International Journal of Transportation Science and Technology 7 (2018) 60–73

Contents lists available at ScienceDirect

International Journal of Transportation

Science and Technology
journal homepage: www.elsevier.com/locate/ijtst

Smart vehicle monitoring and assistance using cloud computing

in vehicular Ad Hoc networks
Yash Agarwal a, Kritika Jain a, Orkun Karabasoglu b,⇑
a
Trackyotech Private Limited, India
b
Department of Industrial Engineering, Yasar University, Izmır, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: The increasing number of on road vehicles has become a major cause for congestion, acci-
Received 1 March 2017 dents and pollution. Intelligent Transportation Systems (ITS) might be the key to achieve
Received in revised form 7 December 2017 solutions that help in reducing these problems significantly. The connected vehicular net-
Accepted 9 December 2017
works stream is a rapidly growing field for research and development of various real-time
Available online 20 December 2017
applications. In this paper, novel techniques have been proposed to serve the speed based
lane changing, collision avoidance and time of arrival (TOA) based localization in Vehicular
Keywords:
Ad Hoc Networks (VANETs). As GPS requires clear line-of-sight for accurate services of
Connected vehicles
Internet of Things (IOT)
positioning and localization applications, we designed a Time of Arrival (ToA) based algo-
VANET rithm for areas where strong GPS signals are unavailable. Collision avoidance using auto-
Traffic management matic braking and camera-based surveillance are a few other applications that we
Intelligent transportation addressed. The feasibility and the viability of the algorithms were demonstrated through
simulations in Simulation of Urban Mobility (SUMO) and Network Simulator-2 (NS-2).
We prototyped a working hardware and tested it on actual vehicles to assess the effective-
ness of the proposed system. We designed a mobile app interface for the on-board unit for
smart, efficient and remote traffic monitoring. The integrated VANET Cloud Computing
architecture acts as the platform for the proposed applications.
Ó 2017 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

Introduction

Vehicles nowadays come with highly intelligent and sophisticated systems to serve the real-time safety applications. The
increase in requirement and funding in the Intelligent Transportation Systems (ITS) stream has led to a great demand for
real-time applications to serve the public safety and reduce the potential hazards from these vehicles. Many countries have
already started the implementation or are planning to implement these features in their areas with the help of giant auto-
motive firms (General Motors, 2015; Johnson, 2016). Vehicular Ad Hoc Network (VANET) has a key role to play in this. VANET
is basically a subset of Mobile Ad Hoc Network (MANET) developed to establish the wireless communication system with the
vehicles (Upadhyay, 2016). The Dedicated Short Range Communication (DSRC) protocol (Upadhyay, 2016) was developed to
support the vehicular communications. The basic elements of VANET include the Vehicle-to-Vehicle (V2V), Vehicle-to-
Infrastructure (V2I), Vehicle-to-Grid (V2G) and the Vehicle-to-Anything (V2X) modes of communication. Cloud computing
is another unique featured service that takes the real-time applications to a whole new level (Mamun et al., 2012). The basic

Peer review under responsibility of Tongji University and Tongji University Press.
⇑ Corresponding author.
E-mail addresses: [email protected] (Y. Agarwal), [email protected] (K. Jain), [email protected] (O. Karabasoglu).

https://doi.org/10.1016/j.ijtst.2017.12.001
2046-0430/Ó 2017 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73 61

elements of cloud computing include Infrastructure as a Service (IaaS), Platform as a Service (PaaS) and Software as a Service
(SaaS). The cloud computing services basically help in providing the shared resources in a network over the Internet (Lee
et al., 2014). The integrated model of Cloud computing and VANET i.e., the Vehicular Cloud (Hussain et al., 2014) plays a
major role for implementing the real-time safety applications in the vehicular networks and the smart traffic monitoring sys-
tems. The flow network of the system is as follows: the vehicles moving in a segment are connected to each other and
exchange important information (viz., location, speed, etc.) among themselves. This information is further passed on to
the Road Side Units (RSU) that are installed at regular intervals of distance on the roads (Upadhyay, 2016). The RSU’s securely
pass on the data to the central server supported by the cloud computing services (Voicu et al., 2014). The central cloud server
monitors and segregates this data so that it can be accessed by the smart traffic monitoring system and corresponding
mobile devices that officers are equipped with. The Fig. 1 below depicts our proposed system architecture. RSU 1 and
RSU 2 stand for roadside units 1 and 2, respectively. Vehicular Network 1 and 2 are denoted by VN1 and VN2, respectively.

In this paper, we propose novel algorithms for the applications described below, simulate in SUMO and NS2, develop a
first level prototype, and test a few of them in actual connected vehicles.

(1) Speed based lane changing: The road segment is classified into lanes with particular speed limits with a tolerance of

10% set to them. The vehicles moving on these lanes will have to follow these speed limits. The vehicles can change
the lanes by increasing and decreasing the speed and also considering the minimum gap that is required between any
two vehicles. An emergency warning message is broadcast to the nearby vehicles and RSU’s if any violation is observed.
(2) Collision avoidance: Based on the location information exchanged among the vehicles, the minimum gap required is
calculated. In case the vehicle is moving and the minimum gap is not met, the on-board unit sends the brake signal
to the vehicle powertrain controller and the brakes are automatically applied in a proper way to avoid collision
(Ajaykumar et al., 2015; Segata and Cigno, 2013).
(3) Video/photo surveillance: The on-board unit is also interfaced with a camera that can be remotely activated from the
central cloud server also (Joshi et al., 2015). As the camera high signal is detected, the camera is activated and then
turned to the direction calculated by the cardinal points and the photo or the video of the designated event is taken.
(4) Vehicle/accident detection: In the road network, whenever a vehicle is in halt position in the middle of the road (except
for the traffic light) an emergency warning is broadcast to all the nearby vehicles and to the vehicles that newly join
the network to avoid any further collisions.
(5) Time of arrival based localization: The GPS receivers require a clear line of sight for accurate services (Abumansoor and
Boukerche, 2012). When a vehicle is in a tunnel or a shady area (areas that are not properly mapped) the GPS signals
and information is not so reliable. Moreover, the GPS receiver can also fail sometimes, or there may be some malicious
attacks that tamper with the data and at the end the data being received is not authentic and reliable. Therefore, some
other technique for localization is required in order to keep a proper track on the vehicles. To address to this problem,
we propose an algorithm based on the Time of Arrival (ToA) and the Time Difference of Arrival (TDoA) based localiza-
tion technique.
(6) Broadcasting emergency messages: The Ad Hoc On Demand Distance Vector Routing (AODV) protocol has been found to
be the most efficient (Paul et al., 2011), fast and reliable (Ashritha and Sridhar, 2015) for the smart broadcast of the
emergency warning messages (Biswas et al., 2012; Jarp et al., 2005). Therefore, we have adopted the AODV protocol
for the smart and time efficient broadcast of the emergency message in our NS2 simulations.

The rest of the paper is organized as follows: in Section ‘‘Literature review”, we have discussed the literature review and
related work done in this field. Section ‘‘Methodology” describes our methodology and gives the detailed algorithm that we
designed. The performance evaluation of the algorithm, its results and analysis have been discussed in Section ‘‘Results and
analysis”. Section ‘‘Conclusions” concludes the paper followed by future works and the limitations in Section ‘‘Limitations
and future works”.

Fig. 1. Proposed system architecture.

62 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73

Literature review

Over the past few years, VANET has been attracting more and more research for on road safety applications and comfort
for the drivers. However, the on-board computation, communication and storage are some of the concerns for various real-
time safety applications (Hussain et al., 2014). Cloud computing services have been gaining more and more popularity due to
some extraordinary features such as real-time computing, virtual data storage etc. The integrated architecture of cloud com-
puting and VANET is a sophisticated system for the real-time applications that require complex computing and large storage.
Many different models, principles and architectures have been proposed for merging VANET with Cloud Computing (Mamun
et al., 2012; Lee et al., 2014). The cloud computing architecture includes SaaS (Software as a Service), PaaS (Platform as a
Service) and IaaS (Infrastructure as a Service) that have the capability to serve potential applications such as real-time com-
puting, mobile storage, remote data access etc. (Hussain et al., 2014). These services can play a major role in solving the basic
difficulties of VANETs that include highly changing and dynamic network mobility, randomly changing path of the vehicles
etc. The vehicular cloud computing model basically has three important elements i.e., Vehicular Cloud (VC), Roadside Cloud
(RC) and the Central Cloud Server (CCS). To support the real-time safety applications in vehicular networks, our proposed
integrated vehicular cloud computing architecture might play a crucial role and the elements of this act as the supporting
pillars.
There is a necessity to structure the elements of vehicular cloud computing architecture in an efficient way. (Kakkasageri
and Manvi, 2012) have proposed the multiagent driven dynamic clustering of vehicles in Vehicular Ad Hoc Networks show-
ing the basic model for dividing the on-road vehicles into small clusters to share resources and exchange information.
(Arkian et al., 2014) have devised a three-step approach to calculate the traffic density in a road segment with the help of
multi hop communication and information exchange techniques.
In addition to the clustering based approach, smart and safe broadcast protocols were also taken into consideration by
researchers for developing a reliable system for the emergency message/warning message broadcast in the network.
(Llaster et al., 2015) studied the greedy algorithms for information dissemination in groups of autonomous vehicles.
(Voicu et al., 2014) worked on the fast and reliable broadcasting in VANET. (Paul et al., 2011) have discussed the advantages
and disadvantages of the VANET routing protocols for inter-vehicle communication. (Allal and Boudjit, 2013) have surveyed
on the existing Geocast routing protocols for VANETs and have proposed one of their own to reduce the number of message
overheads. (Malik and Bishnoi, 2015) looked into the various possible security threats in VANETs. The paper focuses on the
data dissemination, routing protocols and the safety/ emergency message broadcast to vehicles and roadside units in a vehic-
ular network. Ashritha and Sridhar (2015) and (Biswas et al., 2012) worked on efficient vehicle authentication mechanisms
with the help of the fixed on road infrastructure to establish a secure communication link among the vehicles and between
the vehicles and road side units.
With the developing models of vehicular clouds, smart message broadcast etc. many more applications like video capture
service, infotainment, remote surveillance, speed based lane changing etc. were proposed (Joshi et al., 2015; Abduljalil, 2014;
Hussain et al., 2013). GPS systems are considered to be the most accurate and reliable for the location-based services. How-
ever, the GPS systems come with a drawback, they require a clear line of sight for the accurate services. Other than the GPS,
the Radio Frequency Identification RFID assisted systems have also been proposed for the navigation purposes in VANETs.
(Cheng et al., 2012) proposed a technique for the deployment of the RFID based navigation system in VANETs. (Bajaj and
Gupta, 2012) worked on an integrated system that uses both the GPS and the RFID for vehicle tracking. The RFID systems
use the radio frequency to exchange information that can easily be disrupted. Moreover, the range of the RFID systems is
very less compared to that of the GPS systems. This makes the RFID systems more applicable in buildings and not over
the geographical areas. (Wahab et al., 2013) and (Abumansoor and Boukerche, 2012) have investigated on the non line of
sight and GPS free localization system for location verification in VANETs. (Yan et al., 2014) worked on a grid based local-
ization system in vehicular networks with linear error propagation. (Suryawanshi et al., 2015) focused on the inter-
vehicular communication assisted localization based on the previous path detection models. The GPS system uses the Time
of Arrival (ToA) based approach for localization and therefore a system that can efficiently compute the ToA based formulas
depending on the received signals form the target vehicles can be used as an alternative to the GPS system.
Automatic and emergency braking systems is another application that is being looked into for the safety applications. This
application can help in the collision avoidance for the on road vehicles. Ajaykumar et al. (2015) and Segata and Cigno (2013)
have studied and simulated the emergency braking application based on the drivers’ behavior models.
In this work, we have introduced novel techniques for the lane change assistance and the ToA based localization for con-
nected vehicles. Usually, GPS receivers provide with accurate data and are commonly used for the positioning and the local-
ization of the vehicles. Our proposed time of arrival based localization algorithm can serve as an alternative for use in tunnels
and areas where GPS data might become unavailable. The feasibility and the viability of the algorithms is demonstrated with
the help of Simulation of Urban Mobility (SUMO) and Network Simulator-2 (NS-2) software environments. We have also
designed and prototyped the first level working on-board unit and its corresponding mobile app to show the effectiveness
of our proposed approach in real world settings. The in-vehicle integration of the hardware prototype with the vehicle sys-
tem level powertrain is required for applications such as automatic braking. The proposed techniques in this paper with
some modifications might also be employed in autonomous vehicles.
 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73 63

Methodology

In this section we introduce our algorithms and the hardware design for the smart traffic monitoring system. This section
has two parts: (1) proposed algorithms, and (2) first level prototype design.

Algorithm

The algorithm designed to facilitate the speed based lane changing, vehicle detection, accident detection, automatic brak-
ing, video/photo surveillance, vehicle positioning, localization and smart traffic monitoring has been illustrated in this sec-
tion. The algorithm has been divided into three parts based on the platform and the application. The first part of the
algorithm serves the localization and the unknown vehicle positioning applications. The second part covers the lane change,
vehicle and accident detection, camera activation, emergency message broadcast, automatic braking applications. Lastly, the
third part of algorithm basically gives a brief design idea for the smart app for completing the traffic monitoring system.
The following assumption and definitions have been taken into consideration while writing the algorithm.

(1) All the vehicles moving on the road being monitored, are equipped with a working and a fully functional on-board unit
and have a valid, unique subscription to the service.
(2) The on-board unit installed in the vehicles is passing on authentic information to the roadside units and other vehicles.
(3) The GPS receivers are in hot-start condition as long as the vehicle is not in a tunnel or any area where there is no clear
line-of-sight communication.
(4) The app can be accessed only by a registered traffic police official upon entering the correct credentials.
(5) An emergency vehicle represents a vehicle with the highest priority. Ambulance, police vehicles, fire trucks, etc., come
under this category. An emergency vehicle is denoted as Ev .
(6) A minimum gap has to be maintained between any two vehicles moving on the road. For any vehicle to change its lane,
the minimum gap criteria should be met. The minimum gap is denoted as mingap . Assuming the road surface to be pla-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
nar, we estimate the mingap using the basic distance formula: mingap ¼ ðx2  x1 Þ2 þ ðy2  y1 Þ2 (1). The required min-
imum gap between two vehicles is assumed to be M g .
(7) The camera can be activated remotely form the cloud server and the smart app. The camera signal is denoted as C S .
(8) To avoid possible collisions, the automatic-braking application for the vehicle powertrain is proposed. The signal is
generated form the microcontroller based on the mingap criteria. The automatic braking signal is denoted as Ab .
(9) The lane ID for lanes is given as Li ¼ ½L1 ; L2 ; L3 ; L4 . . . the vehicle ID for vehicles is given as V j ¼ ½V 1 ; V 2 ; V 3 ; V 4 . . .. Lli and
Lui are the lower and upper speed limits of any particular lane Li . Lb is the lane buffer ID, where the lane ID of any vehi-
cle V j is stored. Sest is the current estimated speed of any vehicle V j .
(10) W t represents the warning time given to the drivers. After warning the driver to change the lane or the speed, the sys-
tem will wait for W t and then broadcast the warning message.
(11) si represents the time of arrival (ToA) of a broadcast signal at any time instance i. si;j represents the time difference of
arrival (TDoA) calculated by subtracting si and sj .
(12) The hardware unit is interfaced with a green and red LED that light up when the lane change based on the mingap is
allowed and not allowed respectively.
(13) The direction of the target (Chen et al., 2010) is evaluated on the basis of the location of the reference vehicle with
respect to it. We estimate the direction of the target vehicle and the target object with the help of the cardinal points
for tracking the unknown vehicle and recording the event with the camera respectively. Fig. 2 shows the state change
diagram and the computation algorithm of the cardinal points. The directions are abbreviated as NW = North West,
N = North, NE = North East, W = West, SP = Same Position, E = East, SW = South West, S = South and SE = South East.

Fig. 2. State change diagram for cardinal points (Chen et al., 2010).
64 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73

TOA based localization

Assume that a vehicle V 3 2 V j is moving on a clear road and suddenly gets stuck in a tunnel. Another vehicle V 2 2 V j is
tracking this vehicle to help it. Now, we assign the vehicle V 3 as the target vehicle and the vehicle V 2 as the reference
vehicle. The GPS receiver of V 3 has stopped functioning and the on-board unit is just broadcasting messages at a constant
time interval. Any vehicle in the range of the V 3 can receive this signal. The vehicle V 2 has its OBU perfectly working and
can estimate the self-location. The algorithm to track the exact location of the target vehicle from the reference vehicle is
as follows:

Algorithm 1: TOA based localization.

1: While (1)
2: let the current location of V 3 be (x, y);
3: while ðV 3 is broadcastingÞ
4: if (time elapsed or V 2 entered the tunnel)
5: stop;
6: else
7: receive the first message from V 3
8: record the simultaneous position of V 2 as ðx0 ; y0 Þ
9: record the ToA as s0
10: receive the second message from V 3
11: record the simultaneous position of V 2 asðx1 ; y1 Þ
12: record the ToA as s1
13: receive the third message from V 3
14: record the simultaneous position of V 2 asðx2 ; y2 Þ
15: record the ToA as s2
16: receive the fourth message from V 3
17: record the simultaneous position of V 2 asðx3 ; y3 Þ
18: record the ToA as s3
19: Ensure that ðx0 – x1 – x2 – x3 Þ – ðy0 – y1 – y2 – y3 Þ
20: Let ðxi ; yi Þ ¼ ½ðx0 ; y0 Þ; ðx1 ; y1 Þ; ðx2 ; y2 Þ; ðx3 ; y3 Þ
21: Let si ¼ ½s0 ; s1 ; s2 ; s3 
22: for ði ¼ 0; i < n; i þ þÞ
23: calculate TDoA ðsi;j Þ
24: si;j ¼ si  sj
25: Let si;j ¼ ½s0;1 ; s0;2 ; s0;3 
26: Record all these values and substitute them in the formula below to get the approximate values of the location of
vehicle V 3 
27: A1 ¼ 2c  x2sx 0
 x1sx 0

 0;2 0;1

28: B1 ¼ 2c  y2sy 0
 y1sy 0


0;2 0;1
 2 2 2 2 
x þy x y x2 þy2 x2 y2
29: C 1 ¼ c  s0;2  s0;1  1c  2 2s0;2 0 0  1 1s0;1 0 0

30: A2 ¼ 2c  ðx3sx0
 x1sx 0
Þ
 0;3 0;1

y3 y0 y1 y0
31: B2 ¼ c  s0;3  s0;1
2
 
x23 þy23 x20 y20 x21 þy21 x20 y20
32: C 2 ¼ c  ðs0;3  s0;1 Þ  1c  s0;3  s0;1
33: Substituting
 the values of the  above equations in the formula below, we have:
V 3 ðx; yÞ ¼ CB11 A2 C 2 A1 C 1 B2 C 2 B1
A2 B2 A1 ; A1 B2 A2 B1 (2)

Speed based vehicle-monitoring algorithm

Lets assume that the vehicle set V i is moving on the lane set of Li . The vehicles are moving with the speed Sest on these
lanes and the lane buffer ID keeps refreshing itself after every change in lane. The lane change algorithm is structured con-
sidering the specific speed tracks (the road segment divided into lanes), the current speed of the vehicle and the minimum
gap acceptance with the vehicles already present in there. The camera on the OBU can be remotely activated by the person-
nel monitoring the cloud computing server. The automatic braking signal to the vehicle power-train is transmitted based on
the distance and speed of the vehicle with respect to the nearest object (another vehicle in this case). The algorithm is as
follows:
 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73 65

Algorithm 2: Model for lane changing.

1: While (1)
Emergency message/ broadcast/ situation/ vehicle
2: Check the C S , Ab , EV ;
3: if C S is high
4: go to the camera sub routine;
5: else
6: do nothing;
7: if Ab is high
8: go to the automatic braking sub routine;
9: else
10: do nothing;
11: if EV is high
12: broadcast an emergency message to free the lane for emergency vehicle;
13: else
14: do nothing;
Position Verification
15: Extract the values of Li , Lb , Lli , Lui , Sest ;
16: alarm off; timer off;
17: display ‘‘OK” message to driver;
18: elsif Sest > Lui
19: alarm on; timer on;
20: display ‘‘Lane Change or Decelerate” message to driver (For autonomous vehicles, signal the powertrain controller)
21: check the deceleration and lane change sub routines;
22: after W t elapsed;
23: broadcast the warning message;
24: elsif Sest < Lli
25: check engine state
26: if off
27: do nothing;
28: else
29: alarm on; timer on;
30: display ‘‘Lane Change/ Accelerate” message to driver;
31: check the acceleration and lane change sub routines;
32: after W t elapsed;
33: broadcast the warning message;
Lane Change:
34: if Sest > Lui or Sest < Lli
35: check mingap
36: if mingap P M g
37: Light on the green LED indicating Lane Change Allowed;
38: else
39: Light on the red LED indicating Lane Change NOT Allowed;
40: update Lb ;
Deceleration
41: if Sest > Lui
42: Decelerate;
43: check Sest
44: calculate mingap
45: display ‘‘Sest and mingap ” to the driver;
Acceleration
46: if Sest < Lli
47: Accelerate;
48: check Sest
49: calculate mingap

(continued on next page)

66 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73

50: display ‘‘Sest and mingap ” to the driver;

Camera Subroutine
51: if C S is high
52: display ‘‘Camera on” message to driver;
53: obtain the location of the target;
54: extract self location;
55: calculate the distance and direction with the help of cardinal points;
56: turn the camera on and rotate it to point towards the target;
57: take the photo/video and send to roadside unit;
58: set the C S to low;
⁄
Automatic Braking Subroutine:
59: extract self location;
60: calculate the distance with nearest object;
61: calculate mingap ;
62: if mingap /calculated distance very less ð< M g Þ
63: set the Ab to high;
64: display ‘‘Automatic Braking Activated” message to driver;
65: send automatic brake signal to vehicle powertrain controller;
66: else
67: set the Ab to low;

⁄
The automatic-braking signal when activated, is communicated to the vehicle powertrain provided that the on-board unit is inte-
grated with the vehicle in the design phase itself.

Algorithm for smart app

Assume that a traffic police official is a monitoring the traffic of a particular area of radius 3 km. Now to monitor the entire
3 km radius on the smart app, the traffic police official will have to enter the login credentials in the app. Once verified he/she
can get access to the data of that particular area. The app is also an integrated platform that connects the police official with
the nearest emergency services that can be availed just by the press of a button on a real-time basis. The algorithm for the
smart app is as follows:

Algorithm 3: Mobile traffic monitoring.

1: While ð1Þ
2: Obtain the login credentials of the traffic police officer;
3: check the credentials with the registered users’ list;
4: if (not verified)
5: display ‘‘Error: Please check your credentials and try again” message;
6: else
7: detect location with inbuilt GPS;
8: send the request for data to the central cloud server;
9: wait for the data to load;
10: display the data to the respective sections as received from the cloud server;
11: if (Emergency button pressed)
12: display the nearest hospital;
13: display the nearest police station;
14: display the nearest fire station;
15: if (Hospital button pressed)
16: call the ambulance;
17: else
18: do nothing;
19: if (Police station button pressed)
20: call the station;
21: else
22: do nothing;
23: if (Fire station button pressed)
24: call the fire truck;
25: else
 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73 67

26: do nothing
27: else
28: do nothing
29: if (Logout button pressed)
30: logout and take to home page;
31: else
32: do nothing;

Hardware design

In this section we demonstrate our hardware prototype that runs the algorithms described previously. The on-board unit
that is to be designed is a PIC16F877A microcontroller based system with a 9600 bps GPS receiver, 433 MHz radio transcei-
ver and serial JPEG camera interfaced to it. The PIC16F877A provides us with a 256 bytes EEPROM storage memory that can
be useful in storing the important data in the on-board unit. The GPS receiver continuously feeds the data viz. location, speed,
time etc. to the microcontroller. The microcontroller monitors this data and extracts the required information received from
the GPS data (GPVTG- Speed of the object, GPGLL- Position of the object etc.) (Aprs, 2016). The extracted data is then com-
pared with the threshold levels in the microcontroller and a warning is generated when the values do not match. As dis-
cussed in the algorithm in Section 3.1, after the elapsed time, a warning message is broadcast to all the vehicles and the
roadside units in range with the help of the radio transceivers. All the broadcast data is saved in the roadside units and is
also passed to the central server supported by cloud computing. In case of any mishap or an undesired event, the camera
of the OBU is activated and the event is recorded for further monitoring purposes. The rotation of the camera is depended
upon the coordinates of the target that is to be captured. Now the traffic police official who is the in charge of that particular
area can view this traffic situation and can remotely monitor it with the help of the smart app. The smart app has an option of
calling the emergency vehicle also. Therefore, if any accident has taken place on the road, the traffic police official can call the
emergency vehicle from the nearest source just by the press of a button. The Fig. 3 depicts the schematic diagram of the com-
ponents flow chart of the on-board unit connected with the cloud server and the smart app.
We have designed the first level of the on-board unit that performs the basic applications like speed detection, vehicle
detection, location detection, emergency message broadcast and automatic braking signal generation. The Fig. 4 shows
the first level design of the OBU.
We have also designed an interface for the mobile app that can be integrated with the hardware for remote and mobile
traffic monitoring. In the working condition, the app displays a green signal to the user showing that it is active and the data
is fresh. The app displays the location information, date and time, number of vehicles moving, number of vehicles stopped at

Fig. 3. Flow chart diagram of the OBU.

68 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73

Fig. 4. First level design of OBU.

traffic light signals and the number and location of the vehicles with possible problems or complications. Emergency help
like ambulance, fire truck etc. can be called just by the press of a button in the app. The Fig. 5 depicts the sample GUI of
the mobile app.

Results and analysis

In this section, we discuss the performance of the proposed framework and algorithms with the simulations performed on
SUMO (Simulation of Urban Mobility) and NS-2 (Network Simulator-2). SUMO is a road traffic simulation suite that can sup-
port a large number of nodes in VANET. It supports a wide variety of driver models, lane models, and vehicle configurations
etc. The simulations performed using SUMO can be imported into NS2 for performing the network-based simulations (Mittal
and Choudhary, 2014). NS2 is a discrete event simulator that supports routing, multicast protocols etc. over wired and wire-
less networks (Martinez et al., 2011). This section further has two parts: (1) simulation results for the ToA based localization
and (2) the simulation results for the speed based lane changing with minimum gap acceptance and camera activation. The
SUMO map is based on the parameters as discussed below.
The road network was created with the help of the nodes and the segments. Nodes are the starting and ending points of
the segments. The road network is denoted as R ¼ fsðx; yÞ; eðx; yÞ; Li ; V j g. Here, sðx; yÞ is the starting point and eðx; yÞ is the
ending point of any lane Li with any vehicle V j moving on it.
We have done the NS2 simulations for the dynamic environment in which we can have mobile nodes and we employed
the Ad Hoc On Demand Distance Vector Routing protocol (AODV) as it was found to be the most efficient one for the safety
and emergency message broadcasting (Paul et al., 2011). The following parameters were taken into consideration for the NS2
simulations (Table 1).

Simulations for the ToA based localization

In this section we discuss the simulations performed to evaluate the proof of concept for the lane change model. We
designed an ‘I’ shaped map with 6 nodes and 10 edges in SUMO. Two cars with pre-defined routes were run on the ‘I’ shaped
map. The blue car in the simulation acts as the target vehicle and the green car is the reference vehicle. This model was then
imported into NS2 and the data of the two vehicles was collected. We then calculated the location of the unknown vehicle by
substituting this data in the formula (2) as discussed earlier with the help of Perl programming. The Fig. 6.1 depicts the
SUMO simulation for the customized map and routes. The vehicle in the top most lane of the map is stuck and requires help.
Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73 69

Table 1
Parameters used in NS2 simulations.

Parameter Value
Channel type Wireless channel
Antenna model Omni antenna
Network interface type Wireless Phy
Interface queue type Droptail
MAC type Mac/802.11p
Radio propagation model Two ray ground
Max packet in IFQ 50
Ad hoc routing protocol AODV
Number of nodes 10

Fig. 5. Sample GUI of smart app.

Fig. 6.1. SUMO simulation on the custom map for concept of Localization.
70 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73

It broadcasts the emergency message in regular intervals of time. The vehicle in the bottom lane is the reference vehicle that
is tracking the stuck vehicle. As soon as the reference vehicle comes in range of the broadcast messages, it begins to estimate
the position of the unknown vehicle with the help of our (Algorithm 1).
The Fig. 6.2 depicts the NS2 simulations of the same scenario. The node 0 is a static node and is in a fixed position that is to
be determined by node 1. The node 1 is a dynamic node with a particular speed limit and has an irregular trajectory to reach
node 0. As soon as the node 1 comes in the broadcast range of node 0, the values of location of node 1 corresponding to the
time of arrival of the broadcast signals are recorded.
The data (as computed in the Algorithm I: 21–33) from the trace files of the NS2 simulations is as listed in Table 2. We
took the four consecutive time instances when the reference vehicle received the broadcast message as the values for the
time of arrival (si ). We then calculated the values of the time difference of arrival (si;j ) by subtracting the current time of
arrival from the first value of time arrival. The location ðxi ; yi Þ of the reference vehicle was recorded corresponding to the
respective value of time of arrival. This data was then substituted in the formulas as discussed in the Algorithm 1. The com-
putation results have shown that the V 3 ðx; yÞ was found to be (200, 300) and the direction of V 3 with respect to V 2 was found
to be North East that correctly matches with the simulation.

Simulations for lane change model

In this section we discuss the simulations performed to evaluate the proof of concept for the lane change model. The Open
Street Map (OSM) (OSM, 2016) of the G4 Beijing-Hong Kong-Macau Expressway was imported into SUMO and the traffic
flow for random routing with the free lane-driving model was generated on it. The Figs. 7.1, 7.2 and 7.3 depicts the traffic

Fig. 6.2. NS2 simulation for ToA based localization.

Table 2
Data from the NS2 trace files.

Variable Value
ðx0 ; y0 Þ ð46:05; 132:11Þ
ðx1 ; y1 Þ ð52:76; 145:52Þ
ðx2 ; y2 Þ ð59:47; 158:94Þ
ðx3 ; y3 Þ ð64:05; 148:66Þ
s0 7:59295E  07
s1 7:11366E  07
s2 6:63715E  07
s3 6:7812E  07
s0;1 4:79298E  08
s0;2 9:55802E  08
s0;3 8:11755E  08
 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73 71

Fig. 7.1. Earth view and SUMO map of G4 highway.

Fig. 7.2. Lane Change model on the SUMO map of G4 highway.

Fig. 7.3. NS2 simulation of the lane change model.

model generated, the lane changing concept on the map of G4 Beijing-Hong Kong-Macau Expressway with the help of SUMO
and NS2 simulations. The Fig. 7.1 shows the geographic earth view and the SUMO model of the G4 highway respectively.
The Fig. 7.2 of depicts the lane change model on the SUMO map of the G4 highway. The three lanes on the G4 highway are
assigned with particular speed limits and every vehicle moving on these lanes will have to follow these limits. However,
these vehicles can change their lanes by increasing or decreasing the speed and following the minimum gap criteria. The
speed limit is highest in lane 1 followed by lane 2 and lane 3. The red vehicle moving on lane 2 increases its speed to shift
to lane 1. Based on the minimum gap acceptance criteria, the red vehicle is then allowed to shift to lane 1.
72 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73

The Fig. 7.3 shows the NS2 simulation of the lane change scenario. The node 0 and 1 are moving in the lane with a higher
speed limit as compared to the nodes 2, 3 and 4. The node 2 increases the speed and shifts its lane by maintaining the min-
imum gap. The shortest distance from node 2 to any other node is estimated using the distance formula (1) and that partic-
ular node is informed about the movement of node 2. The node closest to node 2 (node 0 in this case) activates its camera to
take a video or a photo of this event.

Hardware prototype

We designed the first level hardware prototype that could perform the speed detection, lane change assistance, vehicle
detection and the emergency message broadcast applications on real-time basis. We placed one unit at a stationary place
that acted as an RSU and two units were installed in two different vehicles (Niu brand electric scooters) for the field trial.
We performed three trials considering different situations on a two-lane road segment of length 900 meters. We have used
433 MHz radio transceivers for broadcasting the emergency and warning messages. The communication range between the
vehicles and the RSU is 1000 meters.
The first trial of the prototype was intended to verify the location and the lane change model following the minimum gap
criteria on the two-lane road with speed limits set to 20 km/h and 40 km/h, respectively. With the help of the data from the
RSU, we verified the position of the vehicles with respect to the lane for comparing the speed limits. The lane change model
was tested according to the Algorithm 2.
In the second trial, we tested the prototype for the emergency message and the other warning signals. One of the vehicles
was halted on the road indicating a possible problem. It was observed that on coming closer to the halted vehicle, the OBU of
the moving vehicle generated the warning signal, the camera activation signal, automatic brake signal and also sent an emer-
gency message to the RSU.
The third trial was done to assess the response of the system when it detects any other object/obstruction on the road. We
introduced another vehicle (obstruction) that was halted on the road without the prototype installed in it and the RSU was
signaled about it. When the emergency message was sent by the RSU to the vehicles with the OBU regarding the obstruction,
it was observed that the automatic brake signal and the camera signal were turned on.
With the help of the field trials, it was observed that the system was working efficiently in terms of generating the warn-
ing signals and the emergency messages on real-time basis. It was observed that the hardware could successfully perform
the lane change model, location verification and all the alert signals including the automatic braking, camera activation and
the emergency message broadcast.

Conclusions

In this work we proposed an intelligent traffic monitoring system based on cloud computing and mobile apps. We intro-
duced algorithms for real-time safety applications in vehicular networks. We investigated on the speed based lane change
model, collision avoidance on roads, accident detections, video/photo surveillance and time of arrival based localization
applications. Proof of theoretical concept of our system design has been demonstrated with the help of simulations per-
formed in SUMO and NS2. A traffic model for the G4 Beijing- Hong Kong – Macau highway was generated using OSM maps
in the SUMO software for a realistic visualization of the scenario. Using the data of the SUMO simulations, we have imple-
mented the lane changing and ToA algorithms in NS2 and verified our proposed technique.
We also developed a working prototype and performed field studies to demonstrate its workings. We introduced inter-
face of a mobile app that is to be integrated with the hardware so that the traffic can be remotely monitored. The framework
and equipment that is proposed in this study can help reducing traffic congestion, on road accidents due to improper lane
change, vehicle tracking and etc.

Limitations and future works

In this section, we discuss about the limitations of our work and the possible future work opportunities associated with it.
We assumed that the road segments are VANET approved that require functional RSU’s at regular distance intervals. How-
ever, it may not be possible that we have the infrastructure at these exact intervals due to the pre designed road structures.
Due to this, we can expect some time delay in the emergency message broadcast and the information exchange taking place
between the vehicles, RSU’s and the cloud server.
We assumed that the vehicular cloud network was fully secure and resistant to any malicious attacks. It was also assumed
that the data coming from the GPS receivers was authentic. But in real life scenario this case cannot be generalized. The
vehicular cloud server is prone to attacks (Malik and Bishnoi, 2015) for data manipulation. And therefore the vehicular cloud
network is to be secured (Yan et al., 2012) using smart and efficient techniques.
We developed the partial hardware for the system design and we intend to completely design the fully functional system
integrated with the app to implement it in the real-time scenarios. As the OBU uses the power battery, there may be some
instances where the battery is down and then OBU doesn’t function as desired. Therefore, we intend to get into optimizing
the power of the OBU and connect it with the vehicle power in order to make the system more efficient and reliable.
 Y. Agarwal et al. / International Journal of Transportation Science and Technology 7 (2018) 60–73 73

A user behavior study to the warnings from the device should be conducted to evaluate the effectiveness and the adap-
tation of technology index.

References

Abduljalil, F.M., 2014. Video capture service in the intelligent transportation system based on cloud computing. Int J Comput Appl 97 (5).
Abumansoor, O., Boukerche, A., 2012. A secure cooperative approach for nonline-of-sight location verification in VANET. IEEE Trans Veh Technol 61 (1), 275–
285.
Ajaykumar, N., Prasad, B.G., Sneha, K., 2015. Simulation of automatic emergency electronic braking system. Int J Eng Develop Res (IJEDR) 3 (3).
Allal, S., Boudjit, S., 2013. Geocast routing protocols for VANETs: survey and geometry- driven scheme proposal. J Internet Serv Inf Security 3 (1/12), 20–36.
Aprs.gids.nl, 2016. GPS – NMEA Sentence Information, [Online]. Available: <http://aprs.gids.nl/nmea/-gll> [accessed: 11.01.16].
Arkian, H.R., Atani, R.E., Pourkhalili, A., Kamali, S., 2014. Cluster-based traffic information generalization in vehicular ad hoc networks. Veh Commun, 197–
207.
Ashritha, M., Sridhar, C.S., 2015. RSU based efficient vehicle authentication mechanism for VANETs. IEEE 9th International Conference on Intelligent Systems
and Control (ISCO), pp. 1–5.
Bajaj, D., Gupta, N., 2012. GPS based automatic vehicle tracking using RFID. Int J Eng Innov Technol (IJEIT) 1 (1).
Biswas, S., Misic, J., Misic, V., 2012. An identity-based authentication scheme for safety messages in WAVE-enabled VANETs. Int J Parallel Emergent Distrib
Syst 27 (6).
Chen, T., Liu, H., Schneider, M., 2010. Evaluation of cardinal direction developments between moving points. ACM GIS’10.
Cheng, W., Cheng, X., Song, M., Chen, B., Zhao, W.W., 2012. On the design and deployment of RFID assisted navigation systems for VANETs. IEEE Trans
Parallel Distrib Syst 23 (7), 1267–1274.
Hussain, R., Abbas, F., Son, J., Kim, D., Kim, S., Oh, H., 2013. Vehicle witnesses as a service: leveraging vehicles as witnesses on the road in VANET clouds.
Cloud Comput Technol Sci (CloudCom), 439–444.
Hussain, R., Son, J., Eun, H., Kim, S., Oh, H., 2014. Rethinking Vehicular Communications: Merging VANET with Cloud Computing. Cloud Comput Technol Sci
(CloudCom), 606–609.
Jarp, S., Bechler, M., Wolf, L., 2005. Evaluation of routing protocols for vehicular Ad Hoc networks in typical road traffic scenarios. Proceedings of the 11th
EUNICE Open European Summer School on Networked Applications, pp. 584–602.
Johnson, D., 2016. Smart City Development in China | China Business Review, [Online]. Available: <http://www.chinabusinessreview.com/smart-city-
development-in-china/> [accessed 08.01.2016].
Joshi, J., Jain, K., Agarwal, Y., 2015. CVMS: cloud based vehicle monitoring system in VANETs. Proceedings of the IEEE 4th International Conference on
Connected Vehicles and Expo, ICCVE.
Kakkasageri, M.S., Manvi, S.S., 2012. multiagent driven dynamic clustering of vehicles in VANETs. J Netw Comput Appl, 1771–1780.
Lee, E., Lee, E.K., Oh, S.Y., Gerla, M., 2014. vehicular cloud networking: architecture and design principles. Commun Mag IEEE 52 (2), 148–155.
Llaster, I., Kuhlmorgen, S., Festag, A., Fettweis, G., 2015. Greedy Algorithms for Information Dissemination within Groups of Autonomous Vehicles.
Intelligent Vehicles Symposium (IV), IEEE, pp. 1322–1327.
Malik, V., Bishnoi, S., 2015. Security threats in VANETs: a review. Int J Recent Res Aspects 2 (1), 72–77.
Mamun, Md. A.A., Anam, K., Onik, Md. F.A., Efsar-E-Alam, A.M., 2012. Deployment of cloud computing into VANET to create ad hoc cloud network
architecture. Proceedings of the World Congress on Engineering and Computer Science.
Martinez, F.J., Toh, C.K., Cano, J.C., Calafate, C.T., Manzoni, P., 2011. A survey and comparative study of simulators for vehicular ad hoc networks (VANETs).
Wireless Commun Mobile Comput 11 (7), 813–828.
Media.gm.com, General Motors Promotes for Connected Vehicles (V2X) Technology Development in China, [Online]. Available: <http://media.
gm.com/media/cn/en/gm/news.detail.html/content/Pages/news/cn/en/2015/Nov/1124_v2x.html/> [accessed 13.01.16].
Mittal, N.M., Choudhary, S., 2014. comparative study of simulators for vehicular ad-hoc networks (VANETs). Int J Emerg Technol Adv Eng 4 (4).
Open Street Map (OSM), Available: <https://www.openstreetmap.org/way/183789599-map=16/39.7326/116.1169/> [accessed 12.01.16].
Paul, B., Ibrahim, Md., Bikas, Md.A.N., 2011. VANET routing protocols: pros and cons. Int J Comput Appl 20 (3), 28–34.
Segata, M., Cigno, R.L., 2013. Automatic emergency braking: realistic analysis of car dynamics and network performance. IEEE Trans Veh Technol 62 (9),
4150–4161.
Suryawanshi, S., Gupta, D., Gupta, S., Jain, S., 2015. On the hybrid augmentation of the inter-vehicular communication assisted localization using previous
path detection. IEEE Int Adv Comput Conf (IACC) 2015, 82–87.
Upadhyay, A., 2016. ‘‘Introduction to VANET”, Slideshare.net, [Online]. Available: <http://www.slideshare.net/Akanksha_Upadhyay/vanet> [accessed:
16.01.16].
Voicu, R.C., Abbasi, H.I., Fang, H., Kihei, B., Copeland, J.A., Chang, Y., 2014. Fast and reliable broadcasting in VANET Using SNR with ACK decoupling. IEEE Int
Conf Commun (ICC), 574–579.
Wahab, A.A., Khattab, A., Fahmy, Y.A., 2013. Two-way TOA with limited dead reckoning for GPS-free vehicle localization using single RSU. ITS
Telecommunications (ITST), 13th International Conference on 5 (7), pp. 244–249.
Yan, G., Rawat, D.B., Bista, B.B., 2012. Towards secure vehicular clouds. IEEE 6th International Conference on Complex, Intelligent and Software Intensive
Systems (CISIS), pp. 370–375.
Yan, T., Zhang, W., Wang, G., 2014. A grid-based on-road localization system in VANET with linear error propagation. IEEE Trans Wireless Commun 13 (2),
861–875.