.q..\.

- -nH.‘.|

.
m
.
.
.
»
u
.
.

 

Gv...
k\.\lh.:u
: ; § - .

gr. kuxmwmpi.‘

 

 

masts

../

HIAGANST ATEU

IIIIIIIIIIIIIII I II

 

 

 

 

 

 

                                           

WIIIIIIIIIIII

301555 0506

This is to certify that the

dissertation entitled

TESTING TRAFFIC CONTROLS STRATEGIES FOR

INCIDENT CONGESTION MANAGEMENT OF A SURFACE STREET SYSTEM
presented by

Sorawit Narupiti

has been accepted towards fulfillment
of the requirements for

Ph.D. Civil Engr.

degree in

 

 

(0%” C 7/347
William C. Taylor

Major professor

 

Date May 28, 1996

 

MSU i: an Affirmatiw Action/Equal Opportunity Institution 0-12771

4.. _._._—~» fl"“

"‘— '

A“ v

 

LIBRARY
Mlchlgan State
Unlverslty

 

 

 

PLACE N RETURN BOX to roman this checkout from your "cord.
TO AVOID FINES Mum on or baton date duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/EM Opportunity Inotltmlon
m1

 

TESTING TRAFFIC CONTROL STRATEGIES FOR
INCIDENT CONGESTION MANAGEMENT
OF A SURFACE STREET SYSTEM

By

Sorawit Narupiti

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

1 996

ABSTRACT

TESTING TRAFFIC CONTROL STRATEGIES FOR INCIDENT CONGESTION
MANAGEMENT OF A SURFACE STREET SYSTEM

By

Sorawit Narupiti

Selecting the most appropriate traffic control strategy for incident congestion
management can have a major impact on the extent and duration of the resulting
congestion. This research investigated the effectivenesses of several control strategies on
various incident conditions. The selected control strategies representing possible ITS
technologies included traffic metering (ATMS), traffic diversion (ATIS), and traffic
diversion with signal timing modification (ATIS/A'I'MS). The analysis was conducted on a
hypothetical dense grid surface street network. Mid-block incidents of various durations
were tested. The results indicated that the ATIS/ATMS based solution can reduce
congestion duration up to 27 minutes, with a saving of 261 vehicle-hours of delay. A
sensitivity analysis was performed to obtain the effectivenesses of various control
strategies under different demand levels. Several alternate diversion plans and control

variables for initiating signal modification were also tested.

This work is dedicated to my parents, Mrs. Sumitda and Mr. Vicham, my brother and
sister, Mr. Noppadol and Ms. Wipawee, and all of my friends,

who all dream to witness my accomplishment.

ACKNOWLEDGMENTS

I am indebted to my advisor, Dr. William Taylor, for his invaluable advice,
guidance, and friendship. He is a true example of a great educator and has added meaning
to the word “supportive”, both official and personal matters.

I would like to express my gratitude to Dr. Richard Lyles, Dr. Thomas Maleck,
and Dr. Rene Hinojosa, my committee members, for their professional suggestions and for
helping me develop a deep insight into the topics.

I would like to show my appreciation to my parents who has always given me
perpetual and unconditional support. Their continuous encouragement helped me achieve
my academic goal..

I would like to extend my gratitude to all my colleagues at Michigan State
University for their unreserved friendships and support. My work experience with them is
very useful for my future.

I would like to thank the Department of Civil and Environmental Engineering, the
International Road Federation, and Michigan State University who contribute financial
support and opportunity to conduct work and research.

I am pleased to acknowledge all my beloved friends at Michigan State University.
Although I cannot refer to them all, I would never forget good experience I shared with

them.

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................................... vii
LIST OF FIGURES .................................................................................................... x
CHAPTER 1
INTRODUCTION ...................................................................................................... l
1.1 Description of the problem ........................................................................ 1
1.2 Statement of the problem .......................................................................... 7
1.3 Objective and scope of the research .......................................................... 7
1.4 Research approach .................................................................................... 8
CHAPTER 2
LITERATURE REVIEW ........................................................................................... 10
2.1 Theoretical traffic signal control strategies ................................................ 11
2.2 Signal control strategies and response to incidents in existing urban traffic
control systems ......................................................................................... 14
2.3 Route guidance and access control strategies ............................................ 17
2.4 Effectiveness of control strategies for incidents ......................................... 18
2.5 Summary .................................................................................................. 23
CHAPTER 3
DESIGN OF STUDY ................................................................................................. 24
3.1 NETSIM simulation model ....................................................................... 24
3.2 Hypothetical network formulation ............................................................. 25
3.3 Traffic demand and initial traffic signal setting ........................................... 27
3.4 Incidents ................................................................................................... 29
3.5 Technological scenarios and control strategies .......................................... 30
3.6 Measures of effectiveness (MOEs) ............................................................ 33
3.7 Study plan ................................................................................................ 36
3.8 Summary .................................................................................................. 36
CHAPTER 4
SIMULATION ANALYSIS AND RESULTS ............................................................ 39
4.1 Effect of type and duration of incident ...................................................... 39

4.2 Effect of different control strategies .......................................................... 51

4.3 Summary .................................................................................................. 64

CHAPTER 5

SENSITIVITY ANALYSIS ........................................................................................ 66
5.1 Sensitivity to traffic demand change .......................................................... 66
5.2 Sensitivity to diversion control alternatives ............................................... 74
5.3 Sensitivity to signal control variable alternatives ........................................ 81
5.4 Sensitivity to signal control coverage ........................................................ 84
5.5 Summary .................................................................................................. 89

CHAPTER 6

CONCLUSION AND RECOMMENDATION ........................................................... 91
6.1 Conclusion ............................................................................................... 91
6.2 Recommendation ...................................................................................... 94

BIBLIOGRAPHY ....................................................................................................... 98

LIST OF TABLES

Table 2.1 Delay and journey time for routes affected by the incident ........................... 19
Table 2.2 Percentage average trip time relative to base case: arterial incident .............. 22
Table 3.1 Traffic demand ............................................................................................ 27
Table 4.1 Total travel time in the network under incidents .......................................... 40

Table 4.2 Start and end time of intersection blockage: various incident types and
durations ..................................................................................................... 41
Table 4.3 Total travel time of traffic passing the incident ............................................ 50

Table 4.4 Total travel time for four control strategy scenarios : lO—minute both-lane

closure incident .......................................................................................... 52
Table 4.5 Total travel time of particular traffic groups: both-lane closure incident ....... 63
Table 5.1 New traffic demand used in the sensitivity analysis ....................................... 67

Table 5.2 Total travel time for three volume scenarios: IlO-minute both-lane closure
incident ...................................................................................................... 68

Table 5.3 Start and end time of intersection blockage: various demand levels, 10-minute
both-lane closure incident ............................................................................ 70

Table 5.4 Total travel time of alternate diversion plans: 10-minute both-lane closure

incident ...................................................................................................... 76

Table 5.5 Total travel time of traffic streams in alternate diversion plans: lO-minute
both-lane closure incident ........................................................................... 82
Table 5.6 Total travel time of alternate signal control variables: lO—minute both-lane
closure incident .......................................................................................... 83
Table 5.7 Total travel time of traffic groups of alternate signal control variables:
lO-minute both-lane closure incident .......................................................... 86
Table 5.8 Total travel time of alternate signal control coverage: lO—minute both-lane
closure incident .......................................................................................... 87

Table 5.9 Total travel time of traffic groups of alternate signal control coverage:

lO-minute both-lane closure incident ........................................................... 89
Table 6.1 Impact of alternative control strategies on various incidents ........................ 93
Table 6.2 Effectiveness of the control strategies under different demand levels ........... 95
Table 6.3 Comparison of alternate diversion plans ...................................................... 96

Table 6.4 Comparison of alternate signal control policies for signal

modification ................................................................................................ 96

LIST OF FIGURES

Figure 3.1 Street network in this study ....................................................................... 26
Figure 3.2 Intersection geometry and volume ............................................................. 28
Figure 3.3 Study plan ................................................................................................. 37
Figure 4.] Statistics from NETSIM for determining congestion period ....................... 44
Figure 4.2 Queue length obtained every four seconds ................................................. 44
Figure 4.3 Congestion duration: no control change ..................................................... 45
Figure 4.4 Affected links and duration of congestion: no control change ..................... 48
Figure 4.5 Congestion duration: different control strategies ........................................ 57
Figure 4.6 Affected links and duration of congestion: different control strategies ........ 59
Figure 5.1 Congestion duration: different demand levels ............................................. 71

Figure 5.2 Comparison of reduction in congestion duration in different demand levels 74

Figure 5.3 Congestion duration: alternate diversion plans ........................................... 77
Figure 5.4 Affected links and duration of congestion: alternate diversion plans ........... 78
Figure 5.5 Congestion duration: alternate signal control variables ............................... 85
Figure 5.6 Congestion duration: alternate signal coverage .......................................... 88

Chapter 1

INTRODUCTION

1.1 Description of the problem

Selecting the most appropriate traffic control strategy for incident congestion
management can have a major impact on the amount and duration of the resulting
congestion. With the implementation of the Intelligent Transportation System (ITS)
technology, several advanced traffic control options are now available. However, this
topic has not been thoroughly studied. In this study, several ITS control strategies are

tested to determine their effectivenesses under various incident and control conditions.

1.1.1 Incidents and their effect on traffic

Traffic incident is a term used for a random, unplanned event which affects traffic
operations. It can be a temporary reduction in roadway capacity caused by accidents,
roadway maintenance, or construction activities, or a temporary increase in traffic demand
as commonly occurs immediately before and after special events. Incidents, generally
referred to as nonrecurrent events, can be divided into three basic categories based on
their uncertainty of occun'ence (Holmes and Leonard, 1993):

a. Normal and generally accepted (although not necessarily desirable)-such as

on-street parking. This type of incident is usually tolerated by drivers as

2

being part of normal traffic conditions on the network.

b. Expected and programmed--includes roadwork and maintenance activities.
Their occurrence is foreseen. and planned, but unexpected to motorists.

c. Unexpected-such as vehicle breakdowns and accidents. Neither drivers nor
the traffic agency is prepared for this type of the incident because the
occurrence is unpredictable.

Incidents generally result in traffic congestion. The magnitude and duration of the
congestion is difficult to predict because, in most cases, neither the location, the time, nor
the severity of an incident is known beforehand. Limited information and knowledge on
incidents can result in non-optimal choices, such as an improper route choice and ill-timed
signal settings.

Each incident has a unique impact on traffic operations in a network. Beaubien
(1994) analyzed the effect of incidents on delay for various incident types. Similar work
was also conducted in Texas (Dudek, 1976) and in London, England (Holmes and
Leonard, 1993). Delay generally ranged from 15-30 vehicle-minutes due to a vehicle
disablement to several vehicle-hours due to major accidents with lane blockages.

Incident management has received considerable attention since incidents are one of
the most pressing traffic problems in urban areas. Lindley (1987, 1989) reported that
congestion due to incidents accounted for more than 61 percent of all urban congestion in
1984, and is expected to cause more than 70 percent of the total congestion by the year

2005 (National Cooperative Highway Research Program, 1991).

3

1.1.2 Measures to alleviate congestion due to incidents

Incidents commonly create two types of congestion: primary and secondary.
Primary congestion is caused by traffic queuing at a bottleneck. Secondary congestion
arises from the blockage of other intersections by primary congestion (Longley, 1968).
The goal of incident management is to restore roadway capacity as quickly as possible, so
as to limit primary congestion, and at the same time avoid or reduce secondary
congestion.

Incident management shares the same traffic treatment philosophy as recurrent
congestion management, although different traffic control strategies may be required.
Recurrent congestion occurs routinely at specific locations and times of day. An example
is the peak hour traffic jam. Both short-term and long-term solutions are considered as
methods to alleviate recurrent congestion. Considering the techniques for reducing
recurrent traffic congestion, which can be divided into three categories (Rathi and
Leibennan, 1989):

a. Increasing the capacity of the road system--through construction of
additional facilities or a physical improvement to provide additional capacity.

b. Reducing traffic demand--through behavioral changes or travel demand
management or traffic restraints.

c. Maximizing the use of available capacity-through traffic engineering
practices aimed at minimizing the capacity reducing factors (e.g. through
traffic regulation) or more efficient use of existing capacity (e.g. signal

control improvement).

4

The first method is not appropriate for incident management. The unpredictable nature of
incidents makes it infeasible to maintain excess capacity at all elements of the network.
Behavioral changes are also not applicable to incident management because of the time
frame involved (long-term solution). However, traffic restrictions in congested areas,
such as metering and traffic diversion, may be applied to incident management. The
efficient use of existing capacity through traffic management is the only response
applicable to nonrecurrent congestion. Therefore, measures to alleviate incident
congestion focus on maximizing the use of available capacity.
Actions to reduce incident congestion can specifically be categorized into three

groups (Van Vuren and Leonard, 1994):

a. Incident control.

b. Behavioral control or demand control.

c. Network control.
Incident control deals with the initial cause of congestion. Techniques include reduction in
incident duration (i.e., incident removal) and local traffic management to reduce the
impact of the incident at the scene. Behavioral control includes the provision of
information to motorists so that they can adjust their travel pattern to avoid the congested
links through route diversion, a departure time change, a mode change, or even trip
cancellation. The route change is the only response to the incident that can be applied to
motorists currently on the streets. Network control is the efficient use of available
network characteristics, including throughput and storage capacity. This method employs

measures such as signal control alterations.

5

1.1.3 Strategies for incident management

Most traffic control measures attempt to improve network efficiency, i.e., reduce
delay for all vehicles using the system. With this goal in mind, a criterion by which
success will be measured and a technique to achieve the goal must be developed for a
specific application. Such a technique is called a strategy.

In the course of incident management, traffic control strategies can be classified
into two main categories:

a. Signal modification.
b. Traffic diversion.

Signal modification strategies are generally implemented at traffic control centers.
Examples of signal modification strategies are longer or shorter cycle time, phase changes
to reflect current demand, changes in the green splits and offsets to maintain equal queues
for conflicting movements, traffic metering to avoid blockage, and reverse progression.
The signal modification strategies require a signal control system that is responsive to
changing traffic demand. Signal control strategies have been extensively reviewed in the
literature (Wright and Huddart, 1989; Montgomery and Quinn, 1992; Quinn, 1992;
OECD, 1981; McShane and Pignataro, 1978).

The strategies for driver information are pre-trip information and route guidance.
These strategies attempt to provide knowledge of traffic conditions to drivers so they can
make route choices which minimize the effect and extent of the incident. Communication

through in-vehicle devices or changeable message signs is required to give the drivers

6

incident and routing information. These control policies were surveyed in Van Aerde and
Rakha (1989) and Van Aerde and Plum (1988).

The effectiveness of each traffic control strategy depends on demand, the network,
and control characteristics. Signal alteration alone is applicable only when the demand
does not exceed the total network reduced capacity after an incident occurs. Traffic
metering requires some links to be designated for queue storage. The effectiveness of
traffic routing depends on the availability of alternate routes and their level of congestion.
Therefore, there is no single most appropriate control strategy that can be applied in all

situations.

1.1.4 Intelligent Transportation Systems (ITS) and incident congestion

In the past decade, the ability to improve performance of the transportation
systems has been made possible through the advancement of computer, communication,
and information technologies. This effort has been pursued in many parts of the world. In
the United States, Intelligent Transportation Systems (ITS), formerly Intelligent Vehicle-
‘iighway Systems (IVHS), is a single phrase describing the use of technologies to
accomplish transportation goals in many functional areas. The European community calls
the program Advanced Transport Telematics (ATI') and Road Transport Informatics
(RTI) (Castling, 1994).

ITS offers potential incident congestion reduction through Advanced Traffic
Management System (ATMS) and Advanwd Traveler Information System (ATIS ). The
primary feature of ATMS is the provision of real-time (dynamic) control to respond to

changing traffic conditions. ATIS provides travelers with information required to ensure

7

that their journey is as efficient and safe as possible. The combined effect of ATMS and
ATIS has been studied by Rakha et. a1. (1989) and Sarakki and Kerr (1994). There are
several demonstration programs employing these two technologies currently underway in
the United States, Europe, and Japan. Thus, the scenario of vehicles equipped with ATIS
devices and circulating in a network with an ATMS traffic control is in the near future.
However, control strategies in. the ITS environment need further investigation to
determine the effectiveness of each strategy under various situations. The benefits of ITS

deployment cannot be fully realized if inappropriate control strategies are applied.

1.2 Statement of the problem

Traffic control strategies aimed at reducing the consequences of incidents have not
been thoroughly developed and tested. The research conducted for this study is original in
that different control strategies developed for a variety of incident situations in an ITS
environment were tested. The effectiveness of controls under each individual ITS element
was obtained. The joint effect of ATMS and ATIS control strategies was also examined
as benefits gained through their interaction may be lost if one of the strategies is

employed.

1.3 Objective and scope of the research
This research addresses the development and testing of traffic control strategies
designed to reduce the consequences of an incident. The scope of the research includes:
a. Identification of measures of effectiveness (MOEs) to indicate the impact of

an incident on an urban street network;

8

b. Investigation of the impacts of incidents on these MOEs under various
conditions;

c. Development of routing and signal control strategies to cope with the
incidents; and

(1. Determination of the limits of the effectiveness of these strategies by varying

degrees of demand, incident severity, and incident duration.

1.4 Research approach

This research was based on traffic simulation since this permits the analysis and
comparison of different control strategies on the same road network and incident.
NETSIM (NETwork SIMulation) was selected for this study because it could be modified
to replicate control and drivers characteristics within the ITS environment. NETSIM is a
microscopic interval-based simulation model of urban traffic on a surface street network.
The model was first developed in the 19705 and has periodically been enhanced. NETSIM
version 5.0, which was used in this study, includes many advanced features on traffic
signal, driving behavior, and turning movement descriptions and can provide data on the
MOEs suitable for the analysis (Federal Highway Administration, 1995).

The research was based on a hypothetical dense grid network with demand
characteristics representative of traffic conditions in the City of Troy, Michigan. When an
incident was introduced into the network, the evolution and dissipation of congestion were
studied. Congestion resulting from an incident in a network without ITS was used as a

base case, with traffic performance analyzed for various types of incidents. For the base

9

case, the signal timing was held constant and the impacts of the incidents on specified
MOEs were determined.

Several traffic signal control and route diversion strategies were developed for
each traffic situation. These strategies were then tested with the simulated network to
obtain performance measures in various incident characteristics. Data for the MOEs were
collected for each control scheme and the results were compared and discussed.

The effectiveness of these control strategies under different demand condition was

then determined. Moreover, variations of control strategies were evaluated.

Chapter 2

LITERATURE REVIEW

Traffic control strategies have been developed and incorporated in urban traffic
signal control systems since the 19605. Along with the advancement in traffic control
systems and technologies, many traffic control strategies have been developed to provide
efficiency, safety, and a reduction in fuel consumption. Most of the signal control policies
in the past were developed for recurrent traffic conditions, both for peak and off-peak
traffic. As the nonrecurrent traffic congestion problem on urban streets increases, recent
interest has shifted to traffic control strategies under incident conditions. Traffic control
strategies for incident conditions were not possible until responsive traffic signal controls
and communication technologies existed, due to the requirement for a prompt response to
the incident.

Studies on control strategies for incidents started in the 19705 with the evaluation
of responsive signal control systems and driver information systems on unexpected
situations. Nonetheless, the first major work was conducted by Hunt and Holland in
1985, where an attempt was made to determine the effect of an incident in a network

controlled by the SCOOT (Spilts, Cycle length, Offset Optimization Technique) traffic

10

11

control system. Since then increased attention has been given to the ability of each
modern traffic control systems to respond with special control strategies for incidents.
Most traffic control strategies for incidents are derived from those for recurrent
congestion conditions (both undersaturated and saturated conditions). Because of this,
the following section reviewed the control strategies originally developed for recurrent
traffic conditions. The transferability to an incident situation was then discussed, along
with the research on effectiveness of numerous control strategies on incident-caused

congestion.

2.1 Theoretical traffic signal control strategies

The first generation of signal control strategies were based on off-line calculations
for a fixed time signal control system. At an individual intersection, Webster and Cobbe
(1958) suggest that a signal split strategy should equalize the degree of saturation (DS) of
all critical approaches to approximately yield the minimum intersection delay. This control
strategy can handle varying demand in a day by having separate settings for different time
periods. This technique became general practice for most fixed time signal controls, in
both undersaturated and saturated traffic conditions.

For arterial and network considerations, Little (1966) introduced an off-line
mathematical technique to maximize the bandwidth. The principle was to maximize the
number of vehicles able to successfully encounter green signals when traveling along a
street. Over the years, several variations of this approach were developed. NCI-IRP
Report 73 (1969) evaluated several of these offset strategies. Off-line control techniques

investigated in this report were Yardeni’s time-space design, Little’s maximal bandwidth,

1")

and delay/difference—of-offset. Three responsive control strategies, namely basic queue
control, cycle and offset selection, and mixed cycle mode, were also evaluated in this
report. The results indicate that the cycle and offset selection method and
delay/difference—of-offsets techniques rank the best for off—peak periods, whereas the
mixed cycle mode and basic queue control were the best for peak periods.

Perhaps the most comprehensive and most widely—applied control strategy for
fixed time control setting is based on a computer optimization method. TRANSYT
(Traffic Network Study Tool) (Robertson, 1968) is an off-line program utilizing modified
Webster’s method to calculate green splits, and a hill-climbing optimization technique to
determine offset and cycle length which minimizes a performance index. The logic for the
offset calculation is similar to the delay/difference-of-offset method evaluated in NCHRP
Report 73. Similar programs to TRANSYT are SIGOP (Signal Optimization program)
and PASSER II (Progression Analysis and Signal System Evaluation Routine), which have
slightly different calculation procedures.

The control strategies obtained from off—line calculations are effective for average
traffic conditions, but are not responsive to changing traffic patterns. When a traffic
pattern changes, the solutions from these programs are no longer optimal. Thus, the
TRANSYT method is suitable only for recurrent traffic conditions. These methods are
difficult to apply to incident conditions, where the incident alters the capacity and demand
pattern.

Several traffic-responsive signal control strategies were developed for an individual

intersection furnished with traffic detectors. Gazis and Potts (1964) developed a

13

technique for time-dependent signal setting under varying demand. They used queue
length as an input to minimize total aggregate intersection delay. The technique is also
called “bang-bang” because the green time is set at a predetermined maximum value for
the queued approach, and at a minimum value in other directions. When a queue in the
first approach is cleared, the setting is reversed. This signal setting does not, in general,
minimize the period during which one approach is congested. d’Ans and Gazis (1976)
furthered this control method by means of linear programming. Church and Revelle
(1978) formulated similar control strategies with consideration of maximum waiting time
and queue length. When the maximum queue length was used as a control objective, they
found that the solution tended to balance the queue lengths on the most saturated
approaches of each signal phase, and the signal frequently switched between phases.
Michalopoulos and Stephanopolous (1977) reported that the queue constraint was
effective when the demand increases to the limiting value. The optimal control strategy at
saturation is simply the balance of input-output to maintain constant queue length.

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

Many control strategies have been developed for oversaturated traffic conditions at

an isolated intersection. Gordon (1969) suggested that the control objective should be to

l4

maintain a constant ratio among the respective storage spaces. Longley (1968) attempted
to balance queue lengths on all approaches. NCHRP Report 194 showed that, although
the Longley control logic yielded lower delay than the off-line calculation, the queue-
actuated control resulted in lower delay when the degree of saturation was above 0.5.
The report stated that the objective of signal control should be to avoid spillback and to
provide equitable service. The report also gave some tactical control strategies to ease

queue blockage at an intersection.

2.2 Signal control strategies and response to incidents in existing urban traffic
control systems

Under a fixed time control system, the TRANSYT method of delay minimization is
a popular method to determine signal timings (Woods, 1993). However, such timings
cannot respond to unexpected incidents. Because the signal setting is fixed for a time-of-
day period, there is no special control strategy for incidents.

Similar control strategies were used in the UTCS (Urban Traffic Control System)
first generation control system as in the fixed time control systems, but signal plans could
be changed every 15 minutes. A signal timing plan suitable for current traffic conditions is
selected from a set of pre-calculated plan. The UTCS first generation system includes
split adjustment for the Critical Intersection Control (CIC). The signal split is altered if
oversaturation is detected, and queue control comes into place at this critical intersection.

The UTCS second generation system computes a new signal plan instead of
“looking up” a selection in the plan library. Signal split and offset are then adjusted at the

critical intersection. The UTCS third generation system controls intersections

15

independently, with fully adaptive split, offset, and cycle length determination. The signals
can be changed every 3-6 minutes. Because the second and third generation systems are
on-line and more responsive to changing traffic patterns, they can respond to incident-
related traffic but the effectiveness of these systems to an incident has not been evaluated
(Kay, Allen, and Bruggeman, 1975).

In the Japanese UTMS (Urban Traffic Management System), five control
strategies are selected based on the level of traffic demand. Three strategies correspond
to levels before network saturation. Signals are optimized similar to TRANSYT in
undersaturated conditions. In oversaturated conditions, the objective is switched to
prevent blockages and to give priority to main roads by restraining access from side
streets. In the case of incidents, the system keeps priority routes clear during particularly
severe congestion (Woods, 1993).

The SCATS (Sydney Co—ordinated Adaptive Traffic System) of Australia also has
different control strategies for various demand levels. Signal splits and offsets are selected
from embedded plans calculated by an off-line program such as TRANSYT, while the
cycle length is calculated every cycle. However, the system also makes use of some
tactical controls at each intersection (Lowrie, 1982). Response to incidents is primarily
activated by traffic operators. When the detectors are covered by traffic for certain
periods of time, an alarm is signaled to the traffic operator who sets the traffic control.
SCATS tactical logic itself can also respond to incident-related congestion. The logic is
the same as the normal recurrent traffic operation. At each intersection, tactical control

strategies include:

16

a. Signal split selection from a library according to degree of saturation;

b. Green time gap-out;

c. Green time early cut-off due to inefficient use of green time; and

(1. Phase skip if no demand is placed in the previous cycle length.
At a “strategic” level of control, offset and cycle time are selected in response to the
current traffic situation based on the plan selection process. However, the plans are not
typically developed for incident situations. In principle, when an incident occurs on a link,
there is a reduction in traffic at the downstream intersection. The reduction of the green
time for that direction will be given to other phases by means of any of the four strategies.
At an intersection upstream from the incident, if blockage exists and reduces the flow, the
green split is reduced by the split plan change and the early cut-off. SCATS does not have
logic to prevent intersection blockage.

The British SCOOT system is simply an on-line version of TRANSYT. The
control strategies are to minimize delays and stops at all intersection in the network in all
ranges of demand. Signal splits change incrementally based on current demand obtained
from detectors every four seconds. Offsets and cycle times are adjusted every few
minutes. The response to incidents relies on the adaptive logic of the system. In the case
of an incident where the traffic demand change is so rapid that SCOOT cannot adapt to it,
two methods can be imposed. The first method is that SCOOT is suspended and falls
back to manual operation. The other method is to invoke a special plan run. SCOOT also

has a gating feature that limits flow into a particular sensitive area.

17

2.3 Route guidance and access control strategies

Van Aerde and Rakha (1989) studied the potential of two route guidance
strategies, namely user-optimum and system optimum assignment. The user optimum
strategy follows Wardrop’s first principle that a driver is assumed to choose a route which
will minimize their journey time through the network, and all other drivers equipped with
the route guidance instrument make their choice by the same criterion. The system
optimum strategy is that each driver with the route guidance equipment is directed to the
path which minimizes the overall travel time to all drivers. They concluded that system
optimized routings are complex and impractical for any but the most trivial networks.

The difficulty of integrating route guidance/signal control was initially reported by
Allsop and Charlesworth (1977 ). They suggested that the route guidance optimal strategy
was dependent on signal control. In normal traffic operation, an optimal routing can be
found for each signal setting. When the signal timings are changed, the optimal routing
changes. This effect is also reported by Charlesworth (1978) and Maher and Akcelik
(1975).

Although optimal route guidance control is difficult to obtain, modern systems
utilize advanced communication systems to obtain real-time traffic data from individual
vehicles. The real-time travel time data are then used to determine the optimal path. The
systems are based on the user optimal strategy. Each equipped vehicle is provided with
information on the shortest time to its destination, based on current travel time on each

road section. The systems include EURO-SCOUT, CACS, SOCRETES (Castling, 1994).

18

Various techniques can also be used to disseminate traffic information to
motorists. These include radio broadcasting, police control, variable message signs, and
pre-trip planning.

The concept of access control can be applied to incidents if a route guidance
system is available. In an incident situation, access to the area is controlled by means of
traffic diversion. Traffic is diverted to other routes to avoid the obstruction. For a wider
area, where there are numerous alternative routes, access control can be done through the
use of route guidance to re-route traffic to non-congested routes.

Traffic metering (gating) is another technique to control access. The traffic is
screened at an entry point to limit the number of vehicles allowed into the congested area.
Many countries have applied this technique for peak-hour congestion (May and Westland,
1979). Rathi ( 1991) conducted a comprehensive study on the traffic metering on a grid
network of Manhattan, New York, which the overall travel time reduction in the order of

20 percent was obtained.

2.4 Effectiveness of control strategies for incidents

Hunt and Holland (1985) studied the effect of a SCOOT control strategy and
traffic diversion on the reduction in congestion due to an incident on a hypothetical
network. A key assumption in the study was that the flows were chosen so that during the
road closure there was no oversaturation under fixed time control. This implies that any
increase in delay is due to lack of responsiveness, rather than lack of capacity. In the
“before” case, demand which created a degree of saturation of 42 percent at any

intersection on the major arterial was selected to conform to the assumption. The volume

19

level was selected to resemble off-peak traffic in Coventry, England. Each junction had a
simple two-phase signal. The control strategies in this study were modifications of green
split and offsets as a result of the SCOOT embedded adaptive control logic and traffic
diversion. All traffic was obliged to turn away from the incident at an upstream
intersection onto one parallel arterial then return to the original route at an intersection
downstream of the incident. The results from the simulation are shown in Table 2.1.
The results indicate a large incident delay reduction for adaptive signal control
over the fixed time system. However, the assumptions on the treatment of right turns on
the diverted route (vehicles are operated on the left side of the road in England) are
questionable. The assumption was made that there was no opposing traffic for the right
turners and thus there was no waiting time for this traffic. The parallel arterials, which
received the diverted traffic, did not connect to other links and thus the need to retain

progression on the adjacent arterials with the outside network was not considered.

Table 2.1 Delay and journey time for routes affected by the incident
(Hunt and Holland, 1985)

 

Delay Journey time

 

Scenario Route / Control » Seconds Increase over Seconds Increase over
per veh normal before per veh normal before

 

case (%) C859 (%)
Before Direct route 28.4 - 84.4 -
incident Diverted route 61.0 115 145.0 72
After Fixed time control 105.6 271 189.6 125
incident SCOOT responsive 44.4 56 128.4 52

(diverted control
route)

 

 

 

 

 

 

 

 

20

Roberg (1995) investigated several dynamic strategies for controlling and
dispersing traffic jams on an idealized one-way grid network. The strategies include the
application of restricted movements on a number of critical junctions in the network. An
incident was placed on a link in the network to study traffic congestion evolution and
migration. Control strategies in this study were a combination of turn bans, ahead bans,
and gating. Traffic was not allowed to turn into selected links to avoid gridlock. An
ahead ban was imposed around the envelope of the congested area to reduce input into the
critical sections of road. The ahead ban forced traffic to reroute away from the jam. In
some experiments, instead of forcing vehicles to turn away from the congested region via
ahead ban, vehicles were queued on the approaches to the jam without being diverted.
This technique is called gating or external traffic metering.

Roberg and Abbess (1994) reported that gating generally yields higher delay than
rerouting because traffic, which is stored on the approach links, may create some
secondary gridlock.

The results of the study indicate the operational domain, in which the chosen
strategies have successfully eliminated the jam, with respect to different levels of demand,
turning percentage, and lane utilization at the intersection. The increase in delay from the
normal situation was presented.

This study did not have a routing procedure for diverted traffic and there was no
guarantee that vehicles returned to their original routes. The study also neglected to
account for the longer journey time due to alternate routes and thus the indicated

reduction in delay was exaggerated.

21

Van Vuren and Leonard ( 1994) studied four scenarios of different control
strategies on incident congestion; signal optimization, gating, diversion, and demand
restraint. They tested these techniques on an irregular shaped network of Kingston-upon-
Thames. A congestion index, defined as the ratio of the travel time to the free-flow travel
time, was used as a measure of effectiveness. The signal downstream of the incident was
timed so that all approaches had equal saturation. Two main entry points to the incident
were metered and gated traffic was stored in the peripheral region. Drivers were diverted
to alternate routes when they faced an unexpected queue, either primary or secondary,
based on travel time along the links. This means the drivers had no prior knowledge of
congestion on the diverted routes. In this study, traffic was forced to reroute at selected
intersections.

The results show that the equisaturation policy at intersections downstream from
the incident lowers the congestion index, a surrogate measure of delay, by 14 to 22
percent. Benefits from the gating strategy were estimated to be up to 7 percent. The
result suggests that rerouting before reaching the congested area has substantial potential
benefit. This however depends on the incident location, demand pattern, position of traffic
signals in the system, and link storage capacity. The secondary jam due to stored vehicles
was not addressed. It was implied that diversion becomes potentially damaging when the
overall level of recurrent saturation in the network increases.

Rakha et. al. (1989) examined the interactions between route guidance and signal
control strategies on a freeway with a parallel signalized arterial. Traffic demand is

assumed to be 1800 vehicles per hour on the 4-1ane freeway and 150 vehicles per hour on

22

the 2-lane arterial, in the direction incidents are introduced. The spacing between
junctions on the arterial is approximately 2300 meters. Two types of incidents were
investigated, one was a one-lane blockage on the freeway and the other was a blockage of
0.75 of a lane on the arterial. The authors did not consider oversaturation and total
blockage during the analysis. Two signal control strategies were evaluated, namely an off-
line signal optimization by holding the signal setting as it was, and an on-line signal
optimization by adjusting signal timings at periodic intervals. For the on-line signal
optimization strategy the signal splits and the cycle length were optimized as isolated
intersections every three minutes without considering signal offset.

Table 2.2 displays the results from this study for the arterial incident scenarios. In
the no incident situation, the on-line signal optimization reduced the average trip travel
time over the off-line control by four percent. However, signal optimization increases trip

time when a higher percentage of vehicles with route guidance capability is assumed. The

Table 2.2 Percentage average trip time relative to base case: arterial incident
(Rakha et. al., 1988)

 

 

 

Scenario Percentage of route guidance (%)
(Duration of incident) 0 20 4o 60 80 1m
No signal optimization
No incident 100 82 79 79 79 79
5 minutes 102 83 80 79 79 79
10 minutes 105 85 81 80 79 79
15 minutes 109 88 82 81 79 79
On-line optimization
No incident 96 81 79 79 79 79
5 minutes 96 82 82 80 80 80
10 minutes 98 87 82 82 80 79
15 minutes 100 89 83 81 81 79

 

 

 

 

 

 

 

 

 

23

authors gave the reason that the signal optimizer was set to calculate timings based on the
non-incident flow rates rather than the reduced flow rates following the incident. The
conclusion of this study was that the combined effect of this signal optimization strategy
and route guidance was not positive on arterial incidents because the signal control

strategy failed to improve traffic performance.

2.5 Summary

Most traffic control strategies, developed for recurrent traffic conditions and
employed in current control systems, may not be effective in an incident situation.
Incident congestion requires a responsive signal control system that can adapt the signal to
changing traffic patterns during the incident. Moreover, other alternatives, such as traffic
restraints and traffic diversion, can be applied to alleviate the congestion. Limited number
of studies were conducted to evaluate the effectivenesses of these traffic control strategies
for incidents. The primary focus of the previous studies was to estimate the impact of one
control type on incident congestion. Although various signal control policies, as well as
routing procedures were tested, they have not been tested in the same street network.
Furthermore, a number of assumptions were imposed and a detailed analysis of the impact

of these assumptions has not been conducted.

Chapter 3

DESIGN OF STUDY

This study was designed to test different traffic control strategies under a
controlled environment using the NETSIM traffic simulation program. The testing was
based on a hypothetical network with a base demand. The normal traffic operations were
assumed to be similar to a peak traffic period, with the optimal signal timing plan
developed by an off-line calculation. Three types of incidents were considered; a one-lane
closure, a two-lane closure, and a reduction of the two-lane capacity to 15 percent of the
original capacity. Three incident durations with various conu'ol strategies applied to these

scenarios were tested.

3.1 NETSIM simulation model

Simulation is a standard tool of engineers in studying existing systems and in
predicting the behavior of projected systems (Gerlough, 1965). Traffic simulation
provides the mechanism for testing theories, modeling concepts, control strategies, and
new ideas prior to field demonstration. For this research, given that incidents are rare and
unplanned event, it is impossible to study the effect of traffic control strategies empirically.

Therefore, simulation is suitable for this research in that it provides an opportunity to

24

25

experiment with possible policies under predetermined incident and controlled traffic
conditions.

The NETSIM traffic simulation model was chosen for this study. NETSIM was
developed by the Federal Highway Administration for simulating traffic operations on a
surface street system. This microscopic model is based on the behavior of individual
vehicles, and the newest version (5.0) includes detailed features on signal and routing
controls. The model is accepted as an official tool for traffic analysis. The model has
recently been modified to incorporate advanced signal and routing control logic, although

these options are not yet available for public use.

3.2 Hypothetical network formulation

There is a general resistance among transport modellers to use hypothetical
networks for simulation work because of the potential for introducing unrealistic features
(Van Vuren and Leonard, 1994). Nonetheless, a theoretical network is essential when the
objective is to test a series of options, including interactions among features which have
not been applied in an existing network. The network can be designed to obtain a direct
measure of the effect of a change by controlling other characteristics that might impact on
the designated measure. The network thus can be used to obtain performance measures
under certain characteristics for several control strategies. Moreover, the results of one
case can directly be compared with another because the network environment is identical.

A postulated surface street network system shown in Figure 3.1 was constructed

for this study. The network is grid system, with each intersection one-quarter mile apart.

26

 

{a

Incident takes place at node 111

Figure 3.1 Street Network in this study

27

Streets are all two-way with a left turn pocket of 200 ft at each intersection approach.

Specification of the network and demand is illustrated in Figure 3.2.

3.3 Traffic demand and initial traffic signal setting

Traffic volumes were selected to be similar to the peak-hour traffic condition. The
traffic volumes and the associated level of service (LOS) is shown in Table 3.1.

The initial traffic signal settings were determined from the TRANSYT signal
optimization program. The program selected signal splits, offsets, and cycle length which
minimizes a performance index. The performance index for this study was system-wide
delay in the network. These initial signal conditions represent the optimal signal setting

for normal traffic condition using the off-line technique.

Table 3.1 Traffic demand

 

 

Level of
Traffic direction Traffic volume service"
(Vph) (Intersection 10)
East Bound (EB)
- Thru and Right turn 1350 B
- Left turn 150 E
West Bound (WB)
- Thru and Right turn 675 D
- Left turn 75 C
North Bound (NB)
- Thru and Right turn 900 B
- Left turn 100 E
South Bound (SB)
- Thru and Right turn 450 B
- Left turn 50 D

 

 

 

 

 

Note: Levels of service are determined at optimal progression based on
Synchro 2.0“ program (1995) and Highway Capacity Manual (1994).

28

 

 

 

 

 

 

 

 

—J L__
f‘—_——'
__\\ 1 f—
s/va '
: .12: it.
- 34%
a- 374-25
.4.
e b7 s
g \ J L 7 0 R J \—
O 6 \37 f
4 9 we: 1
1315 5 83E 75 I
O
7: :35 '
fl :
I

£959.

4
O
a
O

b. Volumes at typical intersections (approximate)

Figure 3.2 Intersection geometry and volume

29

The LOS in Table 3.1 were determined with this optimal signal timing in place.
The LOS is a qualitative indicator describing operational conditions within the traffic
stream, as perceived by motorists/passengers (Highway Capacity Manual, 1985). For
signalized intersections. the average stopped delay is used as a criterion for determining
LOS. The delays were obtained from the SynchroT-H traffic software. This model has a
major advantage over the standard Highway Capacity Manual calculation method in that
the model automatically determines arrival patterns resulting from the signal timings and

network configuration.

3.4 Incidents
The characteristics of the incident introduced in the network must be specified.
‘se characteristics include incident type, location. time of occurrence, frequency, and
ration. The type of incident determines the severity of the incident and thus the impacts
on traffic performance. The types of incident, for example, include accidents, stalled
vehicles, and minor or major roadway maintenance. The incident location characterizes
the storage capacity on the street section and the time it takes to affect the adjacent
intersections. The time of occurrence would change the effect because of the changing
demand pattern over the day, with the most serious impact being during peak hour traffic.
The duration of incident affects the congestion because it extends the time period of
reduced capacity of the system.
Incidents with one-lane closure, an 85 percent reduction of the street capacity, and
both-lane closure at a mid-block location were selected for study. The one-lane closure

incident represents illegal parking, stalled vehicles, or one-lane roadway maintenance in

30

the curb lane. The incident with an 85 percent reduction in the roadway capacity blocks
one traffic lane and reduces traffic in the other lane. This might be a characteristic of an
accident. The traffic in both lanes is assumed to pass the incident location on a first-in-
first-out basis. The discharge during the incident is determined from the maximum flow at
the bottleneck. The both-lane closure incident represents some serious accidents or major
roadway activities which require total blockage. The range of the type and severity of the
incident offers the opportunity to investigate the different impacts of selected control
strategies. The location of the incident in the network is shown in Figure 3.1.

Three durations of the incident were planned for the analysis, namely 5, 10, and 15
minutes. The duration of the incidents was limited by the recovery time to normal traffic
operation in the most severe case, the 15-minute both-lane closure. With this size and
configuration of the network, level of demand, and location of the incident, traffic
operations in the 5 and 10-minute both-lane closure incident recover within one hour after
the beginning of the simulation, while it takes about one and a half hours to recover in the

15-minute both-lane closure case.

3.5 Technological scenarios and control strategies
The study was formulated to test four technologies in addition to the normal
operations. The four options were:
a. Do—nothinguno special control strategies applied;
b. ATMS only-«traffic metering;
c. ATIS only—traffic diversion; and

d. ATIS/ATMS-combined traffic diversion and signal timing modification.

3 1
3.5.1 Do-nothing

The do-nothing case means there is no diversion and no new control strategy
applied during the incident. The signal timing remains the same as before the incident

occurred.

3.5.2 ATMS only
Traffic metering was chosen to represent the ATMS only scenario. It is assumed

that signals on the main arterial are responsive to the queue length. The traffic signal
timing at an intersection is modified when there is a possibility that this intersection will be
blocked in the next cycle. With the signal modification, traffic is released to approach the
incident at the same rate that traffic is discharged at the incident location. This technique
ensures that the queue on the link does not result in intersection blockage. In the both-
lane blockage incident, traffic is prohibited from entering that link. The time to implement
this control is determined by calculating the growth of the queue, a special sensor situated
at an upstream location on the link to detect the presence of queue, or manual monitoring

from surveillance cameras.

3.5.3 ATIS only
Traffic diversion represents the ATIS only scenario. Traffic diversion is a method
to reduce the demand approaching an incident-caused congested area by rerouting traffic
away from the incident. Diversion is the only demand restriction solution that can be
applied to traffic already on the street at the time an incident takes place.
There are many mediums to implement traffic diversion. The route information

can be disseminated via a dynamic route guidance system, a changeable message sign, or

32

even police regulation. With a dynamic route guidance system, the percent of compliance
to the information is one major issue. However, an assumption of 100 percent compliance
is used in this study.

The diversion paths can be determined in many ways. If no information on travel
time on possible diversion routes is available, then traffic can be equally distributed to
parallel arterials. The simplest point of diversion is at an intersection upstream to the
incident. A more sophisticated method includes the calculation of dynamic travel time on
all possible paths and dynamic traffic assignment. However, this method requires
information on dynamic signal controls. To date, there is no known diversion method that
considers the combined effect of the adaptive signal control and dynamic traffic
assignment. The diversion plan can also be determined from the results of simulation runs.

. percentage of the distribution to adjacent parallel arterials can be set to yield the

minimum total travel time.

3.5.4 ATIS/ATMS
The traffic diversion combined with the signal modification on the diversion route
is selected for this category. While there are several methods to initiate a signal
modification, the signal setting based on degree of saturation is chosen in the initial case.
The degree of saturation control variable is a well-accepted method of signal control
setting. In theory, Webster used this criterion for his signal split determination. In
practice, several urban traffic control systems such as SCATS and SCOOT have adopted

this method in their signal modification.

33

At each intersection green splits can be determined independently according to the
local demand. In many “reactive” traffic controllers such as the SCATS strategic control
level, local traffic patterns in a previous cycle are used to determine the split settings.
Traffic counts are then translated to the degrees of saturation (DS). The degree of
saturation is the ratio of the actual amount of green time utilized by traffic to the ideal
(minimum) amount of green time which could serve the same amount of traffic. The DS
can also be viewed as a measure of unused green time and the minimization of the unused

green time means the lowest intersection delays.

3.6 Measures of effectiveness (MOEs)

Several MOEs that are apprOpriate for use in characterizing the control of incident
congestion were used. To be useful, the MOEs must be understandable and easy to
obtain. Moreover, the MOEs must reflect the objectives.

Indicators related to stops, delays, travel time, and productivity are used for this
incident congestion management. These MOEs can be drawn from general MOEs
suggested by Pignataro et. a1. (1978).

The MOEs used in recurrent and nonrecurrent traffic conditions are different due
to different desired objectives. The incident congestion also requires some MOEs which
may not be the same as recurrent conditions.

The MOEs selected for this study are:

a. Total travel time;
b. Delay time;

c. Queue time;

34

(1. Time to dissipate the congestion; and

e. Duration of spillbacks.

3.6.1 Total travel time
Total travel time is the sum of the travel time for all individuals completing their
trips. This measure considers number of motorists, distance of travel, speed, and delay
associated with the travel. In the diversion versus non-diversion cases, this MOE can be
used to determine whether the time saved due to lower delay on the diversion route
exceeds the increased travel time due to the longer distance on the diversion route. The
total travel time may be the most appropriate indicator for a system operator, who seeks

the minimum overall system-wide travel time.

3.6.2 Delay time
Delay is a very good MOE in that it reflects the traffic operation as perceived by
system users. Overall delay over a period of time reflects the efficiency of the system
without considering the impact on individuals. Some control strategies such as gating are
designed to sacrifice some traffic movements in order to yield higher overall productivity
and the impact of this strategy can be reflected in this MOE. This MOE, however, may
not be appropriate for comparing diversion and non~diversion control strategies as the

total delay does not consider the increased distance on the diversion routes.

3.6.3 Queue Time
Queue time is similar to delay time except it considers only the period of time

vehicles spend in a standing or moving queue. Normally the queue time is a proportional

35

to the delay time. However, in NETSIM, queue time is recorded at any time whereas the
delay time is collected only when a vehicle departs a link. With this reporting
characteristic, the queue time is a good representation of the congestion currently on the

streets .

3.6.4 Time to dissipate congestion
This MOE is useful for describing the effect of congestion on traffic conditions as

a whole. One of the objectives of congestion management is to clear the congestion as
soon as possible. This indicator reflects the impacts of a control strategy on congestion
duration. However, this MOE is very difficult to measure in the field due to the fact that it
depends on how congestion is defined. In the NETSIM model, this MOE is not directly
available but a surrogate measure can be determined from the increased queue time on
each link. The queue time as a result of an application of a control strategy can be
compared with the queue time under normal traffic condition to find the duration of

congestion due to the incident and control strategy.

3.6.5 Duration of spillbacks
The duration of spillbacks roughly indicates the duration time of the congestion.
When spillbacks occur following an incident situation, these cause a breakdown in the
system. Spillbacks generate intersection blockages and make the congestion spread very
quickly. The end of spillback duration in the system is a measure of the quickness of the

action to relieve congestion.

36
3.7 Study plan

The study scheme for this research was designed to test the impact of the control
strategies under technological scenarios and variations in traffic flows. A set of control
strategies were selected to represent the technological groups and a detailed analysis of
the effect of these controls on the traffic were performed. Subsequently, alternate signal
control strategies as well as diversion alternatives were included in the experiment to seek
better results under selected incident situations. Sensitivity analyses of the base case
traffic parameters were conducted to determine the impact of these controls on changing

traffic characteristics. The study plan is shown in Figure 3.3.

3.8 Summary

The experiment was designed to test control strategies on a hypothetical network
using the NETSIM program. The base traffic demand represents the peak period, with the
approximate overall network LOS C. The normal traffic was operating with optimal
signal timing obtained from an off-line calculation. Incidents at mid-block with one-lane
closure. both-lane closure, and two-lane closure which reduces the capacity to 15 percent
of its original were introduced in this experiment. Four technological scenarios depicting
no-change in traffic control, the deployment of ATMS alone, ATIS alone, and the
combined ATMS and ATIS were studied. Traffic metering, traffic diversion, and traffic

responsive control strategies were designed in the analysis.

37

5E beam m.m 0.5mm”—

 

 

ummuuofi 38qu my

 

 

12:52 $3.63 Evy—om cm

 

 

336956.253: totem «.an Hominy

 

 

_

_ _
_

 

 

Genet: pans—un—

 

 

oaamot ecu—-53 8:58.0—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

oueago tea—=2. 3 SEE—Sm
25-53 23.58 83-52 25.53
8:262 I E m— 8:268 5:. fl 58:?! I E m— 5:262 I SE 3
€88 «.3 as 2 panacea as 2 seas $3 55 2 beau saw 558
wen—oce :E m ecu—reg E m 25.25 5:. m 25.28 E m
.8582“. 6E 2.
E2288. .Ewm c2335 uEEb merge... USE...
aims: 2% m5. 22.. 95. ”552.8
_ _ _ _

 

 

 

 

 

 

 

 

 

 

 

 

 

8558 33:58. 289:... a. caste .238 Ear...

 

€2.83 2 same

_ 338385 =e ~< T639. zombie :O_ _ o6: meam _ 7:ch 33—. §§§_ _ 5:228 «e fitmen— _

— _ _ _ _
_ _

$838 owemfi 3&5 8535» 35:3 vim—.38.. 938a.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mF<\m2._.< mF<\m2._.<
flame—o ecu—-53 255E6—
B>=an=a 15:8 925.4.
8
3
minute 3.2.2. 2
8:55.. 3.2.5 cceaasaeaefio 8:359... Seam
32688.5 Emu—awn: 23 .< 5583.5 5829? 3. 52.53.25 Booked: .<
I r I I n I H
255m mew—.8 2.2qu
mF<

 

Esme—u ace—-53 83£EA=

852:8... use-5.. mP—Z

Chapter 4

SIMULATION ANALYSIS AND RESULTS

The NETSIM model was used to simulate traffic conditions under several control
scenarios. Measures of effectiveness (MOEs) from the program were collected and
compared. The first section described traffic conditions when incidents were introduced.
Then various control strategies were tested on the same incident situations and their

VIOEs were compared.

4.1 Effect of type and duration of incident

Table 4.1 shows the effect of different types and durations of incidents on traffic
operations when there is no change in traffic control. The three types of incidents have a
significant difference in their effect on traffic congestion as measured by travel time,
intersection blockage and congestion duration. They also have different impacts on

different traffic streams.

4.1.1 Total travel time
As shown in Table 4.1, the one-lane closure incident produces no discernible
impact on traffic operations at this demand level because traffic on the blocked lane is able

to switch to the other lane.

39

40

Table 4.1 Total travel time in the network under incidents

 

 

 

 

Total travel time in one hour (veh-hrs)
Incident type Incident duration (minutes)
0 5 10 15
(no incident

One—lane closure 774.3 774.3 774.2 774.3
85% reduction in 774.3 779.2 793.8 841.4
capacity

Both-lane closure 774.3 788.3 882.5 1193.3

 

 

 

 

 

 

 

In the 85 percent reduction of capacity and both-lane blockage situations, total
travel time increases exponentially as the duration of the incident increases. Traffic
passing an incident which reduces the capacity at the bottleneck to 15 percent of its
original capacity must stop and wait in a queue before proceeding past the incident. Traffic
proceeding through the both-lane closure incident must stop until the incident is cleared.

As expected, traffic experiences longer delay when the duration and severity of an

incident increase.

4.1.2 Intersection blockage
The increase in total travel time is exacerbated by spillback (intersection blockage)
on approaching links and adjacent arterials. The spillbacks and durations are shown in
Table 4.2.
In the 85 percent capacity reduction incident situation, an intersection blockage

occurred at the nearest upstream intersection at 326 seconds after the incident started. The

41

Table 4.2 Start and end time of intersection blockage:
various incident types and durations

 

 

 

 

 

 

 

 

 

 

 

 

 

Time of spillback occurrence after incident (sec)
Intersection number Incident duration (minutes)
5 10 15
Begin End Begin End Begin End
85% reduction in capacity‘
10 326 335 326 636 326 936
9 n/a n/a n/a n/a 917 1132
Both-lane closure 1
10 212 337 212 648 212 974
9 n/ a n/ a 602 725 602 1020
409 n/a n/a n/a n/a 879 1089
14 n/a n/a n/a n/a 721 981
13 n/a n/a n/a n/a 824 898

 

 

 

 

 

 

 

 

 

Note: " The intersections are periodically cleared during the spillback periods.

blockages at this intersection ended in 9, 310, and 610 seconds after the initial blockage
for 5, 10, and 15-minute incidents respectively. Blockage at the prior upstream
intersection also occurred in the 15-minute incident situation, starting at 917 seconds after
the incident was introduced and lasting for 215 seconds. However, these upstream
intersections were intermittently clear as traffic was released at the incident location or at
the downstream intersection. The total duration of blockage was 8, 118, and 237 seconds
at the nearest intersection for 5, 10, and 15-minute incidents and 46 seconds at the second
nearest intersection for the 15-minute incident situation.

In the 5-minute both-lane closure incident situation, the blockage at the nearest

upstream intersection started at 212 seconds after the incident was introduced, and lasted

42

for 125 seconds. In the 10-minute incident situation, the blockage started at the nearest
upstream intersection at the same time as in the 5-minute incident situation but lasted for
436 seconds, and the blockage at the prior upstream intersection began at 602 seconds
after the incident started and ended 123 seconds later. The blockage at the nearest
upstream intersection in the 10-minute incident situation lasted longer than in the 5-minute
incident situation because of the presence of the traffic queue waiting at the entry of the
link.

The 15-minute both-lane closure incident caused spillbacks on five road sections,
three of which were located on the street where the incident occurred and two of which
were situated on a parallel arterial. On the arterial leading to the incident, the blockages
were initiated at 212, 602, and 879 seconds after the incident, and lasted for 762, 418, and
210 seconds, respectively from the nearest to the furthest intersection. One intersection
on a side street (intersection 14 in Figure 3.1) was blocked starting 721 seconds after the
incident and lasting for 260 seconds. This secondary congestion is the result of the
blockage at the nearest upstream intersection to the incident (intersection 10). One
intersection on a parallel arterial (intersection 13) experienced spillback caused by the
blockage at intersection 14. The blockage on this intersection started 824 seconds after

the incident and lasted for 74 seconds.

4.1.3 Congestion duration
The time period and lateral extent of ‘ congestion in the network can be determined
from an analysis of queue time statistics available from the standard NETSIM outputs.

While delay is the primary MOE used in this analysis, the output from the NETSIM model

43

is not suitable for using this measure as a basis for tracking the spread of congestion
through the network. Figure 4.1 demonstrates why delay from the standard NETSIM
outputs cannot be used. This is because the delay for individual vehicles are collected
when they leave a link. Thus, during an incident that closes a link, the delay would go to
zero, even though vehicles are queued on that link. This output attribute results in high
delay being reported after the incident is cleared. As shown in Figure 4.1, the lO-minutc
both-lane closure incident was cleared at 900 seconds after the simulation started. The
incident caused a blockage at the downstream intersection and hence no traffic departed
the link between the 902nd and 1020th second after simulation. The delay was reported
when the spillback dissipated as shown by the high delay between the 1000th and 1500th
second. Travel time from NETSIM also has the same reporting characteristic as the
approach delay.

A more appropriate measure for determining the spread of congestion is stopped
delay. The stopped delay curve in Figure 4.1 was calculated from queue lengths obtained
every four seconds. The queue length is illustrated in Figure 4.2. The queue time from
the standard NETSIM output is found to be highly correlated with the queue length and
stopped time delay and thereby suitable for identifying the congestion period. It is noted
that the shape of the three curves shown in Figure 4.1 are different because of their
definitions.

The effect of incident type and duration on the length of the network-level
congestion periods is shown in Figure 4.3. All incidents started at 300 seconds after the
beginning of the simulation. In the 85 percent capacity reduction incident situation, the

traffic operations returned to normal at time 400, 900, and 1400 seconds after the

Link Delay (veh-min)

Queue length (veh)

44

Link (9,10), Through traffic, lO-rninute both-lane closure incident

 

 

 

 

 

 

 

 

400
3500
300«
250-»
2m“ +QueueTime
+Stopped(calculated)
150 - > +Delay (Approach)
100..
“1h
a
0‘ ::::;..-, : : \:!;;E;g§‘~
0 600 1200 1&0 2400 31110 3600

Time after simulation (see)

Figure 4.1 Statistics from NETSIM for determining congestion period

Link (9,10), Through lanes, lO-minute both-lane closure incident

 

 

A l I A A
v I v v v

d

a
a
j

d

O
a
r

 

 

s-
o a a 4
V V '

 

radlrhlll

 

0 coo 1&0 1000 2400
Time after simulation (lac)

Figure 4.2 Queue length obtained every four seconds

 

45

coon

owes—.6 35:8 6: Because cosmoweeu me 2:9,.“—

ooon 8cm

.3... coca—3:.- ..eae 05:.

83 coup

 

 

 

 

55-2 lit
sci: IT
554+
ascoztxl

 

 

 

 

 

 

2:555 5:262 ....._oenso 2029a mm d

rv8P

(urn-Iran) amp anano

46

83

5.83 3. 23E

.3: 5.3.95? beta 06:.

 

 

 

 

 

 

83 come coon 800 Sea 80— 85 08 o
.1 t« . n 0 0 . .1 o
a 8n
.. 8.
r 42v
I . it I.
.. 08
.. 08
£E-m~+ .. 8°—
seélll
:_E-m+ .I Saw
.2502le
.. 8:
.. 82
.r 08'
88

 

 

“coming: 9:503 ecu—150$ .

tutu-u») «In who

47

incident was cleared, for the 5, 10, and lS-minute incident respectively. With the same
order of incident duration, the both-lane closure incident congestion periods lasted 600.
1900, and 3900 seconds after the incidents were removed.

The spread and duration of congestion in the network can be presented as
illustrated in Figure 4.4. The queue time of all vehicles in a link is used to determine the
congestion period. The congested links in the network are plotted over time to determine
the congestion coverage and duration. In the lO-minute both-lane closure incident
situation, intersection blockages created congestion on both the street where the incident
occurred and the nearby north-south streets. After the incident is cleared and the spillback
dissipated, links downstream from the congested links receive heavy traffic and become
congested. This is illustrated in Figure 4.4a.

Figure 4.4b shows the congestion caused by the lS-minute both-lane closure
incident. Congestion occurs on the mainstream approaching the incident as well as on a
parallel arterial. Although the congestion starts to dissipate on the upstream links within 5
minutes after incident is cleared, the spillbacks produce major congestion on approaching
links as well as downstream links well after the incident is cleared. The congestion in the

network lasted for 80 rrrinutes after the incident was introduced.

4.1.4 Delay on different traffic streams
To better understand the effects of the various lane blockages, an analysis was
made to determine the relative delay to various traffic streams in the network. This
information is essential in determining the control strategies to be used in reducing the

impact of an incident. Table 4.3 shows total travel time of through traffic on the path

48

omen—.6 .9556 e: “5:558 be 5:82. 15 8...: 380...? v6 953...—

EuEuE 5cm mug—Ewe on

 

 

 

 

Ll

Eoofifi code 85:?— mm

 

yr

 

 

C

 

 

 

 

 

 

 

 

 

 

5:82 .822: a

.555 .2: wage. a

 

 

 

 

_

 

 

 

 

 

 

 

4|

«cu—£65 5cm 5558 S

 

 

LI

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.522: 3:5? use—-53 2355.2 .a

 

.555 see maze? 8

__ __ __ :
[3|

 

 

 

 

.855 sec 83:? m

__ ___
I T

 

 

 

IT
I fill

 

 

 

 

49

cataco— 28295 n—

EmEufi Saw 3355 om

 

 

_

 

 

IHI

 

 

 

 

 

EmEuE 33a «3:58 om

 

 

 

 

 

5:53 3. 2:9".

“5295 33m 8558 cm

_
a —

 

 

 

 

 

 

 

 

 

 

 

 

 

.522: “was 8.358 8

:2:

D

 

 

 

 

___:__

.5295 33a 3358 on

D

___:__

.522: 2.520 22.53 2252-2

 

 

 

 

 

 

.5205 his 8356 cu

 

 

 

 

 

IHI

 

 

 

 

 

 

 

 

“.3205 3:5 835:. 9.

 

 

 

___:__

.522: 33a «.2355 an

 

 

 

 

 

 

 

 

 

 

 

50

Table 4.3 Total travel time of traffic passing the incident

(traffic heading from intersection 409 to intersection 12)

 

 

 

 

 

 

 

 

 

Total travel time in one hour (veh-hrs)
Incident type Incident duration (minutes)
0 5 10 15
(no incident)

One-lane closure 62.4 62.4 62.3 62.4
85% reduction in 62.4 64.9 77.3 102.7
capacity

Both-lane closure 62.4 78.7 115.7 152.8

 

 

approaching the incident. The results reveal the same trend as Table 4.1 with travel time
increasing as the incident duration increases. However, there were some differences in the
impact on this traffic stream versus the impact on the entire network. Table 4.3 indicates
that the journey time on this through route in the lO-minute 85 percent capacity reduction
incident situation is lower than the journey time for a 5-minute both-lane closure incident
(77.3 vehicle-hours compared with 78.7 vehicle-hours). However, the total travel time of
all traffic in the lO-minute 85 percent capacity reduction incident situation (793.8 vehicle-
hours) is higher than that of the S-minute both-lane closure incident (788.3 vehicle-hours).
This means that the lO-minute 85 percent capacity reduction incident causes greater traffic
interruption in non-affected traffic streams than the S-minute both-lane closure incident,
although the spillback duration in the first situation is shorter.

As shown in Table 4.3, the lS-minute both-lane closure incident creates

widespread congestion in the network. For this incident, the increase in travel time on the

51

path passing the incident contributes a smaller proportion of the overall network travel

time than in the other cases.

4.2 Effect of different control strategies

After determining the effect of an incident on traffic with no change in traffic
control, different types of control strategies were tested and the impacts were measured.
This simulation experiment allows the testing of these control strategies alone as well as
the combination of two or more controls. In this study, the impact of the control
strategies in three different ITS technological groups were determined. The experiments
were performed on several incident types and durations.

The 85 percent capacity reduction and both-lane closure incidents were selected as
the base cases for determining the effect of different control strategies because they cover
a wide range of congestion levels and offer the opportunity to experiment with various
controls. Since one of the objectives of this study is to compare alternative control
strategies without traffic diversion, with diversion alone, and a combination of diversion
and traffic control, total travel time is a good indicator of performance because it includes

the effect of longer route distances resulting from diversion.

4.2.1 Total travel time
Four major traffic control scenarios were tested and compared. The total travel
time of all links in the network for these different scenarios is shown in Table 4.4.
Traffic metering in the partial lane closure incident situation consists of gating at
the intersection immediately upstream from the incident. Traffic is released to approach

the incident at a rate equal to 15 percent of the link capacity so that no growing queue

52

Table 4.4 Total travel time for four control strategy scenarios:
lO-rninute both-lane closure incident

a. 85 percent capacity reduction incident

 

Total travel time in one hour (veh-hrs)

 

 

 

 

Scenario Incident duration (minutes)
5 10 15
No traffic control change 779.30" 793.80 841.40
ATMS: traffic metering 781.79 792.99‘ 821.74
ATIS: traffic diversion 783.56 803.99 820.00"
ATIS/ATMS: traffic 789.19 804.76 823.27

diversion with signal
change (equalization of
degree of saturation)

 

 

 

 

b. Both-lane closure incident

 

Total travel time in one hour (veh-hrs)

 

 

 

 

Scenario Incident duration (minutes)
5 10 15
No traffic control change 788.30 882.50 1193.30
ATMS: traffic metering 787.55" 847.68 9u.2o
ATIS: traffic diversion 788.45 826.01" 869.76"
ATIS/ATMS: traffic 794.96 839.75 931.83

diversion with signal
change (equalization of
degree of saturation)

 

 

 

 

Note: " indicates the lowest total travel time for an incident situation

 

 

53

developed. In the total-lane closure incident, traffic metering is set at a rate which keeps
the intersection from being blocked by a queue. The green phase is skipped when the
queue stored on any receiving link is full.

ATIS traffic diversion is the assignment of traffic to adjacent arterials to avoid the
incident. Traffic is distributed equally between two adjacent parallel routes at the nearest
upstream intersection to the incident. The middle through lane can be used by both left
and right-tum traffic. No signal adjustment is made in this scenario.

In the ATIS/ATMS scenario, signals along the diversion routes are modified every
cycle, using traffic data of the previous cycle. The signal control change in the
ATIS/ATMS scenario attempts to equalize degrees of saturation of all approaches at each
intersection. This logic is similar to a SCATS network with no system level control. No
offset change was made in these initial simulation runs.

The results from Table 4.4a indicate that traffic metering control yields longer total
travel time than the do-nothing condition in the S-minute incident situation. The increase
in travel time is caused by the fact that gated traffic has to wait at the upstream link for
approximately one cycle (100 seconds), while the spillback occurs for only 9 seconds
(Table 4.2). In this situation, it is better to let the spillback occurs for a short time period.
Traffic metering in the 10 and lS-minute incident situations results in an improvement
over the do-nothing cases.

The traffic metering strategy in the 5 and lO-minute incident situation gives shorter
travel time than traffic diversion. This implies that the longer travel distance caused by the

diversion is greater than the time waiting in the queue at the upstream link. In the 15-

54

minute incident situation, the traffic metering gives slightly higher travel time than traffic
diversion alone.

In the 85 percent capacity reduction incident cases, traffic diversion in the 5 and
lO-minute incident situations does not improve the overall travel time. This is because the
diverted traffic has a greater increase in total time than the delay savings to the traffic
impacted by the incident if they had remained on their original travel path in these incident
situations. Since traffic can be partially released at the incident location, the system
performs better if vehicles wait to get through the incident than if they reroute to adjacent
streets. Although the 5 and lO-minute incidents create periodic blockage at the upstream
intersection(s), spillbacks are relatively short. The rerouting produces traffic disruption at
other intersections, creating higher delays to traffic on the diversion paths, and on overall
network performance. In the lS-minute incident, however, diversion leads to lower
network travel time as the blockage and waiting time to pass the incident adversely affect
the network operation.

Traffic diversion alone yields the lowest travel time for the lS-minute incident.
The addition of ATMS control increases total travel time slightly over the traffic diversion
only case.

The ATIS/ATMS control creates higher total network travel time than traffic
diversion alone for all three incident durations. In fact, in the 5 and lO-minute incident
situations, changing signal timing causes higher delay than the no—control change

scenarios.

55

For the both-lane closure incident (Table 4.4b), traffic metering improves the
traffic operations over the no control change situation. The level of improvement
increases as the duration of the incident increases.

Traffic diversion does not lower total travel time for the 5-minute incident
situation. However, for the lO-minute and lS-minute incident, this control decreases the
total travel time. The reduction is 27 percent from the no-control change situation for the
lS-minute incident since the diversion eliminates all intersection blockages.

The addition of ATMS to the ATIS does not reduce the total travel time for any of

the incident durations.

4.2.2 Intersection blockage
All traffic control strategies eliminated intersection blockage. The traffic metering
at the upstream intersection cut off the green light of the congested direction before the
back of the queue reached. In the diversion situation, traffic was re-routed before the

queue blocked the upstream intersection.

4.2.3 Congestion duration
The duration of congestion is another measure used to express the impacts of these
control strategies. As discussed in Chapter 3, the queue time was an appropriate measure
to identify the beginning and the end of the congestion period for the no control change
scenarios. However, because there is increased travel distance in the diversion control
strategies, queue time may be the best measure to determine the congestion period, but the

total travel time is a measure of traffic efficiency for these control strategies.

56

The 10 and lS-minute both-lane closure incidents were selected for this analysis.
The effect of different control strategies on congestion duration is shown in Figure 4.5. In
the lO-minute incident, the overall congestion ends 1900 seconds after the incident is
cleared if there is no control change. Traffic metering causes longer congestion duration
than the no control change situation by 200 seconds. This is because traffic metering,
although it successfully eliminates the spillback , stores traffic in the network to postpone
the surge of heavy congestion. It spreads the congestion peak but does not reduce the
congestion duration. On the other hand, the ATIS alone and the ATIS/ATMS control
strategies alleviate the congestion 800 seconds sooner than the do-nothing case.

In the lS-minute incident, the no control change creates congestion until 3500
seconds after the incident is removed. The traffic metering shortens the congestion
duration over the do-nothing case by 700 seconds. The ATIS/ATMS shortens the
congestion period by 1600 seconds over the no control change case. Diversion alone
eliminates the congestion period 500 seconds sooner than the ATIS/ATMS, or 2100
seconds sooner than the do-nothing case.

The extent of the congestion is shown in Figure 4.6. The lO-minute both-lane
closure is chosen in the analysis. The traffic metering strategy produces heavy traffic on
the link approaching the incident, as these links are designated to store the spillback
(Figure 4.6a). Traffic diversion creates heavier traffic on diversion routes during the
rerouting, but the duration is much shorter, as noted above (Figure 4.6b). The diversion
with signal modification produces higher delay on approaches to the intersections on the

diverted paths because green time is taken away from other direction for the diversion

57

comm

mofioazm _ozceo Eocotfi E235“. cozmowceu We Bani
.93. sea-.38.. 3:- 05:.
88 8%. 8.: 8w.

by

 

 

d.

 

 

m.2._.< 32.4. +
ace—a mF< III

ace—a mSF< +
excuse 35:8 o2 +
9E8. .2502 l.X.l

 

 

 

 

 

.522: 958.0 ecu—-53 2.5.5.2 .a

W) flab rim «1 we who

58

oovm

000v 8N1

5.83 3. 2:3...

.903 £523.... .0:— 2:3.

800 800 Svu coop

*

4P

 

 

 

 

m2h<\m:.< If
wee—n m:< III

ace—a m2h< Ill
$55 .833 e2 '0'
3:9: 7632 III

 

 

‘

 

.2:

:08—

: coup

. 8:

 

 

.522: 8:83 ecu—-52. 2.5.5-2 .e

mil-nu) crab up" u} an arena

59

822.2% .8250 2202—8 ”acumen—.8 .5 528.6 23 8E: Boo»: ed 952"—

:2232 2822: E

2822: Ban 3328 mv

2822: “was 62:25 3.

 

 

C

 

 

 

 

 

 

 

A _

___2

 

 

 

 

 

 

 

 

 

2322: cuts 832:. cm

 

 

 

 

 

__n__

2522: cuts 352:. mu

 

 

DD

___.___

C

 

 

 

 

 

 

 

 

 

 

 

 

 

:
an
: __

2522: Stu 22:25 m~

CD

 

 

 

_ I

E C
_ : :
2.0222 Ste «252:. on

 

 

 

 

 

 

.flfii

 

 

—
—
—
—

‘

___

 

 

D

 

 

 

 

 

:—

 

 

 

 

_ _.._

$2.22: 058,—. d

.522: note 3525 mm

 

_E

 

 

 

___—

 

C

 

 

 

 

 

::

.522: 25a 6252:. cu

 

EC

 

 

___:__

 

 

 

_ __._ _c_ __._
.5222 even 3325 m

 

D

 

 

 

 

D

 

 

 

 

 

 

—
—
_—

60

.5382 2822: a

28222 Ste 8528 my

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5:53 3. 28E

2822: 83m 8328 mm 28222 note 8525 cu

 

 

 

 

E3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

«522: Ste 22:25 2 28222 Ste 8355 m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_ c 1LT.

 

 

 

 

 

 

 

 

 

 

 

 

 

22802: 052,—. .n

61

E2282 2822: E

28222 32a 3325 ma

 

 

 

 

 

 

 

 

 

 

 

I %.

5.83 3. 2:2".

28222 Ste 3328 mm

___
Jul

4|

28222 Ste 8528 S

 

 

 

 

 

 

 

 

 

 

'T$T_

 

 

 

 

 

 

 

 

 

529% 2.22. 28 528222: meE: :2ch .o

2822: .22“? @2526 cu

 

 

 

 

28222 .22? 32.55 m

 

 

 

I.
I...
II
I

62

routes. The termination of signal modification after the incident is cleared creates heavy

traffic on the downstream links at rerouted intersections (Figure 4.6c).

4.2.4 Delay on different traffic streams
An analysis of the impact of these control strategies on different traffic groups was
performed since each control strategy provides different treatments to traffic groups. It is
possible that a control treatment would give an advantage to a specific traffic group and
sacrifice others, and total travel time alone cannot distinguish the impact on different
traffic groups. Therefore, the data for MOEs of specific traffic groups were obtained and
analyzed.
The traffic was divided into three groups, based on the potential for different
impacts resulting from the control strategies. They were:
a. Rerouted traffic;
b. Traffic competing with the rerouted traffic on diversion route(s); and
c. Traffic at other intersections.
MOEs for these traffic groups were measured to determine the impact of ATIS/ATMS
and ATIS only control strategies. Table 4.5 displays the results of this analysis. As
expected, the delay in making left turns for the diverted traffic is reduced under ATMS
control as the queue of these vehicles increased the degree of saturation for this
movement, resulting in an increased allocation of green time. However, the delay to
traffic in other directions, which are competing with the diverted traffic, increases as a
result of this reallocation of green times. Traffic in other locations also suffer higher delay

due to the interruption of progression caused by the signal adaptation. This interruption

63

 

 

 

 

 

 

 

 

 

 

$335.5 confine EwEuE

 

 

€5.53 .50: use 5 m6: 788.: .33.

 

mm. “mm endow mudmw 8.er 3.va mvdwu ~20?
05.0mm. mmxmv wmdmv mmd; 3.va 3%va 9?: .850 =<
35% a use.
$va mmdmm mmdum mmfium No.53” 2.3m £023 an 95203..an
365.5% 9.5. .2
$2 32 mm: mm: «3 8x EB 23m
Kém aflwm mm.m g 3.2 mad mod ER :3
mume
ufimb 96:3 838
commuozv so 9.63
3:5. c2936 so 359 so??? no 339 .8552... :e
owcanu 1ch 5:5 3:0 Examin— uwcazu in»? 5:5 3:0 coacmzo «wear—u 1:»? 5:5 3:0 c2520
2 S m 956 4::

 

2522: 230.0 25.52. ”museum 2&8. 3.52:3 he 2:: .022. .38. n4» 2me

 

64

of traffic in other directions is the main contributor to longer overall travel time in the

ATIS/ATMS scenario than in the ATIS alone scenario.

4.3 Summary

With the initial traffic and network condition in this study, the one-lane closure did
not affect traffic performance because the level of the demand was well below the reduced
capacity at the incident location. However, the 85 percent capacity reduction and the
both-lane closure did affect traffic operations. The longer the incident duration and the
more severe the incident, the more impact on the network travel time. The traffic stream
passing the incident had the greatest increase in delay, although other directions were
delayed by the spread of congestion.

Control strategies tested in this research had different effects on the incident-based
congestion. Traffic metering reduced travel time in the both-lane closure incident situation
and the longer duration of the 85 percent capacity reduction scenario. However, this
control did not have a major impact on the reduction in the length of the congestion
period.

Traffic diversion did not improve the traffic operation in the incident situations
which had short duration (5 and 10 minutes in the 85 percent reduction, and 5 minutes in
the both-lane closure). Traffic diversion was effective in the lS-minute partial lane
closure, and the 10 and lS-minute both lane closure scenarios. It reduced the network
total travel time despite the fact that some vehicles had to travel longer distances.

Diversion shortened the congestion period compared to the do-nothing case.

65

The traffic diversion with 'signal change along the diversion routes was not
effective in the 5 and lO-minute 85 percent capacity reduction, and the 5-minute both lane
closure incident situation. Although this control tended to favor the diverting traffic and
resulted in an improvement in total travel time over the no control change scenario in
other scenarios, it did not provide congestion reduction beyond the traffic diversion alone

strategy.

Chapter 5

SENSITIVITY ANALYSIS

The analyses in the previous chapter were based on a common set of traffic
parameters and a specified set of control strategies. In this chapter, the impact of the
control strategies under different traffic conditions and control concepts were investigated.
The traffic volumes were altered to depict higher demand, as might be found in the peak
period, as well as a low demand scenario representing the off-peak period. Alternate
diversion schemes were developed and examined. Different treatments for signal tinting,
including the use of different variables (other than degree of saturation) as a basic for
modifying the phase plan, were considered. Finally, the number of signals considered in

the ATMS strategies were varied.

5.1 Sensitivity to traffic demand change

To detemiine the effectiveness of the control strategies under diverse traffic
demand conditions, two traffic volumes were introduced to represent the traffic demand in
a highly-congested peak period and an off-peak period. The new traffic volumes and levels
of service (LOS) are shown in Table 5.1. A 15 percent increase in traffic volume was
used to represent a higher traffic demand and more congested network condition. A 30

percent decrease in traffic volume was selected to represent an off-peak traffic condition.

67

Table 5.1 New traffic demand used in the sensitivity analysis

a) Low demand-~off-peak period

 

 

 

Level of
Traffic direction Traffic volume service"
(VPh) (intersection 10)
East Bound (BB)
- Thru and Right turn 945 A
- Left turn 105 E
West Bound (WB)
- Thru and Right turn 472 C
- Left turn 53 D
North Bound (NB)
- Thru and Right turn 630 B
- Left turn 70 D
South Bound (SB)
- Thru and Right turn 315 B
- Left turn 35 E

 

 

 

b) High demanduhighly-saturated system

 

 

 

Level of
Traffic direction Traffic volume service‘
(V1310 intersection 10)
East Bound (EB)
- Thru and Right turn 1553 F
- Left turn 172 E
West Bound (WB)
- Thru and Right tum 776 B
- Left turn 86 D
North Bound (NB)
- Thru and Right turn 1035 D
- Left turn 115 E
South Bound (SB)
- Thru and Right turn 517 C
- Left turn 58 C

 

 

 

 

Note: Levels of service are determined at optimal progression based on
Synchro 2.0“ program (1995) and Highway Capacity Manual (1994)

 

68

For each demand case, signal timings were recalculated using the TRANSYT program.

The optimal signal timings were then used for normal signal operations.

The new demand levels were input into the NETSIM model. A lO-minute both-

lane closure incident was selected for this analysis.

5.1.1 Impact on total travel time

The impacts on travel time of new volumes are shown in Table 5.2. In all traffic

volume conditions with the lO-minute both-lane closure incident, traffic metering reduces

the travel time over the no traffic control change. The metering is more effective in the

high demand situation, reducing 56 percent of the increased travel time due to incident

under heavy traffic, compared to 36 percent in the low demand situation.

Table 5.2 Total travel time for three volume scenarios:
lO-minute both-lane closure incident

 

 

 

 

 

 

 

 

Total travel time in one hour (veh-hrs)
Scenario Volume scenarios
30% decrease Normal 15% increase
(off-peak) (peak) (runny-congested)
Normal traffic (no incident) 486.79 774.30 1005.10
No traffic control change 523.29 882.50 1287.95
ATMS: traffic metering 510.24 847.68 1128.66
ATIS: traffic diversion 502.36‘ 826.01‘ 1115.09
ATIS/ATMS: traffic 505.84 839.75 1076.09"
diversion with signal
change (equalization of
degree of saturation)

 

Note: ' indicates the lowest total travel time for an incident situation

 

69

Traffic diversion alone also improves traffic operations over the do—nothin g
situation in all demand ranges. However, in the off-peak and normal peak period, the
ATIS/ATMS strategy does not increase the benefits. In fact, modification of signal tinting
increases the travel time by 3.48 vehicle-hours in the off-peak demand case. In the highly
congested traffic condition, the addition of traffic signal control modification reduces the
travel time over diversion only, from 1115 vehicle-hours to 1076 vehicle-hours.

The results shown in Table 5.2 suggest that signal timing modification along with
the diversion is desirable in high traffic demand conditions. This may be because the high
volume of rerouted traffic contributes to longer overall delay in the diversion only
situation. The benefits of signal tinting modification favoring the rerouted traffic exceeds
the increase in delay incurred on other directions, thereby reducing the overall delay in this

scenario.

5.1.2 Impact on intersection blockage

The three traffic demand levels created diverse results when measured by
intersection blockages. The results were shown in Table 5.3.

Spillback occurs at all demand levels for the no control change scenarios. The
traffic volume determines the start time and duration of the spillback(s). The higher the
volume, the sooner the blockage begins. The start time of the blockage at intersection 10
starts at 248, 212, and 208 seconds after the incident occurs for off-peak, peak, and
highly-congested traffic conditions, respectively. The incident in the highly-congested
condition creates intersection blockages on the north-south arterial, as the result of the

spillback on the arterial on which incident occurs. All control strategies successfully

70

Table 5.3 Start and end time of intersection blockages: various traffic demand levels.
lO—minute both-lane closure incident

 

Time of spillback occurrence after incident (sec)

 

 

 

 

Intersection number Traffic demand ICVQI
Off-peak Normal peak Highly-congested
Begin End Begin End Begin End
10 248 634 212 648 208 631
9 n/ a n/ a 602 725 554 1015
14 n/ a n/a n/ a n/a 497 710
13 n/ a n/ a n/ a n/a 1075 1086

 

 

 

 

 

 

 

 

 

Note: The intersections are periodically cleared during the spillback periods.

eliminate the intersection blockages.

5.1.3 Impact on congestion duration

The congestion durations as determined from the queue time are illustrated in
Figure 5.1. Control strategies in each of the traffic demand ranges have similar potential
in reducing the congestion duration. The traffic metering strategy does not have much
effect on congestion duration over the do-nothing cases. Two diversion options, with and
without signal tinting modification, have similar impacts. In the highly-congested traffic
condition, however, diversion with signal modification clears the congestion 300 seconds
sooner than the diversion alone.

A comparison of the reduction in the congestion period for various traffic demand
levels is shown in Figure 5.2. The congestion duration in the highly-congested condition
is projected from Figure 5.1(c) because the simulation approached the NETSIM maximum

number of vehicles in the system. Figure 5.2 shows that the application of control

71

comm

26$. “Ems—ow 2.95:6 Bataan cocmuwcou _.m 2:!"—

.uou. fizz—3.... no... 06:.

 

 

 

 

 

 

 

 

coon covw coop camp 80 o
l. .l . 0 . c
.. 8.
iv am
It ‘1‘ II. t‘ l.

tr 8”

it 8'
m2h<\m:.<+
«co—a m_._.<lIl

28: mzzlxl .. 8m
smear—u 3358 02+
2:9: 36.52 12*}

at com

.. cos

coo

 

339:8 958% E 83836 E083 em .a

(trim-q“) apb tpu u! cum anano

72

comm

080

6.253 2 25E

.33 552:5? a8: 25,—.
oovw 80.. 89 80

 

 

 

 

mE< \m_.—.< lil

ace—a m_._.< +

vac—u mSE.< +

own—Eu 35:8 62 +
652. .2502 lx

 

 

*

 

1. SN

.. 82

 

 

Anne oSmE ma 25$ 335 38.52 .n

comp

(sq-trait) apb qua tn aura anano

73

80v

6.2.5 3 use...

to: 5:233. 33- 05:.

 

 

 

 

 

 

 

00.43 0000 0000 00cm 000— 8a.. 0
- t. 4 n - o 0
r can
m2h<\m:.<+ .. 00v
2.0.- m_._.<+
ace—a m2h<+
owes—u .9203 02+
052. 3.502le f 000
1 08
t \
r”
r 08..
C 009.
a. 8:
000.

 

 

Avenuwzeuécwé 05:50 E 833:. E022. 2 .0

(um-nut '13" Iran III ma mno

74

 

80

7o 0

 

D Do-nothing
50 “ IATlS/ATMS

 

 

 

 

Congestion duration (rain)
8

20 n

10 <>

.._l

 

 

 

 

 

 

 

 

Off-peak Peak Highly-congested

Traffic demand scenario

Figure 5.2 Comparison of reduction in congestion duration in different demand levels

strategies is more effective in reducing the congestion period at higher demand levels. The
ATIS/ATMS control reduces the congestion period to less than a half of the time required

under the no control change scenario.

5.2 Sensitivity to diversion control alternatives
The analyses in the preceding chapter were based on diverting traffic equally onto
the two adjacent parallel arterials at the nearest upstream intersection. Two new diversion
plans were developed and compared in this section. Three diversion alternatives
considered here were:
a. Traffic diversion at the nearest upstream intersection, equally distributed to

two adjacent arterials (the alternative plan from Chapter 4);

75

b. Traffic diversion at the nearest intersection, optimal distribution to the
arterials; and
c. Traffic diversion at the two nearest intersections, equally-distributed.

Under alternative b, the diversion percentage is chosen to yield the lowest network
travel time. This distribution was obtained by trial-and-error, because NETSIM does not
have internal logic for dynamic traffic assignment. The optimal distribution sent 56
percent of eastbound through traffic at intersection 10 to the north arterial (passing
through intersections 10-6-7-11) and 44 percent of the traffic to the south arterial (passing
through intersections 10-14-15-1 1). The last alternative was to divert half of the traffic at
each of the two nearest upstream intersections. At each intersection the traffic was

equally distributed to the two parallel arterials.

5.2.1 Impact on total travel time
The total travel time employing these diversion plans is shown in Table 5.4. The
optimal distribution plan gives the shortest travel time, although the reduction is only 5.18
vehicle-hours or 0.6 percent of the total travel time from the equal distribution diversion
plan. The diversion at two upstream intersections gives only a slight improvement over

the equal diversion at the nearest intersection.

5.2.2 Impact on congestion duration
Figure 5.3 shows the congestion duration of the three diversion procedures. All
three diversion plans create similar incident durations. However, the two-intersection

diversion has 200 seconds longer congestion than each of the one-intersection diversions.

76

Table 5.4 Total travel time of alternate diversion plans
lO-minute both-lane closure incident

 

 

 

 

 

Diversion plan Total travel time in one hour (veh-hrs)
Equally distributed at the 826.45
nearest upstream intersection
Optimal distribution at the 821.27
nearest upstream intersection
Equally distributed at two 825.20
upstream intersections

 

In contrast with the congestion duration, the spread of the congestion on the
network was significantly different. The effect of congestion spread is illustrated in Figure
5.4. The optimal distribution diversion results in a slightly longer congestion period on the
north parallel arterial, to which more traffic is assigned, and a slightly shorter congestion
period on the diversion route passing the south parallel arterial. The two-intersection
diversion causes congestion at the second upstream intersection. The congestion at the
nearest upstream intersection clears sooner than the one-intersection rerouting since less
traffic is approaching this intersection (half of the through traffic is diverted at the prior

intersection).

5.2.3 Impact'on different traffic streams
To understand the impact of the control on different traffic movements, traffic
was arranged into three groups. The first group was the traffic on the diversion routes
including the paths passing the second nearest intersection. The second group was the

traffic competing with the diversion traffic, at the intersections on which the diverted

77

comm

mam—n. 5.862.. 295.... "cos—use cosmowceu fin 2:2...

.3: £23.35.- ..o... 05:.

089 Svu coop

l?

coup So 0

 

 

 

20:332.: Seabed:

63. .a 8:33»... _asumlOl
CO—ouvafleflc

.330: .n 00.25230 .uE..nOl.TI
00:02:20.

7.23: 3 00:25.20 Warm—IT

owe-£0 .0308 02 +

 

 

2:2. .2502 le

 

 

 

comp

 

mutant: Iran "Imam-no

78

was... 00.0020 3203.0 “=0..mow=00 .0 0.0.3.5.. 000 8.0.. 0300»? tn 2:»...

.000.00. 02.0 00.00.00 mu 00.0.00. 00....0 00.00.00 cu

 

 

 

 

 

 

 

 

 

 

 

 

 

00.0000. 000.00. E

 

 

 

 

 

 

.528. a... 8358 m. .838. can 8555 2 .5005 as: 8.2.5 m

_ _ _ __ ___

 

 

 

 

 

 

 

 

 

it:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l
‘1“‘1‘

 

 

 

30.0 00%.". mm 00.3. 00.00082... 0:60.09 ._ .3: 0... .0 00:30.20 .0: .0

13'

79

003000. 0000.00. M

0000.00. 03% 00.00.00 m.

 

 

 

 

 

_

 

 

60:80 «0 2.0.".

0000.00. 000.0 0030.0. mu

 

lrr

 

Ll.

 

C

 

 

 

 

 

 

 

 

4 _

 

«00.0.00. 02.0 00000.00 3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.8005 0% 855a 8

 

___

 

 

 

 

 

 

 

 

0000.00. 00...» 00.00.00 m

 

 

 

 

 

4|

 

 

 

IJI

00.60.0000. 0.02.000 .8000: 05 .0 00.50.0020 .0058 .0

 

 

 

 

80

00.0000. 0000.00. E

000.00. 000.0 0030.5 m.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

€2.80 «.0 2:00

000.00. 000.0 00000.00 mm

 

LT

 

 

_L

 

 

 

 

 

 

 

4|

0000.00. 0000.0 00000.00 3

 

 

 

 

 

000.000.0000... E00003: 030 00 00:00:00... .005 .0

 

l
*‘f‘f—

 

080.0000 0200 80:56 8

 

 

L_

 

 

 

 

 

 

0000.00. 000.00 00.00.00 m

 

 

4

 

 

 

 

 

Jul

 

81

traffic traversed. The last group was all other traffic. The effect of the alternate diversion
plans on different traffic groups is shown in Table 5.5.

Comparing the alternate rerouting plans at the nearest intersections, the optimal
distribution plan results in lower travel time on the diversion routes. Moreover, the
optimal plan produces the minimal interruption to other traffic directions competing with
the diversion routes. By sending more traffic to the north arterial, the rerouting produces
less impacts on traffic in other directions as traffic diverting to the north arterial makes
two right turns, compared with two left turns on the south arterial diversion path.

Comparing the equally-distributed diversion at the nearest intersection and the
diversion at two upstream intersections, the latter plan produces lower travel time (delay)
on the diversion routes. The travel time of diverted traffic is shorter by 3.58 vehicle-hours
in the two-intersection diversion case. The two intersection diversion results in lower
congestion at each intersection. Although the two-intersection diversion creates higher
delay in non-diverted traffic, the overall travel time (delay) is shorter by 1.65 vehicle-

hours.

5.3 Sensitivity to signal control variable alternatives

In the previous analysis, the equalization of degree of saturation was used as a
criterion for signal modification. In this section, two other control variables were
introduced and tested. The control parameters were maximum queue length and stop

time.

82

Table 5.5 Total travel time of traffic streams in altemate diversion plans:
lO-minute both-lane closure incident

 

 

 

 

 

 

 

 

Total travel time in one hour (veh-hrs)
I ink Group Diversion plan
Equally distributed at Optimal distribution Equally distributed at
the nearest upstream at the nearest two upstream
intersection upstream intersection intersections
Links on diversion
routes' where traffic
makes
Left turn 25.57 24.51 23.05
Through 38.19 38.01 37.88
Right turn 15.24 16.24 14.49
All links approaching
intersections at which 513.62 507.08 515.27
traffic is diverted2
(except the diverted
traffic direction)
All other links3 233.83 235.43 234.51
Total 826.45 821.27 825.20

 

 

 

Note: 1. This traffic group contains intersection 5, NB-RT

intersection
intersection
intersection

6, EB-THRU, NB-RT
7. EB—RT
9, 58-1513 EB-RT

intersection 10, EB-LT, EB-RT
intersection 11, SB—LT, NB—RT
intersection 13, SB-LT
intersection 14, EB-THRU, SB-LT
intersection 15, EB-LT
2. This traffic group contains all traffic directions at intersection 5, 6, 7, 9, 10, 11, 13, 14, 15
excluding diverted traffic movements above
3. This traffic group contains all traffic except the two traffic groups above.

83

The maximum queue length in each traffic movement is a good indicator of traffic
demand at the intersection. The signal setting strategy used in this case is to balance the
maximum queue lengths in all approaches, which leads to the postponement of spillback
and the elimination of secondary congestion. The spillback can be either left-tum queue
overflow to the through traffic lane or the intersection blockage at an upstream
intersection.

The maximum queue length in each cycle is not provided in the NETSIM standard
output. However, it can obtained by collecting queue data every four seconds from the
NETSIM intermediate statistics.

The stop time delay is another possible control variable available from NETSIM.

.: equalization of stop time delay in all competing approaches implies equality of the

icvel-of-service (LOS), as defined in the Highway Capacity Manual, in all directions.

5.3.1 Impact on total travel time
Table 5.6 shows the total travel time for the three signal control alternatives. The

signal setting by equalizing of the stop time yields the lowest total travel time. It improves

Table 5.6 Total travel time of alternate signal control variables:
lO—minute both-lane closure incident

 

Signal control variable Total travel time in one hour (veh-hrs)

 

Degree of Saturation 839.75
Maximum Queue length 833.33

Stop time 819.53

 

 

 

 

84
this MOE by 2.4 percent of the total travel time beyond the setting by degree of saturation

for the 10—minute both-lane closure scenario.

5.3.2 Impact on congestion duration
The congestion duration for each of these three signal control settings is shown in
Figure 5.5. The signal settings using the degree of saturation and stop time alleviate the
congestion 300 seconds sooner than that using the maximum queue length. The control
based on the stop time, however, produces the least time in queue during and after the

incident, as reflected in the minimum total travel time measure.

5.3.3 Impact on different traffic streams
Table 5.7 shows the effect of these signal settings on different traffic groups.
Among these three controls, the equalization of stop time concept is the best because it
reduces travel times at intersections on the diversion routes over the degree of saturation
and the maximum queue length criteria. The reduction in travel time on diversion routes is

greater than the increase on other links, resulting in the lower overall travel time.

5.4 Sensitivity to signal control coverage

A further investigation was made to explore the effect of the boundary of the
signal control modifications. In the previous analyses, the signal modification was limited
to the intersections on diversion routes. In this analysis, signals at all internal intersections
in the network were modified. The simulation was then performed and data on MOEs
were obtained The signal tinting for this experiment were controlled using the degree of

saturation as in the preceding chapters.

85

80a

00.00.0000 .00000 .0090 20002.0 ”00.00000 00.000300 Wm 00%.".

San 824.

.00.. 00.00.0000.- 00000 05:.

con.

coup

 

 

 

00... 00%|?

£900. 050:... 00:00.va1'
00.00026 .0 0050a +
0902.0 30.000 02 IOI.

0... 0.: .2502 l.X.l

 

 

 

by
‘

 

 

 

:03

:80

.. coop

coup

Wimbttmurmnmo

86

Table 5.7 Total travel time of traffic groups of alternate signal control variables:
lO-minute both-lane closure incident

 

 

 

 

 

Total travel time in one hour (veh-hrs)
Link' Group Signal control variable
Degree of Maximum Queue Equalization of
Saturation Length Stop time
Links on diversion
routes where traffic
makes 15.33 14.46 13.44
Left turn 11.26 10.26 9.98
Right turn
All links approaching
intersections at which 379.88 374.44 360.72
traffic is diverted
All other links 433.28 433.87 435.39
Total 839.75 833.33 819.53

 

 

 

 

 

 

5.4.1 Impact on total travel time
Table 5.8 shows the total travel time for two scenarios with different signal
coverage. The modification of signal timings at all intersections results in lower total
travel time. By making the signals at these intersections more responsive to the traffic, it

results in better progression and lower total travel time.

5.4.2 Impact on congestion duration
Figure 5.6 shows the congestion durations of these two signal coverage

alternatives. Although the signal modification at all intersections produces slightly lower

87

Table 5.8 Total travel time of alternate signal control coverage
lO-minute both-lane closure incident

 

Signal control coverage Total travel time in one hour (veh-hrs)

 

On diversion routes 839.75

 

All intersections 826.48

 

 

 

delay (queue time) than the signal change on diversion route case, the network recovers

from congestion at the same time.

5.4.3 Impact on different traffic streams

The preceding section shows that the larger coverage area of signal modification
yields better traffic operation, as indicated by shorter overall travel time. The analysis of
travel time on particular traffic streams reveals that the benefit comes from the lower delay
at competing approaches on the diversion routes. The results are shown in Table 5.9. On
:he diversion route, the signal modification at all intersections slightly reduces the travel
time in the diverted traffic directions. However, most of the delay reduction occurs on
traffic in other directions, e.g. through traffic on the parallel arterials. This is because the
adaptation of signals at intersections upstream and downstream from the intersections
receiving diversion traffic results in more response to the cycle-by cycle traffic situation,
leading to better progression on these parallel arterials. The signal modification at all
intersections reduces the delay of competing movements to diverted traffic from 379.88
vehicle-hours to 361.74 vehicle-hours, although it increases the travel on other links from

433.28 vehicle-hours to 438.43 vehicle-hours.

88

0w000>00 .0030 20002.0 ”00.00000 00.0mom00u 9m 0.00mi

.00.. 00.00.00... 00:- 05....

comm 080 8Q“ 80' GON— 80 o

 

48v

 

00030000000. ..<I.|
00.000 00.2030 00+
0m00...0 .000000 02 .IOI
0... 00. .000002 .IXI

 

 

 

 

 

 

 

 

coup

(urn-unmisttrwi-uamnb

89

Table 5.9 Total travel time of traffic groups of alternate signal control coverage
lO-minute both-lane closure incident

 

Total travel time in one hour (veh-hrs)

 

 

 

 

 

 

 

 

 

Link Group Diversion only With signal With
change on signal change at
diversion routes all intersections
Links on diversion routes
where traffic makes
Left turn ' 18.82 15.33 15.05
Right turn 11.33 11.26 11.26
All links approaching
intersections at which 377.93 379.88 361.74
traffic is diverted
All other links 418.33 433.28 438.43
Total 826.01 839.75 826.48
5.5 Summary

The control strategies used in Chapter 4 were tested to determine their
effectivenesses under different demand conditions. The “traffic metering” and “diversion
only” alternatives were based on the optimal fixed time signal plans for each demand level.
The impact of these control strategies under traffic demand representing off-peak, peak,
and highly-congested conditions were obtained for the lO-minute both-lane closure
incident situation. The traffic metering produces lower total travel time (delay) than the
no control change in each demand range. However, the traffic metering does not have
much impact on the congestion duration. Traffic diversion is more effective than traffic

metering, and diversion with signal control change is the most effective at the high demand

90

level. For both diversion with and without signal modification, the reduction in congestion
period increases as the traffic demand increases.

Three different diversion and signal control setting alternatives were analyzed.
The optimal distribution at the nearest upstream intersection yields the lowest delay. The
distribution at two upstream intersections creates less congestion for each intersection but
in a wider area. The duration of overall congestion in the three diversion plans lasts
approximately the same. The optimal distribution produces minimum interruption to
traffic competing with the diverted traffic directions.

Signal control modifications using three control variables were tested. The signal
setting using the stop time produces the lowest total travel time, although it does not
shorten the congestion period beyond the setting using degree of saturation. The
reduction in delay occurs at intersections along diversion routes. Analysis on signal
control coverage indicates that the signal modification at all intersections results in lower
delay than the signal change on diversion routes since it increases progression on the

parallel arterials.

91

Chapter 6

CONCLUSION AND RECOMMENDATION

6.1 Conclusion

This research examined the effectivenesses of several control strategies on various
incident conditions. The aim of these controls was to alleviate the congestion caused by
inese incidents. In this study, the control strategies considered were traffic metering,
traffic diversion, and signal timing modification. The research was conducted using the
NETSIM simulation program. A hypothetical network of a surface street system was
used. Using the same platform, several incident situations as well as control strategies
were tested. The measures of effectiveness used to identify the performance of each
control strategy were total travel time (delay and queue time), congestion duration, and
spillback duration.

The impacts of three incident types, each with three incident durations, were
studied under the assumption of no control change in traffic control system. The
investigation showed that, in the severe incident conditions, the congestion lasted up to 68

minutes and produced 419 vehicle-hours of delay if no special control strategy was

applied.

92

The impact of the control strategies on the various incident situations was
analyzed. It was found that different control strategies had different levels of
effectivenesses in specific incident conditions. These results are shown in Table 6.1. In a
less severe incident situation, such as the one-lane closure and the 5-minute 85 percent
reduction in capacity, none of the control strategies offered any improvement in traffic
operations over the do-nothing case. In the lO—minute partial-lane and 5-minute both-lane
closure incidents, only the traffic metering strategy reduced network travel time. In the
lO—minute partial-lane closure situation, traffic metering did not have a major impact on
'he congestion period.

Traffic diversion was effective when the severity of the incident increased (15-
minute 85 percent capacity reduction, and 10 and lS-minute both-lane closure). This
control strategy substantially reduced the length of the overall congestion period.
Although diversion results in an increase in congestion on the diversion routes for a short
time period, it reduces the congestion on the affected traffic and the overall network. The
diversion with signal timing modification strategy did not offer any improvement over
traffic diversion alone.

The limits of effectiveness of these control strategies were examined by conducting
a sensitivity analysis on their effectiveness at different demand levels. The control
strategies were tested using volumes representing off-peak, peak, and highly-congested
traffic conditions. The results indicated that, when the demand level increases, the control

strategies are more effective in reducing both total travel time and congestion duration.

93

Table 6.1 Impact of alternative control strategies on various incidents .

a. Difference in total travel time from the no control change scenarios (percent)

 

Incident type and duration I

 

Control strategy 85 percent capacity reduction I

Both-lane closure I

 

5 minutes 10 rninutesllS minutesl 5 minutes 10 minute5115 minutesl

 

 

ATMS: traffic metering + 50*
ATIS: traffic diversion + 85
ATIS/ATMS: traffic +193
diversion with signal

timing modification

 

 

- 4*
+52
+56

-29

- 32"

-27

 

 

- 6‘ -32 -60
+ 1 -52“ -77"

+48 -40 -62

 

 

 

 

b. Difference in congestion duration from the no control change scenarios (percent)

 

 

 

(both-lane closure incident)
Control strategy Incident duration
10 minutes 15 minutes
ATMS: traffic metering + 8 . 15
ATIS: traffic diversion . 32* , 36"
ATIS/ATMS: traffic - 320 _ 25
diversion with signal
timing modification

 

 

 

 

 

Note: 1. the negative signs indicate the improvement over the no control change
2. bold numbers indicate the situations where the control strategies are effective
3. the ' means this control strategy produces the greatest improvement in that
incident type and duration.

94

Moreover, in the highly-congested condition, diversion with signal modification became
the most effective. The results are shown in Table 6.2.

Three diversion plans were compared: equal distribution at the nearest intersection,
optimal distribution at the nearest intersection. and equal distribution at two upstream
intersections. As shown in Table 6.3, the optimal distribution diversion plan results in a
slight improvement over the equal distribution plan.

For each control strategy, variations in the techniques to accomplish the controls
were studied. For the signal timing modification (ATMS), three control variables were
used to modify the signal timing; the degree of saturation, the maximum queue length, and
the stop time. The use of stop time yielded the lowest total travel time. The results of this
analysis are shown in Table 6.4.

The signal coverage experiment indicated that the signal modification at all

intersections was better than only changing the signal on the diversion routes.

6.2 Recommendation

This research has contributed to the understanding of the impact of several control
strategies on incident congestion. It also identifies several areas where there is a need for
further research. The study can be extended to cover additional incident types and
durations under various levels of demand. This could be used to develop a set of rules to
make real time decisions on alternative control strategies depending on the traffic
condition existing at the time of the incident and the severity of the incident.

The same study framework can be used to experiment the impact of the incident

on different network geometric configuration to uncover an absolute effectiveness of a

95

Table 6.2 Effectiveness of the control strategies under different demand levels

a. Difference in total travel time from the no control change scenarios (percent)
(lo-minute both-lane closure incident)

 

 

 

diversion with signal
timing modification

 

 

 

 

Control strategy Demand level
Off-peak Peak Highly-congested
ATMS: traffic metering - 36 - 32 - 56
ATIS: traffic diversion - 57* - 52* - 61
ATIS/ATMS: traffic - 43 - 4o - 75*

 

b. Difference in congestion duration from the no control change scenarios (percent)
(IO-minute both-lane closure incident)

 

 

 

diversion with signal
timing modification

 

 

 

 

Control strategy Demand 19“]
Off-peak Peak Highly-congested
ATMS: traffic metering + 0 + 8 - 21
ATIS: traffic diversion - 23" - 32" - 49
ATIS/ATMS: traffic - 23* - 32" - 55"

 

 

Note: 1. the negative signs indicates the improvement over the no control change
2. the ‘ means this control strategy produces the greatest improvement in that

demand situation.

 

96

Table 6.3 Comparison of alternate diversion plans
lO-minute both-lane closure incident

 

Measureof
effectiveness

Diversion plan

 

Equally distributed at the
nearest upstream

Optimal distribution at
the nearest upstream

Equally distributed at
two

 

 

intersection intersection upstream intersections
Difference in total
travel time from - 52 - 57 - 53
the no control
change
Difference in
- 32 - 32 - 24

congestion
duration from the
no control change

 

 

 

 

Table 6.4 Comparison of alternate signal control policies for signal modification
lO-minute both-lane closure incident

 

Measure of
effectiveness

Control variables for signal timing modification

 

Degree of saturation

Maximum queue lengtH

Stop time

 

Difference in total
travel time from
the no control
change

Difference in
congestion
duration from the
no control change

 

-40

-32

-45

-20

 

 

-58

-32

 

 

 

 

97

control strategy under different network circumstances, in addition to the different volume
and incident conditions.

The control strategies should be incorporated into a complete simulation analysis
tool. One of the limitations of this research is the lack of simulation tools which integrates
control strategies. Although the NETSIM simulation program was used in this study, it
lacks of capability to incorporate an adaptive signal logic as well as dynamic traffic
routing. The development of these features will be a valuable contribution to the
extension of these experiments on control strategies.

Since the study was designed to test certain selected control strategies for a wide
range of incident conditions and one incident situation was selected for the sensitivity
analysis, the results from the sensitivity analysis demonstrate the effectivenesses of various
control strategies only in that situation. A full factorial analysis of these control strategies
may be only way to reveal their effectivenesses in all range of traffic and incident
conditions. It is recommended that the full factorial design of the experiment should be
conducted to determine the most appropriate control strategy for a particular incident and

traffic condition.

BIBLIOGPAHY

BIBLIOGRAPHY

Al-Deek, H. and A. Kanafani (1993) Modeling the benefits of advanced traveler
information systems in corridors with incidents. Transportation Research C , Vol
1, No 4. pp 303-324.

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

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

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

Beaubien, R.F. (1994) Incident management notes. Traffic engineering enforcement
committee. Hubbell, Roth & Clarks Inc.

Belcher, PL. (1990) The effects of traffic management and control on vehicle capacity in
London. Third International Conference on Road Traffic Control, Institute of
Electrical Engineers, Conference Publication No 320, pp 67-71.

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

Bretherton, RB. et. a1. (1986) The use of SCOOT for traffic management. Second
International Conference on Road Traffic Control, Institute of Electrical
Engineers, Conference Publication No 260, pp 81-84.

Bretherton, RD. and G.T. Bowen (1990) Recent enhancement to SCOOT - SCOOT
version 2.4. Third International Conference on Road Traflic Control, Institute of
Electrical Engineers, Conference Publication No 320, pp 95-98.

Castling, I. (1994) Advanced technology for road transport: IVHS and A77. Artech
House, Boston.

98

99

Church, R. and C. Revelle (1978) Modelling an oversaturated intersection. Transportation
Research A, Vol 12, pp 185-189.

Cohen, S. (1986) Traffic control strategies during peak hours: The case of Aix-En-
Provence. Second International Conference on Road Trafi‘ic Control, Institute of
Electrical Engineers, Conference Publication No 260, pp 133-136.

Cracknell, I.A. et. al. (1975) Physical restraint: Greater London’s cordon restraint study.
Trafiic Engineering and Control, September, pp 368-372.

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

Drenick, R.F., S.A. Ahmed, and W.R. McShane (1981) Decentralized control of
congested street networks. Transportation Research Record 795, National

Academy of Sciences, Washington, DC, pp 24-27.

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

Dudek, CL. (1976) Better management of traffic incidents: Scope of the problem. ibid.

Dudek, CL. (1988) Solutions to problems generated by incidents and special events.
Strategies to alleviate traffic congestion, Institute of Transportation Engineers,
Washington, DC, pp 24-31.

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

Federal Highway Administration ( 1995) TRAP user reference guide. Office of Safety and
Traffic Operations, March.

Gartner, NH. and D.L. Hou ( 1992) Comparative evaluation of alternative traffic control

strategies. Transportation Research Record I 360, Transportation Research
Board, Washington, DC, pp 66-73.

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

Gerlough, D.L. (1965) Simulation as a tool in traffic control system evaluation. Trafiic
control: Theory and instrumentation, Instrument Society of America, Plenum
Press, NY, pp 71.

100

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

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

Gray, B.M. and L. Ibbetson (1991) Operational traffic control strategies for congestion.
Trafiic Engineering and Control, February, pp 60-66.

Hall, R.W. ( 1993) Non-recurrent congestion: How big is the problem?, Are traveler
information systems the solution?. Transportation Research C, Vol 1, No 1, pp
89-103.

Hobeika, A.G., et. al. (1993) Real-time traffic diversion model: Conceptual approach.
Journal of Transportation Engineering, Vol 119, No 4, pp 515—534.

Holmes, G.N., and DR. Leonard (1993) “The frequency and importance of incidents
which cause congestion in urban areas”, Transport Research Laboratory,
Contractor Report 342.

Hunt, RB. and TR. Holland (1985) The effects of an incident in SCOOT system - in
simulation and on-street. Trafi‘ic Engineering and Control, February, pp 55-58.

Institute of Transportation Engineers (1989) Toolbox for alleviating trafi‘ic congestion.
Washington, DC.

Kagolanu, K. (1994) A comparative study of traffic control system IT E Compendium of
Technical Papers, Institute of Transportation Engineers, pp 374-378.

Kay, J.L., J.C. Allen, and J.M. Bruggeman (1975) Evaluation of first generation
UT CS/BPS control strategy. Report FI-IWA-RD-75-26, Federal Highway
Administration, March.

Lindley, J .A. (1987) Urban freeway Congestion: Quantification of the problem and
effectiveness of potential solutions. ITE Journal, January, pp 27-32.

Lindley, J .A. (1989) Urban freeway congestion problems and solutions: An update. IT E
Journal, December. pp 21-23.

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

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

101

Lowrie, PR. (1982) Sydney co-ordinated adaptive traffic control system-Principle,
methodology, algorithm. International Conference on Road Traffic Signaling,
Institute of Electrical Engineers, Conference Publication N o 207, pp 67-70.

Lowrie, P.R. ( 1990) SCA TS - Sydney co-ordinated adaptive trafi‘ic system: A traflic
responsive method of controlling urban trafl‘ic. Roads and Traffic Authority,
Sydney.

Luedtke, RM. and J.T. Short (1994) User benefit from an arterial incident detection and
response system. IT E Compendium of Technical Papers, pp 523-527.

Maher, MJ. and R. Akcelik (1975) The re-distributional effects of an area traffic control
policy. Traffic Engineering and Control, September, pp 383-385.

May, AD. and D. Westland (1979) Transportation system management TSM-type
projects in six selected European countries. Traffic Engineering and Control,
February, pp18-20.

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

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

McShane, W.R. and LJ. Pignataro (1978) Some results on guidelines for treatment of
traffic congestion on street networks, ibid., pp 19-25.

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

Michalopoulos, PG. and G. StephanOpoulos ( 1977a) Oversaturated signal systems with
queue length constraints-I: Single intersection. Transportation Research A, Vol
11, pp 413-421.

Michalopoulos, PG. and G. Stephanopoulos (1977b) Oversaturated signal systems with
queue length constraints-II: Systems of intersections. Transportation Research A,
Vol 11, pp 423-428.

Michigan Department of Transportation (1992) Metropolitan Detroit Advanced Trafl‘ic
Management System (A TMS). Metropolitan Transportation Center, Traffic&Safety
Division.

102

Miner, RP. (1990) After the incident management: Reducing the impact of traffic
incidents through planned and coordinated response. Technical Papers of ITE ’s
1990, 1989, and 1988 Conferences, Institute of Transportation Engineers,
Washington, DC, pp 78-89.

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

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

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

 

 

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

. (1978) Trafiic control in oversaturated street networks. Report No 194,
Transportation Research Board, Washington, DC.

 

. (1991) Freeway Corridor Management, Synthesis of Highway Practice
1 77, Transportation Research Board, Washington, DC.

 

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

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

OECD (1981) Trafl‘ic control in saturated conditions. Road Research, Paris, France.

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

Pigantaro, LJ. (1974) Oversaturation terms and measurements seem adequate after field
validation. Traflic Engineering, Vol 44, No 9, pp 20-24.

Pooran, FJ. and H.C. Lieu (1994) Implementation of operating strategies for integrated
freeway/arterial control systems. IT E Compendium of Technical Papers, pp 364-
368.

Quinn, DJ. (1992) A review of queue management strategies. Traflic Engineering and
Control, November, pp 600-605.

103

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

Rathi, AK. (1991) Traffic metering: An effectiveness study. Transportation Quarterly,
July, pp 421-440.

Rathi, AK. and EB. Lieberman (1989) Effectiveness of traffic restraint for a congested
urban network: A simulation study. Transportation Research Record 1232,
Transportation Research Board, pp 95-102.

Rathi, A.K. and AJ. Santiago (1990) The new NET SIM simulation program. Trafi‘ic
Engineering and Control, pp 317-320.

Rathi, AK. (1991) Traffic metering: An effectiveness study. Transportation Quarterly,
Vol 45, No 3. PP 421-440.

Reid, DA. and MC. Bell (1990) Junction design and congested route choice. Third
International Conference on Road Traffic Control, Institute of Electrical
Engineers, Conference Publication No 320, pp 81-85.

Roberg, P. (1995) A distributed strategy for eliminating incident-based traffic jams from
urban networks. Traflic Engineering and Control, June, pp 348-354.

Roberg, P. and CR. Abbess (1994) Diagnosis and treatment of congestion problems in
central urban areas. Presented at EURO l3/OR Conference, Strathclyde.

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

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

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

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

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

104

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

Shepherd, SP. (1992) A review of trafiic signal control. Institute of Transport Studies,
the University of Leeds, January.

Taylor, I.G., M.C. Bell, and GM. Geary (1987) Queue volume: A measure of congestion.
Trafi‘ic Engineering and Control, November, pp 582-585.

Transportation Research Board (1984) Development in trafiic signal system.
Transportation Research Circular 282, Nation. -. escarch Council, Washington,
DC.

Transportation Research Board (1987) Management of freeway emergencies and special
events. Transportation Research Circular 326, National Research Council,
Washington, DC.

Transportation Research Board (1989) Traflic management for freeway emergencies and
special events. Transportation Research Circular 344, National Research Council,
Washington, DC.

Van Aerde, M. and H. Rakha (1989) Development and potential of system optimized
route guidance strategies. VNIS’89, IEEE-VTS-V399, pp 304-309.

Van Aerde, M.W. and S. Yagar (1990) Combining traffic management and driver
information in integrated traffic networks. Third International Conference on
Road Trafl‘ic Control, Institute of Electrical Engineers, Conference Publication No
320, pp 11-15.

Van Vuren, T. and D. Leonard (1994) Urban congestion caused by incidents. Traffic
Engineering and Control, Vol 35, No 7/8.

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

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

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

 

 

 

RRIES

"illlllllllllllil