A Review paper on Artificial Neural Network: Intelligent Traffic Management System

NIET Journal of Engineering & Technology (NIETJET)
Volume 10, Issue summer 2022 ISSN: 2229-5828 (Print)
27 | Page
Publisher: Noida Institute of Engineering & Technology,
19, Knowledge Park-II, Institutional Area, Greater Noida (UP), India.
Abstract—This paper provides a brief overview of the Intelligent Traffic Management System based on Artificial
Neural Networks (ANN). It is being utilized to enhance the present traffic management system and human resource
reliance. The most basic problem with the current traffic lights is their dependency on humans for their working.
The technologies used in the making of this automated traffic lights are Internet of Things, Machine Learning and
Artificial Intelligence. The basic steps used in Internet of Things are reported along with different ANN trainings.
This ANN model can be used for the minimization of traffic on roads and less waiting time at traffic lights. As a
result, we can make traffic lights more automated which in turn eventually deceases our dependency on human
resources
.
Keywords- ANN, Artificial Neural Networks, Machine Learning, CSS, Databases, Sensors, Internet of Things,
RBF Neural Network, ANN and Dijkstra, VANET (Vehicle Adhoc NETwork), ISN-IoT, V2V communication,
ITS (Intelligent Traffic System).
1. INTRODUCTION
This paper basically aims for the improvement in our traffic system which has mainly been dependent on human
resource only till now. But now as automation has come up in each and every industry around the world and as it
takes over the world with its more fascinating features, here we have decided to make things which play an
important role in a human’s daily life. Traffic lights. What does come to our mind when we think of these lights.
The only question that comes up in our mind is that why are there traffic jams all around the globe when we have
traffic lights and the main aim of these lights is to control the traffic. The main answer of this problem is the
dependency of our traffic management system on humans. We all know how traffic lights are being operated in
almost every country from the past 50 years. These traffic lights are human dependent and so they are to be operated
by a human. Another thing that makes the old-styled traffic lights less productive is the timer problem. The timer
is set to a specific time limit and it resets to that time even when there is no traffic.
Additionally, the traffic management in our country has gotten worse than excepted. According to research, the
capital of our country, New Delhi has an average length of 3 kms to 4.5 kms every day in their peak hours. This is
a concerning figure. But it is not the problem in our country. China, on an average has an average traffic jam of 2
kms to 3.4 kms every day.
The traffic system proposed here deals with this problem. The traffic system here has been designed in such a way
that it can adjust its timer with the help of four camera’s which have been linked to each other through Internet of
Things and Artificial Neural Networks. Alongside we have also linked the cameras and the LED lights to our
databases which will tell them how much the timer has to be displayed and when the lights can turn green and
adjust the timer according to the traffic conditions being captured in our cameras and through our micro
programmed chips and sensors. Also, the main problem of how to reduce traffic on the four ways is also discussed
A Review paper on Artificial Neural
Network: Intelligent Traffic Management
System
Abhinav Sharma1
, Anushka Trivedi1
, Akshita1
, Dr. Vivek Kumar2
, Dr. Hitesh Singh3
1
Student, 2
Proffesor, 3
Associate Proffesor, Noida Institute of Engineering and Technology, Greater Noida

28 | Page
here and a more accurate and solution has been given to it through the use of Internet of Thongs, Artificial Neural
networks and Machine Learning.
2. LITREATURE REVIEW
At the first glance we only think that the only technology that will get used in a traffic light is Neural Networks.
But this is completely wrong. There are many other technologies which get used in designing a traffic light. The
main technology is machine learning. Machine learning plays an important role in a traffic light. To understand the
working of the timer of a traffic light and to understand how it can give the best result such that the traffic is reduced
is a big thing. We overcame this problem with the help of python. On the other hand, using technologies such as
Java script, SQL, CSS and databases and linking them to each other such that the lights and the cameras can work
properly is not easy. We tried to link the cameras with the lights on such a way that the camera itself calculated the
depth and arranges the timer on the display of the traffic light. The main disadvantage here is that the camera cannot
calculate the depth on curved roads. The camera and the program are installed in such a way that the camera cannot
calculate the depth on curved roads.
The camera can even detect traffic on both the sides of the road. We also have linked all the four cameras with each
other. Linking of the cameras to the program and then the program getting executed and displaying timer on the
timer display on the traffic light was done through machine learning and ANN.
The technology was created to aid traffic engineers in making decisions on how to respond to traffic issues before
they escalated. In Artificial Neural Networks, we used feed-forward backpropagation, cascade-forward
backpropagation, radial basis, and extended regression. Our technique [1] for creating appropriate estimations is
demonstrated by test results based on actual data acquired from the city of Leicester in the United Kingdom.
Artificial Neural Network (ANN) models [2] are constructed to estimate the ships' paths based on the
aforementioned clustering findings and other known criteria (such as ship speed, loading capacity, self-weight,
maximum power, and water level). When compared to real data, the created model is more than 70% right,
according to the results of the experiments. It will also aid in the development of Yangtze River traffic management
rules.
Using data acquired by in-road stationary sensors, this study seeks to analyse and categorize the traffic congestion
state of distinct road segments inside a city. To classify traffic congestion situations, an Artificial Neural Network
(ANN)-based approach [3] is utilized. ITS will automatically change traffic rules based on traffic congestion levels,
such as changing the length of the queue at traffic lights and offering other routes. It also aids the government in
the development of plans for the building of flyovers and alternative routes to aid traffic control.
Machine learning is a technique in which a computer learns and grows on its own using previously processed data.
These algorithms [4] complete the task swiftly and efficiently. By delivering efficient supervised and unsupervised
learning of obtained data, VANET achieves its purpose. We looked at how machine learning algorithms can address
the security, communication, and traffic concerns that plague VANET systems, as well as the infeasibility of their
implementation.
This study [5] proposes a framework for reducing road traffic congestion in an intelligent traffic management
system that makes the most efficient use of existing infrastructure and resources to meet the aforementioned
challenges. Data collection, data storage, data processing, and business application modules are the four key
modules of the proposed framework.
This study [6] aims to construct a short-term traffic forecasting model for a two-lane undivided roadway with mixed
traffic conditions in India using a back propagation artificial neural network. The findings were compared using
random forest, support vector machine, k-nearest neighbor classifier, regression tree, and multiple regression
models. Back-propagation neural networks outperformed other techniques, with an R2 value of 0.9962, which is
an outstanding result.
The proposed solution [7] relies on vehicle-to-vehicle communication and a local view of traffic congestion. As it
goes along a road, the automobile maintains note of the distance it has travelled and the distance it expects to travel
in a free-flowing traffic situation. When a vehicle is unable to calculate alternative routes due to a lack of critical
traffic data, it performs a reactionary manouver. It is discovered that there is information concerning the traffic.
Three literature solutions are compared to the suggested solution: DIVERT, PANDORA, and s-NRR.

29 | Page
The relocated mobile agents acquire and communicate traffic flow metrics (speed and density), historical data,
resource information, spatio-temporal data, and other data, and assess the static agent using emergent intelligence
methodologies. The static agent's emergent intelligence approach uses analytical, historical, and spatio-temporal
data. [8] monitors and forecasts predicted patterns of traffic density (commuters and vehicles) and trip times in each
zone and area. The static agent determines the best traffic redirection routes based on forecasted and analyzed data,
ensuring smooth traffic flow and reducing traffic congestion frequency, density, and trip duration. NS2, SUMO,
Open Streat Map (OSM), and the MOVE tool are used to perform the performance analysis in a realistic setting.
The algorithm of the mechanism [9] is thoroughly defined in order to provide a good solution to the challenge and
assure the system's stability. The car's registered passive RFID tags will contain all of the vehicle's and owner's
information. These tags will be linked to a virtual wallet that can be recharged in order to avoid penalties. The
system would incorporate systems for tracing missing autos and monitoring vehicular movements based on system
records.
The major purpose of this proposed framework [10] is to create a ground-breaking Internet of Things-based Traffic
Management (IoT-TM) system that can make short-term traffic management choices, allowing for precise and
efficient traffic clearance. Data is gathered from a variety of traffic profiles, which include parameters such as time
consumed. The Mixed Ant Colony Glow Worm Swarm Optimization technique is used to accomplish precise
recognition of the traffic optimized features before learning. The proposed research strategy, IoT-TM, was capable
of producing superior traffic management decisions than existing research approaches because the entire
interpretation of the predicted investigation framework was accomplished in the MATLAB environment.
The purpose of this work [11] is to reduce the average response time for events reported by cars by proposing a
dynamic perspective for real-time traffic planning in fog-based IoV systems. To begin, we develop a decentralized
city-wide traffic signal in which fog nodes are cars located near roadside equipment. We then model parked and
moving vehicle-based fog nodes using queueing theory, arriving at the conclusion that moving vehicle-based fog
nodes can be characterized as an M/M/1 queue. By separating the offloading optimization problem into two
problem instances and arranging traffic flows across multiple fog nodes, an approximation technique is developed
to address it. To demonstrate the superiority of our technique, performance reviews based on real-world cab datasets
are undertaken.
[12] Artificial neural networks (ANN) approaches are combined with adaptive learning concepts to create the
prediction methodology. These learning principles lead to a large increase in accuracy and a decrease in
computation times. Furthermore, the study aims to select the best adaptive learning rule among a set of rules. Traffic
data from a number of places across the United Kingdom is used to test the forecast model (U.K.). The findings
show that the suggested multivariate forecasting method can simultaneously predict traffic flow and speed in
freeway or expressway networks while being computationally more efficient.
It's a difficult undertaking to collect and model a large number of time series from all route segments. Data mining
techniques are used to transform complex data into a structural framework and make decision-making easier for
ITMS issues. [13] Furthermore, for deriving analytical conclusions on optimal route design, real-time road section
weights must be generated using continuous data from various time domains for each day of the year. Dynamic
road weights are created or updated using a variety of climate, road, and vehicle-related decision criteria.
Determining the weight of a road segment is complicated due to decision overlap within attribute clusters. A
classification is used to give exact data modelling without creating a chaotic overlapping environment. It's critical
to take a proactive approach.
In addition to the historic and real-time techniques, various time series and artificial neural networks (ANN) models
[14] were constructed in the prediction methodology competition. According to the current study, support vector
machines (SVM), a recently discovered pattern recognition and regression technology, could be utilized to estimate
traffic speed in the near future. Furthermore, an ANN model is built, and the performance of both of these tactics,
as well as real-time and historical approach findings, are compared. The investigation was conducted using data
from San Antonio's roadways.
The primarypurpose of the study[15] was to determine how efficient various smoothing models werein eliminating
anomalies in traffic flow data before employing an artificial neural network to estimate short-term traffic flow
dynamics. in order to make things easier All of the data collected by the loop detector, wavelet filter, moving
average model, and Butterworth filter was thoroughly analyzed. The artificial neural network was then used to
calculate traffic flow across different time intervals, which was then quantified using generally recognized

30 | Page
assessment standards. The results of the research enable us to develop efficient and precise denoising algorithms
for anticipating short-term traffic flow.
In the proposed technique [16], this controller acts as a preventive controller. We also create an artificial neural
network-based intelligent clustering controller that acts as a reactive controller to guide clustering in the spike
ISDN-IoT sensing zone. As a result, an intelligent queuing model is offered to control the spike ISDN-IoT network's
flow table buffer capacity, hence improving the overall network's service quality (QoS). A new training method
was utilized to train the PRSNN to change its weight and threshold. The recommended model increases QoS by
14.36 percent when compared to a deep neural network, according to simulated data.
In this research, we propose the Fog Route Vehicular (FOREVER) mechanism [17] for Intelligent Transport
Systems to aid traffic management in Vehicular Networks (VANET). In order to accomplish this, FOREVER will
discover and provide an alternate route for drivers to bypass past congestion. FOREVER is based on the FOG
computing paradigm, which attempts to compute and alter a vehicle's path in order to avoid traffic jams. As a result,
FOREVER was able to lower CO 2 emissions by 7.9%, stop time by 8.3%, and journey duration by 7.6%, according
to the data.
Using various practical demonstrations, this part [18] discussed and demonstrated an integrated hierarchical multi-
layer traffic management, surveillance, and control system (TMSCS). An example of an integrated multi-layer
multi-criteria traffic management issue in TMSCS was handled in the context of PIACON processes. In addition,
specific ANN and fuzzy logic techniques are used. They enable traffic control options and multi-criteria preference
structures to be selected.
The Dell PowerEdge T430 server was used to test 5 commonly used datasets of varied sizes in air traffic control
big data to verify the technique [19]. According to research, this method can enhance the Spark system's
performance by around 35% on average when compared to the default parameter configuration. The performance
of the algorithm improves as the dataset is larger. This solution effectively ensures the Spark system's parameter
adjustment efficiency and achieves the goal of efficiently processing ATM large data.
As a prediction technique, an Artificial Neural Network (ANN) [20] is employed, and ANN is trained using various
traffic data and environmental variables. The ANN is trained using these features, as well as the measured real-time
arrival time of the bus at various stops over a 10-day period. This trained ANN is used on the server. In this study,
the performance of the proposed IOT-ANN-SPTS system is compared to that of previous comparable
investigations. According to the performance analysis, the proposed system minimized error.
TSC and ITS can take advantage of a wide range of new technologies [21]. Machine learning (CI) allows traffic
systems to overcome nonlinearity and unpredictability by providing flexibility, autonomy, and robustness. This
paper surveys some commonly used CI paradigms, analyses their applications in TSC systems for urban surface-
way and freeway networks, and introduces current and potential issues of control and management of recurrent and
nonrecurrent congestions in traffic networks in order to provide valuable references for further research and
development.
In order to predict location-based speed, stream-based density, and stream speed, erroneous data is fed into an
artificial neural network (ANN) that has been trained with accurate data and the same procedure is used. For a
variety of input error ranges, [22] is employed. The findings show that when trained with high-quality data, the
ANN can withstand automated data errors and provide accurate traffic status estimates, highlighting its utility for
real-time ITS applications in such traffic circumstances.
This paper covers the results of the first six-month of the experiment [23]. ANN approaches, their applications in
transportation and traffic control, prospective data sources, and congestion characterization and management were
all thoroughly examined. A demonstrator application was built based on the review's conclusions to demonstrate
the capabilities of an ANN UTC application. In reality, the demonstration app had a far larger purpose.
The taxonomy for categorizing development studies [24] is based on the type of dataset that was utilized for
development. A variety of motivational subjects to pursue research advancement in this field have been reported.
Recommendations for various stakeholders on a number of crucial topics are offered in order to promote and
expedite progress in the car-following setting. In order to find research gaps in the experimental methodologies
employed in the literature, a detailed examination is carried out.

31 | Page
It has been proven that ANN is good at predicting traffic congestion. [25] The outcomes were improved even further
when a decisional tree was used to incorporate expected congestion state, real-time GPS data, and anomalous
occurrences. Drivers have access to a real-time smartphone application that informs them of traffic conditions at
their destination. Large GPS datasets obtained from autos travelling in a densely populated Tunisian city were used
to test and verify the model.
TABLE I: TECHNOLOGIES USED AND THEIR SUCCESS RATES
References
Number
Technology used Success
Rate
Year Proposed
1 Java and Artificial Neural Systems 90% 2013
2 Artificial Neural Network 95% 2016
3 In Road stationary Sensors 58% 2019
4 VANET 65% 2021
5 ITS 47% 1287-1297
6 Artificial Neural Networks 77% 2018
7 V2V communication 55% 2020
8 Emergent Mobile Techniques 48% 2020
9 ANN Algorithms 78% 2019
10 IoT based traffic management system 58% 2017
11 IoV Systems 88% 2018
12 ANN 77% 2012
13 ITMS 58% 2019
14 ANN, Support Vector Machines 15% 2004
16 ISN-IoT 85% 2020
17 VANET 98% 2017
18 Traffic Management Surveillance and
control system
88% 2006
19 Spark System 77% 2021
20 Artificial Neural Networks, SPTS 85% 2019
21 Computational Intelligence, CI
Programs
54% 2017
22 ANN and Dijkstra 66% 2011
25 Research Mapping 25% 2020
The table above gives a proper technology wise usage and the years in which they were proposed. Also, the table
aims to show how much the technologies were not able to achieve a full hundred percent of their capability. The

32 | Page
most effective technology out of the above technologies was the VANET (Vehicle Adhoc Network), which has
been proven to be a very effective technology in the past, and its currently being used in our current traffic lights
as well.
3. CONCLUSIONS
This paper sums up to summary that we can even make the traffic light more automated and less human dependent
through the use of modern technologies such as Internet of Things, Artificial Neural Networks and some basic
machine learning. The main aim here is to show how we can adapt to the changes in the modern world by evolving
ourselves with the new technologies that are coming every second in our world. The basic knowledge someone
should have for understanding this paper is some basic machine learning and some coding languages. Also, this
paper shows us how we can make the make use of technologies in our modern world and implement them for the
benefit of mankind.
4. FUTURE SCOPE
The future scope of this traffic light is that this traffic light will not only be accessible only in our country but also
in other countries, which in turn will eventually make our country recognizable all over the world. This paper also
aims to mainly free ourself from human dependencies and also to make automation a part of our daily lives. Until
yesterday, people were totally dependent on humans for the operating of traffic lights. These traffic lights as
compared to our traditional traffic lights which are being currently used, are less expensive and also require less
supervision.
REFERENCES
[1] Passow, Benjamin N., et al. "Adapting traffic simulation for traffic management: A neural network
approach." 16th International IEEE Conference on Intelligent Transportation Systems (ITSC 2013). IEEE,
2013.
[2] Gan, Shaojun, et al. "Ship trajectory prediction for intelligent traffic management using clustering and
ANN." 2016 UKACC 11th International Conference on Control (CONTROL). IEEE, 2016
[3] Ashifuddin Mondal, Md, and Zeenat Rehena. "Intelligent traffic congestion classification system using
artificial neural network." Companion Proceedings of the 2019 World Wide Web Conference. 2019.
[4] Khatri, Sahil, et al. "Machine learning models and techniques for VANET based traffic management:
Implementation issues and challenges." Peer-to-Peer Networking and Applications 14.3 (2021): 1778-
1805.
[5] Mondal, Md Ashifuddin, and Zeenat Rehena. "An IoT-Based Congestion Control Framework for
Intelligent Traffic Management System." Advances in Artificial Intelligence and Data Engineering.
Springer, Singapore, 2021. 1287-1297.
[6] Sharma, Bharti, et al. "ANN based short-term traffic flow forecasting in undivided two-lane
highway." Journal of Big Data 5.1 (2018): 1-16.
[7] Gomides, Thiago S., et al. "An adaptive and distributed traffic management system using vehicular ad-
hoc networks." Computer Communications 159 (2020): 317-330.
[8] Chavhan, Suresh, and Pallapa Venkataram. "Prediction based traffic management in a metropolitan
area." Journal of traffic and transportation engineering (English edition) 7.4 (2020): 447-466.
[9] Patel, Meet Pratap, and Sohrab Ardeshar Vakharia. "RFID BASED TRAFFIC MANAGEMENT
SYSTEM USING ARTIFICIAL NEURAL NETWORKS." Advance and Innovative Research (2019):
373.
[10] Eswaraprasad, Ramkumar, and Linesh Raja. "Improved intelligent transport system for reliable traffic
control management by adapting internet of things." 2017 International Conference on Infocom
Technologies and Unmanned Systems (Trends and Future Directions) (ICTUS). IEEE, 2017
[11] Wang, Xiaojie, Zhaolong Ning, and Lei Wang. "Offloading in Internet of vehicles: A fog-enabled real-
time traffic management system." IEEE Transactions on Industrial Informatics 14.10 (2018): 4568-4578.
[12] Dunne, Stephen, and Bidisha Ghosh. "Regime-based short-term multivariate traffic condition forecasting
algorithm." Journal of Transportation Engineering 138.4 (2012): 455-466.

33 | Page
[13] Sumit, Sakhawat Hosain, and Shamim Akhter. "C-means clustering and deep-neuro-fuzzy classification
for road weight measurement in traffic management system." Soft Computing 23.12 (2019): 4329-4340
[14] Vanajakshi, Lelitha, and Laurence R. Rilett. "A comparison of the performance of artificial neural
networks and support vector machines for the prediction of traffic speed." IEEE Intelligent Vehicles
Symposium, 2004. IEEE, 2004.
[15] Chen, Xinqiang, et al. "Sensing data supported traffic flow prediction via denoising schemes and ANN: a
comparison." IEEE Sensors Journal 20.23 (2020): 14317-14328.
[16] Al-Jamali, Nadia Adnan Shiltagh, and Hamed S. Al-Raweshidy. "Intelligent traffic management and load
balance based on spike isdn-iot." IEEE Systems Journal (2020).
[17] Brennand, Celso ARL, et al. "A novel urban traffic management mechanism based on fog." 2017 IEEE
symposium on computers and communications (ISCC). IEEE, 2017.
[18] Adamski, Andrzej. "2.2. INTEGRATED TRAFFIC MANAGEMENT SURVEILLANCE AND
CONTROL SYSTEMS IN URBAN AREAS." (2006).
[19] Wei-jun, P. A. N., et al. "Application of Improved Spark System Based on ANN in Big Data Processing
of Air Traffic Management# br." Computer and Modernization 12 (2021): 78.
[20] Jabamony, Jalaney, and Ganesh Ramaswamy Shanmugavel. "IoT based bus arrival time prediction using
Artificial Neural Network (ANN) for smart public transport system (SPTS)." International Journal of
Intelligent Engineering & Systems 13 (2019).
[21] El Fazziki, Abdelaziz, et al. "An agent-based traffic regulation system for the roadside air quality
control." IEEE Access 5 (2017): 13192-13201.
[22] Zhao, Dongbin, Yujie Dai, and Zhen Zhang. "Computational intelligence in urban traffic signal control:
A survey." IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) 42.4
(2011): 485-494.
[23] Fulari, Shrikant, Lelitha Vanajakshi, and Shankar C. Subramanian. "Artificial neural network–based
traffic state estimation using erroneous automated sensor data." Journal of Transportation Engineering,
Part A: Systems 143.8 (2017): 05017003.
[24] Lyons, Glenn, et al. "Urban traffic management; the viability of short-term congestion forecasting using
artificial neural networks." PTRC, 1996.
[25] Talal, Mohammed, et al. "Review on car-following sensor based and data-generation mapping for safety
and traffic management and road map toward ITS." Vehicular Communications (2020): 100280.
[26] Akahane, Hirokazu, Takashi Oguchi, and Hiroyuki Oneyama. "Saga of Traffic Simulation Models in
Japan." Simulation Approaches in Transportation Analysis. Springer, Boston, MA, 2005. 269-300.
[27] Shaaban, Khaled, et al. "Queue Discharge at Freeway On-Ramps Using Coordinated Operation of a Ramp
Meter and an Upstream Traffic Signal." Procedia Computer Science 170 (2020): 347-353.

A Review paper on Artificial Neural Network: Intelligent Traffic Management System

More Related Content

Similar to A Review paper on Artificial Neural Network: Intelligent Traffic Management System (20)

More from NIET Journal of Engineering & Technology (NIETJET) (20)

Recently uploaded (20)

A Review paper on Artificial Neural Network: Intelligent Traffic Management System