Global dunyoda ilm-fan va ta‘limdagi innovatsion
rivojlanishning zamonaviy trendlari 15 dekabr, 2022 yil.
282
IMPROVING THE
TRANSPORTATION
SYSTEM WITH
AI-BASED
TECHNOLOGIES
TURSUNOV JAVLON,
Master’s student, Tashkent University
of Information Technologies named
after Muhammad al-Khwarizmi
javlontatu96@mail.ru
MEMONOVA GULRUKH,
Assistant teacher, Karshi branch of
Tashkent University of Information
Technologies named after
Muhammad al-Khwarizmi
memonovagulrux@gmail.com
BOBOKULOV SHAKHZOD
Master’s student, Karshi branch of
Tashkent University of Information
Technologies named after
Muhammad al-Khwarizmi
bobokulovshakhzod200@gmail.com
Abstract:
As with many other fields where the involvement of intelligent systems with
the help of Artificial Intelligence is prevalent, transportation system cannot be ignored
when it comes to incorporating AI and automation in order to improve the current trans-
portation system by using new advanced technologies. There have been a lot of new tech-
nologies have been implemented in transportation systems especially in public transpor-
tation, making it convenient to use for commuters and other users. Having said that, de-
spite the fact that new technologies have been used to tackle some of the existing issues,
there are some problems that are still persisting. For instance, traffic jams can be seen
even in developed cities around the world, making specialists think deeply and push the
boundaries of technology to come up with innovative solutions to cure existing hurdles in
transportation. In terms of easing the heavy traffic flow, intelligent traffic lights powered
with AI have their own role to play and huge progress can be seen in this area of research.
Some of the technologies that are used in smart traffic lights are induction loops, micro-
wave radar, and video detection. What has been done in this work is quite different from
previous technologies and methods and proposes a new methodology to be used in traf-
fic lights and with help of this method, traffic jams can be reduced significantly. Precise-
ly, an algorithm based on a convolutional neural network is used to detect vehicles, and
depending on the traffic density determined by live video footage, traffic lights make a
smart decision about which road should be opened more while other another road should
be closed for less time. From the environmental and economic perspective, this technol-
ogy with the proposed methodology reduces greatly gasoline use by cars, thus reducing
carbon dioxide emissions and saving the time passengers waste when stuck in traffic jams.
Key words:
CNN; deep learning; supervised learning; traffic congestion; object detec-
tion; object localization; intelligent traffic lights.
https://doi.org/10.47689/STARS.university-pp282-287
STARS International University
283
INTRODUCTION
Conventional traffic lights control the traffic flow based on rules, but since a
number of cars is increasing on the road, they are proving inefficient and caus-
ing long waiting times with their reliability being under question. According to
available information, the EU estimates that traffic jams cause economic damage
totaling $100 billion per year for its member states [5]. Since carbon dioxide is
the major contributor in terms of greenhouse gases, reducing its emission can
benefit hugely with regard to achieving a green economy. There is some re-
search carried out to evaluate how much extra greenhouse gases are emitted
by each car when there are miles-long queues at the intersections in particular
cities [6]. Apart from carbon dioxide emission, traffic jams are causing the loss of
time when commuting, thus taking away a huge chunk of working hours. Along
with economic impact, long traffic jams and congestion can have a negative
impact on motorists who are caught up in the traffic and from the perspective
of psychology, aggressive behavior and stress are attributable to congestion
[12]. Since there are serious consequences caused by traffic jams, experts and
scholars are working actively on the frontline, with this issue becoming one of
the crucial research areas in the transportation field as well as shifting the focus
to Intelligent Traffic Systems (ITS) [7][4] which play a vitally important role in
terms of tackling the existing urban traffic problems. In order to provide smooth
traffic flow, intelligent traffic lights come into play, and to facilitate their deci-
sion-making, there are some ways to detect and measure the number of cars
waiting at the intersection. For instance, the number of passing or waiting cars
can be measured by magnetic loops placed under the tarmac which can detect
the car when there is motion around it. This works by the principle of induction
when the magnet moves around the coil, thus generating electricity [3]. Another
method is to use radars which are based on whether there is a movement or not
[2]. These are the prevalent means to detect moving vehicles on the road and
there will be more ways coming to detect cars.
LITERATURE REVIEW
There have been a significant number of studies to address the issue of traf-
fic jams, especially in urban areas. Since the world population is increasing, a
rise in urban life can be seen with each day passing which, in turn, is causing
the increasing number of cars with traffic jams being a prevalent issue of to-
day’s era [8]. There have been a large number of studies proposing to use of
smart traffic lights to regulate traffic. Currently, we are living in a world where
advanced technologies are being used to an extent we never imagined and the
most common ones are cameras, phones, and traffic control systems what we
are missing is merging and giving them a brain [9]. One of the easiest methods
is to make special signs for distinctive lanes depending on the weight of the
vehicles and by applying this strategy, heavy cars can be diverted in situations
where the traffic jams can be prevented [10]. The primary objective of devel-
oping technologies based on the Artificial Intelligence which is used in traffic
lights is that they have the capabilities to adapt to situations depending on the
density of cars on the roads, especially during rush hours. They play a huge role
Global dunyoda ilm-fan va ta‘limdagi innovatsion
rivojlanishning zamonaviy trendlari 15 dekabr, 2022 yil.
284
in terms of minimization of traffic congestion which is mainly under the spot-
light in today’s society [11].
Traffic lights can be more optimized with IoT being incorporated like Artificial
Intelligence and their use is increasing day by day. One of the applications of IoT
in terms of usage in smart traffic lights is in the form of communications between
traffic lights [13]. In real-time environments, an adaptation of systems should be
quick to traffic conditions. Apart from the algorithms which are able to control the
flow of the traffic, other sensors mounted on the lanes help such kind of technolo-
gies to make smart decisions which improve consequently the traffic flow further
[17][18] and for this reason, it is important to carry out research on smart traffic
lights. Generally, discussions in a large number of research have been made to
develop smart traffic lights [13-16]. Despite the sheer number of research, what
has been done still needs to be developed further to achieve reliable results, and
also further research is needed for both hardware and software which are indis-
pensable in terms of developing such a technology. Research that has been done
so far on smart traffic lights has caused the creation of traffic lights which is quite
distinguishable from conventional traffic lights in terms of adaptation and this fea-
ture may come in handy in particular traffic conditions. In carrying out research to
develop intelligent traffic lights, their settings can be based on the density of cars
on the road, convenience for emergency vehicles, and movement of pedestrians.
In the following section, what has been done is described in terms of this work and
its contribution to smart technologies.
METHODOLOGY
This work heavily relies on object detection which, in this case, are cars and
other vehicles to be determined on the road in order to count them by the written
program. These days, methods based on convolutional networks are being used
in object detection and localization. Object localization is the process of detecting
objects in the image or video and surrounding the object with bounding boxes
so that it can be distinguished from other objects easily [19]. By using CNN which
stands for the convolutional neural network, this neural network can be trained
to detect particular objects. This kind of neural network can be trained with su-
pervised learning. Supervised learning is a way of training the algorithm which
requires annotations along with data [21]. There are algorithms based on convolu-
tional neural networks which are widely used in object detection and one of them
is YOLO which stands for “You Only Look Once”. YOLO has better accuracy and
performance compared to other alternative algorithms, thus increasing its appli-
cations in tasks that involve detecting a particular object [20]. When traffic lights
are made smarter based on the visual assessment of the environment, high-reso-
lution cameras are employed to capture the live video which can be divided into
frames, and frames are treated as a still image. The same applies in this work as
well and YOLO performs better in still images. In this work, the YOLO pre-trained
model has been made use of that’s why when counting the number of cars, the
model has been imported into the programming environment. In the case of this
work, the still images have been used for prototype and some of images in the
dataset is shown below (Figure 1).
STARS International University
285
Figure 1. Image dataset.
All images were used to evaluate the performance of pre-trained model and it
performed quite well in terms of detecting cars in Figure 2 below, the car detec-
tion process is shown.
Figure 2. Object detection.
Since the YOLO algorithm uses the COCO dataset which was trained to detect
eighty most used objects, now in this case, it is detecting all objects included in the
dataset used for training the YOLO. Now what needs to be done is to only detect
cars and count them. All algorithms are implemented in the python programming
environment and of course, other programming environments can be used as well
as long as they are able to use the opencv library which is especially dedicated to
computer vision tasks. After successfully implementing the algorithm, car count-
ing can be seen in the following Figure 3.
Figure 3. Vehicle counting.
Global dunyoda ilm-fan va ta‘limdagi innovatsion
rivojlanishning zamonaviy trendlari 15 dekabr, 2022 yil.
286
Finally, after completing all steps, the developed model is ready to make decisions
based on car density on a specific road. For instance, if the number of cars is more
on one road compared to another road, the algorithm calculates the optimal time for
both sides of the road in terms of how long the road should be opened.
RESULT AND CONCLUSION
In the near future, algorithms that are capable of self-learning can make a huge
difference when combined with various sensors which help to perceive surround-
ings so that smart decisions can be made, with improved safety, less time loss, and
less carbon dioxide emission. Despite the fact that this proposed methodology
used to determine and make smart decisions based on traffic density detection at
the intersections is still in its infancy, it can be further developed to make signifi-
cant progress in this field as well as serve as a reference for those who are work-
ing in the same field to overcome hurdles in transportation systems. The research
results of this paper have important practical significance in solving the traffic
congestion problem and reducing the waiting time of people at the intersection
of traffic lights. Besides, the upcoming steps of this research aim to incorporate
camera synchronization which will be built on what has been done so far, hence
improving the accuracy which has huge importance in terms of making the right
decision. In addition to eliminating the shortcomings of a proposed method such
as poor visibility in foggy environments and lack of accuracy in miles-long traffic
jams, existing technologies like magnetic loops and radars will also be implement-
ed to improve the proposed system further, with reliability increasing in any kind
of situation.
References
1. Zhong-Qiu Zhao, Peng Zheng, Shou-Tao Xu, Xindong Wu, “Object Detection
with Deep Learning: A Review”, IEEE Transactions on Neural Networks and Learn-
ing Systems (Volume: 30, Issue: 11, November 2019).
2. David F, Jose T, Pablo A, Mateo B, “Vehicular Traffic Surveillance and Road
Lane Detection Using Radar Interferometry, IEEE Transactions on Vehicular Tech-
nology 61(3):959-970, March 2012.
3. Antonio M, José H, Víctor M, Gumersindo J, and Alexander A, “Traffic Control
Magnetic Loops Electric Characteristics Variation Due to the Passage of Vehicles
Over Them”, IEEE transactions on intelligent transportation systems, VOL. 18, NO.
6, JUNE 2017.
4. Hamidi H, Kamankesh A, “An approach to intelligent traffic management system
using a multi-agent system”, Int J Intell Transp Syst Res 2018; 16(2):112-24.
5. https://www.innovationnewsnetwork.com/smart-traffic-light-technolo-
gy-controlled-using-artificial-intelligence/17869/
6. Nathan David, Chinedu Duru, “Evaluating the Emission of CO2 at Traffic Inter-
sections with the Purpose of Reducing Emission Rate, Case Study: The University
of Nigeria, Nsukka”, International Journal of Environmental Science & Sustainable
Development, December 2018
STARS International University
287
7. Tang Y, Zhang C, Gu R, “Vehicle detection and recognition for intelligent traf-
fic surveillance system”, Multimed Tools Appl 2017; 76(4):5817-32.
8. Carley M and Christie I, “Managing Sustainable Development”, Routledge 2017.
9. H. Surden and M. Williams, “Technological Opacity, Predictability, and
Self-Driving Cars”, 2016, Cardozo L. Rev. 38. Hein Online: 121
10. Soh, A., L. Che, G. Rhung, and H. Md Sarkan. (2010). “MATLAB Simulation of
Fuzzy Traffic Controller for Multilane Isolated Intersection.” International Journal
on Computer Science and Engineering 2 (4): 924–33.
11. Srivastava, M. D., S. S. Prerna, S. Sharma, and U.Tyagi. (2012). “Smart Traffic
Control System Using PLC and SCADA.” International Journal of Innovative Re-
search in Science, Engineering and Technology 1 (2):169–72.
12. D. A. Hennessy and D. L. Wiesenthal, “Traffic Congestion, Driver Stress, and
Driver Aggression,” Aggresive Behaviour vol. 25, pp. 409–423, 1999, DOI: 10.1002/
(SICI) 1098-2337(1999)25:6<409: AID-AB2>3.0.CO; 2-0.
13. O. Avatefipour, S. Member, F. Sadry, and S. Member, “Traffic Management
System Using IoT Technology – A Comparative Review,” 2018 IEEE Int. Conf. Elec-
tro/Information Technol, 2018, pp. 1041–1047, doi: 10.1109/EIT.2018.8500246.
14. M. Kabrane, S. Krit, and L. El Maimouni, “Smart Cities: Study and Comparison
of Traffic Light Optimization in Modern Urban Areas Using Artificial Intelligence,”
Int. J. Adv. Res. Comput. Sci. Softw. Eng., vol. 8, no. 2, pp. 1018, 2018, doi:10.23956/
ijarcsse.v8i2.570.
15. R. Hawi, G. Okeyo, and M. Kimwele, “Techniques for Smart Traffic Control:
An In-depth Review,” Int. J. Comput. Appl. Technol. Res., vol. 4, no. 7, pp. 566–573,
2015, doi:10.7753/IJCATR0407.1014.
16. A.H. Ahmed, “A Review of Adaptive Intelligent Traffic Control Systems,” J.
Res. Bus. Soc. Sci., vol. 1, no. 1, pp. 2209–7880, 2018.
17. Mustapha Kabrane, Salah-Ddine Krit, Lahoucine El Maimouni, and Jalal Laassiri.”
Control of Urban Traffic Using Low-Cost and Energy-Saving for Wireless Sensor Net-
work: Study and Simulation.” International Journal of Engineering Research and Man-
agement (IJERM) ISSN: 2349-2058, Volume-03, Issue-04, April 2016.
18. Mustapha Kabrane, Salah-Ddine Krit, Lahoucine El Maimouni, Jalal Laassi-
ri “Energy saving in wireless sensor networks: Urban traffic management appli-
cation” Journal of Theoretical and Applied Information Technology 15th January
2017. Vol.95. No.1.
19. Fatima Zahra Ouadiay, Hamza Bouftaih, El Houssine Bouyakhf, M. Majid Him-
mi, “Simultaneous Object Detection and Localization using Convolutional Neural
Networks”, ISCV 2018, 10.1109/ISACV.2018.8354045.
20. C.S. Asha, A.V. Narasimhadhan, “Vehicle Counting for Traffic Manage-
ment System using YOLO and Correlation Filter”, IEEE Xplore: 08 October 2018,
DOI: 10.1109/CONECCT.2018.8482380.
T. Hastie, R. Tibshirani, J. Friedman, “Overview of Supervised Learning”,
Springer Series in Statistics, 2009, chap-2, part-II.
