You are on page 1of 32

110

CHAPTER 5

VANET CONNECTIVITY ENHANCEMENT USING DYNAMIC


CLUSTERING APPROACH

5.1 INTRODUCTION

With a broad extent to connect and communicate the roadway, the


area of VANET is focusing among the researchers. Due to the real time traffic
data monitoring requirements, VANET has considered as an special network to a
great extent. In VANET, the formation of a stable network and communication
is the most exigent task due to the high mobility and random spatial distribution
of nodes.

The high mobility of vehicles and frequent changes in coverage


and connectivity of moving vehicles along with the deployed infrastructures
necessitated the development of connectivity optimization techniques
(Ejaz et al. 2018). Clustering is one of the most important technique for
organizing ad hoc networks with high system scalability, increased connectivity,
and reduced frequent connectivity breakages. It is hard-hitting to design link
layers for cluster-based vehicular networks. Besides clustering, coverage and
connectivity are two important performance metrics to signify the eminence of
the network communications. The connectivity probabilities are analyzed
between Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) in single
and multi-way scenarios.

In the clustering approach (Craig et al. 2016) each cluster has a leader
vehicle or infrastructure, named cluster head (CH), which is responsible for
governing communication in the cluster. Other vehicles in cluster are considered
111

as cluster member (CM) that can communicate directly with CH or indirectly


communicate via different CM using hop communication. LeiLiu et al. (2018);
Xiang Ji et al. (2018), surveyed many clustering algorithms such as Mobility
clustering algorithm, popular Lowest IDentifier (LID) and direction based
clustering approach for adhoc networks. Recently, many of these algorithms
were simulated for VANET. However, due to some distinction characteristic of
VANET such as high mobility and channel conditions, some algorithms are not
suited for this environment. So, VANET researchers distinctively designed some
clustering algorithms based on their requirements and unique properties. In this
proposed study design the fuzzy based clustering algorithm is used to enhance
the connectivity probability of proposed network.

5.2 NEED OF CLUSTER BASED CONNECTIVITY APPROACH

From the year 1980‟s onwards the clustering algorithms are proposed
by many researchers for efficient wireless networks like wireless sensor
networks, adhoc networks , MANET in general and VANET in particular
(Bali et al. 2014). Grouping of vehicles is a latent approach to improve the
scalability and connectivity of networking protocols for VANET scenarios.

Xueshi Hou et al. (2016) presented that the connectivity of moving


vehicles is the key factors that has a impact on the data transmission process.
Due to the lack of real time mobility traces, connectivity cannot understand
the realistic large scale environment and notably the impact of mobility
over-connectivity is unknown. For that reason, they used some topology metrics
such as the velocity of vehicle and density to analyse the connectivity and
mobility among vehicles. In their analysis, they discovered that there exists a
dichotomy relationship between the node speed and size. The two dichotomy set
that the mobility weaks the connectivity when the component speed is larger
than the threshold. Otherwise, there is no impact on connectivity.
112

Bhaskar et al. (2010) proposed the weighted clustering algorithm with


the help of genetic approach to enhance the performance of the cluster head
selection process. It used the combined parameters such as degree of cluster
head, mobility of node and gap distance to explore dominant set. This scheme
selects the finest number of cluster heads that enclosed all the nodes. Genetic
approach does not give an optimal solution when they reduce the transmission
range because the count of cluster heads increased than the cluster member.

Sherif M Abuelenin et al. (2015) provided the knowledge of vehicle


headway distribution that is essential for estimating the probability of
connectivity in VANET. They consider the vehicles are coming in the single
lane and the consecutive vehicles maintain a particular safe distance between
them. The safe distance improves the vehicles spatial distribution and the
empirical data in single traffic. Finally, the headway distribution of lane traffic in
free flow condition is modelled using a shifted exponential approach which
considers the safe distance between the vehicles. From this, it is studied that how
the minimum distance affects the connectivity probability in low dense traffic
conditions.

Caixing Shao et al. (2015) said that the grouping of nodes into
platoons could improve the connectivity probability of network. Moreover, it is
tough to design a MAC protocol for platoon based VANET. So they formulated
connectivity aware MAC protocol for the platoon based VANET. To derive a
relationship between connectivity and system throughput, they present a multi
priority Markov model. Finally, as a result, the throughput increased with the
connectivity probability.

Dang & Wu (2010) suggested that clustering is an active topology


management approach in adhoc networks, increasing the scalability and lifetime
of the network. They proposed an exponentially weighted moving average
113

(EWMA) protocol for updating and maintaining the clusters. They analysed the
performance of the network using PDR, network overhead, and delay.

Shen et al. (2004) proposed a CLuster-based Topology Control


(CLTC) mechanism for Adhoc networks. The CLTC mechanism presented for
extension of scalability and connectivity of the network. Moreover, They used
three phases, cluster formation, cluster head selection, and cluster maintenance.
The protocol enhanced the cluster maintenance process to avoid frequent cluster
head selection. Moreover, they have analysed the network performance using
cluster overhead, average cluster head selection, PDR and delay.

Zhengmin et al. (2010) discussed a cluster-based distributed gateways


routing protocol for adhoc network. The authors proposed border mobile
terminals as cluster heads. The protocol designed cluster reformation algorithm
to avoid redundant cluster formation. They improved the overall connectivity of
the network. Moreover, the protocol analysed the network performance using
cluster overhead, average cluster head selection, PDR and delay.

Still the research in connectivity enhancement in heterogeneous


networks is continued by many researchers. Existing researchers provide better
connectivity probability only due to the high mobility and low density of
vehicles in VANET. Clustering is an important approach in VANET to reduce
the high mobility impacts during communication. This study proposes an
analytical model to appraise the performance of a density based dynamic cluster
to VANET. The analytical model is developed to evaluate four essential
parameters, such as packet delivery ratio, connectivity probability, cluster head
change rate and end to end delay. The results attained from the analytical model
are accompanied with network simulator. The simulation results are analyzed
with the network simulator NS2 and confirmed that the proposed scheme
prolonged the network connectivity compared to other existing schemes. The
114

proposed clustering algorithm provides better cluster stability and low overhead.
Moreover, this research work estimated the connectivity probability among
V2V, and V2I using network performance metrics.

5.3 THE PROPOSED VANET CONNECTIVITY ENHANCEMENT


USING DYNAMIC CLUSTERING APPROACH

This study aims to implement the clustering among vehicles by using


a fuzzy based dynamic clustering approach to make the connectivity of the
network get more robust and scalable. High mobility of vehicles and scalability
of the network are two critical metrics to be considered while affording reliable
and efficient communication in VANET. Clustering is a rigid impact to handle
the dynamic features of the VANET. Plenty of factors related to network
conditions, application requirements and user preferences such as
velocity and density of vehicles, strength of signal and data rate must be
considered for evaluation process of cluster formation and cluster head selection
(Jane et al. 2005). But, the prominent metrics are taken in this study to design
the fuzzy based clustering algorithm those are relative mobility based on
location and speed of vehicles, acceleration, and direction of moving vehicles.
These parameters are taken as inputs to the proposed cluster formation, cluster
head selection, and maintenance process. The proposed dynamic fuzzy based
clustering approach used to analyze the connectivity probability among V2V and
V2I communication. The main contributions of this study are as follows:

Cluster formation: Cluster is formed using both vehicles and


infrastructures among V2V and V2I using fuzzy inference system (FIS).

Cluster Head Selection process: Cluster head selection process is


designed using the defuzzify approach.

Cluster Maintenance: Using defuzzify approach, role transferring


function, cluster merging and retreating process of CM are used to maintain the
115

cluster stability. These methods are used to elect the next cluster head in
futuristic manner. Further this approach is used to maintain the cluster longevity.

Connectivity Analysis: Connectivity probability of V2V and V2I is


analyzed using fuzzy based clustering approach.

Figure 5.1 Block diagram of proposed PCCV protocol

5.3.1 Cluster Formation

Clustering techniques are employed in the VANET to form the


connected network hierarchy and to enhance the connectivity among the
networks. In VANET, when the vehicles enter in to a highway network, it is
considered as an ordinary node and shares its current mobility related beacons
116

with the nearby vehicles. Similarly, it receives the beacons from its peers. Based
on this received beacons, if a vehicle finds a CH it can connect with existing CH
and it is assigned as CM. In this study, two different clustering algorithms
designed based on two various communications like V2V and V2I. In V2V
clustering scenario, vehicles are taking over the leadership role and make the
cluster with its peers using WAVE standard. When the vehicles come under the
particular transmission range of RSI it is capable to form a cluster and the RSI is
set as the CH and another vehicle becomes as the next CM. Figure 5.1 depicts
the block diagram of proposed PCCV protocol.

Assume in a V2V communication, the cluster is formed using the two


moving vehicles x and y, in which each vehicle looks for a CH, which are at
most K-hops away. In this clustering model, a network can be partitioned into
smaller as well as more stable connected networks in terms of relative mobility
RMV , acceleration (an ) and direction (d) etc. A vehicle calculates RMV with its
neighbors based on their speed and location differences. This information is
updated using beacon data received. The location differences LDV between the
vehicles Lx & Ly , are calculated as follows,

LDV = Lx − Ly (5.1)

Speed differences (SDV ) between the vehicles in highway scenario is calculated


using

SDV = Sx COSθ − Sy COSθ (5.2)

where, θ is vector angle between the vehicles. Now, the RMV value is calculated
from Equation (5.1) and (5.2). ∂ and ∂′ are the weight values of location and
speed of the vehicles.

RMV = ∂LDV + ∂′ SDV (5.3)


117

In this proposed methodology, each vehicle travel within its peers‟


transmission range VTR or RSI transmission range RSITR . The location
coordinate of vehicle and infrastructure is denoted as (LV ) and (LI ) respectively.

5.3.1.1 Algorithm for V2V cluster formation

1. Read location difference 𝐿𝐷𝑉 coordinates of vehicles;


2. Define the vehicle transmission range 𝑉𝑇𝑅(𝑖,𝑗 ) ;
3. Read speed difference SDV coordinates;
4. Calculate the relative mobility RMV (i, j);
5. Generate Fuzzy inference rule (𝑓 𝑥 ) = (𝑅𝑀𝑉 , d, an )
6. Compare Fuzzy inference rule 𝑓 𝑥 ;
If (𝑓 𝑥 ==medium) then
Vi join in cluster as CM;
Else if (𝑓 𝑥 ==high) then
Vi join in cluster as cluster member CM and nominee cluster head
𝐶𝐻𝑁𝑜 ;
And select cluster head using CH select();
Else if (𝑓 𝑥 ==low) then
Vi join in cluster as CM or reject;
End if
7. CH select();
8. End

The affinity values of each vehicle calculated using a fuzzifier in


terms of RMV , an and d. Now, a vehicle can join in a cluster on account of the
affinity values that defines the proximity of the V2V or V2I cluster. If affinity
range of the peer cluster size is bigger than the RSI then the vehicle make a
decision to join in to the clusters of the VTR as CM, else it joins in the V2I
cluster. If the two neighboring vehicles, travel within the same transmission
118

range with same direction and same an those vehicles can form a cluster group
on the highway. At the same time, opposite direction based clustering is
considered only when dact is low. Hence, this approach is not considered to
avoid less cluster life time. Figure 5.2 depicts the structure diagram of fuzzy
inference system for clustering approach.

Figure 5.2 Structure diagram of fuzzy inference system for clustering


approach

5.3.1.2 Algorithm for V2I cluster formation

1. Read location coordinates LI of infrastructures ;


2. Define the transmission range of RSI (RSITR(i)) and vehicle( VTR (i));
3. Read location coordinates LV of vehicles ;
4. Calculate the relative mobility RMV (i, j);
5. The vehicle transmission range is less than RSI transmission range
If( VTR (i) < RSITR(i) ) then
Vehicle Vi join in cluster as cluster member (CM) and set RSIi is as
current cluster head 𝐶𝐻𝐶 ;
Else if
Set as RSIi is next cluster head (𝐶𝐻𝑁 );
Repeat from step 2 until form the cluster with RSI;
Else
Apply V2V cluster();
End if
6. End
119

The cluster formation and cluster head election process is performed


using a fuzzy based multi-attribute decision making (MADM) technique. This
technique plays a vital role to generate the fuzzy logic rules 𝑓 𝑥 for all
variations. In this fuzzy logic based cluster formation algorithm, the
parameters RMV , d, and an are considered as inputs to generate fuzzy rules using
fuzzy inference engine. Figure 5.3 & 5.4 depicts the fuzzy membership function
for an and d .The task of the fuzzifier is to transform the actual quantities into
fuzzy sets.

Figure 5.3 Fuzzy membership function for acceleration (𝐚𝐧)

The crisp set for given input parameters are given as


RMV = High = 65Kmph to 100Kmph , Medium = 35Kmph to 65Kmph , Low =
35Kmph , d = same , oppposite and an = {accelerate, same, decelerate}.
The membership function trim and trapezoid are
used to generate membership values (μ) using the fuzzy set parameters.
Yanan et al. (2013). The product inference rule is generated according to input
values. Then, the next phase is obtained by the reverse process of fuzzifier called
as defuzzifier approach for CH selection process.
120

Figure 5.4 Fuzzy membership function for direction (𝐝 )

5.3.2 Cluster Head Selection

Maintaining the cluster and its stability is the key objective in


clustering concept. Therefore, the cluster head (CH) selection process is also
involved based on maintaining the stability of cluster. It can reduce the overhead
of cluster reorganization, avoid frequent CH selection process and provide an
efficient hierarchical topology. During the selection of CH, only one CH
nominee candidate is selected to be the CH which has more potential than other
nominees. The next phase of FIS is defuzzification approach that initiates
selection of current cluster head (CHC ), next cluster head (CHN ), and cluster head
nominees CHNo in a cluster. Defuzzification is the process of acquiring a crisp
output from the aggregated fuzzy rules by using several methods such as lambda
cut, maxima and centroid methods etc. This study proposes the height of
maxima methods to solve the problem of CH selection from set of cluster
nominees. This defuzzification describes three methodologies such as methods
associated to the minimum (First of Maxima-FOM), methods associated to the
maximum (Last of Maxima-LOM) and methods associated to the mean (Mean of
Maxima- MOM). As per FOM Current Cluster Head 𝐶𝐻𝐶 is selected as,

𝐶𝐻𝐶 = min x f x = maxh f w (5.4)


121

The Next Cluster Head 𝐶𝐻𝑁 is selected based on the MOM method as,

𝐶𝐻𝑁 = max x f x = maxh f w (5.5)

and remaining CM of maxima is consider as CHNo ,

xi ∈M xi
CHNo = (5.6)
M

M is defined as M = xi μ xi = h(c) where M cardinality of set


M. Here, h(c) is the height of fuzzy set. In V2I clustering approach, each vehicle
calculates RMV in terms of RSI location and distance. d, and an are involved to
elect the 𝐶𝐻𝐶 and 𝐶𝐻𝑁 . But, in V2V clustering approach, each vehicle gathers
the information RMV calculated by each vehicle in the route, to select the CH
using FOM method with least RMV and direction flow of vehicles. The CHN , is
calculated using MOM and when the 𝐶𝐻𝐶 exit from the cluster, 𝐶𝐻𝐶 is directly
elected by 𝐶𝐻𝑁 as 𝐶𝐻𝐶 . The 𝐶𝐻𝐶 and 𝐶𝐻𝑁 selection is beaconed among cluster
members. The elected CH act as the intermediary between other CMs. Each
vehicle communicates with its peers either directly (CM) or CH.

5.3.2.1 Algorithm for cluster head selection

1. Start cluster head selection CH select()


2. Read nominee cluster Heads 𝐶𝐻𝑁𝑜 ;
3. Generate fuzzy inference rule 𝑓 𝑥 ;
4. Defuzzify (𝑓 𝑥 ) ;
a. Calculate First of Maxima FOM;
Select current cluster head 𝐶𝐻𝐶 ;
b. Calculate mean of maxima MOM ;
Set nominees cluster head CHN ;
c. Calculate last of maxima LOM ;
Set next cluster head CHNo ;
5. End
122

5.3.3 Cluster Maintenance

In the VANET environment, vehicles can either enter or depart in the


highway at any duration, due to some mandatory reasons. Therefore, the
maintenance of cluster longevity is the main objective of forming the clusters
using durable cluster heads. The following reasons are the main causes of cluster
maintenance:
 Transferring leadership role from 𝐂𝐇𝐂 to 𝐂𝐇𝐍 : When the 𝐶𝐻𝐶 can
no longer be a CH or exit from the cluster then, 𝐶𝐻𝑁 will take over
the role to avoid the cluster reconfiguration process. Ultimately, this
𝐶𝐻𝐶 will change its role as CM and 𝐶𝐻𝑁 as 𝐶𝐻𝐶 to maintain the
uninterrupted connectivity between intra and inter cluster. This
approach is highly feasible as this role changing process does not
create impact on the cluster structure.
 Cluster merging: When two 𝐶𝐻𝐶 s are within the same
communication range it initiates the cluster merging process. In this
scenario, the 𝐶𝐻𝐶 with a higher FOM will continue its leadership role,
while the other 𝐶𝐻𝐶 are consider as CM. The new 𝐶𝐻𝐶 sends beacon
to other CMs in the cluster to maintain connectivity.
 Retreat of CM: Owing to high frequent changes of network
topology and dynamics of vehicles, a CM may exit from the
connectivity range or leave from the highway. Thus, when a 𝐶𝐻𝐶
does not receive any periodical beacons from its CM, it assumes that
the CM to be disconnected or had left. As a result the 𝐶𝐻𝐶 will
remove the record of particular CM from its cluster list and send
beacons to the current CMs.

If no more peers take over that leadership then automatically the role
of 𝐶𝐻𝐶 is carried out by the neighboring RSI. Eventually, this 𝐶𝐻𝐶 will change
its status to a CM or retreat from the cluster. The nominee value of each CM is
123

compared with the 𝐶𝐻𝐶 . If the variation between the compared values is greater
than the 𝐶𝐻𝐶 , then the new CM is elected as 𝐶𝐻𝐶 . The candidate value of 𝐶𝐻𝐶
varies between 0.0 and 1.0 where the F𝐶𝐻𝐶 range 0.0-0.2 indicates the weakest
nominee and the F𝐶𝐻𝐶 range greater than 0.7 is considered as the strongest
nominee in the quantification. Table 5.1 represents the sample fuzzy rules and its
output fuzzy values.

5.3.3.1 Algorithm for cluster maintenance

1. Read fuzzy range of current F𝐶𝐻𝐶 and next cluster head F𝐶𝐻𝑁 ;
2. Compare the ranges of both cluster head
If (F𝐶𝐻𝑁 > F𝐶𝐻𝐶 ) then
Assign next cluster head vehicle as current cluster head 𝐶𝐻𝐶 = 𝐶𝐻𝑁 ;
Else if
Assign RSI as current cluster head 𝐶𝐻𝐶 = RSI(i);
Else
Elect new cluster head using CH select() ;
End if
3. End

Table 5.1 Sample Fuzzy Logic Rule Base of PCCV


Sample Fuzzy Logic Rule Base of PCCV
IF (𝑅𝑀𝑉 is Low) ⋂ (an is Decelerate) ⋂ (d is same) then (Output is 0.3)
IF (𝑅𝑀𝑉 is Low) ⋂ (an is Decelerate) ⋂ (d is same) then (Output is 0.3)
IF (RMV is Medium) ⋂ (an is Decelerate) ⋂ (d is same) then (Output is 0.7)
IF (RMV is Low) ⋂ (an is same) ⋂ (d is same) then (Output is 1.0)
IF (RMV is medium) ⋂ (an is Decelerate) ⋂ (d is opposite) then (Output is 0.5)
IF (RMV is High) ⋂ (an is Accelerate) ⋂ (d is opposite) then (Output is 0.3)
IF (RMV is High) ⋂ (an is Accelerate) ⋂ (d is opposite) then (Output is 0.2)
124

5.3.4 Connectivity Analysis

The second phase of this work analyzes the connectivity probability


among the moving vehicles and the road side infrastructures (RSIs) based on the
fuzzy-based clustering VANET. The connectivity probability is defined as the
probability that the moving vehicles can communicate with its peers or RSI on
the highway within a minimum number of hops. Considering a multilane
highway bounded by optimal placements of RSIs, with moving vehicles are
travelled on the roadway according to HSMLC protocol. Assume, the distance
between the two RSIs is „d‟ and road length is L. Let the transmission range of
vehicles and infrastructures are VTR and RSITR and d′act be the density of vehicles
per meter (l), then the cumulative probability distribution function is defined as,
n
d ′act l ′
f n, l = e−(d act )l
, where n ≥ 0; If the inter vehicle distance or net
n!

distance DG smaller than the (l) then the probability distribution function is
described as ,


P DG ≤ l = 1 − e−𝑑 𝑎𝑐𝑡 l (5.7)

According to that, two different communications approaches of


VANET such as V2V and V2I connectivity probability is analyzed on the four
different scenarios. First, cluster based V2V connectivity is analyzed on both
single and multilane scenarios and second, cluster based V2I connectivity is
analyzed on both single and multilane scenarios.

5.3.4.1 V2V connectivity at single lane scenario

In this single lane V2V scenario, vehicles act as both CM and CH and
travel in the same direction (d ). Here, the inter vehicle distance between two
consecutive vehicles is DGi . If DGi come under the range of 𝑉𝑇𝑅 then the
particular vehicle is joined as CM in this cluster. So the connectivity probability
is calculated as,
125

N−1
P′ = i=0 p DGi ≤ 𝑉𝑇𝑅 (5.8)

V2V based cluster communication scenario is represented in


Figure 5.5. In the single lane connectivity scenario two options are possible to
make a seamless connectivity, either vehicle can connect with the CM and its
probability measure is q or it can directly connect with CH and its probability
measure is p. Now, the connectivity probability is described as

N−1
PV1 = i q. P DGi ≤ R CM } + p. P DGi ≤ R CH } (5.9)

R CM and R CH transmission range of CM and CH vehicles. As per


equation 5.7, the connectivity probability of V2V in single lane scenario (PSV ) is
defined as,

′ ′ N−1
PSV = q. 1 − e−𝑑 𝑎𝑐 𝑡 R CM + p. 1 − e−𝑑 𝑎𝑐𝑡 R CH
(5.10)

5.3.4.2 V2V connectivity at multi lane scenario

Figure 5.5 represents the V2V communication at multi lane scenario.


In this scenario, the adjacent lanes as well as vehicle moving directions d are
considered for connectivity analysis. When the gap between the two consecutive
vehicles DGi is larger than its transmission ranges VTR ,the link breakages
happened so the vehicles can try to make a connectivity with the adjacent lane
clusters. During this scenario, the connectivity link of same lane is considered as
broken link Bl . So the vehicle can try to connect with adjacent lane cluster‟s CM
or CH. As per the Figure 5.5, the connectivity among vehicles is established
using more than one cluster (N). If established connectivity is only formed by
qk .
N−1
CM then the connectivity probability is defined as PBl = k=0 ′ ,
1 − e−𝑑 𝑎𝑐𝑡 R CM
else if it is generated by the CH then connectivity probability is
126

N−1 ′
PBl = k=0 pk . 1 − e−𝑑 𝑎𝑐𝑡 R CH , else if it is established by both combination of
CM and CH then connectivity probability is represented as,

N−1 ′ ′
PBl = DG > 𝑉𝑇𝑅 = k=0 q k . 1 − e−𝑑 𝑎𝑐𝑡 R CM + pk . 1 − e−𝑑 𝑎𝑐𝑡 R CH (5.11)

Sometimes the adjacent lane vehicles directly communicated with


single cluster via healthy links. This link established either using CM or CH.
Thus, as per binomial distribution the total connectivity probability PT k is
defined using two different link qualities such as broken(Bl) and healthy
links(Hl).

N−1
PT = k
PBl k PHl N−1−k
(5.12)

Here, k of N-1 has broken links and PHl is healthy links in a network.

Figure 5.5 Cluster based V2V communication scenario

As per the Figure 5.5, Vsi and Vsj in same lane, and DG is longer than
the 𝑉𝑇𝑅 . Thus they can to communicate with adjacent lane vehicles of Vai
and Vaj . The probability function of Psa (i)and Psa (j) is ,


Psa (j) = Psa (i) = P DG ≤ 𝑉𝑇𝑅 = 1 − e−𝑑 𝑎𝑐𝑡 𝑉𝑇𝑅 (5.13)
127

Thus, connectivity probability among Vsi and Vsj is denoted as,


Ps(i,j) = Psa (j) + Psa (i). Healthy link communication PHl is defined as the two
adjacent lane vehicles are directly connected with single cluster, either by CM or
CH is signified as,

N−1 ′ ′
PHl = DG ≤ 𝑉𝑇𝑅 = k=0 q k . 1 − e−𝑑 𝑎𝑐𝑡 R CM + pk . 1 − e−𝑑 𝑎𝑐𝑡 R CH (5.14)

N−1
PMV = k
PBl k PHl N−1−k
(5.15)

Thus as per Equation 5.15, the total connectivity probability of


multilane V2V communication PMV is calculated.

5.3.4.3 V2I connectivity at single lane scenario

In this V2I communication scenario, vehicles can communicate with


road RSI directly or using peers. Assume that, 𝑉𝑠 can communicate with the RSI
using one hop communication and 𝑉𝑠 to infrastructure communication using
cluster member is termed as two hop mode. The single lane V2I scenario is
depicted in Figure 5.6. From this depiction gap between the two adjacent
infrastructure is ′d′ and its transmission range is RSITR . Let, the connectivity
probability of a 𝑉𝑠 in single lane V2I scenario is P′I .This module of the work
defined using different cases of scenarios.

Case i: If the vehicle can connect directly under the coverage of


adjacent Infrastructure without any CM then the total connectivity probability of
vehicle is equal to 1.In this case, the vehicle can connect using single hop mode.
So the connectivity probability is,

2RSI TR
PC′ = (5.16)
d
128

Case ii: If the vehicle communicate using more than one hop mode
then two possibilities are offered to connect the 𝑉𝑠 with the infrastructure using
either CM of V2I cluster or CH of V2V cluster.
a) Connect using CM of V2I: If the 𝑉𝑠 travels under the coverage barrier
of any infrastructures then it can attempt to communicate with its
peer. If the peer is CM of cluster, then the connectivity probability is,

′ ′
PCM ′ = ∆q 1 − e−𝑑 𝑎𝑐𝑡 l (5.17)

where, ∆= 1 − PC ′ and l′ = 2 RSITR + R CM −d

b) Connect using CH of V2I: If the 𝑉𝑠 travels under the coverage


barrier of any RSI then it can try to communicate with its peers.
If the peer is CH of V2V, then the connectivity probability is,

PCH ′ = ∆p 1 − e−𝑑 𝑎𝑐𝑡 l" . If more than one CH participate to connect
the vehicle with RSI then the connectivity probability is,

N−1 ′
PCH ′ = k=0 pk . 1 − e−𝑑 𝑎𝑐𝑡 l"


PCH ′ = ∆p 1 − e−𝑑 𝑎𝑐𝑡 l" + PCM ′ (5.18)

where, l" = 2 RSITR + R CH − d

From the Equation 5.16 to 5.18, the overall connectivity probability


of this single lane V2I scenario is PSI = PCM ′ + PCH ′ +PC′

′ ′ ′
PSI = ∆ q 1 − e−𝑑 𝑎𝑐𝑡 l + p 1 − e−𝑑 𝑎𝑐𝑡 l" + PCM ′ + PC ′ (5.19)
129

Figure 5.6 Cluster based V2I communication scenario

5.3.4.4 V2I connectivity at multi lane scenario

In this multilane V2I communication scenario, moving vehicles can


try to communicate with other vehicles via RSIs and neighboring lane vehicles.
In this case, assume that 𝑉𝑠 can communicate with the RSI using two or more
hop communications using more than one cluster. Figure 5.6 represents the V2I
communication at multi lane scenario. Let, the connectivity probability of 𝑉𝑠 in
multi lane V2I scenario is PMI .This module of the work is defined using
different cases of scenarios.

Case i: If the vehicle can communicate under the coverage of two


adjacent RSI without relay vehicles and using one hop fashion then the total
connectivity probability of subject vehicle is equal to 1.

Case ii: In this case, 𝑉𝑠 comes under the breach of coverage. Thus
two possibilities are offered to connect the vehicle with the RSI. They can
connect the vehicle using adjacent road peers either CM of V2I cluster or CH of
V2V cluster.
130

a) Connect using CM of V2I: If the Vs travels under the coverage


barrier of any RSI then it can try to build a communication with
peers. If the peer is CM of cluster on the same road, then the
connectivity probability is described as per Equation 5.17. If the CM
is from adjacent lane, then the connectivity probability is,

N−1
PCM " = k
PBl k PHl N−1−k
(5.20)

Then using the connected CM the 𝑉𝑠 communicates with RSI using as


per the Equation 5.17, so the total connectivity probability using CM
of V2I is (PCM −V2I )defines as,

PCM −V2I = PCM " + PCM ′

N−1 ′ ′
PCM −V2I = k
PBl k (PHl )N−1−k + ∆q 1 − e−𝑑 𝑎𝑐𝑡 l (5.21)

b) Connect using CH of V2V: If the subject vehicle travels under the


coverage barrier of any RSI then it can try to create a connection with
peers to reach the RSI. If the peer is CH of V2V, then the

connectivity probability is, PCH ′ = ∆p 1 − e−𝑑 𝑎𝑐𝑡 l" . If more than one
CH is participate to connect the vehicle then the connectivity
probability is,

N−1 ′
PCH ′ = k=0 pk . 1 − e−𝑑 𝑎𝑐𝑡 l" (5.22)

If the connected CH, extend the connectivity to RSI then, the



connectivity probability is PCH−V2V ′ = ∆p 1 − e−𝑑 𝑎𝑐𝑡 l" + PCM ′ ,where,
l" = 2 RSITR + R CH − d. Hence, the overall connectivity probability of multi
lane V2I scenario as per case ii is defined as PMI ′ = PCM −V2I + PCH−V2V ′ .

′ ′ ′
PMI = ∆q 1 − e−𝑑 𝑎𝑐𝑡 l + ∆p 1 − e−𝑑 𝑎𝑐𝑡 l" + PCM ′ + PC′ (5.23)
131

The overall V2V and V2I connectivity of the network is analysed


using the Equations 5.10, 5.15, 5.19 and 5.23 interms of DSRC communication
technology.

5.4 SIMULATION AND RESULTS

5.4.1 Simulation Study

This work is implemented using two simulators (Narendra et al. 2014)


such as network simulator NS2 and fuzzy based experiments simulated using
matlab. The main goal of these simulations is to analyse the performance of
proposed protocol using with and without infrastructures. A VANET of 100
moving vehicles and infrastructures are deployed into 1000mx50m area of the
highway. The simulation results are studied by varying the density of vehicles
from 10 to 100. The experiment is evaluated using various transmission ranges
of infrastructures. The overall performance of the proposed study is compared in
terms of packet delivery ratio (PDR), End to End delay, throughput, Average CH
duration, and connectivity of network with related Analytical model for
clustered vehicular ad hoc network analysis (AMCV) and Efficient and Reliable
Cluster-Based Data Transmission for Vehicular Ad Hoc Networks (ERCV)
schemes. Table 5.2 depicts the parameters used in the simulations.

Table 5.2 Simulation parameters of PCCV


Parameters values
Simulated Area 1000mX50m
Simulation Time 300Sec
Area of Map Pearl City –TN state Highway
Number of Vehicles /RSI 10 -100 /16
Vehicle Speed 35-100 kmph
Simulators Matlab/NS2
PHY/MAC protocol IEEE 802.11p
132

Table 5.2 Continued


Transmission range of OBU /RSI(m) 50 /100
Network Interface/Channel Type WirelessPHY/Wireless channel
Packet size (bytes)/ packet count 512 /100
Antenna model Omni
Mobility model HSMLC

5.4.2 Algorithms used for Comparison

The two latest existing connectivity algorithms of VANET AMCV


(Raghavendra et al. 2018) and ERCV (Xiang Ji et al. 2018), have been
considered for comparison with the proposed algorithm PCCV for evaluating the
connectivity features and performance of the proposed scheme.

AMCV provides solution in the form of an analytical model to


analyse the performance of a clustered VANET. The analytical approach is
developed to appraise the connectivity probability using three important
networking parameters, namely packet delivery ratio, throughput, and delay of
the network. The simulation results of this scheme totally differ from the
analytical approach because the propagation model, channel interference, and
other physical layer properties are not considered in analytical modelling. ERCV
provides the solution for efficient connectivity and reliable data transmission in
VANET. First of all they filter out the unstable neighbours in the network and
consider some stable nodes for communication. They proposed a link reliability
based clustering algorithm to analyse the network condition by assigning ranks
to the road sectors. Routes with the lowest rank values are selected as the
optimal data transmission paths and introduced link node to evaluate the network
condition. The model is developed to evaluate the connectivity probability using
three important networking parameters, of packet delivery ratio, end to end delay
and throughput.
133

5.4.3 Result Analysis and Discussion

This section describes the results obtained through simulation of the


proposed scheme. The experimentation is performed by varying the number of
nodes and its velocities in terms of network parameters. The obtained results are
analysed with each other and analysed the proposed scheme with other existing
schemes. The efficiency of the proposed and related schemes is analysed using
the parameters such as end to end delay, packet delivery ratio, network lifetime,
CH and CM duration, CH change rate, packet loss, connectivity probability and
throughput. The simulation results depict that the proposed cluster based
connectivity approach provides an effective mechanism to give seamless
connectivity when compared with the other related schemes.

The performance of the proposed PCCV clustering algorithm is


analysed using the cluster stability and rate of CH change in the cluster.

Average CH duration: It defines the average lifetime of CH in the


cluster which is responsible for managing transmissions in the cluster.

Average CM duration: It defines the average lifetime of CM in the


cluster which is responsible for transferring message inside of the cluster.

Cluster head change rate: Average rate of change of cluster head


with respect to density and velocity of vehicles.

Furthermore, the connectivity performance of the proposed algorithm


is analysed using connectivity probability of the network, PDR, end to end delay
and throughput.

Connectivity probability: It is the probability that the node can


access other nodes for transferring the messages with minimum hop counts.
134

Packet Delivery Ratio: Packet delivery ratio can be determined as


the ratio of the number of packets successfully received by the destination node
and number of packets sent by the source along with the dedicated link.

End to end delay: End to end delay defines the time taken for a
transferring of a packet across a network from source vehicle to destination
vehicle or Infrastructure in VANET.

Throughput: Throughput is the rate of effective transmission of


message over a network from source node to a destination node.

5.4.3.1 Average CH duration

Figure 5.7 shows the average CH duration of the cluster in seconds


based on the different velocity of vehicles. The transmission range of
infrastructure is assigned as 100m, and the vehicle's transmission range is 50m.
As the vehicle's velocity increases, the CH frequently changes, and the average
durability of CH will also decrease. Compared to existing works, the PCCV
maintains its CH duration nearly 27.1sec during 65Kmph. Compared with other
related works, PCCV‟s CH duration 20sec and 33sec higher than ERCV and
AMCV. But the CH duration of PCCV and existing algorithms get decreases
rapidly when the velocity of vehicles gets increases.

Figure 5.7 Average CH duration in terms of velocity of vehicles


135

5.4.3.2 Average CM duration

Figure 5.8 shows the average CM duration of the cluster in seconds


based on the different velocity of vehicles. The density count is 100 and during
65Kmph, the cluster member duration is analysed. As the vehicle's velocity
increases, the CM role also changed frequently moreover and the average
durability of CM will also decrease. Compared to existing works, the PCCV
maintains its CM duration of nearly 60.5sec during 65Kmph. Compared with
other related works, PCCV‟s CM duration 18 sec and 20sec higher than ERCV
and AMCV. But the CM duration of PCCV and existing algorithms get
decreases rapidly when the velocity of vehicles gets increases.

Figure 5.8 Average CM duration in terms of velocity of vehicles

5.4.3.3 CH change rate

Figure 5.9 shows the average CH change rate. It represents that if


there are more merging points within the cluster, there is a higher chance of
changing the CH role. Moreover, the CH change rate depends upon the
velocities and transmission ranges of participating units. Compared to existing
works, the PCCV maintains its CH for a long duration, so the CH change rate is
0.13 and 0.25 less than ERCV and AMCV.
136

Figure 5.9 CH change rate in terms of velocity of vehicles

From the results, it is observed that the CH change rate increases


when the velocity of nodes gets increased it leads to change the network
topology frequently. Moreover, if cluster members shift out of the cluster or
cluster merging, this may increase the cluster head changing rates. The higher
transmission range of RSI provides better connectivity among the vehicles
within a cluster. Moreover, it also reduces the CH change rate of cluster as a
whole.

5.4.3.4 Connectivity probability

This proposed PCCV scheme is compared with the existing schemes


such as AMCV and ERCV in graph figure 5.10. The V2V connectivity
probability is enhanced from 10 counts of density vehicles with 65Kmph
velocity and the aid of clustered infrastructures with 65Kmph. During high
density, the network connectivity reached above 80% because the peer vehicles
acted as CM, and RSI acted as CH. Thus, the new cluster forming by vehicles
and selection of CH processes are neglected. Compared with other related
works, PCCV has performed 6.3% and 12.17% better than ERCV and AMCV
because of the cluster merging method and transferring the current cluster head
role to CHN leads to extend the cluster lifetime and also avoid cluster
reformation process.
137

Figure 5.10 Connectivity probabilities in terms of density of vehicles

5.4.3.5 Packet delivery ratio

The packet delivery ratio of the four proposed schemes of PCCV is


calculated based on varying density of vehicles. High PDR value simultaneously
increases the throughput of the network also. The overall PDR of PCCV is
compared with the related works such as AMCV and ERCV that is depicted in
Figure 5.11. The PDR of the proposed study is increased from the counts of 10
vehicles. During the dmax scenario, the PDR tremendously increased and
reached 79%. Due to the high maintenance of clusters, the PCCV has performed
8% and 14% well than ERCV and AMCV.

Figure 5.11 PDR in terms of density of vehicles


138

5.4.3.6 Throughput

Figure 5.12 shows the system throughput under three clustering


algorithms in terms of the density of vehicles. The figure demonstrates that
PCCV has given high throughput as 40.4Kbps at 65kmph. The average
throughput of the PCCV is 40.4Kbps it is better than the 36.6Kbps of ERCV and
33Kbps of AMCV. Furthermore, the total throughput of PCCV has provided
3.6% and 6.4%, better performance improvement than ERCV and AMCV,
respectively.

Figure 5.12 Throughput in terms of density of vehicles

5.4.3.7 End to End delay

The end to end delay analysis is carried out using four proposed
approaches of PCCV. V2I (multi lane) model is pinned to be more efficient and
optimal as low end to end values are plotted. The proposed V2I (multi lane) has
a minimum delay than the other three approaches V2V (SL), V2V (ML), and
V2I (ML). Approaches with high end to end delay lead to loss in data packets,
retransmission of data packets, and represent the poor quality of connectivity,
etc. The overall end to end delay of PCCV is compared with AMCV and ERCV
in terms of density of vehicles. Figure 5.13 represents the end to end delay of
proposed and existing schemes. The proposed scheme achieves less delay about
139

11.01% and 16.5% to deliver the packets compared to ERCV and AMCV,
respectively. The proposed scheme increases average delay drastically when the
numbers of nodes get decreased and velocities of vehicles get increased. The
graph shows that the PCCV performed 6% and 9% better than ERCV and
AMCV, respectively, because of the proposed schemes high CH durability that
leads to cluster maintenance.

Figure 5.13 End to End delay in terms of density of vehicles

Table 5.3 Evaluation of PCCV with the existing systems in terms of


density of vehicles
Performance Metrics AMCV ERCV PCCV*
Packet Delivery Ratio-PDR (%) 50.1 61 70.5
Connectivity probability 0.58 0.6 0.62
End to End delay(msec) 2190 2090 1860
Packet loss (%) 49.9 39 29.5
Throughput (Kbps) 25.1 30.2 34.8
Average CH duration (sec) 13.03 17.5 19.06
Average CM duration (sec) 30.4 35.5 43.1
CH change rate 0.65 0.42 0.3

Table 5.3 shows the evaluation of connectivity probability of PCCV


with the existing systems AMCV and ERCV for different parameters. The
140

parameters may be network lifetime, packet delivery ratio, connectivity


probability, packet loss, Average CH duration and an end to end delay are
compared with related schemes when the average of 70 vehicles involved in the
simulation. The reliability of data transmission can be achieved through
seamless connectivity among nodes in the network. In the existing connectivity
schemes connectivity range is increased, when the number of vehicles increased.
In addition the packet loss also increased due to congestion among the clustered
network. The proposed scheme resists the frequent link breakages using fuzzy
based clustering approach and achieving a high packet delivery ratio with
increased network lifetime of the proposed network.

Table 5.4 Evaluation of PCCV with the existing systems in terms of


velocity of vehicles
Performance Metrics AMCV ERCV PCCV*
Packet Delivery Ratio-PDR (%) 65 71.88 79
Connectivity probability 0.69 0.75 0.81
End to End delay(msec) 1854 1710 1500
Packet loss (%) 35 28.12 21
Throughput (Kbps) 33.28 36.80256 40.448
Average CH duration (sec) 2.1 4.5 7.2
Average CM duration (sec) 5 9.2 16.2
CH change rate 0.82 0.73 0.6

Table 5.4 shows the evaluation of coverage and connectivity of


PCCV with the existing systems AMCV and ERCV for different parameters.
The parameters may be network lifetime, packet delivery ratio, connectivity
probability, packet loss, Average CH duration and an end to end delay are
compared with related schemes when the average of 65Kmph velocity vehicles
with dmax density involved in the simulation. The reliability of data
transmission can be achieved through seamless connectivity among nodes in the
141

network. In the existing connectivity schemes, connectivity probability are


decreased too low when the velocity of vehicles increased, compared to the
proposed scheme. Moreover, the packet loss also increased and PDR decreased
due to rapid mobility of vehicles. The proposed scheme resists the frequent link
breakages using PCCV and achieving a high packet delivery ratio with increased
network lifetime of the proposed network compared to other existing schemes.

5.5 SUMMARY OF THE CONTRIBUTION

This work used a fuzzy based clustering approach to form clusters


that achieved efficient cluster formation. It used various metrics like density,
speed of vehicles, distances between vehicles to form dynamic clusters, head
selection, and maintenance also analyzed. In the proposed work, the cluster is
formed using moving vehicles and static infrastructures. When the number of
vehicles gets increased responsibility of CH is to take over by the infrastructures.
But in V2V communication CH is selected based on the relative mobility,
direction of moving vehicles and their acceleration values. The cluster
maintenance also performed well in a manner for the maintenance of cluster
longevity using durable cluster heads. Whenever, the cluster head moves away
from its current cluster the responsibility of cluster head changed over to CHN
using transferring leadership role. Moreover, cluster merging method also
maintains the cluster longevity and uninterrupted connectivity between and
within the cluster. In the proposed study, the cluster is formed in a single lane
and extends its formation up to multi lane road structure. The connectivity of
network is analyzed in four different scenarios such as V2V (Single lane), V2V
(multi lane), V2I (single lane), and V2I (multilane lane). Thus, the proposed
PCCV provide seamless connectivity among the proposed network structure.
The simulation result shows that PCCV achieved better packet delivery ratio,
low delivery latency with less packet loss and maintain better CH durability than
AMCV and ERCV.

You might also like