Essay: Survey on Clustering Schemes in Vehicular Ad-Hoc

Survey on Clustering Schemes in Vehicular Ad-Hoc
Networks(VANETs)
Abstract’Vehicular Ad-hoc Networks(VANETs) are special
kind of Mobile Ad-Hoc Network which has significant role
in the applications related with traffic safety and traffic efficiency.
The process of clustering of nodes in VANET improves
the performance of VANET structure by performing data aggregation.
There are several clustering schemes for Vehicular Ad-hoc
Networks.Each scheme uses different criteria for the formation of
clusters and election of cluster head.In this survey,the objectives
and principles of different clustering schemes for VANETs are
explained.The features that must be considered while designing
a good clustering scheme are also explained.
Keywords’Ad-hoc networking,MANETs,VANETs,Clustering
I. INTRODUCTION
With the heavy increase in population,vehicular traffic
intensity has increased considerably and it has caused a
large increase in the rate of road accidents.A Vehicular
Ad-Hoc Network (VANET)is a network of vehicles
in which direct communication between vehicles takes
place through an ad-hoc network.It ensures efficient
and free flow of traffic by exchanging the current road
condition and traffic related information.Also it guarantees
safety of passengers by exchanging information like road
intersection details,emergency warning information etc.Thus
vehicular networks are secure networks that reduces the
number of road accidents considerably.In addition to safety
applications,vehicular networks are also used for comfort
applications such as automated toll collection systems,fuel
station information,entertainment system etc. A detailed
architecture of VANET is given fig 1.
Each node or vehicle in the VANET structure is
equipped with an On-Board Unit(OBU) and a GPS system.
Road Side Unit(RSU)is the another component in VANET
that forms the infrastructure backbone of the network.The
number and distribution of RSU depends on the communication
protocol used.The Vehicle to road side infrastructure
communication configuration provides a high band width link
between vehicles and RSUs.Vehicle-to-vehicle and Vehicle-toroad
side communications is achieved with the wireless standard
Dedicated Short Range Communications(DSRC). DSRC
is a short to medium range communication service that provides
the wireless connectivity between moving vehicles with
data rate of up to 54Mbps.This service covers a wide range
of applications such as emergency warning system for vehicles,
intersection collision avoidance,vehicle safety inspection,
Fig. 1. VANET architecture
electronic parking payments and emergency vehicle signal priority.
The distributed nature of the vehicular network generates
and propagates large volume of messages in VANET.This
will reduce the network life time and increase the routing
overhead,which can be avoided by clustering.A sort of virtual
groups that are formed by a clustering algorithm are called
clusters.
II. BACKGROUND
Clustering is a process of grouping the nodes or vehicles
in the network according to some rules or criteria.These rules
differs from one clustering algorithm to another. In each cluster,
there must be at least one cluster head(CH).Usually,nodes
which have better features are elected as cluster head.A cluster
head acts as a local coordinator for its cluster by efficiently
performing the operations such as data aggregation,data forwarding
and intra-cluster transmission.The cluster members
forward the message packet to the cluster head within the cluster.
The cluster head will either distribute the packet within the
cluster if the intended destination vehicle is within the cluster
or forwards to another cluster.In a clustered VANET,only the
cluster head node and road side unit is participating in the
communication process instead of individual node to road side
unit communication.
Some of the advantages of clustering in VANET are:
 The distributed network nature of VANET increases
2
the network overhead and reduces its lifetime.Clustering
avoids this problem by performing data aggregation.
 VANET structure reduces the resources for retransmission
process due to collision.
 In a clustered structure,local changes need not be updated
by the entire ad-hoc network.
 Clustering guarantees the efficient utilization of communication
channel bandwidth.
 Spatial reuse of resources can be achieved by clustering.
III. CLUSTERING SCHEMES IN VANETS
Clustering algorithms are used to generate stable clusters
according to some rules or criteria.A stable cluster should
have low rate of cluster head change and long cluster
head duration.Clustering schemes improve the performance
of VANET,due to their dynamic nature,high scalability and
load balancing results.In VANET,some algorithms use position
information of the vehicles obtained from GPS as the clustering
criteria, while some others use direction of the vehicles
measured from digital maps.In this section,features of some
important clustering algorithms in VANET are explained.
A. DMAC Clustering scheme [1]
DMAC(Distributed and Mobility-Adaptive Clustering
in Vehicular Networks)aims to generate stable clusters using
force directed algorithms.Force directed algorithms applies
forces to set of edges and set of nodes in the network.By
applying forces to the nodes,it is possible to pull them closer
together or push them further apart.Based on the distance
and velocities,every node applies a force Frel to all its
neighbours.A positive force is applied to the vehicles that
moves in the same direction and a negative force is applied to
the vehicles moving away from each other.The vehicles moves
in the same direction or towards each other are considered for
clustering and no clustering is performed for the vehicle to
which the total magnitude of forces applied is negative.This
scheme suggests the following stages for clustering.
Cluster formation:Each vehicle periodically broadcasts beacon
messages to identify the neighbours.The message contains
node ID,node location,speed vector,total force,state and time
stamp.The distance between the nodes can be determined
by the location information.Each vehicle must maintain and
update the neighbours set according to the periodic beacon
messages.A vehicle should calculate pairwise relative force for
every neighbours using coulomb’s law,accumulated relative
force applied to it and total magnitude of force ‘F’.The
participating vehicles make decisions about cluster formation
based on current state of the node and relation of ‘F’ to its
neighbour’s ‘F’.
Cluster head election:On receiving beacon messages,a
node calculates force applied to the neighbouring nodes by
using the position and relative mobility.This information is
used to determine the suitability of a vehicle to become cluster
head.Nodes having higher number of positive neighbours and
maintaining closer distances to the neighbours,is selected as
cluster head.
Cluster maintenance:Suppose a node comes within R
distance from a cluster-head and if the relative force Frel
of the cluster head is bigger than that of the new node,then
the new node becomes a cluster member.If a cluster member
finds itself at a certain time,have an Frel value bigger than
that of any nearby cluster heads,then it tries to form its own
cluster.
B. Modified DMAC Clustering scheme [2]
Modified DMAC aims to improve stability by reducing
number of cluster head changes.Depending on the node parameters
such as connectivity,energy level and mobility each node
has assigned a generic weight.The node with highest weight is
selected as cluster head.Modified DMAC avoids re-clustering
when group of nodes moves in different directions.This is
achieved by using a parameter called freshness.Freshness(u,v)
is defined as how long a node ‘u’ will be in the transmission
range of node ‘v’.If the connection time is very short,reclustering
is not triggered.
This scheme is based on periodical sending of HELLO
messages which helps each node to get up-to-date information
about their neighbours’ weight.The proposed scheme is implemented
using the following procedures.
init() is called in the cluster formation phase or when a new
node is added to the network.It is used to find a neighbour
with higher weight.In case of tie that is neighbours with
same weight are found,node with higher ID is chosen. The
node itself becomes cluster head if no neighbour node has
higher weight and it will broadcasts a cluster head message.
Otherwise,it will send a JOIN message.
ReceiveHelloMessage: This procedure is performed after receiving
a HELLO message.Freshness is calculated at this stage.
ReceiveJoinMessage: After receiving a JOIN message,this procedure
is called to compare the weight value.Based on the
comparison,the node is elected either as cluster member or as
cluster-head.
Link failure(u): When a node detects a failure of link with node
‘u’,this procedure is called.Here,node ‘u’ is removed from the
neighbour set of node which detects link failure.If node ‘u’
was the cluster head,a cluster head election has to be initiated.
.
C. Mobility based clustering using Affinity propagation(
APROVE) [3]
Affinity propagation is a new technique for data
clustering which aims to create clusters in short time.It is a
clustering algorithm based on message passing between data
points.Affinity propagation finds exemplars which are the
representative of clusters or cluster heads. To describe the
current affinity of a data point and for choosing another data
point as its exemplar,the data points pass messages to one
another.The input of this algorithm is a function of similarities
3
S(i,j) which represents how well it is suited the data point ‘j’
to be the exemplar of the data point ‘i’.The aim of affinity
propagation is to maximize the similarity function S(i,j) for
every data point ‘i’ and its chosen exemplar ‘j’.
Cluster formation: Here,the two types of messages passed
between the data point and exemplar are
1.responsibility(i,j) :This message sends from node ‘i’ to
exemplar ‘j’.It is similar to similarity function S(i,j).
2.availability(i,j) :It indicates j’s desire to be an exemplar for
‘i’ and is send from exemplar ‘j’ back to node ‘i’.
The decision for clustering is made independently by
each node by transmitting responsibility and availability
messages.In order to create stable clusters,the algorithm
uses similarity function which is a combination of negative
Euclidean distances between the current node positions, and
those in the future.Each node ‘i’ has a self similarity S(i,i).
Cluster head selection: Data points having a larger self
similarity initially have more chances of becoming an
exemplar.Clustering decisions are made after every clustering
interval(CI).Clusters that are found by the algorithm minimizes
both relative mobility and distance between each cluster head
and its cluster members.
D. Direction based clustering(C-DRIVE) [4]
In C-Drive,direction of a vehicle is taken as the criteria
for clustering.This scheme aims to control the traffic at the
intersection by determining the vehicle density in advance.In
order to determine the direction in which each vehicle will
take at the intersection,they should be equipped with a digital
map.For clustering,initially the digital map is divided into
different regions.Each region has its own unique id.
Cluster formation: At the intersection,each vehicle can
take four directions Straight(S),Right(R),Left(L) and UTurn(
U).On a particular lane,four clusters can be formed
based on the above directions-S,R,L and U.Each cluster
contains the vehicles which are travelling in the same
direction.With respect to time,the cluster head is the first
vehicle enter into the region in a particular direction.The
density calculation of vehicles in each cluster is an important
function of the selected cluster head.Density calculation is
very helpful for the effective management of traffic and for
adjusting the signal timings.
In direction based clustering,information exchange
takes place by direction based propagation function.This function
ensures that each vehicle will forward the message only to
the vehicle which is in front of it and moving along the same
direction.If the receiving vehicle belongs to the same cluster
as the transmitting vehicle,receiver will accept the message.
E. Dynamic clustering scheme in vehicular networks [5]
Dynamic clustering scheme aims to increase the
stability of inter-vehicular links within the VANET.
Here,clustering is a 3-step procedure,based on movement
of vehicle,UMTS(Universal Mobile Telecommunication
System)Received Signal strength(RSS) and Inter-Vehicular
Distance(IVD).
 Step 1:Clustering based on direction of vehicles: It has
2 steps.Firstly,clustering is done relative to the direction
and then relative to the position of Base Station
Transceiver(BST).
 Step 2:Clustering based on the UMTS Received Signal
Strength: The UMTS signal strength is measured with
respect to the Base Station Transceiver(BST) position.As
the vehicle move towards the BST,the UMTS RSS
increases accordingly.One of the vehicles within the subclusters(
formed in step 1) which move into 3G active
region can become a Gateway Candidate Node(GWC)
based on their UMTS RSS.Then it will receive intense
UMTS RSS and form a single GWC sub-cluster.
 Step 3:Clustering based on IEEE 802.11p transmission
range:Here, each pair of GWCs compare its intervehicular
distance with IEEE 802.11p transmission
range.If the distance is lesser,it form a new cluster or
join an existing cluster(if any one already belongs to a
cluster).
The drawback of this scheme is that,the cluster formation
is a long process.Also,all participating vehicles are treated
with the same priority.So it is necessary to enable QoS for
differentiating the services according to vehicle priorities.
F. Modified C-DRIVE [6]
The main drawback of C-DRIVE approach is the
high rate of change in cluster head,which affects the cluster
stability.Modified C-DRIVE is designed to overcome this
limitation.Similar to C-DRIVE,Modified C-DRIVE (MCDRIVE)
also performs clustering based on direction metric
and all assumptions are same as that of C-DRIVE.Here,three
imaginary points are defined along the road approaching
the intersection.A start point where cluster formation is
initiated,an end point where all clustering operations are
terminated and a threshold point which lies between start
point and end point.These points are base points for cluster
maintenance and cluster head election.It is assumed that the
distance between the start point and intersection point will not
exceed the communication range.When a vehicle arrives at
the start point and detects that it is not a cluster member,the
same is elected as temporary cluster head or Header.
Cluster formation and head election:In MC-DRIVE,clustering
is based on a threshold distance.This value represents the
optimal length of the cluster and depends on the speed and
radio range of the vehicle.The header calculates the threshold
distance and broadcasts a query packet which contains
the header id,cluster id,position,direction information and
threshold distance.This helps to find if there is any vehicle
which travels in the same direction within the threshold.If the
header receives replies from all the eligible vehicles within a
particular time,the farthest vehicle from the header within the
threshold distance is elected as cluster head.This process is
4
repeated at every reference points till the header reaches the
threshold point.
Once a cluster head is elected,to find the cluster
members,cluster head will send a packet which contains the
same fields that of previous packet and also list of IDs of
eligible vehicles.On receiving the packet,each vehicle verifies
its direction and whether the list include its ID or not.If the ID
is absent,it checks whether it is within the threshold distance.If
so it joins to the cluster.
G. AMACAD Clustering scheme [7]
AMACAD(An Adaptable Mobility-Aware Clustering
Algorithm based on Destination) is specially designed for
urban or city environments.It aims to prolong the cluster
lifetime and to reduce the global network overhead.Also
it tries to overcome the challenges such as temporary
fragmentation,congestion,overhead on the base stations
and local servers.In this scheme,the urban area is divided
into smaller segments or regions.All the participating
vehicles maintain a table containing the information about
the location,speed,final destination,relative destination and
available bandwidth of that vehicle.Relative destination
is the nearest destination with respect to the current
region,and the final destination is the actual destination of the
vehicle.Whenever a new vehicle enters into a region,it updates
its relative destination.Relative positions are calculated
based on the latitude and longitude coordinate points on the
road.This clustering scheme mainly involves the following
stages.
Cluster formation: Suppose a vehicle ‘V’ wants to find
a cluster head.It sends an ‘affiliation message’ to all of
its neighbours.If no reply is received,it starts the clustering
operation through the following steps:
Step 1: ‘V’ exchange hello messages with its neighbours to
gather the table information.
Step 2: ‘V’ calculates the relative values using the parameter
table.
Step 3: ‘V’ computes a weight function value(F(V;Z))(directly
proportional to the distance between V and neighbour vehicle
Z,the difference in speed between V and Z and the distance
between relative destinations of V and Z) F(V;Z) should
be determined for all the vehicles within V’s transmission
range.Selection of cluster member is based on the minimum
value of F(V;Z).
Cluster head selection:After the cluster formation phase,a
cluster-head should be selected for each cluster.Node with
minimum value of Fv = n
i=1 F(i;Z) (where ‘n’
is the number of vehicles in the cluster) is selected as
cluster head,.The selected cluster head sends the updated
parameter table to all the vehicles within the cluster.Cluster
heads keep a table containing vehicle’s id,cluster head
id,speed,location,relative and final destination.
Cluster maintenance:When a mobile node wants to move
to another cluster or in a position to select a new cluster
head,the operations such as add member,delete member,update
mobility changes,change of cluster-head and re-clustering
are required.AMACAD perform all the above operations
by considering the threshold values of speed,density and
bandwidth of vehicles.
AMACAD tries to improve the performance of cluster
by considering the parameters such as cluster-head lifetime,
membership lifetime,number of cluster and re-affiliation
rate.
H. Stability Based Clustering for VANETs(SBCA) [8]
SBCA is a clustering scheme aims to improve the
stability and to reduce the communication overhead with in
vehicular network.This involves two phases,cluster setup and
maintenance phase.Vehicles in the close proximity of each
other are selected as cluster members in the cluster set up
phase.A primary cluster head is selected from each cluster.In
order to achieve stability,SBCA proposed a new concept
called secondary cluster-head.A secondary cluster head is
a backup for primary cluster head and is selected for each
cluster.That is,secondary cluster head takes the role of primary
cluster head when it is no longer available.In a particular
cluster,the clustering architecture has to be reconfigured if a
cluster head becomes absent for any reason and will increase
the message overhead.The concept of secondary cluster head
overcome this limitation.
Cluster setup phase: Initially,each node is in undecided
state.If there is an existing cluster head,it will send an Invite
to join message(ITJ message) in every t(j) time units.Upon
receiving ITJ message,each node will check the received
signal strength(P(r)).And if P(r) is greater than a predefined
threshold value(P(th)),node will send a Request To Join
message(RTJ message)to the cluster head.This message
includes node ID and network address.The cluster head
will send back an ACK and the node which receives ACK
becomes the cluster member.As long as a node receives ITJ
message from its cluster head,it remains as a cluster member.
Maintenance phase: The stability and reliability of the
cluster structure generated in the setup phase is achieved
in maintenance phase by using the concept of secondary
cluster head.Each primary cluster head generates a unique
identifier,cluster(id))for the cluster,which is a hash function
of current time and primary cluster head ID. The node which
produces the minimum sum of distance between primary
cluster head and cluster member, and velocity difference
between primary cluster head and cluster member is selected
as secondary cluster head by the primary cluster head.
I. A trust based clustering for VANETs [9]
This clustering scheme aims to produce stable clusters with
less frequency of re-affiliation.Direction and relative speed of
a vehicle is the criteria for constructing the clusters.Vehicles
which are moving in the same direction and speed that is
less than or equal to a pre-defined threshold value(S(TH)) are
5
considered as cluster members.
Cluster formation:Each vehicle ‘V’ will broadcasts a
hello message to all the neighbours with in its transmission
range.Upon receiving the hello messages,recipient will send
back an ACK containing its position and average speed.Vehicle
‘V’ finds the direction of each vehicle from consecutive beacon
messages receiving from its neighbours.’V’ also compares the
relative speed of each vehicle with S(TH).The vehicles which
moving in the same direction that of ‘V’ and with the average
speed that is less than or equal to the S(TH) is selected as
cluster members.Also,this algorithm will construct clusters
of different size based on traffic density and neighbouring
vehicle’s speed.If traffic density is less and speed of each
vehicle is high,it creates large size clusters.If traffic density
is high and average speed of vehicles is less,crates small size
clusters.
Cluster head election:If the algorithm detects a Road
Side Unit(RSU) within the cluster,then RSU is selected as the
cluster head.Because RSU is static and it has more processing
capabilities.If RSU is not present,the slowest moving vehicle
whose total trust value is greater than the predefined trust
value is selected as the cluster head.Trust value is used to
check whether the vehicle is a normal vehicle or a malicious
one.Trust value defines the level of confidence between
a vehicle and its neighbours which is evaluated from the
transaction history.
J. Clustering based on Minimal Path Loss Ratio(MPLR) [10]
This clustering algorithm aims to improve the spectrum
efficiency and to reduce the congestion in the network.Path loss
ratio is the criteria for constructing stable clusters.It works on
a motorway with 3 unidirectional lanes.These lanes have a
speed of 60 mph,70 mph and 80 mph respectively.In every
4 km in the motorway,base stations are installed.Cluster is
formed by grouping the nodes with in the transmission range
of each other.The path loss parameter between the vehicle and
base station is the criteria for selecting the cluster head.The
vehicles within the cluster is communicating using 802.11p
protocol and the communication between the cluster-head and
base station is using LTE technology.
The two parameters considered for clustering are,path
loss which depends on the distance between the base station
and vehicle,and interference between the vehicles due to
overlapping coverage.A good cluster head should have strong
signal to the base station,minimal path loss and it should also
consider the interference from the surrounding vehicles.Higher
path loss means that the vehicle and the road side infrastructure(
base station) is separated by a large distance.When a node
entering into the motorway,it will check for an existing cluster
head with in its transmission range.If present,the new node
will compare its path loss with that of the existing one.Based
on that the cluster head is selected.
K. Distributed clustering with contention-free/contentionbased
MAC protocols [11]
In this paper, a distributed clustering based multichannel
communication scheme is developed in which
contention-free/contention-based MAC protocols are integrated.
DSRCs 75-MHz bandwidth at 5.9GHz band is divided
into 7 channels,each channel has its own functions in clustering
process.Here,each participating vehicle is equipped with two
transceivers.It can operate simultaneously through different
channels. The vehicles within the nearby proximity form a
cluster. Each vehicle can be in 4 states: cluster head, quasicluster
head, cluster member and quasi-cluster member.Quasicluster
head indicates that the vehicle is neither a cluster head
nor a cluster member.
In the clustering process,the cluster head broadcast an
invite-to-join (ITJ) message every t time unit. Once a quasicluster
head receives ITJ message,it will check the received
signal strength. If it is greater than a pre-defined threshold
value,it will consider it as a valid message.Then,it will send
back a request-to-join (RTJ) message which includes the node
ID and network address. Upon receiving the RTJ message,the
cluster head will send back an ACK and also adds that vehicle
into the cluster member list.
A quasi-cluster head can receive the ITJ message
within t time units if it is within the communication range
of the cluster head. If it cannot receive a valid ITJ message
within t units of time, the quasi-cluster head itself becomes a
cluster head. Suppose, a cluster member cannot receive the ITJ
message every T unit (repetition period) from the cluster head,
then state of the vehicle is changed to quasi-cluster member
to guarantee the timely delivery of the safety messages.
IV. SUMMARY OF CLUSTERING SCHEMES:
Through the detailed study of the 11 clustering
schemes,we conclude that each clustering scheme has different
objectives and different criteria for clustering.Most of the
clustering schemes aim to improve the cluster stability.The
AMACAD clustering scheme is applicable for urban or city
environments.It tries to improve the cluster life time and to
reduce the overhead on base stations by performing the clustering
based on destination positions.DMAC performs clustering
based on force directed algorithms.A modified scheme of
DMAC called Modified-DMAC tries to produce stable clusters
by taking generic weight as the clustering criteria,which
depends on connectivity,energy level and mobility of each
node.This scheme also uses a parameter called freshness to
reduce the re-clustering rate.
APROVE is a clustering scheme which uses a new
technique called affinity propagation to generate clusters in
much short time.In C-DRIVE,direction of vehicle is the key
metric for clustering.It aims to control the traffic at junction
points by determining the traffic density in advance.But the
limitation of this method is that the high rate of change
of cluster heads.In order to overcome this limitation,a new
clustering scheme called MC-DRIVE is developed.Similar to
C-DRIVE,it is also based on direction metric.Here,each vehicle
6
TABLE I. COMPARISON OF DIFFERENT CLUSTERING SCHEMES IN
VANETS
Clustering
scheme
Metric Cluster density
Objectives
1.DMAC Force directed
algorithms
Distance dependant
Improve the cluster
stability
2.Modified
DMAC
Speed, Location and Direction
Traffic
direction
dependant
Improve the cluster
stability by reducing
the cluster head
changes
3.APROVE Distance and Speed
proximity
Traffic
direction
dependant
Create clusters much
in less time
4.C-DRIVE Direction of vehicle Traffic
direction
dependant
Control the traffic at
the intersection
5.Dynamic
clustering
scheme
Movement of vehicle
and UMTS RSS
Not
mentioned
Improve the stability
6.MC-DRIVE Direction of vehicle Traffic
direction
dependant
Improve the cluster
stability
7.AMACAD Location, speed, relative
and final destination
Adjustable Prolong the cluster
life time and Reduces
the network overhead
8.SBCA Radio range Not
mentioned
Imprve the stability
and reduce the communication
overhead
9.Trust based
clustering
Direction and relative
speed
Not
mentioned
Stable clusters with
less frequency of reaffiliation
10.MPLR Path loss ratio Distance
from the
base station
Improve the spectrum
efficiency and reduce
the congestion in the
network
11.Distributed
clustering
Contention free or contention
based MAC protocols
Radio propagation
specific
Not mentioned
calculates a threshold distance for clustering purpose which
indicates the optimal length of the cluster.
A dynamic clustering scheme is introduced for
VANETs in [8].Here,clustering is based on vehicle movement,
UMTS received signal strength and inter-vehicular distance..
MPLR scheme takes path loss ratio as a parameter
for clustering.It tries to improve the spectrum efficiency and
overcome the congestion problem in the network.As the
name indicates,SBCA is a stability enhanced scheme which
reduces the communication overhead within the network.It
selects two cluster heads from each cluster,primary and secondary.
Secondary cluster head is a backup for primary.
By taking direction and relative speed of the vehicle as
the criteria for clustering,a new trust based scheme is proposed
in [11].It tries to improve the stability with less frequency of
re- affiliation.Cluster head election is based on a trust value
which helps to check whether the vehicle is a normal one or
a malicious vehicle.
V. CONCLUSION
As VANETs have attracted more attention in recent
years,many research has been addressing all issues related to
them.Distributed nature of vehicular network reduces the network
lifetime and increases the routing overhead by generating
large amount of messages.A clustered VANET structure efficiently
overcome these limitations.In this survey,we first provided
fundamental concepts about VANET,clustering,including
the definition of cluster head and cluster stability and necessity
of clustering for a vehicular network.Then we explained 11
clustering schemes in VANETs.We discussed each scheme
in terms of objective and criteria for clustering,clustering
mechanism and discussed the factors that has to be considered
for constructing stable clusters.
With this survey we realize that a cluster based VANET
structure has many issues to consider,such as cluster stability,
the overhead for cluster formation,cluster msaintenance and
cluster head election and the energy consumption of vehicular
nodes for cluster based communication.Also different clustering
schemes may have different objectives and criteria for
clustering.However,clustering stability is to be considered always
while discussing a clustering scheme.Because,stability is
the measure for the performance and scalability improvement
of a clustering scheme.With this survey,readers can have a
clear idea about VANET clustering,especially those schemes
explained in this article.No clustering scheme is well suited
for all the road traffic scenarios.We hope that this survey
paper helps the researchers to develop more stable and efficient
clustering schemes for VANETs.
REFERENCES
[1] Leandros A. Maglaras and Dimitrios Katsaros, ‘Distributed Clustering
in Vehicular Networks’, 2nd International workshop on Vehicular Communications
and Networking, (May.2007).
[2] G. Wolny, ‘Modified DMAC Clustering Algorithm for VANETs’, International
Conference on Systems and Networks Communications,page
no.268-273 (October.2008).
[3] C. Shea, ‘Mobility-Based Clustering in VANETs Using Affinity Propagation’,
Global Telecommunications Conference,page no.5-14 (November.
2009).
[4] Mounir Boussedjra and Joseph Mouzna and Labiod Houda, ‘Direction
Based Clustering Algorithm for Data Dissemination in Vehicular networks’,
IEEE confernce on vehicular communications,page no.120-130
(2009).
[5] Abderrahim Benslimane and Tarik Taleb, ‘Dynamic Clustering-Based
Adaptive Mobile Gateway Management in Integrated VANET 3G Heterogeneous
Wireless Networks’, IEEE Journal on Selected Areas in
Communications,Volume 29,page no.120-130 (March.2011).
[6] Joseph Mouzna and Houda Labiod and Manoj Devisetty and Manohara
Pai M.M, ‘Modified C-DRIVE: Clustering based on Direction in Vehicular
Environment’,4th IEEE conference on Intelligent Vehicles Symposium,
Page no. 5-9 (June.2011).
[7] M.M.C. Morales, ‘An Adaptable Mobility-Aware Clustering Algorithm
in vehicular networks’,13th Asia-Pacific Network Operations and Management
Symposium (APNOMS,page no.1-6)(September.2011).
[8] Ahmed Ahizoune and Abdelhakim Hafid, ‘A New Stability Based
Clustering Algorithm (SBCA) for VANETs’,6th IEEE Workshop On User
Mobility and Vehicular Networks,(2012).
[9] Rashmi Ranjan Sahoo and Rameswar Panda and Dhiren Kumar
Behera, ‘A Trust Based Clustering with Ant Colony Routing in
VANET’,International conference on Computing communication and
Networking Technologies,(July.2012).
[10] Y Harikrishnan and Jianhua He, ‘Clustering algorithm based on
minimal path loss ratio for vehicular communication’,International
Conference on Computing, Networking and Communications, Mobile
Computing Symposium,(2013).
[11] Khalid Abdel Hafeez and Lian Zhao and Jon W. Mark, ‘Distributed
Multichannel and Mobility-Aware Cluster-Based MAC Protocol
for Vehicular Ad Hoc Networks’,IEEE Transactions on Vehicular
Technology,Volume-62(October.2013).