Professional Documents
Culture Documents
Akram Final Thesis 15 September
Akram Final Thesis 15 September
NOTES : If the thesis is CONFIDENTIAL or RESTRICTED, please attach with the letter from
the organization with period and reasons for confidentiality or restriction
“I hereby declare that I have read this thesis and in my
opinion this thesis is sufficient in term of scope and quality for the
award of the degree of Doctor of Philosophy (Computer Science)”
Signature : ________________________________
Name of Supervisor : PROF. TS. DR. KAMALRULNIZAM BIN ABU BAKAR
Date : 15 SEPTEMBER 2021
BAHAGIAN A - Pengesahan Kerjasama*
Adalah disahkan bahawa projek penyelidikan tesis ini telah dilaksanakan melalui
kerjasama antara ________________________dengan ________________________
Disahkan oleh:
Tandatangan : Tarikh :
Nama :
Jawatan :
(Cop rasmi)
* Jika penyediaan tesis atau projek melibatkan kerjasama.
Nama dan Alamat Pemeriksa Dalam : Prof. Dr. Shukor bin Abd. Razak
Sekolah Komputeran, Fakulti Kejuruteraan
Universiti Teknologi Malaysia
81310 Johor Bahru, Johor
School of Computing
Faculty of Engineering
Universiti Teknologi Malaysia
SEPTEMBER 2021
i
DECLARATION
Signature : ....................................................
Name : MUHAMMAD AKRAM MUJAHID
Date : 15 SEPTEMBER 2021
iii
DEDICATION
iv
ACKNOWLEDGEMENT
v
ABSTRACT
vi
ABSTRAK
vii
TABLE OF CONTENTS
TITLE PAGE
DECLARATION iii
DEDICATION iv
ACKNOWLEDGEMENT v
ABSTRACT vi
ABSTRAK vii
TABLE OF CONTENTS viii
LIST OF TABLES xii
LIST OF FIGURES xiii
LIST OF ABBREVIATIONS xvi
LIST OF SYMBOLS xx
LIST OF APPENDICES xxi
INTRODUCTION 1
1.1 Overview 1
1.2 Problem Background 2
1.2.1 Cluster Head Instability 9
1.2.2 Cluster Member Instability 13
1.2.3 Cluster Maintenance Instability 16
1.3 Problem Statement 21
1.4 Research Questions 21
1.5 Research Aim 22
1.6 Research Objectives 23
1.7 Research Contribution 23
1.8 Research Scope 24
1.9 Significance of the Research 24
1.10 Thesis Organization 25
viii
LITERATURE REVIEW 27
2.1 Introduction 27
2.2 VANETs 27
2.3 Location Service in VANETs 30
2.3.1 Non Cluster-based Location Service Schemes 30
2.3.2 Cluster-based Location Service Schemes 35
2.3.2.1 Static Cluster-based Location
Service Schemes 40
2.3.2.2 Dynamic Cluster-based Location
Service Schemes 45
2.4 Review of Existing CH Election Schemes 50
2.4.1 Cluster Formation Range-based CH Election
Schemes 50
2.4.2 Link Reliability-based CH Election Schemes 52
2.4.3 Remarks and Observation 59
2.5 Review of Existing Cluster Member Affiliation
Schemes 63
2.5.1 Link Lifetime-based CM Affiliation Schemes 63
2.5.2 Remarks and Observations 66
2.6 Review of Existing Cluster Maintenance Schemes 69
2.6.1 Cluster Merging Algorithms 69
2.6.2 Cluster Splitting Algorithms 73
2.6.3 Remarks and Observations 77
2.7 Findings of the Literature Review 79
2.8 Summary 82
RESEARCH METHODOLOGY 83
3.1 Introduction 83
3.2 Operational Framework 83
3.3 Research Design and Procedure 85
3.3.1 Reliable Cluster Head Election Scheme 85
3.3.2 Dynamic Cumulative Custer Member
Affiliation Scheme 87
3.3.3 Optimized Cluster Maintenance Scheme 90
ix
3.4 Simulation Framework 92
3.5 Network Model and Simulation Setup 93
3.6 Performance Evaluation 95
3.7 Assumptions and Limitations 100
3.8 Summary 100
x
5.4.2 Fraction of Location Information Saved in the
Location Servers 135
5.4.3 Query Success Rate 138
5.5 Summary 140
CONCLUSION 165
7.1 Overview 165
7.2 Research Achievements 165
7.2.1 Reliable Cluster Head Election (RCHE)
Scheme 166
7.2.2 Dynamic Cumulative Cluster Member
Affiliation (DCCMA) Scheme 167
7.2.3 Optimized Cluster Maintenance (OCM)
Scheme 167
7.3 Limitations of the ESCLS Mechanism 169
7.4 Directions for Future Work 170
REFERENCES 173
APPENDIX A 187
xi
LIST OF TABLES
xii
LIST OF FIGURES
xiii
Figure 3.5 Design, Implementation and Evaluation of Cluster
Maintenance Scheme 91
Figure 3.6 The Integration of SUMO and OMNET++ Via Veins 93
Figure 3.7 Part of the Doha Map used in the Simulation 94
Figure 4.1 Flow Chart of the Proposed RCHE Scheme 104
Figure 4.2 Centroid and Centroid Vehicle 105
Figure 4.3 CH Election Range 106
Figure 4.4 RSU Communication Range 109
Figure 4.5 Sensitivity Analysis of Weighting Factors 113
Figure 4.6 Average CH Lifetime at Different Maximum Speed 116
Figure 4.7 Average CH Lifetime at Different Maximum Density 117
Figure 4.8 Average Query Success Rate at Different Maximum Speed
118
Figure 4.9 Average Query Success Rate at Different Maximum
Density 119
Figure 4.10 Average Query Delay at Different Maximum Speed 120
Figure 4.11 Average Query Delay at Different Maximum Density 121
Figure 5.1 Flow Chart of DCCMA Scheme 125
Figure 5.2 Road ID and Direction of the Vehicles 127
Figure 5.3 Cumulative Moving Average Speed of the CH 128
Figure 5.4 Update Intervals from the IV to CH and CH to RSU 130
Figure 5.5 Average CM Lifetime at Different Maximum Speed 134
Figure 5.6 Average CM lifetime at Different Maximum Density 135
Figure 5.7 Information Saved in the Location Server at Different
Speed 137
Figure 5.8 Location Information Saved in Location Servers at
Different Density 138
Figure 5.9 Query Success Rate at Different Maximum Speed 139
Figure 5.10 Query Success Rate at Different Maximum Density 140
Figure 6.1 Flowchart of the Cluster Merging 146
Figure 6.2 Overlapping Range of Two Clusters 147
xiv
Figure 6.3 Flow Chart of the Splitting Algorithm 151
Figure 6.4 Cluster Splitting Around the Intersection 152
Figure 6.5 Query Success Rate at Different Maximum Speed 156
Figure 6.6 Query Success Rate at Different Maximum Density 157
Figure 6.7 Query Response Delay at Different Maximum Speed 159
Figure 6.8 Query Response Delay at Different Maximum Density 160
Figure 6.9 Localization Error at Different Maximum Speed 161
Figure 6.10 Localization Error at Different Maximum Density 163
xv
LIST OF ABBREVIATIONS
xvi
ECB - Efficient Cost Based
ETSI - European Telecommunications Standards Institute
FLISLS - Fraction of Location Information Saved in Location Servers
GH - Group Head
GM - Group Member
GPS - Global Positioning System
GRPs - Geographical Routing Protocols
HCBLS - Hierarchical Cluster Based Location Service
IL - Intersection Leader
IEEE - Institute of Electrical and Electronics Engineers
ITS - Intelligent Transportation System
IV - Isolated Vehicle
IC - Intersection-based Clustering
I2V - Infrastructure to Vehicle
IVC - Inter Vehicular Communication
ID - Identity
IS - Intermediate Server
LBS - Location-based Services
LET - Link Expiration Time
LS - Location Service
LE - Localization Error
LOC - Location
LLT - Link LifeTime
LQ - Link Quality
LTE - Long Term Evolution
LRD - Left Relative Destination
MAC - Media Access Control
MATLAB - Matrix Laboratory
MALM - Mobility Assisted Location Management
MoGLS - Mobile Group based Location Service
MG-LSM - Mobile Group-based Location Service Management
MTT - Merging Threshold Time
NVC - Novel Vehicle Clustering
xvii
NS2 - Network Simulator 2
OBU - On Board Units
OCR - Owned Communication Rate
OSM - Open Street Map
OCM - Optimized Cluster Maintenance
OMNET++ - Objective Modular Network Testbed in C++
PDR - Packet Delivery Ratio
PD - Previous Distance
QoS - Quality of Service
QBLS - Quorum Based Location Service
QLSP - Quorum Based Location Service Protocol
QSR - Query Success Rate
QLS - Quorum-based Location Service
QRD - Query Response Delay
R - Range
RD - Relative Destination
RH - Regional Head
RL - Responsible Leaders
RS - Responsible Servers
RSU - Road Side Unit
RLQ - RSU Link Quality
RVM - Relative Velocity Metric
RLT - RSU Link LifeTime
RRD - Right Relative Destination
RCHE - Reliable Cluster Head Election Scheme
RTS/CTS - RTS/CTS - Request-to-send and Clear-to-send
RWCP - Reputation-based Weighted Clustering Protocol
PLM - Power Loss Metric
SCHR - Service Channel Range
SFLS - Semi-Flooding Location Service
SDN - Software Defined Network
SRD - Straight Relative Destination
SUMO - Simulation of Urban Mobility
xviii
SNR - Signal to Noise Ratio
SV - Source Vehicle
TCP - Transmission Control Protocol
TDMA - Time Division Multiple Access
TDFLS - Totally Distributed Flat Location Service
UN - Undecided Node
UAV - Unmanned Aerial Vehicle
VANETs - Vehicular Ad Hoc Networks
V2V - Vehicle to Vehicle Communication
VI - Variability Index
V2I - Vehicle to Infrastructure
V2V - Vehicle to Vehicle
WAVE - Wireless Access in Vehicular Environments
VEINS - Vehicles In Network Simulation
ZGLS - ZoomOut Geographic Location Service
xix
LIST OF SYMBOLS
α - Alpha
Ωi - Aggregate relative velocity
β - Beta
∈ - Belongs to
𝛥𝑆𝑡ℎ𝑟𝑠ℎ - CM affiliation threshold value
±Vth - CM affiliation threshold value
𝑀(𝑖) - Capability metric
𝑡𝑖𝑏 - Cluster head election time
𝑡𝑖𝑚 - Cluster headship loosing time
𝑡𝑖𝐽 - Cluster joining time
𝛿𝑖 - Number of neighbors
⋀v - Mobility Metric
xx
LIST OF APPENDICES
xxi
INTRODUCTION
1.1 Overview
In fact, VANETs provide various services such as finding the closest parking
slot, gas station or restaurants and discovering the position of any other vehicle
including emergency vehicles, patrolling police vehicles and resource discovery
vehicles. These services mainly depend on the accuracy and reliability of the location
service (Al-Mayouf et al., 2016; Schoch, Kargl, Weber, & Leinmuller, 2008).
1
and sudden network disconnections. The development of a reliable, scalable, and
stable location service is a big challenge in VANETs (Woo and Lee, 2018;
Balouchzahi, Fathy, & Akbari, 2016).
2
To perform different location service functions such as location updates and
location queries, several schemes have been proposed in VANETs. Location service
schemes are divided into various categories based on different characteristics. These
location service schemes are classified into two main classes such as flooding-based
and rendezvous-based schemes (Ayaida, Fouchal, Afilal, & Ghamri, 2012). The
Rendezvous-based schemes are further divided into Quorum-based and cluster-based.
Moreover, cluster-based location service schemes are categorized into static cluster-
based location service and dynamic cluster-based location service. The different
categories of the existing location service schemes are given in Figure 1.2.
3
location updates and queries are simple in these schemes. However, the flooding
schemes are inefficient and not suitable for large scale networks which creates
scalability and network management problems. The flooding schemes suffer from the
high network congestion and low throughput (Balouchzahi et al., 2016; Garg, Pandey,
& Singh, 2014).
4
has been proposed to limit the communication of vehicles within the location servers.
Cluster-based location service is more efficient because aggregated messages are sent
to the location servers.
The static cluster-based location service depends on the static clusters. In the
static clustering, the location of the cluster is predefined through dividing the area into
fixed segments. The group of vehicles in a segment forms the cluster. In some cases,
Road Side Unit (RSU) is also used to form static clusters. In each segment, the vehicle
that is nearest to the center of the segment is elected as a CH. The other vehicles join
the cluster as CMs. Every vehicle in a cluster sends its location to the CH. CH sends
the locations to RSU that acts as a higher-level location server. Static clustering has
some advantages such as simple cluster formation and low signaling cost during CH
election (Singh and Kaur, 2015). However, the CH election depends on fixed locations.
Location servers are elected without considering mobility parameters. When these
location servers leave their respective areas, new location servers are defined. Due to
this, location server election process occurs frequently that affects the performance of
the location service. The CHs depend on intermediate servers to update the locations
to higher level server. Likewise, frequent handover occurs when location server moves
between different areas. Due to the abovementioned issues, CH and CM change
frequently. Consequently, the CH and CM instability increases, which affects the
performance of the location service.
5
The dynamic cluster-based location service relies on dynamic clustering. The
dynamic clustering is also termed as mobile clustering. It depends on the mobility
parameters such as speed and position to make the clusters. In the static clustering,
clusters are defined by dividing the road into fixed segments as shown in Figure 1.5.
A separate CH is elected for each segment. The lifetime of CH is limited to segment.
The CHs and CMs switch between different clusters. Whereas in dynamic clustering,
the clusters are not bounded to the fixed partition of the road as shown in Figure 1.4
and in Figure 1.6, CHs and CMs move with clusters. The main advantage of dynamic
clustering is the reduction of reclustering by moving the cluster with the vehicles
(Abdel-Halim, Fahmy, & Bahaa, 2019; Dhugga, Sharma, & Sharma, 2015; Singh and
Kaur, 2015). In dynamic cluster-based location service, vehicles having similar
mobility characteristics form one cluster and move within clusters as shown in Figure
1.3.
6
clustering parameters, communication overhead and isolated vehicles as mentioned in
Figure 1.4 (Bi et al., 2020; Woo and Lee, 2018).
In the clustering approach, one vehicle is elected as a CH, other vehicles join
the cluster as CMs. The CH not only performs cluster management functions such as
communication with its CMs, inter-cluster communication, information exchange with
RSUs, also performs other functions in the cluster-based location service such as to
keep and update the location information of its CMs to higher-level location servers.
The CH responds to different queries about the location of other vehicles in the
network. The unavailability of the CH due to any reason breaks the cluster. All the
communication between CH and other vehicles disrupted, hence the performance of
the working application halted. The stability of the CH is the basic requirement to
maintain the stability of the cluster.
The stability of the cluster depends on the parameters used to elect CH, CM
and cluster maintenance. Some of the important parameters are speed, distance,
direction, number of neighbours, link lifetime, destination, signal-to-noise ratio, fixed
size road segments, road id and density. Due to rapid topological changes in VANETs,
speed is considered as a most important parameter while forming the clusters. The
combination of various parameters is used to increase the cluster stability, while
keeping the requirements of different applications.
7
Communication overhead is another important parameter that impacts the
stability of a cluster. With the increase of the communication messages, the network
overhead increases which causes the delay and collision of messages. The
communication links between CH and CM breaks, which ultimately affects the
stability of the cluster.
The isolated vehicles are not part of any cluster. These vehicles send CM
affiliation messages continuously to become the member of the clusters. With the
increase of isolated vehicles, the number of CM affiliation messages increases, which
creates extra communication overhead in the network. The communication between
CH and CM is disrupted which reduces the lifetime of the CH and CM.
8
1.2.1 Cluster Head Instability
In the static cluster-based location service, the road is divided into segments.
The vehicle nearest to the center of the segment at any instant is elected as a CH as
shown in Figure 1.5 (Aissaoui et al., 2015). This approach simplifies the CH election
and reduce the management functions by defining the CH election range. However,
the election of the CH is bounded to the fixed point without considering the speed
parameter. Furthermore, the CH is elected without including the link reliability with
the RSU. Also, due to the fixed size road segments, communication range of
neighbouring CHs overlaps each other. The CMs of these segments receive CM
requests from the multiple CHs. The location information of these CMs is not updated
due to the collision of the messages. Due to the frequent change of CHs, non-optimum
cluster formation range and unreliable link with the RSU, the CH instability increases.
9
vehicle will broadcast the information of that vehicle. Same procedure is performed
by all the vehicles, until, a vehicle which has optimum CHEV is elected as a CH as
shown in Figure 1.6. To control flooding, messages have delay more than threshold
value are discarded.
10
The aforementioned location service schemes utilize the clustering approach
to improve the performance of the location service by electing a stable vehicle as a
CH. However, CH instability still exists due to non-optimum cluster formation range
and due to unreliable link of the CH with the higher-level location server. The
following studies based on general clustering (in which clusters are formed without
the requirements of any specific application) tried to address aforementioned issues.
Arkian, Atani, Pourkhalili, and Kamali (2015) depend on the static and
dynamic CHs to enhance the cluster stability. Every RSU is working as a static CH.
The dynamic CH is elected from vehicles on the basis of distance, speed,
neighbourhood degree and RSU Link Quality (RLQ). The proposed scheme improves
the cluster stability by using mobility and Quality of Service (QoS) metrics. The link
reliability with the RSU is considered while electing the CH. The leading vehicle starts
the process of the clustering. The mobility and RLQ metrics are computed
independently without synchronizing each other. This factor does not ensure that a
vehicle that is most stable with its neighbours also has a reliable link with the RSU.
Similarly, it does not guarantee that a vehicle having a reliable link with the RSU is
also most stable with its neighbours.
The Base Station (BS) collects the information about the number of vehicles in
the network and divides the area into different partitions based on the density of
vehicles and number of required clusters (Qureshi, Abdullah, Bashir, Iqbal, and Awan,
2018). The BS defines the centroid of each partition. The CH for each partition is
elected on the basis of signal strength and direction of other vehicles and distance from
the centroid. The partition is based only on density without considering the length of
the road. There may be very high density in a small area which leads to the increase in
number of clusters. Due to the neighbouring clusters, the overlapping area increases.
Besides this, speed of the vehicles and reliable link with the BS are ignored while
electing the CH. Due to static approach, the number of clusters and CH instability
increases.
Pal, Gupta, Prakash, and Tripathi (2018) define the centroid of cluster on the
basis of the current position of vehicles. The CH is elected from the vehicles in the
11
range of the centroid by considering the speed and distance from the centroid. The
uniqueness of the centroid is affected due to different number of vehicles in the
neighbor table of each vehicle, which does not ensure a reliable cluster formation
range. Also, CH is elected without considering the reliability with the RSU.
In the light of earlier mentioned related works, it is observed that the cluster-
based location service schemes are more focused on improving the location service
functions such as location updates and location queries schemes instead of reducing
the cluster instability. Existing CH election schemes which include HCBLS, CBLS
and MoGLS increase the CH instability due to non-optimum cluster formation range
and unreliable link with the RSU. In the aforementioned studies, a Centroid Vehicle
(CV) is defined before the cluster formation range. CV is elected from the vehicles on
the basis of neighbour table. Each vehicle has a different table which compromise the
uniqueness of the CV, which ultimately affects the number of vehicles in range. In the
HCBLS scheme, clusters are defined based on the road partitions. Each CH sends and
receives messages up to 400m on both sides of it. Due to overlapping range, expected
CHs receive multiple messages from neighboring vehicles. Collision of messages take
place which affects the location updates and location queries. In CBLS and MoGLS,
CH is elected without defining a specific range. The messages beyond a certain delay
are discarded. The communication overhead increases during CH election. The
collision of messages increases. The overlapping range and undefined range leads to
non-optimum cluster formation range. These issues calculate unreliable Cluster Head
Election Value (CHEV) which result in unstable CH election. Consequently, these
issues increase the CH instability. The location updates and location queries suffer due
to the CH instability which is acting as a lower-level location server. The performance
of the location service degrades in terms of the query success rate, query response
delay and localization error. Therefore, a reliable CH election scheme should be
incorporated in VANETs that addresses the aforementioned issues effectively.
12
1.2.2 Cluster Member Instability
In the CBLS scheme, the CM affiliation depends on the speed difference with
the CH. The CH allows those vehicles to join the cluster whose speed difference with
the CH is less than 18 km/h. Moreover, the maximum number of CMs in each cluster
is limited to 20 (Asoudeh et al., 2017). The stability of the CM in a cluster is increased
by using speed. However, due to sudden and frequent topology changes, speed varies
instantaneously which does not ensure the long lifetime of a CM in the cluster. Also,
those vehicles whose speed difference is greater than 18 km/h cannot affiliate to the
cluster. The isolated vehicles increase in the network due to fixed CM affiliation
threshold values. The location information of these vehicles are not updated to the
location servers.
13
The MoGLS scheme proposed by Woo and Lee (2018) addresses the CM
VVIP instability by using speed and distance parameters to affiliate the vehicles as CMs.
Every vehicle calculates its sojourn time, the time that a vehicle remains in the range
of a CH for which it is responsible, compares this sojourn time with fixed threshold
value sent by the CH. If the value of the sojourn time is greater than predefined
threshold value, the vehicle joins the cluster as a cluster member otherwise waits for
the messages from the other CH. If there are more than one CHs in neighboring area,
the vehicle joins the cluster with a greater sojourn time as shown in Figure 1.7.
14
The positions of the isolated vehicles are not updated to the server due to which query
success rate decreases.
Cambruzzi et al. (2016) and Huo et al. (2016) vehicles depend on the LLT to
affiliate with the CH. The vehicles join the cluster based on the larger staying time in
a cluster. The CM affiliation threshold value is determined by vehicle itself. Although,
the number of the isolated vehicles decreased, however, CMs switch between different
clusters due to instability of the existing CHs. Moreover, the LLT depends on the
instantaneous value of the speed, which does not confirm a reliable CM affiliation.
The lifetime of a CM in a cluster decreases.
More precisely, with the frequent and rapid changes in VANETs, instantaneous
mobility information obtained through beacon messages results into inaccurate link
estimations (Srivastava, Prakash, & Tripathi, 2020). Isolated nodes continuously
search nearby clusters to affiliate. If the number of isolated nodes in the network
increases, the network congestion and network overhead also increases which lead to
poor network performance (Huo et al., 2016).
15
cluster. Also, each CH defines a threshold value to allow the vehicles with the existing
cluster. However, due to fixed threshold values, some of the vehicles could not affiliate
to any cluster and remain isolated. Instantaneous values lead to unreliable affiliation
with the CH. The affiliated CM change clusters frequently. The location updates and
location queries are affected. Also, due to fixed threshold values, the number of
isolated vehicles increases. Isolated vehicles continuously send and receive messages
to affiliate with the nearby clusters. The exchange of communication messages
between CH and CH increases which leads to extra communication overhead. The
delay and collision of messages increases. The locations of isolated vehicles are not
updated to any cluster which ultimately affects the performance of the location service.
Due to high mobility of the vehicles, the cluster structure and network topology
change frequently. Several events are triggered at cluster level (Cheng and Huang,
2019) such as, existing CM loses its affiliation with the current cluster, the existing
CH deteriorates its stability or disappears due to issues in communication hardware,
overlapping of the clusters in neighboring area and cluster splitting around the
intersection. The stability of the clusters is suffered due to these events (Bi et al., 2020).
16
& Varadarajan, 2019). In this research, the cluster maintenance is performed by
focusing on the cluster merging and cluster splitting.
The stability of the cluster is affected when several clusters co-exist within a
short distance. The clusters overlap each other. The merging of clusters depends on
the overlapping range of the clusters and on the duration (also called Merging
Threshold Time (MTT) before the start of the merging. Due to topological changes in
VANETs, the distance and range change suddenly and frequently. So, in order to make
merging effective, the overlapping CHs wait up to MTT before the start of the merging.
After completing the merging, the next step is to elect the CH for the merged cluster.
The CMs of the merged cluster affiliate with the new CH (Awan et al., 2020; Farooq,
Ali, & Rehman, 2016; Lin et al., 2016).
The merging of the cluster is initiated when a CH determines that all its CMs
are merged into another cluster; the distance between CHs is less than 100m; and
sojourn time is greater than threshold value (Woo and Lee, 2018). The proposed
MoGLS scheme tried to make the merging of the cluster stable by adding range,
distance and sojourn time conditions. However, the MTT does not ensure that merging
clusters move towards each other or move away from each other. During MTT, the
merging clusters may move away from each other while staying in the range of each
other. The merging is started and completed which lead to unstable merging.
Moreover, the new CH is elected without considering the reliable link with the RSU.
These issues increase the cluster instability.
17
Senouci et al. (2019a) merging process is invoked when two neighbouring
clusters overlap each other over a merging time. The CH with large number of the
CMs is elected as a CH of the merged cluster. Although, the proposed study simplifies
the merging criteria, however, the merging range and MTT are not defined. Also, the
new CH is elected without the link reliability with the RSU. Due to these issues, the
merging of the clusters is not affective and cluster instability increases.
When one CH comes into the range of another CH and both CHs are moving
in the same direction, then the cluster merging starts (Haider, Abbas, Boudjit, and
Halim, 2020). The vehicle which is at the middle of the merged cluster is elected as a
CH of the merged cluster. Although merging criteria is simple but without defining the
specific values of the merging conditions such as range and MTT, the merging process
is triggered frequently which increases cluster instability. Moreover, CH depends only
on the distance from the other vehicle without considering link status with the RSU.
Cluster instability occurs around the intersection due to divergent of the traffic
flow towards straight, left and right as arrowed in Figure 1.8. Existing CH moves only
in one direction and CMs of other directions lose their affiliations from the CH. The
following section analyzes the existing cluster splitting studies around the intersection.
Zhao, Liu, Wu, and Liu (2016) depend on the direction of the vehicles to
manage the clustering around the intersection. It is supposed that each vehicle knows
its turning direction before entering into the intersection area. The vehicles are
considered within the intersection area, when the distance between center of the
intersection and vehicles is less than 2R, where R is the transmission range. The most
18
middle vehicle in each direction is elected as a CH. Although the existing study tries
to split the existing cluster based on the direction parameters, but some limitations still
exist such as, when the existing cluster enters into the intersection area, the vehicles of
this cluster is splitted into three sub-clusters on the basis of left, right and straight
direction as shown in Figure 1.8. The CH of the existing stable cluster is re-elected.
The cluster instability increases due to clustering after entering the intersection area
and reclustering after leaving the intersection area. The CHs of the split clusters are
elected without considering the link status with the RSU. These issues increase the
cluster instability which degrades the performance of the location service.
Zhou, Wu, and Wang (2017) attempt to improve the existing schemes by
reducing the clustering in the intersection area. In the proposed algorithm, the Base
Station (BS) elects the CH from the road segment which has large number of neighbors
and lowest Cluster Head Election Value (CHEV). The Weighted CHEV depends on
the relative position between vehicles and the relative distance from the vehicle to the
BS. The existing study tries to reduce the unnecessary clustering around the
intersection by electing only one CH in the intersection area. The CH election is based
only on distance parameter. The link reliability of CH with the BS is included by
considering the distance from the BS. Here, only CH has to managed the whole
intersection traffic. At the intersection, the vehicles are moving in different directions.
The elected CH can move only in one direction. When the distance between CH and
vehicles of other road sections increases beyond the communication range, vehicles
will lose their affiliations; as a result, the cluster instability increases.
19
According to the aforementioned information, existing studies try to manage
the clustering around the intersection. Existing schemes performs the clustering when
cluster enters the intersection area and reclustering when cluster leaves the intersection
area. Even a stable cluster is reclustered. Cluster instability increases due to
unnecessary clustering. New CHs are defined at the cost of cluster instability. Also,
new CHs are elected without considering the link status with the RSU.
20
1.3 Problem Statement
The cluster-based location service relies on the stability of the cluster. Cluster
stability is frequently affected because of high mobility, unreliable communication
links, and frequent changes in the topology. In the cluster-based location service, every
CH updates the locations of its CMs to RSU. In the existing cluster-based location
service schemes HCBLS, CBLS and MoGLS, CH election is based on the reliable link
with other vehicles ignoring the link reliability with the RSU. Moreover, CH is elected
without optimum cluster formation range. The unreliable link with the RSU and non-
optimum cluster formation range surge CH instability. The CM is the vehicle whose
position is cached in the location servers. However, due to the reliance on
instantaneous values of the speed, CMs join and leave clusters frequently. Likewise,
some of the vehicles are not affiliated to any cluster due to the fixed CM affiliation
threshold values. The isolated vehicles in the network increase. These issues increase
the CM instability. After electing the CH and affiliating the vehicles as CMs, the
cluster stability depends on the efficiency of the cluster maintenance schemes.
However, non-optimum cluster merging and splitting; and unreliable links with the
RSU increase the cluster instability. In the cluster-based location service, location
updates and location queries are negatively affected with the CH instability, CM
instability and non-optimum cluster maintenance schemes. The CH and CM stability
is measured by CH lifetime and CM lifetime. The higher value of the CH lifetime
represents that CH is more stable and serve the cluster for longer time. Also, the higher
value of CM lifetime indicates that CM remains affiliated for larger time. The location
updates and location queries are improved which ultimately enhance the performance
of the location service in terms of the query success rate, query response delay and
localization error. The lower values of CH lifetime and CM lifetime leads to CH
instability and CM instability respectively.
21
i. How to improve the process of electing a CH in order to enhance the CH
stability?
a) How to calculate the cluster formation range that increases cluster
stability?
b) How to evaluate the reliability of communication link between the
potential CH and RSU in order to enhance the CH stability?
ii. How to enhance the process of affiliating a vehicle in order to improve the
CM stability?
a) How to evaluate a reliable CH to affiliate a vehicle to improve CM
stability?
b) What are the necessary parameters to calculate CM affiliation threshold
value to reduce the isolated vehicles in the network?
iii. How to enhance the process and conditions of cluster maintenance to improve
cluster stability.?
a) What are the merging conditions need to be optimized to increase the
cluster stability?
b) What are the parameters that should be considered for reducing the
unnecessary clustering to improve the cluster stability?
c) What are the significance parameters that should be considered to enhance
the link reliability with the RSU?
22
1.6 Research Objectives
The objectives of this research are designed based on the research questions
mentioned in the Section 1.4 as follows.
i. The Reliable Cluster Head Election (RCHE) scheme enhances the CH stability
through optimizing the cluster formation range and improving the link
reliability of the CH with the RSU. The performance of the RCHE is evaluated
in terms CH lifetime, QSR and QRD.
ii. The Dynamic Cumulative Cluster Member Affiliation (DCCMA) scheme
improves the CM stability through enhancing the link lifetime with the CH and
using dynamic CM affiliation threshold values. The performance of the
DCCMA is evaluated in terms of CM lifetime, QSR and fraction of vehicles
saved in the location servers.
iii. The Optimized Cluster Maintenance (OCM) scheme enhances the cluster
stability by optimizing cluster merging and splitting. The performance of the
OCM is evaluated in terms of QSR, QRD and LE.
23
1.8 Research Scope
iv. Communication between Vehicle to Vehicle (V2V) and Vehicle to RSU is based
on IEEE 802.11p standard.
v. Focus on Urban VANETs which means network scenario is considered inside the
city.
vi. In this study V2V, Vehicle to Infrastructure (V2I) and Infrastructure to Vehicle
(I2V) communications are considered.
vii. The network scenario does not include complex city topologies such as bridges,
tunnels, and curved roads. Also, the network scenario does not include U-turns.
viii. Security-related issues for vehicles and RSUs are not considered.
24
1.10 Thesis Organization
25
Chapter 7 concludes this work by summarized research achievements and
outlines the future research directions.
26
LITERATURE REVIEW
2.1 Introduction
In this chapter, a critical review of the latest and relevant literature is presented
which provides the necessary background information and establishes the foundation
for the material presented in the subsequent chapters. Firstly, an overview of VANETs
is presented in Section 2.2 which explains the concepts and characteristics of the
VANETs. Section 2.3 describes the different types of location service schemes in
VANET. Subsequently, Section 2.4 critically discusses the existing CH election
schemes in VANETs. Thereafter, Section 2.5 reviews the existing CM affiliation
schemes. Next to that, Section 2.6 investigates the shortcomings of the existing cluster
maintenance schemes. Eventually, based on literature review, the unresolved issues
regarding cluster instability are identified in Section 2.7. Finally, this chapter is
concluded by a summary in Section 2.8. The general structure of the literature review
is presented in Figure 2.1; arrow represents the hierarchy of location service in
VANETs.
2.2 VANETs
VANETs are the networks consisting of a large number of vehicles which have
sensing, processing, and moving capabilities. The unique characteristics of VANETs
include: rapid topological changes, frequent network disconnections, high node
mobility, large number of vehicles and frequent density variations (Haider, Abbas,
Abbas, & Baker, 2019; Sharma and Kaul, 2018). In VANETs, vehicles’
communication is based on two scenarios: Vehicle to Vehicle (V2V) and Vehicle to
Infrastructure (V2I) or Infrastructure to Vehicle (I2V). Infrastructure may be an RSU,
Base Station (BS) or any other similar device. Vehicles depend on V2V
27
communication to exchange information with each other. In V2I, vehicle sends
messages to infrastructure whereas I2V is used to send messages from server to
vehicles as shown in Figure 2.2 (Ullah, Yaqoob, Imran, & Ning, 2018; Younes and
Boukerche, 2015).
28
All vehicles are embedded with communication device called On Board Units
(OBU) to exchange information (Arif, Wang, Bhuiyan, Wang, & Chen, 2019; Manvi
and Tangade, 2017).
The VANETs communications V2V and V2I are provided through Dedicated
Short-Range Communication (DSRC) standard that is based on IEEE 802.11p protocol
(Sheikh and Liang, 2019). The DSRC operates in the 5.9 GHz band in which 75MHz
is reserved for the VANETs applications. The available 75MHz is divided into 7
channels with 10MHz for each channel and a guard band of 5MHz (Shah et al., 2019;
Zemouri et al., 2015). The Control Channel (CCH) is dedicated for the safety
applications while the Service Channels are reserved for the non-safety applications as
shown in Figure 2.3 (Santamaria, Tropea, Fazio, & De Rango, 2018; Ghandour, Felice,
Artail, & Bononi, 2014).
29
The RSU is a fixed communication device deployed on the roadside to manage
the network. RSUs work as a gateway between VANETs and the other types of
networks (Pan, Cui, Wei, Xu, & Zhong, 2019; Vegni and Loscri, 2015). An
infrastructure-based network relies on RSUs to provide different types of services such
as getting the information about nearest restaurant, gas station, petrol pump and
parking stands (Boussoufa-Lahlah et al., 2018; Ali, Malik, Rahman, Iqbal, &
Hamayun, 2016; Hoeft and Rak, 2016).
The RSU based applications need different types of the information such as
weather, vehicles density, location and information about different services. Vehicles
send information to the nearest RSU, which exchanges it with higher level servers to
process the information. (Ali et al., 2019; Al-Mayouf et al., 2018; Ganeshkumar and
Gokulakrishnan, 2015). Different types of applications such as routing, channel access
management, information dissemination, security and location service exploit RSU to
get required information (Awan et al., 2020; Al-Otaibi, Al-Nabhan, & Tian, 2019).
Location service is used to keep the record of the current position of vehicles
in VANETs. Referring to section 1.2, the existing location service schemes are divided
into non-cluster based location service schemes and cluster-based location service
schemes. The critical analysis of existing location service schemes, cluster-based and
non-cluster based location service, is given in the following section and is highlighted
in Table 2.2.
30
schemes depend on the flooding approach to provide the location of a vehicle. The
description of these schemes is given in the following sections.
Wu, Ohzahata, JI, & Kato (2014) depend on 3-hop distance to disseminate the
location information. The backbone vehicle updates and responds the location queries.
31
The Totally Distributed Flat Location Service (TDFLS) scheme reduced the
communication overhead that occurred due to flooding using 3-hop communication.
However, location update and location query are still based on flooding approach.
32
In the Quorum-based location service, Stojmenovic, Liu, and Jia (2008)
vehicles update their location information in north and south directions. All the
vehicles which receive the location updates make a north-south column. The vehicle
which needs the location of another vehicle broadcasts a query in the east and west
direction to search through the whole network. All participating vehicles make west-
east row. The location information of the required vehicle is obtained at the
intersection of the row and column as shown in Figure 2.5.
33
information. In the proposed ZGLS scheme, one-hop neighbours are organized into
relatives. During location update, the relatives act as a chain of location servers. The
location update messages move in front and backward directions using these chain of
location servers. When a vehicle generates a query request, a relative that is nearest to
the querying source responses the query. Location update and location query schemes
are simplified but have issues during quorum constructions. The location updates
transverse the whole network through the chain of location servers without any specific
boundary limits. The location update quorums create extra communication overhead.
Also, the performance of the ZGLS is affected in the absence of any infrastructure
such as RSUs (Yim et al., 2017).
34
2.3.2 Cluster-based Location Service Schemes
35
Rossi, Fan, Chin, & Leung, 2017; Ren, Khoukhi, Labiod, Zhang, & Veque, 2017a;
Ren, Zhang, Khoukhi, Labiod, & Vèque, 2017b; Azizian, Cherkaoui, & Hafid, 2016;
Huo et al., 2016; Singh and Bali, 2015; Arkian et al., 2015; Ucar, Ergen, & Ozkasap,
2013). The CH election schemes are optimized without considering the requirements
of specific applications. Therefore, stability of the cluster is suffered when these
schemes are used in different applications such as location service. The location
service has its own characteristics in which CH has to perform location updates and
location queries apart from other cluster management functions. So, there is a need to
design clustering schemes while keeping with the requirements of location service.
36
The main objective of the Channel Access Management (CAM) is to improve
the performance of different applications such as information dissemination, routing
and security by optimizing the channel utilization. Different approaches such as
RTS/CTS and CSMA/CA are used to manage the channel access. In the cluster-based
channel access management, the CH allocates the bandwidth for its CMs on the basis
of predefined criteria. (Ağgün et al., 2020; Chaurasia et al., 2019; Pal et al., 2018;
Gupta, Prakash, & Tripathi, 2017; Gupta, Prakash, & Tripathi, 2016; Zhang et al.,
2014; Hafeez et al., 2013; Yang, Tang, & Huang, 2013; Almalag, Olariu, & Weigle,
2012; Hafeez, Zhao, Liao, & Ma, 2011). In the channel access management, the main
task of the CH is to assign channels to its cluster members according to the
requirements of the applications. The ultimate objective of the CAM based clustering
is to make efficient use of bandwidth to avoid collisions. However, the location service
clusters aim to maximize their stability by focusing on vehicles’ mobility parameters.
It is obvious that clustering for the purpose of channel management does not consider
into account the requirements of location service.
37
Clustering plays a vital role in position based routing through forming a
hierarchical structure over the existing adhoc network. The use of clustering in routing
protocol minimizes flooding and reduces interfacing between vehicles. The scalability
of the routing protocols is increased through the use of clustering. In the majority of
the cluster-based routing protocols, CH performs the inter-cluster communications
(Aravindhan and Dhas, 2019; Zhang et al., 2019; Abuashour and Kadoch, 2018;
Moridi and Barati, 2017; Abuashour and Kadoch, 2016; Lin et al., 2016; Mohammed,
Abdelgader, Wang, & Shen, 2016; Mehra, Bali, & Kaur, 2016; Satyajeet, Deshmukh,
& Dorle, 2016; Abboud and Zhuang, 2015; Paul, Ibrahim, Bikas, & Naser, 2012; Raw
and Das, 2011). When clustering is used for the purpose of routing, the concern is to
create clusters that are stable enough to deliver the messages from the source to the
destination. In the position-based routing, to find the location of the destination, source
node uses location service and broadcasts location requests to the intermediate nodes
until it reaches the destination. The destination node replies with a geo-unicast packet
(Sandonis et al., 2016). The performance of the position-based routing depends on the
efficiency of the location service.
38
Table 2.1 Location Service Specific Characteristics
Characteristics
Cluster In the cluster based location, at a lower level, CH collects and
Stability maintains the locations of its CMs. Performance of the location
service highly depends on the stability of the clusters.
Minimum Isolated or Undecided Nodes (UN) are those vehicles which could
Isolated not be affiliated to any cluster due to CH or CM affiliation criteria.
vehicles In order to enhance the performance of location service, there
should be minimum isolated vehicles in the whole network.
Efficiency The performance of the location service depends on location
update and location query schemes. Overhead of location updates
and location queries should be kept minimum to improve the
efficiency of the location service.
Robustness Location service utilizes location server to maintain locations.
Location service should not be disrupted due to failure or
disconnection of these server from network.
Load Balancing Location service should utilize load balancing techniques to avoid
congestions in the network.
Locality Distance between location server and querying source should be
Awareness less than the distance between source and destination. Location
query packets should not be propagated unnecessarily over the
whole network.
Accuracy Location service should provide accurate and up-to-date location
of vehicles in the network.
The critical analysis of the cluster-based location service which includes static
and dynamic clustering is given in the following section.
39
2.3.2.1 Static Cluster-based Location Service Schemes
40
update their information to location servers, but it is very difficult for the vehicles
which are away from the dedicated servers to update their position. Location servers
are elected without considering mobility parameters which creates location server’s
instability. Furthermore, when location servers enter or leave the current area,
handover issues are created. Even more to this, an extra communication overhead is
created which degrades the performance of the location service.
41
A real time traffic monitoring system developed by Aissaoui et al. (2014) is
based on the usage of RSUs. A road is divided into fixed size units as shown in Figure
2.7, where each unit defines a cluster. The vehicle nearest to the unit center is
automatically elected as a CH. Every vehicle shares its speed, location and direction
information with the CH periodically. To keep up to date view of the vehicles on the
road, CH updates this information to its nearest RSU. Location information of those
vehicles which are within the cluster but the outside range of CH is not updated to the
CH. CH is elected without considering mobility parameters due to which cluster
instability increases. This factor will degrade the performance of the monitoring
system that is based on the location service.
42
each CM affiliates with a new CH in each segment. The HCBLS did not proposed any
cluster maintenance scheme, which further increases the cluster instability.
For resolving the lack of RSUs, Aissaou et al. (2016) use vehicles nearest to
the intersections as RSU. At the intersections, where RSUs are not available, one of
nearest vehicle to the intersection is elected to perform the functions of an RSU. The
vehicle with the lowest speed within a defined radius is elected as an RSU. Typically,
in normal CM-CH scenario, every CM sends its location information to its respective
CH. The CH updates these locations to an RSU but in this intersection scenario, the
CH sends location information to one of its CM which is not stable as compared to the
CH. An RSU designation is based on lowest speed; and fast speed CHs bypass quickly
without handing over location information of its CMs. High density also affects the
election of vehicle as RSU. Although this study has tried to reduce RSU deployment
cost, but dependency on the instable vehicle affects the location update and location
queries, which ultimately will degrade the performance of the location service.
Van Gulik, Phillipson, and Fouchal (2018) depend on RSU to define the static
clusters. Each RSU does not only provide the location of the target vehicle, but also
provides the best path from source to destination. Congestion is reduced by assigning
penalty cost to the selected roads. The focus of this study is to optimize the routing by
reducing congestion. The performance of the location service is affected due to
depending completely on RSUs. In order to find the location, the RSUs interconnect
with each other without any central location server. Due to which, communication
overhead and location query delay increase.
A static cluster is formed within the range of RSU (Nebbou, Lehsaini, Fouchal,
and Ayaida, 2019). Every vehicle in the range of an RSU updates its location to RSU.
The objective of this study is to reduce the cluster formation overhead and signalling
cost by designating RSU as location server and integrating location service and routing
functions respectively. Positions of those vehicles which are not in the range of RSU
is not updated to RSU due to which the delay and query success rate are affected.
43
On the basis of the aforementioned discussion, it is observed that static cluster-
based location service schemes consider the area division approach to make the
clusters. Vehicles in each division form a cluster. Static clustering has some
advantages such as simplicity of cluster formation. CHs are elected without any extra
signalling cost and complexity. The reference point and boundary of each cluster is
defined. Every CH is only responsible for its designated area. Regarding
disadvantages, CH election depends on specific locations such as high-density
intersections. These schemes did not ensure the proximity between location servers
and vehicles. The position of those vehicles which are outside the range of location
server is not updated. The ratio of isolated vehicles increases on the network. Location
servers are elected without completely considering mobility parameters. When these
location servers leave their respective areas, new location servers are defined. Due to
this, location server election process occurs frequently which affects the performance
of the location service. Update and query packets seem to be forwarded several times
to the location servers at level-2 and level-3. Similarly, each location server updates
the location of vehicles to a higher-level server. Location updates and location queries
are affected due to unreliable link between these location servers. Likewise, frequent
handover occurs when location server moves between different areas. Moreover, when
these servers move out of their designated areas, the hierarchies of location servers
change. The location queries related to these servers are forwarded to the next higher-
level servers which increases overhead and delay until the reformation of the new
location servers is finalized.
44
2.3.2.2 Dynamic Cluster-based Location Service Schemes
In the dynamic clustering, clusters are formed by vehicles moving in the same
direction and have similar characteristics such as speed and distance between vehicles
(Woo and Lee, 2018; Asoudeh et al., 2017; Woo and Lee, 2011). CH is elected from
vehicles having optimum mobility parameters.
45
compared to static clustering, in which lifetime of each CH and CM is limited to road
segment.
46
members, it reduces the usage of intermediate links to update the data to a higher-level
server. Every CM communicates with other vehicles in the network through its
respective CH which reduces the communication overhead and increases the
scalability.
Although, dynamic clustering is more flexible and scalable, but there is still a
need to enhance the cluster stability by improving the CH election, the CM affiliations
47
and cluster maintenance schemes. The critical analysis of the clustering schemes is
given in the sections 2.4, 2.5 and 2.6 in the context of the location service. Table 2.2
highlights the pros and cons of various location service schemes. Table 2.3 compares
the existing cluster-based location service schemes.
48
Table 2.3 Analysis of Cluster-Based Location Service Schemes
Reference LS Election Parameters Dynamicity CH Link CH
Reliability Instability
with CMs
Ashok et al. (2011) High density intersections Static Average High
Garg et al. (2014) Lowest speed and high density Static Average High
Aissaoui et al. (2014) Distance from the center Static Low High
Aissaoui et al. (2015) Distance from the center Static Low High
Aissaou et al. (2016) Distance from the center Static Low High
Asoudeh et al. (2017) Distance, speed, direction and Dynamic Low Average
neighbours
Woo and Lee (2018) No of beacon messages Dynamic Average Average
49
2.4 Review of Existing CH Election Schemes
Aissaoui et al. (2015) divide the road into fixed segments, where each segment
defines a cluster. The vehicle with the minimum distance from the center of the
segment is elected as a CH. The CMs receive multiple cluster member affiliation
requests from the neighbouring CHs due to the overlapping communication range. The
collision and delay of messages increase. Moreover, a separate CH is elected for each
segment which leads to frequent CH changes.
50
Ωi = δi − Vi (2.1)
51
CTVf = Averagespeed (CSLf)= 105 ± Ʋ (2.4)
where Cluster Speed Limit (CSLs) represents the speed limit of the slow speed cluster,
Ʋ is the speed adjustment factor whose value depends on application and varies from
1 to 10. The proposed VANET Cluster Scheme (VCS) has attempted to improve the
cluster stability by electing a separate CH for each lane. However, as the number of
clusters increases, clusters will start overlapping. As a result, the CMs will receive
messages from multiple CHs. Due to these issues, the stability of the clusters suffers
adversely.
The urban area is divided into different segments on the basis of the density
(Qureshi et al., 2018). Base Station (BS) defines the centroid of each segment. The CH
election depends on the minimum distance from the centroid, signal strength and
direction of the vehicles. Vehicles near to the centroid participate in the CH election
process. The CH stability is attempted to be improved by including a limited number
of vehicles, however, the CH Election Value (CHEV) is affected due to high density
and uncertain CHER.
The cluster formation range is based on the centroid (Pal et al., 2018). The
centroid is defined on the basis of the current position of the vehicles. The CH election
depends on the speed and distance from the centroid. The uniqueness of the centroid
is affected due to different number of vehicles in the neighbor table of each vehicle
which affects the cluster formation range.
Link reliability is another important parameter used to elect the CH. This
parameter ensures the link reliability of the elected CH with its CMs and RSU. It
depends on parameters such as LLT, Power and signal strength to determine the quality
of the link. The following section critically analyzes the existing schemes that depend
on link reliability.
52
The two-layer clustering is based on static CHs and Dynamic CHs as shown in
Figure 2.9 (Arkian et al., 2015). Each RSU is working as a static CH. The dynamic
CHs within the range of the static CHs become its CMs. The dynamic CH election is
based on suitability value that depends on Quality of Service (QoS) metric ɸv, and
𝛱 = 𝛼 . ɸ𝑣 + 𝛽 . ⋀𝑣 (2.5)
The vehicle having the highest suitability value Π is elected as a dynamic CH.
The cluster stability is improved by considering the mobility parameters and link status
with the RSU. The CHEV Π is a weighted sum of mobility and QoS parameters. The
CHEV does not confirm that a vehicle which is stable with respect to mobility
parameters also has reliable link with the RSU. The elected CH which is the most
stable vehicle in its neighbours may have poor link quality with the RSU. Likewise, a
vehicle having a high link quality and low stability may be elected as a CH. Moreover,
CH is elected without defining the CHER which affects the computation of the CHEV.
53
A Destination and Interest-aware clustering (DIAC) scheme proposed by
Ahmad et al. (2018) elects a vehicle as a CH which has highest Link Quality (LQ)
with the base station. The cluster formation is based on average relative speed
difference with other vehicles, position, direction, destination of vehicles and LQ. The
developed DIAC scheme has enhanced the cluster stability by considering the mobility
parameters and LQ with base station. However, the CH stability is compromised due
to undefined CHER and the Source Vehicle (SV) which may not be a stable vehicle to
complete the cluster formation process. Moreover, CHEV does not guarantee the link
reliability and stability with neighbours simultaneously.
To enhance the cluster stability, Alsuhli et al. (2019) has developed a Double
Head Clustering (DHC) scheme that is based on consistent neighbours, relative
position, relative speed, direction, average Signal-to-Noise Ratio (SNR) and average
Link Expiration Time. The proposed DHC scheme depends on consistent neighbours
to elect the CH. The consistent neighbours are those vehicles which are moving in the
same direction and have speed difference less than predefined ΔSmax , where the ΔSmax
≤ 21.6 km/h. The direction angle of vehicle i and j is calculated using equation 2.6
and 2.7 respectively. The angle between two vehicles is determined by the equation
2.8. The two vehicles are moving in the same direction if the angle between them is
𝜋
less than , otherwise vehicles are moving in the opposite direction.
2
𝑉𝑦𝑖 (2.6)
𝑑̂(𝑛𝑖 ) = arctan( )
𝑉𝑥𝑖
𝑉𝑦𝑗 (2.7)
𝑑̂(𝑛𝑗 ) = arctan( )
𝑉𝑥𝑗
𝜋
ɸ =𝑑̂(𝑛𝑗 ) − 𝑑̂ (𝑛𝑖 ) < (2.8)
2
where𝑑̂(𝑛𝑖 ) is the direction angle of the vehicle i and 𝑑̂ (𝑛𝑗 ) is the direction angle of
the vehicle j. ɸ is the angle between two vehicles. The eligibility value is calculated
54
on the basis of weighted sum of relative position, relative speed, average SNR and
Link Expiration average LET. The LET is calculated using the Equation 2.9.
Time
where 𝑆𝐶𝐻𝑅 is the service channel range, Δ𝑑ij the distance between vehicles i and j,
ΔSij is the speed difference between vehicles i and j and sign(.) is the sign function
whose value is 1 when node i is faster than j and -1 otherwise. The developed DHC
scheme has advantages of including important parameters in the CH election process
to enhance the CH lifetime. However, the CHEV computation is not based on
contribution of all vehicles in the range, on the contrary, it involves only those vehicles
whose speed difference is less than 21.6 km/h; in addition to that, eligibility does not
ensure that a vehicle which is stable also have high LET. Likewise, the eligibility value
does not base on the optimum values of weights. These issues affect the computation
of CHEV which degrades the performance of the CH stability in terms of low CH
lifetime.
The CH does not only communicate with the CMs, but also connects with the
BS or with other CHs for the exchange of the messages. The elected CH must have
excellent communication capability along with the minimum relative mobility with its
neighbours (Bi et al., 2020). In the developed Affinity Propagation Algorithm (APA),
the CH election depends on relative distance, relative velocity and Owned
Communication Rate (OCR). The OCR represents the maximum amount of data
transmitted through a channel in one second. The APA does not only consider the
mobility parameters but also regards the communication parameters to provide the
communication with the infrastructure. However, the OCR does not ensure a reliable
link with the RSU; a CH may have a high communication rate but very low link
lifetime with the RSU which leads to CH instability. Moreover, CH is elected without
defining the CHER.
55
The CH stability is affected while ignoring the CHER and LLT. The following
schemes do not exploit the CHER and LLT to elect the CH. The CH stability is
discussed in the following sections.
A capability Metric (M) is defined to elect the CH. The metric depends on
Relative Velocity Metric (RVM), transmission Power Loss Metric (PLM) and
direction of the vehicles to compute the value of the M using the Equation 2.10 (Huo
et al., 2016).
The angle between the directions of vehicle i and j is determined using the
Equation 2.11. The vehicle i and j are moving in the same direction if the angle
between them ɸ ≤ п/4, otherwise, vehicles are moving in the opposite directions.
⃗⃗⃗⃗⃗⃗
𝐷𝑉𝑖 . ⃗⃗⃗⃗⃗⃗
𝐷𝑉𝑗 (2.11)
ɸi ,j = 𝑎𝑟𝑐𝑐𝑜𝑠
⃗⃗⃗⃗⃗⃗𝑖 ||𝐷𝑉
|𝐷𝑉 ⃗⃗⃗⃗⃗⃗𝑗 |
where ⃗⃗⃗⃗⃗⃗
𝐷𝑉𝑖 = 𝐷𝑉𝑖𝑥 𝑥 + 𝐷𝑉𝑖𝑦 𝑦 and ⃗⃗⃗⃗⃗⃗
𝐷𝑉𝑗 = 𝐷𝑉𝑗𝑥 𝑥 + 𝐷𝑉𝑗𝑦 𝑦 are the direction vectors
of the vehicle i and vehicle j.
56
The vehicle which has the smallest capability metric is elected as a CH. The
advantage of the proposed scheme is to consider the power loss metric along with
average relative velocity and direction to enhance the cluster stability. The capability
metric computation simply depends on the sum of RVM and PLM without assigning
weights. The Equation applies that a vehicle which is most stable may have more
transmission power loss; also, vehicle may have less transmission power loss but may
not be stable with respect to its neighbours. This factor does not ensure the optimum
value of the capability metric.
Ren et al. (2017a) use vehicle’s movement direction and relative position to
elect the CH. Every vehicle advertises its location at regular intervals through beacon
messages. The vehicle which is geographically at the center of its neighbours is elected
as CH as shown in Figure 2.10. The vehicles which are directly connected to CH are
considered as CMs. The length of the cluster is less than twice of CH’s communication
range. A “Safe Distance Threshold” represented by Dt is defined to reduce CM
disconnections which should be smaller than transmission range of the CH. The
vehicle which is stable with respect to its neighbours advertises itself as a CH. A
similar approach is proposed by Haider et al. (2020). The CH election in these schemes
is based on the simple concept of centrality without any extra communication overhead
involved during CH election process. The centrality of the vehicle is computed on the
basis of neighbours list. The uniqueness of the central vehicle is affected due to
different neighbour list of each vehicle. Moreover, CH election is based on position
without considering other parameters such relative speed, density and link status. Due
to these issues, cluster instability increases.
57
Figure 2.10 CH Election Based on the Centre
Woo and Lee (2018) proposed location update and CM affiliation scheme to
enhance the performance of the location service. The CH election does not consider
mobility parameters, cluster formation range and reliability of the CH with higher-
level location servers. Due to ignoring important parameters, the CH instability
increases which degrades the performance of the location service.
58
vehicles and reputation such that w1 + w2+ w3+ w4 = 1, without mentioning specific
values of each weight. The CH election is performed without defining the cluster
formation range and reliable link with the RSU. The performance of the RWCP is
affected when it is implemented in the context of applications such as routing, security
and location service.
In the static cluster-based location service, CHER is defined by the size of road
segment. The effective diameter of each vehicle is 800m. A separate CH is elected for
each segment. As the communication range of each CH is overlapped with its adjacent
CHs due to the mobility of the vehicles, so CHs receive multiple CM requests from
these overlapped segments. As a result, the collision and delay of location updates and
location queries increase. This causes the CH to change frequently. Similarly, the rate
of change of CH increases with the increase of speed.
59
studies, a leading vehicle starts the process of CH election which is already moving
with high speed but this does not ensure a stable CH election. In some studies, a vehicle
which does not receive beacon messages for a certain time, elects itself a CH without
considering the mobility parameters.
Asoudeh et al. (2017) every vehicle calculates its CHEV and broadcasts it to
neighbouring vehicles, which compare this value with their own CHEV. If the
received CHEV is better than its own CHEV, then it further broadcasted to its
neighbouring vehicles. This process continues until a vehicle which has optimum
CHEV is elected as a CH. The messages of the vehicles whose delay is more than
certain delay are discarded which leads to unreliable and incomplete CHEV.
Moreover, the elected CH also changes frequently. These issues arise due to undefined
clustering reference point and a reference vehicle. There is a need to define a reference
point and a stable vehicle that completes the process of election of CH election by
defining the CHER.
60
Table 2.4 Analysis of the CH Schemes
Reference CH Election Dynamicity Cluster Link Clustering Main issues causing CH instability
Parameters Formation Reliability Application
Range With RSU
Arkian et al. (2015) Mobility and Static, Yes Yes General Lack of synchronization between Mobility and QoS
QoS Dynamic parameters
Singh and Bali (2015) Aggregate Dynamic Leadership No General Undefined cluster formation range, No any link status
relative speed with RSU, fixed selection threshold values
Aissaoui et al. (2015) Distance Static Yes No Location Non-optimum cluster formation range, CH election
Service without mobility parameters, NO reliable link with
the RSU
Cambruzzi et al. (2016) Speed, central Dynamic No No General Undefined CFR, No link status with the RSU, Centrality
vehicle computer without direction
Farooq et al. (2016) Speed Dynamic Yes No General Large number of clusters, cluster overlapping, No link
status with the RSU
Huo et al. (2016) Velocity, dir, Dynamic No No General Undefined weights, Lack of synchronization between
power loss speed and power parameters
Asoudeh et al. (2017) Distance, dir, Dynamic No No Location Undefined CFR, CHEV without specific weights, no
speed, Service link status with the RSU
neighbor
61
Reference CH Election Dynamicity Cluster Link Clustering Main issues causing CH instability
Parameters Formation Reliability Application
Range With RSU
Qureshi et al. (2018) Distance, Static Yes No General Non-optimum CFR, CH election ignored mobility
signal parameters and link with the RSU
strength, dir
Pal et al. (2018) Speed, Dynamic Yes NO General Centroid centrality issue, no reliable link with the RSU
distance
Ahmad et al. (2018) Speed, LQ Dynamic No Yes General Undefined cluster formation range, unstable SV elects
position, dir, CH
destination
Joshua et al. (2019) Velocity, Dynamic No NO General Undefined CHEV weights, no reliable link with the
position, dir, RSU
neighbours,
reputation
Alsuhli et al. (2019) Speed, Dynamic Consistent Yes General CHEV without weights and CFR without including all
position, neighbours vehicles
direction,
Bi et al. (2020) SNR, CN, LET
Velocity, Dynamic No Yes, OCR General Undefined CFR, Lack of synchronization between
direction, OCR mobility and OCR
62
2.5 Review of Existing Cluster Member Affiliation Schemes
The main objective of location service is to store the location of the vehicles in
VANETs. In the cluster-based location service, the CH stores the location of its CMs.
If the CM switches between different CHs or some of the vehicles could not affiliate to
any CH, then locations of the CMs are not updated to the location servers. In order to
improve the performance of the location service, the CM instability should be
minimized. This section analyzes the existing CM affiliation schemes and highlights the
existing issues.
Link lifeTime (LLT) defines the reliability of link between vehicles. It depends
on communication range, speed, distance and time to determine the quality of the link.
High value of the LLT represents a reliable link between vehicles. The following
section reviews the existing LLT-based CM affiliation schemes.
The CM affiliation depends on Link LifeTime (LLT) with the CH, link delay
and number of loss messages (Cambruzzi et al., 2016). The CM Affiliation Value
(CMAV) is calculated using the Equation 2.12.
where S is the LLT with the CH, F represents link delay and E is the number of loss
messages. The value of the LLT is computed using the Equations 2.13, 2.14 and 2.15.
𝑟𝑖 − 𝑑(𝑖, 𝑗) (2.13)
S= , 𝑠𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜
𝑣𝑟
63
| 𝑣𝑟 = |𝑣𝑖 − 𝑣𝑗 | 𝑖𝑓 |𝑣𝑖 − 𝑣𝑗 | ≥ 1 (2.14)
where S is LLT between the vehicle and the CH, ri is the communication range,
d (i, j) is the distance between the vehicle and the CH, vr is the relative speed difference
between the vehicle and the CH. The CM stability is improved by considering LLT
parameter. However, CMAV depends on the instantaneous values of the speed, also
CMAV relies on the fixed threshold values without considering the variables involved
such as location update intervals.
Every vehicle calculates its capability metric M that depends on relative speed,
power and direction, and sends CM requests to the existing CH (Huo et al., 2016). If
the capability metric M of new joining CM is less than the value of the existing CH,
then the CM joins the cluster as a CH. The CH election process is triggered on the
joining of new CM. Isolated vehicles are reduced due to CM affiliation scheme.
However, the existing CH is not stable which results in frequent switching of CMs
between different CHs. Moreover, when there are more than one CHs, CM joins the
new cluster on the basis of LLT. The CM instability increases due to instantaneous
values of the speed while calculating the LLT.
64
direction of the CH to affiliate with the cluster. The CM lifetime of these schemes is
affected due to the dependence on the instantaneous values of the distance and speed.
The vehicle that are currently inside the segment are considered as CMs
(Aissaoui et al., 2015). The CM affiliation depends on the distance of each vehicle from
the CH. The advantage of the proposed scheme is that the vehicles are affiliated with
the CH without any extra exchange of the messages. However, the CM affiliation
ignores the mobility parameters, also vehicles are getting multiple CM requests from
neighbouring CHs which increases the CM instability.
The cluster joining relies on the speed difference with the CH (Arkian et al.,
2015). When a CH receives a CM affiliation request, it compares the speed difference
with a predefined threshold value. If the speed of the vehicle satisfies the Equation 2.16,
the CH will accept the request.
where Vi is speed of the vehicle i and 𝑉𝑐ℎ is the speed of the CH and 𝛥𝑆𝑡ℎ𝑟𝑠ℎ is the
CM affiliation threshold value. The vehicles depend on the instantaneous value of the
speed of CH which does not ensure the stable affiliation with the cluster. Along with
this, the isolated vehicles also increase due to fixed threshold values.
The CM affiliation depends on the speed difference with the CH. Vehicles
calculate their speed difference based onto the instantaneous values of the speed of the
CH (Asoudeh et al., 2017). The CM affiliation threshold value is defined as 18 km/h
with the maximum limit of the CMs as 20. The lifetime of a CM in a cluster decreases
along with the increase in the number of isolated vehicles in the network which is the
main cause of the CM instability.
65
direction as a CH and if its speed difference with the CH lies within the range ∓Vth
also as given in the Equation 2.17.
|𝑉𝑗 − 𝑉𝑐ℎ | ≤ ±𝑉𝑡ℎ (2.17)
where |𝑉𝑗 − 𝑉𝑐ℎ | is the speed difference and ±𝑉𝑡ℎ is the CM affiliation threshold
value. The performance of the proposed scheme is improved by reducing the isolated
vehicles due to flexibility of CM affiliation threshold value, however, CM instability
is affected due to reliance on the instantaneous values. Moreover, CM affiliation
criteria is based without the reliable link with the CH.
Haider et al. (2020), the cluster joining depends on the direction and distance
from the CH. When there are more than one CHs, the vehicle joins that CH which is
moving in the same direction and has less distance. If there is no neighbouring CH, the
vehicle elect itself as a CH. The proposed scheme has reduced the isolated vehicles by
simplifying the CM affiliation which only depends on distance and direction. As the,
the CM affiliation depends on the instantaneous speed of the CH and also ignores the
link status with the CH, as a result, the lifetime of the CM in a cluster reduces, which
reduce the cluster stability.
where parameter may be distance, speed, LLT or any other similar value and parameter
limit is the CM affiliation threshold value to affiliate the vehicles with the cluster. In
the simplest CM affiliations, vehicles depend on the distance between the vehicle and
the CH. The distance parameter is not a reliable parameter to affiliate the vehicles due
to mobility characteristics of the vehicles as the vehicles move frequently and
suddenly. Additionally, the communication range of each vehicle is 400m. When the
66
distance between CH and vehicles is greater than 400, the vehicles cannot update their
location information to CH.
All these three types of parameters depend on the instantaneous speed of the
vehicles and CH to formulate the CM affiliation scheme. Due to high dynamic
characteristics of VANETs, these instantaneous values do not ensure a reliable and
stable CM affiliation. Vehicles join and leave the cluster frequently due to which, the
lifetime of each CM in a cluster reduces. Similarly, the location updates and location
queries are suffered due to high switching of CMs in different clusters.
67
Table 2.5 Analysis of the CM Affiliation Schemes
Reference CM Affiliation Link CM Clustering Main issues causing CM instability
Parameters Reliability Affiliation Application
with CH Threshold
Aissaoui et al. (2015) Distance No Values
Not fixed Location CM switches between different CHs, No reliable link with the
service CH
Arkian et al. (2015) Speed No Fixed General CM affiliation depends on the instantaneous values, fixed
threshold value, no reliable link with the CH, isolated vehicles
Cambruzzi et al. (2016) Distance, speed, Yes Fixed General The link reliability is based on instantaneous values of the CH,
delay, messages Also threshold values based on fixed values, isolated vehicles
Huo et al. (2016) loss
Speed, direction, No Not fixed General CM lifetime in a cluster reduces due to instantaneous speeds
power and frequent switching,
Asoudeh et al. (2017) Speed No Fixed Location CM instability due to frequent switching between CHs and
Mehmood et al. (2017) Speed No Fixed service
General isolated
CM vehicles.
lifetime is reduced due instantaneous speed, unreliable link
clustering with the CH, isolated vehicles increases
Woo and Lee (2018) Speed, distance, Yes Fixed Location Due to non-optimum values of speed, fixed values, CM change
time service CHs frequently,
Haider et al. (2020) Distance, direction NO Fixed General Unreliable link with the CH, fixed threshold values,
clustering dependence on instantaneous speed of the CH, the isolated
vehicles increases.
68
2.6 Review of Existing Cluster Maintenance Schemes
The stability of the cluster is affected due to the occurrence of different events
such as CM losing its affiliation with the CH, the decreasing stability of the CH,
existence of a number of clusters within communication range and disbanded cluster.
The cluster maintenance retains the cluster stability by developing algorithms
accordingly. The following section critically analyzes the existing cluster maintenance
schemes in terms of clustering merging and cluster splitting.
Arkian et al. (2015) depend on range, relative speed and number of beacon
message intervals to merge the cluster. Cluster stability is improved by including
important parameters, however, the merging stability can be further enhanced by
defining the specific number of interval in Merging Threshold Time (MTT) and
electing a new CH based on the reliable link with the RSU. Due to unpredictable
69
vehicular mobility, the merging CHs may approach each other or may move away
during the MTT, subsequently, the existing MTT does not ensure a stable merging.
Cluster merging starts when all the members of one cluster are in the range of
another CH and total number of CMs are less than predefined threshold value
(Cambruzzi et al., 2016). The merging is initiated when distance between CMs and
CH of other cluster is less than the communication range. However, cluster merging is
not stable due to undefined MTT and CH of the merged cluster.
Lin et al. (2016) merging depends on the distance between two overlapping
clusters and the duration in which two clusters remain in the range of each other. When
the distance between two moving clusters is half of the communication range and MTT
is equal to the interval of two messages, the merging starts. The vehicle which is at the
middle of the two merged clusters is elected as a new CH. The merging stability is
improved by considering merging conditions, however, during MTT merging clusters
may move away from each other which causes instantaneous merging. Likewise, CH
is elected without considering the link reliability with the RSU.
If two clusters remain in the range of each other for the duration of 10 seconds,
the merging process starts (Ren et al., 2017a). The MTT is 10 seconds to avoid the
temporary merging, however, this value is fixed and ignored the involving variables
such as beacon messages intervals. In addition, the new CH is elected without
considering the link reliability with the RSU.
70
The MoGLS scheme depends on the distance between two CHs, sojourn time
and MTT. If the sojourn time of the original cluster is greater than or equal to the target
cluster and distance between two CHs is less than 100m up to the duration of the
location update interval, then the cluster merging starts. Although existing study
includes the sojourn time to enhance the cluster stability, however, limitations are still
there due to fixed CM threshold value, ignoring the link reliability while electing the
CH (Woo & Lee, 2018).
All CMs interest based merging depends on the range, interest, number of CMs
after merging and the eligibility of the merged CH (Alsuhli et al., 2019). The two
moving clusters merged subject to the conditions; all the members of the existing cluster
are consistent with the target CH, the total number of CMs after the merging are less
than the maximum limit and the target CH is more stable than the current CH. However,
merging ignores the MTT and link status with the RSU while merging the clusters which
increases the cluster instability.
Haider et al. (2020) depend on the direction and overlapping range of two
clusters. The clusters which are moving in the same direction and are in the range of
each other start the merging without considering MTT. The merging of two clusters is
not reliable in the absence of MTT. Furthermore, the CH is elected without considering
the link reliability with the RSU. The analysis of the cluster merging schemes is
presented in the Table 2.6.
71
Table 2.6 Analysis of the Cluster Merging Schemes
Reference Overlapping Merging New CH Main issues causing cluster maintenance instability
Range Threshold of the
Time(MTT) Merged
Cluster
Arkian et al. (2015) Yes Yes Yes Undefined beacon messages interval, unreliable link with the
RSU
Cambruzzi et al. (2016) Yes Not Not Cluster instability increases due to undefined MTT and CH of
the merged cluster
Huo et al. (2016) Not Not Yes Overlooked overlapping range and MTT and CH is elected
without reliable link with the RSU.
Lin et al. (2016) Yes Yes Yes Non-optimum MTT, the median vehicle is elected as a new CH
without the link status with the RSU and mobility parameters
Ren et al. (2017a) Yes Yes Yes Fixed MTT and unreliable link with the RSU
Woo and Lee (2018) Yes Yes Yes Non-optimum MTT, CH election without considering link
status with the RSU
Alsuhli et al. (2019) Yes Not Yes Without MTT and overlapping distance, unreliable link with
the RSU
Haider et al. (2020) Yes Not Yes Without MTT, CH is elected without the link status with the
RSU
72
2.6.2 Cluster Splitting Algorithms
The main objective of the cluster splitting is to divide the existing cluster into
two or more clusters to maintain the cluster stability. The cluster stability is
compromised due to increased size of the cluster beyond the maximum limit, and
increased density within the defined range (Huo et al., 2016). In addition, due to
different speeds, a vehicle or a group of vehicles goes out of the range of their respective
CHs. In such a scenario, the cluster splits up into more than one clusters (Haider et al.,
2020). Moreover, around the intersection, existing CH moves only in one direction. The
distance between CH and its CMs moving in the other directions increases beyond the
default range. The existing CH initiates the process of the election of the new CHs based
on the predefined cluster splitting algorithm. After splitting, the new CHs send CM
requests to affiliate the vehicles to the clusters. Through the splitting, the average size
of the cluster and issues such as collision and messages loss are reduced.
Zhao et al. (2016) an intersection based clustering algorithm splits the vehicles
into different directions around the intersection as shown in Figure 2.11. The CH
election is based on two algorithms. The CH election in the straight road depends on
speed and distance parameters, while around the intersection direction parameter is used
to manage the vehicles into clusters.
73
Figure 2.11 Cluster Splitting Around the Intersection
Zhou et al. (2017) reduced the clustering around the intersection by electing a
CH. In the proposed Novel Vehicle Clustering (NVC) algorithm, each CH sends its
CMs information to the Base Station (BS) which is acting as a local server as shown
in Figure 2.12. The BS elects the candidate CHs from each road section based on the
74
highest number of vehicles. Then a vehicle which has the lowest weight is elected as
a CH. The weight is based on the relative position between vehicles and the relative
distance from the vehicle to the BS. The CH election is based on number of neighbours
and distance, without considering the speed. Only one CH has to manage the whole
intersection traffic. Traffic is moving in different directions but the elected CH can
move only in one direction. When the distance between CH and vehicles of other road
sections increases beyond the communication range, the vehicles will lose their
affiliations; as a result, their locations are not updated to the CH. Due to this, the query
success rate decreases.
Mehmood et al. (2017) divides the traffic into left, right and straight lanes.
Cumulative weight is calculated using density, connectivity level, and distance and
speed parameters. The core parameter is connectivity level which is number of vehicles
in a specific lane. The CH is elected from the lane which has large number of vehicles
as shown in Figure 2.13. The existing study tries to elect the CH on the basis of number
of vehicles in a lane to affiliate more vehicles, to reduce CM instability around the
intersection. However, existing CH can move only in one direction, leaving the vehicles
of other directions without any CH, due to these issues the cluster instability increases.
75
Figure 2.13 Lane Based CH Election
Woo and Lee (2018), the CM affiliation scheme exploits the intersection to
calculate the sojourn time. Dynamic clusters are formed on the basis of similar
trajectory. To join a cluster, every vehicle calculates its sojourn time P that depends
on the position of the vehicle at the approaching intersection. However, existing study
ignores the cluster splitting issue around the intersection which causes the cluster
instability around the intersection.
The destination based cluster formation Ali et al. (2019), depends on the speed,
distance, direction and interest to elect a CH. After receiving the CM affiliation
requests from the vehicle, the CH confirms the destination and direction of the vehicle
before accepting the request. The cluster stability is improved by including the
destination parameters, however, CH election is overlooked around the intersection.
76
2.6.3 Remarks and Observations
The stable cluster merging depends on the optimum merging conditions such
as overlapping range, Merging Threshold Time (MTT), the duration of the overlapping
range before the start of the merging, and election of the CH for the merged cluster.
Table 2.7 critically analyzes the existing cluster splitting algorithms. The
existing studies split the cluster to maintain the cluster stability around the intersection.
However, limitation still exists due to unnecessary clustering after vehicles enter the
intersection area and reclustering after vehicles leave the interstation area. Even, the
existing stable cluster is reclustered, similarly, in the existing studies, the link
reliability with the RSU is ignored, while electing the new CH. Due to these issues,
cluster instability around the intersection increases. Therefore, an optimized cluster
merging is needed to be designed to enhance the cluster stability.
77
Table 2.7 Analysis of the Cluster Splitting Schemes
Scheme CH Election: CH Election: Main issues causing cluster maintenance instability
Intersection Area Straight Road
Huo et al. (2016) Direction, stay time Speed, power No Cluster splitting, CH instability around the intersection
Zhao et al. (2016) Direction Distance, speed Cluster instability due to twice clustering around the
intersection, CH election ignores the link status with the RSU.
Even existing stable cluster is dispersed
Zhou et al. (2017) Distance, No CH election CH election without mobility parameters and link with the
neighbours RSU, cluster instability increases due to only one CH for the
whole intersection. Existing stable clusters are destabilized
Mehmood et al. (2017) Without CH Density, CL, NO cluster splitting around the intersection
election distance speed
Woo and Lee (2018) Beacon messages Without CH No cluster splitting around the intersection
election
Ali et al. (2019) Without CH Distance, speed No cluster splitting around the intersection
election
78
2.7 Findings of the Literature Review
79
broadcast approach and undefined range. The CH lifetime can be improved by defining
the cluster formation range and increasing the link reliability with the RSU.
At the same time, the CM affiliation threshold values are fixed without
considering the metric involved such as CM to CH and CH to RSU location update
intervals. Due to fixed threshold value, some of the vehicles do not affiliate to any
cluster. As a result, the isolated vehicles in the network increases. Due to the unstable
CM affiliation, the CM change the clusters frequently and number of cluster changes
per vehicle also increases. Therefore, a dynamic cumulative cluster member affiliation
scheme should be designed to reduce the CM instability.
80
cluster based location service, the location updates and location queries are affected
due to the instability of the CM.
Furthermore, the cluster splitting divides the existing clusters into two or more
clusters to maintain the cluster stability. Around the intersection area, the existing
cluster disperses due to the movement of the vehicles into different directions. The
existing studies split the cluster to maintain the cluster stability around the intersection.
However, limitation still exists due to unnecessary clustering after vehicles enter the
intersection area and reclustering after vehicles leave the interstation area. Even, the
existing stable cluster is reclustered. Similarly, in the existing studies, the link
reliability with the RSU is ignored while electing the new CH. Due to these issues,
cluster instability around the intersection increases. Therefore, an optimized cluster
merging needs to be designed to enhance the cluster stability.
81
2.8 Summary
82
RESEARCH METHODOLOGY
3.1 Introduction
This research work is framed into three major stages as shown in Figure 3.1.
The first stage comprises of thorough review and analysis of the available literature
about location service, cluster-based location service, clustering in VANETs, Cluster
Head (CH) election, Cluster Member (CM) affiliation, and cluster maintenance
schemes in VANETs. Afterwards, development stage includes three phases. In the first
phase, a Reliable Cluster Head Election (RCHE) scheme is developed to improve the
CH stability. The output of this scheme is elected CH.
83
Figure 3.1 Operational Framework
84
3.3 Research Design and Procedure
This section presents the concepts to design and develop RCHE, DCCMA and
OCM schemes. Figure 3.2 illustrates the interaction among the three phases which
jointly contribute to the development of the proposed ESCLS mechanism. Table 3.1
shows the overall research plan for each phase.
The cluster-based location service mainly depends on the stability of the CH.
Existing studies try to improve the performance of location service by improving the
CH election criteria, but limitations still exist due to the non-optimal cluster formation
range and unreliable link of CH with the RSU. In order to improve the lifetime of a
CH, appropriate mobility parameters including distance, speed, RLT, direction, and
number of neighbours are considered. The proposed RCHE scheme consists of three
phases such as, Centroid Vehicle (CV) election, formation of Cluster Head Election
Range (CHER) and election of the CH. In the CV election phase, each vehicle
calculates its distance from the centroid. The vehicle nearest to the centroid is elected
as a CV. In the second phase, all vehicles whose distance from the CV is less than 400
defines the CHER. Furthermore, a CH is elected from the vehicles in the CHEV range
based on the Variability Index (VI), RSU Link Time (RLT) with the RSU and number
of neighbours. The Cluster Head Election Value (CHEV) is the weighted sum of the
VI, RLT and number of neighbours. The vehicle which has the largest CHEV is elected
as a CH. The overall procedure and results analysis of the RCHE scheme is described
in detail in chapter 4. Figure 3.3 illustrates the block diagram of the design and
implementation of the RCHE scheme.
85
Cluster head election value
IV Isolated vehicle
CHLT Cluster Head Link Lifetime
RLT RSU
Link Lifetime
86
Figure 3.3 Design, Implementation and Evaluation of RCHE Scheme
87
Table 3.1 Overall Research Plan
Proposal Research Objectives Research Research Methodology Performance Result
Phase Matter Questions Metrics Validation
To develop a reliable CH CH How to improve the (a) Elect a Centroid Vehicle. CH lifetime,
election scheme that reduces election process of electing (b) Define the Cluster Head Election Query success
Phase 1 the frequent CH changes and a CH in order to Range rate, HCBLS,
enhances the link reliability enhance the CH (c) Every vehicle compute its CHEV Query response CBLS
with the RSU to improve the stability? based on VI, RLT and Nnbr delay
CH stability. (d) A vehicle whose CHEV is largest is
elected as a CH
To enhance a CM affiliation CM How to enhance the a) IVs sense CM requests from the CH CM lifetime,
scheme that reduces the affiliation process of b) IVs compute their CHLT for each CH Query success
frequent CM changes and affiliating a vehicle c) CH affiliates the vehicles as a CM rate, Fraction of CBLS,
Phase 2 isolated vehicles to enhance in order to improve based on the comparison of the CHLT location MoGLS
the CM stability. the CM stability? with CM affiliation threshold values information
saved in the
location servers
To optimize a cluster Cluster How to enhance the a) Merging initiates subject to the Query success
maintenance scheme to maintenance process and fulfilment of the merging conditions (b) rate,
Phase 3 stabilize the cluster merging conditions of New CHs are elected based on RLT Query response HCBLS
and cluster splitting to cluster maintenance (c) Specifying the intersection area. (b) delay, CBLS,
improve the cluster stability. to improve cluster New CHs are elected based on highest localization error MoGLS
stability? RLT and destination. (c) CM affiliates
88
other vehicles through beacon messages. Vehicles sense the CM requests from the
neighbouring CH and start the CM affiliation process. Figure 3.4 elaborates the block
diagram of the design and implementation of the DCCMA scheme. Every vehicle
calculates its CH Link LifeTime (CHLT) that is based on Cumulative Moving Average
Speed (CMAS) instead of the instantaneous speed of the vehicles. Vehicle sends that
value to CH. The CH compares RLT value with the dynamic threshold value that
depends on location updates intervals from vehicle to CH and from CH to RSU. If the
value of the CHLT is greater than threshold value, CH affiliates the vehicle as a CM
and updates its location information to nearest RSU.
89
If there are more than one CHs in neighbouring area, then the vehicle sends the CM
request to CH which has higher CHLT value. Moreover, if there is no neighbouring
CH or CHLT is less than the CM affiliation threshold value, then vehicle sends its
location information directly to the RSU. The overall procedure and result analysis of
the DCCMA scheme is described in detail in chapter 5.
The OCM scheme is based on cluster merging and cluster splitting. Due to the
high dynamicity of the VANETs, there is always a possibility that more than one
clusters co-exist in neighbouring area. Existing studies attempt to increase the cluster
stability through enhancing the cluster merging algorithm, however, issues still exist
due to non-optimal merging conditions. The performance of the location service is
improved through optimizing cluster merging. In the proposed cluster merging
algorithm, when there are more than one clusters in the range of each other, the CH
with the highest id invokes the merging process. CH computes and verifies the merging
condition such as overlapping range, Merging Threshold Time (MTT) and distance
between CHs at the start and end of MTT. When merging conditions are satisfied, then
CH sends the merging accepted messages. After merging the clusters, a new CH is
elected on the basis of RSU Link Time (RLT). The CH which has highest RLT is
elected CH for the merged cluster. The detailed process of the cluster merging is
explained in chapter 6.
90
relative destination of each vehicle which may be left, right or straight. After entering
the intersection area, the existing CH advertise its direction and destination. Then
existing CH elects the CHs for other two remaining directions. CH compares the RLT
of CMs in a specific direction (left, right or straight) on the basis of destination. The
CM which has a highest RLT in a particular direction is elected as a CH for that
direction. The same process is repeated to elect the CH for the remaining directions.
The overall procedure and result analysis of the OCM scheme is described in detail in
chapter 6. Figure 3.5 shows the block diagram of the design and implementation of the
OCM scheme.
91
3.4 Simulation Framework
The final phase is the performance evaluation of the proposed work. The
performance of the RCHE, DCCMA and OCM schemes are simulated, tested and
evaluated against existing cluster-based location service schemes. Two types of
simulators, namely network simulator and traffic simulator are used to simulate the
vehicles in VANETs. Network simulators are used to simulate network protocols and
traffic simulators are used to emulate the road traffic behaviour. There are a number of
simulation frameworks have been developed such as Network simulator 2 (NS2),
Objective Modular Network Testbed in C++ (OMNET++), Optimized Network
Engineering Tool (OPNET), Matrix Laboratory (Matlab). However, OMNeT++ is
preferred as it provides extensible, modular, component-based C++ simulation platform
for developing and executing network simulations.
92
Therefore, the proposed ESCLS mechanism is simulated using network
simulator OMNET++ version 5.3 and is integrated with a realistic scenario which is
generated by using traffic simulator SUMO version 0.32.0. VEINS is used as a
middleware framework to connect OMNET++ and SUMO as shown in Figure 3.6.
93
messages for all vehicles are set to 10Hz. The simulation period of 1000 seconds
provides the reliable results. To justify the density of vehicles, one lane in each
direction is used to simulate vehicles. In the proposed ESCLS mechanism, physical
and MAC layers are configured according to the IEEE 802.11p standard. The detailed
simulation parameters are given in the Table 3.2. Simulations are performed by taking
different maximum speed ranges from 10 m/s to 30 m/s (36 km/h – 108 km/h). The
density range of the vehicles is 150-950.
94
average density (13-16 vehicle/km/lane is approximately equals to 620 vehicles in
6km x 6km area). In the second scenario, maximum density of the vehicles varies from
150 to 950 while keeping the average speed 40 km/h. Table 3.4 exhibits the Physical,
MAC and other related parameters which are used to set up the simulation environment
(Woo and Lee, 2018; Asoudeh et al., 2017; Aissaoui et al., 2015).
Parameter Value
Number of RSUs 20
MAC 802.11p
95
CBLS and HCBLS as proposed by ( Woo and Lee, 2018; Asoudeh et al., 2017;
Aissaoui et al., 2015) respectively. The performance of the ESCLS mechanism is
evaluated in three stages based on three proposed schemes RCHE, DCCMA and OCM.
In the first stage, the performance of proposed RCHE scheme is compared with
the HCBLS and CBLS schemes. The proposed RCHE scheme is compared with
existing CH schemes of HCBLS and CBLS which represents that comparison is only
performed between proposed CH scheme and existing CH schemes of HCBLS and
CBLS. The impact of CH lifetime is analyzed in terms of query success rate and query
response delay.
The basic requirement for the second stage is the elected CH. After the election
of one vehicle as a CH, other vehicles start the CM affiliation process. The proposed
DCCMA is compared with existing CM schemes of CBLS and MoGLS. In the first
step, CH is elected for each scheme, then CM affiliation is performed. The
performance of each scheme is measured by CM lifetime. The impact of CM lifetime
is analyzed in terms of query success rate and fraction of location information saved
in RSUs.
The third cluster maintenance stage depends on the first two stages, in which
CH is elected for each scheme and vehicles are affiliated as CMs. The basic
requirement for cluster maintenance is the existence of clusters with CH and CMs. To
compare the proposed OCM scheme with the existing HCBLS, CBLS and MoGLS
schemes. The CH elections and CM affiliations schemes are configured for each
scheme, then results are obtained after configuring OCM schemes. The impact of
cluster maintenance, which also depends on CH lifetime and CM lifetime, is evaluated
in terms query success rate, query response delay and localization error.
96
scheme is compared with only CH election schemes of HCBLS, CBLS and MoGLS.
Same process is repeated for CM affiliation and cluster maintenance.
Once, the CH is elected The selected evaluation metrics for ESCLS mechanism
are used by existing studies to evaluate the cluster stability and the performance of the
location service. The comparison is based on the following metrics: CH Lifetime, CM
Lifetime (Zhang et al., 2018), query success rate, query response delay, fraction of
location information saved in the location servers, and localization error (Aissaoui et
al., 2015). The performance of the proposed ESCLS mechanism is evaluated by
performing simulations at different values of speed and density. The description of the
metrics used is given in the following section.
The stability of the cluster directly depends on the lifetime of the CH. The
larger value of the CH lifetime represents that cluster configuration is more stable.
The CH lifetime is the ratio of the total CH durations time to the number of cluster
heads. The CH lifetime is the average lifetime of CHs for a given scenario as defined
in Equation 3.1 (Zhang et al., 2018).
where n is the number of cluster head formed, (𝑡𝑖𝑏 ) is the time of cluster head election
and (𝑡𝑖𝑚 ) is the time of cluster headship losing.
97
(b) CM Lifetime (CML)
The CM lifetime is the difference between the time when the vehicle joined the
cluster and the time when the vehicle left. The CM lifetime is the average of the
lifetime of the CMs as formulated in Equation 3.2 (Zhang et al., 2018).
∑𝑛𝑖=1(𝑡𝑖𝐿 − 𝑡𝑖𝐽 )
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝐶𝑀 𝑙𝑖𝑓𝑒𝑡𝑖𝑚𝑒 = (3.2)
𝑛
where n is the number of clusters members, (𝑡𝑖𝐽 ) is the time of clustering joining and
(𝑡𝑖𝐿 ) is the time of cluster leaving.
It is the ratio of number of the queries that have been responded to the number
of queries that have been sent from the source as given in Equation 3.3 (Asoudeh et
al., 2017; Aissaoui et al., 2015).
where 𝑄𝑢𝑒𝑟𝑦𝑟𝑐𝑣𝑑 is the received query and 𝑄𝑢𝑒𝑟𝑦𝑠𝑒𝑛𝑡 is the sent query.
The query response delay is the average of the difference between the time
when a vehicle sends a query and the time vehicle receives the respond. It is the ratio
of the total elapsed time of queries to the total number of queries. It is the average of
query response delay of n vehicles as defined in Equation 3.4 (Asoudeh et al., 2017;
Aissaoui et al., 2015).
98
∑𝑛𝑖=1 𝑄𝑢𝑒𝑟𝑦𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒𝑡𝑖𝑚𝑒(𝑖) (3.4)
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑄𝑢𝑒𝑟𝑦 𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝐷𝑒𝑙𝑎𝑦 =
∑𝑛𝑖=1 𝑄𝑢𝑒𝑟𝑦(𝑖)
where Query Responsetime(i) is the query response time of query of vehicle i and
Query(i) is the query of the vehicle i.
It is the ratio of number of the vehicles whose position is saved in the location
servers to the number of vehicles in the network as formulated in Equation 3.5
(Aissaoui et al., 2015).
99
where 𝐿𝑂𝐶𝑐𝑢𝑟(𝑖) is the current position of the vehicle i and 𝐿𝑂𝐶𝑠𝑣𝑑(𝑖) is the saved
position of the vehicel i in the location server.
This research has the following assumption and limitations which are
considered in the network scenario.
3.8 Summary
This chapter presented the research methodology and framework used in the
development of the proposed ESCLS mechanism. The operational framework
explained the proposed phases of the ESCLS such as RCHE, DCCMA and OCM.
Table 3.1 summarized the overall research plan. In addition, this chapter elaborated
the simulation environment and parameters as well as the performance evaluation
metrics that are used to evaluate the proposed schemes. The various assumptions and
limitations were also considered in the simulation setup of this research.
100
RELIABLE CLUSTER HEAD ELECTION SCHEME
4.1 Introduction
This chapter presents a Reliable Cluster Head Election (RCHE) scheme that
addresses the existing Cluster Head (CH) election issues. The main objective of the
RCHE scheme is to enhance the performance of the location service by improving the
CH stability. In the existing cluster-based location schemes, CH election depends on
static clustering or dynamic clustering. In static clustering, road is divided into
segments and a separate CH is elected for each segment. The lifetime of CH is limited
to the segment. Dynamic clustering reduces the frequent CH while considering
mobility parameters during CH election process, but issues still exist due to non-
optimum cluster formation range. Moreover, every CH updates the location of CMs to
RSU. The CH is elected without considering the link reliability with the RSU which
affects the location updates and location queries of vehicles. Due to aforementioned
issues, cluster instability increases which degrades the performance of the location
service. In the proposed RCHE scheme, the stability of the CH is improved by
optimizing CH election range and enhancing link lifetime with the RSU. Section 4.2
describes the design of the proposed RCHE scheme. Afterward, sections 4.3 and
section 4.4 describe the simulation setup, results and analysis of the RCHE scheme
respectively. Finally, the summary of this chapter is presented in section 4.5.
101
stability of the CH. Location service schemes based on static clustering approach
divide the area into various segments and the collection of vehicles in each segment
defines a cluster. A CH is elected based on the position of a vehicle in segment. All
other vehicles in segment join the cluster as CMs. The static cluster-based schemes
suffer from the frequent CH changes and also clusters also overlap each other due to
cluster formation range. Dynamic clustering is more flexible because a CH moves
with its CMs. The election of the CH is based on the values of different metrics such
as distance, speed, direction and number of neighbours. In the existing dynamic
cluster-based location service scheme, each vehicle calculates its CHEV and
broadcasts to its neighbour vehicles. This process continues until a vehicle that has
optimum value is elected as a CH. Messages with delay greater than a threshold value
are discarded. The computation of the CHEV is affected due to undefined cluster
formation range. The computed CHEV elects unstable vehicle as a CH. Moreover,
existing cluster-based location service schemes elect CH without considering the link
reliability with the RSU. Some of the existing studies in the general clustering exploit
link status to elect CH. However, these schemes do not ensure stability of the CH with
the RSU and CMs simultaneously. Due to these issues, CH instability increases which
decreases the lifetime of the CH. The location updates and location queries are also
affected. Therefore, RCHE scheme is incorporated in the clustering that enhances the
reliability of the CH with the RSU. Moreover, the reliability of the CHEV also
increases by defining the CH election range. The design of the proposed approach is
presented in Section 4.3. As discussed in section 2.4, Table 4.1 provides a comparison
between existing CH schemes and the proposed RCHE scheme.
102
4.3 Design of the RCHE Scheme
The CV election phase starts with the identification of centroid. The road is
divided into segments of 800m. The centroid of each segment is already identified.
Each vehicle gets the information about the centroid of each segment through a digital
map. Vehicles share their information such as location, speed and direction through
Beacon Message the exchange of BMs. Moreover, the existing CH advertises its status through BMs.
(Lines 1-5, Algorithm 4.1). If vehicles get CH advertisement messages, then vehicles
advertise their location to existing CH, otherwise vehicles start the process of new CH
election and elect a Centroid Vehicle (CV). A vehicle whose position from the
centroid is minimum is elected as a CV and other vehicles continue to share its
information. Every vehicle calculates its distance from the centroid using the Equation
4.1 and shares it with other vehicles in its range (Lines 5-16, Algorithm 4.1).
103
Figure 4.1 Flow Chart of the Proposed RCHE Scheme
104
𝐷 = √(𝑉𝑏𝑥 − 𝐶𝑥 ) 2 + (𝑉𝑏𝑦 − 𝐶𝑦 ) 2 ) ∀ Vb ɛ [1, Z] (4.1)
where D is the distance between vehicle Vb and centroid, (Vbx, Vby) are the x and y
coordinates of the position of the vehicle Vb, (Cx, Cy) are the x and y coordinates of the
position of the centroid. The vehicle whose distance from the centroid is minimum
advertises itself as CV and starts the process of CH election as shown in Figure 4.2.
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐷𝑉𝐶𝑉 .𝐷𝑉 ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
Vd (4.2)
𝜙(CV ,Vd) = 𝑎𝑟𝑐𝑐𝑜𝑠 ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
∀ Vd ɛ [1, Z]
|𝐷𝑉 𝐶𝑉 ||𝐷𝑉 Vd |
where ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝑫𝑽𝑪𝑽 = 𝑫𝑽𝑪𝑽 𝒙 ⃗ + 𝑫𝑽𝑪𝑽 𝒚⃗ and ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ ⃗ + 𝑫𝑽𝑽𝒅 𝒚
𝑫𝑽𝑽𝒅 = 𝑫𝑽𝑽𝒅 𝒙 ⃗ are the
direction vectors of the CV and vehicle Vd.
105
Figure 4.3 CH Election Range
The most common approach used for CH election is based on calculating the
CHEV that is combination of mobility and proximity parameters. Every vehicle
calculates its CHEV that quantifies its fitness to work as a CH. In the proposed RCHE
scheme, the CHEV is a weighted sum of VI, RLT and number of neighbours. The
relative importance of each metric is determined by adjusting the value of the weight
coefficients.
106
𝑦−1 (4.3)
𝑉𝐼𝑑 = ∑𝑗=0 (𝑣𝑑 − 𝑣𝑗 ) ∀ j € [1, Y-1]
1
𝑉𝐼𝑑 = ∑𝑦−1
𝑗=0 (𝑣𝑑 − 𝑣𝑗 ) ∀ j € [1, Y-1] (4.4)
𝑌
In the cluster based location service, CH updates the locations of its CMs to
RSU periodically. Location updates and location queries depend on a reliable link of
CH with RSU. If a vehicle with the greater link lifetime is elected as CH, it remains in
the range of RSU for a long time. The positions of the vehicles are updated consistently
which ultimately improves the performance of location service. Hence, the link
lifetime is an important metric to evaluate the most suitable CH. Alsuhli et al. (2019)
depend on Link LifeTime (LLT) to enhance the link reliability between vehicles as
mentioned in Equation 4.5.
)|𝑆𝑖𝑔𝑛(△𝑣)
|𝐿𝐿𝑇| = 𝑅−|(𝐷𝑣𝑑 − 𝐷𝑣𝑘 (4.5)
|𝑉𝑑 −𝑉𝑘 |
where R is the communication range, 𝐷𝑣𝑑 − 𝐷𝑣𝑘 is the distance between two vehicles
vd and vk, 𝑣𝑑 − 𝑣𝑘 is the speed difference and 𝑆𝑖𝑔𝑛(△ 𝑣) is the sign function which
depends on the speed variation. If vd is faster than vk, then 𝑆𝑖𝑔𝑛(△ 𝑣=1 and -1
otherwise.
107
It is observed from the Equation 4.6, the vehicle Vr with lower speed have
higher RLT values; ultimately, these vehicle have higher possibility to become a CH.
The lower speed CHs are not stable with respect to its neighbours vehicles and leave
cluster frequently. So CH instability increases
where RSUp is the position of RSU, Vrp is the position of vehicle Vr and sign(△d) is
a sign function which depends on the distance between vehicle and RSU. CHEV
depends on RLT to elect a CH, and a vehicle has greater RLT has more chances of
being elected as CH.
|𝑅𝐿𝑇𝑑 | =
𝑅−|(𝑅𝑆𝑈𝑃 − 𝑉𝑝 (𝑑)) |𝑆𝑖𝑔𝑛(△𝑑)
∀ j € [1, Y-1] (4.7)
1 𝑦−1
| ∑ (𝑣𝑑 −𝑣𝑗 )|
𝑌 𝑗=0
where RLT is the link lifetime of CH with RSU, R is the communication range, RSUp
is the position of RSU, Vp(d) is the position of vehicle Vd and sign(△d) is a sign
function whose value is -1 when vehicles are approaching the RSU and its value is 1
when vehicles are moving away from the RSU as determined by Equation 4.10 (Lines
9-16, Algorithm 4.2). The description of the sign function is given in the following
section.
It can be observed from the Figure 4.4 that the vehicles in the range A have
more link lifetime as compared to vehicles in the range B. The vehicles that enter the
108
range A cover 800m distance before leaving the transmission range of the RSU.
Whereas, existing vehicles in the range B cover 400m distance to get out of the range
of the RSU. In the range A, when vehicles enter the transmission range of the RSU,
the distance between RSU and vehicle starts decreasing from 400m and at the RSU
position, it reaches to zero and starts increasing on other side of the RSU. When the
distance between RSU and vehicles reaches to 400m, vehicles exit the transmission
range of the RSU. During the calculation of LLT, there is a chance that vehicle Vb in Link Lifetime
the range B has larger LLT as compared to a vehicle Va in the range A due to same
distance from the RSU as shown in Figure 4.4. In actual, vehicles in the range B have
smaller LLT as compared to the vehicles in the range A. So, in proposed RCHE scheme
a sign function sign(△d) is formulated whose value is calculated by Equations 4.8, 4.9
and 4.10.
109
4.3.3.3 Calculation of Number of Neighbours (Nnbr)
CHEV of all vehicles in its range. The CV advertises a vehicle as a CH which has
highest value based on the Equation 4.11 (Lines 18-29, Algorithm 4.2).
Input:
Rseg: Road segment
Z: Number of vehicles // number of vehicles in a segment
Cseg: Centroid of road segment // center of each segment
Output:
CHER: Cluster Head Election Range
Start:
1 . Va send BMs ∀ Va ɛ [1, Z]
2 . Va receive BMs // vehicles share information
110
3. If CH exists
4. CH share its information
5. Go to line 1
6. Else
7. Min_distance = Distance_array [0]
8. For (b=1, b ≤ Z, b++) // election of centroid vehicle
9. Compute D(Vb) = distance (Vb, Centroid)
10. Distance_array [b] = D(Vb)
11. If ((distance(D(Vb)) < distance ((Min_distance))
12. Min_distance = D(Vb)
13. CV=Min_distance
14. Else Vb sends and receives BMs as a vehicle
15. End if
16. End for
17. CV advertise its status as a Centroid Vehicle
18. For (c=1, c ≤ Z, c++) // defining the CHER Centroid vehicle
𝑐 ⃗⃗⃗⃗⃗⃗⃗
𝐷𝑉 .𝐷𝑉𝐶𝑉⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
19. ɸ𝑐 = 𝑎𝑟𝑐𝑐𝑜𝑠 |𝐷𝑉
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ ||𝐷𝑉
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ |
// ɸ angle between a vehicle Vc and CV
𝑉𝑐 𝐶𝑉
20. CV_range(Vc) = distance (CV, Vc) ≤ 400 for three consecutive BMs
21. If ( ɸc ≤ Π/4 Ʌ CV_range(Vc))
22. CHER_Array[c]=Vc
23. Else Vc sends and receives BMs as a vehicle
24. End if
25. End For
26. Y=Sizeof(CHER_Array [])
27. End if
28. Return CHER
End
111
14. Sign(△d)= +1
15. End if
16. |𝑅𝐿𝑇𝑑 | = | 𝑅−|(RSUP)−(V|𝑉𝐼
P (d)|( Sign(△d) ))|
// RLT of vehicle Vd
𝑑|
17. Nnbr (Vd)= count (Nnbr_table) // Number of neighbours of Vd
1
18. 𝐶𝐻𝐸𝑉(𝑉𝑑 ) = 𝛼(𝑉𝐼 (𝑉𝑑 ) + 𝛽(𝑅𝐿𝑇(𝑉𝑑 ) + 𝛾 (𝑁𝑛𝑏𝑟 (𝑉𝑑 ))
19. Vd Sends CHEV to CV
20. CHEV_Array [d] = CHEV (Vd)
21. End for
22. X= Sizeof (CHEV_Array [d])
23. CH = CHEV_Array [0] // CH value is initialized at CHEV [0]
24. For (e=1; e < X; e++)
25. If (𝐶𝐻𝐸𝑉_𝐴𝑟𝑟𝑎𝑦[e] > 𝐶𝐻])
26. 𝐶𝐻 = 𝐶𝐻𝐸𝑉_𝐴𝑟𝑟𝑎𝑦[𝑒]
27. End if
28. End for
29. Return CH
End
112
Different weighting factors combinations are evaluated to analyze the effect of
α, β and γ values on the CH lifetime. Combinations CHEV(0.3, 0.4, 0.3), CHEV(0.3,
0.3, 0.4), CHEV(0.4, 0.4, 0.2), CHEV(0.2, 0.4, 0.4) and CHEV(0.5, 0.3, 0.2) are
analyzed to find the optimum combination that provides highest CH lifetime.
CHEV(0.3, 0.4, 0.3) configuration favours the RLT and number of neighbours as
compared to stability of the vehicle. In the CHEV(0.3, 0.3, 0.4) combination, CH
lifetime is effected due to more concentration on the number of neighbours.
CHEV(0.4, 0.4, 0.2) combination although gives better CH lifetime values but still is
affected due to same preference assigned to stability and RLT of the vehicle. In CHEV
(0.2, 0.4, 0.4) combination, the lifetime of the CH is affected due to assigning less
values to mobility parameters.
The weighting factors group of CHEV(0.5, 0.3, 0.2) has shown the best results
in terms of highest CH lifetime as shown in Figure 4.5. The stability of the CH depends
on the VI, and RLT depends on the VI. In particular, if the election of the CH is based
on the stability with neighboring vehicles and highest RLT, it will ensure a most stable
CH.
200
180
160
CH Lifetime(seconds)
140
120
100
80
60
40
20
0
0.3,0.3,0.4 0.3,0.4,0.3 0.5,0.3,0.2 0.4,0.4,0.2 0.2,0.4,0.4
Weighting Factors
113
4.5 Performance Evaluation of the RCHE Scheme
The stability of the cluster directly depends on the lifetime of a CH. The larger
value of the CH lifetime represents that cluster configuration is more stable. The
Average CH lifetime is obtained through dividing the total durations of the CHs in the
network by total number of CHs. This section demonstrates the performance
comparison of the RCHE scheme against HCBLS, and CBLS schemes in terms of
average CH lifetime based on speed and density.
114
(a) Impact of Speed on Average CH Lifetime
The CH lifetime of the HCBLS depends on the size of the segment. The
elected CH will take less time to cover same distance with the increase in speed. The
CH lifetime of the HCBLS decreases with the increase in the speed. In the CBLS,
every vehicle calculates its speed difference, distance difference with neighbouring
vehicles and number of neighbours to compute CHEV. Each vehicle broadcasts this
information. In turn, each neighbour again broadcasts this information. The messages
beyond certain delay are discarded. The links between vehicles further breaks with the
increase in speed, which leads to incomplete and unreliable CHEV. Moreover, the
CHEV is affected due to undefined weights of the metrics.
In the proposed RCHE scheme, the CH election includes only those vehicles
which remain in the range of the CV for the three consecutive intervals, and eliminates
the fast moving and slow moving vehicles from the CH election range which ensures
a stable CHEV. The CH is elected by obtaining average relative velocity of individual
vehicle with respect to neighboring vehicles in the network. The use of the VI in RLT
value further magnifies the stability of the CH. These factors contribute in achieving
the improved CH lifetime with the increase in the speed of the vehicles. For the lowest
speed (36km/h), the average CH lifetime of the proposed RCHE scheme is improved
by 19% and 78% as compared with the CBLS and HCBLS respectively. The increase
in speed affects the performance of all schemes, because of the rapid topological
changes. The RCHE scheme showed the best performance as compared to the HCBLS.
In the HCBLS, the CH lifetime is limited to the segment of the road. In the HCBLS,
the CH takes less time to cover the road segment at higher speeds. At the highest speed
(108km/h), the RCHE achieves 30% and 89% improved average CH lifetime as
compared to CBLS and HCBLS respectively as given in Figure 4.6.
115
RCHE CBLS HCBLS
200
Average CH Lifetime(seconds)
150
100
50
0
36 54 72 90 108
Maximum Speed (Km/h)
Figure 4.7 shows the performance of the RCHE, CBLS and HCBLS schemes
in terms of average CH lifetime against the maximum number of the vehicles in the
network. In HCBLS, CH lifetime depends on the length of the road segment and speed
of vehicles. With the increase in density, the stay time of a CH in a segment increases
which ultimately increases the lifetime in a segment as compared to speed based.
However, lifetime of a CH in static clustering is limited to length of the road segment.
CBLS depends on dynamic clustering. The CH election in CBLS is based on the
broadcasting approach. The messages beyond certain delay are discarded. With the
increase of the density, the number of discard messages increases which ultimately
affects the CHEV. The high vehicular densities result in high congestion and
contention on transmission channels which lead to losing CHEV messages.
116
maximum density 950) as compared to the CBLS and HCBLS respectively, as shown
in Figure 4.7.
200
Average CH Lifetime(seconds)
180
160
140
120
100
80
60
40
20
0
150 350 550 750 950
Maximum Density
It is the ratio of number of the queries that have been responded to the number
of queries that have been sent from the source. The QSR depends the stability of the
CH, CM, location updates and reliable links of the CH with the RSU. Figure 4.8 and
4.9 show the query success rate comparison of RCHE, CBLS and HCBLS.
In the HCBLS, the CH which carries the location information of its CMs
changes frequently due to static clustering approach. Due to high mobility of vehicles,
the lifetime of the CH reduces. In the CBLS, location service depends on CH,
intermediate server and main server, which are all moving location servers. With the
increase in speed, the instability of these servers increases which decrease the query
success rate. The query success rate of the proposed RCHE scheme is improved due
to the stability of the CH and because of reliable links with the RSU.
117
The results show that the RCHE scheme achieves better query success rate as
compared to counterparts. The RCHE scheme achieves 11% and 21% improved query
success rate at lowest speed (36 km/h) as compared to the CBLS and HCBLS
respectively. At the highest speed (108km/h), the query success rate is improved by
17% and 26% against the existing CBLS and HCBLS schemes respectively as given
in Figure 4.8.
100
90
Query Sucess Rate (%)
80
70
60
50
40
30
20
10
0
36 54 72 90 108
Maximum Speed (Km/h)
Figure 4.9 shows the performance of the RCHE, CBLS and HCBLS schemes
in terms of query success rate against the number of the vehicles in the network. In the
HCBLS, with the increase in the density, the lifetime of each CH and CM in a segment
increases; however, lifetime is limited to road segment. The location updates and
location queries are affected due to change of CH in each segment. The CBLS depends
on broadcasting approach without defining any range. The messages beyond certain
delay are discarded. With the increase of density, the contention on communication
channels increases which causes further messages loss. The CHEV is affected which
increases the CH instability.
118
In the proposed RCHE scheme, density is managed by first defining the CH
election range and then electing a CH. Also, the CH election depends on reliable link
with the RSU. The CH showed reasonable stability in the increased density. The query
success rate of the proposed RCHE scheme is improved by 10%, and 47% for the
lowest density (150 vehicles in the network) and 20% and 22%, for highest density
(950 vehicles in the network) as compared to the CBLS and HCBLS respectively.
90
80
Query Success Rate(%)
70
60
50
40
30
20
10
0
150 350 550 750 950
Maximum Density
The query response delay is the difference between the time when a vehicle
sends a query and the time when vehicle receives the respond. It is the average of the
delay that is obtained by the ratio of total elapsed time of the queries to the total number
of queries. It is an important parameter to measure the performance of the location
service in vehicular networks. The main objective of the location service is to provide
the fresh locations of vehicles. The freshness of locations is compromised with the
increase in delay. Figure 4.10 and Figure 4.11 shows the delay comparison of RCHE,
CBLS and HCBLS.
119
(a) Impact of Speed on Query Response Delay
In the cluster-based location service, delay also depends on the stability of the
CH. The lifetime of the CHs in HCBLS is limited to road segment. A new CH is elected
for each segment. The frequent changes in the CHs affects the location updates and
location queries. In the CBLS, the instability of the CH and higher level servers
increases due to change in speeds. The instability of the CHs triggers the CH election
process frequently, which applies that more control messages are transmitted to again
elect the CH, which affects the location updates and location queries.
The delay of the proposed RCHE scheme is improved due to stability of CHs
at higher speeds. The location updates and location queries are improved. The results
show that the RCHE scheme shows less delay as compared to counterparts. The RCHE
scheme achieves 8% and 20% improved delay at lowest speed (36km/h) and 18% and
26% at highest speed (108km/h) as compared to the CBLS and HCBLS respectively
as shown in Figure 4.10.
500
450
Query Response Delay(ms)
400
350
300
250
200
150
100
50
0
36 54 72 90 108
Maximum Speed (Km/h)
120
(b) Impact of Density on Query Response Delay
Figure 4.11 shows the performance of the RCHE, CBLS and HCBLS schemes
in terms of query response delay against the number of the vehicles in the network. In
HCBLS, query response delay increases due to the instability of the CH and CH. The
CBLS depends on the broadcasting approach to elect the CH. The stability of the CH
is affected due to non-optimum values of the speed and distance and ignoring the
proper weightage of speed, distance and number of neighbours. The instability of the
CH further increases with the increase in the density. The network congestion and
contention increase in the network which delays the location queries.
500
450
Query Response Delay(ms)
400
350
300
250
200
150
100
50
0
150 350 550 750 950
Maximum Density
121
4.6 Summary
122
DYNAMIC CUMULATIVE CLUSTER MEMBER AFFILIATION SCHEME
5.1 Introduction
123
could not affiliate to any cluster due to static threshold values. IVs increase in the
network.
Due to frequent change, the position of these CMs is not updated consistently
to higher-level servers. Moreover, location information of the isolated vehicles which
are not part of any cluster is not updated to location servers. The cluster member
instability increases. Consequently, performance of the location service highly
affected due to unavailability of location information of these CMs. Due to missing
location information of the vehicles, query success rate decreases. Moreover, due to
frequent and repeating CMs affiliation requests, location update and location query
delay increases. Therefore, a Dynamic Cumulative Cluster Member Affiliation
(DCCMA) scheme is incorporated in the clustering that addresses the aforementioned
issues effectively. Table 5.1 highlights the comparison of the existing and proposed
DCCMA scheme.
124
link reliability with the CH. Referring to Figure 5.1, the flowchart explains the overall
working of the DCCMA scheme, where the blue colored blocks refer to the
contribution of this research. The algorithm 5.1 explains the overall procedure of the
DCCMA scheme.
125
The proposed DCCMA scheme depends on the Cumulative Moving Average
Speed (CMAS) instead of the instantaneous speed to calculate CHLT. The IVs are
reduced by increasing the flexibility to affiliate the vehicles. The DCCMA scheme is
based on three phases such as CH and RSU sensing phase, CM eligibility computation
phase and CM affiliation phase. The output of each phase is used as input of the next
page.
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐷𝑉𝐼𝑉 .𝐷𝑉 ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐶𝐻(𝑖) (5.1)
𝜙(IV ,𝐶𝐻(𝑖)) = 𝑎𝑟𝑐𝑐𝑜𝑠 |𝐷𝑉
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ ||𝐷𝑉
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ |
∀ i ɛ [1, n]
𝐼𝑉 𝐶𝐻(𝑖)
where ⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐷𝑉𝐼𝑉 = 𝐷𝑉𝐼𝑉𝑥 𝑥 + 𝐷𝑉𝐼𝑉𝑦 𝑦 and ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐷𝑉𝐶𝐻(𝑖) = 𝐷𝑉𝐶𝐻(𝑖)𝑥 𝑥 + 𝐷𝑉𝐶𝐻(𝑖)𝑦 𝑦 are the
direction vectors of the IV and CH(i) respectively.
126
Figure 5.2 Road ID and Direction of the Vehicles
127
Alsuhli et al. (2019) depend on instantaneous speed of the vehicles to calculate
the Link LifeTime (LLT) as mentioned in Equation 5.4.
)| 𝑆𝑖𝑔𝑛(△𝑣)
|𝐿𝐿𝑇| = 𝑅−|(𝐷𝑣𝑑 − 𝐷𝑣𝑘 (5.4)
|𝑉𝑑 −𝑉𝑘 |
The proposed DCCMA scheme utilize the CMAS of the CH(i) to calculate
the CHLT(i) using the Equation 5.5.
CHCR − | DCH(i) − DIV |sign(△CH(i)−IV ) (5.5)
𝐶𝐻𝐿𝑇(𝑖) =
| 𝐶𝑀𝐴𝑆𝐶𝐻(𝑖) − 𝑉𝐼𝑉 |
where CHCR is the Cluster Head Communication Range, | DCH(i) − DIV | is the
distance between CH(i) and IV, 𝐶𝑀𝐴𝑆𝐶𝐻(𝑖) is the cumulative moving average speed
of CH(i), 𝑉𝐼𝑉 is the speed of isolated vehicle and sign(△CH−IV ) is the sign function
whose value is 1 when CH(i) is faster than IV, otherwise -1. IV calculates CHLT with
each of the neighbouring CH(i). IV compares the CHLT of these CHs and sends CM
request to the CH which has the highest CHLT (Lines 23-33, Algorithm 5.1).
128
5.3.3 CM Affiliation Phase
In the existing studies, CM affiliation threshold value 𝑇𝑟𝑒𝑠ℎ depends on the fix
values while ignoring the vehicle to CH and CH to RSU update intervals which causes
the increase in IVs in the network. In the proposed DCCMA scheme, 𝑇𝑟𝑒𝑠ℎ is dynamic
that depends on the 𝐼𝑉_𝐶𝐻_𝑈𝑝𝑑𝑎𝑡𝑒_𝐼𝑛𝑡𝑒𝑟𝑣𝑎𝑙 and 𝐶𝐻_𝑅𝑆𝑈_𝑈𝑝𝑑𝑎𝑡𝑒_𝐼𝑛𝑡𝑒𝑟𝑣𝑎𝑙 as
shown in Figure 5.4. The CH(i) compares the CHLT value with the CM affiliation
threshold value 𝑇𝑟𝑒𝑠ℎ that is based on the equation 5.6 (Lines 34-38, Algorithm 5.1).
129
Figure 5.4 Update Intervals from the IV to CH and CH to RSU
130
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐷𝑉𝐼𝑉 .𝐷𝑉 ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝐶𝐻(𝑖)
8 𝜙(IV ,𝐶𝐻(𝑖)) = 𝑎𝑟𝑐𝑐𝑜𝑠 |𝐷𝑉
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ ||𝐷𝑉
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ |
𝐼𝑉 𝐶𝐻(𝑖)
9 If ((𝜙 ≤ π/4 and IV_Road-ID == CH(i)_Road-ID))
10 IV compute CHLT(i) // for all in the range of IV
V𝑛+1 + n.CMAS(i)n
11 𝐶𝑀𝐴𝑆(𝑖)𝑛+1 = n+1
12 IV computes sign(△CH−IV )
13 If (𝐶𝑀𝐴𝑆𝐶𝐻(𝑖) > 𝑉𝐼𝑉 )
14 sign(△CH−IV )=1
15 Else
16 sign(△CH−IV )= -1
17 End if
CHCR−| DCH(i) −DIV(j) |sign(△CH−IV )
18 𝐶𝐻𝐿𝑇(𝑖) = | 𝐶𝑀𝐴𝑆𝐶𝐻(𝑖) −𝑉𝐼𝑉 |
19 CHLT-A[]= CHLT(i)
20 Else print CH(i) has different direction
21 End if
22 End For
23 U= Sizeof(CHLT-A[])
24 If (D=0)
25 Go to line 40
26 Else
27 Highest=CHLT-A[0]
28 For (i=1; i ≤ U; i++)
29 If (CHLT-A[i] > Highest
30 Highest= CHLT-A[i]
31 Else CH(i) has lower CHLT
32 End if
33 End For
34 IV sends CM request to Highest CH(i)
35 If (CHLT-A(i) ≥ Tresh) // CH compares CHLT with Tresh
36 CH sends the CM_reply to IV
37 IV acknowledge the CM_reply
38 CH(i) updates location_table
39 Else
40 If distance(IV(j), RSU) ≤ to 400)
41 IV sends its position to RSU
42 RSU updates location table
43 Else go to Line 1
44 End if
45 End if
46 End if
47 Return CM
End
131
5.4 Performance Evaluation of the DCCMA Scheme
132
5.4.1 Average CM Lifetime
Figure 5.5 presents the average CM lifetime of CBLS, MoGLS and DCCMA
schemes against different maximum speed of the vehicles. The CM lifetime of all
schemes decreases with the increase in speed as shown in Figure 5.5. It can be
considered normal because when vehicles move faster, the network topology becomes
more dynamics, this makes it difficult for the CMs to remain affiliate with clusters.
The CBLS depends on the instantaneous speed difference with the CH. With the
increase in speed, the links between CH and CMs break frequently which reduces the
lifetime of each CM in cluster. In the MoGLS, the lifetime of the CM is improved by
considering sojourn time which depends on the speed and distance of the vehicles.
However, the CM lifetime is still affected due to depending on instantaneous speed
and ignoring relative speed and distance with the CH.
The proposed DCCMA scheme showed the best average CM lifetime due to
the enhanced stability of the CM in the cluster. The stability of the CM also increases
with the stability of the CH. In the DCCMA scheme, the stability of the CM with the
CH is further enhanced by considering the reliable CMAS instead of instantaneous
speed. The CMs remain more stable with their CHs as compared to the CBLS and
MoGLS which leads to enhanced CM lifetime. The results show that the DCCMA
scheme achieves better average CM lifetime as compared to compared to the CBLS
and MoGLS. At the lowest speed (36km/h), the average CM lifetime of the proposed
scheme is improved by 22% and 26% as compared with the MoGLS and CBLS
133
respectively. The increase in speed affects the performance of all schemes, however,
due to its improved CM affiliation scheme, the DCCMA showed better performance.
At the highest speed (108km/h), the DCCMA achieves 32% and 41% improved
average CM lifetime as compared to MoGLS and CBLS respectively as given in
Figure 5.5.
200
180
Average CM Lifetime (sec)
160
140
120
100
80
60
40
20
0
36 54 72 90 108
Maximun speed (km/h)
Figure 5.6 shows the performance of the DCCMA, MoGLS and CBLS
schemes in terms of average CM lifetime against the maximum number of the vehicles
in the network. In the CBLS scheme, only those vehicles affiliate with the CH, which
have maximum speed difference with the CH equal to 18km/h. Also, the maximum
number of vehicles in a cluster is limited to 20. In the MoGLS scheme, every vehicle
calculates its sojourn time and compares it with the with the fixed threshold value that
is 62.5 seconds. Only those vehicles affiliate with the CH whose sojourn time is greater
than 62.5. Although the existing schemes attempted to improve the CM lifetime by
synchronizing the speed and sojourn time with the CH, however, the number of
isolated vehicles in the network increases due to the fixed threshold values. Moreover,
the CM lifetime is affected due to CH instability and due to considering instantaneous
values of the speed. The isolated vehicles continuously send CM affiliation messages
134
continuously to affiliate with the clusters. These issues are further magnified with the
increase in the density of the vehicles. The network congestion and communication
contention increase. The collision and loss of the messages increase which reduces the
lifetime of the existing CMs.
180
160
Average CM Lifetime (sec)
140
120
100
80
60
40
20
0
150 350 550 750 950
Maximun density
The fraction of vehicles saved in the location server is the number of vehicles
saved in the location table of the location servers divided by the real number of vehicles
135
in the map. It is important parameter to measure the performance of the location service
in vehicular networks. The vehicles are the entities whose position is maintained in the
location servers. The performance of the location service directly depends on the
number of vehicles saved in the CHs and RSUs, which are acting as location servers.
The fraction of the vehicles saved in the location servers depends on the performance
of the CM affiliation schemes. Figure 5.7 and 5.8 illustrate the performance
comparison of the DCCMA scheme against CBLS and MoGLS schemes in terms of
fraction of vehicles saved in the location servers.
136
DCCMA MoGLS CBLS
Figure 5.8 shows the performance of the DCCMA, MoGLS and CBLS
schemes in terms of fraction of the vehicles saved in location servers against the
location information of the vehicles saved in the network. In the CBLS scheme,
vehicles whose speed difference with the CH is greater than 18km/h could not affiliate
with the cluster. Furthermore, in the MoGLS schemes, vehicles whose sojourn time is
less than 62.5 seconds are unable to join the cluster. Due to fixed CM threshold values
in these schemes, the isolated vehicles in the network increases. These vehicles send
continuously send CM requests on the network to affiliate with the cluster causing
extra communication messages on the network which is further amplified with the
increase in density of vehicles. The network congestion increases which results in the
loss of CM location update messages.
The proposed DCCMA scheme showed better performance in high density due
to flexible and reliable CM affiliation scheme. The number of the isolated vehicles in
the DCCMA scheme reduces due to dynamic CM affiliation threshold and the stability
of the CM in cluster. Moreover, vehicles which could not find any CH in their range
and could not affiliate with the cluster sends their location information directly to the
nearest RSU. Figure 5.8 shows the performance of the three schemes in terms of
137
fraction of the vehicles saved in location servers against the number of the vehicles in
the network. The results show that the fraction of the vehicles saved in the location
servers is improved in the DCCMA scheme by 8% and 9% (with maximum 150
vehicles in the network) and 15% and 23% (with maximum vehicles 950 vehicles in
the network) as compared to CBLS and MoGLS respectively.
100
95
90
85
80
75
70
65
60
55
50
150 350 550 750 950
Maximun density
The query success rate is defined as the ratio of total queries to the correctly
responded ones. Query is initiated by a vehicle who needs the position of another
vehicle. Vehicle sends the query to its CH, if CH has the requested location, it responds
otherwise it fulfils the request by sending it to high level location server. Figure 5.9
and 5.10 show the query success rate comparison of DCCMA, CBLS and MoGLS.
138
instability of these servers increases. The reliability of the CMs with the CH and the
location information of the vehicles saved in the location service reduces with the
increase in the vehicles. In the MoGLS scheme, due to unreliable links with the CM
that depends on the instantaneous values of the speed, the connections with the CH
increases with the increase in speed. In these schemes, the query success rate is
affected due to CH and CM instability. The location information saved in the RSU
decreases with the increase of the speed.
The query success rate of the proposed DCCMA scheme is improved with the
increase in the speed due to a reliable link with CH that is based on the CMAS. The
results show that the DCCMA scheme achieves better query success rate as compared
to its counterparts. The RDCMA scheme achieves 7% and 9% improved query success
rate (with maximum speed 36 km/h) and 13% and 17% (with maximum speed
108km/h) as compared to the CBLS and MoGLS respectively as shown in Figure 5.9.
95
90
Query Success Rate (%)
85
80
75
70
65
60
55
50
36 54 72 90 108
Maximun speed (km/h)
Figure 5.10 shows the performance of the DCCMA, MoGLS and CBLS
schemes in terms of query success rate against the number of the vehicles in the
network. In CBLS and MoGLS, the number of isolated vehicles increases due to CM
139
affiliation mechanisms. These vehicles continuously send CM request to neighbouring
CHs. These messages further increase with large densities. The network congestion
and communication contention on the channels increase. The collision of the messages
increases which results in the loss of messages. The query success rate of the CBLS
decreases with the increase in density due to high contention and congestion on the
network.
In the proposed DCCMA scheme, effect of the density reduces due to decrease
in isolated vehicles. This is because of the dynamic and reliable CM affiliation
schemes. The query success rate showed reasonable stability in the increased density.
The DCCMA scheme achieves 10% and 14% improved query success rate (with
maximum 150 vehicles in the network) and 17% and 24% (with maximum 950
vehicles in the network) as compared to the CBLS and MoGLS respectively.
95
90
85
Query Success Rate (%)
80
75
70
65
60
55
50
150 350 550 750 950
Maximun density
5.5 Summary
140
the CH. Moreover, in the proposed scheme, the isolated vehicles were reduced by
proposing dynamic threshold values instead of static values. Also, in the absence of
the nearby CH, the isolated vehicles updated the location information directly to the
RSU without announcing itself as a CH. This contributed, in improving the
performance of the location service by enhancing the cluster stability. The
performance of the DCCMA mainly depended on the CM lifetime. The CM lifetime
of DCCMA at 36km/h is 173 seconds as compared to MoGLS and CBLS CM lifetimes
142 seconds and 137 seconds respectively. The CM lifetime of DCCMA is improved
due to CMAS and dynamic threshold values. The high value of CM lifetime indicated
that a CM remains affiliated with a CH for a longer time. The location of CMs were
updated to CH efficiently. The location information of CMs in location servers
increases, which ultimately enhance the performance of the location service in terms
of fraction of vehicles location information saved in location servers and query success
rate. Due to rapid and unpredictable topological changes in VANETs, there is no
baseline and upper line for acceptable performance. A CM may leave the affiliation
soon after joining a cluster due to collision or delay of Beacon Messages (CM) with
CM lifetime 0. The CM lifetime depends on CM affiliation scheme. The higher values
of CM showed a long life affiliation with the CH. Simulation results showed that the
DCCMA scheme significantly improved the performance of the location service in
terms of increasing cluster member lifetime by an average of 31%, increasing the
fraction of the vehicles saved in the location servers by 16% and increasing the query
success rate by 14%.
141
OPTIMIZED CLUSTER MAINTENANCE SCHEME
6.1 Introduction
It has been observed from the past literature that all roads have intersections
at the different positions of the roads. The vehicles in the cluster disbands into different
143
directions on reaching the intersection due to their different destinations. The Cluster
Head (CH) of the cluster move only in one direction while leaving the CMs in other
directions without any CH. The cluster instability increases around the intersection.
Existing studies split the cluster into three clusters after entering the intersection area
and reclustering after leaving the intersection area. The clustering process is triggered
twice after entering and after leaving the intersection area. These studies even
reclustered the stable cluster in the intersection area. The splitting is performed at the
cost of cluster instability due to unnecessary clustering. Furthermore, new CHs are
elected without considering the link status with high level location servers. Table 6.1
differentiates the existing and proposed OCM scheme.
144
information and reliable link of the CMs. New CHs are elected while considering the
destination and reliable link with the location servers.
Referring to Figure 6.1, the flowchart explains the overall working of the
cluster merging algorithm where the blue colored blocks refer to the contribution of
this research. The algorithm 6.1 explains the overall procedure of the cluster merging.
CHs send and receive Beacon Messages (BMs) to share their information. The CH
with the highest id invokes the merging process by sending merging requests to CHs
which have same road Identity (ID) and direction. Upon receiving the
acknowledgements from the neighbouring CHs, the CH initiates the merging process
and computes the merging conditions (Lines 1-11, Algorithm 6.1). The combination
of the road ID and direction ensures that CHs are moving on the same road and in the
same direction. Otherwise, CH moving in the same direction but onto another parallel
road will also participate in the cluster merging which will result into temporary and
unreliable merging that further enhance the cluster instability.
In most of existing studies, the merging is started when all of the CMs are in
the range of each other. It happens when one CH is moving faster as compared to other.
Although merging is completed, however, it does not ensure a stable merging due to
difference of speeds. In the proposed cluster merging algorithm, when the distance
between two CHs is less than or is equal to R as given in the Equation 6.1 and shown
in Figure 6.2, the merging is invoked because it ensures that half of the both cluster
overlap each other.
145
Figure 6.1 Flowchart of the Cluster Merging
146
Figure 6.2 Overlapping Range of Two Clusters
The time during which overlapping CHs wait before the start of the cluster
merging is called Merging Threshold Time (MTT) represented by Tmerge. In the
existing studies Tmerge relies on a fixed value or location update interval of one CH
which lead to unreliable merging. The location update interval is the time in which
every CH periodically updates the location of its CMs to RSU. Tmerge is not sufficient
to ensure a valid merging. The Tmerge time should be defined with minimum limits such
that the participating CHs may contact with the higher-level location server to update
the location information of their CMs. It also confirms that RSU receives location
information from multiple CH within the same overlapping area which increase extra
communication overhead and management functions to RSU. Therefore, to make
stable merging, the Tmerge depends on the update interval of both merging CHs which
is an enhanced version of the existing studies, as defined in Equation 6.2.
If CHs remain in the range of each other up to Tmerge, then the merging process
will start. Although, the existing condition certifies that the merging CHs remain in
the range of each other but it did not ensure that merging clusters move towards each
other or move away from each other. If during Tmerge, the merging clusters move away
from each other, the merging process will be started and completed during Tmerge which
results in temporary and unreliable merging. In the proposed cluster merging
algorithm, a new condition is included that depends on the distance at the start of the
Tmerge and the end of the Tmerge.
147
Previous Distance (PD) is calculated at the start of the Tmerge using the Equation
6.3 and Current Distance (CD) is calculated at the end of the Tmerge based on the
Equation 6.4. Equation 6.5 is the combined version of Equation 6.3 and Equation 6.4.
Equation 6.5 guarantees that merging CHs move towards each other. The merging of
the clusters is started subject to the execution of the conditions defined by Equation
6.1,6.2 and 6.5. The optimized merging conditions ensure a reliable cluster merging as
compared to the existing studies.
If the merging conditions are not satisfied, existing clusters do not merge and
move with their CMs (Lines 12-24, Algorithm 6.1).
After merging, the next step is to nominate the CH of the merged cluster from
the existing CHs. Every involving CH computes its RLT using Equation 6.6. The CH
which has highest RLT is elected as a CH of merged cluster (Lines 25-36, Algorithm
6.1).
The new CH registers with the RSU, advertises itself as a new CH and sends
CM requests messages. The CH affiliates the CMs and updates their location in the
location table. Then new CH sends the location information of its CMs to the nearest
RSU as shown in Figure 6.2 (Lines 37-44, Algorithm 6.1).
148
In the proposed cluster merging algorithm, the Tmerge and overlapping distance
at the start and end of Tmerge are optimized which ensures a stable merging. Moreover,
the CH of the merged cluster is elected from the existing CHs based on the higher RLT
with the RSU. The collision and delay of the location updates and location queries are
reduced due to optimized cluster merging.
149
30 For ( p=1; p ≤ D; y++)
31 If(Merging_array[p] > Maximum_RLT)
32 Maximum_RLT=Merging_array[p]
33 Else CH(p) has a lower value
34 End if
35 End For
3 6 Maximum_RLT is elected as CH for the merged cluster
3 7 CH(p) register itself with the nearest RSU
3 8 CH(p) sends CM advertisement messages
3 9 CMs reply acknowledge messages
4 0 CH(p) affiliate new CMs
4 1 CH(p) sends location information to RSU
42 return merged cluster
43 End if
44 Else Merging is not possible
4 5 End if
End
Referring to Figure 6.3, the flowchart explains the overall working of the
splitting algorithm, where the blue colored blocks refer to the contribution of this
research. The algorithm 6.2 explains the overall procedure of the splitting. Every
vehicle gets its Relative Destination (RD) through the usage of the digital map. The
RD locate the next approaching intersection in the route of each vehicle. Moreover,
the turning direction of each vehicle is determined through RD. Therefore, vehicles
have advance information about their turning directions. Vehicles send and receive
beacon messages to share location information. The area equal to radius (½)R, from
the centre of every intersection is designated as intersection area, where R is the
communication range as shown in Figure 6.4. After entering the intersection area,
existing CH advertises its ID and RD and turning direction on the basis of RD (lines
1-5, Algorithm 6.2). After advertising its RD and turning direction, the CH starts the
process to elect CHs for remaining directions. The CMs of CH compute their RLT
value using Equation 6.7. The RLT metric ensures that a stable vehicle which has
highest reliability with the RSU is elected as a CH for a particular direction. Every CM
sends its RLT and RD to CH. The existing CH forms two groups on the basis of RD
(lines 6-10, Algorithm 6.2). The existing CH compares the RLT and RD of each CM,
and elects two CHs for other two directions on the basis of highest RLT and RD.
150
Figure 6.3 Flow Chart of the Splitting Algorithm
151
Figure 6.4 Cluster Splitting Around the Intersection
where R is the communication range, | DRSU – DVj | is the distance between RSU and
CMj, |𝑆𝑖𝑔𝑛( △ 𝑑) | is a sign function whose value is -1 when CM is approaching
RSU and 1 when CM is moving away from the RSU; and |𝑣𝑚 – 𝑣𝑗 | represents the speed
difference between CMm and CMj (Lines 11-35, Algorithm 6.2)..
New CH register itself with the nearest RSU. After registering, CH sends CM
advertisement messages. The CMs in the range of new CH reply with the
acknowledgement messages. The CH affiliates the CMs and updates their location in
the location table. Then new CH sends the location information of its CMs to the
nearest RSU (lines 36-41, Algorithm 6.2).
Input:
C: Cluster
CH: Cluster Head
CMs: Cluster Members
RSU: Road Side Unit
Output:
Splitted Cluster
Start
1 CH(i), CM(j) // CH with id (i) and CM(j) send and receive BMs
2 If (d(CH(i), CoI)) ≤ ½(R)
3 CH(i) advertise its ID, Relative destination and turn direction
152
4 Else go to line 1
5 End if
6 Every CM(j) send its RLT and relative destination to CH(i)
𝑅 −| DRSU –DVj | 𝑆𝑖𝑔𝑛( △𝑑) |
7 𝑅𝐿𝑇𝐶𝑀(𝑚) = 1 𝑛−1
∑
𝑁 𝑗=0
|𝑣𝑚 –𝑣𝑗 |
8 N= number of CMs in the neighbour table of CH(i)
9 Destination-B-Array[]=RLT.CM(m) // array of CMs having same direction-B
10 Destination-C-Array[]=RLT.CM(m) // array of CMs having same direction-C
11 P= Sizeof (Destination-B-Array [])
12 If (P=0)
13 No any CM for this direction
14 Else
15 CH-B=Destination-B-Array[0] // CH-election for direction -B
16 For ( y=1; y ≤ P; y++)
17 If (Destination-B-Array[y] > CH-B)
18 CH-B=Destination-B-Array[y]
19 Else CM has lower LLT
20 End if
21 End For
22 End if
23 CH(i) Advertise CH-B as a CH for Destination-B // CH for Direction-B
24 Q= Sizeof (Destination-C-Array [])
25 If (Q=0)
26 No any CM for this direction
27 Else
28 CH-C=Destination-C-Array[0] // CH-election for direction -C
29 For ( x=1; x ≤ Q; x++)
30 If (Destination-C-Array[x] > CH-C)
31 CH-C=Destination-C-Array[x]
32 Else CM has lower LLT
33 End if
34 End For
35 End if
36 CH(i) Advertise CH-C as a CH for Destination-C // CH for Direction-C
37 CH-B, CH-C sends CM advertisement messages
38 CMs reply acknowledge the advertisement messages
39 CH-B, CH-C affiliate new CMs and update their information in location table
40 CH-B, CH-C sends location information to RSU
41 return split clusters
End
153
electing CHs for remaining directions. The same CHs perform the clustering functions
after leaving the intersection. In the optimized cluster splitting, the ratio of reclustering
is reduced at each intersection. Also, the extra CH election communication overhead
decreases. The rate of change of CHs around the intersection is reduced. Moreover,
the new CHs are elected on the basis of link reliability with the RSU. These
enhancements have positive effect on the performance of the location service.
The performance of the OCM is evaluated and compared with existing cluster-
based location service schemes such as HCBLS, CBLS and MoGLS. These schemes
are most related, latest, ISI indexed and designed specifically for the location service.
The OCM scheme depends on two algorithms such as cluster merging and cluster
splitting to provide cluster maintenance. These two algorithms are implemented
independently. The performance of OCM is measured by only cluster merging
algorithm or by cluster splitting algorithm or by the implementation of both algorithms
simultaneously based on the requirement of the cluster maintenance.
154
The CBLS and HCBLS schemes only proposed CH election and CM
affiliation schemes while overlooked the cluster maintenance. The MoGLS scheme
has improved the stability of the cluster by introducing cluster merging algorithm, but
the problem of the cluster splitting around the intersection is still not proposed. In the
existing studies, due to the clustering and reclustering around the intersection, the CH
election schemes, unreliability of the CH with the location server and the cluster
instability increases. The performance of the cluster-based location is affected in terms
of the QSR, QRD and LE.
The query success rate is defined as the ratio of total queries to the correctly
responded ones. It is an important parameter to measure the performance of the
location service in vehicular networks. In the cluster-based location service, query
success rate depends on the reliability of the location servers, fractions of vehicles
location information saved in the location servers, stability of the CH and CM and on
the efficiency of the cluster maintenance schemes Figure 6.5 and Figure 6.6 shows the
query success rate comparison of RCHE, CBLS and HCBLS.
Figure 6.5 shows the performance of the OCM, MoGLS, CBLS and HCBLS
schemes in terms of query success rate against the high speed of the vehicles in the
network. HCBLS is based on static clustering. When vehicles move faster, the lifetime
of each CH and CM in a segment reduces. In each segment, every CM affiliates with
155
a new CH. The instability of the CH as well as of the CM increases. In the CBLS,
location service depends on CH, intermediate server and main server, which are all
moving location servers. With the increase in speed, the instability of these servers
increases which decrease the query success rate. In the MoGLS scheme, the query
success rate is affected due to instability of the CH and reduced lifetime of the CM.
The query success rate of the proposed OCM scheme is improved at higher
speeds due to stability of CHs and CMs, reliable links with the CH and RSU; and
efficient cluster maintenance schemes. The results show that the OCM scheme
achieves better query success rate as compared to counterparts. The OCM scheme
achieves 8%, 12% and 22% improved query success rate (with maximum speed 36
km/h) and 21%, 29% and 39% (with maximum speed 108km/h) as compared to the
MoGLS, CBLS and HCBLS respectively as shown in Figure 6.5.
100
95
Query Success Rate (%)
90
85
80
75
70
65
60
55
50
36 54 72 90 108
Maximun speed (km/h)
Figure 6.6 shows the performance of the OCM, MoGLS, CBLS and HCBLS
schemes in terms of query success rate against the number of the vehicles in the
network. In the HCBLS, query success rate is improved as compared to speed. With
the increase in density, the CH and CM stay in the segments for more time. The
156
lifetime of the CH and CM increases, however it is limited to segment lengths. Due to
frequent change of CHs and CMs, the query success rate decreases. In the CBLS and
MoGLS schemes, the number of isolated vehicles increases due to CH and CM
schemes. Isolated vehicles continuously send CM affiliation messages to join a cluster.
With the increase of density, the network congestion and contention increases. The
query success rate of the both schemes decreases. However, the MoGLS scheme shows
better performance due to the inclusion of cluster merging algorithm in the cluster
maintenance.
The proposed OCM scheme showed better performance with the increase of
density, due to defining CHER, reliable links with the RSU and CMs and including
cluster merging and cluster splitting algorithms in the cluster maintenance scheme.
The OCM scheme achieves 13%, 14% and 23% improved query success rate (with
maximum density 150) and 26%, 30% and 32% (with maximum density 950) as
compared to the MoGLS, CBLS and HCBLS respectively.
100
95
Query Success Rate (%)
90
85
80
75
70
65
60
55
50
150 350 550 750 950
Maximun Density
157
6.4.2 Average Query Response Delay
The delay indicates the required time to process a source request and to respond
back to it. It is an important parameter to measure the performance of the location
service in vehicular networks. The main objective of the location service is to provide
the fresh locations of vehicles. The freshness of locations is compromised with the
increase of delay. Figure 6.7 and 6.8 show the delay comparison of RCHE, CBLS and
HCBLS.
The HCBLS scheme is based on static clustering, the rate of change of CH and
CMs increases with the increase in speed. The instability of the CH and CMs increases,
due to which delay also increases. In the CBLS, the instability of the CH and CMs
increases due to non-optimum limits of speed and distance, unreliable links of the CH
with the RSU and CMs. In the MoGLS scheme, the communication links between CH
and CMs and between CH and RSU breaks frequently due to the CH and CM
instability. With the increase in the speed, the instability of the CHs and CMs causes
more disconnections in the network which mean that more control messages
transmitted to again form the cluster again which affects the query response delay
negatively.
The proposed OCM scheme has shown better performance in terms of the
query response delay, in the increased mobility due to the enhanced stability of the
CH, CM and cluster maintenance schemes. The results represent that the OCM scheme
has less delay as compared to counterparts. The OCM scheme has reduced delay by
14%, 17% and 25% (with maximum speed 36 km/h) and 17%, 23% and 29% (with
maximum speed 108km/h) as compared to the MoGLS, CBLS and HCBLS
respectively as shown in Figure 6.7.
158
OCM MoGLS CBLS HCBLS
500
450
Query Response Delay(ms)
400
350
300
250
200
150
36 54 72 90 108
Maximun Speed(km/h)
Figure 6.8 shows the performance of the OCM, MoGLS, CBLS and HCBLS
schemes in terms of query response delay against the number of the vehicles in the
network. In the HCBLS, query response delay has improved as compared to speed.
This is so because of the stability of the CHs and CMs in high density. The election
and affiliation of the CHs and CMs does not involve too much control signalling.
However, the lifetime of the CHs and CMs is limited to length of the segment. In each
segment, the CMs affiliate to a new CH. Due to the instability of the CH and CMs, the
query response delay is affected. In the CBLS and MoGLS schemes, due to fixed CM
affiliation threshold values, a number of vehicles could not affiliate to any cluster.
These isolated vehicles further increase with the increase of the density. These
schemes are proposed without considering the reliable link with the RSU and cluster
maintenance schemes except the MoGLS which included the cluster merging
algorithm. Due to these issues, the cluster instability increases. The network
congestion and communication contention on the channels increases, which increases
the delay.
159
15%, 17% and 28% improved query response delay (with maximum density 150) and
25%, 28% and 29% (with maximum density 950) as compared to the MoGLS, CBLS
and HCBLS respectively.
500
450
400
350
300
250
200
150
150 350 550 750 950
Maximun Density
In the HCBLS, the location updates are affected due to the frequent change of
CHs and CMs. Also, the CH depends on the intermediate vehicles or CHs to update
the location information of the vehicles to the nearest RSU. With the increase in the
160
speed, the instability of these servers increases which delays the location updates and
ultimately the localization error increases. The CBLS relies on moving vehicles to play
the role of lowest level location server, the Intermediate Server (IS) and the Main
Servers (MS). With the increase in the speed, the communication links between these
servers break frequently which ultimately affects the location updates. The location
updates in the MoGLS scheme are affected due to instantaneous values of the speed
and unreliable links with the location servers. The instability of the CH and CM
increases at higher speed which increases the delay in the location updates and hence
increases localization error.
The proposed OCM scheme depends on reliable links with the CH and the
higher level location servers. The stability of the CH and CM is improved through
using CH election range, CMAS (Cumulative Moving Average Speed) and dynamic
threshold values. The performance of the location updates increases which reduces the
localization error. The OCM scheme achieves 20%, 27% and 46% improved
localization error (with maximum speed 36 km/h) and 33%, 40% and 50% (with
maximum speed 108km/h) as compared to the MoGLS, CBLS and HCBLS
respectively as given in Figure 6.9.
40
Average Localization Error(m)
35
30
25
20
15
10
5
0
36 54 72 90 108
Maximum Speed (km/h)
161
(b) Impact of density on Localization Error
Figure 6.10 shows the performance of the OCM, MoGLS, CBLS and HCBLS
schemes in terms of average LE (Localization Error) against the maximum number of
the vehicles in the network. The LE in the HCBLS scheme is improved as compared
to LE based on speed due to increase in the density in the network. The lifetime of the
CH and CM is increased in each segment due to obstacles of more vehicles. Due to the
stability of the CH and CMs, the performance of the location updates is increased
which ultimately reduce the localization error. However, due to limitation of the
segments size, the number of the CH increases; and also CMs switch CH frequently,
so the performance of the localization error is restricted. In the CBLS and MoGLS,
due to fixed threshold values, the number of isolated vehicles increases. These isolated
vehicles continuously send CM request to join a cluster. With the increase in density,
the network congestion and communication contention in the network increases. This
leads to losing location update messages and failing in providing updated traffic status
information. MoGLS shows better performance as compared to HCBLS and CBLS at
high density due to including merging algorithm.
In the proposed OCM scheme, due to reliable links of the CH with the CMs
and RSU, reduced isolated vehicles, optimized cluster merging and cluster splitting in
the cluster maintenance, the performance of the location updates is improved which
ultimately reduces the localization error. The OCM scheme achieves 25%, 35% and
43% improved localization error (with maximum 150 vehicles in the network) and
41%, 46% and 50% (with maximum speed 950 vehicles in the network) as compared
to the MoGLS, CBLS and HCBLS respectively.
162
OCM MoGLS CBLS HCBLS
40
6.5 Summary
163
CONCLUSION
7.1 Overview
The focus of this research was to develop ESCLS mechanism to enhance the
performance of the location service through improving cluster stability. The main
function of the location service is to maintain the location of vehicles. The CH stores
165
the locations of its CMs. The performance of the location service depends on the
stability of the CH, CM and optimum cluster maintenance schemes. The cluster
stability is improved by enhancing the CH lifetime, CM lifetime and maintaining the
cluster by optimizing the cluster splitting and cluster merging. This research has
focused on improving the cluster stability while keeping in view the inherited
characteristics of VANETs such as vehicles high mobility and limited communication
links lifetime. The CH lifetime is improved in the proposed RCHE scheme (chapter 4)
while considering the cluster formation range and link reliability with the RSU.
Moreover, this research has enhanced the CM lifetime by developing DCCMA scheme
(chapter 5) which depends on reliable link with the CH and dynamic CM affiliation
threshold values. In addition, after the cluster formation, the OCM scheme (chapter 6)
is developed to enhance the cluster stability. The subsections 7.2.1, 7.2.2 and 7.2.3
summarize the research contributions.
The RCHE scheme is proposed to address the research questions (i) as given
in section 1.4. Existing cluster-based location service schemes focused on location
update and location query schemes while overlooked the optimization of the cluster
stability. The number of vehicles participating in the CH election process is a very
important prerequisite of cluster formation. The Cluster Head Election Value (CHEV)
is affected due to undefined cluster formation range which leads to unreliable and
incomplete CHEV. Also, every CH, which is also acting as a lower-level location
server, updates the location information of its CMs to RSU. The location updates are
affected due to unreliable link between the CH and the RSU. Due to the above
mentioned issues, cluster instability increases which ultimately degrades the
performance of the location service. The main contribution of the RCHE scheme is the
optimization of the cluster formation range before the election of the CH and
considering a reliable link with the RSU. The CH lifetime is improved which
ultimately enhances the performance of the location service performance in terms of
Query Success Rate (QSR) and Query Response Delay (QRD).
166
7.2.2 Dynamic Cumulative Cluster Member Affiliation (DCCMA) Scheme
Once cluster is formed based on CH and CM, the next phase is the cluster
maintenance. Due to rapid mobility changes in VANETs, there is always a possibility
of overlapping of more than one clusters in neighbouring area. Also, around the
intersection, existing cluster splits into three clusters at the cost of clustering after
entering the intersection area and again reclustering after leaving the intersection area.
The cluster instability increases due to non-optimum cluster merging conditions,
unreliable links of merged and splitted CHs with the RSU and unnecessary clustering.
The OCM scheme is proposed to address the research question (iii) as given in section
1.4. The main contribution of the proposed scheme is to reduce the clustering around
167
the intersection by using destination information of each vehicle. The merging
conditions are optimized before the start of cluster merging. As each CH updates the
location information to RSU, so CH is elected based on the link status with RSU. The
stability of the cluster increases, which ultimately improves the performance of the
location service of QSR, QRD and Localization Error (LE).
The proposed OCM scheme improved the cluster stability by enhancing cluster
merging and cluster overlapping schemes. The performance of the cluster maintenance
depends on the stability of the CH and CM. In the existence of stable CHs and CMs,
OCM scheme further enhance the cluster stability by reducing contention and
congestion issues. The ratio of successful location updates and location queries
increased, which ultimately improved the query success rate, query response delay and
localization error.
168
On the basis of aforementioned discussion, it is concluded that RCHE Scheme
enhance the cluster stability by improving CH stability, DCCMA increase the cluster
stability by improving CM stability. The OCM further strengthen the cluster stability
by optimizing cluster merging and cluster splitting. The overall combine impact of
these scheme is the enhanced performance of the location service in terms of query
success rate, query response delay and localization error.
169
7.4 Directions for Future Work
Every vehicle that needs the location of another vehicle sends a query to its
CH. In VANETs, locations of the source and target vehicles change suddenly and
rapidly. Therefore, it is necessary to improve the existing location query schemes in
order to provide accurate and timely location services while considering the hierarchal
structure of the cluster-based location service.
Recently many studies have been proposed regarding the usage of cellular
technologies such as Long Term Evolution (LTE) and 5G in VANETs. In LTE, base
stations are designed and deployed while keeping the requirements of cellular
networks. The deployment of base stations as location servers in the context of the
location service, cluster formation challenges with reference to considering LTE
connection parameters and interfacing challenges of lowest-level servers with base
stations can be further explored.
In the recent years, researchers have explored the usage of Unmanned Aerial
Vehicles (UAVs) to assist VANETs in performing and monitoring different functions.
A UAV may play the role of a moving location server by storing the current positions
of the vehicles. The performance of the location service increases with the usage of
mobile location servers that can provide wide coverage. However, there are yet
numerous challenges such as interfacing, communications, connectivity, mobility in
the implementation and integration of UAV-VANETs architecture.
170
To meet the demanding requirements of the future ITS, cloud computing and
fog Computing are expected to be future candidate technologies for 5G VANETs. Fog
computing may play the role of lower-level servers to cater the location service
requirements towards the edge of the VANETs. Integration of the fog computing
towards the edge of network will improve the performance of the location service by
providing real time location service through minimizing location updates and location
query delay. Cloud servers can be used as a regional location server to store the
locations of vehicles. There is a need to consider different challenges with reference
to scheduling of location updates from lower level servers, location query responses,
accuracy and privacy of locations saved on location servers.
171
REFERENCES
Abbas, G., Abbas, Z. H., Haider, S., Baker, T., Boudjit, S. and Muhammad, F. (2020).
PDMAC: A Priority-based Enhanced TDMA Protocol for Warning Message
Dissemination in VANETs. Sensors, 20(1), 1-45.
Abboud, K. and Zhuang, W. (2015). Impact of Microscopic Vehicle Mobility on
Cluster-based Routing Overhead in VANETs. IEEE Transactions on Vehicular
Technology, 64(12), 5493-5502.
Abdel-Halim, I. T., Fahmy, H. M. A. and Bahaa-El Din, A. M. (2019). Mobility
Prediction-based Efficient Clustering Scheme for Connected and Automated
Vehicles in VANETs. Computer Networks, 150, 217-233.
Abuashour, A. and Kadoch, M. (2016). A Cluster-based Life-Time Routing Protocol
in VANET. In Proceedings of the 4th International Conference On Future
Internet of Things and Cloud (FICLOUD), IEEE, 213-219.
Abuashour, A. and Kadoch, M. (2018). Control Overhead Reduction in Cluster-based
VANET Routing Protocol. Ad Hoc Networks, 223, 106-115.
Ağgün, F., Çibuk, M. and Ur-Rehman, S. (2020). A New Self-Organizing Multi-
Channel MAC Schema for RSU-Centric VANETs. Physica A: Statistical
Mechanics and its Applications, 551, 1-15.
Ahmad, I., Noor, R. M., Ahmedy, I., Shah, S. A. A., Yaqoob, I., Ahmed, E. and Imran,
M. (2018). VANET–LTE based Heterogeneous Vehicular Clustering for
Driving Assistance and Route Planning Applications. Computer Networks,
145, 128-140.
Ahwazi, A. Z. and Noorimehr, M. (2016). MOSIC: Mobility-aware Single-hop
Clustering Scheme for Vehicular Ad Hoc Networks on Highways.
International Journal of Advanced Computer Science and Applications, 7, 424-
431.
Aissaou, R., Dhraief, A., Belghith, A., Menouar, H., Filali, F. and Mathkour, H.
(2016). VALS: Vehicle-Aided Location Service in Urban Environment. In
Proceedings of the IEEE Wireless Communications and Networking
Conference on, IEEE, 1-6.
173
Aissaoui, R., Dhraief, A., Belghith, A., Menouar, H., Mathkour, H., Filali, F. and Abu-
Dayya, A. (2015). HCBLS: A Hierarchical Cluster-based Location Service in
Urban Environment. Mobile Information Systems, 2015, 1-16.
Aissaoui, R., Menouar, H., Dhraief, A., Filali, F., Belghith, A. and Abu-Dayya, A.
(2014). Advanced Real-Time Traffic Monitoring System based 0n V2X
Communications. In Proceedings of the International Conference on
Communications (ICC), IEEE, 2713-2718.
Al-Mayouf, Y. R. B., Abdullah, N. F., Ismail, M., Al-Qaraawi, S. M., Mahdi, O. A.
and Khan, S. (2016). Evaluation of Efficient Vehicular Ad Hoc Networks
based on a Maximum Distance Routing Algorithm. EURASIP Journal on
Wireless Communications and Networking, 265(2016), 1-11.
Al-Mayouf, Y. R. B., Mahdi, O. A., Taha, N. A., Abdullah, N. F., Khan, S. and Alam,
M. (2018). Accident Management System based on Vehicular Network for an
Intelligent Transportation System in Urban Environments. Journal of
Advanced Transportation, 2018, 1-11.
Al-Otaibi, B., Al-Nabhan, N. and Tian, Y. (2019). Privacy-Preserving Vehicular
Rogue Node Detection Scheme for Fog Computing. Sensors, 19, 965.
Ali, G. M. N., Chong, P. H. J., Samantha, S. K. and Chan, E. (2016). Efficient Data
Dissemination in Cooperative Multi-RSU Vehicular Ad Hoc Networks
(VANETs). Journal of Systems and Software, 117, 508-527.
Ali, M., Malik, A. W., Rahman, A. U., Iqbal, S. and Hamayun, M. M. (2019). Position-
based Emergency Message Dissemination for Internet Of Vehicles.
International Journal of Distributed Sensor Networks, 15(7), 1-12.
Almalag, M. S., Olariu, S. and Weigle, M. C. (2012). TDMA Cluster-based Mac for
VANETs (Tc-Mac). In Proceedings of the IEEE International Symposium on
a World of Wireless, Mobile and Multimedia Networks (WoWMoM), IEEE, 1-
6.
Alsuhli, G. H., Khattab, A. and Fahmy, Y. A. (2019). Double-Head Clustering for
Resilient Vanets. Wireless Communications and Mobile Computing, 2019, 1-
17.
Aravindhan, K. and Dhas, C. S. G. (2019). Destination-aware Context-based Routing
Protocol with Hybrid Soft Computing Cluster Algorithm for VANET. Soft
Computing, 23, 2499-2507.
174
Arif, M., Wang, G., Bhuiyan, M. Z. A., Wang, T. and Chen, J. (2019). A Survey On
Security Attacks in Vanets: Communication, Applications and Challenges.
Vehicular Communications, 19, 1-36.
Arkian, H. R., Atani, R. E., Pourkhalili, A. and Kamali, S. (2015). A Stable Clustering
Scheme based on Adaptive Multiple Metric in Vehicular Ad-Hoc Networks.
Journal of Information Science and Engineering, 31, 361-386.
Ashok, D., Pai, M. M. and Mouzna, J. (2011). Efficient Map based Location Service
for VANETs. In Proceedings of the 11th International Conference on ITS
Telecommunications, IEEE, 387-392.
Asoudeh, S., Mehrjoo, M., Balouchzahi, N.-M. and Bejarzahi, A. (2017). Location
Service Implementation in Vehicular Networks by Nodes Clustering in Urban
Environments. Vehicular Communications, 9, 109-114.
Awan, K. A., Din, I. U., Almogren, A., Guizani, M. and Khan, S. (2020). StabTrust A
Stable and Centralized Trust-based Clustering Mechanism For IoT Enabled
Vehicular Ad-Hoc Networks. IEEE Access, 8, 21159-21177.
Ayaida, M., Barhoumi, M., Fouchal, H., Ghamri-Doudane, Y. and Afilal, L. (2014).
Joint Routing and Location-based Service in VANETs. Journal of Parallel and
Distributed Computing, 74, 2077-2087.
Ayaida, M., Fouchal, H., Afilal, L. and Ghamri-Doudane, Y. (2012). A Comparison
of Reactive, Grid and Hierarchical Location-based Services for VANETs. In
Proceedings of the Vehicular Technology Conference (VTC Fall), IEEE, 1-5.
Azizian, M., Cherkaoui, S. and Hafid, A. S. (2016). A Distributed D-Hop Cluster
Formation for VANET. In Proceedings of the Wireless Communications and
Networking Conference, IEEE, 1-6.
Bali, R. S. and Kumar, N. (2016). Secure Clustering for Efficient Data Dissemination
in Vehicular Cyber–Physical Systems. Future Generation Computer Systems,
56, 476-492.
Bali, R. S., Kumar, N. and Rodrigues, J. J. (2014). Clustering in Vehicular Ad Hoc
Networks: Taxonomy, Challenges and Solutions. Vehicular Communications,
1, 134-152.
Balouchzahi, N.-M., Fathy, M. and Akbari, A. (2016). An Efficient Infrastructure
based Service Discovery in Vehicular Networks Using P2P Structures. The
Journal of Supercomputing, 72, 1013-1034.
175
Benkerdagh, S. & Duvallet, C. 2019. Cluster‐based Emergency Message
Dissemination Strategy for VANET using V2V Communication. International
Journal of Communication Systems, 32(5).
Bi, X., Guo, B., Shi, L., Lu, Y., Feng, L. and Lyu, Z. (2020). A New Affinity
Propagation Clustering Algorithm for V2V-Supported VANETs. IEEE Access,
8, 71405-71421.
Bilal, S. M. and Ali, S. 2017. Review and Performance Analysis of Position based
Routing in VANETs. Wireless Personal Communications, 94, 559-578.
Boban, M., Misek, G. and Tonguz, O. K. (2008). What is the Best Achievable QoS for
Unicast Routing in VANETs? In Proceedings of the Globecom Workshops,
IEEE, 1-10.
Boumerdassi, S. and Renault, E. (2016). A Flooding-based Solution to Improve
Location Services in VANETs. In Proceedings of the International Conference
on Communications (ICC), 2016. 1-6.
Boussedjra, M., Mouzna, J., Bangera, P. and Pai, M. M. (2009). Map-based Location
Service for VANET. In Proceedings of the International Conference on Ultra
Modern Telecommunications and Workshops, IEEE, 1-6.
Boussoufa-Lahlah, S., Semchedine, F. and Bouallouche-Medjkoune, L. (2018).
Geographic Routing Protocols for Vehicular Ad Hoc NETworks (VANETs):
A Survey. Vehicular Communications, 11, 20-31.
Brahmi, N., Boussedjra, M., Mouzna, J., Cornelio, A. K. and Manohara, M. (2010).
An Improved Map-based Location Service for Vehicular Ad Hoc Networks. In
Proceedings of the 6th International Conference on Wireless and Mobile
Computing, Networking and Communications, IEEE, 21-26.
Bylykbashi, K., Elmazi, D., Matsuo, K., Ikeda, M. and Barolli, L. (2019). Effect of
Security and Trustworthiness for A Fuzzy Cluster Management System in
VANETs. Cognitive Systems Research, 55, 153-163.
Cambruzzi, E., Farines, J.-M., Kraus, W. and Macêdo, R. (2016). A Cluster
Management System for VANETs. International Journal of Intelligent
Transportation Systems Research, 14, 115-126.
Chaurasia, B. K., Alam, M. I., Prakash, A., Tomar, R. S. and Verma, S. (2019).
MPMAC: Clustering based MAC protocol for VANETs. Wireless Personal
Communications, 108, 409-436.
176
Chen, Y., Fang, M., Shi, S., Guo, W. and Zheng, X. (2015). Distributed Multi-Hop
Clustering Algorithm for Vanets based On Neighborhood Follow. Eurasip
Journal on Wireless Communications and Networking, 98(2015), 1-12.
Cheng, X. and Huang, B. (2019). A Center-based Secure and Stable Clustering
Algorithm for VANETs On Highways. Wireless Communications and Mobile
Computing, 2019, 1-10.
Chou, L.-D., Yang, J.-Y., Hsieh, Y.-C., Chang, D.-C. and Tung, C.-F. (2011).
Intersection-based Routing Protocol For VANETs. Wireless Personal
Communications, 60, 105-124.
Cooper, C., Franklin, D., Ros, M., Safaei, F. & Abolhasan, M. 2016. A Comparative
Survey of VANET Clustering Techniques. IEEE Communications Surveys &
Tutorials, 19, 657-681.
Daeinabi, A. and Rahbar, A. G. (2014). An Advanced Security Scheme based on
Clustering and Key Distribution in Vehicular Ad-Hoc Networks. Computers
and Electrical Engineering, 40, 517-529.
Daoud, O., Hamarsheh, Q. and Damati, A. (2020). Enhancing the LTE-based
Intelligent Transportation System's Performance. Wireless Personal
Communications, 2020, 1-19.
Darwish, T. and Bakar, K. A. (2015). Traffic Density Estimation in Vehicular Ad Hoc
Networks: A Review. Ad Hoc Networks, 24, 337-351.
Darwish, T. and Bakar, K. A. (2016). Lightweight Intersection-based Traffic Aware
Routing in Urban Vehicular Networks. Computer Communications, 87, 60-75.
Dhugga, P. K., Sharma, M. and Sharma, A. (2015). An Algorithm for Static
Geographical Clustering in VANET. In Proceedings of the 3rd International
Conference on MOOCs, Innovation and Technology in Education (MITE),
IEEE, 420-426.
Dua, A., Kumar, N. and Bawa, S. (2017). ReIDD: Reliability-aware Intelligent Data
Dissemination Protocol for Broadcast Storm Problem in Vehicular Ad Hoc
Networks. Telecommunication Systems, 64, 439-458.
Fan, P., Haran, J. G., Dillenburg, J. and Nelson, P. C. (2005). Cluster-based
Framework in Vehicular Ad-Hoc Networks. In Proceedings of the
International Conference on Ad-Hoc Networks and Wireless, Springer, 32-42.
177
Farooq, W., Ali Khan, M. and Rehman, S. (2016). A Novel Real Time Framework for
Cluster based Multicast Communication in Vehicular Ad Hoc Networks.
International Journal of Distributed Sensor Networks, 2016, 1-18.
Fatemidokht, H. and Rafsanjani, M. K. (2020). QMM-VANET: An Efficient
Clustering Algorithm based on QoS and Monitoring of Malicious Vehicles in
Vehicular Ad Hoc Networks. Journal of Systems and Software, 165, 1-16.
Ganeshkumar, P. and Gokulakrishnan, P. (2015). Emergency Situation Prediction
Mechanism: A Novel Approach for Intelligent Transportation System using
Vehicular Ad Hoc Networks. The Scientific World Journal, 2015, 1-12.
Garg, A., Pandey, K. and Singh, (2014). B. Hierarchical Map-based Location Service
for Vanets in Urban Environments. In Proceedings of the Seventh
International Conference on Contemporary Computing (IC3), 2014. IEEE,
199-205.
Ghandour, A. J., Di Felice, M., Artail, H. and Bononi, L. (2014). Dissemination of
Safety Messages in IEEE 802.11 P/WAVE Vehicular Network: Analytical
Study and Protocol Enhancements. Pervasive and Mobile Computing, 11, 3-
18.
Ghazi, M. U., Khattak, M. A. K., Shabir, B., Malik, A. W. and Ramzan, M. S. (2020).
Emergency Message Dissemination in Vehicular Networks: A Review. IEEE
Access, 8, 38606-38621.
Ghodrati, A. D. (2013). Reduces Broadcast Storm using Clustering in VANETs.
International Research Journal of Applied and Basic Sciences, 7, 979-987.
Guo, M. H., Liaw, H. T., Chiu, M. Y. and Deng, D. J. (2016). On Decentralized Group
Key Management Mechanism for Vehicular Ad Hoc Networks. Security and
Communication Networks, 9, 241-247.
Gupta, N., Prakash, A. and Tripathi, R. (2016). Clustering based Cognitive MAC
Protocol for Channel Allocation to Prioritize Safety Message Dissemination in
Vehicular Ad-Hoc Network. Vehicular Communications, 5, 44-54.
Gupta, N., Prakash, A. and Tripathi, R. (2017). Adaptive Beaconing in Mobility Aware
Clustering based MAC Protocol for Safety Message Dissemination in VANET.
Wireless Communications and Mobile Computing, 2017, 1-15.
Hafeez, K. A., Zhao, L., Liao, Z. and Ma, B. N.-W. (2011). Clustering and OFDMA-
based MAC Protocol (COMAC) for Vehicular Ad Hoc Networks. EURASIP
Journal on Wireless Communications and Networking, 117(211), 1-16.
178
Hafeez, K. A., Zhao, L., Mark, J. W., Shen, X. and Niu, Z. (2013). Distributed
Multichannel and Mobility-Aware Cluster-based MAC Protocol for Vehicular
Ad Hoc Networks. IEEE Transactions on Vehicular Technology, 62, 3886-
3902.
Haider, S., Abbas, G., Abbas, Z. H. and Baker, T. (2019). DABFS: A Robust Routing
Protocol for Warning Messages Dissemination in VANETs. Computer
Communications, 147, 21-34.
Haider, S., Abbas, G., Abbas, Z. H., Boudjit, S. and Halim, Z. (2020). P-DACCA: A
Probabilistic Direction-Aware Cooperative Collision Avoidance Scheme for
VANETs. Future Generation Computer Systems, 103, 1-17.
Hajlaoui, R., Alsolami, E., Moulahi, T. and Guyennet, H. (2019b). Construction of A
Stable Vehicular Ad Hoc Network based on Hybrid Genetic Algorithm.
Telecommunication Systems, 71, 433-445.
Hoeft, M. and Rak, J. (2016). How to Provide Fair Service for V2I Communications
in VANETs? Ad Hoc Networks, 37, 283-294.
Hsu, C.-S. and Wu, S.-W. (2017). An Efficient Cost-based Location Service Protocol
for Vehicular Ad Hoc Networks. In Proceedings of the International
Conference on Communication, Networks and Satellite (ComNetSat), IEEE,
93-97.
Huang, H. and Zhang, S. (2013). A Routing Algorithm based on Dynamic Forecast of
Vehicle Speed and Position in VANET. International Journal of Distributed
Sensor Networks, 2013, 1-9.
Huo, Y., Liu, Y., Ma, L., Cheng, X. and Jing, T. 2016. An Enhanced Low Overhead
and Stable Clustering Scheme for Crossroads in VANETs. EURASIP Journal
on Wireless Communications and Networking, 2016, 1-13.
Jerbi, M., Senouci, S.-M., Rasheed, T. and Ghamri-Doudane, Y. (2009). Towards
Efficient Geographic Routing in Urban Vehicular Networks. IEEE
transactions on vehicular technology, 58, 5048-5059.
Ji, X., Yu, H., Fan, G., Sun, H. and Chen, L. (2018). Efficient and Reliable Cluster-
based Data Transmission for Vehicular Ad Hoc Networks. Mobile Information
Systems, 2018, 1-15.
Joshua, C. J., Duraisamy, R. and Varadarajan, V. (2019). A Reputation based
Weighted Clustering Protocol in VANET: A Multi-objective Firefly Approach.
Mobile Networks and Applications, 24, 1199-1209.
179
Kchaou, A., Abassi, R. and Guemara, S. (2018). Towards A Secured Clustering
Mechanism for Messages Exchange in VANET. In Proceedings of the 32nd
International Conference on Advanced Information Networking and
Applications Workshops (WAINA), IEEE, 88-93.
Khan, A. A., Abolhasan, M. and Ni, W. (2018). An Evolutionary Game Theoretic
Approach for Stable and Optimized Clustering in VANETs. IEEE
Transactions on Vehicular Technology, 67, 4501-4513.
Latif, S., Mahfooz, S., Jan, B., Ahmad, N., Cao, Y. and Asif, M. (2018). A
Comparative Study of Scenario-driven Multi-Hop Broadcast Protocols for
VANETs. Vehicular Communications, 12, 88-109.
Lee, E., Choe, H., Thirumurthi, P., Gerla, M. and Kim, S.-H. (2013). Quorum-based
Location Service in Vehicular Sensor Networks. In Proceedings of the 9th
International Wireless Communications and Mobile Computing Conference
(IWCMC), IEEE, 1744-1749.
Lin, D., Kang, J., Squicciarini, A., Wu, Y., Gurung, S. and Tonguz, O. (2016). MoZo:
A Moving Zone based Routing Protocol using Pure V2V Communication in
VANETs. IEEE Transactions on Mobile Computing, 16, 1357-1370.
Liu, J., Wan, J., Wang, Q., Deng, P., Zhou, K. and Qiao, Y. (2016). A Survey on
Position-based Routing for Vehicular Ad Hoc Networks. Telecommunication
Systems, 62, 15-30.
Liu, W., Qin, G., He, Y. and Jiang, F. (2017). Distributed Cooperative Reinforcement
Learning-based Traffic Signal Control That Integrates V2X Networks’
Dynamic Clustering. IEEE Transactions on Vehicular Technology, 66, 8667-
8681.
Lu, R., Lin, X., Liang, X. and Shen, X. (2011). A Dynamic Privacy-Preserving Key
Management Scheme for Location-based Services in VANETs. IEEE
Transactions on Intelligent Transportation Systems, 13, 127-139.
Manvi, S. S. and Tangade, S. (2017). A Survey on Authentication Schemes in
VANETs for Secured Communication. Vehicular Communications, 9, 19-30.
Martinez, F. J., Toh, C. K., Cano, J. C., Calafate, C. T. and Manzoni, P. (2011). A
Survey and Comparative Study of Simulators for Vehicular Ad Hoc Networks
(VANETs). Wireless Communications and Mobile Computing, 11, 813-828.
180
Mchergui, A., Moulahi, T., Othman, M. T. B. and Nasri, S. (2020). Enhancing
VANETs Broadcasting Performance with Mobility Prediction for Smart Road.
Wireless Personal Communications, 2020, 1629-1641.
Mehmood, A., Khanan, A., Mohamed, A. H. H., Mahfooz, S., Song, H. and Abdullah,
S. (2017). ANTSC: An Intelligent Naïve Bayesian Probabilistic Estimation
Practice for Traffic Flow to Form Stable Clustering in VANET. IEEE Access,
6, 4452-4461.
Mehra, R., Bali, R. S. and Kaur, (2016). Efficient Clustering based OLSR Routing
Protocol for VANET. In Proceedings of the Colossal Data Analysis and
Networking (CDAN) Symposium on, IEEE, 1-7.
Mo, Z., Zho, H., Makki, K. and Pissinou, N. (2008). Mobility-assisted Location
Management for Vehicular Ad Hoc Networks. In Proceedings of the Wireless
Communications and Networking Conference, IEEE, 2224-2228.
Mohammed Nasr, M. M., Abdelgader, A. M. S., Wang, Z.-G. and Shen, L.-F. (2016).
VANET Clustering based Routing Protocol Suitable for Deserts. Sensors,
16(4), 1-23.
Moridi, E. and Barati, H. (2017). RMRPTS: A Reliable Multi-Level Routing Protocol
with Tabu Search in VANET. Telecommunication Systems, 65, 127-137.
Mühlethaler, P., Renault, E. and Boumerdassi, S. (2020). Design and Evaluation of
Flooding-based Location Service in Vehicular Ad Hoc Networks. Sensors,
20(8), 1-16.
Nebbou, T., Lehsaini, M., Fouchal, H. and Ayaida, M. (2019). An Urban Location
Service for Vehicular Area Networks. Concurrency and Computation:
Practice and Experience, 31(24).
Pal, R., Gupta, N., Prakash, A. and Tripathi, R. (2018). Adaptive Mobility and Range
based Clustering Dependent MAC Protocol for Vehicular Ad Hoc Networks.
Wireless Personal Communications, 98, 1155-1170.
Pan, J., Cui, J., Wei, L., Xu, Y. and Zhong, H. (2019). Secure Data Sharing Scheme
for VANETs based on Edge Computing. EURASIP Journal on Wireless
Communications and Networking, 2019, 1-11.
Paul, B., Ibrahim, M., Bikas, M. and Naser, A. (2012). VANET Routing Protocols:
Pros and Cons. International Journal of Computer Applications, 20(3) 28-34.
Qi, W., Song, Q., Wang, X., Guo, L. and Ning, Z. (2018). SDN-enabled Social-Aware
Clustering in 5G-VANET Systems. IEEE Access, 6, 28213-28224.
181
Qureshi, K. N., Abdullah, A. H., Bashir, F., Iqbal, S. and Awan, K. M. (2018). Cluster‐
based Data Dissemination, Cluster Head Formation Under Sparse, and Dense
Traffic Conditions for Vehicular Ad Hoc Networks. International Journal of
Communication Systems, 31(8), 1-16.
Ramakrishnan, B., Selvi, M., Nishanth, R. B. and Joe, M. M. (2017). An Emergency
Message Broadcasting Technique using Transmission Power based Clustering
Algorithm for Vehicular Ad Hoc Network. Wireless Personal
Communications, 94, 3197-3216.
Raw, R. S. and Das, S. (2011). Performance Comparison of Position based Routing
Protocols in Vehicle-To-Vehicle (V2V) Communication. International
Journal of Engineering Science and Technology, 3, 435-444.
Regin, R. and Menakadevi, T. (2019). Dynamic Clustering Mechanism to Avoid
Congestion Control in Vehicular Ad Hoc Networks based on Node Density.
Wireless Personal Communications, 107, 1911-1931.
Rehan, M., Hasbullah, H., Faye, I., Rehan, W., Chughtai, O. and Rehmani, M. H.
(2018). ZGLS: A Novel Flat Quorum-based and Reliable Location
Management Protocol for VANETs. Wireless Networks, 24, 1885-1903.
Ren, M., Khoukhi, L., Labiod, H., Zhang, J. and Veque, V. (2017a). A Mobility-based
Scheme for Dynamic Clustering in Vehicular Ad-Hoc Networks (VANETs).
Vehicular Communications, 9, 233-241.
Ren, M., Zhang, J., Khoukhi, L., Labiod, H. and Vèque, V. (2017b). A Unified
Framework of Clustering Approach in Vehicular Ad Hoc Networks. IEEE
Transactions on Intelligent Transportation Systems, 19, 1401-1414.
Ros, F. J., Martinez, J. A. and Ruiz, P. M. (2014). A Survey on Modeling and
Simulation of Vehicular Networks: Communications, Mobility, and Tools.
Computer Communications, 43, 1-15.
Rossi, G. V., Fan, Z., Chin, W. H. and Leung, K. K. (2017). Stable Clustering for Ad-
Hoc Vehicle Networking. In Proceedings of the IEEE Wireless
Communications and Networking Conference (WCNC), IEEE, 1-6.
Sadou, M. and Bouallouche-Medjkoune, L. (2017). Efficient Message Delivery in
Hybrid Sensor and Vehicular Networks based on Mathematical Linear
Programming. Computers and Electrical Engineering, 64, 496-505.
182
Sandonis, V., Soto, I., Calderon, M. and Urueña, M. (2016). Vehicle to Internet
Communications using the ETSI ITS GeoNetworking Protocol. Transactions
on Emerging Telecommunications Technologies, 27, 373-391.
Sanguesa, J. A., Fogue, M., Garrido, P., Martinez, F. J., Cano, J.-C. and Calafate, C.
T. (2016). A Survey and Comparative Study of Broadcast Warning Message
Dissemination Schemes for VANETs. Mobile Information Systems, 2016, 1-
18.
Santamaria, A. F., Tropea, M., Fazio, P. and De Rango, F. (2018). Managing
Emergency Situations in VANET through Heterogeneous Technologies
Cooperation. Sensors, 18(5), 1-24.
Satyajeet, D., Deshmukh, A. and Dorle, S. (2016). Heterogeneous Approaches for
Cluster based Routing Protocol in Vehicular Ad Hoc Network (VANET).
International Journal of Computer Applications, 134, 1-8.
Schoch, E., Kargl, F., Weber, M. and Leinmuller, T. (2008). Communication Patterns
in VANETs. IEEE Communications Magazine, 46, 119-125.
Senouci, O., Aliouat, Z. and Harous, S. (2019a). MCA-V2I: A Multi-hop Clustering
Approach over Vehicle-to-Internet Communication for Improving VANETs
Performances. Future Generation Computer Systems, 96, 309-323.
Senouci, O., Harous, S. and Aliouat, Z. (2019b). A New Heuristic Clustering
Algorithm based on RSU for Internet of Vehicles. Arabian Journal for Science
and Engineering, 44, 9735-9753.
Shafi, S. and Ratnam, D. V. (2019). A Cross Layer Cluster based Routing Approach
for Efficient Multimedia Data Dissemination with Improved Reliability in
VANETs. Wireless Personal Communications, 107, 2173-2190.
Shah, S. S., Malik, A. W., Rahman, A. U., Iqbal, S. and Khan, S. U. (2019). Time
Barrier-based Emergency Message Dissemination in Vehicular Ad-Hoc
Networks. IEEE Access, 7, 16494-16503.
Shah, Y. A., Habib, H. A., Aadil, F., Khan, M. F., Maqsood, M. & Nawaz, T. (2018).
CAMONET: Moth-Flame Optimization (MFO) based Clustering Algorithm
for VANETs. IEEE Access, 6, 48611-48624.
Sharma, S. and Kaul, A. (2018). A Survey on Intrusion Detection Systems and
Honeypot based Proactive Security Mechanisms in VANETs and VANET
Cloud. Vehicular Communications, 12, 138-164.
183
Sheikh, M. S. and Liang, J. (2019). A Comprehensive Survey on VANET Security
Services in Traffic Management System. Wireless Communications and
Mobile Computing, 2019, 1-23.
Singh, A. and Kaur, M. (2015). A Novel Clustering Scheme in Vehicular Ad Hoc
Network. International Journal of Applied Information Systems (IJAIS), 10(3),
1-5.
Singh, J. P. and Bali, R. S. (2015). A Hybrid Backbone based Clustering Algorithm
for Vehicular Ad-Hoc Networks. Procedia Computer Science, 46, 1005-1013.
Srivastava, A., Prakash, A. and Tripathi, R. (2020). Location based Routing Protocols
in VANET: Issues and Existing Solutions. Vehicular Communications,
23(2020), 1-30.
Stojmenovic, I., Liu, D. and Jia, X. (2008). A Scalable Quorum-based Location
Service in Ad Hoc And Sensor Networks. International Journal of
Communication Networks and Distributed Systems, 1, 71-94.
Sugumar, R., Rengarajan, A. and Jayakumar, C. (2018). Trust based Authentication
Technique for Cluster based Vehicular Ad Hoc Networks (VANET). Wireless
Networks, 24, 373-382.
Toor, Y., Muhlethaler, P., Laouiti, A. and De La Fortelle, A. (2008). Vehicle Ad Hoc
Networks: Applications and Related Technical Issues. IEEE Communications
Surveys and Tutorials, 10, 74-88.
Ucar, S., Ergen, S. C. and Ozkasap, O. (2013). VMaSC: Vehicular Multi-Hop
Algorithm for Stable Clustering in Vehicular Ad Hoc Networks. In
Proceedings of the Wireless Communications and Networking Conference
(WCNC), IEEE, 2381-2386.
Ucar, S., Ergen, S. C. and Ozkasap, O. (2015). Multihop-cluster-based IEEE 802.11P
and LTE Hybrid Architecture for VANET Safety Message Dissemination.
IEEE Transactions on Vehicular Technology, 65, 2621-2636.
Ullah, A., Yaqoob, S., Imran, M. and Ning, H. (2018). Emergency Message
Dissemination Schemes based on Congestion Avoidance in VANET and
Vehicular FoG Computing. IEEE Access, 7, 1570-1585.
Usha, M. and Ramakrishnan, B. (2020). Robust MPR: A Novel Algorithm for Secure
and Efficient Data Transmission in VANET. Wireless Personal
Communications, 110, 355-380.
184
Van Gulik, K., Phillipson, F. and Fouchal, H. (2018). Congestion Control in a Location
Service for VANETs. In Proceedings of the International Conference on
Mobile, Secure, and Programmable Networking, Springer, 289-293.
Vegni, A. M. and Loscri, V. (2015). A Survey on Vehicular Social Networks. IEEE
Communications Surveys and Tutorials, 17, 2397-2419.
Wahab, O. A., Mourad, A., Otrok, H. and Bentahar, J. (2016). CEAP: SVM-based
Intelligent Detection Model for Clustered Vehicular Ad Hoc Networks. Expert
Systems with Applications, 50, 40-54.
Woo, H. and Lee, M. (2011). Mobile Group based Location Service Management for
Vehicular Ad-Hoc Networks. In Proceedings of the IEEE International
Conference on Communications (ICC), IEEE, 1-6.
Woo, H. and Lee, M. (2018). A Hierarchical Location Service Architecture for
VANET with Aggregated Location Update. Computer Communications, 125,
38-55.
Wu, C., Ohzahata, S., JI, Y. and Kato, T. (2014). Toward a Totally Distributed Flat
Location Service for Vehicular Ad Hoc Networks. In Proceedings of the 79th
Vehicular Technology Conference (VTC Spring), IEEE, 1-6.
Xie, Y., Wu, L., SHEN, J. and Alelaiwi, A. (2017). EIAS-CP: New Efficient Identity-
based Authentication Scheme with Conditional Privacy-Preserving for
VANETs. Telecommunication Systems, 65, 229-240.
Yang, F., Tang, Y. and Huang, L. (2013). A Novel Cooperative MAC for Broadcasting
in Clustering VANETs. In Proceedings of the International Conference on
Connected Vehicles and Expo (ICCVE), IEEE, 893-897.
Yim, Y., Cho, H., Kim, S.-H., Lee, E. and Gerla, M. (2017). Vehicle Location Service
Scheme based On Road Map in Vehicular Sensor Networks. Computer
Networks, 127, 138-150.
Younes, M. B. and Boukerche, A. (2015). A Performance Evaluation of an Efficient
Traffic Congestion Detection Protocol (ECODE) for Intelligent Transportation
Systems. Ad Hoc Networks, 24, 317-336.
Zeadally, S., Hunt, R., Chen, Y.-S., Irwin, A. and Hassan, A. (2012). Vehicular Ad
Hoc Networks (VANETs): Status, Results, and Challenges.
Telecommunication Systems, 50, 217-241.
185
Zemouri, S., Djahel, S. and Murphy, J. (2015). A Fast, Reliable and Lightweight
Distributed Dissemination Protocol for Safety Messages in Urban Vehicular
Networks. Ad Hoc Networks, 27, 26-43.
Zhang, D., Ge, H., Zhang, T., Cui, Y.-Y., Liu, X. and Mao, G. (2018). New Multi-Hop
Clustering Algorithm for Vehicular Ad Hoc Networks. IEEE Transactions on
Intelligent Transportation Systems, 20, 1517-1530.
Zhang, L., Liu, Z., Zou, R., Guo, J. and Liu, Y. (2014). A Scalable CSMA and Self-
Organizing TDMA MAC for IEEE 802.11 P/1609. X in VANETs. Wireless
Personal Communications, 74, 1197-1212.
Zhang, W., Zheng, R., Zhang, M., Zhu, J. and Wu, Q. (2019). ECRA: An Encounter-
aware and Clustering-based Routing Algorithm for Information-centric
VANETs. Mobile Networks and Applications, 25(2020), 632-642.
Zhao, H., Liu, J., Wu, J. and Liu, W. (2016). An Intersection-based Clustering
Algorithm for Vehicular Ad Hoc Networks. In Proceedings of the 8th
International Conference on Wireless Communications and Signal Processing
(WCSP), IEEE, 1-5.
Zhou, Y., Wu, X. and Wang, P. (2017). A Novel Vehicle Clustering Algorithm in
Intersection Scenario. In Proceedings of the 9th International Conference on
Advanced Infocomm Technology (ICAIT), IEEE, 431-437.
186
APPENDIX A
Sample Results
Table A.1 RCHE CH election results with varying speed and density 620
Table A.2 RCHE CH election results with varying density and speed 40km/h
Table A.3 CBLS CH election results with varying speed and density 620
Table A.4 CBLS CH election results with varying density and speed 40km/h
187
Table A.5 HCBLS CH election results with varying speed and density 620
Table A.6 HCBLS CH election results with varying density and speed 40km/h
Table A.7 DCCMA CM affiliation results with varying speed and density 620
Table A.8 DCCMA CM affiliation results with varying density and speed 40km/h
188
Table A.9 MoGLS CM affiliation results with varying speed and density 620
Table A.10 MoGLS CM affiliation results with varying density and speed 40km/h
Table A.11 CBLS CM affiliation results with varying speed and density 620
Table A.12 CBLS CM affiliation results with varying density and speed 40km/h
189
Table A.13 OCM Cluster maintenance results with varying speed and density 620
Table A.14 OCM Cluster maintenance results with varying density and speed 40
Table A.15 MoGLS Cluster maintenance results with varying speed and density 620
Table A.16 MoGLS Cluster maintenance results with varying density and speed 40
190
Table A.17 CBLS Cluster maintenance results with varying speed and density 620
Table A.18 CBLS Cluster maintenance results with varying density and speed 40
Table A.19 HCBLS Cluster maintenance results with varying speed and density 620
Table A.20 HCBLS Cluster maintenance results with varying density and speed 40
191
LIST OF PUBLICATIONS
193