_I

-!

-!

-!

-!

0 0!
-1

-!

4.3 CL
vo VK!
-I

-!

-!

-!

-1

0 0!

80+0

DS

VO

RL
VK!
—!
-!
-!
-!
—!
-!
-!
-!
RL
VK!
-!
—!
-!
—!
-!
-!
-!

_I
RL
VK!
-l
-!
-!
-!
-!
-2
-!
-!
RL
VK!
-!
-!
-!
-!
—!
—!
-!

86
DS

DS

' SA 12 D8119
VO VK!ADS

-I
O
-I
O
I
-I
.

35
31
O
O
46

->102

0
34

SA 12 D8115
VO VK!ADS

37
32
0
O
60

—>110

O
41

' SA 12 DS
VO VK!ADS

' SA

31
4O
0
O
45
87
0
29

8 DS

VO VK!ADS

39
61

51

87

49

intersection. It also contains signal control information like the current mode of operation,
tactical or strategic, the primary split plan and cycle length, the controlling strategic
approach, and the intersection degree of saturation. The third through tenth lines reveal the
important statistics for each of the eight strategic approaches. Strategic approach one and
three represent the through and left turn movements for northbound Pontiac Trail. The
through and left turn information for south, east, and westbound approaches are shown as

lines 2 and 4, 5 and 7, and 6 and 8, respectively.

The first column showing “35" is the code number for the Pontiac Trail IT en Mile Road
intersection. The second column, beginning with the letter “S,” indicates the strategic
approach number and vehicle group. The next column, under the heading “PT,” indicates the
length of the green phase dedicated to each movement. The following column, under “DS,”
indicates the degree of saturation for the critical approach during that cycle. The next
column, under the heading "V0, " shows the actual volume of traffic which was recorded at
the approach location by the Autoscope sensor. The next column in the group, designated
with the heading "VK" indicates the number of vehicles estimated by the SCATS control
program when the algorithm selected the appropriate green phase length. The final column
for each of the critical movement lines indicates the average degree of saturation for the
movement. The average is calculated from the degrees of saturation recorded for the current
and previous two signal cycles. This process is then repeated for each successive signal

cycle.

Since all six of the signalized intersections were converted to SCATS control, all of the
necessary traffic volume information was collected remotely, using the SCATS/Autoscope
data processing system. Unfortunately, not all of the intersections were equipped to collect
all of the data that was required for analysis in this study. The SCATS configuration at the
intersections of Pontiac Trail at McHattie Street and Reynold Sweet Parkway did not allow
for the collection of left turn movement data. The traffic volumes collected by the SCATS

system at these locations are limited solely to the through movements. Using this

50

information signal timings are interpreted by SCATS in the local controller. The
coordination of the cycle length and offset between adjacent intersections is computed at the
Troy traffic operations center. The lack of recorded traffic signal output files for these two
intersections made the delay parameter assessment impossible in this study. Delay
parameters were calculated only at the four intersections of Pontiac Trail and Nine Mile
Road, Pontiac Trail and Ten Mile Road, Pontiac Trail and Eleven Mile Road, and Reynold
Sweet Parkway and Ten Mile Road.

W
Interest in the performance of the FAST-TRAC system has resulted in the completion of a

number of concurrent studies. As a result, many different types of traffic data have been
collected by the Road Commission for Oakland County and other research teams conducting
performance evaluation studies in South Lyon. One particular study‘”’ involved the
collection and analysis of empirically recorded approach traffic volume and delay data. This

information was particularly useful in this study.

The collection of this intersection approach directional turning movement and delay data
took place prior to the activation of SCATS in December, 1995 and later after the activation
of SCATS in May, 1996. The data were collected with the use of video cameras positioned
to record the arrival characteristics of oncoming traffic. Specific approach movements at the
intersections of Nine Mile and Eleven Mile Roads with Pontiac Trail were video taped on
Wednesday, December 6th, Thursday, December 7th, and Friday, December 8th 1995. The
delay data was recorded during the hours of noon to five o’clock pm. Taping during the
usual peak hour of 5:00 pm. to 6:00 pm. was not possible due to the low visibility

conditions after 5:00 pm.

Two cameras were positioned at the southeast comer of the Nine Mile Road/Pontiac Trail
intersection. The cameras were aimed to record oncoming traffic on the north and west

approaches to the intersection. Two different cameras were positioned on the southwest

5]

corner of the Eleven Mile Road/Pontiac Trail intersection. These cameras were aimed to
record oncoming traffic on the north and east approaches to the intersection. Data reduction
took place when the video tapes were reviewed using a television, video cassette recorder,
and stop watch. Although time consuming, this method of data collection and reduction
resulted in accurate and detailed time measurements of individual vehicles. Tapes could be
played, rewound, and replayed to record the arrival and delay time Characteristics for
simultaneously arriving vehicles. To accomplish the same results in a field setting would
have required up to eight people per approach. Summarized results showing the “before”

and “after” volume, delay, and signal information is presented in Table 4.1.

4.2 Development of the Traffic Analysis Models

The original intent of this study was to use the SOAP traffic signal analysis software to
directly model the "after" delay conditions in South Lyon on an incremental, cycle-by-cycle,
basis. Unfortunately, certain aspects of the SOAP program make the use of this approach
impossible. To complete an affective comparison, a new computerized methodology was
developed. After further data analysis it was determined that this system would also have
to be used to analyze the “before” conditions. This section details the methodology of the
SOAP delay calculation procedure as well as highlighting some of the limitations that
prohibited the use of the SOAP program for the completion of this study. The development

of the new analysis methodology will also be described in this chapter.

IZIII SQIEE II 1.. .
SOAP was developed for the analysis of traffic signal design alternatives at four-legged

intersections with or without protected left turning phase intervals in the signal sequence,
including fixed time, semi-actuated, and fully actuated control. One of the benefits of the
SOAP program is that it allows the Operating characteristics of an individual intersection to

be expressed in terms of specific measures of effectiveness (MOE). SOAP MOEs include

52

wctouiwcm _S=oE=o.:>:m 2:. :30 no Ens—Enum— Szfloth 83w Sawing ”350m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mg: hmdm 29303 .3 on 52832:— uwfio><

6:- omén 36m «3:. mm. _m mm E5. :3. . L
IE. «0.2 we? 3.: «E: o E: :3 2:25.58
vm; ov. _N o _ .vm um; _ 3:— cv nwzoE-r ucaoofizow

Gaol 2:2 5.5—”.— EE =95. Caz—Sm
no.5 0:: 20383 .3 on 550225 omfio><
E .N 9‘.— vwém 3.2 3.9. 2.3 2 mm mm ow oi 03 ES. :3 ucsoemalmj
.2. .2- 33 8.3 8.3 2.: w -2. 3 8 mm 3. E5... :3 258558
_ _.~ 9: Ran-N N~.wm whm. 2.2 On wv mw em «on new swash; 65525209
18.2.... $.83: 13.2.. _.u._o..un.. 12.3.. .222... 5.2%.. ._u._£on.. Luca: zeowon: Luca: ..E£un..
RE. 105 20383 05:0 £0383 80$ DE:- Auomv Ewan... €5.33 18¢ 0:2 oEZ “Eu :9; Caz—5.—
»EDD 30,—. E DEF owacu>< >200 owEo>< 520 omfiu>< v.26 owfio>< uEEo> owflo><

 

 

 

 

 

 

 

 

 

 

 

:55 =2§=§u 8.5 956m .63 5:8

 

 

55 689.9 62:86: 23... 2. 2.3.

53

various vehicle approach delays, number of stopped vehicles, fuel consumption, queue
lengths, and other flow characteristics.” A further benefit to the system is that it allows
these MOE to be calculated for relatively short time intervals. Its design allows a user the
ability to analyze up to 48 separate time increments in a single program execution. Thus, it
was originally hoped that a series of forty five 80 second time increments could be used to

evaluate one hour of delay at intersection with an 80 second signal cycle length.

Unfortunately, it was discovered that this approach would not be possible with SOAP. The
initial plan was to use SOAP to model the “before” South Lyon intersections based on the
RCOC fixed traffic signal timings and the traffic volumes recorded in the SCATS output
files. The “after” conditions would be modeled on a cycle-by-cycle basis using the same
traffic volumes and modifying the signal timings after each signal cycle. The minimum
allowable SOAP analysis time interval is five minutes. SCATS control makes modifications
to the traffic signal phasing plan every signal cycle. As a result, some of the cycle length

settings for intersections in South Lyon were as short as 50 seconds.

It was also recognized that the SOAP approach delay calculation procedure uses a
deterministic modeling approach that would be very useful in this study. The deterministic
approach lends itself to a situation like a fixed time/SCATS comparison because it has the
ability to make a delay calculations for any length analysis period. The SOAP computational
strategy calculates the critical movement delay for a specified time period of signal operation.
The amount of delay is a function of the conditions which exist during the analysis interval,
independent of the time period length. For example, SOAP calculates identical average
delays for a volume of 50 vehicles arriving during a five minute interval and 600 vehicles
arriving during an hour long interval. The SOAP procedure does not differentiate between
the two as long as the traffic proportions, Signal timing, saturation flow, vehicle headway,
and lost time were the same. The total intersection delay is calculated by using the sum of
the average incremental delays throughout the full analysis period. While this delay

calculation methodology is quite useful, it is not as comprehensive as some of the more

54

sophisticated traffic simulation systems. Some of these shortcomings are addressed later in

this chapter.

The delay calculation procedure that is used by SOAP employs a modified version of the
Webster and TRANSYT-7F delay calculation algorithms. The Webster algorithm was
developed during the 1950's and 1960's for calculating delay at signalized intersections.“

The Webster equation, shown below, contains three separate delay terms.

(I, = [C(1-A)2/2(l-Ax)] + [x2/2q(l-x)] - [0.65(C/q2)"3 * x261]

where,

d, = average delay per vehicle on the particular approach of the intersection (sec/veh)

C = cycle time (seconds)

A. = proportion of the cycle that is effectively green for the phase under consideration (i.e.,
g/C)

q = flow (vehicles per hour)

8 = saturation flow (vehicles per second of green)

x = degree of saturation (i.e., x = q/As)

The equation calculates the average delay per vehicle on an intersection approach, heretofore
Classified as “approach delay.” It includes both stopped delay as well as lost time due to
accelerating from a stop and decelerating from the free flow travel speed. Each of the three
terms in the Webster equation represent the contribution of a different delay component
apparent at signalized intersections. The first term represents the average delay for an
approach assuming uniform arrivals at a fixed-time signal controlled intersection. It was
derived from deterministic queuing theory equations. The second delay term is added to the
first to account for random arrivals. The third term is subtracted from the first two and can
vary from zero (in the case of purely random arrivals) to the value of the second term (in the
case of purely uniform arrivals). The third term was empirically derived from simulation and

field measurement.“

55

To calculate approach delay, SOAP has made a modification to the original Webster
equation. The change involves the elimination and replacement of the second and third terms
of the algorithm. The replacement term comes from an algorithm developed in England in
the late 1970's by Robertson for use in the TRANSYT signal design and analysis program!“
Robertson retained Webster’s first term for uniform vehicle arrivals but replaced the last two
terms with a single term to give more accurate predictions of approach delay during higher

degrees of saturation. The Robertson equation is in the form:

D2 + D3 = [(Bnle)2 + rvrledP’2 - BH/B,l

where,
B”: 2(1- x) + xz
Ba: 42 ‘ 22

z = (2x/v) * (60/T)
v = approach volume (veh/hr)

T = period length (min)

The Robertson modification to the Webster equation was incorporated to more accurately
assess delay for traffic volumes which approach or exceed the design capacity of the
intersection. The Webster equation, by itself, is useful when analyzing approaches with a
degree of saturation between approximately 15 and 97 percent. When the degree of
saturation exceeds 97 percent, the Webster equation produces delay values which are
negative and do not become positive until a degree of saturation in excess of 130 percent is
attained. To illustrate this concept the following example is presented. When the volume
of a particular approach with a saturation flow of 1,600 vehicles per hour is increased from
1,300 to 1,400 vehicles per hour, the average delay predicted by the TRANSYT model
increases form 20.6 to 27.3 seconds per vehicle. The same average delay value calculated
with the unmodified Webster equation shows an increase from 23.9 to 74.5 seconds per

vehicle.“

56

The SOAP delay calculation procedure for left turn traffic involved a more complicated
version of the through movement calculation methodology. The left turn delay calculation
strategy expands the basic through movement calculation procedure by employing a delay
prediction based on a calculated left turn capacity. The SOAP methodology assumes the left
turn capacity for a permissive/protected left turn phase during the signal cycle is the result
of three separate capacity and delay calculations. The first regime determines the delay
resulting from the protected phase of the signal plan. The capacity and delay for this interval
are relatively simple to calculate. The delay can be calculated using the basic
Webster/T RAN SYT equation where the input parameters are those assumed or recorded for
the left turn movement. The second delay regime comes from the number of left turns
which can be completed during the permissive phase. This delay is calculated by first
determining the portion of permissive left turn green time in which there is no through traffic
opposing the left turn movement. This is found by subtracting the actual opposing through
volume from its theoretical saturation flow volume. The “unsaturated” or remaining green
time can be used to accommodate left turns unopposed by through traffic. The third regime
of left turn clearance capacity is determined from the number of turns that can complete the
left turn maneuver during the "all red" clearance interval. In the SOAP documentation these
clearance interval left turners are referred to as "sneakers." In South Lyon, it was assumed

that one sneaker vehicle would be able to complete a left turn during the clearance interval.

The calculation of left turn delay in South Lyon was completed only for permissive/protected
left turn conditions. No "permissive only" left turn delay could be analyzed since the SCATS
data collection and analysis architecture collects left turn volume and phasing statistics only

for protected left turn movements.

After understanding the internal procedure of the SOAP program’s delay calculation
methodology, it became possible to duplicate the procedure and incorporate additional
features for use in this study. The new procedure would also incorporate a mechanism to

shorten the allowable analysis window from the SOAP imposed minimum of five minutes,

57

to a time interval of 50 seconds. Thus, it would be possible to calculate average approach
vehicle delays and total intersection delays for the intersections in South Lyon directly from

the SCATS output file data.

4 u 99
Recognizing the merits of this type of deterministic delay calculation procedure, a computer
program incorporating a similar approach was developed to calculate the delay of critical
movements within the South Lyon under SCATS control. The program used the SOAP
calculation methodology as well as a data extraction system to read traffic and signal timing
information from a SCATS output data file. To analyze the “after” traffic conditions, the
program used the Autoscope recorded volumes and SCATS signal phase timings to calculate
average approach delay for each signal cycle. Total delay was calculated using the product
of hourly vehicular volume and average approach delay. The same program configuration
was used to complete the “before” analysis. Instead of using the SCATS signal timings, an
optimized fixed time signal plan, in which the signal timings did not vary from cycle to cycle

wmumd

The analysis models were divided into two separate groups. The first set of models were
developed to analyze the “after” conditions Of South Lyon. The “after” analysis was
completed prior to the “before” analysis because of the manner in which the traffic volume
and traffic signal data were collected. The “before” models also required the completion of
a SOAP signal timing optimization analysis to determine the most effective signal phase
split arrangement to accommodate the hourly traffic volumes. The same analysis technique
and cyclical traffic volume data were used for both the “before” and “after” periods. The

sole variant between the two groups was the traffic signal timing.

Four separate computer programs, one for each of the four intersections analyzed in South
Lyon, were coded using the Fortran programming language. Each intersection required a

separate program to accommodate various SCATS output data file formats. Each

58

intersection had a different number and order of the critical approaches within each data set.
The intersection of Ten Mile Road and Pontiac Trail incorporated eight critical approaches.
Each of the four intersection approach legs included a single shared through/right turn lane
and a single exclusive left turn lane. The intersections of Nine and Eleven Mile Roads with
Pontiac Trail included through lane information only for two of the four approach legs.
Separate left turn data were recorded for the north and southbound approaches at both
intersections. The Ten Mile Road/Reynold Sweet Parkway intersection data set included
through and left turn data for the westbound approach, through data for the eastbound
approach, and a single unopposed left turn approach volume and signal data on the
northbound approach. The data collection configuration at the Ten Mile Road/Reynold
Sweet Parkway location was different from the other three because it has only three approach

directions.

An output listing of the program source code for the program written to analyze the
intersection of Pontiac Trail and Ten Mile Road is included in Appendix A of this report.
The program is composed of three separate components of data retrieval, delay calculation,
and delay output information. The program was designed to read the SCATS output data file
and extract the traffic volume detected and the green time dedicated to each of the eight
critical intersection approach movements. This information was used to “feed” the modified
Webster/TRANSYT delay equation. The average approach delay was calculated separately
for each movement during each cycle. The hourly average of this delay as well as the total
of all delay for the intersection experienced during the full hour were calculated and written
to an output file. A sample segment of this output file has been included as Appendix B of
this report. As shown in the sample output, the program structure allowed the data to be
extracted, delay to be calculated, and statistics to be written continuously until the end of the
SCATS file was encountered. While the length of the output file was the same for each
intersection, the number of SCATS data records that were extracted and used in the
calculation process to determine the delay times under SCATS controlled conditions varied

from approximately 30,000 to over 45,000 records.

59

Three input parameters used in the “after” models were not taken directly from the SCATS
output data files. These were the vehicle start-up lost time, headway, and critical movement
saturation flows. Each of these parameters is critical to determining the delay at signalized
intersections. The headway and saturation flow rates determine the capacity that a particular
lane group has to service the traffic demand. A traffic demand in excess of the capacity of
the approach will experience greater delay. The lost time value reflects the amount of time
that a green signal phase can not service the approach traffic demand. The lost time
parameter is also critical because its effects are felt during each signal phase. A few seconds
of lost time for a four phase signal with a one minute cycle can result in over nine hours of
wasted green time every day. The values for lost time, headway, and saturation flow for each
of the intersection approach locations were collected from Road Commission for Oakland

County Traffic Operations Center records.

SCATS data files were recorded for the South Lyon intersections during a one week time
period starting at midnight Monday May 5, 1996 and ending at midnight Saturday May 10,
1996. From the recorded data it was possible to analyze twenty separate periods for each
critical intersection approach movement as part of this study. As a result the total size of the
comparison sample data set included 480 records for the “before” condition and 480 records

for the “after” condition.

The analysis program calculated the average approach delay and total intersection delay for
each hour of the day. To analyze a varying sample of traffic conditions, four separate daily
analysis periods were selected. The first was the hour between 12:00am and 1:00am. This
hour represents one of the low traffic demand periods of the day. Often, several traffic signal
cycles would occur before a single vehicle was detected on a particular approach. Under
these conditions it was thought that SCATS would function at its best. Only the minimum
green time would be allocated to the low volume movements. Correspondingly, it was

expected that delays for the remaining approaches would also be minimized.

60

Two peak traffic demand periods, 7:00am to 8:00am and 5:00pm to 6:00pm, were also
analyzed. These intervals were selected for the opposite reason. The comparison of SCATS
to fixed time signal control would help to illustrate the differences in operation during high
demand periods in which SCATS could implement a longer cycle length and better allocate
green time to the high demand movements in an effort to minimize delay. The final analysis
period was the hour from 12:00pm to 1:00pm. During this hour the demand was between
the high demand volume of the morning and evening peak hours, and the midnight low level

volume.

4 “ ” e

The second phase of the South Lyon traffic analysis involved the completion of the “before”
portion of the “before/after” study. The “before”delay analysis incorporated a two step
approach. Rather than using the actual RCOC signal timings for South Lyon, a set of
“optimized” fixed time settings were used to determine the “before” delay conditions. The
decision to use optimized fixed time was made to assess the benefit of real time adaptive
signal control against a well timed fixed system. A comparison against the original timing
plans may have showed a benefit based on a poorly timed fixed system. Therefore, the
benefit would be from improving a bad condition instead of comparing two different, though

equally good, systems.

The SOAP program was also used to determine an “optimized” fixed timing plan to compare
the adaptive and fixed control strategies. The explicit SOAP optimization strategy attempts
to minimize delay by using a cycle length, green split, and dial assignment optimization
process. All of these parameters can be calculated automatically by the program. SOAP
also allows an implicit optimization method. The implicit method requires some judgement
on the part of the user with regard to the use of permissive versus protected left turn control,
actuated control, and phase sequence selection. All of the implicit optimization methods and
some of the explicit ones were based on some of the SCATS minimum parameters to assure

a fair comparison.

6]

The SOAP optimization process works by attempting to minimize the total intersection delay
for the time period under analysis. Earlier it was shown that SCATS minimizes the delay by
equalizing the degree of saturation for each of the constituent intersection approaches. By
contrast, SOAP attempts to reduce stops and delay on a “global” scale. This philosophy is
best illustrated by a major street/minor cross street intersection situation. In this situation
the major street volumes are significantly higher than the minor street traffic. SOAP would
minimize the intersection delay by allocating a proportionately higher percentage of green
time to the major street approaches. However, the benefit realized by the heavier movement
delay reductions would be gained at the expense of the minor street traffic. In extreme cases
the minor street approach traffic could experience average delay greatly in excess of those
experienced by the major street approaches. The total delay is minimized because the

inordinately average high delays are experienced by a relatively low number of vehicles.

As noted in the SOAP manualm’ this method results in an unfair allocation of green time,
especially in situations where one of the minor street approach volumes is relatively high.
However, this method can result in substantial savings in vehicle emissions and operating

costs because the total amount of stops and delay at the intersection in minimized.

The optimization procedure used to determine the “before” fixed time signal plan did not
take full advantage of the SOAP optimization process. The signal cycle lengths used for the
fixed plans were the same as the SCATS average in each of the four analysis periods. The
similar cycle lengths were used to account for the need for coordinated progression between
the signals. SOAP analyzes each intersection in isolation. In reality SCATS could have been
limited in its selection of a more suitable cycle length by the need to maintain coordination
with adjacent signals. Minimum green time restrictions were also placed on the SOAP
system. In several instances SOAP attempted to allocate very low green times to minor street
movements. However, these green times would not allow adequate time for pedestrian
clearance and could result in safety problems with drivers attempting turning movements into

a traffic stream with inadequate gaps.

62

Two other restrictions were placed on the SOAP selection process. The same
permissive/protected left turn phasing was maintained between both of the test groups. Once
again, this decision was based on real-world safety considerations in which protected left turn
phasing would be warranted by the lack of acceptable left turn gaps. The final restriction
was the forced use of the same phase sequence pattern. It was felt that this would result in
a more balanced comparison without significantly effecting the signal performance

parameters.

Separate optimization trials were completed for each time period at each of the four
intersections. These signal timings were based on the average traffic volume recorded during
the week of data collection. The fixed time plan did not vary from day to day. The traffic
volumes and delay calculation procedure used during the “before” analysis periods did not
change. The program was modified so that the timing plan was fixed throughout the analysis
while the same SCATS traffic volumes were read for each cycle. It is recognized that a
direct one to one correspondence may not have existed during each fixed signal phase and
the volumes from each SCATS phase period. However, it was felt that the total number of
SCATS signal cycles would be approximately equal to the number of fixed time signal
phases for each separate analysis hour. The signal timing plans used to analyze each of the

various analysis periods are shown in Tables 4.2a through 4.2d.

4.3 Model Limitations
Like all traffic models, the modeling technique used to calculate delay conditions in South
Lyon had some inherent weaknesses. These weakness were the result of two primary

sources. The first came from the SOAP program and delay calculation procedures. The

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Am<OmV umS—uan— £922.44 was 8.23.530 mama 05 win: cuLuEEoo was couaflEuno in: 8.25 ”So:
om mm mm om €03 5&5... 296 .52.
683 N 4.3 m Nam m 6N. .6 m mates aaau_o..8m._2..
Scam mm $0.3 _N LSN-am ON 68:: a ,3:th ncaoemam
68va mm 48m»: 5 SEN 0N $2.: o swag; 3888?.
689m . m. chow m 650w m 68:: w. ES. :3 ucscnfizom
N55 E 6605 N.m. 683$ 7 wm “SN-we 1, 3. @225. uzsonfisom
650m. m Show m Now. m 68:: m Es... :3 unacnnzoz
NEWS 3 l SmN-o NV.- 1 680.3 _ wm $53 3. 595::- wasoncglzr
226 2.. No one 1126 o 65 No 82 -l 263 .35 No I. 68 265 2: .6 485 .l . .
uwaEuBum GE; uwficoueum DE:- umanuBm DEE. uwgcoubm GEE.
8:822 SEE 580 confine—Z emana— cun-zo .5282? 83m 520 5:82? “WEE :35 NEEDS—2 5353‘
En oono . E 8% EWoo; . Em ooumN En ooux - Ea 00$ Ea co; . Ea comm.

 

 

 

 

 

 

mMCEF Gama BEEENO - 23% 0:2 052 EB 22% cannon— No 5:03.25
3.5 5:335 55 2.5% .63 5:8

 

 

63

5.....30 can. use... visage 5 56m 8:: 526 Au... 63:.

 

 

 

 

 

 

 

 

 

 

 

 

$2va owmxoam 295151 Ea coufiEuqO 5“me 2: was: 8.03960 23 conga—sac Em... own—E ”20:
8 mm mm mm 65 53.3 265 Eek
SE... m 65: N Sod m @405 N 4.4.20.5 quEDNUEuméfl.
68M:- > SSW N- .NNCSW h 68mg v 5: c3 ccaoemam
6&6». Nm Sada 3 68 N .3 N._ 6.80.3 I cwzocfi. 2:588”.—
CNSwN- N. Emma h 680w N. 680N- 4. 5&- anfi 320833
Swhm mm 62.3 3 S» N .8 E 680.3 2 ”NM—BE... 35088?
News > Saw N. 686w h 6.3.5 v EB. c3 usaonfisom
689.84 9. 80.9. D. Rm. 5 mm $0.3 mm @385. ucsonfinom
6N1:- N. SEW N. 6.3.x 5 ~35 v E: c3 3:82:02
“$va 9. 68on 3 68m. 5 mm 650$ mm smacks-N. ucsonnflloR—
20.3 05 .8 863 225 65 No N83 2&0 2: .8 N83 2%0165 No 803
”92:83; DEC. ownEuEuL DEC. owficouuon— "DEC. omEcoEum DEC.
5532? 35.5 520 5:82;... “mann— 520 coua8=< 3.3m 520 5:32? SEE 520 £25.62 88:51.4
Em cone . Emoonm En co; . Em ooHS Ea comm - Ea oouN- Ea co; - Ea SHE

 

 

 

 

 

 

SEEP Ema 6886.5 - 23. oz: 59 65 =5 :65; L6 888.25
6.5m 523...:— E»: 8.5m :63 5:8

 

 

5:225 GEE 13:. .6955 5 £5 as: :35 a... “any

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E<Om. 093.8,. 29:22 05 803.250 acwfi u... was: 020388 33 gouging 5% 82.». “0.0:-
fl 2 fl 2 .83 593 206 as
$5N N $o.N N $N..N N $0.m N 39.25 oo§80.0..0u~.-..<=
$o.oN 0. $50 . .. $50. 0. $o.o. 0 53.001. 0:802:02
$52. mm $m.§ N0 $0.2. vm $N.m> .3 095...... 0:385
$50 0 $ . N 0 $50 0 $2. v Es... c9. 0:80.83
1 $52 mm $53. N0 $0.2. am $52. 3. smacks... 0:28.33
2qu 2.. .o 003‘ 0.92.0 2.. .0 303 0.50 2.. .o 38. uflfiu on. .0 Com.
omega». 08.... 090.529. 08.... $5580.. 2...... 092580.". 05....
.5032? 82E 520 5030.3. 342.. 520 5.88.2 0.85. 520 8:80.}. 3.57. 520 Equ>22 58...“?
55.8.0-2...on Em8..-EmoouN. Swoonw-8moouN. Eng“.-EmoouN.
$25.... Rama 03:50.5 - Box 2.3. no... 05 .35.»... Beam 29:3”. .0 85038:.
Sam £52.33. 35 £28 .33 5:8
8:825 2...... Ba...— 35500 a saw 08.: :85 3.. 93.;
3.109 owaxumm marge 05 co..a~.EunO «:90 2.. mafia 020.988 33 803.80% 5% 08.... “So:
mm 0.. mm mm
§N N «film N evN N .Mmm m
$3.. N. $0.0. N. $v.NN a. $0.0. a
$3.. N. $90. N. $v.NN a. $0.0. m
$N.w N. $.50 m $o.m 0 $0... 0
$0.2. 00 $0.2. N0 $0.2. o0 $5¢0 mm
$N.m N. $m.0 n $o0 0 $0... 0
$0.2. m0 $0.2. N0 $0.2. 00 $500 an
2&0 2.. .0 38. 0.0.6 o... .8 .08. Buxu o... .o 38. 3%.... on. .o .08.
093583. 05.... 03538.. 08.... omficoobm 0E... 03.529. 08....
8.38.? 32.... 520 .5082? 03...”. 520 8.38.2 82E 520 8.38.? can... 580 608263. 008...“?
Emoo”0-Emoouw Bacon. -quouN. $80-5.“ oouN. Bacon. .80 SUN.
£55.... .330 03.8.30 - 03¢ 2.2 0“.05.“... 05 :8... 80:0... .0 50838:.
:56 5:333. 35 8.5m 83 5:8

 

 

 

 

 

25a. 605,.— 60533 5 =— m 3.5: 909.0 and. 03:8

65

second main area of limitation came from the architecture of the SCATS system. The
SCATS/Autoscope architecture is designed to collect and process the data required to most
efficiently implement signal timings. Unfortunately, the data necessary for the operation of
the system is not perfectly designed for the completion of comprehensive and detailed delay

studies.

The SOAP analysis package was developed specifically for the analysis of isolated
intersections. It also allows the calculation of a number of signalized intersection measures
of performance. The program is limited, however, in some aspects which were important to
this study. The most important limitation of the SOAP environment is that it lacks a
mechanism to adjust the approach traffic arrival distribution. SOAP considers distribution
of vehicle arrivals to be uniform with respect to any cycle. No attempt is made to model
coordinated progression and the arrival of vehicles is assumed not to be influenced by
upstream signals. As a result a direct assessment of the benefits of coordinated progression

that SCATS can demonstrate was not possible in this study.

Another limitation of the SOAP program was the lack of ability to assess the conditions on
a stochastic basis. Unlike analysis packages like NETSIM, SOAP does not differentiate
between drivers and their various driving habits. Differences in these parameters would
primarily impact the start—up lost time and vehicle headway. While the impact of variability
of these parameters was not accounted for in this study, the large sample size helped to

minimize the effect of variation on the average values.

Another feature that is available in systems like NETSIM are detailed descriptions of
geometric features and their relationship to traffic operations. For example, the methodology
used for this analysis assumed that there were no "over saturated" left turn or right turn lanes.
This condition would result in through lane blockages and increased delays due to traffic
unable to move. The analysis methodology employed in this study assumed an infinite

length of turn lanes and no potential for "spill-back" queues from downstream intersections

66

or through lane blockages. In reality, there were a limited number of occurrences of through
lane blockages on the east and westbound approaches to the intersection of Pontiac Trail and

Ten Mile Road.

The operational configuration of the Autoscope/SCATS system in South Lyon also presented
some limitations in this study. Most notable was the lack of the ability to calculate of delay
for all movements at the intersections. The system was configured to detect vehicles and
modify signal timings for only the critical approach movements. Left turning traffic was
detected only in locations where an exclusive left turn phase was present. In all cases the
right turning was combined with the through traffic movement. Therefore, no delay
parameters were calculated exclusively for right turning traffic at any of the intersections.
However, the impact of the lack of this data was not critical since the right turn traffic shared
the through traffic lane with the through traffic. Therefore, the SOAP delay calculation
procedure remained useful for all of these locations. The only exclusive right turn lane in
the South Lyon system was located on the northbound approach of Reynold Sweet Parkway

to Ten Mile Road. No delay data was calculated for this movement.

While the utility of using the SCATS output data files was obvious, the use of these files also
presented other limitations in the study. The original plan was to assess the delay conditions
at all six of the South Lyon SCATS controlled intersections. Due to the data collection
configuration of the system the intersections of Pontiac Trail at Reynold Sweet Parkway and
McHattie Street could not be analyzed. The SCATS data input requirements at these two
intersection was such that the approach through volumes were all that were necessary to
control the timing of the signals. The cycle lengths at these two locations is coordinated at
the RCOC Traffic Operations Management Center. The phase splits, however, are controlled
locally at the intersection signal controller. No volume or timing data is transmitted or stored

at the control center. As a result, there were no data files to analyze in this study.

67

An additional factor which limited the model somewhat did not become apparent until after
the first attempts at delay modeling were completed. It was found that the
Webster/TRANSYT delay equation within the model predicted inordinately high delay
values for approach traffic levels in excess of the saturation flow. This phenomenon
occurred despite the SOAP correction of adding the Robertson model terms to the Webster
formula to limit such conditions. While over saturated conditions occurred infrequently in
South Lyon, this condition did present aproblem at several locations. The most noticeable
events were observed during the low volume conditions when vehicles on the minor street
approaches were delayed in excess of one signal cycle although the approach volumes could
not have resulted in a loaded cycle. To correct this problem the maximum value for the
degree of saturation in the delay calculation equation was limited to 100%. This limitation
resulted in predictions of delay which were more realistic compared to the field observed

values.

4.4 Delay Model Output Data

After all of the programs were debugged and fully operational they were executed to produce
the output statistics required to compare the delay conditions in South Lyon with and without
SCATS signal control. A sample segment of data output produced by the model is illustrated
in Figure 4.3. The comparative MOE for average approach delay and total intersection delay
are presented in Tables 4.3a through 4.6f. The first set of tables, 4.3a through 4.3f, are for
the intersection of Pontiac Trail and Ten Mile Road. Tables 4.4a through 4.4f, 4.5a through
4.5f, 4.6a through 4.6f, illustrate the same statistics for the intersection Pontiac Trail and
Nine Mile Road, Pontiac Trail and Eleven Mile Road, and Ten Mile Road and Reynold

Sweet Parkway, respectively.

The tables are structured to illustrate the hourly volume and delay statistics for each of the
critical approach movements. The tables are further divided into four sets of columns, one

for each of the analysis intervals. Below each of the hourly approach delay values is the total

68

Figure 4.3 Sample Segment of Delay Model Output

Pontiac Trail/T en Mile Road Intersection

Period begins at 7:00

Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.

Movement #7 hrly vol
Movement #8 hrly vol

2447.000 ave. delay = 43.215
=401.000 ave. delay = 38.613
= 17.000 ave. delay = 24.958
= 16.000 ave. delay = 37.307
=194.000 ave. delay = 31.360
=336.000 ave. delay = 52.589

. = 50.000 ave. delay = 34.943
. = 67.000 ave. delay = 30.712

Intersection total delay: 17.606 veh-hrs
ave. cycle length = 85.8105ec

Period begins at 8:00

Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.

Movement #7 hrly vol
Movement #8 hrly vol

=353.000 ave. delay = 29.904
=324.000 ave. delay = 32.991
= 16.000 ave. delay = 30.338
= 42.000 ave. delay = 23.932
=l93.000 ave. delay = 37.659
=265.000 ave. delay = 34.587

. = 22.000 ave. delay 2 25.188
. = 22.000 ave. delay = 26.563

Intersection total delay: 11.197 veh-hrs
ave. cycle length = 80.9558ec

Period begins at 9:00

Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.

Movement #7 hrly vol
Movement #8 hrly vol

2369.000 ave. delay = 29.272
=389.000 ave. delay = 32.122

= 17.000 ave. delay = 65.652
= 14.000 ave. delay = 23.884
=186.000 ave. delay 2 29.813
=228.000 ave. delay = 31.621
. = 23.000 ave. delay = 24.237
. = 27.000 ave. delay = 25.658

Intersection total delay: 10.765 veh-hrs
ave. cycle length = 80.733sec

69
intersection delay statistics for both the fixed time and SCATS adjusted signal control delay
groups. Under the total intersection delay, the average cycle length for the analysis interval
is also shown.

,9

The first five tables for each location, designated “a” through “e, illustrate the delay
statistics for the days of Monday through Friday. The final table for each location, labeled
“f,” shows the weekly average for all of the volume, delay, and cycle length statistics for the
week. In the following chapter these statistics will be more closely analyzed and compared

in an effort to assess their significance and relevance to one another.

70

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

UEE. E: 59.. 0030: 2:25.80. 00%: smsocfi “20:
Wow 0.00 00» 9mm n.0m 00w 0mm. me .l 303 50:3 3&0 umfiufiqw
00% 5.8 8.2 ova NE: on: 02 ME :53 8835:;
NS. 0.00 0N. N.w~ 0.3. 3. Nfim <30 mm v.2 W: 0_ ES. :3 05.00.me
wdm Ohm MON msm 0.0m mom man mag wmm w.m~ :H mm cwsoih uczoemam
«:3 0.0m mv 0.m~ o._ m 0m 06m 0.2. VM NM. :2 v E3. :3 00:00.83.
0%.» n.0m :mv 0.0m v.3 :H mamm 03 00— w. _ N 0NM 00 5?th 050983.
NS NS an :N 0.3 0N 0.; n.0m N 0.: Nm v E=._. 03 2:505:00
_.m0 meow 00m Naum 0.: 03 flow 0.0N 50¢ 0.0. _ .0 «.2 :msohf. 9:605:00
mdm N00 3“ v.0m 0.m~ 0m mdN w.mm 0N M: v.m v Esh 03 0:302:02
w.wm 0.3. am 5.0m NE 0pm 0.3 ”mm 03. 0.2 Em mm. imwzoih Emma—E02
.000—EU :05 203 65:00 055 203 3:50 :3.“— 205 6.2000 ENE 303 I ll l.‘ [ ‘
mh<Um 03:530.. UE30> mH<Um .boEEaaO: uE=_o> wH<Um 03.5000: uE=_0> mH<Uw 035500.. 0620>
A003 Nina zuaohmm<1l 052... 303 wagon zunoxazx 055% y 303 0:00 nuaoemm’x 0580. 133% 058.”. 32:30.2 concanz
E 000.2520 E 00; .Emoofi. anoquENOAE Ea00; .Ea00“m_
000— w an: ”ammo—E
031 0:2 cup. 05 29;. 80:0“— .«0 3000085
330 30.3.20 EX. £28 53 5:8
35:0 .0102 wagon and 03:.
0.0m 0.00 0.; 0.2 w.mm 9mm 0.? 0.mm A003 50:3 2&0 uwfiu><
wmdm 0n.0~ m0.~_ 3.0 _0.: 06: 3.0 3.0 05.003 .339 0000032:—
Ndv van m0 0.3” 0.? 3 5.0m adv 50 0.0 0.0 0 E5... 03 0030003
«.mm Yum m: m.wN v.3 D: 0.~m 0.0m 0mm 0.0m w.0~ 0_ 00:05... 00009210
_.mv 93‘ Em m.m~ n.0m m: 0.3” 0.50 0m 0.: _.w_ N E3. cut. 00:00.33
0: m.~0 wvv m.mm 0.0m 0mm v. 0 m mSN v2 b.0m m.0m 2 case—E. 00:00:03.
m0.» 0.; 00 m.0m TVN vM new 000 0~ hm 06 0 EFF 00‘— 0:305:00,
0.05 fiwm own msm 0.2 3v 0.mm 5.2 SV n6 0.m >0 :9..th 00:85:09
0.5m 0.0m 3.. Nam w.0v _N 0.3 fig 2 ms mé 0 ES. cow— 0.500552
m.mv 5mm m0v 0.0a 0.2 00v NS. m.0m 54v #0 Nam N0 .3320. EsonstomL
3250 SE A :13 3250 5: no: 3580 SE 2100 328 as 205
mF<Um .boNE—E‘O: 0830> mH<Um Eng—.000: 0E20> mh<om .bodEUQO: oEio> wk<0m 03:000.. uEgo>
A003 hing £89340. 058% N83 hing acacia; 0000;. 303 .315 20005“? 0555. A003 3.00 0000591 0595. E20032 580%?
a 80.53% E SUTEQSHE .580.§9§ S85E82

 

 

08m 0:2 cub 05 :20. 80:00 00 0000302:—
h.56 532:0 .53 35m .83 saw

002 0”“: ”20:02

 

 

33:0 .0002 .2309 .06 03:.

71

 

 

 

 

 

 

 

 

 

 

 
  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

052. E: .00.: 00.00: 3:0E0>0E 2.00.: 0w=0..0. 0.0:
3... 98 m. .w 3... 3.. ohm 0mm 30 Ga. 5.33 2&0 .052
03% 00.0N No.2 «0.0. 00.... SN. EN 0w.N 30.00.». 0000 :ogufiec.
«.3 5.0 ms OWN m..m Nm 0.mN ..mN 3. v... 0.0. m E30. 000 0:30.80
N. . m ..0m 08 0.oN 0.. m 00. v.3 m.wm mvm N.0m 0.: 3 09.90% 050008...
.0N 0.0N 0m 0.mN W? on ON». 0.Nm .m w... .0. 0 E: 03 05.00.33
0.: 0.00 >0... 0.00 53 omN 0.? OR 00. 0.0m 600. 2. 00.00.... 05.00.33
04mm 0.3 0.. 0.: 0.mN om 0.mm 000 2 0.0. .0 ON E:... 000 05.000300
0. . 0 00m 000 .Nm 2: 00v 0.0m .NN E... 0,0. mm m0. 0w=0..0... 0::000.=ow
0.3. v. .e. um 00». m.mN 0N NSN m...N NN 0.2 0.» N E: 03 0:300:02
mew 0.0N 0:. 00m 50. m0... v.3 ...Nm 03. 00 N0 oMN .3320... 0:800:07.
.0::00 :20 ~00; 30:00 :30 00>. .0...:00 :30 0.3 60:00 :400 2.0[5 1‘:
mk<0m 000.5500: 0E=.0> m...<0m ._00§E.:0O.. 0E=.0> m...<0m ..00~.E.:0O.. 0E:.0> w...<0m ..00N.E:0O: 0E=.0>
303.3...5 0000. < 0.02... .003 Eonflommm... . 0.03... .003 >000 0000.00... 2.08% Sum. 0000 08200.0 2.02% E0E0>02 008.00....
E0000.::10..n E 00“. .E000HN. 050000.E.w 00;. E0 00.. .0200 N.
000. o 02 «00::
003. 002 :0... 05 :2... 30:00 00 000009.25
05% 80.5.0... 33 0.28 53 5.3
2.3.5 .003). >509 .00... 0.00...
000...: EP Ewt 005.0: 3:0Eu>oE 0000.: 00:20. .20:
3o 98 w; 3... 00.. on. >3 9?. .80 50:3 2&0 omega
.0.0m wmwN 0N.N. v0.0 v0.2 0. .m. MNN 0N.N C0003 Fin. 0000030....
0.00 m.mm 00. MEN 0.00 mm 0.0m 0.3 F 0.0m ...wN N E0... 00.. 05.00.30
hoN m.0N .0. 0.mN Eom 0... N00 mém 0.0m m.MN v.mN 0N 03.20... 05.00.3m
N.mN m.mN 3 ON 0.; .N N.0m VS» 3 .0. .0. 0 Es... 000 05.00.33
.00. v.0v Em 0...... EN.“ mNN N..m m.mN 00. v...” 0.3 m0. €580... 05.00.33
m.0v 3% <0 MEN m.mN mN men 000 .N 0.0 Na. 0 0.50. 000 0:305:09
0.00 5.0m 30 N0m E: .9. as.” m.mN 00v M0. 0.0 00. 09.20% 0:80:50?
va v.00 5m 0.0m w.wN v. 0.3 .00 w. 0.0 0.0 0 Es... cox. 05.000002.
0.: mNV :0 m.0m 0.... Ev 0.? 6mm 03. 0.0 5.0 mm. 09.80... 0::00000E
.228 S... 22. .380 5: 32,. .858 a... 3010 .858 a... Eula
m...<0w ..00~.E00O.. 0E=.o> m...<0m 03:000.. 0:5.o> m...<0m ..0o~.E00O.. 0E...o> m...<0m ..0u~.E00O._ 0E...0>
Com >200 00000004. 0002... an. Exam 00.8.00... 0002... .003 main 0000:00< 000:... 30$ 00.00 00020.3. 0.00.; E08052 00080040
E 0000:0000 E 00..-E000.N. E00007E00.H E000...E000N.

 

 

 

0000 0:3. :0... 05 :2... 00.0000 00 500022....
3.5m 5.0.330 .55 308m .33 520

 

 

3&5 0.32 590 on... as:

72

QEE. EB Ewt 822...

A23539: Sta: swash: “Bo:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

9% o? .3 OS 03 ca 0.? SH to: 3:3 226 oweué
wvam mmem 5.: v0.0 En: ONE SN 3: 22-593 QED c2335.:—
odv Woo ow mam odw Om mgm Ewm o0 v.2 0.: o 5.5 c3 2:35am
Wm». 00.... o_ m m.w~ o. _ m .2 9mm cam .3. New v.2 Tm FREE... vcaoemam
Nam odm 3 mew 5.; am ~.mm on hm m.m_ _.w_ v EFF :3 9.8933
5.05 we? o? Vow Emv ova 1mm 0.0m E.— o.o~ bow M2. cmzohc. 9.8283
0.; mam 3 Nam new 3 n.3, 0.3 c. Q? on 2 EB. a3 9.255:an
NZ. New. mum. ms». 2: mmv 0.xm mam 3m 0.2 on mm: @3th 322.52%
m6». :7. m». _.cm We». vN bum NNM Om 5w vé m EFF c3 9:52:02
Yam Wm». www o. _ m 5.2 1! How ‘39? ~.m.m m3. ”2 m.m o: cm3_:+wmmwnctoz
33:00 5E 293 33:00 55'] EB; 35:00 55 Sub 6:50 .53 296
w._.<Um .boNEEQO: oE:_o> mH<Um .budEsaO: uE=_o> mH<Um 53:530.. uE30> mP<Um .bufiEiO: 0:515
303 mafia pagan/x 058p. A83 mica :uaoamrwzl 055C. r 33; Nfloolzmaaammxil 058% 303 $30 2899; 058,—. 52:30.2 £399?
Em 86-:5 comm Emoo; -Emooufl Ea oouw-Em comb Eu 8; -Ea CONE
omEu>< ESQ - cozy—E ; :95 33% .8 x83
30m 2:2 :8. 23 EC. gasp—om .8 8:823:—
%.55 nouégm 55 93% 83 58m
:- 25 3.52 309 an... 033—.
Btu: E2 Ewt ova—u? 85896.: uEab F3385 ”Bo:
:5 98 n. a OS 9% cam on cam 62: £33 226 omen;
3.3 3.: 3.2 3.9 S .w_ 2.2 vvd 5N €5-53 have 5:822:—
w.om QC. m: 05 me mm 5.3 QNM me WE _.w_ E E3. «3 canoemam
9mm. v.3” Omm ofim 53 :2 mac 5mm cmm wfim o. _ m mm cwsoE... 2:833
”mm 0.3 ov _.mm 92 cm wdv firm mm vd 3: N E3. «3 2:09.33
can can mmv adv 9w.“ com msm 9mm 0: Ymm m. _ v 3 awash; 230533
Wm». ode E. v.5 5mm wm ~.mm own mm v.2 vs 2 E3. cox— anacondaom
Numb Eon 2m mév mdm wwv w.wm w. _ N 3m v.0 o6 wE 29.8—E. vcaonfiaom
_.wm v.3 no v.3 oém om oém OSN E Q: o6 N E3. c3 2:22:02
adv msm 5v Omm he. 2? NS fimm Hmv 0:2 o.m hmw :wsohfi. 9:52:52
.obccU 55 23% 75:90 .85 205 _o::oU 52m 393 .0550 55 39$
mh<Um .boEEcaO: uE30> mH<Um .boEEamO: uE2o> mF<Um .bofiEsmO: oE:_o> mF<Um .bufiEcaO: 25:5
A83 >300 suaoamfl..l 053% 393 QED £889: 058... xAuumv QED 58594. 058,—. A83 “Au—om cumoama 952,—. E2532 2829?
En cob - Em comm Em co; - Em 092 Ea comm - Ea cog. Ea co; - Em comm—

 

 

 

 

 

 

 

 

caom 2:2 sub 98 EC. amazon be 8:838:—
35 832.3% 35 $28 83 58m

 

:-

com“ .2 as: Santa

   

«:0 .0632 £09 and 03:.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

73

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3%: E2 Ewt ova—u: scuEu>oE Sta: swag...“ H80:
Woo 35 3: 9% 3m o? n: a? 303 3:3 226 owe”;
00.3. 3:— 050 Sun woe. mm...— omd $8 $5.55 E00 5585::
odm few 02 mdm Now mwfl 9.3 man 3.». WE NS. 2 a :02... 930833
_.om arm or. :h mam 9. 2h mdm hm 9mm «Sm N smart. canoemum
Nov mam Om Wm. ma. 2 0.2 w.m_ 2 0.0 od o E3. 51 ccsonfisom
9mm ca: 3.0 ~12 ~32 9% N; n. _ m 3.“. Wm _.v me. $32.? ocsonfiaow
«awn oém hm mém Wow 2 man mdo o. _.m we m Ezh c3 9.25552
Qwv Sum 05 v.3 wb 5v 5.: as mmm o.m _.v S— zwsofih 9522qu
3:50 55 29$ 35:00 :mE 2?: 63:00 :55 - 3?: SEED 55 293
mH<Um ,._3~E:EO.. u:_=_c> mk<Um .buEEsz: oEBo> mF<Um .boEEcaO: oE=_o> mh<Um ..B~E=EO.. uE:_o>
30$ have zuacEmfll Stat. 33; Mien 280.3%! QESH Auowvfifloa 58:36. 2&8... 33¢ wfloa 58.9?le uEafi. 22:26: £805“?
£386-25 oonm Em ocu— -Emcoum. Ea ooncha 8;. En 0o; -anonfi
03— .N. >22 Nanwosb
naom 2:2 2:2 93 32,—. 85:0; ho 5:832:—
35 85.3.3. 55 mg; .33 5:8
5 .90 ~03: £09 £3. 93.;
059: E2 Ewt 330:. 355265 Stab swaohfi ”Bo:
Ndo Odo Q; odm «.3 ohm 9mm odm 88v Swan: 03an owfio><
2.0m o5: 3.0 mm.m 3N— mw.o_ cod 2 .o €2-53 haw—on 8335::
ode adv mmm odm fiwm N: 0.; v.2. onm CNN v.3 c n :85. 950833
mdm fiwm v0 OR 5.3 we v.w~ 9mm _w od od o .3385. usaooumam
Wmm Nov om ow. wd. w NE fia v o.o od o E3. aux. acaonfisow
own ”fl Ea Yam me: new NS 0.: an 2 ma mm smacks... canoncusom
0.0m 0.3. NV 9M: 92 2 own 0.? mm vé .1.“ m Ea. £3 2.59552
5.9. Tmm 0mm 0.: No mmv mSm as 3m «4 Qm E. .3385. 9509552
65:00 :aE 29$ .2500 55 39$ 35:00 55 29$ 33:00 Sun 205
mF<Um .bofiEch: o:_:_o> mH<Um .bufiEcaO: uE=_o> mk<Um .bufiEaaO; uEEo> mH<Um .buflEcaO: uE=_o>
€03 xfloo cuaoanz uEafi. 38V mafia :uwwmmm< 058,—. 803 $30 282%? UEEF 33; 5200 coaoam< uEEH 52:26: cuaoam<
Emocb-Emoonm Em8u_-Emooum_ Emoouw-ano# SSH—-anoufl

 

 

 

 

 

 

33— 2:2 052 can 28% 898; go :ouuufiBS
3.5m Susan 53 938 83 58m

 

 

.5 .5 two—Z 30n— avé «Ssh

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

058. ES Em: 00300.. $00539: 058. smacks“ “80:

30 0.8 2: 98 new 0.3 a: 30m A30 393 200%“on
3.0— f .2 RN Sam Omar.— m>.__ 36 Inc 05.53 N200 00:03.8:—
o.0v M; OmN 0.9. m6? o: Own 0.2. EN 0mm 1mm 2 55th 0030833
0. _ m _.wN me m.wN 0.0N cm VON 0.0m 3 N.mN 0.2 c .3395. 0030235
m.mm 50: m: 0.2 Nd. N Yew o.wN_ 0N 0.0 0.3 0 F5... to; ccsonfisom
N.NN 2: .00 odN o.N_ Sm 52 _.NN emu NN _.v 0oN cwsohf. 0:509:qu
o.0N th mN Om.— o.w_ w m.__ S: C mN mN. o 53. an; 35005002
Wm? oNN N; 0.2 0.0 000 ._.‘Z ms Rm 0._ 0.0 m: cmsthncsonztoz

30:00 :20 30> 6:000 :20 2.0.3 6:000 SHE 305 32:00 020 205 i

mH<Um .buNESE 0E20> mH<Um 53.8500: 0E20> mhdnuw 53:560.. 0E20> mF<Um .buEESQO: 0E20>

A003 x200 zoaoanz 058% [303 0200 500009? 058,—. 30.3 EEO zuaoha‘q. 055,—. 303 >630 20005.3. 052% Equ>02 £08594.

Em 86-80 00% En co“. .E0 092 Ea comm - Ea cog. Ea co“. -Ea oonN_

 

 

 

 

 

 

82 no 3: fiazfi
000."— 0:2 2:2 05 1.3.... 80:00 .0 00:08.3:—
Sam 80.5.20 E5 0300 .83 525

 

74

394.0 .232 500 3... 03¢

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0590 EB Ewt 0020: 3008965 0E3. nwsefi ”30:

So coo a: 98 13 0.3 v. _ n q? 33.0 5?“: 2&0 ”was?
_ #2.. No.2 :6 3.0 3 .m. vNN— wmd no.0 €5-53 $30 00008035
N9V 0.2. 0mm fwm «1mm 00. 0.0m 0.: cm N 0.0m ndm N. swsoFF 0020933
mNN 0.0N ow NNN m.0N 3 0.0N Q3. 3‘ 0. _ m 0.00 _ _ smack: ccsoemam
w.0m mam MN N.w_ 0.: N_ 0: 0m. 0 0._ mg N 0.5.0 a3 00:00:38
0.0m ON. 50 N.0_ We. 0mm 93 0.3” 0N0 oN 0.m 0m. nwsohc. ncsonfisom
b. _ m fen Om W: 0.: I 0.9... 0.02 NN m4 EN N_ 52.0 a3 0:302:02
0.0m N.mN Kw 0.2 m0 03. 0.: Wm 0mm 04 N0 .2 $20.2... 00:02:02

30:00 :30 290 63:00 Si 29: 65:00 :30 393 3550 SE 220

mp<0m .2256: 25:5 whom aging? 2:25 Wham $05.50.. u53> mbfim ..B~_§ao.. 2:25

303 59:00 zuaohmma‘. 0E8... 393N300 zuaohmmt. 052... 30$ Afloo 200000“? 058,—. 303 330 2000590 058% 52:30.2 gaging

Eudochm 00% ENG?— -Em oonN_ Ea ooncha 8;. En oo; .50 COUN—

 

 

 

 

 

 

82 .m :2 .3355?

003. 0:2 2:2 05 :80. 25:00 00 000888:—
Eam noagaé 590 095m .33 580

 

 

 

 

75

00.00 EB Em: 00:.0: 3:00.260. 0002. £00.00 “0.0:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~00 :00 0.00 0.00 30 0.00 in 0.00 33. 50:3 0.2.0 0022;.
0... .N :3... 000 Saw. Nw.m. Nm... 3:0 who $0.00.». 00.00. 000008000.
..00 0.9m 0mN 0.0m .00 0... 6mm ..mb mmN w...“ .00 N. 0 3000... 000000003
0.0N ..xN 00 m.wN 0.0N 00 m...N 0. . m .0 ..NN v.0N m. :w:0.=.... 000000000
0...». 0 m... 0N 0h. Wm. 0 0.0. 0.0m . . 0.0 0.0 0 E0... 00.. 00000500w
m.wN 0... 3.0 N... v... 3.0 :3 N..N 0: ..N 0... mm. :wzoih 00000500m
m..m 0.0m 0m 0.... msN 0. mam 0.5. mm m.m NM .. E0... 00.. 000005002
. 0.0m “MN 0nmllun0wm. m0 .- .3. -.. .m. ..w 5.». 0.. ..0 w... cmsfiflwswmfioz
.00000 00.0 30ij .00000 00.0 305' .00000 00.0 20>. .00000 :20 30>.
mb<0m ..00~_E:0O.. 0E:.0> mh<0m ..00~_E:0O.. 0E:.0> m...<0m ..0056000: 0:5.0> m0.<0m ..000E00O: 0:5.0>
140.00. >0300000000¢I1 000:0... waflm..0mifimol:00¢i 0.0.0:... 140% . >0.0QI_H010050|<WI. 0.0.0:... 1400ma.0m[00000m0fll 0000:... 0005255. 000903.
E00000-Em00”m E000“. .00 SN. E0 000;:0 00“.. E0 00“. .5: oouN.

 

 

 

 

 

 

0020:. 0:8 - 020:0 : 8:: 0000 .o 0.83.
0000 0...). 00.2 000 .000... 000000 .0 000002000.
0.50 80.5.9.0 500 0:00 .33 5.50

 

 

 

 

 

.0 «00 .000—2 0.09 00.0 0.00...

00.00 0.00 Em... 000.00- 30080.65 00000 @0000 “0.00

 

000 0.00 :0 0.00 .00 0.00 0:. 0.00. 33. 50:3 226 00.222

 

0N.NN M0,: 0m.0 mmh 00.0. .0... 0m... .00 30-00.). N200. 0000000000.

  

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.0m 0.Nv MMN 0.; 0.0m 00. 0.00 M00 ..VN N.0N 000 N. 0 000.... 000000003
w.0N va .w WNM 0.0m 0m 0.0N 0.0m 3. 0.wN YEN .0 :wsofi... 00:00:00.0
0.0m N00 NN 0.0. 0.... 0 0.0. N... 0 0.0 0.0 0 0.5.0 00.. 00:000.:0w
N.wN ..N. 00.0 0.0. 0.0. m0... ..N". 0.0. m0» ..N N0 00N 03:00.... 00000500m
0.0m 0.0m 0m m0. 0.NN 0». 0.0.0 0.N.. ..v 0.N 0.N .. E0... 00‘. 000005.52
0.0m 00m ~00 . ..m .llt 0.0 000 IV... N0 .0». ..N .0... N: $000.... 000005002

.0000 0 00.0 30.20 .00000 00.0 4003 .0000 0 00.0 EPQI .0000 0 00.0 god

m...<0m ..00~::00O.. 0E:.0> m...<0m ..00~::00O.. 0:3.o> m...<0m ..00~.E:0O.. 05:.0> mh<0m ..00000000: 0:3.0>

.000. 00.00. 00.00000<l1 050:... 30%.00 00.00031 0000:... .000. QED 0000000... 00.00... .000. 3.00 0000.000fl1 00000... 00000963. £000.00...

050000- E0 006 E0 00“. .50 SN. E0 000-000 00”.. E: 00“. .E0 00HN.

 

 

 

 

 

 

 

000. H0. 002 £00.:

 

0000 0...). 00.2 000 .02... 000000 ..0 0000000000.
0050 803.20 30: 0.5.00 .53 580

 

 

 

 

.0 30 .000—2 0.99 00.0 0.00...

76

00.00 50. 0.00 000.00

3005055 00.00 @002. 0.00

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a. _ w 0.3 02 0.8 S... 0.3 0.? mm .8... 2.080 00.6 ”Mam-am.
0N0 5:0 00N 00N 0m.0 mm...» ..00 5:0 €0-00>31.~0!.00 000000005.
m..m N..0 0 00m 060 0. 0.0N n.0m 0 0.0N 0.0m N 0 00:: 00000303
0.00 0.0 00 m...” ..N0 00 0N0. ..Z. m0. 0.0m 0.00 .. $005.0 000000000
0..... 0.0. .- N... 0... h m... 0.0 N 0N 0N 0 E0... 000-. 000000.00m
0.. . 0..... m00 0.0 N0 mmm m0 m0 00N 0.0 mm mm $000.... 0000000000
..0. m0. 0 0.0 0.0 0 m... 0.0. E 0.. 0.. N 500 000 000005002
0N. 0.0 N...- 00 0.m 00». 0.0. 0.0 NN». W0 0.m mN. 0.005.0Pr 0:0WW0002
0:80 :20 2.5 8:80 8:. 23W .280 03: 20.0 8080 80 30>. - ||
mP<Um ..0005000: 0E0.0> m...<Um 0005000.. 050.0> m..-<Um ..0025000: 0E0.0> m...<Um ..000500bh 0E0.0>
.003 000.00 “0000.00.01 00.5... .003 0.0.00 0000.00... 00.00... .003 >-0.00 00000-00< 00.5..- Juqu-me 00mm“? 0000:... 50505.). 0000:00<
50000-50000 5000“. -Em00UN. Ea Sum-5000.0 5000“. -5000HN.
000. K .32 N033...
000w. 0...). 090.0 000 .02... 000000 ..0 00000305.
0050 80.3.20 500 0:00 83 .080
. .00 .0003. 0.00 00.0 0.00...
0500 50. 0.00 000.00. 0.005055 00.00 50050 “0000
«S 0.8 now 0.00 2.0 0.3 0.2 mm 33. £08.. 2&0 ”mew-H
3.0 00.0 00N 0m.N 00.0 00.... m. .0 0. .0 0..-003 M200 00000305.
N.0m N..0 0. 0.Nm 0.0m N. 0.. m 0.; N. 0.0 0.0 0 $000.... 00000303
0.00 0mm .0. n.0m N00 00 0N0 ..N0 00. N..0 N..w m .3000; 000000000
Wm. ..N. 0. m0 0.0 0. Wm. ..N. N N.m ..m N E0... 00‘. 000005000
0.. . 0.0 00.0 ..0. 0.0 0Nm Wm 0.0 00N wh 0.m mN $080.0 000005000
..N W0. 0. 0.0 0.0 0 0.0. 0.0. on 0... 0.. N E0... 00‘. 000000002
0.0 0.0. NNO 0..- ..0 n. m 0.0 m0 0mm 00. Wm 00 00005.0 000000002
.00000 00.0 2:30 .9500 00.0 20.3 .00000 00.0 20.0 .00000 00.0 4003
0:00 .8580. 2:25 0500. 03880.. 28.; 050m 03880.. 28:5 0500 .8258? 2:25
.003 000.00 00000003 0000:... .003 0.0.00 00000000. 00.2... .003 $.00 0000000,... 00.8... A 003 020.00 0.000590 00.00... 50505.). 00000000.
50000-50000 5000“. -EN00N. 50 000.50 00.0 50 00“. -5000HN.

 

 

 

 

 

 

 

000w. 0...). 000.0 0:0 .02..- 000000 ..0 00000805.

0050 802.030 0.00 0:00 83 .080

 

.0 8.000.). 0.00 am. 0 030,—.

08. .0 32 .088

   

77

05:: EB Em: 0020::

35:56:. 052. 5:85 “20:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5; ohm New odw mgm 93 ~43 mm 303 Ewcox. 2&0 uwabfim.
Ed maN NNN vmd Ed 36 wmd omd AE-:0>0J?BD 5:832:—
mém N; 3 o.vm 0.; a aNm _._m 2 ..3 0.3 w z :85. 050933,
mdm Wm: we Nwm 5.2. we wfim 000 a: :3 .mHN v :wzofih 2398‘”.—
vw as m. 0.x 0.0 v od od 0 w.~ EN 2 EE. :3 0:002:30m
ow «mm m _ v Wm 9v «mm Yw Go Fm .0 Wm 02 @025. uszonsgom
N6 0v m vN o._ N v.0. w.m 3. ..m hm n 50H :3 0:202:02
«lull 0.0 on». m: 3.: mm». n.0,: os ommll , iwwc [104m I: @225 0:202:02
7550 ENE 20.; v _:0::00 ENE 29$ .930. Si 293 1:200] :20 305
m._.<0m ..:u~_E::O.. ino> mk<0w ..BEEEO; uE=_0> mh<0m ..uoEEzaoz 0E:_0> mk<0m ..00~::::O_. uE=_o>
Auomvfiglmuaiegfl 052% 33.0 Nquflmmmmfll 050C. 1. 33.0 A200 Logan? 053% 303 gammam’x 052,—. E05302 £08591
EmooHo-Emoonm Emoo; .8352 En 00%-:3 8;. En co; -anonfl
000— .o 52 “335E.
03¢ 2.2 .535 0:: .8; 30:0: 00 cocoons“:—
055 808.2: 5»: 956m .33 5.5m
:— 30 .0602 Sun .66 03:8
0590 EB Ewc 0030:.” 355958 gnaw: .3085 “80:
néw ohm ..ow odw on own «..Nn mm 303 choq 2&0 ammuo><
ww.m mo.»V mm: mm; ww.m C .v 95 Rd 03-55 $30 50896::
vdm N. _ v v oém vawm 2 03 ~._ m 2 me 0.9“ m 3003.: 05.0983,
ode 5.; oo— mdm 5.3 mm mdv 0% «2 03 0.3 v .3085. 95083
od. 0.2 N. 9.0 o.o o No 0.x 0 od ed o ES. to; ccaonfigw
m.w :6 wmv v.0 Qm wmm m6 No wow 0.0 5m 2‘ :9.an 0:00:50om
0.: cs v Wm _.m m w.: _.o_ 9. od 0.0 o E3. :04 0:255:02
{ 9m Wm. mm: E ..v 2 m 5.0 v.0 03 m0 9m 03 3385. 0:00:50on
7 33:00 :30 30.qu 65:00 :20 Sub 30:00 1E0 303 65:00 :mE EPQ
miom ..B~§_ao.. 2:2; mh<0m ..BEEEO; 2:23, mbfim ..B~E&o.. 2:25 mSOm .3052? 85:5
303 >615 0.82%: 052.: 303 $30 5005“? 058% 303 have 5859‘ 058,—. 30$ 3.00 5833‘ 053% 0:08:02 28030.4:
Emguoécmoonm Emoo; -Emoong Ea ooquba 8K. Encou— -Ea comm—

 

 

 

 

 

 

uaom ~32 :o>o_m 0:: E: 80:00 00 5:03:85
:25 803...»: 3»: 0.50m 53 58:

 

.5 :0—082 209 uni 033—.

82 .w :2 508283

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

78

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

050:: E3 Ew: 0030:. 00:00:06.: 050.: $095 :06:
20 3.0 200 0.00 .0: 3.0 New 0.? 33.: 50:3 020 0020.5.
20.0 8.0 0mN MNN hmé 00.0 30 Omd 24.63 40:00 6:000:06:
n :0 v 5: O: m mm : 00 m: : on N :m : : w oN n 0N m :05th 0:000:03
N.wm. N.mw cm :00 0.00 no 53. Owe 00: :.om wbm m finch; 0:300:03
:0: 0N: :: N0 Ne w :8 od— m n: 0.: 0 E3. 50: 0:000:30m
:.o: 0.0 w; 0:. Ne. mNm 0:: m0 wnN :0 Wm mm :w:0::.:. 0:300:000m
rim: w.:: :. 0N 9m N o.:: m; mm h: m: N E00 50.: 0:605:02
m0 N.:. 3:. 0:. :0 on :0: 0.0 «NM m0 Om OM: r3095. 0:255:02
6:60 0 ENE 20.: 60:0 0 :05 20.»: 6:60 0 :05 20 C 6:600 ENE 2103‘
m.:.<0m ..0006500: 0E06> mF<0m ..00~:::00.. 0E=6> mh<0m ..00EE::O.. 0E:6> m.:.<0m ..00~::::O.. 0E06>
:003 >702: 20:00:92:l 058,—. Jmmvh: 3:00-:00dmmfll 052,—. A000: N200 603% .1 050;. A000,: .375 60049:; 058:. 60:66:): 60050.4
a 000 - En: comm. Emdo: - EmoonN: Ea ocuw - E0 8;. Eu ooh: - E0 CONN:
0m0:0>< x65 - 005:6 : :35 006% 6 0:003
000% 0:2 :0>0:m 0:0 ::0:.:. 000:0: 6 0000080.:—
330 50.330 :20: 0:00 83 580
«a .00 .0003: Son um... 030,—.
050:: ES Em: 0030:. 06050.60: 050:: @085 ”0:0:
000 30 0.; 0.00 30 0.8 0% mm :83 50:3 2&0 00:02
Nmé 00.». EN EN £0 woe 000 mmd 50:03 00:00 00:03:00:—
vdo N6: N: m.mm 00m 0: 0.1m N. : m :: N.0N 0.3 N 30:5. 0:00:60?
0.? 0.3: 2. 000 0.5. mm wdo 0.2. w: QwN 0.0m N :wsoEb 0:000:80
06.: WE h: ma :.w w: 0:: 0.0: 0 od od o E3. 50: 0:000:30m
wd m0 wow wd: N.v Nmm 00 N0 00N oh 00m 8: :wsoEF 0:000:53.
v.3 Fm: 0 wb N.:: 0 mso: 00 m: o: n: N 0.5:. 50.: 00000562
No 1] 5w 2.0 N6 N0 0mm 0:: 0.0 M: m :6 0.». ow: :m:0::.:. 0:085:02
6:6.0UII :50 20> 0 6:060 ENE 30.»: 6:60 0 :05 205 6:60 0 :05 E0>fl
m.:.<0m ..00~:E::O.. 0E06> m.:.<0m ..00~:::::O.. 0E06> m.:.<0m ..00~:E::O.. 0E:6> mh<0m ..00~:E:&O.. 0E:6>
600: .960 603091 058,—. 60$ 50:00 60060.1 050:.:. A003 N0:0Q :08??? 050:. 603 .360 20:00:93. 050:. 62:06:): :000:::<
E: 80-:5 00% En ooh: - Em oouN: E0 oouw-::0 cog. Em con: .80 092

 

 

 

 

 

 

 

08: .2 32 .50:
000m 0:2 0305 0:0 28:. 000:0: 6 5:000:06:
0030 802.2,: E»: 0:00 83 5.50

 

 

«59:0 600$: .33 0m.v 030,—.

79

0000:: E0: 20:: 0030.:

3:506 000:: 2000:: 0:0:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

     

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E: 000 - :3 00;».

 

E: 00:: - En: 00N:

 

E0 9:: - E: 8K.

 

Ease: -anonN:

 

 

00:. 0.0:. 0.00 0.0: 0.: 0.2 0.; 00 :80 5003 2.10 00:02
0:: 2.: 00.0 8: 0m: 00.: 00.0 00.0 02.20.: .300 00000036:
finn 9m”. 0 mNn 0.0In :.: m.:n N.:m m 0.0 0.0 0 E00 00.: 0:035:02
00 «am Em 0 v 0.0. 00N we 0.0 0:. N0 0N MN 5:00:20 0:00;: .
0.0 v. :: 3 N0 0:: 0m 0.0: :0: m: 0.0 0.0 0 E0: :0: 0:03:00
0:.” :.:. m5. . 0.: v 0 EN 00 N0 00N Nc 0N 0m 2000:: 000053.

”It. {I .1 I. ‘i .lell .|.l|; . \

75:00 00:: 200: 05:00 :05 20.0 :00000 00:: 20>: 6:000 5:3 20.:
mh<0m 03:530.. 0E0:0> m:.<0.n ..03::_30.. 0_::0:0> mF<0m 0036.30. 9:20.) ”5.00 ..00~:::30.. 050:5
300: 00:00 90003. m<. i 2.05:. :00fl30m: 2000l::. 4 0.02:. :00: 50:00 2000.39: 00:0::. J30: >0_0n:..:000::mm.|. 0.02:. 308952 2000.33:
E0000 -5500 0000: 00000:: .5000 .5000“: 52.0.: -5000:
000: .0 ~02 N00000003
000M: 002 00:. 05 0:30:33 30:6 30:00:: :0 5:80:25
0035 60:05:03: :03 83.0% :93 5.5m
: DO 1102 0:00 006 030:.
2.03: E0: Ewt 00200 3020900: 0.00:: 2000:: 00:0:
0.0: 0.2 0.: 0.00 00:. 0.0: 0.; 00 :80 5003 30:0 aspi—
000 2.: :0 3.0 00.0 02 00.0 0: .0 0.220 :05 50802::
0.9. 0.2. N: 0.3 0N... o N.:0 0.3m v 0.0 0.0 0 E0:- :61: 00035.5
0.0 :.n 00N Nd : n. 00N 0.0. 0.0 MNv N0 :.N on 2000:; 0:000:
0.n :20: E. N... 0.0 0n 0.: 0N: 0». 0.0 0.0 0 50:. 00.: 0:000:00
0.: 2.: 0:n . 0.0 :.:. 00N 0.0 0.0 mmN N0 :.N 0N 20022.0 0:000:00:
60:00 50:: 20>: 2:000 :03 20.: 20:00 :33 2'0: 6:4:00 :3: 20.0
m.:.<0m 000.5030. 2:005 mh<0m 0806:) m:-<0m .0038030. 0E0:0> mh<0m .0050030. 0.::0_0>
:000: 00:00 2000: < 058:. :000: >500 2000: 8.02:. :000: 00:00 200030.»: 2.02.: :000: .300 200059... 052:. 6080.002 20003:?
E: 000 . E: 000m 6900: - E: 00N: E00000 - E0 000 E: 00: - Ea 00.2
000: .0 0:02 N00000:.
000.: 002 00:. 0:: 0:30:33 .0005 0:00:05: :0 :00000025
inflow cat's—gm 01:00 WH<Um :0»..— 5.5m
:auao .0002 "0:09 006 030:.
8.00:: E0: 50:: 000:0: 3:050:50: 8.00:: 20002: :30:
0.0:. 0.2. 02. 0.00 0.2. 0.2. 0.: 0m :08: 50:01: 0:000 00203:
0.0.0 0:..N 000 N60 000 Nm: 00.0 00.0 02-20.00 00:00 00:03:80:
0.00 0.9. 0|: 1mm 0.00 0 «..Nn ndm N 0.0 0.0 0 50:. 00.: 0:009:00
m6. Wm mom 040. N0 0mN m0. 0.0 09.. N0 0N 0N
0.0 0.0: 00 Wm 00 :v 0:: 0.2 :0. 0.0 0.0 0 E00 0.0.: 0000.000
0.: N.m: 0mm. 0.0 mi mnN 0.0 0.0 .02 N.0 0N 0m 20020:. 0:002:00
"H” ..
33:00 :07: 20>: 75:0 0 00:: 20.3 :0:::0 0 :05 20.0 1:300 :07: 20>:
m.:.<0m ..0Q~::.::30.. 050:5 mk<0m ..0000031P 080:5 mh<0m 03:00.30. 0::0:0> mh<0m 0000030.. 00:075
:000: 00:00 20005.9... 053% :03: 00:00 2000:0300. 050:0 300: .4300 2000300. 00:00. A000: . 0:00 20:03:? 9.08:. 5qu.002 2000.334:

 

 

0:3: 0:2 :0:. 0:0 003:0: 6906 0:093”: 6 5:08:80:
0.30 8031.0 :38 0.5.00 .33 .080

 

 

000:6 a2 «0:0

 

8O

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

  

 

 

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

      

 

 

 

 

 

 

 

 

 

 

 

33:0 .032 ":8 3.9 033—.

3:5: EB Em: £535 35.:ng 3:5: .332. 2::
m: a: o: :2. a: a: c: o a so: 5ch 3.5 as;
3.. S. m 2.0 end bx : 2; sec m: o ”2-5.: .35 5.59.6.5
«.2 a: N. a 3: a 34 v.9. q a: a: o E: :3 233552
wé 5v. Rm Wm a... 3m «an 3 K. m; m c c m cm sunfish acacniam
5.0 w; _ .2 ed 56 c». n a. c E xw : c c o c 53h :5 35:57.9
.36. m. a .99.. - x: w v 19m x 2 .5. . ,m w. 9% ..n c - bu. 3. €312.29 3:39.795
7550 :aE 9?: 3:50 E43 :3: 75:: u- :3... :3: 7590 4'43; _ 33.: If: I
when; ..3:E_EO.. 9E2?) ”3.0m , ..uuwzczao. 98:75 mF/fiwm _ 53:550.. 2:232 mkrnvm ..32533: n 2:33)
1.0!; hammer €3.34. Btu; 1., Wad .9203 .5325“? 3:8... {mm}. 1.12. 22: :97... _ 2.22% r.-. we! 33% 51.3.»34l4. BEEP 805.952 52:9?
55593;; J 59:: .EE.....~_ .Efx; .9531 535.? 6.35 2 . -
8292 ZED . 022?. s :35 «52>. .0 383
33. 9:2 5F can x3313 .326 Begum .0 5:362:—
€am Suzié 5x. whim .33 5:8
. EC 1.82 .2 36 03-h.
9E2. E3 :3. 33.3. 356ch 958. 3:25 “39:
3h :2 a? 2: m: o: In on 33. 2.5 933 $23.
3: cm: 3d xoc on: ma _ 8d 8.: CF29: >25 c2688.:—
sdn 0.32 m~ New ”.2 c 53. N3... N 9.9 c: c 53. :3 333551
at c K. 5%. MA m4 5». ca. m c m3. N: c H mm swash; vcgéwam
56 c4..— cc 0... _.m 7. cc v.2 nv. at c c c 53. cow— 9:553.
:2 n 3 Em. - [ x: f. ._ mm x c x q 3H m: .ca 2 .3325. 2525.
1550 :3; Eu: .3250 :nE By; :5ch . J 5.5 293 3:50 :55 2»;
3.40m .33.:5a0. 223.5 $5.9m ..BEEEO. 2:23» akam ..33530: 25.75 mhxfivm ..vozEsacz 0:575
3ng wind :92...me¢ . k .453». Cod >EuQ 22...:QQ/«éL watch Jud: iguflhuacamzx uEEP 3qu ..NE& Loafing»: 4 uEEF .5525: 5329.?
En cope - £5me Eu 8; . Edocfi. Ea och . Em" ccfl En £11m Ea coa—
c©2 .E 3: Naming
was”. 012 Ed. «Ea Aanfim .326 3053— he :2.3€3£
3.5 553.2“.— 58 $58 .53 5.5m
53:9 1602 ”Son 006 03:.
053. Es. Ew: ova—us 35839: 955. .3225 ”89.
0.: 9.2. ”no ads man on» «In on 393 59.64 3250 owiu><
no; 3..— owd Etc Sue on; cod No.3 C955 haun— c2833.:—
adn Gee a win. wdm 2 3.9. mdn o 3.: cc 0 En... c3 vcsfizcoz
ma. Fr. 3.». N6 N... 3m 5m 3.: a: NC on an .3325. vcacfiflm
:. a n cc c4. ca .3. 0.". x6 Q. N: 3,: 0 ES. :5 959:3
3A: ll. 7c Ic._ c 3 ~ we. 14mm .. Cw -. 7m. emu c: en fin 5.52:; 3:37.43
1555 5.5 3.93! 7550 can [3.9; TECOU 5.3 3...: 752.0 5;; Eu:
3.60m ..33E:a0.. 082.5 m.r<Um 53:530. uE=_o> mh<Um ..uudEsaoz 2:275 mH<Um ..UuEEzaO: 2:275
Cu: >675 gossip—fill 9E5; ‘3.qu Ziolmunwzdmfll uEEk 39$ 33$ muzzgwfll 958;. 33.. .330 zuncflwm< 958,—. 28.5.52 5.332
ca :3 - ca 8;. E 9: - 2.392 E.“ 8.x . E of. E... Q: - Ea :92
oafl .a an: nave—E...
93m 32 sub 93 .3313 895m Raina .0 5:89.35
:55 5:31.”... 38 mini .63 5.6m

 

 

Chapter 5.
Comparison and Interpretation of Modeled Delay

In the preceding chapter, intersection approach delay was calculated for the four South Lyon
intersections when they were under the control of both SCATS and fixed time traffic control.
The delay for the critical approach movements was presented in Tables 4.3a through 4.6f.
Each of these tables contained information for the four daily time intervals of 12:00am to
1:00am, 7:00am to 8:00am, 12:00pm to 1:00pm, and 5:00pm to 6:00pm. The goal of this
chapter will be to analyze these data to assess the differences in delay resulting from the two
control strategies. The significance of these differences will be determined by using various
methods of comparison, including direct and indirect statistical tests. The comparisons will
also assess the change in delay both within and between the various time and location data
groups. It is expected that the analysis of these differences will allow interpretations to be
made relative to the change of signal control in South Lyon. Taken together, these
comparisons will also demonstrate how the control philosophy of SCATS impacted the delay

parameters of intersections in South Lyon.

81

82

5.1 Total Intersection Delay

The analysis of intersection delay was accomplished through various comparisons of the total
intersection delay in the two data sets. The definition of total intersection delay is the sum
of the delay experienced by all vehicles at all approaches to an intersection. It is usually
calculated from the product of the average vehicle delay and the total number of vehicles
during a particular analysis interval. In this study, an average delay value was calculated for
each approach movement during each signal cycle. The total delay came from the
summation of these cycle-by-cycle delays recorded during the full analysis period. The use
of total delay as a measure of effectiveness is helpful because it illustrates the overall impact
of SCATS control on the operation of an intersection. The comparison of the total delay
statistics also helps to illustrate some of the philosophical differences between fixed and

adaptive signal control.

The underlying philosophy of SCATS is to balance the degree of saturation on all of the
approach legs to the intersection. The goal is not necessarily to minimize total intersection
delay. SCATS manages the signal cycle by strategically distributing green time to each of
the approaches. This green time allocation does not, however, guarantee shorter delays to
major and minor street traffic. The benefits of this strategy are numerous, including a more
effective use of available green time, shorter minor street delays, and reduced driver
frustration. The obvious drawback to this approach is the potential for increases in total
intersection delay and decreases in signal efficiency on a system wide basis. Since SCATS
does not minimize total delay, traffic delay may increase resulting in additional wasted fuel

and increased exhaust emissions.

The fixed time phase split plans for each intersection and each analysis period were
determined with the SOAP software package and were shown in Tables 4.3a through 4.6f.
The SOAP optimization structure attempts to minimize total intersection delay by

distributing a higher percentage of green time to the highest volume movements. The SOAP

83

delay minimization often comes at the cost of inordinately high delays to minor street
movements. The underlying philosophy of SOAP is to inconvenience a relatively small
number of drivers for the overall benefit of the whole intersection. The comparison of
SCATS to an optimized fixed time system will ensure that if a performance improvement
is documented, it will result from a comparison of two efficient systems rather than a good
one to a bad one. In addition to analyzing the delay difference, the comparison of total delay
in SCATS and an optimized fixed time operation will help to contrast the objective function

of the two systems.

The comparison of total intersection delay used both a percentage comparison and specific
statistical testing. The tests, detailed in the following paragraphs, resulted in many expected
and unexpected outcomes. To assess the intersection delay differences on a system wide and
hourly basis, a statistical comparison of total intersection delay was conducted. The delay
data at the four intersections, for the four separate time intervals, during the five days of data
recording, resulted in a sample size of 80 records for each “before” and “after” period. The
two samples were contrasted using a paired comparison t-test. In a paired comparison the
difference between the two separate samples is assessed. The computed means, variances,
and standard deviations for total intersection delay in the entire South Lyon system as well

as each separate time period are shown in Table 5.1.

The mean difference in total delay for the total population was +1.46 vehicle hours. The
positive value indicates that mean total intersection delay parameter increased after the
implementation of SCATS when compared to fixed time operation. A paired comparison
t-test was conducted to determine the statistical significance of this result. First, a 95%
confidence interval was calculated to determine the upper and lower bound of the statistical
mean. Using the standard deviation of 2.67 vehicle hours, it was determined with 95%
confidence that the average change in total delay was between +2.05 and +0.87 vehicle
hours. To test the hypothesis that SCATS caused the observed change in delay, an inference

test was also completed.

84

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ma... © £8.93»: mad © 3853.»: ma... © £852.»: m3. © £8525: mad © a:
:52 05 BUM—Hwy— ==Z 2: HUME—~— ==Z 05 hum—4mm :52 Q5 HUM—Hm FUN—dd BOZ O:
ofim wim 2am mm.m and Co 2:9 votompo
mm; 2: mad mod acd 95:3 05 Lo 2:? 182.5
Sub 285095
3.0 mm; ovd amd 8.0- 9.25 595
mo.m Q .m E; mm.~ m ~ .o 225 $304
BEBE vacuum—80 0&3
ow on om om ON 35 295m
rod :6 3: mod mud coarsen— wen—654m
2 H 3.2 mod mmé nod 859$
8“.— omdi 5; on; IN 86 :82
=Eo>0 En cone - :15 8% En co; - Ea so”? :8 oonw - Em 8K % co; - Em ooumm

 

 

cog .53 - 50 .32 .Cnéofi chm—om 5:08:55 :38. $325. 5 85:55
beam :oflaggm >209 mB<Um :omu— 5.5m

 

ES .3..." a. 58358 .8933 E 2.3.

 

85

A statistical inference test uses the data to determine whether a change between two data
groups is the result of an applied treatment (SCATS) or if is the result of random variation
within the data. A rejection region at a 95% level of significance was used to test the “no
change” null hypothesis. The critical value of the t-distribution and the t value for the
observed delay data for the full data set are shown near the bottom of Table 5.]. A
comparison of the t values shows that the value for the observed data (3.70) is greater than
the critical t value (1.98). An observed t value which is greater than the critical value of the
t distribution provides evidence that the observed difference in the two data sets is outside
of the range of random data variation. Therefore, the null hypothesis of “no significant
change” is not accepted and it can be concluded that the observed increase in total delay is
statistically significant and is not merely the result of normal fluctuations in total intersection

delay conditions during the week.

Additional paired comparison t-tests were conducted to analyze the hourly differences in
total intersection delay. Values for the mean, variance, and standard deviation for each of
these groups were calculated for each of the four analysis periods. Table 5.1 shows that after
SCATS, the mean value of intersection delay increased during each of the analysis periods.
To test the statistical significance of these findings 95% confidence intervals and hypothesis
tests were conducted for each of the four analysis periods. The confidence intervals showed
that during the 12:00am to 1:00am period the mean difference in total delay could reside
within a range of performance improvement or degradation. The confidence interval for the
mean difference in delay for the other three periods were all greater than zero, or within the
region of delay increase. To test the statistical significance of these values inference tests

were conducted for each.

As shown in each column at the bottom of Table 5.1 the hypothesis tests conducted for the
morning, midday, and evening periods all showed that the observed difference between the

two data sets to be outside of the range of random variation. Thus, like the conclusion for

86

the overall data set, it was possible to reject the “no change” null hypothesis for each. The
conclusion that can be drawn from these tests is that there is evidence to support the claim
that SCATS signal control increased the total intersection delay over that of fixed time

during these intervals.

The inference test for the midnight period showed an inconclusive result for the effect of
SCATS on the total intersection delay from 12:00pm to 1:00pm in South Lyon. The increase
was 0.02 vehicle hours at all intersections, or an average increase of approximately 0.30
seconds per vehicle during this period. The primary reason for the inconclusive finding in
analysis of total delay during this period was the combination of a very small delay change
and a comparatively high degree of variability within the data set. The variation resulted in
wide confidence interval and a low observed value of t at which the null hypothesis could

not be rejected.

The data were also analyzed using non-statistical methods to identify trends within and
between the data sets. These analyses included comparisons of the percentage difference
between the two data sets and the calculation of various weekly averages for the groups of
the modeled delay output. A tabular comparison of average intersection delay statistics is
shown in Table 5.2. The delay difference in the two systems is shown in terms of a percent
change in the right column for each time period. This value represents the optimized delay
minus the SCATS delay, divided by the optimized delay. Thus, a negative percent change
value indicates a delay savings in the “after” SCATS period and a positive value represents
a delay increase. The composite hourly difference for all locations are shown at the bottom
of each column group. The daily average difference is shown at the far right side of the table.
This table is particularly useful because it illustrates some of the general trends within and

between the two sets of modeled delay data.

87

 

    

ow8o>< note.“—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

whhm No.5 2 .0 $02 3.3 3.: $0.2 and 006 0.3.8 Ed 9:. $0N 05.0 00.0
$2.? 3.0 0 _ ._ 0.3.09 no; Sam «3.0m- 3.0 $0 «0.9“. 5.0 mm; $0.00? 00.0 3.0
$0.0 06m chm $00. 10;. 00.”. m. _ .m 0m.~ mmd $0.” Rd 03.. $63.. 3.0 0m.0
mica 062 3.2 $~N~ mfmm xmbm $0.an _N.m_ v0.0 3 _ .vm 00.: 0”: $0.3 v0...” 3:
"..N? $.0— mm.» men. 3 on, _ N 212 $0.8 00.0 mph $0.0m 9.3.. mm, 2 $0 0v. 9:0 3:0
30:00 5.2.0 7556 :20 7503b :mE 75:30 0.1.5 759.0 Sam

856 Exam ..Bséao. ““55 WEB ..B:§ac.. “wage 35m .3520. 3.56 whim ..B:EE0.. 9?..5 MSGW ..ceésao.

58.80 4.4 3.. .EuQ 53:35 .5980 N509: >25 .Euowwuflm. E850 5-53 mfioa 533331 Euubm :45: main 5.5325 EH50 330312.49 .5832:—

uwflo>< EGO E 000 - Em 006 E 8; - E90012 Ea comm - Ea 00;. En 00; - Ea 09mm

 

 

 

 

 

.3. 9:2 not>3xm gm 2 M
08m oz: 5.65:3; 8.:ch
081 902 5.5.8:. 30:20
053. 2:2 .457:an unseen

       
 

8:93.35

 

 

03—90 00.30322 .30... 0o soar—«9:00
“new seize 35 95.8 53 5:8

 

son 8:85.... .sfi .9 ._

  

 

88

The percentage comparison of the data shows a trend toward increased total delay during
periods of increased traffic demand. This trend is evident both in the time of day and
location of comparison. The values of total delay difference shown at the bottom of Table
5.2 show an increase in total delay as the time interval moves to periods of increasing traffic
demand. The delay difference in the midnight period is +2.6%. However, the delay
difference in the morning, noon, and evening peak traffic periods increases to +20 to +27
percent. Overall, SCATS control showed a system-wide increase of 23.7% in the average
value of total intersection delay. The last column of the table shows the overall average
percentage change in total delay by location. In this column it can be seen that the greatest
increase in total delay occurred at the highest volume intersections of Pontiac Trail and Nine
Mile Road and Pontiac Trail and Ten Mile Road. The delay increase is considerably lower
at the intersection of Pontiac Trail and Eleven Mile Road, an intersection with considerably

lower volumes than the Nine and Ten Mile Road intersections with Pontiac Trail.

While the predominant trend is toward a total delay increase under SCATS, the percentage
comparison table shows that total delay also decreased during several of the analysis periods
and at several of the study locations. In total, intersection delay decreased during 7 of the 16
sample periods. Given the control strategy of SCATS, this reduction in total intersection
delay is somewhat surprising, especially since the optimized signal timing plans were based

on known volumes.

The most notable of the decreases occurred at the intersection of Ten Mile Road and Reynold
Sweet Parkway. At this location total intersection delay decreased during all four of the
analysis periods and showed an overall average decrease of 47.7%. Total intersection delay
also decreased during the morning and evening peak periods at the intersection of Pontiac
Trail and Eleven Mile Road, 2.0% and 0.5% respectively, and during the midnight period
at Pontiac Trail and Ten Mile Road when total delay decreased by 40.0% after SCATS.
When taken in combination with the areas of notable total delay increase, the areas delay

reduction appear to illustrate a relationship between volume and total delay.

89

The comparison table shows that the locations and time intervals with the highest volume
also exhibited the highest total delay increase percentage. Table 5.2 presents the
intersections in descending order in terms of total intersection volume. The last column
under the “Daily Average” heading shows a direct correlation to the decreasing difference
in delay. The same correlation can be seen across the bottom of the table, although the

correlation is not directly one to one as in the case of location.

A review of the daily output statistics shows a likely reason for the apparent connection
between volume and total delay. Tables 4.3a through 4.6f presented the daily delay statistics
for each location and analysis interval. These tables also illustrate a classic example of the
difference between the effect of fixed and SCATS signal control. Under SCATS, the delays
for the critical approach movements decrease in nearly every case, except for the highest
volume approach movements in which the average delay per vehicle increased measurably.
This is consistent with the objective function of SCATS which is to equalize the degree of

saturation.

An example of this concept is illustrated in Table 4.4c, the delay comparison table for the
intersection Pontiac Trail and Nine Mile Road on Wednesday, May 8th. In 4 out of the 6
approach movements the average vehicular delay decreased. The two exceptions to this
trend are the north and southbound through movements of Pontiac Trail, not coincidentally
the two heaviest volume movements. The northbound through movement experienced an
average delay increase of 3.6 seconds per vehicle for 356 vehicles. The 727 vehicles in the
southbound through movement experience an increase of 12.1 seconds per vehicle. While
the additional delays experienced by these drivers would barely be noticed in a 30 minute
commute, the cumulative effect of this delay totals approximately 2 hours and 48 minutes.
This is the effect of SCATS. This point will be become more evident in the approach delay

analysis section of this chapter.

90

5.2 Approach Delay

The strategy used to conduct the analysis of approach delay was similar to that used for the
analysis of total intersection delay. They both incorporated statistical and non-statistical
procedures to analyze the differences between the various locations and time periods, as well
as the internal differences within an individual location and time period. The primary
difference between the two was the increase in the amount of data in the approach delay
sample. Since the approach delay data set was larger, additional comparative analyses were

required to assess the interrelationships between the various data sub-groups.

The first set of analyses were conducted to determine if, from a statistical standpoint, SCATS
was able to lower approach delay. Other comparisons were completed using data averaging
techniques. The following sections detail the methods used to conduct the analyses, the

outcomes of the comparisons, and their interpretations and conclusions.

Paired t-tests were used to determine the difference between the fixed time and SCATS
approach delay data sets. The difference in approach delay was calculated at each location,
during each analysis period for the week of data collection, including all critical turning
movements at the intersections. These resulted in a comparison sample set of 480 trials.
Values for the mean, variance, and standard deviation for each data group were calculated
for each of the four analysis periods. To test the statistical significance of these findings 95%
confidence intervals and hypothesis tests were conducted for each of the four analysis

periods. The results of the paired t-tests are shown in Table 5.3.

Overall, the mean decrease in approach delay for all turning movements was 1.59 seconds
per vehicle. The 95% confidence interval for all approach delay observations was between
-0.33 and -2.85 seconds per vehicle. From this confidence interval it appeared that the
decrease in the modeled average approach delay was statistically significant. To test this
theory, a hypothesis test was conducted using a t-test. A rejection region at a 95% level of

significance was used to test the “no change” null hypothesis. The values for the critical

91

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m0.0 © £853»: m0.0 © 0: m0.0 @ 0: m0.0 © 0: m0.0 ® 0852::
0:2 2: 0.00an hum—.5— 892 On hum—.5— 802 on BUN—H.5— HOZ On ==Z 2: Hum—:5—
0v.~- av. T 00.0- 2;- ohm- Co 2:? octane
00. T 00.7 00.7 00.7 00. 7 0,500-0 05 0o 02? Route
08,—. $8500.?”
93. WE- 03. 02.- wow. 280 5&0
mm.0- 3.0 00; 3”.— 0m.0- 055m .533
3385 8:000:00 00mm
9:. 02 OS 80 om. 9% 235
8.0 03: 8.2 8.2 m3 802.60 0323
3.9: 3.3% 8.02 3va 3.2 8893
0m. 7 0v.m- i 00.0- In 7 cm. _ - :82
=Eo>O 0!:— 000 -80 00% 30?— -En ooumfi £8000 .55000- 8000; .80092

 

 

0000 .53 - 50 .002 .Eotoomv 0209 08809». 0005.3. 5 855005
:55 803:3”.— 55 92% 83 58m

 

 

3.5 =82 < a. :83 55 350% am as:

 

92

value of the t distribution (- l .98) and observed value of t for the overall approach delay data
set (-2.49) are shown in Table 5.3. The table shows that the calculated value of t for the
observed data was significantly larger than the value of the test statistic taken from the t-
distribution table. Based on these values the null hypothesis can be rejected and we can
conclude the observed decrease in approach delay resulting from the implementation of

SCATS in South Lyon is statistically significant.

To determine the difference in approach delay values for the four separate analysis periods
four additional t-tests were conducted. The mean values for the difference in delay for each
of the four analysis periods are also shown in Table 5.3. A set of 95% confidence intervals
for the mean values were computed. The mean difference and confidence interval data give
evidence that the difference in the delay was due to chance during three of the four data
analysis periods. The midnight hour analysis period was the only time segment where both
boundaries of the confidence interval were below zero. The average decrease in delay
between the fixed and adaptive control modes from 12:00am to 1:00am was 1.34 seconds
and the confidence interval range spanned from a decrease of 0.30 seconds to a decrease 2.38
seconds of delay per vehicle. The morning, noon, and evening periods all showed a
confidence interval ranges that spanned mean values of average approach delay increases and

decreases.

Hypothesis tests were completed to further analyze the differences in intersection approach
delay. The hypothesis tests were consistent with the results of the confidence interval
calculations. They showed that during three of the four analysis periods; morning, noon, and
evening, the decreases in average intersection approach delay were not statistically
significant. The calculated t statistic for the observed data were under the critical value of
the t distribution. Therefore, the “no change” hypothesis could not be rejected nor was it was
possible to conclude that SCATS was successful at reducing the intersection approach delay
during these periods. The only period which passed the hypothesis test was the hour from

12:00am to 1:00am in South Lyon. The calculated value of the t statistic for this hour (-2.56)

93

fell below the level of the critical value of the t distribution (-1 .98). Thus, it was concluded
the average approach delay decrease of 1.34 seconds per vehicle was not the result of chance
variation within the data sample, rather it was likely related to the implementation of SCATS

control.

As an additional measure of comparison, the percentage change in the weekly average
approach delay statistics for each critical movement was also determined. Tables 5.4a, 5.4b,
5.4c, and 5.4d present the results of this comparison. Similar to the percentage change
comparison of total delay, the delay difference was computed by subtracting the SCATS
delay value from the fixed delay value and dividing that difference by the fixed delay.
Overall, the tables show that the use of SCATS control resulted in a lower average approach
delay at all four intersection locations. Once again, the greatest improvement occurred at the
two lowest volume locations. The improvement was an average of 14.6% at the intersection
of Pontiac Trail and Eleven Mile Road and 12.0% at the intersection of Ten Mile Road and
Reynold Sweet Parkway. The average approach delay decrease at the Ten and Nine Mile

Road intersections with Pontiac Trail were 2.5% and 6.3%, respectively.

The comparative results also illustrate the contrast in delay change which occurred between
the low and high volume approach movements. Under SCATS control, low volume
movements are less likely to experience prolonged delay compared to the heavy volume
through movements. This condition is best illustrated once again by the morning peak hour
approach delay differences at the intersection of Pontiac Trail and Nine Mile Road. Table
4.4f showed the average daily traffic volumes and modeled approach delay information for
the week of May 6th through the 10th. During the hour between the 7:00am and 8:00am the
total intersection delay were fairly similar for the optimized fixed time and SCATS control
conditions. However, the individual critical approach movement delays that precipitated
these similar total delay values were very different. This condition is illustrated in Table

5.4b.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.500. _ mm 0.0». we 2. N.- 63 $0.? NNN 0 Nu $0.2. m.~m 0.0m N: 3. m 0_ 5.2 $234. 02:0..—
.-...m.v_. ..nv 0mm $02. 0.0». :3 0.02. 0.0m 0.3 .023. 5 mm ..mn (0.3. in 0.0». ESPEF 05.5833
«in mdm _. mm 5... v. r. M: 0 0m $0.0. r z 00m ”3. _ _. m R 0 _m .3. o. _ mm v.3 0052...: 0:22:51”—
: cm.- cSm ch» 500 0.2 me.— 80.~ c 2 m 2 E m? :1 0.2 $0 0 «.0 v0 50.0 :3 0:95:33.
”.0an a: 0.2 30.3, NE v.2 $0.0m S: a Z .bmdx fixm EN 3%:- _.~ 0.0. :3.th 0:95:38
«.0 w. m .m 0 «m $0.3: 0.0. 0.3“ .300? m. E 05 35 um. m,mm 0.5m “:0 m.»- ~,m E: :3 0.52502
3". 00_ 0% n.m~ $0 3.. 0.: H mm ..N S. of «so gem: #2 _m »R mm. a: 3. 0022:: .zaizzfi
EEC-.0 :nE SEE-u :2: 33:0 :50 6.050 5.5 2:50 :2:
umEfiU mk<Um ..0uu_E:mO.. 005:0 mk<Um 00.55000: 00:55 mk<Um .093530; 20:30 m._.<Um ..03=:::O. mac—EU 3.40m ..0u~_::._n_0..
538: 303 .500 :02:&< E33; 33.. Egon 52:0,? E350 l?u$>a_on 55:004. 59.5.: 395. gun 5310-01. E350 vifluwci: suiccca. E2552 59:0,?
min...< 3:5 E 006.0520. £000.. .E000.r._ E0000.E500> E: 00; .anofl

 

 

 

 

 

 

 

 

uuaho>< ::..C 2:...» .1. t... 19:; .3 :33
0.5M «.02 052 0:: :2... 23:00 00 5:03:25
23.: .5031»: .3»: 2.50 53 5.5

 

 

94

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. t a . ~ 1 L 4 .- avd 033—.
$0”. 0.2 EX .9:- ndv 0.3 $06. fiwm 0.0m $92- 0.5. 0.3. «VON. 0.2 NA: umncufix 03.5.”—
wmdm. 60». 5.; RWNN. 00v n00 $20. m.0~ 0.0m E0.- m._m 5.0m $0.2- 3.: 0.: :SH 0.: 05.9.75»
8:. 0.«.m c 2 wow 0: w, Nm «.3 $0 0. m mm 0 S 000:” 0 mm 0 mm 00 f. N 0m 0 _m :quE. 0::oemam
m. E. v R 0 mm mi. N m mm 0 0m 5. mm. m «a N. 0m «.05. ~ mm 0 mm «.0 mm. WC _ 2 EB. co..— 0:25.33
«.0 E. 0 3. \- vm 60.. mm 5 S. m. hm. «.0 n. v ow n 9. $03.. v mm n 0» 0R mm. 0 0m 5 0v 00:95. 055933
.00 f. m 0m 0 on $0 0. 0 5‘ N mm. $~ 0m m an m «N Edv- m 3 0 no $0 uh 0 0_ 0 m. :50. :3 0:025:0m
$0.3 ~.0v 0.2 RF 0m ~.: New .5. ~0_ x hm 5: $0.2. 0 mm m mm $0 m2 0 n. 0 m swag; 0:25:53
«Aw. 0.0m Wow «#:3- 000 5.0m $NC 70m 0.0m 00.2. :H Nam .323 5.0 0.0 E3- 03 0502552
50.00 mém 0.3 60.0w Row 3: $0.5: 0. E be. tbmxm 0.3. Non 9300 0.0. mm 00:22.0 0:250:02
8:56 :2: 6:000 :20 35:00 :30 60:30 :20 EEBU SE
2:36 mbbm 03258. 9:56 mbom 32:58.. ”:56 2.5m 05.880. 92:25 whom 02.508. 0:26 whom .8503.
33:00 0.53 >200 000202 E0290 Sui aion 503004... E330 303 0..—on 53:9? E35“— 33. mien 5:803 E330 lAfl 0..—on 55:03. 0:25.52 c.3202
umEu>< >55 E.cc.¢ EES. . E 09750002 E08H0.500fl.> E500 c.0002
omsuo.»< 31.... ESE . 4 0c Acckr

000% 0:2 :90. 0:: :EF 000:0: 0: 9:53:85
2.50 503...;— 30: 2.46m .53 5:8

 

 

 

 

.51: .. . n .I t .. nvfi oEIH

95

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

$0.2. a 2 02 «.0 «A. p 2 «.mm «.0 m m 3 0.: n5 2 m f _ Z $0.0m. _ 0 0 _ umio>< 0:5“.
60.0 53 «.33. v mm new $0.0m :3. 0,: $0.3 0.3. n 0m 30.0 0.0 0.. ES. :3 0:255:52
$90K h v 682. z w hr. 35:. Wm «6. .2. “N. 1m. 0 5 $0.00. a c 0 n :mgifi Enofimam
$0.0m. fix 5:? w c O: 3.0? 0.: En $33. \. a c 2 30.0 00 00 Ea: :5 055833
LE. . 3. i i. o n a a 3;. x c 3. .3 E, I! x c m w. E33. m c c n Eaefifigg
:aE 75:90 5.5 7550 5.7. 3:95 :13 3:90 :20 . .
9:26 mtcm .8253? $56 $5: .3530. 2:25 whom .8530; 9:56 mbbm ..uaéio. 2:26 mtg ..3..E_ao.
E330 3>£unucuncam< E33; . Cut Jumflamamz E880 33.1200 52:9? 23:00 39: “win 59:0qu Evacum 3.4: :zun 53:93 E08962 22552
I .u Enigma- 1 Em 00.x» .Em%,m...t I \ Egg”. .EQMO . 5:0ch .Eacfih j! E: 00; am,— I

 

 

 

33. on: :u... an: b35500 .ooam 2::on E 503235
.236 5:315 3»: 35m 53 Guam

ouio>< ... .C

 

955. £2.55 €02? ..c mug;

 

 

##2.

uni—U

02¢ “.22 ESE 0:: :20. 25:3 .3 50.40225
25% 5.332: 35 $28 .53 £25

3325

t . I 3.“ 03:.

52532 5:29;.

 

 

96

At the Pontiac T rail/Nine Mile Road intersection the primary through volume movements
are northbound and southbound. Table 4.4f shows that the weekly average total of the east
and westbound through volumes are a quarter of the averaged north-south through movement
total. As such, the SOAP optimization strategy allocated a shorter green phase length to the
east-west movements. Table 4.2b in chapter four showed that the fixed time model used a
green phase length of 58 seconds for the north—south through movement and 20 seconds for
the east-west through movements. This difference, approximately 35%, is roughly
proportional to the difference in volume ratios. The resulting average delays were very

different.

As shown in Table 5.4b, the approach delay under fixed time control ranged from a
minimum average of 8.1 seconds for the northbound through movement to 75.1 seconds for
the westbound through movement. The approach delay for the same movements under
SCATS control were 18.] seconds and 53.7 seconds. While a sizeable difference between
the two delay was apparent under both control strategies, the difference in SCATS was only
53% as much as optimized fixed time operation. Table 5.4 shows that the average approach
delay for the six critical approach movements averaged 67.0 seconds per vehicle in optimized
fixed time mode and only 35.6 seconds per vehicle in SCATS mode. From this we see that

the modeled approach delays in SCATS are more equally distributed.

The advantage to the optimized fixed strategy was that it resulted in a reduction to the level
of overall intersection delay. However, the significance of the a more equalized distribution
of delay is substantial. The delay for the morning northbound through movement at Nine
Mile Road and Pontiac Trail approximately doubled in the adaptive control mode. While the
increase sounds substantial, it was only an average 10 seconds per vehicle. During a morning
commute trip of fifteen or twenty minutes, the additional 10 seconds of delay would be
barely perceptible. By contrast the delay savings for the westbound through movement was
almost 21 seconds per vehicle. While this savings appears small, a greater savings would

be realized if the trip were to traverse a series of SCATS adaptive controlled intersections.

97

The same effect, with varying levels of success, can be seen at the other intersections within

South Lyon.

Another less tangible advantage to the SCATS strategy for green time allocation is the
possible reduction of driver frustration on minor cross streets. The study results showed
some minor street approach delays in excess of one minute. Many of these delays were
experienced in moderate flow situations. In an operational situation such delays would be
unacceptable. Delay times in excess of multiple signal cycles could lead to both safety and
operation complications. Safety concerns would be most apparent in left turn situations.
Drivers waiting in lengthy queues would become more likely to accept dangerously small gap
sizes in the opposing traffic stream. Such a condition could lead to angle and rear end
collisions at an intersection. The equalization of degrees of saturation in SCATS could
lessen the possibility of forced left turns because left turning traffic would not be allocated

a short left turn phase.

5.3 Delay Data Comparison Conclusion

In this chapter the delay resulting from SCATS real-time adaptive control and fixed time
traffic control in South Lyon were compared. The two types of delay that were analyzed
were the total intersection delay and the average approach delay. While these two methods
of delay measurement are similar in some respects, they are quite different in the manner in
which they assess delay. It is because of these similarities and differences that an apparent

contradiction exists in the experimental outcome.

The fundamental conclusion of the study was that approach delay decreased in South Lyon
while total intersection delay increased. The overall increase in total delay after the
implementation of SCATS was 1.46 vehicle hours. However, the overall decrease in
approach delay under SCATS control was 1.59 seconds per vehicle. The conflict arises when

it is recognized that total delay of an intersection is the sum of its constituent approach

98

delays. The experimental results showed that while total delay increased, the approach delay
decreased. After careful analysis of the data it becomes apparent that this conflict is not
unreasonable. It is logical based on the approach traffic volumes and the objective functions

of the two control strategies.

The objective of SCATS is to balance the degree of saturation on the approach legs of an
intersection. Its goal is not to minimize total intersection delay; although this often occurs.
It is also possible to increase one while decreasing the other. The data showed this to have
been the case in South Lyon. The data also showed that in many cases both total and
approach delay reduced under SCATS control. During several time intervals at a small
number of locations the increase in delay was substantial enough to raise the overall level of

change in a negative direction.

Table 5.1 shows an overall increase in mean total intersection delay. However, it also
illustrates that total delay decreased during the midnight and morning hours, although
through inference testing these were shown not to be statistically significant. Table 5.2
shows that total intersection delay decreased during 7 of the 16 sample periods. The
percentage comparison table also shows that on an averaged daily basis, total delay increased

at l of the 4 study intersections.

The delay reduction results shown in both of these comparison tables were, however, heavily
influenced by the poor performance of SCATS at the Pontiac Trail/Nine and Ten Mile Road
intersections. At these locations total delay increased substantially when compared to fixed
time operation. The primary reason for the significant increase in delay can be seen in Table
5.4b. At this intersection the heavy northbound and southbound Pontiac Trail through traffic
experienced an substantial increase in approach delay under SCATS control. It is also
apparent that the lower volume traffic movements on Nine Mile Road experienced an
improvement in delay conditions. However, the actual number of vehicles on Nine Mile

Road was very small in comparison to the Pontiac Trail volume. Therefore, the sum of the

99

delay and volume resulted in a significant decrease in overall performance for the

intersection.

The apparent conflict of a total delay increase with an approach delay decrease is further
clarified with an understanding of the comparison procedure. The approach delays were
calculated based on the number of vehicles using the approach during the analysis period.
No consideration was given to contrasting levels of demand at an intersection. For example,
Table 4.4f shows that Nine Mile Road services 323 through vehicles while Pontiac Trail
services 1,520 through vehicles during the same evening peak hour. In the comparative
analyses no additional consideration was given to the Pontiac Trail approaches based on
traffic demand. Approach delay values were compared on an equal terms rather than on a
relative basis. In this way, the impact of SCATS on the approach delay could be measured

and compared on a direct basis.

Chapter 6.

Summary and Conclusion

The purpose of this study was to document the change in traffic delay in the South Lyon,
Michigan traffic network resulting from the installation of a SCATS traffic control system.
One of the primary goals was to demonstrate how computer modeling and analysis
techniques could be used to model and evaluate the effect of real-time adaptive traffic
control. The other primary goal was to determine whether the introduction of SCATS control
would result in a change to the traffic delay characteristics of the system. Both of these goals
have been achieved. The study has also demonstrated some of the limitations in the existing
delay models and the need for the development of more sophisticated modeling techniques

to more effectively assess the merits of real-time adaptive traffic signal control.

6.1 Summary of the Research Study
The study involved a four step approach to evaluate the effect of real time adaptive control
using traffic simulation models. First, a literature review was conducted to determine the

existing level of knowledge in the field of adaptive signal control system modeling and

100

101

evaluation. The review of existing literature on subjects related to traffic modeling and the
use of real time adaptive traffic signal control demonstrated that a substantial amount of
work had been done to investigate the development of traffic models as well as the
operational effect of adaptive traffic control. Past field studies showed that systems like
SCATS and SCOOT demonstrated significantly improved performance over fixed time
systems. The benefits were realized in traffic flow parameters like travel time, stopped delay,
and the number of stops. These comparative evaluation studies were conducted using strictly
empirical field observation and numeric data analysis. No computer models were available
to evaluate these real-time adaptive traffic signal systems. This lack of effective tools
demonstrates the need for the development of traffic modeling methods which could be used

to more effectively evaluate existing and proposed traffic signal systems .

From the literature it appeared that the deficiency of traffic modeling methods and the
shortage of experimental data comparing adaptive and fixed time control systems was due,
in part, to a lack of appropriate tools to affect a two-way movement of computer data
required for simulation of real-time adaptive control strategies. To overcome this deficiency
this study used a approach that took advantage of output data from an operational SCATS
system. From this data an analytical computer modeling routine capable of predicting delay

parameters for specified operating conditions was developed.

The predictive delay model was developed using a combination of the Webster and
TRANSYT delay equation to assess the delay conditions after SCATS was operational in
South Lyon. The added dimension of this new model was its ability to read a SCATS output
file and compute the resulting approach delay on a cycle-by-cycle basis. Separate models
were coded to represent delay conditions at four South Lyon intersections during four
different time periods of the day. The modeled delay output data was also compared to field
data recorded in South Lyon after the activation of the SCATS control system. The
comparison of modeled and field data showed that the “after” model represented a reasonably

close match to the field conditions.

102

Next, a “before” model was constructed to simulate traffic flow conditions in South Lyon
prior to the introduction of SCATS control. To make an assessment of the true benefit of the
real time adaptive aspects of SCATS, the original South Lyon fixed time signal plans were
adjusted to minimize intersection delay. Identical cycle-by—cycle traffic volumes were used
for the analysis of both “before” and “after" conditions. The computer generated values of
total intersection delay and approach delay were then used to make a comparison of delay

under the two different control strategies.

Various statistical comparisons of the “before” and “after” data sets highlighted the changes
which resulted from the implementation of SCATS in South Lyon. These comparisons
showed that adaptive control resulted in an improvement to the performance measures under
certain conditions. The comparison showed that the SCATS controlled signals resulted in
a 1.56 second per vehicle decrease in the average value of approach delay for the 24 critical
approaches in the South Lyon road network. The most significant improvement was shown
to occur at the Ten Mile Road Reynold Sweet Parkway intersection. The comparative
analysis also showed that SCATS was out performed by fixed time control in several
instances. The comparison showed a 1.46 vehicle hour increase between the “before” and
“after” values of total intersection delay. However, this difference was not statistically
significant. The total intersection delay change was not statistically significant during all of

the analysis time periods.

6.2 Conclusions

There are two main conclusions which can be drawn from this research study. First, it can
be concluded that existing traffic delay models can be used to successfully model delay
conditions at intersections controlled by real time adaptive traffic signals. Second, it was
concluded that under certain demand conditions the implementation of SCATS control in

South Lyon can result in improvements to traffic delay at the network intersections.

103

The level of success demonstrated in this study are encouraging. However, the next logical
step is to extend this knowledge to other areas and to different systems. The application of
this new traffic model or the realization of significant delay reductions compared to fixed
time signal control may not always be possible. The following paragraphs address some of
the factors which relate to these two issues. This section will also summarize some of the
various benefits and costs associated with the FAST-TRAC system and SCATS in particular

which have been learned during the completion of this study.

‘6 ' ”

The most obvious and direct conclusion which can be drawn from this study is that the use
of SCATS resulted in an improvement to the modeled delay in South Lyon. However, the
improvement was not universal. The amount of improvement varied and it was more
apparent during some traffic volume situations than others. The location and the degree to
which improvements were realized was also relative to the other time periods and

intersection locations within the data set.

Three main conclusions were drawn from the various analyses conducted in the previous
chapter. The first was that the average total intersection delay, on a system wide basis, was
increased. This is not surprising since the objective of SCATS is to equalize saturation flows
rather than to minimize total intersection delay. This increase, as indicated by the statistical
tests, was not universal. The conclusion of an overall increase in total delay is somewhat
misleading since one of the four intersections experienced a reduction in delay. However,
the significant increase in delay at the Pontiac Trail/Nine Mile and Ten Mile Road

intersections lowered the overall system performance.

The second primary conclusion was that SCATS more equally distributes average approach
delays to the various approach movements. This is consistent with the SCATS control
objective in which the approach degrees of saturation are equalized. The third and final

primary conclusion was that SCATS appeared to be more effective at reducing delay during

104

low volume periods compared to high volume periods. This was disappointing because a
reduction of delay during high volume periods is more critical to the overall efficiency of a
traffic signal network and is considerably more difficult to obtain through adjustments in

signal operation.

The primary questions of “if” and “by how much” SCATS improved delay in South Lyon
have been addressed in this study. However, the question of whether or not similar outcomes
could be expected in other networks is difficult to answer. The traffic volumes within the
South Lyon road system are relatively low compared to those found in the other more
congested suburbs in Oakland County. The results of the study indicate that SCATS
improved total intersection delay most significantly during lower flow periods. Thus, in
areas where the traffic volumes approach or operate above capacity, real time adjustment to
signal timings may not necessarily enhance the operation of the network to measurable
degrees. The delay experienced in these situations is not the result of poor signal timings,
rather, it results from a lack of capacity provided by the approach geometry or the network
road segments. Flow improvement and delay reduction could be more economically realized

through the implementation of approach geometry and roadway capacity improvements.

The contribution of SCATS can not be discounted during all high demand periods. In high
volume situations where heavy traffic volumes are moving primarily in one direction, a real
time adaptive control system has some definite advantages. Among other features, systems
like SCATS allow the implementation of coordinated progression on the major arterial
street. The major street green phases are interrupted only when minor street traffic is
present. This has been used in traffic networks in Australia and Toronto, Canada with

measurable success.

The questions of applicability for the procedure developed in the study are not as difficult
to answer. The procedure used to complete this study was developed specifically for use on

SCATS output data files and the experiments and analysis conducted were completed after

105

the SCATS system was operational. The primary difficulty of employing this procedure at
another location or another traffic control system is that the “after” data file containing both
traffic volume and signal data must be available to conduct the study. Predictive simulations,
in which a variety of traffic conditions are tested prior to field installations, are not possible
with this approach. To conduct a predictive study, the SCATS control algorithms for
determining phase splits, cycle length determination, and coordinated operation must be
known. Due to the proprietary nature of the SCATS control software the algorithms can only

be approximated and not used directly at this time.

The delay calculation models within the data processing routine are universal and can be
applied to nearly any intersection or system of intersections. The Webster/TRANSYT model
does contain some inherent weakness that could be improved through additional work. The
most obvious limitation of the equation is that it does not allow for the adjustment of vehicle
arrival patterns. The equation assumes that vehicle arrivals are always uniform. As a result
it can be difficult to accurately assess the delay reduction benefit that SCATS can bring about

through coordinated signal progression.

6‘ ' ’9

The general conclusion of “system improvement” afforded by real-time adaptive signal
control systems must also be contrasted with the significant costs which are involved in the
implementation, operation, and maintenance of such systems. Recent studies have shown
that the implementation the FAST-TRAC system architecture in various locations within
Oakland County has resulted in measurable benefits to the motorists driving within the
SCATS networks. It has also brought the expenditure of over $70 million in combined
public and private resources into the road system of Oakland County. However, given the
limited availability of public resources dedicated to transportation enhancement projects, the
question of whether the benefits gained from the FAST-TRAC project are worth the price

must also be addressed.

106

The cost of FACT-TRAC can not be measured solely in terms of the difference in installation
cost between it and a fixed time system. Other more indirect and difficult to measure costs
must also be evaluated. For example, the traffic safety aspects of the system should also be
factored in the cost-benefit equation. Recent studies of the Troy FAST-TRAC system
showed that the frequency and rate of traffic accidents increased after the implementation of
SCATS, although the severity of injuries decreased.“ The benefits derived from SCATS
should also be compared to the cost for the engineering, design, operation, maintenance,

research, and training required to keep the FAST-TRAC system operating.

The benefits and costs of FAST-TRAC systems can be debated. However, it is difficult to
dispute the need to take advantage of advanced technologies to deal with the problems of
urban congestion. Real-time adaptive signal control allows signal to respond to constantly
changing conditions in a predetermined manner. Since urban traffic flow is a variable
process, flow conditions can change quickly based on the occurrence of accidents,
fluctuations in volume, and travel patterns. In the past, traffic engineers have been able to
meet the challenges of arterial street mobility reasonably well with optimization tools which
were available. In the future, it will be necessary to take advantage of advancements in the
fields of traffic engineering, computer science, and electrical engineering to better utilize the

existing transportation resources.

Projects such as the Oakland County FAST-TRAC program may be expensive and
complicated. However, they are a necessary and integral part of the development of
transportation engineering knowledge in the United States. The successes and failure or
costs and benefits of this and future advanced traffic systems can and should be debated.
However, there should be no debating the need for the development of new technologies and

creative solutions to maintain and increase the state of traffic mobility.

107

6.3 Future Research Needs

This study has taken another step toward the use of more comprehensive and sophisticated
simulation techniques to model adaptive signal control systems. While useful by itself, this
study has also identified a number of needs and questions relating to the modeling of larger
and more complex systems and the comparative assessment of alternative traffic control

strategies; including real-time adaptive, actuated, and fixed time.

i 3 l 5 l l 5' l . S
The most obvious area of need to forward the knowledge into the analysis of adaptive control
systems is currently under development. The TRAP-NETSIM simulation system is the most
widely used network simulation package. Unfortunately, it does not yet permit the
simulation of adaptive control systems. The Federal Highway Administration has recently
awarded contracts to incorporate a communication architecture into NETSIM that permits
the simulation of adaptive traffic control. The goal of this research will be the creation of
a modeling system capable of relaying traffic information to an external traffic control
routine on a real time basis. The control routine could be modified to permit the use of any
signal control algorithm, including SCATS. All currently available NETSIM output statistics
and MOE information including delay, travel time, and stops could be collected for a
proposed system. In contrast to the procedure used in this study, the proposed NETSIM
system would also be able to model the effect of varying driver behavior, lane blockages,
transit systems, on-street parking and pedestrian movements. Unfortunately, the operational

version of this system remains years away.

The simulation of various types of traffic control is helpful in a number of respects. On the
application level it allows traffic professionals the luxury of comparing and evaluating
benefits and costs of alternative systems without the time and expense of field trials. In
theoretical applications it can be used for the experimentation, testing, and refinement of
existing and proposed control strategies. Future research studies could take advantage of the

procedure detailed in this study to compare different strategies like SCATS, SCOOT, and

108

OPAC against fixed time. Such comparisons can be used to determine the merits of each and
the ways in which one system may be better suited to accomplish certain delay reduction

objectives than another.

W

The results of this study could also be used to make improvements to the existing SCATS
control algorithms. The study showed that under SCATS the intersection delay parameters
were increased up to 1.6 vehicle hours during certain time intervals when compared to fixed
time. Further refinements of the cycle, phasing, and coordination selection algorithms,
coupled with this testing procedure could be used to analyze methods to increase signal

efficiency.

Another logical extension of using the real-time adaptive signal control modeling techniques
is their incorporation into combined models of adaptive signal control with real-time incident
detection and route guidance structures. Truly effective traffic responsive traffic systems
need to include the ability to adapt to normal fluctuations in travel demand conditions as well
as special event situations. Systems with the ability to detect an incident, then implement
signal control changes and route guidance in response to the incident would be very useful.
The use of advanced simulation modeling techniques accompanied by the incident producing
capability of NETSIM or route guidance algorithms could be used to test proposed ITS

systems prior to full scale field trials.

WW

Future research also needs to address the economic impacts of technically sophisticated
systems like FAST-TRAC. FAST-TRAC has been shown to increase the efficiency of
certain performance measures. Future research will need to include assessments of the
economics of these benefits through comprehensive cost/benefit studies. This study and
others have made steps in this direction. However, more factors need to be understood

before the engineering community encourages large-scale and wide-spread implementation

109

of these systems.

One final area of future research concerning the use of adaptive signal control strategies
would be to assess the safety impacts of their use. Information gained from empirical
observation and law enforcement personnel during the completion of this study has shown
that systems like SCATS require increased driver awareness. SCATS, as explained earlier,
has the ability to shorten, lengthen, add, remove, and rearrange signal phases to better
accommodate the traffic flow through a single or group of intersections. Drivers who are
familiar with only fixed signal timings are required to go through a “learning curve” to
acquaint themselves with the new strategy. Undoubtably, this can lead to safety problems.
Research to determine the length of time or amount of advanced warning and education that
is required for drivers to become aware of the operation of these signals and their varying

operation would also be important.

APPENDIX A
DELAY CALCULATION PROGRAM FOR THE
INTERSECTION OF PONTIAC TRAIL AND TEN MILE ROAD

0063000OOOOOOOOOOOOOOOOOOOOO

lll

******************************************************************

South Lyon SCATS Signal Delay Evaluation Project

P. Brian Wolshon
Department of Civil and Environmental Engineering
Michigan State University Fall, 1996

******************************************************************

Pontiac Trail/T en Mile Road Intersection

RCOC Traffic Operations Center SCATS Output File
May 6th - May 19th 1996

SCATS SIgnal Timing Control

******************************************************************

Variable List:
Gfaz = "green" phase for a particular movement per cycle (sec)
vol = recorded traffic volume per cycle (veh)
C = signal cycle length (sec)
q = critical approach volume (veh)
h = discharge headway (sec)
8 = saturation flow (veh/lane/hour of green time)
lam = proportion of green time available to a movement
g = effective green time
ltime = "lost" time -> 3.0 sec
X = degree of saturation -> q/lamda*s
V = hourly approach volume (veh/hr)

INTEGER hr,min,Gfaz(8),vol(8),faztot,g(8),ltime,icount,hri

REAL q(8),h,s,X(8),Dl(8),z(8),Bn(8),Bd(8),term(8),D2D3(8),delay(8)
REAL avedelay(8),totdelay(8),avecyc,totvol(8),bigdelay(8),intdelay
REAL lam(8),term1(8),term2(8),V(8),aaa(8),bbb(8),C,totcyc

REAL Gsat(8),Gunsat(8),adjvol(8)

CHARACTER DUMMY*25

OPEN (1 ,FILE='pt 10-aft.txt',STATUS='OLD')

OPEN (2,FILE='pt10-aft.out',STATUS='NEW')

intdelay=0
hri=100
totcyc=0
faztot = 0
h = 2.2

s = 1600
ltime = 3

112

READ(1,101)hr, min

C Main Loop

DO 60 ii: 1,21000

C When the hour has been completed print out hourly volume,
C average movement delay and average cycle length for that hour.

C

80

11

IF (hr .GE. hri) THEN
WRITE (2,201) hri-l
DO 80 ia = 1,8
IF (totvol(ia) .NE. 0) THEN
avedelay(ia)=bigdelay(ia)/totvoI(ia)
ELSE
avedelay(ia)=0
ENDIF
WRITE (2,202) ia,totvol(ia),avedelay(ia)
totintdelay = totintdelay + (totvol(ia) * avedelay(ia))
totvol(ia) = 0
bigdelay(ia) = 0
CONTINUE
avecyc=totcyc/(icount*8)
totcyc = 0
intdelay = totintdelay / 3600
WRITE (2,204) intdelay
WRITE (2,203) avecyc
intdelay=0
totintdelay = 0
icount=0
hri=hr+l
icount = icount + 1
READ (1,102) dummy
DO 11 ia = 1,8
READ (1,103) Gfaz(ia), vol(ia)
faztot = faztot + Gfaz(ia)
CONTINUE
C = faztot/2
If (C .eq. 0) then
Write (2,888)
C=120
endif

 

000000

000000000

113

Use the SCATS outfile data to determine delays with the use of
the first term of Webster's delay equation and the TRAN SYT delay
model.

First determine the critical hourly volume. Reducing the left turn
volume by 1 "sneaker" vehicle.

D021 ia= 1,8

IF ((ia .EQ. 3) .OR. (ia .EQ. 4)) THEN
adjvol(ia) = vol(ia) - 1
IF (adjvol(ia) .LE. 0) THEN
q(ia) = 0
ELSE
q(ia) = adjvol(ia)*(3600/C)
ENDIF
ELSEIF ((ia .EQ. 7) .OR. (ia .EQ. 8)) THEN
adjvol(ia) = vol(ia) - 1
IF (adjvol(ia) .LE. 0) THEN
q(ia) = 0
ELSE
q(ia) = adjvol(ia)*(3600/C)
ENDIF
ELSE
q(ia) = vol(ia)*(3600/C)
ENDIF

The volume of left turn traffic that can be accomodated during a
particular cycle is a function of the permissive period and the
protected period. Therefore, the effective green time for the

left turn phase must be computed from both time intervals.

The Gsat variable is the time segment of the permissive left period
which is saturated by the opposing through traffic. The Gunsat
variable is the remaining time of permissive turn period in which
the left turning traffic is unopposed.
IF ((ia .EQ. 3) .OR. (ia .EQ. 7)) THEN
Gsat(ia) = Gfaz(ia-l) * X(ia—1)
Gunsat(ia) = Gfaz(ia—l) - Gsat(ia)
IF (Gunsat(ia) .LE. 0) THEN
g(ia) = GFaz(ia) -ltime
ELSE
g(ia): Gfaz(ia)+Gunsat(ia)-ltime
ENDIF

114

ELSEIF ((ia .EQ. 4) .OR. (ia .EQ. 8)) THEN
Gsat(ia) = Gfaz(ia-3) * X(ia-3)
Gunsat(ia) = Gfaz(ia-3) - Gsat(ia)
IF (Gunsat(ia) .LE. 0) THEN

g(ia) = GFaz(ia) -ltime
ELSE

g(ia): Gfaz(ia)+Gunsat(ia)-ltime
ENDIF

ELSE
g(ia) = Gfaz(ia)-ltime

ENDIF

if (g(ia) .Eq. 0) then
g(ia) = 7
Write(2,999)

endif

lam(ia) = g(ia)/C

X(ia)=q(ia)/(lam(ia)*s)

21 CONTINUE

C Calculate lst term of the Webster delay model (D l) for each
C critical movement.
DO 31 ia = 1,8
D1(ia)=(C*(1-lam(ia))**2)/(2*(1-lam(ia)*X(ia)))

#

CONTINUE
Calculate 2nd and 3rd terms of the Webster delay model (D2 & D3)
using the coverted TRANSYT delay model single term for each
critical movement.
DO 41 ia = 1,8
[F (q(ia) .GT. 0) THEN
z(ia) = ((2*X(ia))/q(ia))*(60/C)
Bn(ia) = (2*(1-X(ia)))+(X(ia)*z(ia))
Bd(ia) = (4*z(ia))-z(ia)**2
aaa(ia) = (Bn(ia)/Bd(ia))**2
bbb(ia) = (X(ia)**2)/Bd(ia)
term 1 (ia) = SQRT(aaa(ia)+bbb(ia))
term2(ia) = (Bn(ia)/Bd(ia))
term(ia) = term1(ia)-term2(ia)
D2D3(ia) = (term(ia)*3600)/q(ia)
delay(ia) = D1(ia)+D2D3(ia)
totvol(ia) = totvol(ia) + vol(ia)
totdelay(ia) = vol(ia) * delay(ia)
bigdelay(ia)=bigdelay(ia)+totdelay(ia)

000W

115

ELSE
totvol(ia) = totvol(ia) + vol(ia)
delay(ia) = 0

ENDIF
totcyc = totcyc + C
faztot = 0

41 CONTINUE

C If the hour is not complete continue to process the statistics

C on a cycle by cycle basis
C

ELSE
hri=hr+1
icount = icount + 1
READ (1,102) dummy
DO 10 ia = 1,8
READ ( 1,103) Gfaz(ia), vol(ia)
faztot = faztot + Gfaz(ia)
10 CONTINUE
C = faztot/2
If (C .eq. 0) then
Write (2,888)
C=120
endif

Use the SCATS outfile data to determine delays with the use of
the first term of Webster's delay equation and the TRAN SYT delay
model.

First determine the critical hourly volume. Reducing the left turn
volume by 1 "sneaker" vehicle.
DO 20 ia = 1,8
IF ((ia .EQ. 3) .OR. (ia .EQ. 4)) THEN
adjvol(ia) = vol(ia) - 1
IF (adjvol(ia) .LE. 0) THEN
q(ia) = 0
ELSE
q(ia) = adjvol(ia)*(3600/C)
ENDIF

000000

000000000

116

ELSEIF ((ia .EQ. 7) .OR. (ia .EQ. 8)) THEN
adjvol(ia) = vol(ia) - 1
IF (adjvol(ia) .LE. 0) THEN
q(ia) = 0
ELSE
q(ia) = adjvol(ia)*(3600/C)
ENDIF
ELSE
q(ia) = vol(ia)*(3600/C)
ENDIF
The volume of left turn traffic that can be accomodated during a
particular cycle is a function of the permissive period and the
protected period. Therefore, the effective green time for the
left turn phase must be computed from both time intervals.

The Gsat variable is the time segment of the permissive left period
which is saturated by the opposing through traffic. The Gunsat
variable is the remaining time of permissive turn period in which
the left turning traffic is unopposed.
IF ((ia .EQ. 3) .OR. (ia .EQ. 7)) THEN
Gsat(ia) = Gfaz(ia-l) * X(ia-l)
Gunsat(ia) = Gfaz(ia-l) - Gsat(ia)
IF (Gunsat(ia) .LE. 0) THEN
g(ia) = GFaz(ia) -ltime
ELSE
g(ia): Gfaz(ia)+Gunsat(ia)-ltime
ENDIF
ELSEIF ((ia .EQ. 4) .OR. (ia .EQ. 8)) THEN
Gsat(ia) = Gfaz(ia-3) * X(ia-3)
Gunsat(ia) = Gfaz(ia-3) - Gsat(ia)
IF (Gunsat(ia) .LE. 0) THEN
g(ia) = GFaz(ia) -ltime
ELSE
g(ia): Gfaz(ia)+Gunsat(ia)—ltime
ENDIF
ELSE
g(ia) = Gfaz(ia)-ltime
ENDIF
if (g(ia) .Eq. 0) then
g(ia) = 7
Write(2,999)
ENDIF

117

lam(ia) = g(ia)/C
X(ia)=q(ia)/(lam(ia)*s)
20 CONTINUE
C Calculate lst term of the Webster delay model (D1) for each
C critical movement.
DO 30 ia = 1,8
D1(ia)=(C*(1-1am(ia))**2)/(2*(l-lam(ia)*X(ia)))
30 CONTINUE
C Calculate 2nd and 3rd terms of the Webster delay model (D2 & D3)
C using the coverted TRANSYT delay model single term for each
C critical movement.
DO 40 ia = 1,8
IF (q(ia) .GT. 0) THEN
z(ia) = ((2*X(ia))/q(ia))*(60/C)
Bn(ia) = (2*(1-X(ia)))+(X(ia)*z(ia))
Bd(ia) = (4*z(ia))-z(ia)**2
aaa(ia) = (Bn(ia)/Bd(ia))**2
bbb(ia) = (X(ia)**2)/Bd(ia)
term 1(ia) = SQRT(aaa(ia)+bbb(ia))
term2(ia) = (Bn(ia)/Bd(ia))
term(ia) = term 1 (ia)-term2(ia)
D2D3(ia) = (term(ia)*3600)/q(ia)
delay(ia) = D1(ia)+D2D3(ia)
totvol(ia) = totvol(ia) + vol(ia)
totdelay(ia) = vol(ia)*delay(ia)
bigdelay(ia)=bigdelay(ia)+totdelay(ia)
ELSE
totvol(ia) 2' totvol(ia) + vol(ia)
delay(ia) = 0
ENDIF
totcyc = totcyc + C
faztot = 0
40 CONTINUE
ENDIF
READ (1,101) hr, min
60 CONTINUE

1 01
102
103
201
202
203
204

118

FORMAT (14X,I3, 1X,12)
FORMAT (A25)
FORMAT (22X,12,5X,I2)
FORMAT ('Period begins at ',I2,':00')
FORMAT ('Movement #',Il,' hrly vol. =',F7.3,' ave. delay =',F7 .3)
FORMAT (' ave. cycle length =', F7.3,'sec')
FORMAT ('Intersection total delay=', F7.3,' veh-hrs')
CLOSE (1)
CLOSE (2)
STOP
END

APPENDIX B
DELAY CALCULATION PROGRAM OUTPUT FOR THE
INTERSECTION OF PONTIAC TRAIL AND TEN MILE ROAD

 

120

Period begins at 0:00
Movement #1 hrly vol. = 62.000 ave. delay = 9.091
Movement #2 hrly vol. = 67.000 ave. delay = 7.693
Movement #3 hrly vol. 4.000 ave. delay = 7.767
Movement #4 hrly vol. 9.000 ave. delay = 8.656
Movement #5 hrly vol. 13.000 ave. delay = 26.661
Movement #6 hrly vol. 10.000 ave. delay = 24.947
Movement #7 hrly vol. 2.000 ave. delay = 17.002
Movement #8 hrly vol. .000 ave. delay = .000
Intersection total delay: .505 veh-hrs

ave. cycle length = 52.9708ec
Period begins at 1:00
Movement #1 hrly vol. = 20.000 ave. delay = 1.444
Movement #2 hrly vol. 41.000 ave. delay = 1.341
Movement #3 hrly vol. .000 ave. delay = .000
Movement #4 hrly vol. .000 ave. delay = .000
Movement #5 hrly vol. 2.000 ave. delay = 21.489
Movement #6 hrly vol. .000 ave. delay = .000
Movement #7 hrly vol. .000 ave. delay = .000
Movement #8 hrly vol. = .000 ave. delay 2 .000
Intersection total delay: .035 veh-hrs

ave. cycle length = 50.1 18sec
Period begins at 2:00
Movement #1 hrly vol. = 28.000 ave. delay = 11.382
Movement #2 hrly vol. = 10.000 ave. delay = .187
Movement #3 hrly vol. = .000 ave. delay = .000
Movement #4 hrly vol. 2.000 ave. delay = 19.136
Movement #5 hrly vol. .000 ave. delay = .000
Movement #6 hrly vol. .000 ave. delay = .000
Movement #7 hrly vol. .000 ave. delay = .000
Movement #8 hrly vol. .000 ave. delay = .000
Intersection total delay: .100 veh-hrs

ave. cycle length = 50.9853ec
Period begins at 3:00
Movement #1 hrly vol. 2 18.000 ave. delay = 2.170
Movement #2 hrly vol. - 22.000 ave. delay = 2.278
Movement #3 hrly vol. .000 ave. delay = .000
Movement #4 hrly vol. .000 ave. delay = .000
Movement #5 hrly vol. 2.000 ave. delay = 23.749
Movement #6 hrly vol. 4.000 ave. delay = 22.839
Movement #7 hrly vol. .000 ave. delay = .000
Movement #8 hrly vol. .000 ave. delay = .000
Intersection total delay: .063 veh-hrs

121

ave. cycle length = 53.206sec
Period begins at 4:00
Movement #1 hrly vol. = 17.000 ave. delay = .166
Movement #2 hrly vol. 20.000 ave. delay = 2.312
Movement #3 hrly vol. .000 ave. delay 2 .000
Movement #4 hrly vol. .000 ave. delay = .000
Movement #5 hrly vol. 7.000 ave. delay = 23.921
Movement #6 hrly vol. 5.000 ave. delay = 25.968
Movement #7 hrly vol. .000 ave. delay = .000
Movement #8 hrly vol. .000 ave. delay = .000
Intersection total delay: .096 veh-hrs

ave. cycle length = 53.597sec
Period begins at 5:00
Movement #1 hrly vol. =1 15.000 ave. delay = 15.315
Movement #2 hrly vol. 88.000 ave. delay = 12.940
Movement #3 hrly vol. .000 ave. delay = .000
Movement #4 hrly vol. .000 ave. delay = .000
Movement #5 hrly vol. 26.000 ave. delay = 27.359
Movement #6 hrly vol. 93.000 ave. delay = 25.413
Movement #7 hrly vol. 5.000 ave. delay = 15.010
Movement #8 hrly vol. 6.000 ave. delay = 15.554
Intersection total delay: 1.706 veh-hrs

ave. cycle length = 62.186sec
Period begins at 6:00
Movement #1 hrly vol. =210.000 ave. delay = 20.428
Movement #2 hrly vol. 183.000 ave. delay = 19.559
Movement #3 hrly vol. 6.000 ave. delay = 16.015
Movement #4 hrly vol. 2.000 ave. delay = 21.873
Movement #5 hrly vol. 62.000 ave. delay = 23.186
Movement #6 hrly vol. =231.000 ave. delay = 33.126
Movement #7 hrly vol. 6.000 ave. delay = 21.226
Movement #8 hrly vol. 16.000 ave. delay = 16.828
Intersection total delay: 4.860 veh-hrs

ave. cycle length = 79.889sec
Period begins at 7:00
Movement #1 hrly vol. =447.000 ave. delay = 43.215
Movement #2 hrly vol. =401.000 ave. delay = 38.613
Movement #3 hrly vol. = 17.000 ave. delay = 24.958
Movement #4 hrly vol. = 16.000 ave. delay = 37.307
Movement #5 hrly vol. =194.000 ave. delay = 31.360
Movement #6 hrly vol. =336.000 ave. delay = 52.589
Movement #7 hrly vol. = 50.000 ave. delay = 34.943
Movement #8 hrly vol. = 67.000 ave. delay = 30.712

122

Intersection total delay: 17.606 veh-hrs
ave. cycle length = 85.810sec

Period begins at 8:00

Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

=353.000 ave. delay = 29.904
=324.000 ave. delay = 32.991
= 16.000 ave. delay = 30.338
= 42.000 ave. delay = 23.932
=193.000 ave. delay = 37.659
=265.000 ave. delay = 34.587
= 22.000 ave. delay = 25.188
= 22.000 ave. delay = 26.563

Intersection total delay: 1 1.197 veh-hrs
ave. cycle length = 80.9553ec

Period begins at 9:00

Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

=369.000 ave. delay = 29.272
=389.000 ave. delay = 32.122
= 17.000 ave. delay = 65.652
= 14.000 ave. delay = 23.884
=186.000 ave. delay = 29.813
=228.000 ave. delay = 31.621
= 23.000 ave. delay = 24.237
= 27.000 ave. delay = 25.658

Intersection total delay: 10.765 veh-hrs
ave. cycle length = 80.733sec

Period begins at 10:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

=358.000 ave. delay = 28.185
=373.000 ave. delay = 26.923
= 25.000 ave. delay = 24.269
= 42.000 ave. delay = 31.461

=188.000 ave. delay = 27.691
=199.000 ave. delay = 30.695
= 21.000 ave. delay = 30.972
= 26.000 ave. delay = 20.148

Intersection total delay: 9.597 veh-hrs
ave. cycle length = 80.000sec

Period begins at 11:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.

=419.000 ave. delay = 34.509
=433.000 ave. delay = 34.664
= 26.000 ave. delay = 29.684
= 34.000 ave. delay = 30.437
=218.000 ave. delay = 29.341
=200.000 ave. delay = 29.852
= 35.000 ave. delay = 22.239

Movement #8 hrly vol.

123

= 22.000 ave. delay = 21.156

Intersection total delay: 12.468 veh-hrs
ave. cycle length = 80.0443ec

Period begins at 12:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

2400.000 ave. delay = 29.585
=441.000 ave. delay = 37.316
2 21.000 ave. delay = 53.181
= 34.000 ave. delay = 30.289
=250.000 ave. delay = 33.341
=187.000 ave. delay = 28.290
= 18.000 ave. delay = 23.549
= 28.000 ave. delay = 34.599

Intersection total delay: 12.626 veh-hrs
ave. cycle length = 81.0003ec

Period begins at 13:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.

2372.000 ave. delay = 28.793
=495.000 ave. delay = 40.429
= 37.000 ave. delay = 29.607
= 22.000 ave. delay = 26.577
=246.000 ave. delay = 40.649

Movement #6 hrly vol
Movement #7 hrly vol
Movement #8 hrly vol

. =192.000 ave. delay = 28.375
. = 36.000 ave. delay = 23.206
. = 38.000 ave. delay = 32.366

Intersection total delay: 13.866 veh-hrs
ave. cycle length = 80.200sec

Period begins at 14:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

=445.000 ave. delay = 47.543
=524.000 ave. delay = 54.310
= 30.000 ave. delay = 38.908
= 63.000 ave. delay = 37.799
=298.000 ave. delay = 62.220
=211.000 ave. delay = 36.153
= 35.000 ave. delay = 28.293
= 55.000 ave. delay = 55.973

Intersection total delay: 23.167 veh-hrs
ave. cycle length = 85.8815ec

Period begins at 15:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.

=477.000 ave. delay = 43.850
=518.000 ave. delay = 44.285
= 44.000 ave. delay = 35.484
= 53.000 ave. delay = 38.420
2360.000 ave. delay = 64.899
=212.000 ave. delay = 32.064

Movement #7 hrly vol.
Movement #8 hrly vol.

124

= 29.000 ave. delay = 27.762
= 73.000 ave. delay = 54.948

Intersection total delay: 22.897 veh-hrs
ave. cycle length = 89.0503ec

Period begins at 16:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

=462.000 ave. delay = 46.145
=582.000 ave. delay = 68.489
= 37.000 ave. delay = 37.811
= 88.000 ave. delay = 40.612
=378.000 ave. delay = 69.939
=267.000 ave. delay = 42.614
= 56.000 ave. delay = 33.641
= 85.000 ave. delay = 55.856

Intersection total delay: 30.722 veh-hrs
ave. cycle length = 89.293sec

Period begins at 17:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

=463.000 ave. delay = 45.301
=589.000 ave. delay = 78.856
= 44.000 ave. delay = 37.980
= 90.000 ave. delay = 45.328
=445.000 ave. delay = 75.455
=218.000 ave. delay = 33.171
= 37.000 ave. delay = 43.064
= 65.000 ave. delay = 43.181

Intersection total delay: 32.883 veh-hrs
ave. cycle length = 89.625sec

Period begins at 18:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.
Movement #6 hrly vol.
Movement #7 hrly vol.
Movement #8 hrly vol.

=531.000 ave. delay = 53.251
=545.000 ave. delay = 58.467
= 34.000 ave. delay = 36.424
= 87.000 ave. delay = 53.181
=33 1.000 ave. delay = 56.870
=176.000 ave. delay = 29.846
= 32.000 ave. delay = 25.568
= 55.000 ave. delay = 57.128

Intersection total delay: 26.123 veh-hrs
ave. cycle length = 88.9753ec

Period begins at 19:00
Movement #1 hrly vol.
Movement #2 hrly vol.
Movement #3 hrly vol.
Movement #4 hrly vol.
Movement #5 hrly vol.

=421.000 ave. delay = 29.229
=482.000 ave. delay = 42.888
= 16.000 ave. delay = 28.417
= 35.000 ave. delay = 31.973
=270.000 ave. delay = 41.660

125

Movement #6 hrly vol. =164.000 ave. delay = 27.242
Movement #7 hrly vol. = 16.000 ave. delay = 19.648
Movement #8 hrly vol. = 37.000 ave. delay = 31.990
Intersection total delay: 14.379 veh-hrs

ave. cycle length = 82.068sec
Period begins at 20:00
Movement #1 hrly vol. =370.000 ave. delay = 28.586
Movement #2 hrly vol. =425.000 ave. delay = 32.464
Movement #3 hrly vol. = 18.000 ave. delay = 28.846
Movement #4 hrly vol. = 16.000 ave. delay = 34.437
Movement #5 hrly vol. =185.000 ave. delay = 28.492
Movement #6 hrly vol. =115.000 ave. delay = 25.698
Movement #7 hrly vol. = 17.000 ave. delay = 19.991
Movement #8 hrly vol. = 10.000 ave. delay = 25.838
Intersection total delay: 9.519 veh-hrs

ave. cycle length = 80.156sec
Period begins at 21:00
Movement #1 hrly vol. =322.000 ave. delay = 26.303
Movement #2 hrly vol. =316.000 ave. delay = 22.297
Movement #3 hrly vol. = 11.000 ave. delay = 20.424
Movement #4 hrly vol. = 7.000 ave. delay = 29.026
Movement #5 hrly vol. =198.000 ave. delay = 38.874
Movement #6 hrly vol. = 71.000 ave. delay = 23.412
Movement #7 hrly vol. = 16.000 ave. delay = 18.221
Movement #8 hrly vol. = 12.000 ave. delay = 23.243
Intersection total delay: 7.187 veh-hrs

ave. cycle length = 79.733sec
Period begins at 22:00
Movement #1 hrly vol. =186.000 ave. delay = 17.841
Movement #2 hrly vol. =158.000 ave. delay = 16.188
Movement #3 hrly vol. = 9.000 ave. delay = 13.807
Movement #4 hrly vol. = 8.000 ave. delay = 19.803
Movement #5 hrly vol. =100.000 ave. delay = 33.852
Movement #6 hrly vol. = 47.000 ave. delay = 26.175
Movement #7 hrly vol. = 15.000 ave. delay = 17.628
Movement #8 hrly vol. = 10.000 ave. delay = 18.234
Intersection total delay: 3.117 veh-hrs

ave. cycle length = 75.917sec

LIST OF REFERENCES

1)

2)

3)

4)

5)

6)

7)

8)

9)

10)

11)

127

LISIQEREEERENQES

Road Commission for Oakland County, "FAST-TRAC is on the Cutting Edge of
Traffic Management Technology," Beverly Hills, Michigan, 1993.

Lowrie, P. R., "SCATS, Sydney Co-Ordinated Adaptive Traffic System, A Traffic
Responsive Method of Controlling Urban Traffic," Roads and Traffic Authority
Sydney, New South Wales, Australia, 1990.

Lowrie, P. R., "The Sydney Co-Ordinated Adaptive Traffic System - Principles,

Methodology Algorithms," International Conference on Road Traffic Signaling,
pp. 67-70, London, 1982.

Taylor, William C. and Thomas L. Maleck, “Interim Report on Intersection Delay,”
Department of Civil and Environmental Engineering, Michigan State University,
1994.

Carty, Patrick J ., “Lessons Learned from the Oakland County IVHS Project,” IVHS
America International Conference, Washington, DC, 1993.

Hunt, P.B., Robertson, D.I., Bretherton, R.D., R.I. Winton, "SCOOT - A Traffic
Responsive Method of Coordinating Signals," Transport and Road Research
Laboratory, Crownthome, Berkshire, England, 1981.

Curtis, Owen, "Time-Based Coordination of Traffic Signals," Federal Highway
Administration Report No. FHWA/PU85/001, Washington, DC, 1984.

Stillman, LL. and HM. Maynard, ”Development of a Traffic Flow Sensing and
Surveillance System," IEEE Automotive Conference, Detroit, Michigan, 1967.

Chirstensen, A. “Use of a Computer and Vehicle Loop Detectors To Measure Queues
and Delays at Signalized Intersections," Highway Research Record, Number 211,
Washington, DC, 1967.

Genovese, Joseph A., “SCOOT in the USA,” 1994 Compendium of Technical
Papers, Institute of Transportation Engineers, Washington, DC, 1994.

Sims, AG. and K.W. Dobinson, “S.C.A.T. - The Sydney Co-Ordinated Adaptive
Traffic System Philosophy and Benefits,” International Symposium on Traffic
Control Systems, Vol 2B, 1979.

12)

13)

14)

15)

16)

17)

18)

19)

20)

21)

22)

23)

128

Bielefeldt, Christian, "HERMES - High Efficiency Roads with Rerouting Methods
and Traffic Signal Control," Compendium of Technical Papers, Institute of
Transportation Engineers Annual Meeting, The Hague, Netherlands, 1993.

Bretherton, RD, and GT Bowen, "Incident Detection and Traffic Monitoring in
Urban Areas," Proceedings of the DRIVE Conference, Brussels, Belgium, 1991.

Morello, E. and G. Sala, "Automatic Incident Detection in HERMES," Centro Studi
sui Sistemi di Trasporto, Torino, Italy, 1991.

Fambro, Daniel B. and CJ. Messer, "Estimating Delay at Coordinated Signalized
Intersections," Highway Capacity and Level of Service, Rotterdam, The Netherlands,
1991.

Gartner, Nathan H., "Impacts of Smart Signals on Intersection Capacity and Level of
Service," Highway Capacity and Level of Service, Rotterdam, The Netherlands, 1991 .

“Highway Capacity Manual,” Special Report 209, Transportation Research Board,
National Research Council, Washington, DC, 1985

Bartle, Richard, M., Val Skoro, and D.L. Gerlough, “Starting Delay and Time
Spacing of Vehicles Entering Signalized Intersections,” Highway Research Record,
112, Washington, DC, 1956

Kell, James H., "Analyzing Vehicular Delay at Intersections Through Simulation,"
Highway Research Board, Bulletin 356, Washington, DC, 1962.

Kell, J arnes H., "Results of Computer Simulation Studies as Related to Traffic Signal
Operations," 33rd Annual Meeting of the Institute of Traffic Engineers, Toronto,
Ontario, Canada, August, 1963.

Needham, R.G., "A Simulation Program of a Road Network Controlled by Traffic
Responsive Signals," Road Research Laboratory, Crowthome, England, 1970.

United States Department of Transportation, Federal Highway
Administration,“Signal Operations Analysis Package User’s Manual,” Washington,
DC, 1979.

United States Department of Transportation, Federal Highway Administration,
“TRAP-NETSIM User’s Manual,” Washington, DC, 1995.

24)

25)

26)

27)

28)

29)

30)

31)

32)

33)

34)

129

Van Aerde, Micheal, J. Voss, and G. McKinnon, “User’s Guide to the Student
Macintosh Version of the INTEGRATION Simulation Model,” Department of Civil
Engineering, Queens University, Kingston, Canada, 1991.

Van Aerde, Micheal and Sam Yagar, “Dynamic Integrated Freeway/'1‘ raffic Signal
Networks: Problems and Solutions,” Transportation Research Volume 22A, Number
6, 1988.

Hsin, Victor J.K. and Paul Wang, "Modeling Concepts for Intelligent Vehicle
Highway Systems (IV HS) Applications," Proceedings of the 1992 Winter Simulation
Conference, 1992.

Wang, Paul T. R. and William P. Niedringhaus, “Distributed/Parallel Traffic
Simulation for IVHS Applications," Proceedings of the 1993 Winter Simulation
Conference, Washington, DC, 1993.

Codelli, David K., William P., et. a1., "User's Manual for Traffic and Highway Object
for Research Analysis, and Understanding (THOREAU) Volume [1, Revision 1," The
MITRE Corporation, McLean, Virginia, 1993.

Rathi, Ajay K. And Alberto Santiago, "Simulation and Intelligent Vehicle Highway
Systems," US. Department of Energy Project Number DE-AC05-840R21400,
Washington, DC, 1992.

Gruba, John L., Beaubien, Richard, F., "Combining Real Time, Adaptive Traffic
Control and Machine Vision Detection for Advanced Traffic Management Systems,"
Compendium of Technical Papers, Institute of Transportation Engineers Annual
Meeting, The Hague, Netherlands, 1993.

Beaubien, Richard F ., “Advanced Traffic Management Systems for Arterial Streets,
A European Perspective,” Institute of Transportation Engineers International
Conference, 1992.

Michalopoulos, Panos G. et. a1., “Automatic Incident Detection Through Video
Image Processing,” Traffic Engineering and Control, February, 1993.

Rajendra, Kunwar and Robert E. Maki, “The IVHS Strategy in Michigan,” IVHS
America International Conference, Washington, DC, 1993.

Sobolewski, Mike and James L. Wright, “Rural Applications of IVHS in Minnesota,”
IVHS America International Conference, Washington, DC, 1993.

35)

36)

37)

38)

39)

40)

41)

42)

43)

44)

45)

46)

47)

130

Wallace, Charles E. And Andrew K. Kilpatrick, “IVHS Applications for Rural
Highways and Small Towns,” IVHS America International Conference, Washington,
DC, 1993.

Luk, J.Y.K., A.G. Sims and PR. Lowrie, “SCATS - Application and Field
Comparison with a TRANSYT Optimized Fixed Time System,” International
Conference on Road Signaling, London, 1982.

Moore, S.E., A.G. Sims and RR. Lowrie, “Three Traffic Signal Control Philosophies
Applied to an Arterial Traffic System,” Australian Road Research Board Proceedings,
Volume 8, 1976.

Luk, J .Y.K., “Two Traffic-Responsive Area Traffic Control Methods: SCAT and
SCOOT,” Traffic Engineering and Control, 1984.

Kelman, W. Les, et. al., “The Metropolitan Toronto SCOOT Demonstration Project -

A Preliminary Evaluation Report,” Transportation Research Board, 73rd Annual
Meeting, Washington, DC, 1994

Grubba, John et. a1., “Selecting Real-Time Adaptive Traffic Control for Oakland
County, Michigan,” Transportation Research Board, 72nd Annual Meeting,

Washington, DC, 1993.

Bretherton, RD. and 6.1. Rai, “The Use of SCOOT in Low Flow Conditions,” Traffic
Engineering and Control, December, 1982.

May, Adolf, “Traffic Flow Fundamentals,” Prentice Hall, Englewood Cliffs, New
Jersey, 1990.

Taylor, Micheal and W. Young, ”Traffic Analysis, New Technology and Solutions,”
Hargreen Publishing Company, North Melbourne, Victoria, Australia, 1988.

McShane, William R. and Roger P. Roess, “Traffic Engineering,” Prentice Hall,
Englewood Cliffs, New Jersey, 1990.

Bhattacharyya, Gouri K. And Richard A. Johnson, “Statistical Concepts and
Methods,” John Wiley & Sons, New York, New York, 1977.

Mendenhall, William et. a1., “Mathematical Statistics and Applications, Fourth
Edition,” PWS-KENT Publishing Company, Boston, Massachusetts, 1990.

Taylor, WC. and Ahmed Abdel-Rachman, Working Paper, Michigan State
University Department of Civil Engineering, East Lansing, Michigan 1996.

 

48)

49)

50)

131

Webster, RV. and BM. Cobbe, “Traffic Signals,” Road Research Laboratory
Technical Paper No. 56, London, England, 1966.

Robertson, DI. and P. Grower, “Use of TRAN SYT Version 6,” Transport and Road
Research Laboratory Supplemental Report 255, Crowthome, England, 1977.

Taylor, W.C., J. Wu and T. Hein, “Deployment of Intelligent Transportation Systems
in Michigan - Impact on Traffic Accidents,” Proceedings, World Congress on
Intelligent Transportation Systems, Orlando, Florida, October, 1996, p. 184

HICHIGQN STRTE UNIV. LIBRRRIES
lllll Hill 111 llllllllllll11111 ll llllllllllllllll
31293015706942