Intelligent Traffic Management System for Smart Cities

Abstract

In present-day times, the number of vehicles has increased drastically, but in contrast, the capabilities of our roads and transportation systems still remain underdeveloped and as a result, fail to cope with this upsurge in the number of vehicles. As a consequence, traffic jamming, road accidents, increase in pollution levels are some of the common traits that can be observed in our new age cities. With the emergence of the Internet of Things and its applicability in Smart Cities, creates a perfect platform for addressing traffic-related issues, thus leading to the establishment of Intelligent Traffic Management Systems (ITMS). The work presented in this paper talks about an intelligent traffic management system that lays its foundation on Cloud computing, Internet of Things and Data Analytics. Our proposed system helps to resolve the numerous challenges being faced by traffic management authorities, in terms of predicting an optimum route, reducing average waiting time, traffic congestion, travel cost and the extent of air pollution. The system aims at using machine learning algorithms for predicting optimum routes based upon traffic mobilization patterns, vehicle categorization, accident occurrences and levels of precipitation. Finally, the system comes up with the concept of a green corridor, wherein emergency services are allowed to travel without facing any kinds of traffic congestion.

Full text

Intelligent Traffic Management System for Smart Cities
Abhirup Khanna
The University of Melbourne
abhirupkhanna@yahoo.com
Rohit Goyal
Himgiri ZEE University
rohitgoyalhzu@gmail.com
Manju Verma
Himgiri ZEE University
mv.verma2007@gmail.com
Deepika Joshi
Himgiri ZEE University
deepika.joshi0401@gmail.com
Abstract. In present-day times, the number of vehicles has increased drastically, but in
contrast, the capabilities of our roads and transportation systems still remain underdeveloped
and as a result, fail to cope with this upsurge in the number of vehicles. As a consequence,
traffic jamming, road accidents, increase in pollution levels are some of the common traits that
can be observed in our new age cities. With the emergence of the Internet of Things and its
applicability in Smart Cities, creates a perfect platform for addressing traffic-related issues, thus
leading to the establishment of Intelligent Traffic Management Systems (ITMS). The work
presented in this paper talks about an intelligent traffic management system that lays its founda-
tion on Cloud computing, Internet of Things and Data Analytics. Our proposed system helps to
resolve the numerous challenges being faced by traffic management authorities, in terms of
predicting an optimum route, reducing average waiting time, traffic congestion, travel cost and
the extent of air pollution. The system aims at using machine learning algorithms for predicting
optimum routes based upon traffic mobilization patterns, vehicle categorization, accident occur-
rences and levels of precipitation. Finally, the system comes up with the concept of a green
corridor, wherein emergency services are allowed to travel without facing any kinds of traffic
congestion.

Keywords: Smart Cities; Internet of Things; Traffic
Management System; Machine Learning
1 INTRODUCTION
In today’s times, traffic management has become one of the core concerns for an
urban city. The constant increase in the number of vehicles has led to the recurring
problem of traffic management. An increase in the infrastructure growth is a possible
solution but turns out to be costly in terms of both time and effort. Countries all over
the world are looking forward to developing efficient traffic management systems by
making use of ICT technologies. The recent advancements in Wireless Sensor Net-
works(WSN) and low-cost low power consuming sensors have strengthened the re-
gime towards creating an intelligent traffic management system[2]. Governments are
trying to capitalize the power of present-day computing, networking and communica-
tion technologies for building systems that are able to improve the efficiency of cur-
rent roads and traffic conditions. The advent of the Internet of Things and high avail-
ability of Cloud resources are helping us create mechanisms that can automate the
transportation systems and enhance utilization of existing infrastructures[4].
The tiny sensors which we have today have their applicability across various fields
such as health, surveillance, home automation and industrial practices. A network of
such sensors is able to map an entire city and collect minutest of the details with min-
imum time and cost overhead. With IPv6 becoming more and more popular it be-
comes easy to allocate a sensor node with an IP address for its tracking and localiza-
tion purposes[12]. Traffic systems can make use of such sensor nodes for gathering
real-time information regarding traffic conditions like traffic flow, traffic congestion,
etc. These sensors are also capable of vehicle classification, speed calculation and
vehicle count[5]. The data being collected from these sensor nodes is diverse in nature
and humongous in size. We are fortunate to live in times where we have efficient data
analytics through machine learning algorithms for extracting information or say
knowledge from these huge chunks of data. Machine learning algorithms are capable
of making predictions regarding the levels of traffic congestion in a particular area of
a city. They can very well depict patterns with respect to traffic flow and suggest
measures that authorities can take to curb traffic-related problems. A traffic manage-
ment system can only be successful when all of its actors work and communicate in
sync with another. Talking about our work, we present an Intelligent Traffic Man-
agement System that caters to all traffic related issues of a smart city. Our model sug-
gests an optimum route which it takes into consideration parameters like, travel time,
travel cost (fuel consumption) and travel distance. Our system in use of machine
learning algorithms predicts levels of traffic congestion at various time intervals. It
also comes up with the concept of a green corridor catering to emergency vehicles.
The rest of the paper is categorized as follows: Section 2 elucidates the algorithm
that depicts the workflow of the entire system, whereas its complete layered architec-

ture and mathematical model are discussed in Section 3. Finally, Section 4 exhibits
the implementation and simulation of our proposed system.
2 ALGORITHM
The working of our Intelligent Traffic Management System could be explained
through the illustration of the algorithm that forms the core for it. The algorithm de-
picts the workflow of the system by representing the relationship between various
actors and the information that they share in form of parameters among themselves.
Step 1: Start()
Step 2: Traffic_Management_Controller() Initialization Block
Step 3: Traffic_Moinitoring_Unit ()
Step 4: ORS()
Step 5: On Road Sensors collect information from every road and intersection
Step 6: Vehicle Nodes transmit there location information
Step 7: All information is sent to the respective Gateways. The entire city is divided
into areas and each are has a gateway assigned to it.
Step 8: Gateways transmits all the information to respective TMU and TMC.
Step 9: Data is stored and processed at the Cloud end. KNN based anomaly detection
algorithm is used for categorizing incidents as an accident or not. Features such as
traffic density, moving traffic velocity, vehicle presence, average waiting time and
levels of precipitation are taken into consideration. Levels of precipitation have been
divided into three categories: 0 -10 cm; 10 - 20cm and above 20cm.
Step 10: Random Forest algorithm is used for traffic estimation and predicts traffic
congestion levels across various time intervals.
Step 11: End user enters Source and Destination.
Step 12: Optimum Route is computed considering factors like, average waiting time,
total travel time, travel distance, moving traffic velocity, number of intersections and
intended fuel consumption. A vector space model is constructed based on all of these
parameters for all routes leading to the desired destination. A route whose vector lies
in the region of optimization is considered as optimum route.

Step 13: Results are communicated to the specific Vehicle Node
Step 14: End()
3 PROPOSED WORK
In this section, we discuss our proposed Intelligent Traffic Management System and
all the various actors that constitute it. We present a layered architecture that depicts
the functionalities of our traffic management system and showcases all the different
entities which it comprises. The core of our proposed system is based upon presenting
an optimum route followed by traffic estimation.
3.1 Design Objectives
In this subsection, we elucidate some of the prominent objectives which we intend to
achieve through our proposed work. These objectives can also be considered as driv-
ing forces for designing our proposed intelligent traffic management system.
• Traffic Monitoring: It can be considered as one of the key components of a smart
city. Traffic monitoring allows the local authorities to monitor the flow of traffic per-
taining to a particular area, route or street. It helps in keeping track of the inflow of
traffic from other neighboring cities during specific days or a particular time of the
year. Historical data of traffic monitoring can be very useful in smart city planning
and city infrastructure development.
• Pollution Avoidance: Rising pollution levels pose a threat to the environment as
along with having adverse impacts on human health and wellbeing. The extent of air
and noise pollution are directly proportionally to the intensity of traffic congestion in
a city. Long-standing queues of vehicles result in the exorbitant emission of pollutants
resulting in an increase in temperatures, a decrease in rainfall, respiratory problems,
etc.
• Route Optimization: In recent times, it has been observed that the shortest route
doesn’t seem to work well in terms of total travel time, fuel consumption and average
waiting time. In such scenarios, an optimum route is the best option for travel as it
considers factors such as traffic congestion, distance traveled, total travel time and
fuel consumption. An optimum route comprises of a tradeoff between all these pa-
rameters and sits well for a traveler in context to its time and money being spent on
travel.
• Green Corridor: It’s been a couple of years since the concept of a green corridor
has seen the light of the day. It is a corridor which in reality is a route from a source to
the destination comprising of various traffic signals all of which having a green sig-

nal. The green corridor is used to cater to the emergency vehicles by allowing them to
reach their desired destination without any waiting time and at maximum speed.
• Accident Detection: The overcrowded streets of present-day roads have given rise
to the number of accidents. Accident detection is a crucial part of a traffic manage-
ment system as it not only informs the medical services to attend to the accident hit
personnel’s but also has an impact on the traffic flow and congestion levels of a pa r-
ticular region.
• Jamming: Prevention of traffic jams and reduction in average waiting time are the
two most important functionalities of an efficient traffic management system.
• Vehicle Tracking: It helps the local administration in keeping track of vehicles in
terms of the areas they are traveling, time of travel, speed, places visited and vehicle
type. All of these parameters prove to be fruitful when it comes to maintaining a state
of law and order in the city.
3.2 Layered Architecture
In this subsection we would be discussing the layered architecture of our proposed
intelligent traffic management system. We would also be talking about the various
actors along with their functionalities that constitute the system. Following is the dia-
gram that depicts the layered architecture for our proposed system.
Fig. 1. Intelligent Traffic Management System Architecture
 Traffic Management Controller(TMC): The purpose of the controller is to man-
age and govern the entire system[16]. It is the controller which orchestrates the
functionalities of other application modules and entities within the system. The

controller resides at the Cloud end and has detailed information regarding every
vehicle, traffic signal, gateway, On Road Sensors and Traffic Management Unit.
All of this information is stored and processed by the controller in order to generate
optimized routes between the specified source and destination. The controller es-
tablishes a one to one connection with the middleware and circulates all of its or-
ders through it. It is the controller which generates prediction data concerning with
levels of traffic congestion at varying time intervals. The TMC is the one which
uses a hop counter based flooding algorithm for broadcasting notifications regard-
ing an accident, change of routes, road developmental activities and adverse cli-
matic impact. The occurrence of an emergency vehicle and creating a green corri-
dor for it is all done through the traffic management controller.
 Gateways: All the information that has been sensed and collected by the on-road
sensors are transmitted to the gateways[6]. Gateways act as a common point of
contact wherein diverse kinds of information coming from heterogeneous types of
sensors gets collected. The gateways use greedy based data collection algorithm for
collecting data from various data sources. It is the gateway which is responsible for
the global addressing of Vehicle Nodes (V) by making use of IPv4 addresses[9].
Each gateway is allotted a coverage area, wherein each on-road sensor and vehicle
node has been given an IP address thus facilitating efficient identification of ob-
jects within that area. Every gateway is allocated more than one area so as to en-
hance the granularity of vehicle identification. The gateway also keeps track of its
neighboring gateways along with the total number of vehicle nodes traveling in its
area. Finally, the gateway transmits all forms of unstructured information to its
subsequent traffic management controller.
 Traffic Monitoring Unit(TMU): It acts as an intermediary node between On
Road Sensors and Gateways. The purpose of adding a TMU is to enhance the re-
sponse time of the system as communicating directly with the TMC could lead to
increased latency cost.TMU provides a communication link between TMC and the
rest of the system and also offers local processing and storage capabilities in order
to boost the efficiency of the system[8]. Any information coming from an on-road
sensor or vehicle node is addressed by the TMU which then subsequently informs
the Controller and other devices on the network. All the instruction given by the
Controller are communicated through the TMU to the respective vehicle nodes and
local authorities. The traffic monitoring unit can also be considered as a Fog com-
puting element as it resides at the edge of the network making its access both easy
and efficient. It is the TMU which at regular intervals updates the traffic manage-
ment controller about information regarding every entity involved in the system.
 On Road Sensors(ORS): Sensors are the eyes and ears of the system as they de-
tect the occurrence of events, surrounding conditions and transmit the collected in-
formation. The work of the on-road sensors is to monitor and perceive events or
phenomena that take place on road. Every ORS can be categorized on the basis of
three parameters namely, sensor type, methodology, and sensing parameters. Sen-

sor type defines which type of sensor it is i.e. whether it is a homogeneous or a
heterogeneous sensor or it is a single dimensional or a multidimensional sensor.
Methodology talks about the ways in which a sensor gathers information[16]. It
can be either active or passive in nature. Sensing parameters are the number of pa-
rameters which a sensor can sense. A sensor might just sense one parameter like
body temperature or many parameters like in the case of an ECG. Each sensor
node is provided an IP address which helps in its unique identification. Every sen-
sor node communicates all of its sensor data to its subsequent gateway. Entities
starting with the letter “S” represent the On Road Sensors in the physical topology.
In case of our work, we have used inductive loop sensor technology. The following
are the functionalities that an On-Road Sensor provides.
o Vehicle Count
o Vehicle Presence
o Vehicle Speed
o Vehicle Classification
o Low Bandwidth Consumption
 Vehicle Node: It is the vehicle for whom an entire transportation system is con-
structed in order to provide an effortless and convenient traveling experience. It
can also be seen as a moving sensory node which continues to receive and transmit
information while traveling. Each vehicle node is provided an IP address which
helps in its unique identification. Every sensor node communicates all of its sensor
data to its subsequent gateway. Entities starting with the letter “V” represent the
Vehicle Node in the physical topology. Every transportation vehicle has an LED
display installed that informs the pilot about the most optimum route and the con-
stantly changing levels of traffic. All messages or notifications such as accident
alert or prevention of entry in a particular area from the TMC can be seen on the
LED display.
3.3 Mathematical Model
In this subsection we would be discussing the mathematical model for our pro-
posed system. Following is the nomenclature table that describes all the various enti-
ties that have been used in this mathematical model.
Table 1. Nomelclature Table
Symbol
Meaning
Γ
On Road Sensors
V
Vehicle Node
r
Road
R
Route
Α
Fuel Consumption
ß
Traffic Density

Ф
Moving Traffic Velocity
T
Average Waiting Time
H
Total Waiting Time
Ψ
Vehicle Type
Δ
Vehicle State
N
Number of Intersections
E
Vehicle Priority
Θ
Optimum Vehicle Speed
RC
Road Capacity
RS
Route Selection Function
F
Travel Cost Function
W
Traffic Management Controller
X
Traffic Signal
Ω
Traffic Flow Percentage
V={Ψ,δ,E} (1)
Ψ ∈ (0, 1)// Vehicle type i.e. 0 for light vehicle and 1 for heavy vehicles
δ ∈ (0, 1)// Vehicle state i.e. 0 for stationary and 1 for moving
E ∈ (0, 1)// Vehicle priority i.e. 0 for normal vehicles and 1 for emergency vehi-
cles
H = T / N (2)
Ф = ∑ (D / t) - (∑D) / T (3)
Total Distance = ∑D (4)
ß ∝ 1 / Ф
ß = K / Ф
Ф = (K/ D) X t (5)
t = (Ф X D) / K (6)
τ = ∑t + T (7)
α = (mileage X (t)2 X (speed)2) / Total Distance (8)

α ∝ speed Vehicle Speed > θ
α ∝ time
α ∝ 1 / speed Vehicle Speed <= θ
α ∝ time
F ← (α,t, D) (9)
RC > threshold value
W → V{speed = 0} (10)
Ψ = 1
In case the road capacity exceeds the threshold value, the traffic management con-
troller will prevent entry of all heavy vehicles into that zone. Any such vehicle in the
affected zone will be directed to stop until further directions from the central control-
ler.
X = 3// Traffic light goes green 10 seconds prior to arrival of emergency vehicle at
the intersection
V(E) = 1
ω = ((D/Speed - T) X 100) / (D / Speed) (11)
4 IMPLEMENTATION & SIMULATION
The above mentioned algorithm is implemented on iFogSim framework. In
iFogSim[17] there are various predefined classes that provide a simulation environ-
ment for Internet of Things combined with the benefits of cloud computing. It is a
java based simulation toolkit and can be implemented either using Eclipse or
NetBeans IDE. In our case we would be using the eclipse IDE. To run iFogSim on
eclipse, we first need to download the eclipse IDE and install it. After successful in-
stallation of eclipse IDE, download the latest iFogSim package, extract it and import
it in eclipse. Talking of our proposed work we have created our own classes in
iFogSim and have portrayed our algorithm in form of java code.
In terms of implementing machine learning algorithms, we have made use of We-
ka[15], which is a popular open source tool for executing machine learning algo-
rithms.

The following tables depict the simulation environment for our paper along with
the improvements that can be seen after successful implementation of our model.
Table 2. Vechile Count per Road
Road 1
Road 2
Road 3
Road 4
45
92
70
60
53
75
4
78
90
12
80
77
103
13
95
66
13
20
78
82
9
54
20
95
33
88
28
86
Table 3. Average Waiting Time per Road
Road 1
Road 2
Road 3
Road 4
602.76
460.94
526.07
561.25
531.72
473.87
712.51
443.89
446.38
717.13
487.73
511.73
463.87
664.86
483.43
557.89
689.59
622.82
457.14
412.45
664.12
524.49
618.45
332.98
658.03
464.25
629.67
458.07
Table 4. Improved Average Waiting Time per Road
Road 1
Road 2
Road 3
Road 4
552.72
422.84
507.13
522.15
501.76
433.77
695.16
417.09
408.35
687.11
452.03
486.43
403.57
614.74
428.53
559.39
669.47
612.32
462.03
402.15
644.14
501.42
602.05
311.08
628.15
424.17
611.07
419.17
It is very much evident that the average waiting time has reduced for each road
over all time intervals. Although, the degree of reduction in average waiting time
varies from road to road but the predominant trend over the entire data set remains
the same i.e. a decrease in the average waiting time.
As earlier as discussed in section 2, our traffic estimation is based upon the Ran-
dom Forest algorithm and below is a graph illustrating the correctness of our feature
selection along with the algorithm that we have chosen. The graph depicts the com-

parison between the actual and estimated values of traffic in terms of vehicle count.
As inferred from the graph below, the estimated traffic count may not be the same as
that in actual but in a larger prospective the increase and decrease in the levels of
traffic over all time intervals turnout to be the same for both estimated and actual
values.
Fig. 2. Traffic Estimation
Till now we have discussed how our proposed intelligent traffic management sys-
tem turns out to be beneficial in terms of reducing the average waiting time for a giv-
en road along with making correct predictions with respect to varying levels of traffic.
The following graphs present an analysis of our accident detection mechanism and
presents a comparison between the estimates and actual number of accidents occurred
for a particular road at different time intervals.
Fig. 3. Accident Detection
0
10
20
30
40
50
60
70
80
90
0
2
4
6
8
10
Number of Vehicles
Time
Traffic Estimation
Traffic Count
Estimate
Traffic Count
Actual

Fig. 5. Accident Detection
As per the above two figures, the estimated number of accidents always turn out to
be greater than the actual number of accidents. This implies, that our system was able
to detect real accidents but it also categorized other scenarios having similar charac-
teristics as that of an accident as an accident itself. Thus resulting in an increase in the
number of accidents as compared to the ones that actually happened.
5 CONCLUSION
Traffic Management System is one of the many domains of a Smart City wherein
significant research can be seen. It is an area of work which has answers to many
current day problems pertaining to traffic management of smart cities. We propose a
novel Intelligent Traffic Management System for Smart Cities which facilitates Wire-
less Sensor Networks, Internet of Things, Cloud Computing and Data analytics. The
work discusses the ways in an optimum route is suggested to the end user. The opti-
mum route turns out to be beneficial than the shortest route in most cases in terms of
fuel cost and total travel time. Through our research, we were successful in generating
an optimum route along with making predictions regarding traffic congestion levels.
The system also talks about events of accidents and how they may have an impact on
the traffic flow of a region. Levels of precipitation, an occurrence of an accident, con-
cept of a green corridor, the rate of fuel consumption, % flow of traffic and use of
machine learning algorithms are some of the novel features of our work. In future, we
intend to introduce the vehicle to vehicle communication and impact of speed break-
ers on traffic flow and congestion.

REFERENCES
1. Miz, V., & Hahanov, V. (2014, September). Smart traffic light in terms of the cognitive
road traffic management system (CTMS) based on the Internet of Things. In Design &
Test Symposium (EWDTS), 2014 East-West (pp. 1-5). IEEE.
2. Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of Things (IoT): A
vision, architectural elements, and future directions. Future generation computer sys-
tems, 29(7), 1645-1660.
3. Foschini, L., Taleb, T., Corradi, A., & Bottazzi, D. (2011). M2M-based metropolitan plat-
form for IMS-enabled road traffic management in IoT. IEEE Communications Maga-
zine, 49(11).
4. Yu, M., Zhang, D., Cheng, Y., & Wang, M. (2011, May). An RFID electronic tag based
automatic vehicle identification system for traffic IOT applications. In Control and Deci-
sion Conference (CCDC), 2011 Chinese (pp. 4192-4197). IEEE.
5. Zhou, H., Liu, B., & Wang, D. (2012). Design and research of urban intelligent transporta-
tion system based on the internet of things. Internet of Things, 572-580.
6. Khanna, A., & Anand, R. (2016, January). IoT based smart parking system. In Internet of
Things and Applications (IOTA), International Conference on (pp. 266-270). IEEE.
7. Lingling, H., Haifeng, L., Xu, X., & Jian, L. (2011, December). An intelligent vehicle
monitoring system based on internet of things. In Computational Intelligence and Security
(CIS), 2011 Seventh International Conference on (pp. 231-233). IEEE.
8. Kyriazis, D., Varvarigou, T., White, D., Rossi, A., & Cooper, J. (2013, June). Sustainable
smart city IoT applications: Heat and electricity management & Eco-conscious cruise con-
trol for public transportation. In World of Wireless, Mobile and Multimedia Networks
(WoWMoM), 2013 IEEE 14th International Symposium and Workshops on a (pp. 1-5).
IEEE.
9. Khanna, A., & Tomar, R. (2016, October). IoT based interactive shopping ecosystem.
In Next Generation Computing Technologies (NGCT), 2016 2nd International Conference
on (pp. 40-45). IEEE.
10. Tarapiah, S., Atalla, S., & AbuHania, R. (2013). Smart on-board transportation manage-
ment system using gps/gsm/gprs technologies to reduce traffic violation in developing
countries. International Journal of Digital Information and Wireless Communications
(IJDIWC), 3(4), 430-439.
11. Parwekar, P. (2011, September). From Internet of Things towards cloud of things.
In Computer and Communication Technology (ICCCT), 2011 2nd International Confer-
ence on(pp. 329-333). IEEE.
12. Zhou, J., Leppanen, T., Harjula, E., Ylianttila, M., Ojala, T., Yu, C., ... & Yang, L. T.
(2013, June). Cloudthings: A common architecture for integrating the internet of things
with cloud computing. In Computer Supported Cooperative Work in Design (CSCWD),
2013 IEEE 17th International Conference on (pp. 651-657). IEEE.
13. Rajan, M. A., Balamuralidhar, P., Chethan, K. P., & Swarnahpriyaah, M. (2011, February).
A self-reconfigurable sensor network management system for internet of things paradigm.
In Devices and Communications (ICDeCom), 2011 International Conference On (pp. 1-5).
IEEE.

14. Tomar, R., Khanna, A., Bansal, A., & Fore, V. (2018). An Architectural View Towards
Autonomic Cloud Computing. In Data Engineering and Intelligent Computing (pp. 573-
582). Springer, Singapore.
15. Hall, M., Frank, E., Holmes, G., Pfahringer, B., Reutemann, P., & Witten, I. H. (2009).
The WEKA data mining software: an update. ACM SIGKDD explorations newslet-
ter, 11(1), 10-18.
16. Fore, V., Khanna, A., Tomar, R., & Mishra, A. (2016, November). Intelligent supply chain
management system. In Advances in Computing and Communication Engineering
(ICACCE), 2016 International Conference on (pp. 296-302). IEEE.
17. Gupta, H., Vahid Dastjerdi, A., Ghosh, S. K., & Buyya, R. (2017). iFogSim: A toolkit for
modeling and simulation of resource management techniques in the Internet of Things,
Edge and Fog computing environments. Software: Practice and Experience, 47(9), 1275-
1296.