Content uploaded by Rajeev Kumar Patial
Author content
All content in this area was uploaded by Rajeev Kumar Patial on May 11, 2020
Content may be subject to copyright.
RESEARCH ARTICLE
Performance enhancement of traffic information gathering
(PEnTInG) algorithm for vehicular ad‐hoc networks
Rajeev Kumar
1,2
| Dilip Kumar
3
| Dinesh Kumar
4
1
Department of Electronics and
Communication Engineering, IKG‐Punjab
Technical University, Kapurthala, India
2
School of Electronics and Electrical
Engineering, Lovely Professional
University, Phagwara, India
3
Department of Electronics and
Communication Engineering, Sant
Longowal Institute of Engineering and
Technology, Longowal, India
4
Department of Information Technology,
DAV Institute of Engineering and
Technology, Jalandhar, India
Correspondence
Dilip Kumar, Department of Electronics
and Communication Engineering, Sant
Longowal Institute of Engineering and
Technology, Longowal, India.
Email: dilip.k78@gmail.com
Summary
Vehicular ad‐hoc networks (VANETs) play a vital role in today's context of
vehicular traffic. In this paper, clusters of vehicles are created on the basis of
average speed of the vehicles. One cluster communicates with the next cluster
through a cluster head and also share the same information with next cluster
heads and installed road side units (RSUs). By using this technique, we can
solve the problem of rough driving behavior and road terrorism which is due
to speed variation of vehicles and fake information dissemination by the
drivers. Many a times, drivers may spread fake accident‐related information
into the network which is a serious cause of concern in VANETs. It is ensured
that such drivers are not allowed to spread wrong information in the network
to avoid accidents. To solve this problem, we developed performance enhance-
ment of traffic information gathering (PEnTInG) algorithm that selects only
those drivers/vehicles as cluster heads in a cluster who has maximum value
of the cluster head factor (CHF). The CHF is derived by considering different
weights in range of 0 to 1 of relative average speed, time to leave, trust factor,
and neighborhood degree. Further, the elected cluster head shares and stores
the same information with the RSUs. In case, a driver wants to disseminate
fake or wrong information in a network, then that vehicle driver can be easily
tracked by the local authority by accessing RSU data. Simulation results show
that the stability of PEnTInG is increased by 25% against the existing schemes
viz. lowest‐ID, MCMF, and cluster‐based technique.
KEYWORDS
clustering, DSRC, MANETs, RSU, VANETs
1|INTRODUCTION
Vehicular ad‐hoc networks (VANETs) are considered as one of the specialized areas of mobile ad‐hoc networks
(MANETs). Recently, intelligent transport system (ITS) applications are developed based on VANET communication
domains and protocols as shown in Figure 1. The vehicle domains comprise an OBU and AU with wireless connectivity.
The OBU provides communication link to the AU so as to execute the applications. The ad‐hoc domain consists of vehi-
cles equipped with OBUs and a station along the RSU. VANETs have the capability to improve road traffic safety and
driving efficiency. There is a possibility of single hop communication and multihop communication in VANETs.
Whereas, infrastructural domains are connected to the Internet, allowing the OBU to access the infrastructure network.
Received: 4 June 2018 Revised: 19 June 2019 Accepted: 28 June 2019
DOI: 10.1002/dac.4111
Int J Commun Syst. 2019;e4111.
https://doi.org/10.1002/dac.4111
© 2019 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/dac 1of15
OBU can also communicate with other hosts for nonsafety applications, using the communication of cellular radio net-
works such as GSM, GPRS, UMTS, HSDPA, Wi‐Max, and 4G.
1-3
VANETs have gained a lot of attention from researchers, academicians, and industry leaders, especially in USA, EU,
and Japan. Vehicular networks have a tremendous scope in safety applications, public services, improved driver behav-
ior, and entertainment applications, etc.
A large number of vehicles and lack of RSUs may become bone of contention to the pious idea of vehicular networks.
There are number of other challenges viz., security, mobility, sparse connectivity of vehicles on highways, and large
density of vehicles in the city environment. Many researchers have presented different ideas to solve the various issues
of vehicular networks, but the one that employs the division of vehicles in to clusters has developed lots of interest
among researchers.
4
To solve the above problem, an efficient clustering approach is needed along with dedicated
short‐range communications (DSRC) frequency band for effective communication among commuting vehicles and
RSUs. There is still a scope of developing a new stable clustering structure for efficient information dissemination
among the vehicles in VANETs.
1.1 |Our contribution
The main objective of this paper is to create a new stable cluster structure by observing drawbacks of existing
approaches. We consider a road segment that can be occupied by two types of vehicles in a given time:
i. Vehicle whose frequency of occupancy on a road segment is maximum or commercial vehicle or office goers in a
given time slot, and
ii. Vehicle whose frequency of occupancy on this road segment is minimum or not a local vehicle or long route vehicle
that does not use this road segment quite often.
It is also assumed that all the vehicles are equipped with global positioning system (GPS) module and compliant with
IEEE802.11p.
Let us say, a low‐frequency vehicle somehow manages to become a cluster head (CH) and propagates fake accident
information on the network, then it will be very difficult for police to track a vehicle that is not a frequent traveler of
this road segment. To solve this problem, PEnTInG algorithm is developed that selects a CH in a cluster of vehicle with
maximum value of the CHF. This research contribution ensures that routine vehicles should be selected as a CH; this
helps to improve cluster structure stability by 25% as compared with the existing state‐of‐the‐art approaches.
FIGURE 1 Communication domains
and communication types in VANET.
1-3
(AU: Application unit, GW: Gateway unit,
OBU: On board unit, HS: Hot spot, RSU:
Road side unit)
2of15 KUMAR ET AL.
1.2 |Organization
This paper is organized as follows: Section 2 presented an analysis on the research contributions of various authors in
related areas. Section 3 discussed about architecture of PEnTInG algorithm using flow diagrams. Result analysis and
simulations with existing techniques are presented in Section 4, and Section 5 briefed about conclusion, future scope,
and research gaps of the PEnTInG algorithm.
2|RELATED WORK
A broad guideline to various types of applications is tabularized by Schoch et al.
1
A number of information dissemina-
tion techniques have been developed for VANETs.
5
A brief review of some of these techniques is given below.
2.1 |Traffic information systems
Two types of information system structures such as structure and structure free aggregation are discussed in Fonseca
and Vazão
6
and Kumar and Dave.
7
Wischhof et al
8
proposed a fixed hierarchy traffic information system based on aver-
age speed. Each vehicle is equipped with GPS receiver, digital map, simple digital radio, and a small data processing
unit. It is claimed that, as distance keeps on increasing, accuracy of the information keeps on decreasing, but in support
no mathematical or practical analysis is provided. Nadeem et al
9
proposed that nearby nodes data were aggregated
together, and the ID of the road segment was not aggregated. Authors have reduced the size of stored records to 50 bytes
per record. So, memory limitation was not a concern. This was a structure free aggregation technique. Caliskan et al
proposed different segments of parking spots that were compared with a road segmentation approach in Caliskan
et al.
10
This approach is found similar to a tree‐based approach for road segmentation. This was a hierarchical aggrega-
tion approach that takes data input based on a quad tree structure applied on a realistic model of a German City. This
approach needed to elaborate more on periodic time interval of beacons. The relevance of the information to be prop-
agated was calculated on the basis of distance and age of the resource. Lochert et al
11
used landmarks instead of quad
tree structure, and it was difficult to implement parking space application in fixed hierarchy quad tree structure. Sleet
et al
12
proposed a cluster‐based location service management scheme for VANETs. It was indicated that high efficiency
of location‐based queries with minimum overhead was needed to track the locations of mobile nodes. Lee et al
13
pro-
posed density estimation for rural, highway, and urban scenarios. Authors also took care of redundancy of information.
Cluster key was generated to ensure the identity of vehicles. Bilal et al
14
proposed that fixed segment length was not
good for better density estimation because the range of communication of all vehicles is in a circle. For any two consec-
utive segments, there was an area of the road that is not covered in any of the segments. Ucar et al
15
proposed multihop
structure‐based aggregation for data to be disseminated further. Authors claimed the superior performance of vehicular
stable cluster‐based data aggregation (VeSCA) in comparison to previously existing
12
cluster‐based data aggregation
methods. It was observed that cluster stability played an important role in aggregation ratio. Ucar et al
16
addressed
the issue of broadcast storm and disconnected network problem at high and low vehicle densities, respectively. This
work combined vehicular multihop algorithm for stable clustering‐long term evolution (VMaSc‐LTE) and IEEE
802.11p for multihop clustering which needs further investigation for urban scenario. Similar hypothesis was also pre-
sented in research contributions.
17-21
Das et al
22
proposed a solution to take care of connectivity issues in VANET. The
geographical routing technique follows greedy algorithm in which each node has information of its position and posi-
tion of its immediate neighbor. Hence, optimal path is selected for propagation of the message. The decision to join or
leave the message transfer process is taken on the basis of incentive received by the node. Network performance is
improved by 40% with reduction of message propagation delay. The technique does not work to identify different types
of acceptable and unacceptable driver behaviors. D'Andrea et al
23
developed a real‐time detection system to fetch traffic
information from Twitter stream. It has been found that people write local language in English that conveys no mean-
ing in English. Such cases cannot be identified through the technique provided, and authenticity is also doubtful.
Mouna Elloumi et al
24
developed a traffic monitoring system using multiple unmanned aerial vehicles. Accuracy of
implementation is certainly improved, but cost of implementation is also increased. Roy et al
25
suggested safety as a ser-
vice to the vehicles on the road but at a cost to drivers, and authenticity of the information is also doubtful. Rath et al
26
identified solutions for metropolitan cities where people deal with traffic congestion problems every day. Authors pro-
posed a traffic management system to identify the traffic congestion and to regulate the traffic using a mobile agent and
KUMAR ET AL.3of15
internet of things. Rath et al
27,28
proposed a nano robot with an Intelligent Swarm Smart Controller (ISSC) module to
divert the traffic away from congestion points identified during traffic jam situation depending upon predefined thresh-
old of congestion level. Authors also discussed soft computing techniques based on approximation and probability and
also highlighted aggregation errors when discussed models are not followed. Rath et al worked on these techniques for
wireless sensor networks.
29
2.2 |Clustering techniques
Nowadays, VANETs are upcoming wireless networks where information is disseminated to a set of vehicles and also
shared with nearby RSUs as shown in Figure 2.
21
Vehicle has been communicating through RSUs or CHs. Vehicles and RSUs needed to be equipped with dedicated
hardware to provide safety, security, and infotainment to the people in the vehicle. Therefore, a number of standardi-
zation efforts were taking place in the last decade through research initiatives on VANETs, both at algorithmic level
as well as protocol development efforts such as IEEE 802.11p and IEEE 1609.1‐4 standards. Exact operational details
of all infotainment applications are not yet standardized for many of the vehicular applications.
Wang et al
30
proposed a cross layer algorithm wherein authors collected data through hierarchical and geographical
data collection methods. Here, election of CH was based upon priority decided by calculating estimated travel time left
and speed deviation on a particular road segment. The idea was to create a cluster of vehicles on a given road segment.
Fan et al
31
proposed utility function consisting of position, velocity, and acceleration parameters to measure vehicle
dynamism for previously available traffic. The author extended the existing definition of spatial parameters which
was initially proposed by Bai et al.
32
Here, a vehicle calculated its factor is called cluster relation (CR), and a vehicle
with the highest CR value is elected as CH. Maslekar et al
33
proposed a new cluster‐head election policy based upon
threshold distance (TH
distance
) and radio range parameters for direction‐based clustering algorithm.
34
Authors have been
able to maintain cluster stability thereby helping to achieve accurate density of vehicles. Authors ensured stability to
take care of the negative effect of overtaking vehicle. It has also been found that by reducing the radio range of commu-
nication up to a predefined value, traffic volume estimation accuracy can be increased. This needed a detailed analysis
in highway and city scenario. Wolny
35
optimized the existing distributed and mobility matrix‐based approach presented
in Basagni
36
to correctly represent the mobility matrix. The idea was problems such as cluster instability and re‐
clustering become difficult to handle when vehicles move in different directions. The algorithm was based upon beacons
and forms k‐cluster. This was achieved with the addition of Time‐To‐Live parameter in beacons. But multihop cluster-
ing drastically reduced the accuracy of information. Modified‐DMAC also estimated the lifetime of the established con-
nection (called freshness in DMAC) of two moving nodes. The performance of this method remained to be tested in
sparse and jammed traffic scenarios. From the abovementioned proposals, the protocols based upon positions were
found to be the most adequate to vehicular networks due to their ability to handle the variability in node position. Main-
taining secure communication in vehicular networks is also very difficult and is addressed by Rath et al.
37-39
Stable CH is required for effective communication between the nodes in the network. Various researchers contrib-
uted in this direction to obtain a stable CH. Parameters used in most of the techniques are signal strength received, node
coordinates, average speed of the nodes, direction, and destination of the node. The need of the hour is to create a strong
and trustworthy CH so that fake information is not disseminated on the road segment.
FIGURE 2 Overview of data gathering
system
21
4of15 KUMAR ET AL.
3|CLUSTERING METHODOLOGY
3.1 |Approach formulation
We assume a highway scenario wherein all vehicles are equipped with GPS module and storage device. All vehicles are
compliant with IEEE 802.11p transceivers. The GPS module will provide the position coordinates of the vehicle and will
also store information about number of times a vehicle has run on a given road segment in a given time slot. It is
observed that vehicles that are passing through the same area will gather similar kind of information. Hence, we divide
the vehicles into clusters, and CH will be responsible for propagation of information. So, we will have two types of vehi-
cles, cluster member (CM) and CH. The CH would be responsible for collecting this information from CMs and aggre-
gate it and pass on to RSU and next CH. In this way, a communication chain between various clusters will be created.
CH will have a storage capability to support the data that is to be delivered to RSU and will also help to maintain a
record of number of times a vehicle has crossed the same road segment. RSU consists of a road side cloud to have a local
server, all RSUs will be connected to internet, and its data can be shared with government or public transport systems,
as and when needed. This cloud will help us to compile the data of traffic density in a given time slot.
3.2 |Cluster formation
Vehicles under consideration create a dynamic homogeneous group of clusters on the basis of distributed parameters
viz. time to leave, relative average speed, neighborhood degree, and trust factor (TF). The group of homogeneous vehi-
cles has similar speed levels. The clusters moving along with high speed vehicles result in relatively stable topology as
long as the velocity of the vehicles is more or less same in a cluster.
When a new vehicle V
new
enters into the road segment, first it would search for any available cluster by broadcasting
request to join cluster M
RJC
. If the new vehicle has not received any response after waiting time t
w
, then it initiates the
process of cluster formation and broadcasts M
in
message to form its own CMs. To build neighborhood relationship mes-
sage, M
hello
is broadcasted to other vehicles within the communication range of V
new
.M
hello
message comprises current
speed and position of the vehicle. A primitive group of clusters is formed by moving vehicles in the same direction and
in the vicinity of each other, as illustrated in Figure 3 and detailed algorithm in Figure 4. However, in certain regions of
road segment, the speed levels of vehicles might be very high. Therefore, they are not suitable to be included in a single
cluster. Thus, the creation of small clusters consisting of few vehicles is prevented by considering a member's
threshold M
th
.
First, V
new
compares the speed difference with all its neighbors with a threshold parameter ΔS
th
for selecting its CMs.
A primitive group member is selected if the speed difference of the corresponding neighbor of V
new
is less than the ΔS
th
.
In case this parameter is similar for a number of vehicles, then those vehicles will be part of the cluster.
In case the number of CMs is more than M
th
, then V
new
broadcast cluster message M
cls
to its members to notify its
identification. Otherwise, it discards the cluster formation process and waits for t
w
.The nonclustered vehicle members
react after receiving the M
cls
message and set the CH ID temporarily in the next step. Table 1 indicates all types of sym-
bol and message types used in PEnTInG algorithm.
3.3 |Procedure for election of cluster head
We propose a new CH election algorithm that elects strong and stable CHs to avoid false information dissemination in
the network. The information of CH election of any node is limited to the node that is within its vicinity or in transmis-
sion range. The node priority to become a CH is decided by the value of its CHF. Thus, initially, each node starts
FIGURE 3 Cluster formation, circles in left side indicated unclustered nodes, right side indicates groups formed between nodes
KUMAR ET AL.5of15
calculating its CH factor to become a CH and after that broadcasts M
CHF
messages that contain their CHF values. A
special case to include discarded nodes in the clusters is indicated in Figure 5.
As indicated in Figure 6, every node will cast its vote M
vote
for any other nodes in its range having maximum CHF.
Once the CH election process is completed, then the elected node acknowledges its CH status by sending an acknowl-
edge message M
AK
to its neighbors. Subsequently, the neighboring vehicles update the cluster ID of the new CH.
3.4 |Procedure of cluster maintenance
There are many cluster formation techniques available, but we require a cluster maintenance technique that can with-
stand the topology changes caused by frequent joining and leaving of cluster nodes/vehicles. So, it is very difficult to
FIGURE 4 Cluster formation
6of15 KUMAR ET AL.
ensure that CH stays for a longer duration. The maintenance technique consists of three different scenarios
40
: cluster
joining, leaving, and merging. When a new vehicle joins the cluster, it sends a membership request to the existing
CH so as to become the member. Cluster head checks its relative speed with the threshold of the cluster; if so, then
the CH will accept the vehicle as its member, and hence update ID of the vehicle as the CH loses the contact with
its member when it moves out of the cluster. Thus, CH update its CM list. If two CHs have same relative speed and
are in the transmission range of each other, then the CH with a lower value will give up its CH status and becomes
a CM of other one.
3.5 |Cluster head factor
The stability of the cluster structure is defined by the CH factor. As we know, an elected CH is required to stay con-
nected with the CMs for the long duration; thus, the vehicles having a higher CH factor value are qualified for winning
the dynamic status of CH. The CHF is derived by considering four matrices given by Equation (1); it enables the selec-
tion of CM as CH only in case a vehicle is a frequent flyer on a particular road segment. This frequent flyer selected as
CH is expected to stay on this particular road segment for a long time. CHF is calculated on the basis of four metrics as
given below:
TABLE 1 Symbol and message types
Message Description
t
w
Response time from cluster head
V
new
Leading vehicle entering to new road segment
V
nbr
Vehicle in the transmission range of leading vehicle
S
nbr
Speed of V
nbr
S
init
Speed of V
new
δS
th
Speed difference with neighbors
M
in
Initiate to create a cluster
M
hello
Current position and speed data of nodes
M
th
Number of member's threshold
M
RJC
Request to join cluster
M
cls
Notification to other members upon becoming a temporary cluster head
M
CHF
Message of cluster head factor of vehicle
M
AK
To acknowledge being selected in the cluster
M
vote
Voting right of vehicle
V
i
Vehicle in cluster member list
CHF
max
Maximum value of cluster head factor
CHF
i
Value of cluster head factor of vehicle i
CHF ¯VAL(V
i
) Method to calculate cluster head factor of V
i
using Equation 1
A Temporary state
FIGURE 5 Square represents a cluster head; hollow circle represents primitive group that is yet to form a cluster
KUMAR ET AL.7of15
CHF ¼d1×TTleave þd2×RAS þd3×ND þd4×TF (1)
where, d
1
,d
2
,d
3
,and d
4
are weights whose values vary from zero to one (d
1
+d
2
+d
3
+d
4
= 1) and can be decided by local
agency based on member behavior and road condition.
Time to leave (TT
leave
) parameter indicates time for which a vehicle is present on the given road segment. Parameter
TT
leave
is decided by the current location obtained from GPS estimates. Parameter TT
leave
ensures to select CH with con-
siderable distance to the end of the current road segment. This metric contributes in enhancing the lifetime of cluster
and is computed by Equation (2).
TTleave ¼L−D
ðÞ
DtD(2)
where, Lis the length of the road segment, Dis the distance remaining to be covered by the vehicle, and t
D
is the time
required by the vehicle to cover distance D.
It is our necessity to use normalization so that one parameter does not dominate other factors present in CHF. The
vehicle will be moving on the road at a certain speed with respect to the neighboring vehicle. This is called relative aver-
age speed (RAS). This parameter helps us to identify vehicles having similar speeds and ready to be part of a group of
vehicles. Such vehicles are always near to each other on the road. The driving pattern of drivers of such vehicles is
awarded with a certain value beta (β). In case a driver is found driving with a speed variation above a certain threshold,
then such case is penalized by decrementing the value of RAS by β.
FIGURE 6 Cluster head election
8of15 KUMAR ET AL.
RAS t þ1
ðÞ
¼RAS t
ðÞþβ;St
ðÞ−Savg
≤δSth and RAS t þ1
ðÞ
¼RAS t
ðÞ−β;∣St
ðÞ−Savg∣>δSth (3)
where, RAS(0) = 1 and β= 0.01 in our simulations.
The vehicle whose neighborhood degree is better is more in the communication range of the vehicles and has higher
probability of becoming a stable CH. If a vehicle has speed difference with respect to its neighbor that is less than δS
th
,
this infers that the neighborhood degree is high.
In case the speed of the vehicle is a random variable, then a normal distribution curve is followed to calculate the
speed of the vehicle with mean m and variance (σ
2
) as given in May
41
:
fssðÞ¼ 1
σffiffiffiffiffiffi
2π
pe−
s−mðÞ
2
2σ2
:(4)
Two neighboring vehicles with speed difference ΔSalso follow the normal distribution with probability distribution
function as under in Kakkasageri and Manvi
42
:
fΔsΔsðÞ¼1
2σ2
Δs
e−
Δs−mΔsðÞ
2
2σ2(5)
where, Δs¼s1−s2;mΔs¼m1−m2;and σ2
Δs¼σ2
1þσ2
2:
In order to avoid high variation in the speed of neighbors, the threshold is set as function of
2s
σ. The same is indi-
cated in Rawashdeh and Mahmud
43
;asσ
Δs
increases, f
Δs
will decrease.
A vehicle whose frequency of visiting the same road segment is high, then TF is high. We do not want a nonregular
traveler vehicle to be selected as a CH; only the vehicle that moves regularly on to the road is more probable to be
selected as CH. There is very less probability that vehicles with high frequency of travel will send forged information
to the users, as they will be easily caught by the police. If nis the total number of vehicles (measured through road side
cloud) on the road segment, fis the number of times a vehicle has traveled on a given road segment (L), then
TF ¼1
n
f
L
:(6)
3.6 |Chaining of clusters
The road side cloud is responsible for measuring the number of vehicles in a given road segment. We created a virtual
chain between clusters in order to obtain the vehicular traffic information received from clusters. The vehicular message
propagation encounters many hindrances for communication between nodes such as scattering, reflection diffraction,
and even blocking of radio signals. Signal fading in VANET can be characterized for short as well as long distances
according to Rician and Rayleigh distributions, respectively.
44-46
The Nakagami distribution model fits well for VANETs because its parameters can be adjusted for different ranges of
communication. Figure 7 indicates ras a service channel transmission range and Ras a control channel transmission
range. Received signal power is represented using probability density function for Nakagami distribution as given
below:
PZ2xðÞ¼ m
Pr
nxm−1
ΓmðÞ
e
−mx
Pr;x≥0 (7)
FIGURE 7 Dedicated short range communication (DSRC)
KUMAR ET AL.9of15
where, Γ(.) is gamma function, P
r
is average received power, ptK
rα;ris distance in meters, αis path loss exponent, and
K¼GtGrC
4πfc
no
2, where cis speed of light, f
c
is career frequency 5.9 GHz, G
t
,G
r
are transmitted and received antenna
gains respectively, and mis fading factor.
DSRC uses 75‐MHz bandwidth which is divided into seven channels of equal width. It has six communication chan-
nels and one control channel. The control channel has more signal range in comparison to communication channels for
vehicles. The control channel is used by governing authority.
3.7 |Traffic information generalization
Here, we are calculating traffic volume for the vehicles that are equipped with GPS module and normal vehicles. Let us
consider that the number of vehicles with GPS module is α%. Hence, the vehicles with GPS module are β= (100 −α)%
but not compliant with 802.11p. There exists a virtual cluster of vehicles with each real cluster having 802.11p compliant
vehicles. Thereby, we will calculate the total volume of vehicles in position of real cluster.
47
Let “p”be the number of
real clusters, and virtual cluster “i”has no associated vehicle(i∉p); then, by using the volume data of all real clusters,
we need to calculate the volume of cluster i. Here, similarity between virtual cluster iand real cluster jis given by
S1i;jðÞ¼ 1
dis i;jðÞ
;where dis i;jðÞis the distance between two different cluster types i;j
S2i;jðÞ¼ 1
sav i;jðÞ
;where sav i;jðÞis the average speed difference between virtual and real cluster
S3i;jðÞ¼ 1
Num i;jðÞ
;where Num i;jðÞis the difference between relative real cluster of virtual cluster i and real cluster j:
Now, assimilated similarity measure is T
o
(i,j) is given by
Toi;jðÞ¼S1i;jðÞ
u1 ×S
2i;jðÞ
u2 ×S
3i;jðÞ
u3
;where uk indicates whether Skis to be considered for assimilated
similarity or not;wherein;k¼1;2;3(8)
uk ¼1;if measure Skto be considered;otherwise;0:(9)
For all the pclusters, we will calculate the traffic volume of cluster iby weight coefficients of 802.11p compliant
vehicles. The weight coefficient of cluster jto estimate virtual cluster i,w
i
(j) is normalized aggregated similarity for all
vehicles of pclusters and is shown below:
wijðÞ¼ Toi;jðÞ
∑∇kpToi;kðÞ
∃j∈p:(10)
The traffic volume of iis the weighted average of all real clusters using the weight w
i
(j), ∀j∈pis
ΘiðÞ¼α
δ×∑∀j∈pwijðÞ×θjðÞ (11)
where
,
θ(j)is the traffic volume of equipped vehicles of jcluster.
4|SIMULATIONS AND RESULT ANALYSIS
4.1 |Simulation model
In‐depth analysis is conducted for performance evaluation of approach through MATLAB using parameters of Table 2.
We have analyzed the performance of our approach for different vehicle densities. We used several parameters for
10 of 15 KUMAR ET AL.
analysis, such as separation between vehicles, flow rate, and density. Flow rate is the number of vehicles crossing an
area at a given time, in unit vehicles/h.
48
Vehicular occupancy per unit area is the density of vehicles, where the mea-
surement unit is vehicles/km. We are taking a case where vehicle flow rates are low, medium, and high for 90, 180, and
270 vehicles in 180 seconds, respectively. Vehicle speed follows a normal distribution curve. Speed limits on the high-
way vary from 60 to 120 km/h.
To avoid high speed variation of vehicles on the road, we assume that δS
th
speed variations of the vehicle help to
determine the speed threshold. The normal distribution curve is shown in Arkian et al
21
and has speed distribution
around mean of 100 km/h and σ= 2.77.
Figure 8 shows the evaluation model, which is based on VANET architecture. It consists of the road side units
(RSUs), moving vehicles, and also the clustering based on the connections from the vehicle to the RSUs. The cluster
selections can be seen from the above figure among the red, green, and blue clustered vehicles for the particular time
stamp. This structure is the developed model in the MATLAB environment which is used for the performance evalua-
tions of the VANET network in terms of the CH durations and cluster stabilities.
In the proposed approach, the clustering is chosen in the hierarchical manner to have the communication among the
RSUs in the ordered manner without any disorders which is an efficient and effectual approach of broadcasting infor-
mation or packets.
4.2 |Clustering performance
Strong CH selection is the major objective of this work. Different metric values are taken to evaluate the performance
for selection of the CH technique. We have presented the comparative analysis of PEnTInG with three algorithms, such
as clustering,
21
lowest‐ID,
49
and MCMF.
50
The survival time of the CH on a given road segment is related to CH dura-
tion. This head duration parameter helps us to evaluate stability of the clustering technique. Figure 9 illustrates the CH
time under different flow rates. Hence, CH time depends upon the dynamic nature of vehicle speeds. As vehicle speed
increases, CH becomes less stable and vice versa.
Figure 9 shows the CH duration with respect to the speed of the vehicles. As due to high mobility and dynamic
nature of the vehicles, it is very difficult to manage the routing among the RSUs and the vehicles to respond to changes
in topology. So, as vehicle speed increases, the cluster performance starts decreasing. And it can be noticed from the
above discussed result that the network performance has a great impact on the flow rate of the vehicles and the packet
survival time decreases for the selected CH at that point as the speed of the vehicle increases. Results indicated that as
the speed of the vehicle varies from 60 to 120 km/h, the duration of the CH increases by about 25% for the PEnTInG
algorithm, and also, as vehicle density increases, clusters get merged, and this leads to reduced CH duration. In the
FIGURE 8 Simulation scenario
TABLE 2 Simulation parameters
Parameter Value
Simulation time 180 s
Highway length 3000 m
Speed of vehicles 60 to 120 km/h
Transmission rate 6 Mbps
KUMAR ET AL.11 of 15
PEnTInG scheme, the duration of the CH is improved as indicated in Figure 9, and re‐clustering requirement is less
because CH duration is more. Hence, the number of vehicles involved in re‐clustering is also reduced. It is indicated
that re‐clustering is reduced and CH duration is improved in comparison to three existing state‐of‐the‐art methods.
In order to study the effect of TF on CH duration, the length of the road segment is kept at 2 km. It has been noticed
that as Lis reduced to 2 km or transmission range is reduced, the stability of the cluster is improved, and cluster stability
is minimum in case of high flow rate.
Another important performance metric is formation of the number of dynamic clusters. A large number of formation
of clusters is not desirable. The above figure shows the formation of the clusters for different flow rates. So, it is having
high impact in the VANET network. It is because when the number of cluster increases, then there is a heavy chance of
colliding packets from one spectrum to another spectrum which is a huge problem of increasing the overhead and col-
lisions in the clusters among the CHs. So, the number of clusters must be reduced to have less overhead and congestions
among the communications of the vehicles to the RSUs for the effectual information exchanging procedure. It has been
found that in the PEnTInG scheme, clusters of large size are created, and hence numbers of clusters are small. But as
the speed of the vehicles goes on increasing, the number of clusters also increases due to re‐clustering as shown in
Figure 10. And also to review the effect of TF on the number of clusters, Lis reduced to 2 km; then, the number of clus-
ters is also reduced. The impact of TF on the number of clusters is due to two changes in the algorithm: (a) new vehicles
on the road segment check neighbors for clustering at first and (b) the method does not entertain creation of small
clusters.
Figure 11 indicates that packet loss increases as the speed of the vehicles increases from 60 to 120 km/h; hence, the
number of packets is reduced from 4 × 10
4
to 2.4 × 10
4
. As the velocity of the vehicles is increased, multiple operations
of cluster formation and a vehicle leaving a cluster would be taking place. In this process, CHs will not find a CH nearby
to share the information packet, and due to time lag packet will be lost. In the whole process, a large number of packets
will be lost, and the only solution is to carry out exchange of packets at lower speed or increase the number of RSUs.
FIGURE 9 Cluster head duration: A, low flow rate; B, mid flow rate; C, high flow rate; and D, PEnTInG at L = 2 km
12 of 15 KUMAR ET AL.
5|CONCLUSION AND FUTURE SCOPE
In this work, we have presented a traffic information system that provides correct data to the drivers and RSUs. Here,
the driver selected as a CH is selected if he is a frequent flier on the given road segment. And the probability of driver
bullying is minimum. Because such frequent fliers can be easily caught by the governing authorities. This claim is illus-
trated in result analysis as well. There are certain issues that still remain to be investigated in this work. It has been
proved that the cluster structure developed is even stronger, but this is only applicable in highway scenario where there
is no shortage of vehicles. In case of hilly terrain or area of less vehicle density, we need to work on other solutions.
Then, memory and computation overhead is another issue that needs further investigation.
ORCID
Dilip Kumar https://orcid.org/0000-0002-0826-2086
REFERENCES
1. Schoch E, Kargl F, Weber M, Leinmuller T. Communication patterns in VANETs. IEEE Commun Mag. 2008;46(11):119‐125.
2. Bali RS, Kumar N, Rodrigues JJ. Clustering in vehicular ad hoc networks: taxonomy, challenges and solutions. Vehic Commun.
2014;1(3):134‐152.
FIGURE 11 Number of packets in PEnTInG
FIGURE 10 Number of clusters in PEnTInG: A, L = 3 km and B, L = 2 km
KUMAR ET AL.13 of 15
3. Salhi I, Cherif MO, Senouci SM. A new architecture for data collection in vehicular networks. In: 2009 IEEE International Conference on
Communications. IEEE; 2009, June:1‐6.
4. Al‐Sultan S, Al‐Doori MM, Al‐Bayatti AH, Zedan H. A comprehensive survey on vehicular ad hoc network. J Netw Comput Appl.
2014;37:380‐392.
5. Dietzel S, Petit J, Kargl F, Scheuermann B. In‐network aggregation for vehicular ad hoc networks. IEEE Commun Surv Tutorials.
2014;16(4):1909‐1932.
6. Fonseca A, Vazão T. Applicability of position‐based routing for VANET in highways and urban environment. J Netw Comput Appl.
2013;36(3):961‐973.
7. Kumar R, Dave M. Knowledge based framework for data aggregation in vehicular ad hoc networks. In: International Conference on Com-
putational Intelligence and Information Technology. Berlin, Heidelberg: Springer; 2011, November:722‐727.
8. Wischoff L, Ebner A, Rohling H, Lott M, Halfmann R. SOTIS‐a self‐organizing traffic information system. In: The 57th IEEE Semi‐annual
Vehicular Technology Conference, 2003. VTC 2003‐Spring. (Vol. 4 IEEE; 2003, April:2442‐2446.
9. Nadeem T, Dashtinezhad S, Liao C, Iftode L. Trafficview: a scalable traffic monitoring system. In: IEEE International Conference on
Mobile Data Management, 2004. Proceedings. 2004. IEEE; 2004:13‐26.
10. Caliskan M, Graupner D, Mauve M. Decentralized discovery of free parking places. In: Proceedings of the 3rd International Workshop on
Vehicular Ad Hoc Networks. ACM; 2006, September:30‐39.
11. Lochert C, Scheuermann B, Wewetzer C, Luebke A, Mauve M. Data aggregation and roadside unit placement for a VANET traffic infor-
mation system. In: Proceedings of the Fifth ACM International Workshop on Vehicular Inter‐Networking. ACM; 2008, September:58‐65.
12. Saleet H, Basir O, Langar R, Boutaba R. Region‐based location‐service‐management protocol for VANETs. IEEE Trans Veh Technol.
2009;59(2):917‐931.
13. Lee B, Jeong E, Jeong Y. Message propagation based on three types of density classification for smooth and secure vehicular traffic flow.
Int J Multimedia Ubiquitous Eng. 2014;9(12):383‐404.
14. Bilal SM, Khan SU, Madani SA, Nazir B, Othman M. Road oriented traffic information system for vehicular ad hoc networks. Wirel Pers
Commun. 2014;77(4):2497‐2515.
15. Ucar S, Ergen SC, Ozkasap O. VeSCA: vehicular stable cluster‐based data aggregation. In: 2014 International Conference on Connected
Vehicles and Expo (ICCVE). IEEE; 2014, November:1080‐1085.
16. Ucar S, Ergen SC, Ozkasap O. Multihop‐cluster‐based IEEE 802.11 p and LTE hybrid architecture for VANET safety message dissemina-
tion. IEEE Trans Veh Technol. 2015;65(4):2621‐2636.
17. Kumar R, Dave M. A framework for handling local broadcast storm using probabilistic data aggregation in VANET. Wirel Pers Commun.
2013;72(1):315‐341.
18. Lochert C, Scheuermann B, Mauve M. Probabilistic aggregation for data dissemination in VANETs. In: Proceedings of the Fourth ACM
International Workshop on Vehicular Ad Hoc Networks. ACM; 2007, September:1‐8.
19. Lochert C, Scheuermann B, Mauve M. A probabilistic method for cooperative hierarchical aggregation of data in VANETs. Ad Hoc Netw.
2010;8(5):518‐530.
20. Flajolet P, Martin GN. Probabilistic counting algorithms for data base applications. J Comput Syst Sci. 1985;31(2):182‐209.
21. Arkian HR, Atani RE, Pourkhalili A, Kamali S. Cluster‐based traffic information generalization in Vehicular Ad‐hoc Networks. Vehic
Commun. 2014;1(4):197‐207.
22. Das B, Misra S, Roy U. Coalition formation for cooperative service‐based message sharing in vehicular ad hoc networks. IEEE Trans Par-
allel Distrib Syst. 2015;27(1):144‐156.
23. D'Andrea E, Ducange P, Lazzerini B, Marcelloni F. Real‐time detection of traffic from twitter stream analysis. IEEE Trans Intell Transp
Syst. 2015;16(4):2269‐2283.
24. Elloumi M, Dhaou R, Escrig B, Idoudi H, Saidane LA. Monitoring road traffic with a UAV‐based system. In: 2018 IEEE Wireless Commu-
nications and Networking Conference (WCNC). IEEE; 2018, April:1‐6.
25. Roy C, Roy A, Misra S, Maiti J. Safe‐aaS: decision virtualization for effecting safety‐as‐a‐service. IEEE Internet Things J.
2018;5(3):1690‐1697.
26. Rath M, Pati B, Pattanayak BK. Mobile agent‐based improved traffic control system in VANET. In: Integrated Intelligent Computing, Com-
munication and Security. Singapore: Springer; 2019:261‐269.
27. Rath M, Pati B. Communication improvement and traffic control based on V2I in smart city framework. Int J Vehic Telemat Infotainment
Syst (IJVTIS). 2018;2(1):18‐33.
28. Rath M. Smart traffic management system for traffic control using automated mechanical and electronic devices. In: IOP Conference
Series: Materials Science and Engineering (Vol. 377, No. 1, p. 012201 IOP Publishing; 2018, June.
29. Rath M, Pati B, Pattanayak BK. Relevance of soft computing techniques in the significant management of wireless sensor networks. In:
Soft Computing in Wireless Sensor Networks. Chapman and Hall/CRC; 2018:75‐94.
30. Wang Z, Liu L, Zhou M, Ansari N. A position‐based clustering technique for ad hoc intervehicle communication. IEEE Trans Syst Man
Cybern Part C Appl Rev. 2008;38(2):201‐208.
14 of 15 KUMAR ET AL.
31. Fan, W., Shi, Y., Chen, S., & Zou, L. (2011). A mobility metrics based dynamic clustering algorithm for VANETs.
32. Bai F, Sadagopan N, Helmy A. The IMPORTANT framework for analyzing the impact of mobility on performance of RouTing protocols
for Adhoc NeTworks. Ad Hoc Netw. 2003;1(4):383‐403.
33. Maslekar N, Mouzna J, Labiod H, Devisetty M, Pai M. Modified C‐DRIVE: clustering based on direction in vehicular environment. In:
2011 IEEE Intelligent Vehicles Symposium (IV). IEEE; 2011, June:845‐850.
34. Maslekar N, Boussedjra M, Mouzna J, Labiod H. A stable clustering algorithm for efficiency applications in VANETs. In: 2011 7th Inter-
national Wireless Communications and Mobile Computing Conference. IEEE; 2011, July:1188‐1193.
35. Wolny G. Modified DMAC clustering algorithm for VANETs. In: 2008 Third International Conference on Systems and Networks Commu-
nications; 2008, October:268‐273, IEEE.
36. Basagni S. Distributed clustering for ad hoc networks. In: ISPAN (Vol. 99; 1999, June:310‐315.
37. Rath, M. (2018). An analytical study of security and challenging issues in social networking as an emerging connected technology. In Pro-
ceedings of 3rd International Conference on Internet of Things and Connected Technologies (ICIoTCT) (pp. 26-27).
38. Rath M, Oreku GS. Security issues in mobile devices and mobile adhoc networks. In: Mobile Technologies and Socio‐Economic Develop-
ment in Emerging Nations. IGI Global; 2018:180‐202.
39. Rath M, Swain J, Pati B, Pattanayak BK. Network security: attacks and control in MANET. In: Handbook of Research on Network Forensics
and Analysis Techniques. IGI Global; 2018:19‐37.
40. Zhang Z, Boukerche A, Pazzi R. A novel multi‐hop clustering scheme for vehicular ad‐hoc networks. In: Proceedings of the 9th ACM Inter-
national Symposium on Mobility Management and Wireless Access. ACM; 2011, October:19‐26.
41. May, A. D. (1990). Traffic Flow Fundamentals. Englewood Cliffs, N.J.: Prentice Hall.
42. Kakkasageri MS, Manvi SS. Multiagent driven dynamic clustering of vehicles in VANETs. J Netw Comput Appl. 2012;35(6):1771‐1780.
43. Rawashdeh ZY, Mahmud SM. A novel algorithm to form stable clusters in vehicular ad hoc networks on highways. EURASIP J Wirel
Commun Netw. 2012;2012(1):15.
44. Hafeez KA, Zhao L, Liao Z, Ma BNW. The optimal radio propagation model in VANET. In: 2009 Fourth International Conference on Sys-
tems and Networks Communications. IEEE; 2009, September:6‐11.
45. Intl, A. S. T. M. (2003). Standard specification for telecommunications and information exchange between roadside and vehicle systems‐
5 GHz band Dedicated Short Range Communications (DSRC). Medium Access Control and Physical Layer specifications, E2213‐03.
46. Hafeez KA, Zhao L, Ma B, Mark JW. Performance analysis and enhancement of the DSRC for VANET's safety applications. IEEE Trans
Veh Technol. 2013;62(7):3069‐3083.
47. Aslam J, Lim S, Pan X, Rus D. City‐scale traffic estimation from a roving sensor network. In: Proceedings of the 10th ACM Conference on
Embedded Network Sensor Systems. ACM; 2012, November:141‐154.
48. Arbabi H, Weigle MC. Monitoring free flow traffic using vehicular networks. In: 2011 IEEE Consumer Communications and Networking
Conference (CCNC). IEEE; 2011, January:272‐2760.
49. Gerla M, Tsai JTC. Multicluster, mobile, multimedia radio network. Wirel Netw. 1995;1(3):255‐265.
50. Venkataraman H, Delcelier R, Muntean GM. A moving cluster architecture and an intelligent resource reuse protocol for vehicular net-
works. Wirel Netw. 2013;19(8):1881‐1900.
How to cite this article: Kumar R, Kumar D, Kumar D. Performance enhancement of traffic information
gathering (PEnTInG) algorithm for vehicular ad‐hoc networks. Int J Commun Syst. 2019;e4111. https://doi.org/
10.1002/dac.4111
KUMAR ET AL.15 of 15
A preview of this full-text is provided by Wiley.
Content available from International Journal of Communication Systems
This content is subject to copyright. Terms and conditions apply.
