You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/312383633

Adaptive Traffic Control System Using Dynamic Slot Allocation To Achieve


Minimal Congestion In A Stand Alone Road Junction

Conference Paper · January 2017

CITATIONS READS

0 211

1 author:

Md Mohsinur Rahman Adnan


Shahjalal University of Science and Technology
7 PUBLICATIONS   2 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Adaptive Traffic Signal Control View project

Biomedical Instrumentation and Signal Processing View project

All content following this page was uploaded by Md Mohsinur Rahman Adnan on 16 January 2017.

The user has requested enhancement of the downloaded file.


ICERIE_ MIE_XYZ

Proceedings of the
International Conference on Engineering Research, Innovation and Education 2017
ICERIE 2017, 13 ̶ 15 January, SUST, Sylhet, Bangladesh

Adaptive Traffic Control System Using Dynamic Slot Allocation To


Achieve Minimal Congestion In A Stand Alone Road Junction

Md. Mohsinur Rahman Adnan1,*, Zunaid Omair1 , Raiyan Abdul Baten1 , Shouvik Musavvir1,
Nur Hossain Khan1, Risalat Ahmed1

1
Department of Electrical and Electronic Engineering,
Bangladesh University of Engineering and Technology, Dhaka-1000, Bangladesh

Keywords: Abstract: The optimization of city traffic scenario is an important issue to


 Traffic Jam; be considered. Hence modern techniques of traffic management is necessary
 Image Processing; for improvement of traffic deadlock problem. A new control of traffic lights of a
 Arduino; simple intersection, taking into account vehicle behavior has been proposed in this
 Vehicle Control; paper. Here, the control of traffic light signal is performed taking into account the
 Area Percentage; number of vehicles and the area covered by them in each lane of a four way
 Road Junction; junction. The intersection is considered as a resource shared between vehicles of
two roads. Implementation is done to have efficient and safe traffic flow control
along with reduced waiting time at signal junction; giving priority for
emergency cases and for heavy traffic build ups. There is also provision for
communication with central traffic control authority instantaneously. Image
processing adds an extra feature to the reported work. On one hand, this allows to
distinguish different kinds of vehicles depending on their size and position
variation. On the other hand, since the main objective is to minimize the
evacuation time from a deadlock, the waiting time of each lane's vehicles is
dramatically reduced. This work can easily be extended to incorporate mutual
communication between road junctions to improve accuracy and robustness in
decision making by treating the city traffic system as a whole.

1. INTRODUCTION

Traffic Congestion in motorways is becoming an ever more pressing problem in countries all over the world. One
approach to tackle this problem could be construction of new roads to enlarge the capacity of existing traffic
infrastructure. This approach is very costly and is often not possible to due to environmental or societal constraints.
In addition this is a long term approach. So there is a need of more readily available solutions. These solutions may
be controlling traffic in such ways that congestion is solved, reduced or at least postponed. One such solution is the
use of traffic lights at probable congestion positions such as intersections and junctions.

Traffic lights are signaling devices that are used for controlling movements of vehicles at road junctions. Traffic
light controllers are designed and programmed to assign time to certain directional movements using red, green and
yellow lights. As shown by Ganiyu et. al. (2014), present traffic light controllers use microcontrollers to do so.

* Corresponding author: : adnanmohsin27@gmail.com


These controllers have limitations as they use predefined hardware, which functions according to a program that
does not have flexibility of modification on real time basis. As the numbers of vehicles on road increases, because
resources provided by current infrastructures are limited, proper control of traffic becomes very important. To
manage traffic flow, a new technique called Dynamic Traffic Signal Controller emerged, as was reported by Sayyed
et. al. (2014). This model uses the idea of central control depending on real-time feed from the junctions by
adjusting the times of lights. Thus, optimal switching of traffic lights will control traffic flow with respect to road
capacity and prevent congestions.

Operators can control traffic from control centers by monitoring the traffic situation based on video images and
other measurements from loop detectors. Especially during rush hours, the interpretation by experts of obtained
information about traffic intensities, average velocities, weather conditions, incidents on the motorways can lead to
advice for the travelers and/or actions in order to solve, to prevent or to postpone congestion, as was reported by T.
Bellemans et. al. (2002). Adaptive traffic model, which has a cluster of variables, allows for a simulation of future
traffic densities and average velocities. These predictions can help traffic operators to determine which of the control
measures to take. This work is an attempt to produce such an adaptive decision maker and implement it in a small
scale model which will observe the traffic condition and automatically allocate sufficient green or red time to
different directions such that congestion can be reduced.

2. CURRENT DRAWBACKS

The drawbacks of traditional traffic light controllers are highlighted below:-

I. As was cited by Gong et. al. (2011), under ordinary condition when the traffic lane waits for green light, the
time setting is always the same and fixed. The increasing number of vehicles on road substantially
increases waiting time for vehicles on road and results in heavy congestion.

II. At some junctions even if there is no traffic the light remains red. The road users have to wait till the light
turn green. This results in an increase of waiting time for vehicles.

III. During traffic jam an emergency vehicle like ambulance, fire brigade car or police car may get stuck and
may have to wait for the light to turn green. No priority scheme is available in traditional systems, as was
reported in the work of Farooq et. al. (2009).

3. GOALS OF THIS WORK

I. To determine the capacities of the approaches at the intersection.

II. To determine the durations of green, red and yellow lights (signal timing) at the intersection.

III. To adaptively control the durations with change in congestion such that congestion is minimized.

4. DECISION MAKING ALGORITHM

The primary objective of this work is to design a user interfaced program and implement hardware for dynamic
traffic light control system that is suitable for real life implementations. The work implementation also aims to have
efficient and safe traffic flow control along with reduced waiting time at signal junction, priority for emergency case
i.e. heavy traffic jams and provision of information to traffic control authority instantly. The traffic jam will be
reduced by increasing the green signal time on busy road and decreasing this time on non-busy road.
3 | Adnan et. al., I C E R I E 2 0 1 7

To do so, a maximum time for a full iteration of the four approaches of the intersection was preset. This maximum
value of time at first is ratio-ed for the four incoming lanes, depending on the congestion they are presenting, this is
the primary allocated time. Then the controller went on checking the current condition via snapshots of the lanes at
every millisecond gap, with the criteria of change in control decisions as imposed by congestion situations in the
incoming and outgoing lanes. An outgoing lane may get blocked up by traffic, or an incoming one may get empty
within its allocated time. If there is fewer cars in one approaching lane, the controller will get the cars to pass before
the allocated time is up and thus will be able to save the extra time which will then be allocated for the next set of
four way iteration. If for an approaching lane, its outgoing options (front/right) are blocked up, the controller will
adjust the allocated time to zero by stopping cars from going in that direction as to reduce the pressure in that
outgoing lane. In common literature, the problem of adaptive control in general is dealt with setting a minimal time,
and going on checking if the path has cleared out. If it has not cleared, additional time is provided for the lane. But
this model has its problem in the assumption that the path will clear out if some more additional time provided,
which may not be the case in a high demand root or at peak hours. Rather, it is a good idea to set a maximal time,
which may be adjusted as the situation demands. That is why this assumption is used in working principle of the
controller.

Fig. 1: Right and Front Turn is Allowed

The road ways are defined as in fig.1. The entire work flow and decision making is as seen in the pseudo code in
fig.2.

5. IMAGE PROCESSING TECHNIQUES

First, an array is defined that indicates the webcams that hold the details of the ways; in the serial of the first way,
then the second and so on. Then a light adjustment loop is run which delays processing till the cameras are fully
functional and ready for taking quality image. When the cameras are up and running, the snapshot of background
image without any car is taken as a standard. This image is used in further processing including resizing, RGB to
Gray conversion and saving, as was done by Choudekar et. al. (2011). This ends the first part called Calibration.
Calibration is done after every few runs or whenever there is a restart of the whole system to ensure the webcams are
feeding correct way’s image properly and they have not slid away from their actual position.
Fig. 2: Pseudo Code for Decision Making Fig. 3: Flow Chart for Image Processing

In the image processing part, each webcam takes input snapshot of its way and do resizing, RGB to Gray conversion
and saving. Then it subtracts the way’s initial empty standard from the input image to determine changes, increases
contrast to make the blurs and shadows disappear and finally fills holes in the resultant image. Then the image is
converted to binary and connected areas are found out with their left top point’s x, y coordinate and x length and y
length in PIXEL. These values are given in a matrix with a row for each connected area and the values in columns.
From this matrix further processing is done to find out number of cars and percentage congestion in each way,
which is trivial mathematical calculation.

The divider of two lanes in a road is defined to be the connected area that is at center or closest to center of the input
image, with a length more than three times the width. This was not implemented previously in work of Abbas et. al.
(2013) and is an improvement over their work. Thus the divider is found and from the x-y coordinates of pixel
values incoming and outgoing cars are identified. Also, Skewness of the webcams are compensated for by ratio
conversion from measurements, taking the area read at the closest point to be four times that of the furthest point and
normalizing all the areas with the equation :

Where, = mid height of the area, = height of the resized image. The total procedure is shown in a flow chart in
fig.3.
5 | Adnan et. al., I C E R I E 2 0 1 7

6. HARDWARE IMPLEMENTATION

The heart of the system is an Arduino Uno (ATMEGA62) coupled with a computer. The computer, using image
processing defines the congestion in the roads and does the decision making, while the Arduino implements the
change of lights in order. The whole system is shown in fig. 4.

Fig.4: Hardware Implementation.

The webcams capture snapshots of the roads, and sends them to central server computer where image processing is
done to determine the congestion depending on PIXEL values that indicate externally connected areas in a given
background, which obviously are the cars. This count of cars and the percentage area that they take up in an
otherwise empty road is processed and a 4 X 2 matrix; containing the data of congestion for incoming lanes in first
column and for outgoing lanes in second column; is provided to a second program. The code takes these information
and with the just described algorithm makes decision of allocating green times to each way. The allocated times are
saved into a matrix and sent to Arduino using serial communication. Then Arduino implements the made decisions
by controlling the traffic lights. The Arduino keeps on searching for updated time values in the input matrix from
central computer, which the computer calculates using the webcam snapshots. Car arrivals and departures are
recorded by webcams continuously. Found new data are re-fed to computer. The whole process goes on in an
interconnected loop, with the delays happening mostly in the processing of the images taken.

7. EXAMPLE CASES

Some example cases, as a proof of the effectiveness of the algorithm, are described with respect to figure 1 as
below-
I. The most general case, with all the incomings and outgoings having cars in a way that did not cause
any blockade in outgoing or emptiness in incoming with time. The algorithm starts from road 1 and
goes on allocating green time as calculated in the primary ratio system, till all 4 roads get their
allocated times.
II. Incoming of road 1 is empty, in this case the traffic system did not give any green time to road 1, rather
skipped it and went to serve the next road.
III. Road 1 has cars, but road 2 (or road 3) is blocked. The road 1 had its allocated time, but only the Green
for road 3 (or Green for road 2) was shown, telling the cars to which way it could go and to which it
could not.
IV. Road 1 has cars, but road 2 and 3 both were blocked. The traffic system did not give any green time to
road 1, rather skipped it and went to serve the next road.
V. A dynamic case, with the number of cars changing during the green time of road 3. The result was that
road 3’s green signals were offed as soon as data of its emptiness reached the Arduino.

8. FUTURE WORKS

The positional or size variation of the vehicles are taken into account by accumulating their percentage of
normalized area to that of path’s, rather than actually counting their numbers. This normalization is done here as an
improvement over the work done by Farooq et. al. (2009). The image processing can detect size and placements of
vehicles in an approach, but due to the complexity of code and increase in decision making delay that would result if
the numbers were to be integrated in algorithm and implemented by Arduino, it was left out of this work. This gives
a scope for future improvement. Moreover the image processing code is implemented with the mindset of presenting
a demonstration in laboratory, which can be made more efficient if real world implementations are to be done. For
example, the image processing algorithm is solely dependent on the snapshots of the roads and works by counting
the connected areas from the image’s lighter and darker parts, which could pose a problem with real world’s shadow
effects. A solution to this problem would be to do precision analysis of imagery which would be possible with
higher processing rated computing unit. This work can easily be extended to incorporate mutual communication
between road junctions to improve accuracy and robustness in decision making by treating the city traffic system as
a whole.

10. CONCLUSION

The optimization of city traffic scenario is an important issue to be considered. Hence modern techniques of traffic
management is implemented to solve the traffic problem. The Dynamic Traffic Signal Controller is introduced in
this work having specific functions along with hardware interface. The first part is designing of program which
consists of data collection and calculation of percentage area and therefore automatic evaluation of signal time. After
that the second part is the hardware implementation in a small model version of intersection. This system aims at
saving a large amount of waiting hours caused by traffic deadlocks, using dynamic decision making algorithm.

REFERENCES

Ganiyu R. A., Arulogun O. T., Okediran O. O., “Development Of A Microcontroller-Based Traffic Light System
For Road Intersection Control”, International Journal of Science and Technology Research, Volume 3, Issue 5,
May 2014

Shabnam Sayyed, Prajakta Date, Richa Gautam, Gayatri Bhandari, “Design of Dynamic Traffic Signal Control
System”, International Journal of Engineering Research and Technology, (IJERT) Vol. 3 Issue 1, January 2014

T. Bellemans, B. De Schutter, B. De Moor, “Models for traffic control”, Journal A, vol. 43, no.3–4, pp.13-22, 2002
7 | Adnan et. al., I C E R I E 2 0 1 7

Yue Jinag Gong and Jun Zhang, “Real Time Traffic Signal Control for Modern Roundabouts by Using Particle
Swarm Optimization-Based Fuzzy Controller”, Technical Report, SYSU, March 2011

Umar Farooq, Hafiz Muhammad Atiq, Muhammad Usman Asad, Asim Iqbal and Zeeshan Azmat, “Design and
Development of an Image Processing Based Adaptive Traffic Control System with GSM Interface”, Second
International Conference on Machine Vision, pp. 166-170, December 2009

Pallavi Choudekar, Sayanti Banarjee and M.K.Muju,, “Real Time Traffic Light Control Using Image Processing”,
Indian Journal of Computer Science and Engineering (IJCSE), Volume 2, Issue 1, pp. 6-10, February 2011

Naeem Abbas, Muhammad Tayab and M. tahir Qadri, “Real Time Traffic Density Count Using Image Processing”,
International Journal of Computer Applications (0975-8887), Volume 83–No 9, December 2013

Umar Farooq, Hafiz Muhammad Atiq, Muhammad Usman Asad, Asim Iqbal and Zeeshan Azmat, “Design and
Development of an Image Processing Based Adaptive Traffic Control System with GSM Interface”, Second
International Conference on Machine Vision, pp. 166-170, December 2009

View publication stats

You might also like