Page 1

LETTERS

International Journal of Recent Trends in Engineering, Vol 2, No. 2, November 2009

Time Optimization for Traffic Signal Control

Using Genetic Algorithm

4

Leena Singh1, Sudhanshu Tripathi2, Himakshi Arora3

1Amity School of Engg. & Tech., Lecturer in Computer Sc. & Engg. deptt., New Delhi, India

Email: leenasafya@gmail.com

2Amity School of Engg. & Tech., Lecturer in Instrumentation &Control Engg. deptt. , New Delhi, India

Email: sudhanshutripathi14@gmail.com

3Amity School of Engg. & Tech., Student of Computer Sc. & Engg. deptt., New Delhi, India

Email: himakshi.arora@gmail.com

Abstract— In this paper, a “real-time” traffic signal control

strategy is provided using genetic algorithms to provide near-

optimal traffic performance for intersections. Real-time

traffic signal control is an integral part of the urban traffic

control system and providing effective real-time traffic signal

control for a large complex traffic network is an extremely

challenging distributed control problem. The developed

“intelligent” system makes “real-time” decisions as to

whether to extend green time for a set of signals. The model

is developed using genetic algorithm implemented in

MATLAB. A traffic emulator is developed in JAVA to

represent dynamic traffic conditions. The emulator conducts

surveillance after fixed interval of time and sends the data to

genetic algorithm, which then provides optimum green time

extensions and optimizes signal timings in real time. The

optimization parameters are - total number of vehicles in a

road and importance of the road in the intersection. In the

end, by comparing the experimental result obtained by the

fixed time and real time based traffic systems which

improves significant performance for intersections, we

confirmed the efficiency of our intelligent real time based

control system.

Index Terms — Intelligent System, Traffic Emulator, Genetic

Algorithm.

I. INTRODUCTION

The increase in urbanization and traffic congestion creates

an urgent need to operate our transportation systems with

maximum efficiency. One of the most cost-effective

measures for dealing with this problem is traffic signal

control. Traffic Signal Control is a system for

synchronizing the timing of any number of traffic signals

in an area, with the aim of reducing stops and overall

vehicle delay or maximizing throughput. It provides

control, surveillance, and maintenance functions i.e.

control of traffic by adjusting and coordinating traffic

signals at intersections, surveillance by monitoring traffic

conditions with vehicle detectors and cameras; and

maintenance of equipment by monitoring for equipment

failures. These functions allow a traffic management

agency to service traffic demand, share traffic status with

other agencies and operate and maintain the traffic signal

control system. Traffic signal control varies in complexity,

from simple systems that use historical data to set fixed

timing plans, to adaptive signal control, which optimizes

timing plans for a network of signals according to traffic

conditions in real time [1].

Although traffic signal control has been studied for many

years, it remains an active research topic. A summary of

recent advancements is provided in [2].Kirschfink et al.

introduce intelligent models to catch as much as possible

from vehicle traffic features [3]. Papageorgiou et al. give

an overview of the main traffic control problems and their

approach methods [4]. Some studied the reserve capacity

of a road network under fixed time traffic control [5].

Hong & Lo [6] developed a methodology to analyze the

Phase Clearance Reliability (PCR) of a signalized

intersection and describe the performance of traffic signal.

Han & Zhang [7] proposed an approach to detect and

count vehicles at an intersection in real- time to increase

efficiency on traffic control.

In this paper, we developed emulator for representation of

traffic conditions at an isolated intersection with the

following silent features: Graphical User interface (GUI)

developed in JAVA, random generation of vehicles,

random vehicular direction, collision avoidance, and traffic

signals with fixed phase sequence, surveillance of traffic

conditions(stopped vehicles) at specified intervals, traffic

signals with minimum green length duration. Genetic

algorithm is used for traffic signal timing optimization.

Factors considered for genetic optimization are weights

allotted to each road (depending upon their usage and

traffic capacity etc), fixed maximum and minimum green

timings, fixed cycle timings and total stopped at each

incoming lane.

The Figure 1 shows the traffic flow behavior in the

network depends on control inputs that are directly related

to corresponding control devices i.e. traffic lights, variable

message signs, and disturbances etc. The function of the

control strategy module is to specify the control inputs in

real time based on available measurements (e.g. from loop

detectors or traffic cameras). Surveillance system provides

real time status to the control algorithm which decides

© 2009 ACADEMY PUBLISHER

Page 2

LETTERS

International Journal of Recent Trends in Engineering, Vol 2, No. 2, November 2009

5

Figure 1. Modeled diagram of traffic signal control.

Figure 2. Depicts the implemented algorithm of the Genetic Algorithm

control inputs. A human interface is required to monitor

the control strategy [8].

II. MODELING TRAFFIC EMULATOR

The traffic emulator is modeled to implement a fully

concurrent emulator of cars and traffic signal lights

interaction at an intersection. Traffic emulator consists of a

four legged isolated inter section with corresponding four

traffic lights for controlling straight and right turn traffic,

while the left turn is free. The car generation speed and car

speed can be changed as per desired.

A Collision Avoidance

Collision avoidance is implemented by the concept of locks.

A lock is a fixed sized space that can be occupied by a car. A

car can occupy a lock ahead of it only if it’s unoccupied and

the signal is green if it is the first car in the lane. Similarly,

while turning, before moving, a car must grab all lock objects

it needs, otherwise, it will be blocked. The traffic emulator

implements fixed cycle length and fixed phase sequence to

ensure that all the roads gets their turn and no road is

neglected for a very long time.The emulator conducts the

surveillance and sends the data to the control algorithm for

evaluation. Traffic light management at an intersection is an

extremely challenging and complex. Normal traffic behavior

even though seems pretty normal, is however extremely

difficult to predict & simulate in an artificial environment.

The number of different factors affecting the sequence and

duration of traffic light signals can be very wide. Several

assumptions had to be made, in order to reduce the overall

complexity. The various assumptions made are as follows:

(i) The intersection is assumed to be relatively “busy” and

under-saturated with significant demand variations in all the

approaches.

(ii) The intersection is assumed to be four-phased with a

phase for each approach.

(iii) The phase sequence does not change from cycle to

cycle.

(iv) The cycle time remains fixed.

(v) All cars are assumed to have same speed.

(vi) Cars can take a free left turn provided they do not

have a vehicle in front. There are no unnatural traffic

situations such as accidents, disruptions etc.

III. MODELING TRAFFIC CONTROL PROBLEM

The current Traffic Management system is designed

scientifically but usually fails to provide an optimum

throughput of vehicles through an intersection. Providing

effective real time traffic signal control for a large

complex traffic network is an extremely challenging

distributed control problem. We aim to develop an

efficient traffic adaptive control strategy that identifies the

real time traffic scenario in small steps (surveillance

interval), and gives appropriate green time extensions to

minimize a fitness function consisting of linear

combination of performance indexes of all the four lanes .

Fitness function, f = P.I.1 + P.I.2 + P.I.3 + P.I.4 The

Performance Index (P.I.) for each road depends upon

weight of the each road (i.e. capacity of the road and

priority of the road assumed same), the total number of

vehicles on the road given by S. S = S1 + S2 + S3 +S4

Performance Index (P.I.i) = Wi * Si /GTi i=1,2,3,4;

where Wi is weight allotted to road i respectively; Si is

number of vehicles at road i respectively; GTi is sum of

minimum green time (Gmin) and green extension time (g).

IV. PROPOSED SOLUTION TO THE PROBLEM

USING GENETIC ALGORITHM

A Genetic algorithm (or GA) is a search technique used in

computing to find true or approximate solutions to

optimization and search problems. MATLAB Genetic

algorithm application interfaces are used to implement the

algorithm. The Genetic algorithm is constrained with a

fixed cycle length of 70sec and green extension times (g)

with the bounds of 0 to 5 seconds. g1 + g2 + g3 + g4 – 10;

where (gi represents green extension time, i=1, 2, 3, 4) and

10 is total extension time of the entire signal.

Gmin = 15sec (Fixed green time for each road). G.T. = Gmin

+ x (Green time allotted to the road). The surveillance data

© 2009 ACADEMY PUBLISHER

Page 3

LETTERS

International Journal of Recent Trends in Engineering, Vol 2, No. 2, November 2009

from the emulator is sent to GA. They produce a set of

green time extensions, which minimizes the fitness

function, simultaneously satisfying the constraints. The

Figure 2 illustrates the process schematically as follows:

(1) An emulator is developed which shows dynamic

conditions of traffic on an isolated four-way intersection.

Code for surveillance is written which gives the total

number of cars on each road in the intersection. (2) After

each predefined surveillance interval, genetic algorithm is

executed with the input as total number of cars on each

road as determined by S. (3) The Genetic algorithm (GA)

executed as to obtain the best possible solution. The steps

of algorithm are as: (i) Generate the random population

(i.e. green time extension) which is selected by a random

function within the specified range (0-5). (ii) Evaluate

objective function (i.e. fitness function used). The fitness

function is to be minimized (i.e. it gives small values for

better generations) . (iii) Check, if termination criteria is

satisfied (i.e. either the predefined maximum number of

generation have reached or fitness function is not satisfied,

the performance is tested with both 100 and 6 generation.

A small improvement in case of 100 generation is

observed but to increase the speed of algorithm, 6

generations have used). (iv) If the termination criteria are

not satisfied; selection is performed from the given

population to obtain fitter parents, which can lead to fitter

sons. (v) These parents, thus selected are mated to produce

fitter children and this phenomenon is called crossover or

recombination. (vi) Some mutation is performed (i.e. some

bits of children are altered from the above result). This

emulates the real life as children may have some traits

different but the chances are generally kept very low. (vii)

After mutation we have a new set of generation, now go

back to step (ii). (viii) If the termination criteria in step

(iii) are satisfied, get the solution (i.e. the current

generation). (4) The result is received from GA is the

green extension times for all the four roads. These

extension times are added with predefined fixed green

times and applied to the emulator.

6

V. SIMULATION RESULT

In this section, we compare the results obtained by

proposed real time based system with traditional fixed

time system. Both the systems are tested on the setting of

fixed green signal time of 15 sec. , with a green extension

of up to 5 sec. in case of real time based system. The

comparison parameter considered is the total number of

exit vehicles at a fixed car generation speed and car speed

settings for fixed intervals of times. The results obtained

are as follows: for car generation speed of 200 ms and car

speed of 200 ms is given in Table I. The sample results

shown in Table I gives the output of 573 vehicles and 662

vehicles in the case of fixed time based system and real-

time based system respectively thus showing a significant

performance increase of 21.9 % in case of real time based

system.

CONCLUSION

In this paper, an “intelligent” isolated intersection control

system was developed. The developed intelligent” system

makes “real time” decisions as to whether to extend (and

how much) current green time. The system applications

appear to be very promising. The system shows significant

performance improvement compared to fixed time based

system within experimental limits (computation power,

random path selection, emulator settings) under the given

assumptions. The model developed is based on the genetic

algorithm, which optimizes traffic signal timings in real

time and provides a set of optimum green time extension

for all the four phases depending upon the surveyed traffic

conditions.

REFERENCES

[1] J. Lee, Baher Abdulhai, A. Shalaby, E-H Chung, “Real-

Time Optimization for Adaptive Traffic Signal Control Using

Genetic Algorithms” Journal of Intelligent Transportation

System, Volume 9, Issue 3, pp. 111 – 122, 2005.

[2] S. C. Wong and H. Lo, “Advanced algorithms in traffic signal

control: Editorial,” Journal Intelligent. Transportation System,

Volume 8, No. 2, pp. 61– 62, April– June, 2004.

[3] Kirschfink., H., Hernández, J., Boero, M. “Intelligent Traffic

Management Models”, in the Proc. of ESIT 2000, Aachen,

Germany, pp. 36-44, 2000.

[4] Papageorgiou, M., Diakaki, C., Dinopolou, V., Kotsialos, A.,

Wang, Y. “Review of Road Traffic Control Strategies”. In

Proceedings of the IEEE, Vol. 91, No.12, pp. 2043 -2067, 2003.

[5] H. Ceylan and M. G. H. Bell, “Reserve capacity for a road

network under optimized fixed time traffic signal control,”

Journal Intelligent Transporation System, Volume 8, no. 2, pp.

87– 99, April – June, 2004.

[6] Hong K. Lo,”A Reliability Framework For Traffic Signal

Control, IEEE Transactions On Intelligent Transportation

Systems, Volume 7, No. 2, pp: 250- 259, June 2006.

[7]Chong Han, Qinyu Zhang,” Real-Time Detection of Vehicles

for Advanced Traffic Signal Control”, International Conference

on Computer and Electrical Engineering, pp 245- 249, 2008.

[8] D. Srinivasan, M. C. Choy and R. L. Cheu, “Neural Networks

for Real-Time Traffic Signal Control”IEEE Trans. On Intelligent

Transportation Systems, Volume 7, No. 3, pp. 261-272, 2006.

TABLE I.

COMPARISON BETWEEN REAL TIME & FIXED TIME SYSTEM

Time

( in min.)

Out Traffic

(Fixed Time System)

Out Traffic

(Real -Time Based System)

1 30 37

2 75 73

3 120 141

4 158 181

5 190 230

Total: 573 Total: 662

© 2009 ACADEMY PUBLISHER