Sei sulla pagina 1di 4

Simulation of "Time-based" Versus "Sensor-based" Traffic light System

Abdul-Yasser Abd-Fatah Faieza Abdul Aziz


College of Science and Technology Dept. of Mechanical and Manufacturing Engineering
UTM International Campus, Universiti Putra Malaysia
Jalan Semarak, Serdang, Malaysia
Kuala Lumpur, Malaysia dr.faieza.upm@gmaiI.com
yasser@ic.utm.my

Norzima Zulkifli
Rosnah Mohd. Yusuff Dept. of Mechanical and Manufacturing Engineering
Dept. of Mechanical and Manufacturing Engineering Universiti Putra Malaysia
Universiti Putra Malaysia Serdang, Malaysia
Serdang, Malaysia norzima@eng.upm.edu.my
rosnah@eng.upm.edu.my

Ahstract- Simulation of time-based and sensor-based traffic but higher in installation and maintenance costs (Coleri et. al,
light control system at four-way intersection using Arena 2004). There has been a study of implementing infrared
simulation software was studied. Introduction of sensors in sensors in the two phase traffic light system with real model
traffic light system has been developed to reduce waiting time done in HUM, Malaysia (Albagul et. aI, 2006). Several
of vehicles at intersection. Simulation with Arena is done to simulations modeling of traffic light system has been
observe if the improvement of sensor-based system is conducted, such as fuzzy logic control system to mimic
significant over the current system ( fhed-time) and to evaluate
human intelligence in controlling traffic conditions of two
the effects of implementing sensor-based system in busy and
phase system (Tan, et. aI, 1996); CELLSIM traffic
normal road condition. Algorithm of traffic light sequence is
simulation that controls traffic based on calculation of spaces,
done first prior to development of simulation. Then, the
acceleration and movement of cars using cellular automata
simulation of traffic light system is developed for both normal
and busy traffic condition and the data obtained is analyzed
(CA) and car-following (CF) (Bham and Benekobal, 2001);
using statistical process control. From the result, average
comparison of FRESIM (Freeway simulation using
waiting time for sensor-based system was 62.5% less than time­ CORSIM) and Car-Following Steady State behavior (Rakha
based system for each vehicle in the normal traffic condition, and Crowthe, 2001); Fuzzy Signal Control (FUSICO) based
and 15% less for busy traffic condition. This shows significant on HUTSIM traffic signal controller that mimic human
improvement based on the paired T-test done. thinking in traffic control using computer algorithm
(Koskinen et. aI, 2007) and solving traffic problem using
Keywords- Traffic light control system, Arena, simulation Arena simulation on multiple intersections two-phase system
modeling to determine the best green light timing on certain traffic
condition (Yang and Wen, 2008).
I. INTRODUCTION The objective of this study is to simulate four-phase
traffic light system at four-way intersection using Arena
Fixed time control of traffic light system used in simulation software to see the difference and improvements,
Malaysia is not very efficient, especially at the roads where if any, between time-based traffic light system and sensor­
not all of the junctions are busy (Tan et aI., 1996). Therefore, based traffic light system in normal and busy traffic
all vehicles have to wait for their turn to move at the condition. Data obtained from simulation analysis were
junctions, even though there are no vehicles moving from the compared using paired T-test to see if implementation of
green-lighted junction. This can lead to impatient vehicles to sensors can help to minimize waiting time of vehicles at the
break the law and run over the red lights, which resulted in intersection.
serious consequences such as accidents or fines. However,
with the usage of the sensor-type traffic signal, and II. METHODS
application of some intelligent control for the sensor-type
The methods used in the study were as follows:
traffic light, the time for vehicles to wait in the traffic light
can be minimized. A. Programming Sequence of Traf
f ic
Several studies have been done on implementation of
The programming for sequence of traffic light can be
sensors in traffic light. There are two types of sensors done by understanding the criteria and characteristics of
implemented in the traffic light system, the intrusive traffic light system involved. For this study, a traffic light
(induction loop detectors, micro-loop detectors and system of four-way intersection involved (North, South, East
pneumatic road tube) and non-intrusive (infrared, video and West), with up to six phases of traffic lights;
image processing, microwave radar and laser) type. Intrusive • Phase 1 - East and West green, the rest red
type provides higher accuracy than the non-intrusive type,

978-1-61284-486-2/111$26.00 ©2011 IEEE

789
• Phase 2 East green, the rest red
- represented in first block of model as shown in Figure 1.
• Phase 3 North green, the rest red
-
Here the arrival rate of the vehicles is determined by the
• Phase 4 North and South green, the rest red
-
road condition of the intersection, whether it is busy or not.
• Phase 5 South green, the rest red
-
Therefore, the arrival rate of 1.7 seconds is used for a
• Phase 6 West green, the rest red
-
normal and less busy road condition, and the arrival rate of
However, phases that has been used in this study only 1 second is used for busier road condition, based on the
allow one junction to move, therefore, Phase 1 and Phase 4 study done by Yang and Wen (Yang and Wen, 2008).
is eliminated from the study to reduce the complexity and There is a little difference for simulation of traffic light
confusion of the system for developers and users. state that would differentiate between the time-based system
The configuration of traffic light system has been done as and the sensor-based system. The same simulation model of
follows: the state of traffic light is used with a little configuration
added for the inclusion of the sensor-based system. The
• The green light sequence is based on clockwise
movement, starting with east junction, followed by addition of modules in sensor-based system is to change the
north, west and south. duration of traffic light. If sensor detects three or less
vehicles in the system, it changes duration to two seconds, if
• For time-based system, the green light is always be
more vehicles detected, it changes to 20 seconds.
20 seconds for each junction.
• Sensor is used to scan the condition on the road, if C. Analysis ofData
there are three or less vehicles waiting in the There were two sets of data that has been analyzed: data
junction, indicating that system is not busy and from time-based system and the sensor-based system. These
allowing for two seconds of green light. data obtained from the reports generated by Arena after the
• If there are more vehicles waiting in junction, entire simulation fUll . They were analyzed to see if the
indicating that road is busy, 20 seconds of green improvement in waiting time of vehicle is succeeded or not.
light is allowed. The data is compared by using the paired T-test, a way to
Based on configuration above, algorithms have been compare two sets of data to see if significant improvement
developed for time-based and sensor-based system. With the has been made or not. The paired T-test of the data acquired
algorithms developed, the simulation model can be done. from the simulation is done in two tests: the paired T-test of
E. Simulation Modeling of Traffic Light System Using
the data from the simulation in normal traffic condition and
Arena the paired T-test of the data from the simulation of busy
traffic condition.
Models of time-based system and sensor-based system
As the simulation generates vehicles in random arrival
are constructed using Arena. The difference between both
time based on exponential distribution, average waiting time
systems are the timing of green lights, therefore the model
is different in different simulation run.
should be built to reduce the redundancy of the work. There
would be two main sections within the work frame of the
Arena software; the simulation of traffic junction (Figure 1)
and the simulation of traffic light state (Figure 2).
Simulation of the traffic junction represents the arrival of
vehicles into the system and the queue situation in each of
the junction, while the simulation of traffic light state
determining the state of the green lights for the intended
junction. Therefore, the framework of the simulation of
traffic condition is similar for both time-based traffic light
control system and sensor-based traffic light control system,
but for simulation of the traffic light state, it would be a
Figure 1. Model of the traffic junction situation
little bit different with the inclusion of different timing
because of the sensor function involved. Furthermore,
simulation is modeled based on the traffic condition, busy
and normal. The busy and normal road condition can be
manipulated by the parameters of the arrival entity;
therefore there is no change of the model made. To analyze
the performance of the time-based and sensor-based traffic
light system, the simulation has the run length of 200
seconds for each run and replicated to 25 runs for statistical
purpose.
The vehicles arriving into the system are generated in the
Create module called Car Arrives at Intersection which is

790
Figure 3. Traffic light simulation models for (a) time based system (b
sensor-based system)

All statistical data from simulation and paired T-test is


compiled in Table 1. From here, t-value is calculated at
10.087 for normal traffic condition and 4.5115 for busy
Figure 2. Simulation model of state of traffic light (a) for time-based (b)
traffic condition. With t-value as calculated and degree of
for sensor-based
freedom is 24 (n-l), the p-value for normal traffic condition
Therefore, the number of simulation run is the sample is 4.l495TIO, whereas the p-value for busy traffic condition
size of the experiment and needed to be determined so that is 1.438y4. As both p-value is less than 0.05, it is safe to
data can be accepted with certain level of confidence. Law assume that there are different values in both normal and
of large numbers and central limit theorem stated that for busy condition, which suggests there are improvements if
large number of sample size n, the mean of the outcomes of the sensor is implemented in the system.
an experiment, II, is approximately distributed as normal Result of the study from simulation modeling of the
distribution with true mean J.l and variance d1/n (Smaldone, traffic light control system with time-based system and the
2001). Therefore, by running the simulation several times, sensor-based system showed that there are a lot of
the expected value can be considered as true mean with improvements on waiting time of vehicles in the junctions if
some level of confidence. Number of runs, or sample size is the implementation of sensor is done to the traffic light
determined to be 25 for this experiment. For paired T-test, system. Mean waiting time of time-based traffic light
as data accumulated from simulation runs, the test is done system is around 20 seconds for both normal traffic
using Microsoft Excel and the p-value obtained from paired condition and busy traffic condition, while for sensor-based
T-test should be less than 0.05 to accept that the two sets of traffic light system, the mean waiting time for vehicles in
data is different enough, indicating that there are the junctions is about 7.5 seconds for normal traffic
improvements within the two sets of data tested. condition, and about 17 seconds for busy traffic conditions.

III. RESULTS AND DISCUSSIONS TABLE 1. STATISTICAL DATA FROM SIMULATION AND PAIRED T­
TEST
Figure 3 (a) and (b) show the simulation models for time­
DATA FROM NORMAL CONDITION Busy CONDITION
based and sensor-based during run. SIMULATION
Simulation has been run for 200 seconds for 25 run for AND Time- Time- Sensor-
each time-based and sensor-based system in normal and busy ANALYSIS Based Sensor-Based Based Based
..
condItIon. Data from the reports generated in Arena is mean Sample size 25 25 25 25
waiting compiled in Microsoft Excel before the paired T-test Mean 20.0139 7.5102 19.97 17.4334
is done using Microsoft Excel. Figure 4 shows the time (in Variance 3.4906 44.0932 2.2394 8.5886

seconds) of time-based system and sensor-based system in t-value 10.087 4.5114


yl 0
normal and busy condition. p-value 4. l l.44-04

791
junction and optimal traffic light duration is different from
one intersection to another. These parameters need to be
studied first to obtain the optimal solutions for the intended
traffic light system. However, the simulation done within
the study provided general solution for implementing
sensors in the traffic light control system.
Therefore, by implementing sensor in traffic light system,
the waiting time for vehicles in the junctions at the
intersection can be reduced significantly as has been proven
by statistical method. Further improvements can be made in
the system by modifying the parameters in the simulation
suitable for intended studies.

REFERENCES

[I] Albagul, A., Hrairi, M., Wabyudi, and Hidayatullah, M.F.. "Design
and Development of Sensor Based Traffic Light". Science
Figure 4. Mean waiting time of time-based and sensor-based system in
normal and busy traffic condition Publications 2006. American Journal of Applied Science, 3 (3), pp
1745-1749.
[2] Altiok, T. and Melamed, B. "Simulation Modeling and Analysis with
Within this study, the Arena simulation software, which
Arena". Elsevier, UK. 2007.
is more appropriate for simulation of manufacturing system,
[3] Bham, G.H. and Benekohal, R.F. "A high fidelity traffic simulation
is used. While there are a lot of specific tools or simulation model bas sed on cellular automata and car-following concepts."
software intended for traffic light studies available, the Transportation Research Part C, 1-32 Elsevier. 2004.
Arena software proves to be more general purpose [4] [Caristi, J. 1988. "The Simulation Language SIMAN on
simulation software that can be used in simulation of the Microcomputer and Mainframe". Journal of Applied Mathematics
and Simulation Vol I Number 3 1988.
traffic light control system. Therefore, in formulating
[5] Coleri, S., Cheung, and S.Y., Varaiya, P. "Sensor Networks For
optimal solutions for each of the traffic light in the Monitoring Control". Department of Electrical Engineering and
intersections, more appropriate tools can be used to study Computer Science, University of California, Berkeley. 2004.
the traffic light system and to improve the traffic condition [6] Furstenberg, K., Hipp, 1. and Liebram, A. 2003. "A Laser Scanner
at the intersection intended. However, some of the tools For Detailed Traffic Data Collection and Data Control". 7th World
Congress on Intelligent Transport System
used in studying the traffic light system are intended for
[7] Gonzalez-Calleros, lM., Martinez-Carballido, l Muoz-Arteaga, J.
specific purpose of the study or scarce and hard to get.
and Guerrero-Garcia, l, "An Iterative Method To Desigu Traffic
Hence a more general purpose tool such as Arena simulation Flow Models". Digital Society. 2009
software can be configured in order to get the optimal [8] [Kelton, W.D., Sadowski, R.P. and Sturrock, D.T. 2007. "Simulation
solutions for the intended studies. With Arena". McGraw Hill, New York. NY. 2004.
[9] Koskinen, K., Setala, N. and Kosonen, 1. "Fuzzy Sigual Control -
IV. CONCLUSION FUSICO". Laboratory of Transportation Engineering, Helsinki
University of Technology. 2007.
Traffic light control system involves a very complex
[10] Rakha, H. and Crowther, B.. "Comparison and calibration of
study. Even implementing one traffic light in a single FRESIM and INTEGRATION steady-state- car-following behavior".
junction involved a lot of studies to be done and there is no Transportation Research Part A 1-27 Elsevier Science Ltd. 2003
obvious optimal solution. In the study there are four [11] Smaldone, T. ''Traffic Light Simulation: A Study Into How Long It
junctions of an intersection involved, therefore, a more Takes Vehicles Stopped at Red Lights To Go and To Cross the Light
After It Turns Green". Modeling and Simulation Independent Study
complex work needs to be done as the state of one light
Course, Rutgers University, Camden, N.J. 2001.
influences the traffic flow and conditions of other junctions
[12] Tan, K.K., Khalid, M. and Yusof, R. "Intelligent Traffic Light
as well. Control by Fuzzy Logic". Malaysian Journal of Computer Science,
While the work is to see if the implementation of sensor Vol. 9 No. 2, December 1996, pp. 29-35
in the traffic light control system can improve the waiting [13] [Wiering, M., Veenen lV., Vreeken l and Koopman, A. "Intelligent
time of vehicles in the junction of an intersection, it does not Traffic Light Control". Intelligent System Group, Institute of
Information and Computer Science, Utrecht University, Utrecht, The
provide an optimal solution for the improvement of the Netherlands. 2004.
waiting time of vehicles in junctions as each parameter of [14] Yang,C.L. and Wen, W. "Solving the Traffic Problem Using a
the traffic light system such as arrival rate of vehicles at the Simulation Model". International Conference on Information
intersection, the average number of vehicles waiting in the Resource Proceedings (CONF-IRM 2008)

792

Potrebbero piacerti anche