Optimized Communication in 5G-Driven

A Dissertation submitted in fulfilment of the
requirements for the Degree of Doctor of Philosophy
Optimized Communication in 5G-Driven
Vehicular Ad-hoc Networks (VANETs)
Ammara Anjum khan
Faculty of Engineering and Information Technology (FEIT),
School of Electrical and Data Engineering (SEDE),
University of Technology Sydney
2019

Supervisor
A/Prof. Mehran Abolhasan
Co-Supervisor
A/Prof. Justin Lipman

Declaration Of Authorship
I, Ammara Anjum Khan, declare that this thesis titled, Optimized Communication
in 5G-Driven Vehicular Ad-hoc Networks (VANETs), is submitted in fulfilment of
the requirements for the award of doctor of philosophy, in the Faculty of Engineering
and Information Technology (FEIT), at the University of Technology Sydney.
I confirm that:
• This thesis is wholly my own work unless otherwise referenced or
acknowledged
• The work is done solely while in candidature for a research degree at this
University.
• All information sources and literature used are indicated in the thesis.
• This document has not been submitted/published for qualifications at any
other academic institution.
• This research is supported by the Australian Government Research Training
Program (RTP).
Signature of Student:
Date: October 14, 2019
i
Production Note:
Signature removed prior to publication.

Dedication
I would like to dedicate this work to my beloved mother (Najma Khan) and my
late beloved father (Salah Uddin Khan) whose dreams for me have resulted in this
achievement. I thank my mother with all my heart for all her prayers and un-
conditional love and care that kept me ﬂourishing throughout the journey of my
PhD.
ii

Table of Contents
Declaration Of Authorship i
Dedication ii
List of Figures vii
List of Tables x
List of Algorithms xi
List of Abbreviations xii
List Of Abbreviations xiii
List of Parameters xiv
List Of Parameters xv
ABSTRACT xvi
ACKNOWLEDGEMENTS xviii
1 Introduction 1
1.1 Thesis Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Objectives and Overview of Thesis . . . . . . . . . . . . . . . . . . . 5
1.3 Thesis Outline and Contributions . . . . . . . . . . . . . . . . . . . . 6
1.4 Related Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 Literature Review 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
iii

2.2 Vehicular Ad-hoc Networks (VANETs) . . . . . . . . . . . . . . . . . 12
2.2.1 Applications of VANETs . . . . . . . . . . . . . . . . . . . . . 13
2.2.2 Vehicular Communication (VC) . . . . . . . . . . . . . . . . . 13
2.2.3 Vehicular Communication Infrastructure VCI . . . . . . . . . 14
2.2.4 Features of VANETs . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.5 Heterogeneous Vehicular Ad-hoc Networks (HetVANETs) . . . 17
2.2.6 Challenges of Heterogeneous Vehicular Ad-hoc Networks (Het-
VANETs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Background and related work on 5G-Driven VANET Architectures . . 25
2.3.1 5G-Driven Technologies . . . . . . . . . . . . . . . . . . . . . 25
2.3.2 Cloud Radio Access Network (C-RAN) . . . . . . . . . . . . . 29
2.3.3 Network Function Virtualization(NFV) . . . . . . . . . . . . . 36
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3 5G Next generation VANETs using SDN and Fog Computing Frame-
work 40
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2 Background and Related Work . . . . . . . . . . . . . . . . . . . . . . 41
3.3 5G next generation VANET Architecture . . . . . . . . . . . . . . . . 43
3.3.1 Topology Structure of Fog Computing (FC) Framework, C-
RAN and the SDN controller: . . . . . . . . . . . . . . . . . . 43
3.3.2 Logical Structure of proposed 5G next generation VANET ar-
chitecture: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4 Simulation Methodology . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.5 Comparison of Throughput, Transmission delay and Control overhead
on controllers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4 An Evolutionary Game Theoretic (EGT) Approach for Stable and
Optimized Clustering in VANETs 54
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.1 VANET Clustering Protocols . . . . . . . . . . . . . . . . . . 57

4.2.2 Game Theory: . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3 Proposed EGT framework . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3.1 Proposed EGT Framework . . . . . . . . . . . . . . . . . . . . 64
4.4 System Model and Stability analysis . . . . . . . . . . . . . . . . . . 69
4.4.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.4.2 Solution Approach: . . . . . . . . . . . . . . . . . . . . . . . . 69
4.4.3 Replicator Dynamics and Stability of evolutionary equilibrium 70
4.4.4 Complexity Analysis . . . . . . . . . . . . . . . . . . . . . . . 76
4.5 Simulation set up scenarios and results . . . . . . . . . . . . . . . . . 77
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5 A Hybrid-Fuzzy Logic Guided Genetic Algorithm (H-FLGA) Ap-
proach for Resource Optimization in 5G VANETs 88
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.3 Challenges and Key enabler Technologies for 5G Driven VANETs . . 91
5.4 Resource Optimization in 5G Driven VANETS . . . . . . . . . . . . . 95
5.4.1 Minimise the number of FC-BBUCs (Min-BBUC) . . . . . . . 96
5.4.2 Minimize Delay (Min-Delay) . . . . . . . . . . . . . . . . . . 97
5.4.3 Capacity Load Balance(Cap-LB)) . . . . . . . . . . . . . . . 98
5.4.4 Number of FC-ZCs per BBUC Balance Algorithm (FC-ZC-
per-BBUC-Bal) . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4.5 Constant Traﬃc Load (CTL) . . . . . . . . . . . . . . . . . . 101
5.4.6 Multi-Objective Optimization . . . . . . . . . . . . . . . . . . 103
5.5 Hybrid-Fuzzy Logic guided Genetic Algorithm (H-FLGA) . . . . . . . 105
5.6 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . 109
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6 An End-to-End (E2E) Network Slicing Framework for 5G Vehic-
ular Ad-hoc Networks 119
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.2.1 Network Slicing in 5G Architecture . . . . . . . . . . . . . . . 122

6.3 End-to-End Network Slicing framework in 5G-driven VANETs . . . . 124
6.3.1 Hierarchy/levels of slicing for proposed E2E slicing framework: 124
6.3.2 Edge Cloud (EC) and CN Cloud: . . . . . . . . . . . . . . . . 126
6.4 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.4.1 Objective function . . . . . . . . . . . . . . . . . . . . . . . . 132
6.5 Simulation results and Discussions . . . . . . . . . . . . . . . . . . . . 133
6.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7 Conclusion and Future Directions 141
7.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.1.1 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.1.2 5G Next generation VANETs using SDN and Fog Computing
Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.1.3 An Evolutionary Game Theoretic Approach for Stable and
Optimized Clustering in VANETs . . . . . . . . . . . . . . . . 143
7.1.4 A Hybrid-Fuzzy Logic Guided Genetic Algorithm (H-FLGA)
Approach for Resource Optimization in 5G VANETs . . . . . 144
7.1.5 An End-to-End (E2E) Network Slicing Framework for 5G Ve-
hicular Ad-hoc Networks . . . . . . . . . . . . . . . . . . . . . 144
7.2 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
References 147
vi

List of Figures
1.1 Optimized Communication in VANETs . . . . . . . . . . . . . . . . . 1
1.2 Scope of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Vehicular Communication Infrastructure in the ITS systems [1] . . . 15
2.2 Evolution of mobile networks [2] . . . . . . . . . . . . . . . . . . . . . 19
2.3 SD-IoV Architecture [3] . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 How will the network look like with SDN [4] . . . . . . . . . . . . . . 27
2.5 Centralized Control Plane . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Traditional Networks Vs Software Deﬁned Networks [4] . . . . . . . . 29
2.7 Cloud RAN Infrastructure [5] . . . . . . . . . . . . . . . . . . . . . . 31
2.8 Traditional cellular architecture [6] . . . . . . . . . . . . . . . . . . . 32
2.9 Base Station with RRH [6] . . . . . . . . . . . . . . . . . . . . . . . . 33
2.10 Cloud RAN with RRH [6] . . . . . . . . . . . . . . . . . . . . . . . . 34
2.11 Cloud RAN architecture for mobile networks [6] . . . . . . . . . . . . 35
2.12 FV Infrastructure [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.13 Architecture showing integration of NFV, SDR and SDN [2] . . . . . 38
3.1 Topology Structure of 5G next generation VANETs using SDN and
Fog Computing (FC) Framework . . . . . . . . . . . . . . . . . . . . 44
3.2 Hierarchy of SDN controller, Cloud-RAN and Fog computing framework 46
3.3 Logical Structure of proposed 5G next generation VANETs . . . . . . 47
3.4 Throughput Comparison . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.5 Comparison of Throughput using average and adaptive bandwidth
allocation schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.6 Delay Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.7 Comparison of Control overhead on controller . . . . . . . . . . . . . 53
vii

4.1 Proposed EGT Framework . . . . . . . . . . . . . . . . . . . . . . . 58
4.2 Flow Chart of Proposed EGT . . . . . . . . . . . . . . . . . . . . . . 68
4.3 An illustraion of equilibirum point ne . . . . . . . . . . . . . . . . . 72
4.4 Equilibrium point for Population share ni/N . . . . . . . . . . . . . . 73
4.5 Boundary of equilibrium in the region of ni = ne ±δ for all 0 ≤ ni ≤ 1
where δ 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.6 Stability of equilibrium point for n1/N within n1 = ne ± δ . . . . . . 76
4.7 Simulation snapshots using Static Scenarios . . . . . . . . . . . . . . 77
4.8 Simulation snapshots using Manhattan Grid Mobility . . . . . . . . . 78
4.9 Stability convegence of System with 15 clusters in static scenario . . . 80
4.10 Stability convergence of System with 15 clusters using Manhattan
grid mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.11 Comparison of Switching rate of Proposed EGT with ALM [8] . . . . 81
4.12 Comparison of Average switching rate of proposed EGT with ALM [8] 81
4.13 Throughput maximization for static scenario and Manhattan grid . . 82
4.14 Optimum no. of clusters for static scenario and Manhattan grid . . . 82
4.15 Comparison of Throughput maximization at different speeds for N=100 83
4.16 Comparison of Throughput maximization at different speeds for N=200 83
4.17 Complexity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.18 Scalibility anaylysis for thorughput maximization at different popu-
lation sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.19 Throughput Vs Speed . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1 Minimize number of BBUC pools . . . . . . . . . . . . . . . . . . . . 96
5.2 Minimize Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.3 Capacity Load Balance . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.4 Number of FC-ZC per BBUC Balance . . . . . . . . . . . . . . . . . 99
5.5 Constant traffic load per BBUC . . . . . . . . . . . . . . . . . . . . 102
5.6 Flow Chart of Hybrid-Fuzzy Logic Guided Genetic Algorithm (H-
FLGA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.7 Variation of multi-objective function value for different numbers of
generations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

5.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.12 Multi objective Optimization using Optimized weights . . . . . . . . 116
5.13 End-to-End Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.1 An End-to-End (E2E) Network Slicing Framework for 5G-driven VANETs125
6.2 E2E mission critical slicing including CN and RAN in 5G-VANETs . 126
6.3 E2E Network Slicing Between EC and CN Cloud . . . . . . . . . . . 129
6.4 Mission Critical Resource Block . . . . . . . . . . . . . . . . . . . . . 130
6.5 Program Flow Chart of Proposed E2E slicing Scheme . . . . . . . . . 131
6.6 Optimum Utilization of resource for Mission Critical slice . . . . . . . 134
6.7 Optimum Utilization of resource for non-Critical slice . . . . . . . . . 135
6.8 Combined resource utilization for both Critical and non-Critical slices 135
6.9 Optimization summary as output from MATLAB . . . . . . . . . . . 136
6.10 Optimization using GA . . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.11 Optimized Front-haul distances of RRHs with BBUCs using equation
6.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
6.12 Optmized Front-haul connections of RRHs with BBUCs using equa-
tion 6.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.13 Comparison of E2E Latency of proposed scheme with 5G VANET
architecture [9] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
ix

List of Tables
2.1 Comparison of high speed Wireless Communication Technologies for
Vehicular Networks [1] . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.1 Requirements of Proposed architecture . . . . . . . . . . . . . . . . . 49
3.2 Simulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1 Basic components of proposed EGT with respect to VANET clustering 65
4.2 List of parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Network conﬁguration parameters in static scenarios and mobility
using Manhattan grid . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.1 Possible Type of Service (ToS) Values . . . . . . . . . . . . . . . . . 106
5.2 Type of Service (ToS) Vs. Priority ω for Fuzzy Inference System . . 106
5.3 Simulation Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 117
x

List of Algorithms
1 : H-FLGA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
xi

List of Abbreviations
VANETs Vehicular Ad-hoc Networks
VCNs Vehicular Communication Networks
VCI Vehicular Communication Infrastructure
HetVANETs Heterogeneous Vehicular Ad-hoc Networks
5G Fifth generation
IoT Internet of-Things
SD-IoV Software Deﬁned Internet of Vehicles
ITS Intelligent Transportation Systems
V2V Vehicle to Vehicle
V2I Vehicle to Infrastructure
RSU Road Side Units
CPRI Common Public Radio Interface
D2D Device-to-Device
E2E End-to-End
QoS Quality-of-Service
LTE-A Long Term Evolution Advanced
EGT Evolutionary Game Theory
SDN Software Deﬁned Networking
NFV Network Function Virtualization
C-RAN Cloud-Radio Access Network
MEC Mobile Edge Computing
BBU Base Band Unit
RRH Remote Radio Head
OTN Optical Transmission Network
xii

List Of Abbreviations
FC Fog Computing
FC-ZCs Fog Computing-Zone Controllers
FC-CHs Fog Computing-Cluster-heads
FC-Vehicles Fog Computing-Vehicles
FC-BBUCs Fog Computing BBU Controllers
ZC Zone Controller
D Delay
T Throughput
C Cost
CN Core Network
eNB Evolved node B
GA Genetic Algorithm
Min-BBUC Minimise number of FC-BBUCs
Cap-LB Capacity Load Balance
Min-Delay Minimize Delay
FC-ZC-per-BBUC-Bal FC-ZCs per BBUC Balance Algorithm
CTL Constant Traﬃc Load
H-FLGA Hybrid-Fuzzy Logic guided Genetic Algorithm
FIS Fuzzy Inference System
ToS Type of Service
EC Edge Cloud
CN Core Network Cloud
KPIs key performance Indicators

List of Parameters
G = N, H, S, uCh
EGT game
S = {Ch, M} Strategy set for vehicular nodes
si Current strategy of node i
N = {1, 2, ...., n} Set of vehicular nodes
H = {1, 2, ......, j} with j ⊂ N Set of clusters
uCh
net utility of a cluster head
pi(si) Cost function
TTC Total throughput of cluster
c1 link capacity between the cluster head and the RSU
cj, j ⊂ N link capacity between a member j within the cluster and CH
dH distance between the cluster head and the RSU
dM,j distance between a member j from the cluster head
γ Speed to convergence
Ū(t) average payoﬀ of the entire population of clusters
uChi
(t) payoﬀ to become a cluster head
pHi
(t) proportion of vehicles choosing cluster Hi
TTC average total throughput capacity of a given cluster
ne Equilibrium point
ZC = {ZC1, ZC2, ..., ZCn} set of Fog Computing Zone Controllers
nZC Number of FC-ZCs
nBBUc Number of BBUCs
Links = {BBUCi, ZCj} set of possible link pairs between FC-BBUCs and FC-ZCs
Costi,j link cost for linking ZCs j and BBUCi
D average load demand across all BBUCs
Di ith
element indicating the total load demand in BBUCi
Ni ith
total number of FC-ZCs connected to BBUCi
(ToS) = {D, T, C} requirement of customers based Throughout, Delay and Cost
xiv

List Of Parameters
ω1 Weight of Min-BBUC cost function
ω2 Weight of Cap-LB cost function
ω3 Weight of Min-Delay cost function
ω4 Weight of FC-ZC-per-BBUC-Bal cost function
ω5 Weight of CTL cost function
dfronthaul maximum front-haul distance
υfronthaul link propagation speed
δRTT Round Trip Time
τOWD(ms) one way delay

Abstract
Next generation Vehicular Ad-hoc Networks will be dominated by heterogeneous
data and additional massive diffusion of Internet of Things (IoT) traffic. To meet
these objectives, a radical rethink of current VANET architecture is essentially re-
quired by turning it into a more flexible and programmable fabric.
This research endeavours to provide next generation 5G-driven VANET architec-
ture, with solutions for efficient and optimized communication.
This thesis first introduces an innovative 5G-driven VANET architecture to pro-
vide flexible network management, control and high resource utilization, leveraging
the concepts of SDN, C-RAN and Fog Computing. A new Fog Computing (FC)
framework (comprising of zones and clusters) is proposed at the edge of the network
to support vehicles and end users with prompt responses, and to avoid frequent
handovers between vehicles and RSUs. The key results are improved throughput,
reduced transmission delay and minimized control overhead on the controller.
Furthermore, a novel Evolutionary Game Theoretic (EGT) framework is presented
to achieve stable and optimized clustering in the Fog Computing Framework. The
solution of the game is presented to be an evolutionary equilibrium. The equilibrium
point is also proven analytically and the existence of an evolutionary equilibrium
is also verified using the Lyapunov function. The results are analysed for different
number of clusters for different populations and speeds. An optimal cost is suggested
that defines an optimum clustering thus reducing an overhead of frequent cluster
reformation.
xvi

In addition, this thesis provides a Hybrid-Fuzzy Logic guided Genetic Algorithm
(H-FLGA) approach for the SDN controller, to support diversified quality of service
(QoS) demands and dynamic resource requirements of mobile users in 5G-driven
VANET architecture. The proposed Fuzzy Inference System (FIS) is used to opti-
mize weights of multi-objectives, depending on the Type of Service (ToS) require-
ments of customers. The results proved that the proposed hybrid H-FLGA performs
better than GA. The results improve spectral efficiency and optimizes connections
while minimizing E2E delay and further facilitates the service providers to imple-
ment a more flexible customer-centric network infrastructure.
Furthermore, an end-to-end (E2E) network slicing framework is proposed to sup-
port customized services by managing the cooperation of both the RAN and Core
Network (CN), using SDN, NFV and Edge Computing technologies. A dynamic
radio resource slice optimization scheme is proposed to slice the overall bandwidth
resources for mission critical and non-mission critical demands. The results meet
ultra reliability and E2E latency of mission-critical services.

ACKNOWLEDGEMENTS
Thanks to THE ALMIGHTY for giving me strength, opportunity and patience to
undertake this research and ability to learn and complete this adequately. No doubt
he is the best disposer of all aﬀairs. Without his blessings, this achievement would
not have been possible and today I can stand proudly with my head held high due to
ALMIGHTY support.
Despite the toughness of the entire experience and struggles of PhD, there are al-
ways some people who played very important roles to keep this process going and
making it successful. Every person is a paragon in his entirety and it is important
to treasure people. Being thankful gives us an appreciation for what we have.
In my journey towards this degree, I consider myself extremely fortunate to be
mentored by the soul of honour Prof. Mehran Abolhasan. He has always been
standing as a pillar of support throughout the good, the bad, and the downright
nasty days of my PhD. He has provided me all the support and freedom to pursue
my research, while ensuring that I stay on course and do not deviate from the core
of my research. Without his able guidance, this thesis would not have been possible.
I overwhelmingly pay my immeasurable appreciation and deepest gratitude to Prof.
Mehran Abolhasan for his never-ending guidance and supervision throughout the
entire journey. His motivational talks throughout thick and thins of PhD journey
can never be forgotten.
I have a great pleasure in acknowledging my gratitude to Dr. Wei Nei for his
great assistance and valuable time whenever I approached him and showing me the
way ahead with his great ideas throughout the entire journey. Without his instruc-
tions and help, this work might not have been accomplished.
I would like to thank my co-supervisor Dr. Justin Lipman whose support and
motivation helped me achieve my goal.
Bundle of thanks also goes to my colleagues and friends whom I have beneﬁted
xviii

from their support, healthy and technical discussions, their friendship and advices
over the years. Their support, encouragement and credible ideas have been great
contributors in the completion of the thesis.
I would also like to thank my sister and my brothers for their encouragement and
belief in my abilities that leads me to achieve my goal.
I owe thanks to a very special person, my husband, Adnan for his continued and
unfailing love, moral support and understanding throughout the entire journey of
PhD. You were always around at times I thought that it is impossible to continue,
you helped me to keep things in perspective. I could not have gotten through the
last moments of this degree without his love and support. I am lucky to have him
in my life.
I appreciate my adorable kids Usman and Rameen for enduring my ignorance
throughout this PhD and the patience they showed during my thesis writing. Their
smiles and hugs served as a livener and braced me up to get back to the work.
Thanks for the special prayers they made for me to accomplish this work. Words
would never say how grateful and luckiest I am in the world to have such a lovely
and caring family, standing beside me with their love and unconditional support.
Finally, I acknowledge the people who mean a lot to me, my parents, for show-
ing faith in me and providing their lasting support in all terms and aspects of my
life. Thank you Mum for your endless prayers and thank you Dad. I salute both of
you for the selfless love, care, pain and sacrifice you made to shape my life. I would
never be able to pay back the love and affection showered upon by my parents.
My Dad is no more in this world but he once taught me how to survive with the
impossible and I still remember that. I will be enthralled by his prayers till the day
I breath my last.

Chapter 1
Introduction
N ext generation Vehicular Ad-hoc Networks (VANETs) will be dominated by
heterogeneous data traffic and additional massive diffusion of Internet of Things
(IoT) traffic. To support the exponential growth of heterogeneous data traffic and
5G -Driven VANET Architecture
Using SDN and Fog Computing Frame Work
Stable and Optimized Clustering
approach for Fog Computing (FC)
Framework
Resource Optimization in 5G
VANETs
An End-to-End (E2E) Network Slicing
Framework for 5G-driven VANETs
v-RAN-runs
RAN, Core
and MEC
v-RAN
Edge Cloud (EC)
(NFV)
Hypervisor
VM/VNF v-RAN
Commercial
Server
Hypervisor
VM/VNF CN
Commercial
Server
Low bandwidth
CN Cloud (NFV)
Mission Critical
Applications Slice
Ultra Reliability,
Lowest delay
SDN enabled
Virtualisation Controller
(Network Connections
between VMs)
Critical
l
t delay
Hypervisor
Hypervisor
Non- Critical
Applications Slice
MEC
Server
MEC
Server
Fronthaul
Idle/
off
Idle/
off
Idle/
off
CN
(UP)
CN
(CP)
CN
(UP)
CN
(CP)
CN
(UP)
EC
(CP)
EC
(UP)
EC
(CP)
EC
(UP)
BBUC
Pool 1 BBUC
Pool 2
SDN Controllers
FC-ZC1
FC-ZC2
FC-ZC3
FC-ZC5
FC-ZC6
FC-ZC4
C
H
C
H
C
H
C
H
RSU
C
H
C
H
C
H
( ) g
k for 5G- s
s
s
s
s
s
s
s
s
s
s
s
driv
riv
iv
riv
iv
iv
v
v
v
v
ven V
en V
en V
en V
en V
en V
en V
en V
en V
en V
en V
en V NET
ET
ET
T
ANET
ANET
ANET
ET
ANET
NET
ANET
ANET
ANETs
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
v
v
v
v
v
v
v-
-
-RAN
RAN
RAN
RA
RAN
RA
RAN
R
RAN
RA
RAN
RA
RAN
R -
-
-ru
ru
runs
runs
runs
uns
uns
uns
runs
runs
runs
uns
runs
RAN, Core
and MEC
n 5
5
5G
SDN
N
approac
An EGT Approach for Stable and
Optimized Clustering for
FC- Frame Work
Figure 1.1: Optimized Communication in VANETs
diversified quality of service (QoS) demands, network resources need more flexible
1

2 Chapter 1
and optimized resource allocation strategies. In recent years, VANETs have rapidly
evolved and gained significant attention from both research and industry, as they
can provide a platform to connect a massive number of sensors for Internet of Things
(IoT) applications through wireless communication infrastructures. Integrating dif-
ferent wireless access networks such as cellular (3G, 4G, 5G, LTE, LTE D2D, 3GPP)
and IEEE 802.11p/DSRC, the Heterogeneous VANETs (HetVANETs) are expected
to be a good platform to meet diversified communication requirements of next gen-
eration VANETs. Besides these challenges, another major challenge is an ever in-
creasing network size, density and highly evolved physical layer technology, which
becomes an underlying bottleneck hindering the performance of HetVANETs. This
is due to the highly dynamic mobility, heterogeneity and handoffs between differ-
ent wireless infrastructures and the inflexibility of protocol deployment. Therefore,
conventional heterogeneous Vehicular Ad-hoc Network architectures lack in flexi-
bility on the large scale and cannot efficiently deal with the increasing demands
over different access networks. Nowadays, Software Defined Networking (SDN), an
enabler of 5G technology, is leading towards a revolutionary paradigm to facilitate
flexible network management and optimization on the large scale with unified ab-
straction [10], [11]. Besides SDN, Cloud Radio Access Network (C-RAN) has also
been widely accepted to be a promising solution for heterogeneous networks [10].
In addition, Cloud Radio Access Network (C-RAN) is expected to be a potential
candidate of next generation radio access networks that can facilitate rapid and in-
expensive network deployments by exploiting the extensive computation resources
offered by cloud platforms [12], [13], [2], [14]. Moreover, the idea of Network Func-
tion Virtualization (NFV) is also proposed to solve many problems caused by the
proprietary nature of existing hardware appliances. NFV decouples the software
implementation of network functions from the underlying hardware and has the po-
tential to lead to significant reductions in operating expenses (OpEx) and capital
expenses (CapEx).
In order to meet the before mentioned challenges, a radical rethink of current
VANET architecture is essentially required. According to our vision, this evolu-
tion can be achieved by turning it into a more flexible and programmable fabric, by
integrating 5G enabled technologies such as Cloud-RAN, SDN and NFV. Besides

Chapter 1 3
Scope of
Research
Figure 1.2: Scope of Thesis
these technologies, Edge Computing and Fog Computing technologies are also aimed
at offering ultra low latency, high resource utilization, and real-time access to radio
access that can be used by differentiated services and QoS optimization platforms.
These 5G-driven technologies can jointly be used to provide a multitude of diverse
services and are expected to substantially improve resource sharing over a common
underlying physical infrastructure. Other than these challenges, accommodating
high volumes of traffic with a diverse set of performance and service requirements is
also a major challenge. Reserving radio resources for a particular application may
lead to over-provisioning of resources [15]. There is a need for efficient on-demand
and instant resource allocation strategies [16]. With the evolution of 5G-driven tech-
nology, network slicing has also emerged as a major new networking paradigm, which
is considered to achieve high utilization of both communication and computing re-
sources and minimize the infrastructure deployment cost for operators [17], [18].

4 Chapter 1
The proposed 5G-driven research in VANETs will emerge in an attempt to address
the following challenges;
• Improved throughput
• Minimized overhead on Controller
• Minimized transmission delay
• Stable Cluster sizes
• Optimum Clustering
• QoS Provisioning
• Minimized E2E latency
• Capacity enhancement
• Optimized Front-haul connections
• Optimum resource allocation
• Implementation of Customer centric infrastructure depending on dynamic cus-
tomer needs.
• Ultra reliability
• Reductions in operating expenses (OpEx) and capital expenses (CapEx)
1.1 Thesis Statement
Vehicular Ad-hoc Networks (VANETs) have been promoted as a key technology to-
wards the evolution of upcoming 5G networks. Next generation 5G-driven VANETs,
dominated by heterogeneous data, bring new challenges like diversified QoS demands
including efficient resource management and resource optimization. To meet these
objectives, a radical rethink of current VANET architecture is essentially required
by turning it into a more flexible and programmable fabric. This can be achieved

Chapter 1 5
through technological improvements facilitated by emerging technologies like Soft-
ware Defined Networking (SDN), Network Function Virtualization (NFV), Cloud-
RAN (C-RAN) and Fog/Edge Computing. These technologies can jointly be used
to provide a multitude of diverse services and resource sharing using a globalized
view over a common physical VANET infrastructure. Researchers have provided
multiple solutions for optimized communication in VANETs, but no work has been
yet proposed that suggest efficient resource allocation and optimized communication
with the 5G perspective.
1.2 Objectives and Overview of Thesis
The main objective of this thesis is to provide efficient solutions for optimized com-
munication in 5G driven VANETs, to support the exponential growth of hetero-
geneous data traffic and to meet diversified quality of service (QoS) demands and
dynamic resource requirements of users.
An innovative 5G-driven VANET architecture is proposed, leveraging the concepts
of SDN, C-RAN and Fog Computing technologies. A novel lightweight and semi-
distributed approach, entitled the Evolutionary Game Theoretic (EGT) approach,
is proposed to achieve stable and optimized clustering in VANETs. To achieve
optimization of resources in 5G-driven VANETs, a Hybrid-Fuzzy Logic guided Ge-
netic Algorithm (H-FLGA) approach is proposed for the SDN controller, to solve a
multi-objective resource optimization problem. Five different objectives of network
resource provisioning are formulated and the problem is solved using the Fuzzy
Logic Guided Genetic Algorithm. The Fuzzy Inference system (FIS) is introduced
to optimise weights of multi-objectives, depending on the Type of Service (ToS)
requirements of customers. The H-FLGA scheme improves spectral efficiency and
optimizes connections while minimizing E2E delay and further allows the service
providers to implement a more flexible customer-centric network infrastructure. To
support customized services in 5G-driven VANETs, the proposed E2E network slic-
ing framework manages the cooperation of both the RAN and Core Network (CN),
using SDN, NFV and Edge Computing technologies. A dynamic radio resource slice
optimization scheme slices the overall bandwidth resources for mission critical and

6 Chapter 1
non-mission critical demands, by keeping in view resource elasticity requirements.
The proposed slicing solution meets ultra reliability and E2E latency of mission-
critical services.
1.3 Thesis Outline and Contributions
This section provides an outline of the thesis and summarizes the main contributions.
• In Chapter 2, an extensive review of relevant literature is presented - in
particular, features and challenges of Vehicular Ad-hoc Networks (VANETs)
comprising of heterogeneous infrastructures such as, cellular (3G, 4G, LTE,
5G, LTE D2D, 3GPP) and IEEE 802.11p/DSRC. Further to this, some chal-
lenges of current Vehicular Communication Networks (VCNs) including Het-
erogeneous Vehicular Ad-hoc Networks (HetVANETs) are explained in detail.
An overview of 5G-driven technologies such as Software Defined Networking
(SDN), Cloud-Radio Access Network (C-RAN), Network Function Virtualiza-
tion (NFV) along with their implementation in VANETs is also discussed.
• In Chapter 3, an innovative next generation 5G VANET architecture is pro-
posed by employing the concepts of SDN, C-RAN and fog computing technolo-
gies. In particular, a description of the high level design of proposed archi-
tecture along with the description of architecture components and their roles
contributed in the architecture are discussed. Moreover, a new Fog Computing
(FC) framework is proposed at the edge of the network, to support vehicles
and end users with prompt responses. Furthermore, some benefits of the pro-
posed architecture associating its feasibility in HetVANETs, are also discussed.
Using SDN and C-RAN technologies, the proposed architecture provides flexi-
bility, programmability and effective resource allocation, thus leading towards
significant reductions in OpEx. The performance of the proposed architecture
is investigated by comparing the transmission delay, throughput and control
overhead on the controller with other architectures. Simulation results show
improved throughput, reduced transmission delay and minimized control over-
head on controllers.

Chapter 1 7
• In Chapter 4, we look into the problem of cluster instability in VANETs for
the proposed FC framework in chapter 3. We propose a novel Evolutionary
Game Theoretic (EGT) approach to model the interactive decision making
process between vehicular nodes, in order to automate the clustering of nodes
and nomination of cluster heads, to achieve stable and optimized clustering
in VANETs. The equilibrium point is proven analytically and the existence
of evolutionary equilibrium is also verified using the Lyapunov function. Two
performance evaluation approaches are used to investigate the behaviour and
performance of the proposed game, under different populations, speeds and
cost functions. Our first approach is based on static scenarios and in our
second approach, we use the Manhattan grid as a mobility model to inves-
tigate the behaviour of our proposed game. Simulation results show that
the proposed framework is able to maintain cluster stability, as the clusters
evolve towards balanced sizes and the system converges with an average to-
tal throughput of clusters. Furthermore, the results reveal that the proposed
approach is lightweight, semi-distributed and allows faster convergence, thus
reducing the signalling overhead and complexity in large scale VANETs.
• In Chapter 5, further extending the work of our proposed 5G-driven VANET
architecture, a Hybrid-Fuzzy Logic guided Genetic Algorithm (H-FLGA) ap-
proach is proposed for the SDN controller, to solve a multi-objective resource
optimization problem. The proposed approach formulates five different ob-
jectives of network resource provisioning, particularly focussing on network
aspects such as capacity, delay, number of FC-BBUCs and the traffic load.
The Fuzzy Inference system (FIS) is proposed to optimise weights of multi-
objectives, depending on the Type of Service (ToS) requirements of customers.
Different options are weighted using the proposed FIS and multi-objective
weights are optimized, to provide an optimal solution. The results of the pro-
posed H-FLGA approach are compared with GA and our propsed 5G driven
VANET architecture in [9]. The proposed approach shows the minimized
value of multi-objective cost function when compared with GA. Results show
that the proposed H-FLGA approach minimizes E2E delay in comparison with
GA and 5G driven VANET architecture. The proposed scheme will provide

8 Chapter 1
the network service providers with an opportunity to implement a more flex-
ible customer-centric network infrastructure, by improving spectral efficiency.
Moreover, the proposed approach can also be used to support energy efficient
optimization, as some idle BBUC’s may be switched off without any adverse
effect on the overall system, thus reducing OpEx.
• In Chapter 6, an end-to-end E2E network slicing framework is proposed to
achieve the desired level of QoS provisioning for customized services in 5G-
driven VANETs. The proposed scheme considers managing the cooperation
of both RAN and Core Network (CN), using SDN, NFV and Edge Com-
puting technologies. The proposed framework distributes some services of
5G core close to cell sites using Mobile Edge Computing (MEC) technology
and keep other services with centralized processing, to meet desired levels of
KPIs. The distribution of both mission critical and non-critical demands is
achieved through SDN-enabled NFV technology. Furthermore, a dynamic ra-
dio resource slice optimization scheme is proposed, handling a mixture of both
best-effort traffic and mission-critical traffic. The problem is solved using the
Genetic Algorithm (GA). The overall bandwidth resources are sliced for mis-
sion critical and non-mission critical demands, by keeping in view resource
elasticity requirements. The results are compared with the previously pro-
posed VANET architecture. Simulation results show the effectiveness of the
proposed network slicing framework for the 5G network.
• In Chapter 7, the key contributions and results of the thesis are summarised.
Some possible future research directions are also discussed.
1.4 Related Publications
The publications relating to the work of the thesis are as follows;
• A new hierarchical 5G next generation VANET architecture is proposed by
employing the concepts of SDN, C-RAN and fog computing technologies, to
effectively allocate resources in VANETs with a global view. The transmis-
sion delay, throughput and control overhead on the controller are analyzed

Chapter 1 9
and compared with other architectures. Simulation results indicate improved
throughput, reduced transmission delay and minimized control overhead on
controllers. (Chapter 3). Ammara Anjum Khan, Mehran Abolhasan, and Wei
Ni. 5G next generation VANETs using SDN and fog computing framework.
In Consumer Communications Networking Conference (CCNC). 2018 15th
IEEE Annual, pages 1-6. IEEE, 2018 [9].
• An Evolutionary Game Theoretic (EGT) approach is presented to solve the
problem of cluster in-stability in VANETs. The proposed approach automates
the clustering of nodes and nomination of cluster heads and achieve optimum
clustering by using the cost function. The equilibrium point is proved analyti-
cally and the stability of equilibrium point is tested using the Lyapunov func-
tion. (Chapter 4). Ammara Anjum Khan, Merhan Abolhasan, and Wei Ni. An
Evolutionary Game Theoretic Approach for Stable and Optimized Clustering
in VANETs. IEEE Transactions on Vehicular Technology, 67(5):4501-4513,
2018. [19]
• A Hybrid-Fuzzy Logic guided Genetic Algorithm (H-FLGA) approach is pro-
posed for the SDN controller, to solve a multi-objective resource optimization
problem for 5G driven VANETs. The results of the proposed hybrid H-FLGA
approach are compared with GA and 5G driven VANET architecture in [9].
The proposed hybrid H-FLGA approach shows the minimized value of multi-
objective cost function when compared with GA. (Chapter 5). Khan, A.,
Abolhasan, M., Ni, W., Lipman, J., Jamalipour, A. (2019). A Hybrid-
Fuzzy Logic Guided Genetic Algorithm (H-FLGA) Approach for Resource
Optimization in 5G VANETs. IEEE Transactions on Vehicular Technology.
• An E2E network slicing framework is proposed to achieve QoS provisioning
among customized services in 5G-driven VANETs, by considering both RAN
and Core Network (CN) using SDN, NFV and Edge Computing technologies.
Furthermore, a dynamic radio resource slice optimization scheme is formulated
mathematically. Simulation results reveal that the proposed slicing framework
is able to optimize resources and deliver the targeted KPIs of mission critical
demands.(Chapter 6). Ammara Anjum Khan, Merhan Abolhasan, Justin Lip-

10 Chapter 1
man, Wei Ni and Abbas Jamalipour. An End-to-End (E2E) Network Slicing
Framework for 5G Vehicular Ad-hoc Networks (Under review in IEEE Journal
on Selected Areas in Communications - Special Issue on Network Softwariza-
tion Enablers.)

Chapter 2
Literature Review
2.1 Introduction
T his thesis explores different solutions to provide optimized communication in
5-driven Vehicular Ad-hoc Networks. This chapter includes general discussions of
Vehicular ad-hoc Networks including heterogeneous VANETs and their challenges.
Moreover, motivations behind using 5G-driven technologies are also discussed. The
key topics of this chapter includes:
• Vehicular Ad-hoc Networks, including their features, applications, and com-
ponents of Vehicular Communication (VC);
• Vehicular Communication Infrastructure (VCI);
• Heterogeneous Vehicular Ad-hoc Networks (HetVANETs);
• Vehicular communication Infrastructure (VCI) and HetVANETs based on
their advantages and disadvantages;
• Limitations of HetVANETs and description of Software Defined Internet of
Vehicles (SD-IoV) VANET architecture.
• Detailed description 5G-Driven technologies including;
– Software Defined Networking (SDN) and its applications in Vehicular
Ad-hoc Networks.
– Cloud Radio Access Network Architecture (C-RAN)
11

12 Chapter 2
– Network function Virtualization (NFV)
2.2 Vehicular Ad-hoc Networks (VANETs)
With recent advances in Intelligent Transportation Systems (ITS), Vehicular Ad-
hoc Networks (VANETs) have attracted a large interest in both academia and in-
dustry. VANETs can be considered as a potential core of ITS that is envisioned
to offer a wide variety of versatile services ranging from transportation and road
safety to infotainment applications like web browsing, video streaming, file down-
loading [20]. Smart vehicles are expected to heavily influence daily life and to
motivate a huge market in the near future. With the rapid development of wire-
less communication technologies, vehicles can utilize Vehicle-to-Infrastructure (V2I)
and Vehicle-to-Vehicle (V2V) communications with the help of on-board devices to
provide wireless communication services among vehicles and vehicle to road side
infrastructure [21], [22].
The era of the fifth generation (5G) cellular networks is rapidly evolving. Fifth
generation (5G) wireless communication networks emerge as a strong platform to
support V2V and V2I connections efficiently and securely as well as the integration
with V2X scenarios. The Internet of Vehicles (IoV) uses the wireless communi-
cation infrastructures to allow vehicles to be connected to new radio technologies,
and can be supported by 5G networks. 5G networks are anticipated to support
a number of vertical industries characterized by diversified applications including
future IoV applications and Intelligent Transport Systems (ITS) in scenarios like
high mobility, dynamic network topology, and high data volume with varying QoS
demands [23]. With the increasing demands of new techniques in Vehicular Ad-hoc
Networks, several new applications are emerging in the field of VANETs to integrate
the capabilities of next generation wireless networks to vehicles [24]. However, these
emerging applications require larger, more secure storage and complex computation
capabilities, hence bringing new resource challenges to Vehicular Ad-hoc Networks
(VANETs). To meet the increasing demands of radio and computing resources,
Vehicular Ad-hoc Networks take the advantages of cloud computing and fifth gener-
ation technologies allowing them to evolve towards next generation VANETs. Next

Chapter 2 13
generation Vehicular Ad-hoc Networks are envisioned to carry computing and com-
munication platforms, and will have enhanced sensing capabilities that will facilitate
transportation safety and efficiency.
2.2.1 Applications of VANETs
The applications of VANETs can be classified as;
1. Safety Warning Applications: These applications aim to broadcast mes-
sage alerts about dangerous events on the road with wireless communication
technology and also warn drivers receiving such alerts. These applications have
a strict delay requirements for safety and time critical messages dissemination.
These applications mostly rely on Vehicle to Vehicle (V2V) communication.
Examples include the emergency electronic brake light, the highway merge
warning, lane changing assistance, traffic signal violation warning including
accident avoidance such as cooperative collision avoidance, crash warning and
roll-over warning.
2. Entertainment Applications and General Information Services: The
main objective of these applications is to provide entertainment services to
the passengers and to improve traffic efficiency. Examples include interac-
tive communication services (such as internet access, music download, inter-
active games while travelling), including traffic information systems, weather
information, optimum route selection, value added services and gas station or
restaurant location [25], [21]. These applications usually rely on Vehicle to
Infrastructure (V2I) communication.
2.2.2 Vehicular Communication (VC)
There are two main components of Vehicular Communication;
1. Road-Side Units (RSUs):
RSUs are static components positioned at strategic positions across the roads
and serve as central controllers to provide direct wireless communication ser-
vices to the Vehicles. RSUs are sensors that are connected to the back-
bone networks to provide reliable communication. Furthermore, RSUs are

14 Chapter 2
equipped with network devices to support Dedicated Short Range Commu-
nication (DSRC) using IEEE 802.11p. Examples include GSM, WLANs and
WiMAX [25].
2. On Board Units (OBUs’):
Each vehicle is equipped with On Board Unit that act as a central processing
unit (CPU). With the help of OBUs’, vehicles can send and receive packets
and perform routing functions. These OBUs’ enable the vehicles to send and
receive messages to other vehicles or RSUs within their range using a wireless
communication medium. Nowadays most of the applications provided by In-
telligent Transportation systems depend on the geographical locations of the
sender and receiver, therefore, OBUs’ are equipped with a Global Positioning
System (GPS) or Diﬀerential Global Positioning System (DGPS) receivers.
2.2.3 Vehicular Communication Infrastructure VCI
Vehicular Ad-hoc Networks do not rely only on the ﬁxed infrastructure to provide
ubiquitous connectivity between vehicles [21], [24]. Vehicular Communication In-
frastructure (VCI) is categorized as follows;
• Vehicle to Vehicle (V2V) Communication or Inter Vehicle Com-
munication (IVC): V2V or IVC communication uses a multi-hop multicast
or broadcast mechanism for message dissemination. This type of communi-
cation is adopted when a vehicle is not directly connected to the RSU. V2V
requires less bandwidth for message dissemination as compared to V2I.
• Vehicle to Infrastructure (V2I) or Infrastructure to Vehicle (I2V)
Communication or Roadside Vehicle Communication (RVC) : In
V2I, I2V or RVC communication, the message is disseminated using the RSUs
and the vehicles. The RSU sends or broadcasts a message to all the vehicles
within its vicinity using a single hop transmission and uses multihop trans-
mission to broadcast message to vehicles that are not coming directly under
its vicinity. Sparse Roadside Vehicle Communication (SRVC) and Ubiquitous
Roadside Vehicle Communication (URVC) are subcategories of V2V or I2V
infrastructure. V2I communication requires higher bandwidth as compared to

Chapter 2 15
bandwidth demand for V2V Communication. For instance, to broadcast speed
warnings or broadcasting speed limits, the RSU will determine the appropriate
speed limit by checking internal database and will broadcast the speed limit
warning message to vehicles periodically. Furthermore, the RSU will also re-
quire additional bandwidth to issue an audio or visual warning message to
intimate the vehicle to reduce its speed, if a vehicle violates the desired speed
limit rules [25], [26].
A: Message to B
Figure 2.1: Vehicular Communication Infrastructure in the ITS systems [1]
• Hybrid Vehicle Communication (HVC): HVC uses both inter-vehicle
(V2V) communication and road-side (V2I) or (I2V) communication. Nowa-
days there are variety of ITS cooperative applications that use HVC infras-
tructure. These services include traﬃc management, road accidents warning,
interactive games, infotainment services including road condition sensing and
many other applications [26]. Routing-based (RB) Communication is used in
cases where the path is not directly provided and the packet will be routed
using RB Communication as shown in Fig. 2.1.

16 Chapter 2
2.2.4 Features of VANETs
VANETs have some other intrinsic features that distinguish them from other Ad
hoc Networks [25], [24], [26]. The following are the characteristics of VANETs;
• Frequently changing topology and Frequently disconnected Network:
In VANETs, the speed of vehicles is frequently changing and the topology is
dependent on the mobility of vehicles. Moreover, due to the rapid changes
in the topology, the connectivity between vehicles is hardly maintained. This
frequent network disconnection problem should be considered while designing
a VANET protocol.
• Delay Constraints: In some of the ITS applications, such as emergency
brake warning, the dissemination of the message is very time critical to avoid
a car crash. For these applications, it is more important to control delay
constraints instead of providing high data rates.
• Predicting the Mobility of Vehicles: In VANETs, the mobility of ve-
hicles is constrained by the road directions and traffic patterns. In order to
design a VANET protocol, these mobility models and predictions of the future
directions of the vehicles is important to be considered.
• Urban and Highway Traffic Scenarios: Since VANETs operate in two
different types of traffic scenarios such as Highway and Urban. The highway
scenario is simple due to having no obstacles, whereas the urban traffic scenario
is complex comprising of streets, buildings and obstacles. While designing a
VANET protocol, both types of traffic scenarios should be considered, includ-
ing the affects caused in communication due to shadowing and path loss in
urban traffic environments.
• Rich in Resources: VANET nodes are equipped with OBUs and have
enough resources like power, memory and processing capabilities. OBUs are
usually mounted on-board a vehicle and are used for exchanging information
with RSUs or with other OBUs. The OBUs are comprised of a Resource
Command Processor (RCP) and resources contain a read/write memory that
is used to store and retrieve information. OBUs also contain a user interface

Chapter 2 17
that connects to other OBUs and a network device for short range wireless
communication based on IEEE 802.11p technology.
2.2.5 Heterogeneous Vehicular Ad-hoc Networks (HetVANETs)
Heterogeneous Vehicular Ad-hoc Networks integrate Direct Short Range Commu-
nication (DSRC) with other cellular networks like 3G, 4G, LTE, LTE with D2D,
3GPP and 5G networks. HetVANETs are a potential solution to meet wide variety
of future ITS applications and services.
Challenges of Vehicular Communication Networks (VCNs)
In a typical VCN scenario, vehicles equipped with on-Board-Units (OBU’s) commu-
nicate with adjacent peer vehicles using Vehicle-to-Vehicle (V2V) communication,
and receive services from infrastructures (such as : Road Side Units (RSUs), Cellular
Base Stations (BSs) and Wi-Fi Access Points (APs) using Vehicle-to-Infrastructure
(V2I) communication. In recent years, Vehicular Communication Networks (VCNs)
have gained significant attention from both research and industry as they play a
fundamental role in enabling smart vehicles to get connected to surrounding vehi-
cles and the internet through wireless communication infrastructures [27], [28]. For
example, in case of a traffic accident, emergency information will be disseminated to
all vehicles in that area via V2V communication and detour routes will be provided
to the all other vehicles moving towards that site. Moreover, by exploiting the bene-
fits of cloud computing, real time traffic data is collected from vehicles and deployed
sensors across roadside infrastructure, and optimal decisions are made by the travel
planner applications. However, to meet optimal decision making on a large scale
and to facilitate flexible network control and optimization in VANETs, the concept
of Internet of Vehicles (IoV) comes into play. This is the platform where all drivers
and passengers can enjoy the services of ITS through internet. IoV is expected to
gain a large share of future market based on its applications and tremendous market
demands [3]. Moreover, there is an assumption that is further supported by a report
from Gartner forecasting that there will be more than 250 million connected vehicles
on road by 2020. Despite the efforts made in the field of VCN in recent years, there
are still many challenges left unaddressed.

18 Chapter 2
1. The current key research challenge of VANETs is to identify how to efficiently
exploit the heterogeneous VCNs [29]. The heterogeneity of VCNs can be
further discussed as;
• Wireless LANs (Wi-Fi and DSRC): Wireless LANs (WLANS) in-
cluding both WiFi and Direct Short Range Communication (DSRC) both
have the potential to be easily utilized in this scenario. WLANs have their
own advantage over other cellular technologies but there is a lack of con-
sistency in providing services for VCN scenarios. For example, WLANs
offer high data rates but at the same time their availability is dependent
on local infrastructures with limited coverage.
• Dynamic Spectrum Access (DSA) technologies over TVWS (TV
White Space):Dynamic Spectrum Access (DSA) technologies over TVWS
(TV White Space) can be utilized in VCN scenarios [30], [31]. TVWS
spectrum has a great potential to deal with VCN scenarios but, unfor-
tunately, consistency in services cannot be provided for VCN scenarios,
since, the resources required for TVWS are opportunistic and location
based [3].
• Cellular Technologies: Compared to WLANs, cellular networks pro-
vide the widest communication coverage area and most reliable connec-
tions, while concurrently their capacity is limited with increasing traffic
load in VCN scenarios. Future vehicular networking is expected to adopt
current cellular solutions like 3G, 4G, LTE, 3GPP and is expected to be
heterogeneous in terms of resources and network topology. 4G and LTE
are proposed to support vehicular communication scenarios [32]. For ex-
ample, a vehicle user can take benefits from wide coverage, low latency
and high throughput of LTE cellular networks. However, due to high ve-
hicle mobility and dynamic network topology, it is relatively challenging
to provide satisfied ITS services only through LTE. Especially in VCN
scenarios, where the number of vehicle users increases in the cell, the strict
latency requirements of safety related ITS applications cannot be guaran-
teed by LTE networks alone. Heterogeneous Vehicular NETworks (Het-
VNET), integrating different access networks technologies like DSRC and

Chapter 2 19
LTE, are expected to be a potential solution to meet various ITS service
requirements [27]. A unified wireless access standard called the Continu-
Figure 2.2: Evolution of mobile networks [2]
ous Air Interface for long and medium range (CALM M5) is also defined
for vehicular communication [33]. This standard is a product of unifica-
tion efforts made by the International Standards Organization-Technical
committee (ISO-TC 204 WG16) to define a single uniform standard to
support unified wireless access to improve VCN performance through in-
creased capacity, flexibility and redundancy in packet transmission and
reception. They combined several related air interface protocols and pa-
rameters on top of IEEE 802.11p architecture with support for exist-
ing cellular technologies. However, the current key research challenge of
VCNs is the lack of the central communication co-coordinator associated
with all the existing wireless access infrastructures related to vehicular
communication set-up, implementation and deployment [3], [1].
• Heterogeneity of V2I and V2V transmission modes: On the hand,
the heterogeneity related to both V2I and V2V transmission modes also
poses difficult challenges with high resource utilization efficiency and en-
hanced capacity on the large scale. Since, V2I can provide internet access
to vehicles and V2V can be used direct transmission same like Device-to-

20 Chapter 2
Device (D2D) communication in LTE networks.
• IEEE 802.11p based standards: IEEE 802.11p technology is widely
encouraged by vehicle manufacturing industries across the world. In the
USA, it is promoted through VII and VSCC, Japan through Advanced
Safety Vehicle project (ASV), Europe through C2C-CC and Germany
through SeVeCOM [10]. Compared to cellular technologies, the esti-
mated deployment cost of IEEE 802.11p (WAVE (Wireless Access in
Vehicular Environments)) is predicted to be relatively low. This pro-
tocol is a work-in-progress by the IEEE working group. The medium
access control (MAC) and physical (PHY) layers are based on the IEEE
802.11a standard. DSRC based on the IEEE 802.11p has gained popu-
larity due to its easy deployment, low cost and capacity to support V2V
communication [30]. However, it is dependent on ubiquitous deployment
of road side infrastructures and suffers from scalability issues related to
the limited radio range and unbounded delay [34]. Table. 2.1 also shows
a comparison of different wireless communication technologies in vehicu-
lar networks with respect to bandwidth, allocated spectrum, support for
mobility, bitrate, transmission power and communication range [1].
2. Challenge to meet diverse QoS requirements in VCN: In a VCN sce-
nario, each service request has different QoS requirements. For example, safety
related ITS services require low latency and high reliability requirements, some
applications like delay tolerant applications are more bandwidth consuming,
video streaming services have strict constraints on stable connection and high
speed. Future connected vehicles are expected to support diverse QoS require-
ments for different ITS services with a global view of all service requests, so as
to make optimum decisions for resource sharing for all ITS service requests.
3. Lack of management and control of VCNs on the large scale: With
an ever increasing vehicular network size and density as well as highly evolved
physical layer technology, the control and management of VCNs becomes
highly challenging, impeding the performance of Heterogeneous VCNs. Due
to the high mobility and rapidly changing topology of VCNs, the handoffs

Chapter 2 21
among different wireless access infrastructures are more frequent as compared
to traditional wireless networks, thus causing service interruptions. A large
number of wireless network infrastructures and spectrum resources may be
wasted and thereby lead to the low Quality of Experience (QoE) of vehicle
users. Therefore, the access and admission control mechanism should be fairly
coupled with the QoS driven resource allocations process. There is a need to
develop some unified ways to deal with control and management issues rising
in Heterogeneous VANETs on a large scale. To deploy new services and proto-
cols, a large amount of underlying network devices need to be configured and
modified by network operators [10].
Although mobile cellular technologies have a great potential of providing wide cov-
erage to vehicle user to provide wide variety of ITS service requirements, strict
requirements for real-time services cannot always be guaranteed by cellular net-
works [1]. This is due to changing traffic demands and mobility of vehicles in VCNs.
Therefore, the Heterogeneous Vehicular Network (HetVNET), which integrates dif-
ferent cellular communication networks with DSRC, is a potential solution to meet
communication requirements of different ITS applications and services without the
need of pervasive roadside infrastructures [27]. All of the above mentioned issues
related to current VCNs are considered as the building blocks of progress towards
more efficient future architectures to support future ITS applications.
2.2.6 Challenges of Heterogeneous Vehicular Ad-hoc Net-
works (HetVANETs)
Building Heterogeneous Vehicular Networks, requires a deep understanding of het-
erogeneity and its associated challenges in VCNs. Therefore, Heterogeneous Ve-
hicular Ad-hoc Networks (HetVNETs), which integrate DSRC with other cellular
technologies, can meet different future ITS service requirements [27]. Recently the
Qualcomm Snapdragon automotive development platform was released to support
auto manufacturers and suppliers to test, deploy and transform vehicular applica-
tions. This platform supports not only LTE but also IEEE 802.11p for DSRC. In
this section, we discuss some major advantages and challenges that are related to

22 Chapter 2
Table 2.1: Comparison of high speed Wireless Communication Technologies for
Vehicular Networks [1]
Wireless Fea-
tures
Wi-Fi 802.11p(WAVE) Infrared Cellular
Standards IEEE IEEE, ISO, ETSI ISO ETSI, 3GPP
Channel Band-
width
1-40MHz 10MHz, 20MHz N/A(Optical
carrier)
25MHz(GSM),
60MHz(UMTS)
Allocated Spec-
trum
50MHz at
2.5GHz
300MHz at
5GHz
30MHz (EU)
75MHz (US)
N/A(Optical
carrier)
(Operator-
dependent)
Frequency
Band(s)
2.4GHz,
5.2GHz
5.86GHz-5.92GHz 835-1035nm 800MHz,
900MHz,
1800MHz,
1900MHz
Communication
Range
100m 1000m 1000m(CALM
IR)
15km
Mobility suit-
ability
Low High Medium High
Bit rate 6-54Mbps 3-27Mbps 1Mbps,2Mbps 2Mbps
Transmission
power for mobile
node
100mW 2W EIRP (EU)
760mW (US)
12800 W/Sr
pulse peak
380mW
(UMTS)
2000mW
(GSM)
technologies used in HetVANETs followed by the description of some applications
for ITS.
• Various heterogeneous wireless access technologies exist in HetVANETs like
3G, 4G, LTE, LTE with D2D, 3GPP which cannot be easily well cooperated
under the traditional VCN architectures. Consequently, a many of wireless
network infrastructures and spectrum resources may be wasted thereby leading

Chapter 2 23
to the low quality of experience (QoE) for vehicle users. This situation may
become worse with the increase in scale of network. It may increase the cost
for network operators to deploy new services, as they need to configure or
modify a large amount of underlying devices.
• The management and control of VCNs on large scale becomes an underly-
ing bottleneck for performance of VCNs due to the ever increasing vehicular
network size and highly evolved physical layer technology [3].
• Moreover, the handoffs between different cellular infrastructures in VCNs are
more frequent due to intrinsic characteristics of VANETs (including high mo-
bility and rapidly changing topology) as compared to traditional cellular net-
works.
• To provide consistency in services with the frequent topology changes and vary-
ing QoS demands in VCNs, the heterogeneous substrate cannot have a global
view of all service requests, to make compromise and provide cooperation be-
tween all services. All the above aforementioned challenges of HetVANETs
call for rethinking of current HetVANET architectures to support all network
functionalities more efficiently on a large scale.
The following is a summary of advantages and challenges of candidate techniques
(including cellular and DSR) in V2V and V2I modes for HeTVANETs;
1. V2I Communications
• Advantages of LTE/LTE D2D
– Large coverage
– High downlink and uplink capacity
– Centralized and flat architecture
– Robust mechanism for mobility management
• Challenges of LTE/LTE D2D
– Lack of efficient scheduling schemes for ITS scenarios
– Users in idle state cause high delay in disseminating messages

24 Chapter 2
– Easily overloaded in high density environments
• Advantages of DSRC
– Low cost and easy deployment
– Suitable for local message dissemination
• Challenges of DSRC
– Serious channel congestion on large scale
– Unbalanced link
– Prioritization and service selection
– Hidden node problems and broadcast storm
2. V2V Communications
• Advantages of LTE/LTE D2D
– High Spectrum Efficiency
– High Energy Efficiency
– Efficient resource Scheduling on D2D
• Challenges of LTE/LTE D2D
– Interference between D2D pairs and other users
– Time consuming peer discovery
– Performance degradation with high mobility
• Advantages of DSRC
– Ease of deployment and low cost
– Ad-hoc mode
– Low overhead
• Challenges of DSRC
– Serious channel congestion on large scale
– Hidden node problems and Broadcast storm

Chapter 2 25
2.3 Background and related work on 5G-Driven
VANET Architectures
Heterogeneous vehicular networks have been regarded as a key enabling technology
to meet various QoS requirements for future Intelligent Transportation Systems
(ITS) services. However, conventional heterogeneous vehicular network architectures
lack in flexibility on large scale as discussed in section. 2.2.6 and therefore, cannot
efficiently deal with the increasing demands of data offloading over different access
networks. SD-IoV is leading towards a potential solution to provide ubiquitous
connectivity in integrated VCNs comprising of heterogeneous VCNs.
Data centres with millions of physical and virtual hosts are considered as a
valuable resource for providing many services related to internet and cloud com-
puting environments. To provide an efficient communication and cooperation on
large scale VANETs, millions of vehicles are widely spread in the environment of
Software Defined Internet of Vehicles (SD-IoV), where drivers and passengers can
enjoy all ITS services through the internet. Moreover, future connected vehicles
that are connected through IoV are playing an important role in ITS, by accom-
modating differentiated service requests and with different QoS requirements [3].
SD-IoV study has successfully demonstrated its superiorities to facilitate future ITS
services [3]. There are several papers in which different solutions have been proposed
for vehicular networks using the concepts of SDN and Internet of Vehicles (IoV) as
cited in [35], [28], [36], [37], [38]. Due to its potential to solve different problems in
vehicular scenarios as stated earlier, IoV is expected to be the potential solution to
meet all the aforementioned challenges related to HetVANETs. Despite the efforts
made in this field, this concept is still in its infancy and more refined and holistic
architectures for Vehicular networks are expected in future.
2.3.1 5G-Driven Technologies
Software Defined Networking (SDN)
Software Defined Networking (SDN) is leading towards a revolutionary paradigm
that is mainly differentiated due to separation of control and data plane with con-
trol plane having a centralized control, which dynamically defines forwarding rules in

26 Chapter 2
Figure 2.3: SD-IoV Architecture [3]
to switches in the data plane. Therefore, SDN helps in facilitating flexible network
management and optimization on large scale with unified abstraction [3]. Addition-
ally, network operators can also exploit the benefits of programmable SDN controller
to easily configure network devices and quickly deploy new applications [39].
SDN Approach
The network consists of a set of white boxes (programmable Switches). One or
more SDN controllers are connected to the white boxes via an out of band network.
Control and management is performed via a separate interface. Switches become
simple forwarding devices, obeying rules from the controller(s).
Current Networking Architectures: Limitations and Future
Applications
• Each network device has to be configured separately using low-level and of-
ten vendor-specific commands which is prone to errors. Many configuration

Chapter 2 27
O
Controller Platform
Open South bound API
Network Application(s)
Open Northbound API
Network Infrastructure
Figure 2.4: How will the network look like with SDN [4]
O
SDN Controller
Data
Figure 2.5: Centralized Control Plane

28 Chapter 2
changes are done manually.
• Networking protocols are distributed among devices (switches, routers, fire-
walls and middle boxes).
• Many complex functions are embedded into the infrastructure
– OSPF, BGP, NAT, TE, MPLS, Firewalls, multicast.
– Redundant layer services
– Unique differentiation
• Difficult to implement new protocols and features as it will change the control
plane of all devices which are a part of topology or network.
• There is no common view of the network.
• Expensive network up gradation as new features are introduced via expensive
and hard-to-configure equipment (aka middles boxes)
• Network capital costs have not been reducing fast enough and operational
costs have been growing, putting excessive pressures on network operators.
• Networks continue to have serious known problems with security, robustness,
manageability and mobility.
• Even vendors and third parties are not able to provide customized cost effective
solutions to address their customers’ needs.
• Need innovative ways to manage extremely large and dynamic networks.
Traditional Networks Versus Software Defined Networks
• The key difference is how SDNs handle data packets. In a traditional network,
the way a switch handles an incoming data packet is written into its firmware.
• With SDNs, management becomes simpler and middle boxes services can be
delivered as SDN controller applications

Chapter 2 29
Figure 2.6: Traditional Networks Vs Software Defined Networks [4]
• Most switches particularly used in commercial data centres respond to and
route all packets in the same way. SDN provides granular control over the way
switches handle data, giving network administrators the ability to automati-
cally prioritize or block certain types of packets.
• This technology allows for greater efficiency and control without the need to
invest on expensive, application-specific network switches and devices.
SDN has been instigated in different network scenarios either theoretically, practi-
cally or experimentally due to its benefits of programmability and flexibility. The
scenarios include, WLANs [40], wireless mesh networks [41], wireless sensor net-
works [42], cellular networks [43], narrow sense IoT [44] and wired data center net-
works [45].
2.3.2 Cloud Radio Access Network (C-RAN)
Recently, with the surge of mobile internet traffic, the mobile operators are facing
difficulties in solving the pressure of ever-increasing capital expenditures (CapEx)

30 Chapter 2
and operating expenses (OpEx) with much less growth of revenue [46]. To facilitate
rapid and inexpensive network deployments, the extensive computation resource of-
fered by the cloud platform can be exploited. Cloud Radio Access Network (C-RAN)
is expected to be a potential candidate of next generation radio access networks that
can facilitate inexpensive network deployments [5]. The average revenue per user
(ARPU) cannot catch up with the increasing expenditures, therefore, to meet user
requirements the network operators must find new solutions to maintain a healthy
profit and provide better QoS to customers. There are several ways to cope with
the increasing traffic requirements in an energy-efficient way [5]. The first choice
is to improve the efficiency of spectrum by employing more advanced transmission
techniques like MIMO and beam forming. The second choice is to use Dynamic
Spectrum Access technologies (DSA) such as, cognitive radio to exploit spectrum
holes. However, these techniques cannot provide consistent and reliable services,
and the increasing growth of data capacity is also limited. The third choice is to in-
troduce more small sized cells and take full advantage of frequency reuse. However,
this will cause more interference and there will be an increase in the operation and
maintenance cost of deployed small cell infrastructure. Recently, there are multiple
air interface standards introduced. Server-side cooperative-MIMO is used to facil-
itate rapid and inexpensive network deployments, by jointly processing geographi-
cally distributed base stations with overlapping coverage areas. This technique also
helps in reducing complexity, size and power requirements of base stations that are
geographically distributed.
Traditional Cellular Architecture
In the traditional cellular or Radio Access Network (RAN) architecture, there are
many stand-alone base stations (BTS). Each BTS covers a small area, whereas a
group of BTS provides coverage over a continuous area. Each BTS is responsible for
radio and baseband processing functionalities and transmits its own signal to and
from the mobile terminal, and forwards the data payload to and from the mobile-
terminal and out to the core network via the backhaul. There are few limitations of
traditional RAN architecture. Firstly, each BTS is costly to build and operate, as
it has to perform all radio and baseband processing functionalities. Secondly, when

Figure 2.7: Cloud RAN Infrastructure [5]
more BTS are added to a system to improve the capacity, interference among BTS
becomes more severe as BTS are closer to each other and most of them are using
the same frequency. Thirdly, the traffic of each BTS is continuously fluctuating
(called ’tide effect’) due to mobility of users. Consequently, the average utilization
rate of individual BTS becomes lower. However, these processing resources cannot
be shared with other BTS. Therefore, all BTS are designed to handle the maximum
traffic, not the average traffic that results in wastage of resources and power at times
when there is no traffic or the system is idle. There is an antenna that is generally
located within the proximity of few meters of the radio module as shown in Fig.2.8
as coaxial cables that are used to connect them exhibit high losses. X2 interface is
defined between base stations, S1 interface connects a base station with mobile core
network. This architecture was popular for 1G and 2G mobile cellular networks.
Base Station with RRH
This architecture was introduced when 3G networks were being deployed and re-
cently it is used by majority of base stations. In a base station with Remote Radio
Head (RRH) architecture, each base station is divided into two parts as shown in
Fig. 2.9.

32 Chapter 2
Figure 2.8: Traditional cellular architecture [6]
• RRH or Remote Radio Unit (RRU): RRH provides the interface to the
fiber and performs different functions like digital processing, digital to analog
conversion, analog to digital conversion, power amplification and filtering [47].
• BBU or Data Unit (DU): The baseband signal processing part is called a
Base Band Unit (BBU) or Data Unit (DU). The distance between a RRH and
a BBU can be extended up to 40 km. Optical fiber and microwave connections
can be used between RRHs an BBUs. One BBU can serve many RRHs. Com-
pared to cellular traditional architecture, where a BBU needs to be placed close
to the antenna, the BBU equipment is placed in a more convenient and easily
accessible place thus reducing cost for site rental and maintenance. RRHs are
statically assigned to BBUs that is similar to the traditional cellular or RAN
architecture. RRHs can even be placed up on poles or rooftops. RRHs can be
connected to each other in a daisy chained architecture.
Common Public Radio Interface (CPRI): RRH is connected to BBU by
Common Public Radio Interface (CPRI) interface. CPRI is the radio inter-
face protocol widely used for IQ data transmission between RRHs and BBUs

Chapter 2 33
on Ir interface [48]. It is a constant bit rate, bidirectional protocol that re-
quires accurate synchronization and strict latency control between RRH and
BBU. Other recommended protocols are the Open Base Station Architecture
Initiative (OBSAI) [49] and Open Radio equipment Interface (ORI) [50].
Figure 2.9: Base Station with RRH [6]
C-RAN Architecture
Cloud Radio Access Network (C-RAN) is a novel mobile network architecture where
baseband resources are pooled, so that they can be shared between base stations.
C-RAN is basically designed to be applicable to most typical RAN scenarios that
is from macro cell to femtocell as shown in Fig. 2.7. C-RAN has the following
components [5], [6];
• Baseband Unit (BBU): The BBU acts as a digital unit that is responsible
for implementing the base station functionalities from baseband processing
to packet processing. Several BBUs are placed in a central physical pool to
distribute RRHs according to RF strategies. Using the BBU pool, network op-
erators can dynamically deploy real-time virtualization technology that maps
radio signals from/to one RRH to any BBU processing entity in the pool.

34 Chapter 2
• Remote Radio Head (RRH): RRHs are responsible for performing radio
functions, including frequency conversion, amplification, and A/D and D/A
conversion. The RRHs also send and receive digital signals to and from the
BBU pool via optical fiber. Moreover, antennas are equipped with RRHs to
transmit and receive radio frequency (RF) signals.
• Optical Transmission Network (OTN): Optical Transmission Network
(OTN) is responsible for transmitting and receiving digital signals to and from
the BBU pool via optical fiber.
Figure 2.10: Cloud RAN with RRH [6]
The following are the benefits of C-RAN [5];
• Reduces Cost: It allows pooling the Baseband Units (BBUs) by aggregating
multiple base stations into a centralized BBU Pool, thus offering statistical
multiplexing gain by shifting the burden to the high-speed wireline transmis-
sion of In-phase and Quadrature (IQ) data. All the computational resources
are aggregated in a few big rooms, and are managed centrally and leaves sim-
pler functions in RRHs.

Chapter 2 35
• Improved Energy Efficiency: All processing functionalities of BSs are em-
bedded in a remote data-center. As a result, C-RAN reduces the burden on
base stations by dynamically allocating baseband functionaries in a BBU pool
by introducing energy efficient network operations. Power consumption can
also be reduced by dynamically allocating processing capability tasks between
BSs and also performing some migrating tasks between different BSs in a cen-
tralized BBU pool. Several BSs can be turned to low power or even be shut
down remotely in a BBU pool. Moreover, C-RAN architecture has made it
very convenient and cost effective for network operators to cover more service
areas or split the cell for higher capacity. Therefore, they only need to install
new RRHs that connect with the BBU pool.
• Better Spectrum utilization: C-RAN also improves network capacity by
performing load balancing and cooperative processing of signals originating
from several base stations. C-RAN allows sharing of Channel State Informa-
tion (CSI) of each Base station-mobile station BS-MS link, traffic data and
control information of mobile services among cooperating BSs. Consequently,
by using multipoint cooperation in this scheme, the system capacity is im-
proved, as more streams are multiplexed on the same channel with little or no
mutual interference.
Figure 2.11: Cloud RAN architecture for mobile networks [6]

36 Chapter 2
To address flexible network control, optimization and efficient data offloading in
heterogeneous Vehicular Ad hoc Networks (HetVANETs) on large scale, the cen-
tralization and flexibility of Cloud Radio Access Network (C-RAN) and Software
Defined Networking (SDN) can be integrated with Network Function Virtualiza-
tion (NFV) to support the dynamic nature of HetVANETs where multi domain
resources (like video streaming file downloading, Web browsing and others.) can
also be exploited to support future ITS applications. Interoperability among differ-
ent co-existing wireless infrastructure can also provided using C-RAN.
2.3.3 Network Function Virtualization(NFV)
The rapidly growing market demands have posed many challenges to the traditional
mobile broadband network architectures. On one hand, it is becoming difficult to
accommodate exponentially growing amount of network equipment of operators by
using limited machine room space. On the other hand, the heterogeneity caused by
different specifications of wireless access equipment has triggering many problems
related to management and optimization of networks [2]. Network function virtu-
alization (NFV) is recently proposed to improve the flexibility of network service
provisioning [51]. The idea of NFV is proposed along with other emerging tech-
nologies, such as software defined networking (SDN) and cloud computing to solve
many problems caused due to the proprietary nature of existing hardware appli-
ances. NFV decouples the software implementation of network functions from the
underlying hardware and it has the potential to lead to significant reductions in
operating expenses (OpEx) and capital expenses (CapEx). This technology is still
emerging and there are lot of opportunities for researchers to develop new architec-
tures and applications and to evaluate design trade-offs in emerging technologies for
its successful deployment. Moreover, this technology also facilitates the deployment
of new services with increased agility and faster time-to-value [7]. Some of the fu-
ture challenges for the deployment of NFV include the guaranteed performance of
networks for virtual appliances, their dynamic instantiation and migration as well
as their efficient placement.
It is well known that bringing new services into today’s networks is becoming diffi-
cult day by day due to the proprietary nature of existing hardware devices. This task

Chapter 2 37
does not only require highly and rapidly changing skills of professionals to operate,
manage and integrate these devices, but also requires dense deployments of network
equipment. NFV has been proposed to address all these challenges in an innovative
way to design, deploy and manage networking services by leveraging virtualization
technology. The main idea of NFV is to decouple the physical network equipment
from the functions that run on them [7]. The architecture we propose is based on in-
Figure 2.12: FV Infrastructure [7]
tegrating the centralization and ﬂexibility of SDN and C-RAN with NFV to support
the dynamic nature of Heterogeneous HetVANETs to support future ITS applica-
tions. Fig. 2.13 shows the relationship between SDR, SDN and NFV. Recently
Network Function Virtualization (NFV) has also emerged as a way to decouple
software implementation of network functions from the underlying hardware and
enable software to run in a virtualized environment. This improves the ﬂexibility
of network service provisioning [52] and facilitates the deployment of new services
with increased agility and faster time-to-value. Nevertheless, current VANET archi-
tectures cannot meet the latency requirements of future ITS applications in highly
congested and mobile scenarios. The future trend of autonomous vehicles drives
current VANET architectures, broadening their limits with challenging real-time
requirements. In addition, the maturity of cloud computing has adapted the in-

38 Chapter 2

- +(.;/8,4

xz[N$˜pi¥£¥
}¥

rS]T›¥
63
eWvLkG¥ dmH1d8“

X¥
!uƒtb€¥
Mˆ^¥
’“!¥
gA•Œ*…¥_¢9¥
:¥
qf”~s¥
pŸ2¥

#':

!1
¥dp-ž^5¥
¥-=S5¥

f5šZ¥DlNu¥

O‰^¥ Š`¥ O‹a¥
Cf'V¥ E)pf¥
$7.„¥¥

5 PhB¥+¥ 2=*

Q†œ^¥ Š`¥ Q†œa¥
c#JU6¥ % cJNi1¥
[¥

;€Y¥ 9)0 pc|—¥
,?1¥ / 3¥ ,!4¥
K¤R%[¥Sj@‚Ž(¥5¥

¥ w5¥05¥o¥Ff¥‘x{‡¥ ¥

–¥
yq#6UI¥

z™n¥
7¡kPpi¥

Figure 2.13: Architecture showing integration of NFV, SDR and SDN [2]
vasion of vehicular space with cloud-based services. The cloudification of network
resources through SDN and C-RAN is another promising enabler for 5G Next gen-
eration vehicular networks. SDN is leading towards a revolutionary paradigm which
controls the network in a centralized and programmable manner by decoupling the
forwarding functions (data plane) and network controls (control plane). Moreover,
due to its potential to offer flexibility, programmability and centralized knowledge,
it facilitates flexible network management and control on large scale, with unified
abstraction [13], [2], [14].
2.4 Conclusion
A comprehensive literature review is presented explaining the features and chal-
lenges of Vehicular Ad-hoc Networks (VANETs) comprising of heterogeneous in-
frastructures such as cellular (3G, 4G, LTE, 5G, LTE D2D, 3GPP) and IEEE
802.11p/DSRC. Furthermore, some challenges of current Vehicular Communication
Networks (VCNs) including heterogeneous Vehicular Ad-hoc network (HetVANETs)

Chapter 2 39
are also explained in detail. Current Vehicular Ad-hoc Network architectures util-
ising 5G-driven technologies are also discussed.
Diﬀerent 5G-driven technologies including Software Deﬁned Networking (SDN),
Cloud-Radio Access Network (C-RAN), Network Function Virtualization (NFV)
are also presented including their advantages and disadvantages. Furthermore, tra-
ditional networks are also discussed with 5G driven technologies including including
their limitations and future applications to highlight the importance of 5G-driven
architectures.

Chapter 3
5G Next generation VANETs
using SDN and Fog Computing
Framework
3.1 Introduction
The growth of technical revolution towards 5G next generation networks is ex-
pected to meet the various communication requirements of the future Intelligent
Transportation Systems (ITS). Motivated by the consumer needs for variety of the
ITS applications, researches are currently exploring different network architectures
and techniques, which could be employed in the next generation ITS. In recent
years, VANETs are rapidly evolving. The number of connected vehicles is predicted
to reach 250 million, by 2020 [53]. Moreover, by 2020, smarter and secure ITS are
expected to be operational as a VANET cloud [54]. Nevertheless, current VANET
architectures can not meet the latency requirements of future ITS applications in
highly congested and mobile scenarios. The future trend of autonomous vehicles
drives the current VANET architectures, broaden their limits with hard real-time
requirements.
This main objective of this chapter is to present a new hierarchical 5G next gen-
eration VANET architecture to provide flexible network management, control and
high resource utilization in the VANETs on a large scale. The key idea of this
holistic architecture is to integrate the centralization and flexibility of Software De-
40

Chapter 3 41
fined Networking (SDN) and Cloud-RAN (C-RAN), with the 5G communication
technologies, to effectively allocate resources with a global view. Moreover, a fog
computing framework (comprising of zones and clusters) has been proposed at the
edge, to avoid frequent handovers between the vehicles and the RSUs.
The major contributions and results of this chapter can be summarized as follows;
• A new hierarchical 5G next generation VANET architecture is proposed, util-
ising the idea of SDN, C-RAN and the fog computing technologies.
• To support vehicles and end users with prompt responses, a new Fog Com-
puting (FC) framework is proposed at the edge of network. The details of FC
framework are discussed further.
• The control functionality deployment of controller is divided in a hierarchical
manner to reduce control overhead on the centralised controller.
• The transmission delay, throughput and control overhead on the controller
are also analyzed and compared with other architectures. Simulation results
reveal the minimized transmission delay and control overhead on the controller,
considering different vehicle densities.
• Moreover, the throughput of proposed architecture is also analyzed, using av-
erage bandwidth allocation scheme and adaptive bandwidth allocation scheme
(i.e., by keeping in view different bandwidth demands of users). Simulation
results reveal the improved throughput.
The rest of the chapter is organized as follows. In section 3.2, background and some
related work is presented, to describe the motivations towards the 5G enabler tech-
nologies for the VANETs. Section 3.3 describes the topology and logical structure
of architecture. In section 3.5, the performance of proposed architecture is analyzed
and compared with the other architectures. Finally, the work is concluded.
3.2 Background and Related Work
Due to high mobility and rapidly changing topology of VANETs, it is difficult to
realize next generation ITS services by using a single wireless infrastructure. Exten-
sive research and efforts have been made from both industry and academia, in the

42 Chapter 3
field of next generation Vehicular Communication Networks (VCNs), to get Smart
Vehicles connected with the surrounding vehicles, road-side infrastructures, and in-
ternet through different wireless communication infrastructures [27], [28]. Therefore,
next generation vehicular networking is expected to adopt current cellular solutions
such as 4G Long Term Evolution (LTE), 3GPP, and is expected to be heteroge-
neous in terms of resources and network topology. LTE systems offer the benefits
of large coverage, high throughput and low latency [32]. However, due to high
vehicle mobility and dynamic network topology, it is relatively challenging to pro-
vide satisfied ITS services only through LTE systems. Integrating different access
networks technologies like DSRC and LTE, is proposed to be a potential solution
to meet various ITS service requirements [27]. However, building heterogeneous
vehicular networks (integrating IEEE 802.11p with cellular technologies like 3G,
4G/LTE systems) requires a deep understanding of heterogeneity and its associ-
ated challenges in VANETs. Due to high mobility and rapidly changing topology
of VANETs, the handoffs among different wireless access infrastructures are more
frequent, as compared to traditional wireless networks thus causing service inter-
ruptions. There is a need to develop some unified ways to deal with control and
management issues rising in heterogeneous VANETs on large scale. Furthermore,
to provide consistency in services with the frequent topology changes and varying
QoS demands in VANETs, the heterogeneous substrate cannot have a global view
of all service requests, to make compromise and provide cooperation between all
services. Inspite of all the efforts made in the field of heterogeneous VANETs, there
is still a dramatic gap between the practical requirements of ITS services and what
can be offered by existing heterogeneous VANETs. All of these issues above, call
for rethink of the current network architecture for VANETs. Consequently, the
research and development for the fifth generation (5G) systems have already been
started [55], [56], [2], [14], [57], [58], [59], [60], [61]. On the other hand, SDN has been
proposed as a promising technique that will play a key role in the design of 5G wire-
less communication networks [14]. SDN is proposed to be an effective technology to
be capable of supporting the dynamic nature of VANETs and ITS applications, by
facilitating flexible network management and optimization on large scale with uni-
fied abstraction [13]. In order to meet the demanding requirements of future ITS,

Chapter 3 43
SDN, Cloud Computing, Fog Computing are expected to be future candidate tech-
nologies for 5G VANETs. Some initial studies have also been carried out to integrate
either of these technologies into Vehicular Communication Networks [13], [10], [38]-
[62]. Nevertheless, the performance of SDN technology becomes limited in RSUs,
when the number of vehicles connected with RSU increases [38]. The frequent han-
dover problem in dense scenarios of VANETs, reduces the performance of SDN at
RSUs [63]. However, it is also realised that the scalability of Wireless Distributed
Networks (WDNs) is improved by using techniques like; clustering, multichannel
routing and zoning [22] and [64].
Nowadays, C-RAN has been widely accepted to be a promising solution for het-
erogeneous networks [10]. In C-RAN, all RAN functionalities are performed in the
centralized BBU pool, in cloud based infrastructure, which are connected to RRHs
via fibre. The separation between the data plane and control plane via SDN can be
built upon the open platform of Cloud-RAN by keeping in view the service demands
of different users, thus reducing operational cost. In this chapter, a hierarchical 5G
next- generation VANET architecture is proposed by employing the concepts of
SDN, C-RAN and fog computing as shown in Fig. 3.1.
3.3 5G next generation VANET Architecture
3.3.1 Topology Structure of Fog Computing (FC) Frame-
work, C-RAN and the SDN controller:
To support vehicles and end users with prompt responses, FC framework is config-
ured at the edge of network. FC framework is comprised of the following compo-
nents;
• Fog Computing-Zone Controllers (FC-ZCs): The FC-ZCs are the com-
puting enhanced (i.e., CPU and storage) wireless access infrastructures such
as, RSUs, Base Stations (BSs) connected with the BBU controllers, through
broadband connections. In our case, a zone is defined as a group of vehicles
that is registered with one RSU or a BS. Therefore, one FC-ZC is responsible
for controlling one zone. Most of the data at edge is processed and saved by

44 Chapter 3
/
SDN Controllers
WiMax
4G/LTE
Cloud-RAN
FC-BBUC Pool
FC-CH
FC-ZC
FC-ZC
FC-ZC
FC-BBUC
FC-BBUC
FC-CH
FC-CH
FC-CH
FC-CH
FC-CH
H
H
H
FC-
FC-ZC
WiMax
3G
FC-Vehicle
FC-Vehicle
FC-Vehicle
FC-Vehicle
FC-Vehicle FC-Vehicle
FC-Zone
FC-Zone
CH
H
H
H
H
H
H
H
H
3G
-C
-
-C
-
-CH
CH
CH
CH
CH
C
CH
C
F
F
FC-Vehicle – FC-Vehicle
Communication
FC-ZC – FC-CH
Communication
FC-CH -FC-CH
Communication
FC-BBUC - FC-ZC Communication
Distributed Control Plane
FC-CH -FC-Vehicle
Communication
4G/LTE
C-CH
FC
H
H
H
H
H
H
H
H
FC-CH
FC-CH
FC-ZC
C
C
C
C
WiMax
x
x
x
x
FC
FC
FC-Vehic
c
c
c
c
c
cle
l
l
le
le
le
l
FC-Ve
V
V
V
V
V
V
V h
h
hicle
le
FC-Vehicle
e
H
H
H
H
H
H
Figure 3.1: Topology Structure of 5G next generation VANETs using SDN and Fog
Computing (FC) Framework
FC-ZCs. Moreover, FC-ZC devices are SDN- enabled, meaning they are un-
der the control of SDN controller. SDN controller can control functionalities
such as, packet forwarding, and transmitting, as well as operations related to
infrastructures such as, power control, channel assignment and resource allo-
cation. SDN controller collects and forwards the state information of FC-ZCs
into the C-RAN, via Fog Computing BBU Controllers. The control overhead
of vehicles remains in their own vicinity i.e., FC-zones or FC-Clusters, and is
not sent to the SDN controller, unless required. Hence, FC-ZCs and FC-zones
play an important role in minimizing the overhead in the control plane. These
devices act as both control pane and data plane elements.
• Fog Computing-Cluster-heads (FC-CHs:) Further, FC-zones are divided
into Fog Computing-Cluster heads (FC-CHs. Each FC-CH is controlled and
managed by FC-ZC. FC-CHs are the vehicles, equipped with SDN-enabled On
Board Units (OBUs). The potential functionalities of OBUs include, packet
forwarding, power control, channel selection, interface selection and transmis-
sion mode (i.e, V2V or V2I communication). FC-CHs are also control plane
and data plane elements. FC-CHs collects and forwards the state information
of FC-Vehicles within a FC-ZCs.

Chapter 3 45
• Fog Computing-Vehicles (FC-Vehicles): FC-Vehicles act as end users,
and are also equipped with SDN-enabled OBUs. The potential functionalities
of OBUs include, packet forwarding, sensor localization system like Global Po-
sitioning system (GPS), power control, channel selection, interface selection
and transmission mode (i.e, V2V or V2I communication). Moreover these
OBUs are also equipped with radio transceivers for Wireless Access in Vehic-
ular Environment (WAVE) and other wide-range radio transceivers such as,
3G/4G/LTE for communication with cellular BSS.
• Fog Computing BBU Controllers (FC-BBUCs): FC-BBUC connects
mutiple FC-ZCs with the backhual links. The FC-BBUC acts a as digital
unit that is responsible for implementing the base station functionalities, from
baseband processing to packet processing. Several FC-BBUCs are placed in a
central physical pool, to distribute FC-ZCs according to RF strategies. The ad-
vantage of using SDN-based virtualization for C-RAN, in our proposed frame-
work is that resource allocation and scheduling can be eﬀectively and simply
managed by the central controller, with a global view. Therefore, FC-BBUCs
act as a bridge, connecting VANET infrastructure with the SDN controller.
The FC-BBUC collects the state information of diﬀerent FC-ZCs connected
with it, and by using its own local intelligence, it can make forwarding deci-
sions, thus reducing the overhead on centralised controller. FC-ZC will com-
municate with the FC-BBUC, for inter FC-zone communication. Therefore,
FC-BBUCs are the data plane as well as control plane devices.
• SDN Controller: As a core component, SDN controllers are responsible
for network management and operations such as, rule generation, resource
allocation and mobility management. Moreover, they can also perform some
advanced network functionalities like, learning, network analysis and data pre-
processing. In our case, SDN functionalities are distributed and also shared
among local controllers i.e., FC-ZCs and FC-BBUCs and FC-CHs in a hier-
archical manner. Moreover, the SDN controller is also responsible for Fog
Orchestration and resource management of fog.
• Optical Transmission Network (OTN): Optical Transmission Network

46 Chapter 3
Fog Computing (FC)-Framework at Edge
(Comprising of FC-BBUCs, FC-Zones and FC-
CHs )
Vehicular Ad hoc Network (VANET)
Infrastructure
SDN Controller and C-RAN
Figure 3.2: Hierarchy of SDN controller, Cloud-RAN and Fog computing framework
(OTN) is responsible for transmitting and receiving digital signals to and from
the FC-BBUC pool via optical fiber. The FC-ZCs send and receive digital
signals to and from the FC-BBUC pool via optical fiber.
3.3.2 Logical Structure of proposed 5G next generation VANET
architecture:
The logical structure of proposed architecture is divided into data plane, control
plane and application plane as shown in Fig. 3.3. The data plane includes FC-
Vehicles, FC-CHs, FC-ZCs and FC-BBUCs. Functionalities include data collection,
quantization and then forwarding data to the control plane [65], [37]. The data
plane devices can be configured in to the following function modules;
• Information gathering module of FC-Vehicles, FC-CHs, FC-ZCs,
FC-BBUC: This module uses different sensors to record information related
to position, speed and direction of vehicles and CCTVs, network cameras, lane
checking cameras.
• Communication module of FC-Vehicles, FC-CHs, FC-ZCs, FC-
BBUCs: This module further includes V2V and V2I communication module.
V2V provides wireless communication between two adjacent vehicles, that may

Chapter 3 47
be two FC-Vehicles or two FC-CHs or a FC-CH and a FC-Vehicle, by using
WiFi/WAVE. V2I communication provides wireless communication between
FC-CHs and FC-ZCs. Further, the communication module of FC-BBUCs in-
cludes two types of communication modules, one is between FC-BBUC to
FC-ZC and, other module is between FC-BBUC to SDN controller. Further-
more, inter-(FC-ZCs) and inter-(FC-BBUCs) communication is also performed
by this module.
Dynamic Offloading for
HetVANETs Access Control
Mobility Management
Application Layer
Entertainment services
Network Traffic Monitoring
Controller Interface
Security Management and
Authentication
Control
Plane
Network Status monitoring
Module
SDN Cloud Computing Module
Network Status
monitoring Module
Fog Computing Module/Inter-Zone
Communication module
Inter- FC-Zone
Communication Module
Interface Compatibility
Module
Topology Information gathering Module
FC-BBUC
FC-ZC
FC-CH
SDN controller
FC-Vehicles
Data gathering Module
Communication Module
FC-BBUC
FC-ZC
FC-CH
Data
Plane
Figure 3.3: Logical Structure of proposed 5G next generation VANETs
The control level of SDN decides the flow rules or policy rules [66]. Since, we
are using fog architecture at edge, therefore, the SDN controller will operate in
Hybrid Control Mode as shown in Fig. 3.1 and 3.2. The control plane includes
SDN controllers, FC-BBUCs, FC-ZCs and FC-CHs. The FC-BBUC is the main
control center or fog controller of fog framework. SDN controller functionalities are
shared at the edge of network, between FC-BBUC, FC-ZCs and FC-CHs. The SDN
controller will not take full control of the network. Instead of sending specific flow
rules, the SDN controller will send abstract policy rule. The specific behaviour of
policy rules will be decided by FC-ZCs, FC-BBUCs and FC-CHs, depending on their
own local intelligence [35]. Following are the control plane function modules;
• Information gathering modules of FC-BBUC, FC-ZCs and FC-
CHs: To draw global view map of network, based on data information pro-

48 Chapter 3
vided by the data plane.
• Computing and Storage modules: These modules are deployed in fog
computing framework devices and cloud computing centres.
• Network status monitoring module: Responsible for monitoring the links
of 5G SDN-based VANET architecture.
• Inter-FC-zone communication module: Configured in FC-ZCs to pro-
vide inter zone communication in fog network.
• Inter-FC-BBUC communication module: Configured in FC-BBUCs to
provide inter BBU communication in Cloud-RAN.
• Intra-FC-zone communication module: Configured in FC-CHs to pro-
vide communication between FC-CHs within a FC-Zone.
The application plane is responsible for generating rules and strategies, based on
different application requirements of users/vehicles, and forward these rules to the
control plane. Details are in Fig. 3.3. Table 3.1 shows some of the requirements of
proposed architecture using Cloud computing (SDN controller cloud and C-RAN)
and fog computing framework.
3.4 Simulation Methodology
The performance of proposed architecture is investigated, by analysing the through-
put, transmission delay, and control overhead on controllers, using MATLAB. Some
simulation set-up details are presented in Table 3.2. All bandwidths are averagely
assigned by FC-ZC, to vehicles within a zone. However, every vehicle needs differ-
ent bandwidths in practical ITS scenarios and applications. Considering the real
bandwidth requirements of vehicles, an adaptive bandwidth allocation scheme is
also used to optimize the throughput of fog framework.

Chapter 3 49
Table 3.1: Requirements of Proposed architecture
Requirements Cloud Computing
framework (C-RAN and
the SDN controller)
Fog Computing Frame-
work
Mobility Support Limited Supported
Geographical Distribution Centralised and Distributed Centralised and Distributed
Security Undeﬁned Can be Deﬁned
Location of Server Nodes Within internet Edge Network
Distance between vehicle
and servers
multiple hops single hop/very few hops
Location Awareness No Yes
Delay High Very Low
Control functionality de-
ployment
Hierarchical Hierarchical
Controller Operation Mode Hybrid (shared between Fog
computing devices and the
SDN controller)
Hybrid (shared between
zone controllers and CHs)

50 Chapter 3
Parameters Values
Transmission range 100m
RSU range 570m
No. Of Vehicles 10, 20, 30, 40, 50, 60
Transmit power 24dBm
Receiver Sensitivity -80dBm
RSU range 570m
Time per slot 0.5 ms
Bw 100 Mbps
No. of RSUs 9
Clustering ALM [8]
Table 3.2: Simulation parameters
3.5 Comparison of Throughput, Transmission de-
lay and Control overhead on controllers
The performance of our proposed architecture is analyzed and compared with two
architectures i.e., traditional architecture and 5G VANET architecture proposed
in [67], named as 5G SD VANETs for our comparison. In traditional architecture,
every vehicle communicates with the RSU directly, whereas in [67], each node has to
send signalling information to a node closer to RSU. The performance of proposed
architecture is investigated, by analysing the throughput, transmission delay, and
control overhead on controllers, using MATLAB.
Considering the real bandwidth requirements of vehicles, an adaptive bandwidth
allocation scheme is also used to optimize the throughput of fog framework. Sim-
ulation results in Fig. 3.4 show improved throughput, as compared to throughput
in [67], and throughput of traditional architecture. It is shown in Fig. 3.5 that the
throughput of a FC framework using both average and adaptive bandwidth alloca-
tion scheme is improved as compared to throughput of fog cell in [67]. We analyse
and compare the transmission delay of vehicle, in fog framework, considering diﬀer-
ent vehicle densities. In [67], as the complexity of handovers between vehicles and
RSU is increased, with an increase in multihop relay vehicles, the propagation delay

Chapter 3 51
10 15 20 25 30 35 40 45 50 55 60
No. of Vehicles
0
100
200
300
400
500
600
700
800
Throughput
(Mbps)
Traditional Architecture
5G SDN-based VANETs using FC-Framework
5G Software Defined Vehicular Networks [67]
Figure 3.4: Throughput Comparison
increases. Using the concept of zones and clusters, the number of multihop relay ve-
hicles is reduced, thus reducing delay. For analysis, we use ALM [8], as a clustering
strategy. Fig. 3.6 shows, there exist a minimum transmission delay of 0.06ms, as
compared to transmission delay of traditional architecture and 5G Software Defined
Vehicular Networks architecture [67]. The reason is, in our proposed FC-Framework,
the control functionalities are divided among different controllers and data process-
ing and applications are concentrated in devices/vehicles at the network edge, rather
than existing almost entirely in the cloud. Moreover, devices/vehicles communicate
peer-to-peer to efficiently share/store data and take local decisions, thus reducing
delay. Another reason is that due to more than one FC-CHs within FC-zones, the
FC-CHs are directly communicating with the RSUs, thus reducing number of relay
hops for transmission and reducing delay. It is also seen that, when the density of
vehicles is low, the distance among adjacent vehicles is far away, therefore, the suc-
cess transmission probability of link is low, thus, delay will be increased. Increasing
vehicle density, will decrease the distance among adjacent vehicles and therefore,
the success of transmission probability will be increased. Therefore, delay will be
minimized. We also analyse the control overhead on controllers. Fig. 3.7 shows that
the control overhead on controller is significantly reduced as compared to the con-
trol overhead on controller using traditional architecture and 5G Software Defined
Vehicular Networks architecture in [67]. This is due to hierarchical distribution of
controllers in control plane, and practical use of zones and clusters in our proposed

52 Chapter 3
10 15 20 25 30 35 40 45 50 55 60
No. of Vehicles
0
100
200
300
400
500
600
700
800
900
1000
Throughput
(Mbps)
Average Bw (Proposed)
Adaptive Bw (Proposed)
Average Bw (5G SD VANETs [67])
Adaptive Bw (5G SD VANETs [67])
Figure 3.5: Comparison of Throughput using average and adaptive bandwidth allo-
cation schemes
10 15 20 25 30 35 40 45 50 55 60
No. of Vehicles
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Delay
(ms)
Figure 3.6: Delay Comparison
FC-framework.

Chapter 3 53
10 15 20 25 30 35 40 45 50 55 60
No. of Vehicles
0
50
100
150
200
250
Control
Overhead
(No.
of
packets)
Figure 3.7: Comparison of Control overhead on controller
3.6 Conclusion
This chapter presents a new hierarchical 5G next generation VANET architecture,
by employing the concepts of SDN, C-RAN and fog computing technologies. The
topology and logical structure of architecture is also discussed in detail. Moreover, a
detailed background and overview of 5G enabler technologies for VANETs including
SDN, Cloud-RAN and fog Computing technologies is also presented. Furthermore,
a new Fog Computing framework is presented that offers delay-sensitive, location-
awareness and mobility-based real time services suitable for future ITS scenarios.
Using SDN and C-RAN technologies, the proposed architecture provides flexibility,
programmability and effective resource allocation using control plane and centralised
global knowledge, thus leading towards significant reductions in operating cost of
operators.
It is concluded from the simulation results that the proposed architecture can pro-
vide improved throughput, reduced transmission delay and minimized overhead on
controllers.

Chapter 4
An Evolutionary Game Theoretic
(EGT) Approach for Stable and
Optimized Clustering in VANETs
4.1 Introduction
Discovering and maintaining efficient routes for data dissemination in Vehicular Ad
hoc Networks (VANETs) has proven to be a very challenging problem. Clustering
is one of the control protocols used to provide efficient and stable routes for data
dissemination. However, the rapid changes in network topology in the VANETs
creates frequent cluster reformation, which seriously affects route stability.
The main objective of this chapter is to present a novel Evolutionary Game Theoretic
(EGT) framework to automate the clustering of nodes and nomination of cluster
heads, to achieve the cluster stability in the VANETs.
The main contributions of this chapter can be summarized as follows;
• An EGT framework is presented for proposed FC Framework to solve the
problem of cluster in-stability in VANETs. Using this approach, the clustering
of nodes and nomination of cluster heads is automated in VANETs.
• Our proposed approach is lightweight and semi-distributed, and allows faster
convergence. Our proposed approach reduces the signalling overhead and com-
plexity, and increases cluster stability in large scale VANETs. In our proposed
54

Chapter 4 55
approach, significantly low signalling i.e., the average throughput of all clus-
ters, is handled in a centralized manner, and the decision-making process
(i.e., the automated adjustment and nomination of cluster heads) is performed
in a decentralized evolutionary fashion.
• The solution of the game is presented to be an evolutionary equilibrium. The
equilibrium point is also proven analytically and the existence of evolutionary
equilibrium is also verified using Lyapunov function.
• The proposed game is analysed with different number of clusters for different
populations and cost functions. An optimal cost is suggested that defines an
optimum clustering.
• We present two performance evaluation approaches to test and analyse the
behaviour and performance of our proposed game. Our first approach is based
on static scenarios and in our second approach, we use Manhattan grid as a
mobility model to analyse the behaviour of our proposed game.
The rest of the chapter is organized as follows. In Section 4.2, the current VANET
clustering schemes presented in literature are briefly reviewed and a summary of
VANET clustering problems is presented. Moreover some background on evolu-
tionary games is also discussed in this section. Section 4.3 presents the details of
our proposed EGT framework. Section 4.4 presents system model, solution ap-
proach and analytical proofs regarding convergence and stability of evolutionary
equilibrium. In section 4.5, simulation set-up scenarios, results and discussions are
presented. Finally, the work is concluded.
Due to an ever increasing demand on transportation management and safety in
Intelligent Transportation Systems (ITS), the need for an efficient data dissemina-
tion framework has grown to the point where it is clearly understood that many
future ITS systems should be developed with a stable underlying data communica-
tion network. To this end, clustering plays a vital role to provide an efficient and
steady state routing in VANETs [68], [69], [70], [71], [72]. Clustering has emerged

56 Chapter 4
as an important research topic in VANETs to organize and manage the network in
a more efficient way. Clustering can help different applications by improving the
reliability of the reported measurements. For example, several WSN applications
require an aggregate value to be conveyed to the observer thus reducing commu-
nication overhead in the network, leading to significant savings in resources [73].
In this case, sensors collect data of specific regions by providing more accurate in-
formation about their local regions. Other applications include habitat monitoring
applications [74], defence systems [75] and WSN routing [68]. Clustering vehicles
into different groups offers many benefits such as: stabilizing the dynamic topol-
ogy of VANETs, making an optimum utilization of network resources, improving
the routing efficiency by providing hierarchical routing, providing fast convergence
rates with minimum overhead and saving power consumption [70], [71], [76]. Clus-
tering improves network scalability of large scale VANETs by limiting the number
of globally propagating control messages. Moreover, stable clustering in VANETs
makes the dynamic topology of VANETs appear less dynamic and hence the struc-
ture of the network becomes more manageable. Research has shown that routing
on the top of clustering architectures is more scalable and stable as compared to
flat routing [77], [70], [73], [71]. Clustering in VANETs creates a hierarchy within
the network, which helps in reducing the routing overheads and contention during
route discovery and data forwarding. Clustering in VANETs is also considered to
be one of the control schemes used to organize/coordinate the media access and
to support reliable and scalable multihop communications in VANETs [78]. More-
over, clustering in VANETs can assist in providing supports for Quality of Service
(QoS) requirements for both delay tolerant (road and weather information) and de-
lay intolerant (safety messages) applications [79]. It is also shown that clustering
in VANETs can effectively reduce data congestion [80]. Several clustering schemes
have proposed to improve routing performance in VANETs [68], [73], [71], [81]- [82].
However, very few works have been performed for investigating the stability of the
clustering itself. Furthermore, many of the proposed clustering protocols are based
on greedy algorithms, which do not often provide an optimal/network-wide solution.
Hence, the highly dynamic intrinsic characteristics of VANETs seriously affects clus-
ter stability and results in frequent cluster reformation and reorganization in [8]. We

Chapter 4 57
believe that clustering strategies should consider a whole-of-network approach when
creating a hierarchy in the network. The benefit of such an approach is that the
overall routing performance and stability of the network would be improved. One
approach to achieve this is through the integration of game-theoretic strategies in
clustering algorithms. In this chapter, an EGT framework is proposed to model
the interactive decision making process between vehicular nodes in order to provide
stable and optimized clustering in VANETs as shown in Figure 4.1.
The proposed work investigates the performance of proposed protocol by providing
stable and optimized clustering. The payoff of the proposed game is determined by
the net utility. The utility of head is computed from the difference between total
throughput of the entire cluster and the cost function. Cost is defined as a func-
tion of cluster size. Shannon’s capacity is used to calculate the throughput of each
node in the cluster. The objective of the utility function is based on maximizing
the utility function. Each member is attached to one of the cluster heads which
provides the highest SNR to the member. This criterion applies to any propagation
scenario. We use cluster size to implement the cost function. This cost function is
implemented at different values of cluster size to achieve the objective of optimized
clustering for our proposed game framework. Further details about the payoff and
utility function are presented in section 3.1. In next section of this chapter, VANET
clustering protocols are discussed.
4.2.1 VANET Clustering Protocols
The VANET clustering protocols generally vary in the selection of metric for the
cluster formation [83]. The cluster formation (grouping of vehicles) is based on a
single metric and a multi metric cluster formation criteria. As illustrated in [70],
VANET clustering protocols are also categorized as Centralized Clustering and De-
centralized or V2V Clustering. In centralized clustering, the cluster formation is
achieved via Road side Units (RSUs) based on periodic message exchange between
RSU and the vehicular nodes. In Decentralized Clustering or V2V Clustering proto-
cols, the cluster head election and cluster formation is usually achieved via exchange
of Hello Messages between vehicles.
An overview of VANET clustering protocols is provided as follows.

58 Chapter 4
CH
CH
CH
CH
CH
RSU
CH
CH
CH
CH CH
CH
RSU
CH
CH
CH
Figure 4.1: Proposed EGT Framework

Chapter 4 59
1. Lowest ID Clustering: In [81], the cluster formation is done based on the
lowest ID. The mobile nodes broadcast beacon messages in which node IDs
are encapsulated. The node which has the lowest ID in its neighbourhood is
selected as the cluster head node, while the other nodes are selected as cluster
member nodes. This scheme does not take into account any of the dynamic
characteristics of the network (e.g. node mobility, or node degree).
2. Mobility based clustering: The MOBIC [84] scheme uses a signal power
level mobility metric to represent the relative mobility of nodes which are at
one hop distance. An aggregate local mobility metric is the basis for clus-
ter formation. When a mobile node receives two consecutive beacon messages
from its neighbouring nodes, it measures the relative mobility between the two
nodes as the ratio of the received signal strength of the new beacon message and
the received signal strength of the old beacon message. The mobile nodes then
calculate the aggregate mobility metric based on relative mobility. The mobile
node having the least aggregate mobility is selected as a cluster head node.
This scheme is most commonly used for comparison with other VANET clus-
tering protocols. In [85], they propose a distributed mobility based data clus-
tering algorithm. Aﬃnity PROpagation for Vehicular networks (APROVE)
that forms clusters with both minimum distance and minimum relative veloc-
ity between each cluster head and its members that helps to cluster nodes in a
distributed manner by assuming vehicles know their positions using GPS. It is
observed that APPROVE shows signiﬁcant improvement in cluster stability,
if compared with other scheme such as MOBIC. In [86], they propose two al-
gorithms named as Distributed Clustering Algorithm (DCA) and Distributive
and Mobility Adaptive Clustering DMAC. In these algorithms the nodes are
grouped by following a new weight-based criterion that allows cluster head
selection based on link quality and mobility-related parameters. The mobile
nodes having the highest weight are selected as cluster head nodes. The DCA
is used for clustering quasi static ad hoc networks, whereas DMAC algorithm
adapts to the changes in the topology of network due to the mobility of nodes,
therefore, more suitable for mobility based environments. In DCA, the weight
is calculated thus having the possibility to express preferences on which nodes

60 Chapter 4
are better suited to be cluster heads. In DMAC, each node reacts locally to
any variation in the neighbouring topology, by changing its role (either cluster
head or member node) accordingly. Moreover, It is also observed that the time
complexity of the DCA is bounded by a network parameter that depends on
the change in network topology rather than on the size of the network.
Another modified DMAC [87] is also proposed to improve the original DMAC.
The goal of this algorithm is to improve cluster stability by avoiding re-
clustering when two vehicles meet in different directions. The process of re-
clustering is avoided if vehicles are moving in opposite directions. For the
implementation of the modified features, each vehicular node needs to know
its current location, velocity and moving direction as received from GPS or
other location services. A new parameter called freshness is introduced for
excessive re-clustering. The value of this parameter is calculated between two
vehicular nodes by receiving hello messages and their movement direction data.
The time to live (TTL) parameter helps in the construction of multi-hop clus-
ters.
An Adaptive Mobility Aware Clustering Algorithm (AMACAD) is also pro-
posed in [88] that aims to accurately follow the mobility patterns of vehicles
in VANETs. This algorithm also tries to prolong cluster lifetime and reduce
global overheads. The clustering metric considers the current location, speed
and both relative and final destinations of vehicles.
Aggregate Local Mobility ALM [8] represents a new beacon based clustering
approach that uses aggregate mobility as a clustering metric. This clustering
protocol is aimed at prolonging the lifetime of a cluster in VANETs. The
ALM weight is calculated similar to [84] except the difference that instead of
using the Received Signal Strength RSS, which is highly unreliable, it uses
location information of nodes using GPS or any other location services. A sig-
nificant improvement in cluster lifetime and reduced node state/role changes
is observed as compared to previous popular clustering algorithms. In [89]
they proposed Density Based Clustering (DBC) to provide stable and long life
clustering with a complex metric which takes into account the density of con-
nection graph, traffic conditions and link quality for reliable communication.

Chapter 4 61
3. Direction based clustering: A direction based clustering approach was
proposed in [90] that is suitable for urban areas for VANETs. Vehicles are
grouped into a clusters based on the prediction of directions of vehicles before
intersections.
4. Multi-hop Clustering: Another multi-hop clustering scheme was presented
in [91] that uses relative mobility as a metric between vehicles that are at
multi-hop distance. In this scheme, a radio propagation delay based on bea-
coning is calculated at each node and is aggregated and propagated back to
other vehicular nodes. The node with smallest aggregate mobility value is
chosen as an appropriate cluster head. Moreover, cluster stability is increased
by postponing the process of re-clustering for some interval of time when two
cluster-heads come within the communication range of each other. The per-
formance of the protocol is evaluated using diﬀerent mobility models and by
using 2, 3 and 5 hop clustering. Results show that cluster life time is prolonged
using this scheme. In [92], the authors propose a clustering algorithm called as
Vehicular Clustering based on weighted Clustering Algorithm (VWCA) based
on a Weighted Clustering Algorithm technique (WCA). It consists of a com-
plex metric calculated from vehicle movement direction and the number of
neighbours that are based on dynamic transmission range. The VWCA tech-
nique is mainly aimed at improving the Cluster Head duration, membership
duration and security. In [93] they propose an adaptive service provider infras-
tructure for VANETs (ASPIRE) that focuses on local network criticality and
clustering in a distributed fashion. A fast randomized clustering and schedul-
ing algorithm called as Hierarchical Clustering Algorithm (HCA) is presented
in [94] that forms clusters in a hierarchical manner. HCA creates clusters
within a diameter of at most four hops. HCA is robust in a sense that it
does not rely on localization systems. In [95], a speed-overlapped clustering
method is presented for highways that deﬁnes stable and unstable clustering
neighbors depending on their speed and relative direction. A lane-based clus-
tering algorithm is presented in [75] that is designed to provide stability in
lifetime of clusters in urban scenarios. The process of cluster formation is
based on selecting a cluster head from the lane where there will be a high

62 Chapter 4
traffic flow. However, detecting the lane is a challenging task as each vehicle
is assumed to know its exact lane by using a supplementary system such as,
visual lane recognition, LIDAR etc. In [96] a new multi-metric cluster head
election scheme has also been developed. The vehicles having similar mobility
patterns, speed and travelling direction are grouped together within a cluster.
This creates more stable clusters with increased cluster lifetime. The above
presented clustering algorithms focus on different performance metrics and are
optimized for different goals and objectives such as cluster stability, overhead
minimization and fast cluster formation and etc. with the most predominant
among them being cluster stability.
There is a need to put more research efforts to refine and optimize the cluster
head election policy to present a stable clustering scheme. Moreover, there is
a need to develop a clear definition of the generic terms of performance evalu-
ation metrics (like cluster head stability, cluster head changes, average cluster
stability and etc.) for clustering algorithms with respect to VANETs to pro-
vide consistency between different scientific studies [83]. Moreover, multi-hop
and multi-homing capable clustering solutions require further research. In the
next section, we discuss some motivations of using game theory as a solution
approach for our proposed scheme.
4.2.2 Game Theory:
Game theory has emerged as a solution of various problems in the field of radio
resource management, network formation, admission control, network selection and
many others [97]. There are many games proposed in literature [98], but the game
we propose to solve the problem of unstable clustering in VANETs has the theoret-
ical grounds of EGT.
Our objective is to improve the throughput and stability of an entire network by
adequately clustering nodes and nominating cluster heads. Bearing in mind the
combinatorial nature of clustering and subsequently prohibitive complexity for cen-
tralized optimization, we propose to achieve the objective by formulating an evolu-
tionary game theoretic framework. One reason for adopting the evolutionary game
is due to the fact that each cluster forms a population, and the utility of the entire

Chapter 4 63
population is to be maximized. An evolutionary game is a population game, as op-
posed to many classical games, where each player selfishly maximizes its own gain.
Another reason is, because each individual node has little rationality in a large net-
work, such as VANETs, and the rationality of a node is based on instant knowledge
of the responsive strategies that all other nodes take. Unfortunately, the knowledge
grows with the network size and is impossible to acquire in practice. Replicator
dynamics is adopted in our evolutionary game theoretic approach, to automate the
nomination of cluster heads and refine the clusters, given little rationality at the
nodes. With an increase in the number of nodes to large scales, replicator dynamics
can help reduce overhead and complexity, and increase cluster stability. Specifically,
the clusters with low throughput can replicate those with high throughput by en-
couraging cluster-edge nodes to switch between clusters. Within a cluster, a node
maximizing the throughput of the cluster is nominated to be the cluster head. This
continues until the network stabilizes, i.e., no cluster would enlarge or shrink and
the cluster heads stop changing.
Note that cooperative games typically carried out in a centralized manner, have
also been used to solve coalition formation problems and distribute the total gains
to collaborative players in a fair fashion, such as the Shapley value method [98]. A
cooperative game is typically suitable for small numbers of players, as it requires enu-
meration of possible coalitions and evaluation of their corresponding worths/payoffs
to form coalitions. This is unsuitable for clustering in VANETs, where the network
can be large and the number of possible coalitions is combinatorial to the network
size. There are many games proposed in literature using EGT for solving differ-
ent application in wireless networks. In [99], the authors presented an evolutionary
game to model the problem of routing. There are other few approaches presented in
heterogeneous wireless access networks that considered pricing or cost as a mecha-
nism for resource allocation, admission control and network selection. Mainly three
different approaches namely auction based [100], optimization based [101] and de-
mand supply based [102] are applied to solve different problems in heterogeneous
wireless access networks. Another approach is presented in [100] in which the mo-
bile users used a bidding scheme for radio resource allocation from multiple radio
access technologies by informing the service providers about the price and quality

64 Chapter 4
of service requirements. The service providers make resource allocation decisions in
different wireless access networks to maximize the revenue. In short, cellular and
broadband wireless access traditional systems such as 3G and WiMax or upcoming
technologies like 5G or femtocell networks provoke a number of technical challenges
arising from competitive and cooperative behaviour from wireless devices, making
them potential candidates to be modelled using game theoretic tools [97].
4.3 Proposed EGT framework
In this section, an EGT framework is proposed to automate the clustering of nodes
and nominations of cluster heads in VANETs. Initialized by a randomly generated
clusters based on proximity, the proposed evolutionary game can achieve stable
clusters and cluster heads, as shown in Fig. 4.1. Our proposed solution can also
fine-tune the number of clusters between the games, until an adequate number of
cluster heads are achieved with the highest total capacity. The stability of proposed
EGT is also confirmed by Lyapunov stability analysis.
4.3.1 Proposed EGT Framework
The evolutionary game for clustering between vehicular nodes in VANETs is formu-
lated as G = N, H, S, uCh
, where N = {1, 2, ...., n} is the set of all vehicles and
H = {1, 2, ......, j} with j ⊂ N is the set of randomly deployed clusters.
Utility function:
The net utility of a vehicular node i playing a strategy si from the strategy set
S = {Ch, M} is determined by its payoff, where Ch indicates the strategy the cluster
head uses, and M indicates the strategy thecluster member uses. Depending on the
total throughput of the entire cluster, the net utility of a clusterhead i is defined by
uChi
= c1/n +
n

j=2
1/n
1
c j
+ 1
c 1
− pi(si) (4.1)
where n is the number of nodes within the cluster, si is the current strategy of node
i and pi(si) ≥ 0 is a cost function. Here, c1 is the link capacity between the cluster
head and the Road Side Unit (RSU), and cj is the link capacity between a member

Chapter 4 65
Components of proposed
EGT framework
Components of VANETs
EGT Framework An Evolutionary Game
Theoretic (EGT) frame-
work G = N, H, S, uCh
as
discussed in section 4.3.1.
Players set of clusters H =
{1, 2, ......, j} with j ⊂ N
and N = {1, 2, ...., n} is the
set of all vehicles
Population The population is assumed
to be ﬁnite as represented by
N = {1, 2, ...., n}
Utility function The net utility of a vehicle is
determined by its payoﬀ that
depends on total throughput
of entire cluster and cost as
mentioned in section 4.3.1
Objective of Utility func-
tion
The objective of a node play-
ing Ch is to maximize utility
uCh
of cluster.
Table 4.1: Basic components of proposed EGT with respect to VANET clustering

66 Chapter 4
Parameter Description
G = N, H, S, uCh
EGT game
N Total number of vehicular
nodes
S = {Ch, M} Strategy set for vehicular
nodes
si Current strategy of node i
N = {1, 2, ...., n} Set of vehicular nodes
H = {1, 2, ......, j} with j ⊂
N
Set of clusters
uCh
net utility of a cluster head
pi(si) Cost function
TTC Total throughput of cluster
c1 The link capacity between
the cluster head and the
Road Side Unit (RSU)
cj, j ⊂ N The link capacity between a
member j within the cluster
and the cluster head
dH The distance between the
cluster head and the RSU
dM,j The distance between a
member j from the cluster
head
Table 4.2: List of parameters

Chapter 4 67
j within the cluster and the cluster head. Consider prevailing contention-based
random access techniques, namely, CSMA/CA, as standardized in IEEE 802.11p
VANET. Transmission collisions need to be taken into the consideration of c1 and
cj. Any pair of nodes with the received powers higher than the detection sensitivity
(typically, around the noise floor) can suffer from transmission collisions. A pair of
nodes can transmit concurrently with negligible interference, if the received power
is lower and submerged in the noise. Exploiting Markov modelling techniques [103],
the link capacity can be readily given by
c1 =τ(1 − τ)Nh
log2(1 + φ(dH)),
cj =τ(1 − τ)Nh
log2(1 + φ(dM,j)
(4.2)
where Nh is the number of neighbors within sensitivity range of a node, dH is the
distance between the cluster head and the RSU, and dM,j is the distance between
a member j from the cluster head. For analysis tractability, herein every node is
assumed to have the same number of Nh neighbors within its sensitivity range. τ is
the transmit probability of the node as per timeslot. Only depending on Nh, τ can
be numerically evaluated a priori by solving [ [103], eqs. 27 28]. We note that
τ(1 − τ)Nh is a common constant coefficient across all clusters, and does not affect
clustering or the nomination of cluster heads. Therefore, it is suppressed in eq. 4.4
and onwards.
Objective of Utility function:
The objective of a node i playing a strategy Ch is to maximize the utility that
is represented by the total throughput of cluster as represented by eq. (4.1). The
expected utility of node acting as head is to evolve towards balanced network thus,
achieving high throughput. This objective reflects the benefit gained by a vehicular
node i to become a cluster head and the cost paid for resources for the cluster head.
Cluster Formation:
Each member node is associated with one of the cluster heads which provides the
highest SNR to the member node.

68 Chapter 4
Cost Function:
As discussed in utility function, the utility of cluster head is the diﬀerence between
the reward of a selected strategy and the cost incurred by the cluster head from RSU.
We use cluster size to implement the cost function. This cost function drives the
clustering towards adequate sizes by keeping the cluster sizes and usage of capacity
at an optimum.
Apply Proposed EGT
Cluster with minimum Throughput is
selected for cluster head re-election
Every member of cluster is checked
for Throughput suitability to become
a cluster head
Member with highest Throughput in
a cluster is selected as a cluster head
Calculate Total Throughput of
every cluster
Check Throughput
of Clusters
Take a tentative set of randomly
deployed clusters /Cluster
reformation
If Throughput is maximized
(system converges)
Apply Cost as a function of cluster
size
Pick
up
clusters
with
low
Throughput
Start
End
If
Throughput
of
clusters
is
not
maximized
Check Throughput of
members
Equilibrium
Figure 4.2: Flow Chart of Proposed EGT

Chapter 4 69
4.4 System Model and Stability analysis
4.4.1 System Model
In this section, a detailed explanation of the proposed game framework is presented.
List of used variables is also given in Tab. 4.2. A set of vehicles N = {1, 2, ......., n}
and clusters H = {1, 2, ......, j} with j ⊂ N are assumed to be deployed in a two
dimensional grid of roads

= [xmin, xmax] × [ymin, ymax]. We assume that the
vehicles act as clients and RSU acts as the receiver. The vehicle play their strategies
and make decisions, based on utility maximization of clusters, using a decentralized
evolutionary fashion. We analyse the long term behaviour of the interactions of
vehicular nodes in terms of automation of clustering of nodes and nomination of
cluster heads in VANETs. The flow chart of our proposed EGT is presented in Fig.
4.2.
4.4.2 Solution Approach:
We deal with the problem of finding an evolutionary Nash equilibrium as a solution
of our game. In a cluster, every node knows its neighbouring links and its own
links towards RSU or sink, and this information is available for every node. In our
proposed approach, the decision making process (i.e., the automation of adjustment
of cluster-heads and nomination of cluster-heads) is performed in a decentralized
evolutionary fashion. For game theory, many games need to have centralized
coordination like the one in coalition games. In proposed approach, centralized
signalling is used to reduce the signalling overhead and to have an overall ob-
servation of the system. Furthermore, significantly low signalling is used on the
overall throughput of the system. The RSU is broadcasting to all clusters to remove
instability. A vehicular node gradually learns and adapts some decisions, until it
reaches the point of evolutionary equilibrium that is a stable state with improved
throughput. The vehicles adapt their strategies from a finite set of action profiles
through payoff based strategy adjustment process [35]. In payoff based distributed
learning, at any stage t, the vehicles know only their own actions and payoffs from
t − 1 previous stage and vehicles have no information about the actions taken by
other vehicles. Therefore, at each time t ≥ 0, each vehicle i ∈ N selects an action

70 Chapter 4
profile si ∈ S to maximize its expected utility. At every time step t, this game is
repeated and after a sufficiently large number of repetitive stages, vehicles action
profile reaches an evolutionary Nash equilibrium. It is important to note here that
our proposed game is a non-cooperative game that is formulated based on the group
behaviour of the vehicles for cluster formation and reorganization rather than de-
pending upon the individual behaviour of nodes unlike other classical games. The
payoff of the vehicle is the total net utility of a group of vehicles in a cluster [104].
Therefore, no vehicle has an incentive to deviate unilaterally at the point of evolu-
tionary Nash equilibrium and at this state of the game, we achieve the objective of
stable and optimized clustering in VANETs.
4.4.3 Replicator Dynamics and Stability of evolutionary equi-
librium
In a dynamic evolutionary game, the strategy used by a vehicle from the population
can be replicated by other vehicles through information received via centralized
controller RSU. We use centralized signalling here to have an overall observation of
the payoffs of all vehicles at RSU. The RSU calculates the average payoff of entire
population of clusters and then broadcasts the information to all clusters. When the
process of replication takes place over time, this can be modelled by a set of ordinary
differential equations, called as replicator dynamics [98]. In our game, the replicator
dynamics can be derived for population share of each cluster that is cluster size.
In this scenario, a replicator with a higher payoff will replicate itself faster. The
replicator dynamic equation for analysing the size of cluster is given as
˙
pHi(t) = γpHi
(t)[uChi
(t) − Ū(t)] (4.3)
where γ is used to control the speed of convergence of strategy adaptation and
γ 0. pHi
(t) = ni/N denotes the proportion of vehicles choosing cluster Hi and
is also referred to as population share or the size of cluster. The population share
of clusters can be denoted by the vector pH = [pH1 , pH2 , ..., pHi
, ..., pHm ]. uChi
(t)
is the payoff to become a cluster head. Ū(t) represents the average payoff of the
entire population of clusters. The evolutionary equilibrium is defined as a set of
fixed points of replicator dynamics that are stable. These fixed points are obtained

Chapter 4 71
∂ph1 (t)
∂t
= γ
n1(t)
N

⎧
⎨
⎩
log2(1 + φ(dH(t))
n1(t)
+
n1(t)

j=2
1
n1(t)

k{H,(M,j)}
1
log2(1+φk(t))
− p1(t)
⎫
⎬
⎭
−
⎧
⎨
⎩
log2(1 + φ(dH(t))
n2(t)
+
n2(t)

j=2
1
n2(t)

k{H,(M,j)}
1
log2(1+φk(t))
− p2(t)
⎫
⎬
⎭

(4.5)
numerically and at these fixed points the rate of strategy adaptation γ is zero or
the first order derivative of the proportion of vehicles choosing cluster Hi is ˙
pHi
= 0.
None of the vehicles will change their strategy at these fixed points, since their payoff
is equal to the average payoff of the entire population of vehicles. The evolutionary
equilibrium of our proposed evolutionary game exists, while the equilibrium might
not be unique. Any initial random clustering leads to a stable clustering result,
although the stable clustering can be different due to different initial clustering. The
existence of an evolutionary equilibrium is of paramount importance to VANETs and
can be verified by using Lyapunov Stability analysis. Lyapunov function can be used
to evaluate the willingness of a node to deviate from a fixed point. To evaluate the
stability of fixed point say pHi
∗
, obtained by ˙
pHi
= 0, the eigenvalue values of the
Jacobian matrix that corresponds to the replicator dynamics need to be evaluated.
If all eigenvalues have a negative part, the fixed point is stable [105].
To simplify the problem we first investigate the stability of evolutionary equilibrium
for two clusters that is H = 2 (as often considered in [106] and [107]). We first
show that the average total throughput capacity TTC of a given cluster declines
monotonically with the increasing size of clusters. Now eq. 4.3 can be rewritten as
∂phi
(t)
∂t
= γ
ni
N

ci
ni
+
ni

j=2
1
ni
1
cj
+ 1
ci
− pi

−

k

ck
nk
nk

j=2
1
nk
1
cj
+ 1
ck
− pk

(4.4)
For analysing the stability of two clusters eq. 4.4 can be rewritten in the form of
eq. 4.5 and eq. 4.6. For H = 2, to ﬁnd the equilibrium point for n1 and n2, the

72 Chapter 4
∂ph2 (t)
∂t
= γ
n2(t)
N

⎧
⎨
⎩
log2(1 + φ(dH(t))
n2(t)
+
n2(t)

j=2
1
n2(t)

k{H,(M,j)}
1
log2(1+φk(t))
− p2(t)
⎫
⎬
⎭
−
⎧
⎨
⎩
log2(1 + φ(dH(t))
n1(t)
+
n1(t)

j=2
1
n1(t)

k{H,(M,j)}
1
log2(1+φk(t))
− p1(t)
⎫
⎬
⎭

(4.6)
above equations can be rewritten as
∂ph1 (t)
∂t
= γ
n1
N
f1(n1) = 0
∂ph2 (t)
∂t
= γ
n2
N
f2(n2) = 0
(4.7)
We are able to ﬁnd the equation for equilibrium point, since fi(ni) is monotonic
with respect to ni. Since n1 = 1 − n1, we can calculate the equilibrium point ne by
solving
f1(n1)
n1
=
f2(1 − n1)
1 − n1
(4.8)
The equilibrium point ne exists, since f1(n1)
n1
and f2(n2)
n2
are monotonic. Fig. 4.3 gives
an illustration of how we get the equilibrium point. The equation for equilibrium
point ne is given as
ne = n1 =
f1(n1)
f1(n1) + f2(1 − n1)
(4.9)
Figure 4.3: An illustraion of equilibirum point ne

Chapter 4 73
For our case, we define a candidate Lyapunov function as V (ni) = 1
2
(ni − ne)2
such
that value of V (ni) ≥ 0 around the equilibrium point ne of the system. To check
the first order derivative of V (ni) along the trajectories of system with respect to
time t, the equation is given by
∂V (ni)
∂t
=
∂V (ni)
∂ni
∂ni
∂t
(4.10)
Hence, we get the result as
∂V (ni)
∂t
= (ni − ne) γni (TTCi(ni) − TTCi
(1 − ni)) (4.11)
where TTCi is the payoff of cluster i and TTCi
is the average payoff of all other
clusters in the network. Fig. 4.4 shows the equilibrium point for the system. It
0 0.2 0.4 0.6 0.8 1
n1/N
0
50
100
150
200
250
300
350
400
450
TTC
TTCi
TTCi'
Figure 4.4: Equilibrium point for Population share ni/N
is observed that ∂V (ni)
∂t
≤ 0 for any 0 ≤ ni ≤ 1. It is clearly shown from Fig. 4.4
that as ni/N grows, the factor (TTCi(ni) − TTCi((ni)
)) and (ni − ne) will always
take opposite signs. The equality holds if and only if ni = ne as shown in Fig. 4.3
and Fig. 4.4. This conclusion may be generalized for H ≥ 2 for which Fig. 4.4 will
become multidimensional. For example, for H = 3, the condition of ∂V (ni)
∂t
≤ 0 for
all 0 ≤ ni ≤ 1 will still hold and the curves will be replaced by planes with a point
intersecting at equilibrium point ne. It is worth mentioning that the equilibrium
point ne would be different for different initial deployment and mobility of nodes

74 Chapter 4
in clusters. Since the value of fi(ni) depends on node positions and their mobility.
Therefore, a generalized expression for fi(ni) cannot be derived. Our above analysis
confirms that our proposed game is stable under any initial random deployment
of clusters and this is due to the monotonicity of average payoff TTC of clusters
as shown in Fig. 4.4. To check the stability of equilibrium, point and to find the
ni=ne ± σ
RSU
CH
ne
ni
σ
σ
Figure 4.5: Boundary of equilibrium in the region of ni = ne ± δ for all 0 ≤ ni ≤ 1
where δ 1
boundary conditions for the equilibrium point, we check the value of V̇ (ni) in the
region of ni = ne ± δ where δ 1 as shown in Fig. 4.5. Our analysis confirms that
by applying ni = ne ±δ to eq. 4.11 where δ 1, the condition V̇ (ni) ≤ 0 still holds,
when number of nodes is finite and ne is not an integer. By using TTC1 and TTC2
as
TTC1 =
1
n1(t)
⎧
⎨
⎩
log2(1 + φ(dH(t)) +
n1(t)

j=2
1

k{H,(M,j)}
1
log2(1+φk(t))
− n1(t)p1(t)
⎫
⎬
⎭

Chapter 4 75
TTC2 =
1
n2(t)
⎧
⎨
⎩
log2(1 + φ(dH(t)) +
n2(t)

j=2
1
1

k{H,(M,j)}
1
log2(1+φk(t))
− n2(t)p2(t)
⎫
⎬
⎭
Reforming above expressions in terms of ni
N
and 1−ni
N
we get
TTC1 =
1
n1(t)N
⎧
⎨
⎩
log2(1 + φ(dH(t)) +
n1(t)N

j=2
1
1

k{H,(M,j)}
1
log2(1+φk(t))
− n1(t)Np1(t)
⎫
⎬
⎭
TTC2 =
1
1 − n1(t)
⎧
⎨
⎩
log2(1 + φ(dH(t)) +
1−n1(t)

j=2
1
1

k{H,(M,j)}
1
log2(1+φk(t))
− (1 − n1)(t)Np2(t)
⎫
⎬
⎭
Further assuming
TTC1 =
1
n1(t)N
{Temp1 − n1(t)Np1(t)}
TTC2 =
1
1 − n1(t)
{Temp2 − (1 − n1)(t)Np2(t)}
Hence eq. 4.11 becomes
∂V (n1)
∂t
=
±δγn1
N
(Temp1 − n1(t)Np1(t)) −
±δγn1
(1 − n1)N
(Temp2(1 − n1)(t)Np2(t))
Applying n1 = ne ± δ.
∂V (n1)
∂t
=
±δγ
N

Temp1 − Temp2

ne ± δ
1 − ne ± δ

− (ne ± δ) N (p1(t) − p2(t))

(4.12)
It is clear from eq. 4.12 that ∂V (n1)
∂t
will be zero if δ = 0. At equilibrium point
Temp1 = Temp2

ne
1−ne

,hence eq. 4.12 becomes
∂V (n1)
∂t
=
±δγ
N

Temp2

ne
1 − ne

−

ne ± δ
1 − ne ± δ

(4.13)
From eq. 4.13 if we take positive sign of δ, applying n1 = ne + δ
∂V (n1)
∂t
=
+δγ
N

Temp2

ne
1 − ne

−

ne + δ
1 − ne − δ

76 Chapter 4
The equation yields a negative result since the factor

ne
1−ne

−

ne+δ
1−ne−δ

is less
than zero. In the same way if we take negative sign of δ, applying n1 = ne − δ eq.
4.13 becomes
∂V (n1)
∂t
=
−δγ
N

Temp2

ne
1 − ne

−

ne − δ
1 − ne + δ

The equation yields positive result since −δ 0. Hence the above mathematical
analysis confirms the stability of our proposed clustering game. Due to the discrete
nature of cluster size in VANETs a ping-pong effect is observed within a small local
area of equilibrium point that is usually common to evolutionary games [106] and
[105]. Results show that this ping-pong effect remains within a small neighbourhood
of equilibrium point ne and will never deviate from this area as shown in Fig. 4.6.
Hence our results strengthen the stability of evolutionary equilibrium of our proposed
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
n1/N
0
50
100
150
200
250
300
350
400
450
TTC
TotalTC1
TotalTC2
n1min
ne
n1max
ne
Figure 4.6: Stability of equilibrium point for n1/N within n1 = ne ± δ
game.
4.4.4 Complexity Analysis
To analyse the complexity of our proposed protocol, we consider a graph G := (N, H)
comprising a set N of vertices together with a set H ⊂ N × N of edges. There are a
total of HN
possible configurations of signalling. We analyse the number of control

Chapter 4 77
packets exchanged between RSU and cluster-heads. Let the average number of head-
RSU control packets be represented by S. Then the total number of subsets of S
is given by 2S
. Equating the total possible conﬁgurations of signalling to the total
number of subsets of S, we get
2S
= HN
S ≈ N log2(H) (4.14)
Hence, the complexity of control overhead for head-RSU signalling is O(N log2 H)
with respect to the number of nodes (N) and number of heads (H) which is also
supported by the result of the simulation in Fig. 4.17.
4.5 Simulation set up scenarios and results
We present two approaches to analyse the performance of our proposed game. Per-
formance evaluations for both static and mobility based scenarios are made via
simulations using MATLAB as shown by running simulations in Figs. 4.7 and 4.8.
The wireless standard used for simulation is IEEE 802.11p. The mobility model we
chose to run the set of experiments is the Manhattan grid model. This model oﬀers
(a) EGT simulation with
2 clusters
(b) EGT simulation with
5 clusters
(c) EGT simulation with
15 clusters
(d) EGT simulation with
20 clusters
Figure 4.7: Simulation snapshots using Static Scenarios
more realistic mobility patterns on streets and in urban areas. The geographical
area of VANET is partitioned into two dimensional bidirectional grids (assuming
two way roads). The grid of roads is placed after every 250m. Initially all the
vehicles are deployed randomly in an area of 1000X1000m. After a node begins to
move and reaches at the next intersection, the direction of vehicular node is decided
probabilistically. A node has 50% chance of continuing in the same direction and

78 Chapter 4
(a) EGT running simula-
tion with 2 clusters
(b) EGT running simula-
(c) EGT running simula-
(d) EGT running simula-
Figure 4.8: Simulation snapshots using Manhattan Grid Mobility
25% chance of turning to the west/South directions and an equal 25% chance of
turning to east/north direction. Vehicles are assumed to be randomly deployed in
the network. All vehicles act as clients and the RSU acts as a receiver. The details
of network simulation parameters are given below in Tab. 4.3. We first take a tenta-
tive set of randomly deployed cluster-heads in the network. All vehicles start at the
same time and the vehicles within the range of RSU establish connection with the
RSU. In the same way, each member joins one of the cluster heads which provides
the highest SNR to the member. We apply our proposed EGT approach based on
utility maximization by using eq. (4.1). We assumed cost p as a function of cluster
size and we analysed results on different values of cost function. We initially assume
different numbers of clusters, that is 2, 5, 10, 15, 20 and apply our proposed EGT
game to investigate the performance of clustering for both static and mobility based
scenarios.
The trajectory of evolutionary equilibrium in figures 4.9 and 4.10 shows that the
system converges to a certain point where the stability of clusters is retained and
the clusters evolve towards balanced sizes with converged average total throughput.
Moreover, at the point of evolutionary equilibrium, there is no more role switching
between vehicles (i.e., from cluster head to member or member to cluster head) takes
place. In the same way, we tested the system with different inputs of clusters as
25, 30, 35, 40... We investigate the point where the average total throughput is maxi-
mized both for static and mobile scenarios as shown in Fig. 4.13 and Fig.4.14. This
is the point where we achieved the optimum number of clusters for our proposed
scenario as the throughput is maximized at this point. It is also observed that we

Chapter 4 79
Parameter Static Scenarios Mobility
Length of Road 1000m 1000m
Number of Vehicles 100 100
Position of RSU x = 500, y = 500 x = 500, y = 500
Transmission range
of RSU
500m 500m
Mobility Model Random and Static Manhattan Grid
Mobility
PHY and MAC layer
protocol
IEEE 802.11p IEEE 802.11p
Normalized Trans-
mit power PTx
20mW 20mW
Rxnoise(−90dBm) 1e − 9mW 1e − 9mW
Wavelength λ 0.125m 0.125m
Average Speed of ve-
hicles
0m/Sec 20m/sec, 30m/sec,
45m/sec, 65m/sec
Simulation interval .01sec .01sec
Table 4.3: Network conﬁguration parameters in static scenarios and mobility using
Manhattan grid

80 Chapter 4
0 5 10 15 20
Number of Evolutions
0
20
40
60
80
100
120
140
160
180
200
Total
throughput
Capacity
of
Clusters
No. of clusters=15, N=100
Figure 4.9: Stability convegence of System with 15 clusters in static scenario
0 5 10 15 20
Number of Evolutions
0
1
2
3
4
5
6
7
8
9
10
Total
throughput
Capacity
of
Clusters
No. of clusters=15 ,N=100
Figure 4.10: Stability convergence of System with 15 clusters using Manhattan grid
mobility

Chapter 4 81

!

#
$
%

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

No. of Switches

Figure 4.11: Comparison of Switching rate of Proposed EGT with ALM [8]
50 100 150 200 250
No. of Nodes
5
10
15
20
25
30
35
Average
Switching
PROPOSED EGT
ALM [8]
Figure 4.12: Comparison of Average switching rate of proposed EGT with ALM [8]

82 Chapter 4
2 6 10 14 18 22 26 30 34
Number of Clusters
0.055
0.06
0.065
0.07
0.075
0.08
0.085
Average
Total
Throughput
of
Clusters Average Total Throughput, N=100,
Speed= 0m/sec (Static)
Average Total Throughput,
N = 100, Speed = 30m/sec
Figure 4.13: Throughput maximization for static scenario and Manhattan grid
2 6 10 14 18 22 26 30 34
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
Number of Clusters
Average
Total
Throughput
of
Clusters
N = 100, Speed = 0m/sec (Static)
N = 100, Speed = 30m/sec
Figure 4.14: Optimum no. of clusters for static scenario and Manhattan grid

Chapter 4 83
0 5 10 15 20 25 30
Number of Clusters
0.055
0.06
0.065
0.07
0.075
0.08
0.085
AverageTotal
Throughput
Capacity
of
Clusters
65m/sec, N=100
30m/sec, N=100
20m/sec, N=100
45m/ sec, N=100
Figure 4.15: Comparison of Throughput maximization at diﬀerent speeds for N=100
0 5 10 15 20 25
Number of Clusters
0.06
0.065
0.07
0.075
0.08
0.085
Average
Total
Throughput
Capacity
of
Clusters
Speed=20m/sec, N=200
Figure 4.16: Comparison of Throughput maximization at diﬀerent speeds for N=200

84 Chapter 4
100
101
102
Number of heads
101
102
103
104
10
5
106
Control
Overhead
Number of Nodes = 81
Head-RSU (worst case)
Head-RSU (average case)
Head-RSU (best case)
N
1
log
2
H
100
101
102
Number of heads
102
10
3
10
4
105
106
Control
Overhead
N1
log2
H
100
101
102
Number of heads
10
3
104
10
5
106
Control
Overhead
N
1
log2
H
Figure 4.17: Complexity Analysis
5 10 15 20 25 30
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Average
Total
Throughput
No. Of Clutsers
N=20
N=50
N=100
N=200
N=150
Figure 4.18: Scalibility anaylysis for thorughput maximization at diﬀerent popula-
tion sizes

Chapter 4 85
30 40 50 60 70 80 90 100 110
7.8
7.85
7.9
7.95
8
8.05
8.1
8.15
Average Speed of Vehicles (m/sec)
Throughput
Figure 4.19: Throughput Vs Speed
get better results at higher price pi = 0.5 applied at cluster sizes of 5 or less. There-
fore, for an optimum use of bandwidth allocated by the RSU, better performance is
achieved at higher price by achieving utility maximization for the optimum number
of clusters in our network. We also investigate our proposed game at different speeds
for different population sizes, such as N = 100 and N = 200. Our results conclude
that the average throughput of clusters is maximized and system converges at dif-
ferent speeds as shown in Fig. 4.15 and Fig. 4.16. This shows that the resulting
protocol is extremely efficient and robust and is capable to deal with different levels
of speeds. Fig. 4.19 shows the throughput graph of clusters that shows a decrease in
throughput with increasing speed. Moreover, Our proposed game is also analysed
for scalability at different population sizes, such as N = 20, N = 50, N = 100,
N = 150 and N = 200. Simulations reveal that the evolutionary convergence of the
clusters in a network of different population sizes can be achieved within hundreds
of milliseconds, as shown in Fig. 4.18. It is worth mentioning that the numbers of
nodes N and heads H can be time-varying in practice. Nevertheless, given N and H,
the proposed evolutionary game stabilizes fast and the clusters are formed quickly in
a distributed, automated fashion. In most cases, the network topology of a VANET
would have barely changed before the proposed game stabilizes. Hence, our results
conclude that the average total throughput is maximized and system converges at

86 Chapter 4
different population sizes and therefore, we are able to get the optimum number of
clusters at different population sizes as shown in Fig. 4.18. Therefore, our simula-
tion results in Figs. 4.15, 4.16 and 4.18 reveal that the system converges at different
speeds and population sizes. We also compare our proposed game with Aggregate
Local mobility (ALM) [8] as an existing clustering strategy in VANETs. Our results
in Fig. 4.11 clearly demonstrate that almost 40% of switching rate of nodes (change
of roles from heads to member or member to heads) remains zero as compared to
ALM. Moreover, using our proposed EGT, rate of switching is almost reduced by
50% as compared to ALM. Average switching rate is also reduced as compared to
the compared clustering strategy as shown in Fig. 4.12. Therefore our proposed
game ensures more stable clusters with increased cluster lifetime as shown in Fig.
4.11 and Fig. 4.12. Moreover, to study the complexity of our proposed protocol, we
calculate the complexity of control packets exchanged between RSU and heads. The
control overhead analysis is conducted using Monte Carlo method. Our simulation
results in Fig. 17 show that for a given number of nodes N, the control overhead
increases logarithmically (i.e., O(N log2 H), as the number of heads H increases.
Our results also show that all cases i.e., best case, worst case and average case scale
equally with the number of nodes. Hence, our proposed protocol is lightweight and
computationally efficient, as the complexity of control overhead grows very slowly.
4.6 Conclusion
In this chapter, an EGT approach is proposed for FC framework for Stable and Op-
timized Clustering in VANETs. Our proposed framework is able to maintain cluster
stability, as the clusters evolve towards balanced sizes and system is converged with
an average total throughput of clusters.
The equilibrium point is proved analytically and the stability of equilibrium point
is also tested using a Lyapunov function. Two performance evaluation approaches
are used in this chapter, to investigate the efficiency of our proposed game under
different populations and speeds.
The performance of proposed evolutionary game is empirically investigated with
different cost functions using static and mobile scenarios.

Chapter 4 87
It is concluded from simulation results that the proposed protocol can create more
stable clustering and is able to achieve optimum clustering by using the cost func-
tion. It is also concluded from simulations that the proposed protocol is robust and
is eﬀective for diﬀerent populations and speeds of vehicles.

Chapter 5
A Hybrid-Fuzzy Logic Guided
Genetic Algorithm (H-FLGA)
Approach for Resource
Optimization in 5G VANETs
5.1 Introduction
T o support diversified quality of service (QoS) demands and dynamic resource
requirements of mobile users in 5G driven VANETs, network resources need more
flexible and scalable resource allocation strategies. Current heterogeneous vehicular
networks are designed and deployed with a connection-centric mindset with fixed
resource allocation to a cell regardless of traffic conditions, and static coverage and
capacity.
In this Chapter, a Hybrid-Fuzzy Logic guided Genetic Algorithm (H-FLGA) ap-
proach is proposed over our proposed 5G VANET architecture in chapter 3, to
provide an efficient resource allocation in 5G driven VANETs.
The idea behind using Fuzzy logic is to make the protocol more suitable for partic-
ularly implementing customer-centric network infrastructure with varying types of
service requirements. Since, fuzzy logic is flexible and tolerant of handling imprecise
data and contradicting inputs, the Fuzzy Inference System (FIS) rules can handle
the dynamic customer needs in a highly dynamic environment of the VANETs, by
88

Chapter 5 89
providing a flexible and optimum solution [108], [109]. The proposed protocol is flex-
ible and is a multi-criteria scheme optimized by using the fuzzy logic. Fuzzy logic
is used to make the decision on the appropriate weightage of different objectives
which will help the providers to tune the protocol to work for different scenarios by
modifying the fuzzy membership functions and fuzzy rules. The major contributions
and the results of this chapter can be summarized as follows;
• The proposed Hybrid-Fuzzy Logic guided Genetic Algorithm (H-FLGA) allows
the network service providers to implement a more customer-centric network
infrastructure thus improving their spectral efficiency. The network can auto-
matically adapt to dynamic customer needs and capacity demand fluctuations
of mobile users in VANETs. To the best of our knowledge, this is the first work
in the area of 5G driven VANETs, which uses a hybrid Fuzzy Logic guided
GA approach for resource optimization.
• Five different scenarios of resource optimization are formulated in this chapter
which focus on different network aspects, such as, capacity, minimising number
of FC-BBUCs, minimising delay, the number of FC-ZCs which one BBUC
handles, the traffic load of each FC-ZC and consequently of each BBUC Pool.
In addition, this approach supports energy efficient optimization for service
providers, as some idle BBUC’s may be switched off without any adverse
effect on the overall system thus reducing OpEx.
• Realizing the service oriented view, input and rules of the proposed Fuzzy
Inference System (FIS) are defined, for optimizing weights of multi-objectives,
depending on the Type of Service (ToS) requirements of customers. Using
proposed FIS, different options are weighted and multi- objective weights are
optimized, to provide the optimal solution.
• The results of the proposed hybrid H-FLGA approach are compared with the
GA and the 5G driven VANET architecture in [9].
The remainder of the chapter is organized as follows: Section 5.3 provides some
challenges and the key enabler technologies for 5G Driven VANETs. Section 5.4

90 Chapter 5
describes formulation of resource optimization scenarios in 5G driven VANETs, sec-
tion 5.5 explains details of proposed H-FLGA approach. Section 5.6 provides the
results and discussions and finally, the results of the chapter are concluded.
In the recent years, VANETs are expected to utilize 5G cellular networks to deliver
broadband services and enhance traffic and road safety to the users. In the next
few years, there will be a dramatic increase in Machine-to-Machine (M2M) com-
munication due to the massive diffusion of Internet of Things (IoT) traffic. This
dramatic increase will boost innovation and generate economic growth across wide
range of verticals such as automotive, energy, media, food and agriculture, health-
care, management, manufacturing, public transportation [110]- [111]. On the other
hand, Vehicular Social Networks (VSNs) [112] are also emerging where passengers
can share user centric information with each other using mobile devices and can
exchange data related to infotainment, utility, and emergency services [113]- [114].
By 2020, smarter and secure Intelligent Transportation Systems (ITS) are expected
to be operational as a VANET cloud [115]. With this view, the emerging scenario
of VANET implementations is expected to be heterogeneous in terms of resources,
network topology, contents [116] and traffic types (including legacy voice and data
traffic, as well as those generated by emerging M2M connections), all with different
quality-of-service (QoS) requirements [117]. Also, current heterogeneous VANET
architectures using cellular systems such as 4G and recent LTE Advanced systems
have been designed and deployed with a connection-centric mindset with fixed re-
source allocation to a cell regardless of traffic conditions, static coverage and capac-
ity [118], [116], [119]. Furthermore, they lack in flexibility to efficiently deal with the
data off-loading over different access networks [120] and to provide reconfigurability
of RAN equipment to adapt to varying traffic and QoS demands of users. In order
to support the exponential growth of heterogeneous mobile data traffic of new ITS
applications and to support a platform for IoT applications and social networking, a
radical rethink of current VANET architecture is essentially required. According to
our vision, this evolution can only be achieved by turning it into a more flexible and

Chapter 5 91
programmable fabric, through technological improvements enabled by next gener-
ation emerging technologies like Cloud-RAN, Software Defined Networking (SDN)
and Fog Computing, which can jointly be used to provide a multitude of diverse
services and resource sharing over a common underlying physical infrastructure. In
our previous study, we proposed a 5G driven VANET architecture in [121], which of-
fers more flexible and programmable fabric, leveraging the concepts of SDN, C-RAN
and Fog Computing. In this study, we propose a hybrid optimization approach over
our 5G VANET architecture, to provide an efficient resource allocation using Fuzzy
logic guided Genetic algorithm. Fuzzy logic is one of the most well-known tools used
to solve problems in dynamic and constantly changing systems. To address decision
making process in VANETs, fuzzy logic has been used in different scenarios such as
a broadcast protocol in Vehicular Ad hoc Networks where the fuzzy logic system de-
cides if the node is required to rebroadcast or not [108]. In [109], a fuzzy logic-based
scheme is proposed in VANETs to select backbone nodes, which consider the veloc-
ity of vehicles, the number of neighboring vehicles moving in the same direction and
the height of the antenna. The idea behind using Fuzzy logic is to make the protocol
more suitable for particularly implementing customer-centric network infrastructure
with varying type of service requirements. Furthermore, given the large number of
combinations of linking FC-ZCs with BBUCs and capacity demand fluctuations, in
our proposed architecture [121], an efficient resource allocation becomes increasingly
difficult to tackle, using the conventional brute-force techniques. Since fuzzy logic
is flexible and tolerant of handling imprecise data and contradicting inputs, using
a hybrid Fuzzy logic guided Genetic Algorithm approach can provide us a better
solution.
5.3 Challenges and Key enabler Technologies for
5G Driven VANETs
One of the promising techniques to support 5G cellular networks is Ultra Dense
Networks (UDNs) [122], in addition to macro cells which provides wide cover-
age [123], [119]. By deploying more small cells in a fixed region, the average dis-
tance between the users and the BS can be significantly reduced and hence system

92 Chapter 5
capacity can be increased by improving the spatial reuse of radio resources. Also,
to mitigate the drastic interference generated by the neighboring small cells, the
inter-cell interference coordination (ICIC) scheme of current 4G cellular networks
assigns different blocks of resources to cell-edge user equipment (UE) from neigh-
boring cells. However using this scheme the Base stations cannot make effective use
of resources of neighbor cells, when there are no cell-edge UEs in the neighboring
cells. Hence, another challenge is to design efficient dynamic Radio Resource Man-
agement (RRM) in 5G networks which will adapt to distinct traffic and interference
variations in small cells [123]. Different approaches namely auction based [124], op-
timization based [125], demand supply based [126], Evolutionary Game Theoretic
(EGT) based [19] are applied to solve different optimization problems in heteroge-
neous wireless access networks and VANETs. Furthermore, current Heterogeneous
VANET (Het-VANETs) implementations allocate fixed resources to a cell regard-
less of traffic conditions in other cells. To achieve these goals in 5G VANETs, more
flexible and optimal resource allocation methodologies must be devised to enhance
network capacity for highly mobile users, by keeping in view the different QoS re-
quirements of users.
Also, mobile operators are constrained by the inflexibility and reconfigurability of
Radio Access Network (RAN) equipment with respect to distinct traffic and QoS
demands of users. To meet these challenging requirements a revolution of tech-
nologies in both Radio access networks and the mobile core network is required.
Cloud-RAN has recently been identified as a leading candidate for 5G mobile net-
work architecture which enables the sharing of network resources in a centralized
data center, being cost-effective to operators, and enhances the spectrum efficiency
of next generation networks [119], [127]. In C-RANs, a large number of low-cost
Remote Radio Heads (RRHs) are randomly deployed and connect to the Base Band
Unit (BBU) pool through the fronthaul links. The operations of RRHs and the
computing resources of the BBU pool can be dynamically controlled in order to
adapt to the capacity demand fluctuations, which leads to significant reductions
in capital expenditures (CapEx) and operating expenses (OpEx) with much higher
growth of revenue [46]. Additionally, C-RAN also allows integration of Long-Term
Evolution Advanced (LTE-A) technologies and evolutions of novel 5G radio ac-

Chapter 5 93
cess and WiFi [80]. On the other hand, Software Defined Networking (SDN) has
emerged as one of the possible solutions for combining the management of base sta-
tions and access networks due to the separation of control and data plane [3]. In
an SDN-enabled network, all devices are managed and controlled by a centralized
controller, and the network operators can dynamically assign network virtualization
strategies and forwarding rules to the controller instead of defining rules at different
devices [128]. In addition, SDN allows operators to quickly configure and deploy
new network services and provide fine-grained traffic engineering control for each
user, using a policy-based management paradigm running on commodity hardware.
For example, bandwidth allocation can be dynamically designed by operators on a
per-flow basis instead of using generic origin-destination criteria [129] and operators
can employ different policies for diversified service demands of users.
In 5G C-RANs, resource allocation and the RRH-BBU mapping problem has been
addressed in a number of research works in the literature [130]- [131], However,
in [130] a dynamic RRH-BBU mapping algorithm is developed. However, the ser-
vice provider’s profit is not a focus of attention. Similarly a resource allocation
problem with a bargaining solution is proposed in [132] by employing the news-
vendor game model. However, this scheme requires additional time for resource
reconfiguration, which can deteriorate QoS requirements. In SDN based VANETs,
resource management and allocation are very important since they can significantly
affect the QoS and resource utilization. However, the relationship between QoS
satisfaction and resource limitation including the interaction among various types
of resources have not yet been fully studied, due to the new hierarchical framework
of SDN [133]. There is a need to develop flexible and scalable resource allocation
strategies to support diversified QoS demands in VANETs.
Furthermore, the Ultra-Dense Networks (UDN) are also envisioned to be a highly
promising technology used to enhance network capacity and spatial multiplexing.
Some state-of-the-art research works in UDN and Computation offloading in the
field of VANETs can be found in [134], [135]. The authors study the MECO prob-
lem in UDN and propose a heuristic greedy offloading scheme [136].
Furthermore, collectively SDN and C-RAN will provide service providers with an
opportunity to implement a more customer-centric network infrastructure, where

94 Chapter 5
the network can automatically adapt to dynamic customer needs and capacity de-
mand fluctuations of users in VANETs. The success of virtualization and cloud
technologies provides one of the possible solutions. However, there are many direc-
tions needed to investigate to support SDN based VANETs. SDN based migration
is inevitable and unless a network is built from scratch, there is a need to manage
both legacy and SDN based framework ensuring service delivery and performance
across all domains. The key to all of this is going to be the availability of inter-
operable virtual network functions (VNFs). Furthermore, due to the exchange of
security related data between the vehicles and the RSUs over a separate channel also
impose different challenges such as identity protection and data integrity because of
the expected heterogeneous network architecture in 5G networks [137].
Recently, Mobile Edge Computation offloading (MECO) is also emerging as a key
technology toward 5G to achieve lower latency and higher reliability [138]. However,
the existing MECO research only focus on the resource allocation between the Mobile
devices (MDs’) and the MEC servers and ignored the huge computation resources
in the centralized cloud computing centers. With increase in growth of mobile ap-
plications and MDs’, the resource bottleneck of MEC servers has been becoming
more and more prominent, and is affecting the network operators’ capital expendi-
ture (CapEx) and operating expense (OpEx). In [138] the problem of collaborative
computation offloading with centralized cloud and multi-access edge computing is
studied. Similarly in [139], they studied the collaborative task offloading problem
in vehicular edge computing networks to fully utilize the computing resources of the
remote cloud center and MEC servers. In [140], a distributed and adaptive resource
management controller is designed and tested, which allows the optimal utilization
of Cognitive Radio and soft-input/soft-output data fusion in VANETs
In our opinion, since fuzzy logic is based on natural language and is tolerant of
handling imprecise data, hence, combining Fuzzy logic with GA can provide us
with a better solution for optimal resource utilization in VANETs. Furthermore,
in a highly dynamic VANET environment, an optimal solution is dependent on the
network environment such as bandwidth, vehicle mobility and link status. The so-
lutions based on any mathematical modelling are non-flexible and complex to derive
for rapidly changing environments [141], [142]. Therefore, we use a hybrid approach

Chapter 5 95
using Fuzzy Logic guided Genetic Algorithm for optimum resource allocation in
5G driven VANETs. Our proposed Fuzzy Inference system (FIS) is used to opti-
mize weights of multi-objectives. These optimized weights are then used by Genetic
Algorithm to optimize connections between BBUCs and FC-ZCs.
5.4 Resource Optimization in 5G Driven VANETS
We propose an extension of our previously proposed 5G Next generation VANET
architecture [9]. In our proposed architecture, there are Fog Computing-Zone Con-
trollers (FC-ZCs), Fog Computing BBU Controllers (FC-BBUCs), Fog Computing-
Cluster-Heads (FC-CHs) and Fog Computing-Vehicles (FC-Vehicles). The purpose
of this study is to optimize the allowable connections between FC-ZCs and FC-
BBUCs and also to support cost and energy efficiency by switching off the idle FC-
BBUCs. In this section, we formulate five different scenarios of network resource
optimization in 5G driven VANETs.
Problem Formulation
Let ZC = {ZC1, ZC2, ..., ZCn} with cardinality |ZC| = nZC represents the set
of Fog Computing Zone Controllers (FC-ZCs) which are distributed in an area.
nZC is the number of FC-ZCs and nBBUC represents number of Fog Computing
BBU Controllers (FC-BBUCs). Let BBUC = {BBUC1, BBUC2, ..., BBUCn} with
cardinality |BBUC| = nBBUC represent the set of FC-BBUCs, such that nBBUC ≤
nZC. Let Links = {BBUCi, ZCj} represents the set of possible link pairs between
FC-BBUCs and FC-ZCs.
Variables
Zij =
⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
1, if ZCj is served by BBUCi
0, otherwise
where(i, j) ∈ Links
(5.1)
Yi =
⎧
⎨
⎩
1, if BBUCi is Chosen
0, otherwise
(5.2)

96 Chapter 5
5.4.1 Minimise the number of FC-BBUCs (Min-BBUC)
The objective of this problem is to minimize the number of FC-BBUCs serving FC-
ZCs which are requesting for resources as shown in Fig. 5.1. It is assumed that
FC-ZCs can connect to any of the BBUC pools, which means that there are no
restrictions concerning distance. Given as an input data to the problem includes:
the capacity row vector for BBUCs, capacity demand row vector for FC-ZCs and a
binary link matrix indicating allowable connections between BBUCs and FC-ZCs.
Objective function
The objective function is given by
Minimize CnBBUC
=
nBBUC

i=1
CBBUCi
Yi (5.3)
where nBBUC is the number of FC-BBUCs in the pool, CBBUCi
is ith element with a
value equal to the total available capacity (Aggregated Link Capacity) of FC-BBUCi
in capacity row vector CBBUC.
BBUC Pool 1
BBUC Pool 2
SDN Controllers
FC-ZC1
FC-ZC2
FC-ZC3
FC-ZC4
FC-ZCn
Figure 5.1: Minimize number of BBUC pools

Chapter 5 97
5.4.2 Minimize Delay (Min-Delay)
The objective of this problem is to minimize the delay by connecting FC-ZCs closer
to the possible BBUC Pool location, as illustrated in Fig. 5.2 by using the Min-
Delay algorithm. The SDN controller has all the possible locations of BBUC pools,
thus, knowing all the distances between possible link connections between FC-ZC
and BBUC. Given as an input data to the problem is the available capacity row
vector for BBUCs, capacity demand row vector for FC-ZCs, a binary link matrix
indicating allowable connections between FC-ZCs and BBUCs and cost associated
with each link. Since, the delay is considered to be directly proportional to the
distance between FC-ZCs and BBUCs which in turn is related to the cost associated
with linking BBUC and FC-ZCs.
Objective function
BBUC Pool 1
BBUC Pool 2
SDN Controllers
FC-ZC1
FC-ZC2
FC-ZC3
FC-ZC4
FC-ZCn
1
Figure 5.2: Minimize Delay
Minimize CDelay =
nBBUC

i=1
nZC

j=1
Costi,jZij (5.4)

98 Chapter 5
where nBBUC is the number of BBUCs in the pool, Costi,j is the link cost for linking
ZCs j and BBUCi in the cost matrix Cost.
5.4.3 Capacity Load Balance(Cap-LB))
The Cap-LB algorithm aims to balance the traffic load in every BBUC Pool. The
information of traffic load of each BBUC Pool is always available and updated by the
SDN controller. Before evaluating the decision, the controller has the information of
traffic load of the possible BBUC Pool connections. Thus, the controller will check
not only the maximum capacity limit of the BBUC Pool, but also the possible load
on the BBUC pool. With this approach, as illustrated in Fig. 5.3, it is guaranteed
that the BBUC pools have a capacity balance in what concerns to traffic load.
Objective function
Minimize CcapL
=
1
nBBUC

nBBUC

i=1
(D − Di)2 (5.5)
where Di =
nZC
j=1 ZijCZCj
is ith
element indicating the total load demand in BBUCi
in the total load demand vector D and D = 1
nBBUC
nBBUC
i=1 Di is the average load
demand across all BBUCs. The idea behind the objective function is to reduce the
standard deviations of the total load demand vector D. Under ideal conditions, if
the load demand is the same in all BBUCs, the positive objective function value
must be equal to zero.
5.4.4 Number of FC-ZCs per BBUC Balance Algorithm
(FC-ZC-per-BBUC-Bal)
The objective of this problem is to balance the number of connections to BBUCs
serving FC-ZCs which are requesting for resources using the FC-ZC-per-BBUC-
Bal algorithm. Under ideal conditions, the algorithm should produce a connection
arrangement for BBUCs and FC-ZCs such that the number of FC-ZCs in every
BBUC should be balanced as illustrated in Fig. 5.4. The information on the number

Chapter 5 99
BBUC Pool 1
BBUCPool2
40Mbps
SDN Controllers
FC-ZC1
FC-ZC2
FC-ZC3
FC-ZC5
FC-ZC6
50Mbps
10
Mbps
FC-ZC4
FC-ZC7
FC-ZCn
Figure 5.3: Capacity Load Balance
of FC-ZCs which each BBUC pool has, is always available and updated by the SDN
controller.
BBUC Pool 1
BBUC Pool 2
SDN Controllers
FC-ZC1
FC-ZC2
FC-ZC3
FC-ZC5
FC-ZC6
FC-ZC4
Figure 5.4: Number of FC-ZC per BBUC Balance

100 Chapter 5
Objective function
Minimize CZCperBBUC
=
1
nBBUC

nBBUC

i=1
(N − Ni)2 (5.6)
where Ni =
nZC
j=1 Zij is the ith
element indicating the total number of FC-ZCs
connected to BBUCi in the total connections vector N and N = 1
nBBUC
nBBUC
i=1 Ni
is the average number of FC-ZCS connected across all BBUCs. The idea behind
the objective function is to reduce the standard deviations of the total connections
vector N. Under ideal conditions, if the number of connection is the same in all
BBUCs, the positive objective function value must be equal to zero.
Constraints for Problem 5.4.1, 5.4.2, 5.4.3, 5.4.4
Problems 5.4.1, 5.4.2, 5.4.3, 5.4.4 are subject to the following constraints:
1. The Fog Computing Zone Controller ZC j should atleast be connected to one
FC-BBUC.
nBBUC

i=1
Zij = 1 (5.7)
∀j
{1, 2, ..., nZC} and ∀(i, j)
Links. nZC is the number of FC-ZCs and Zij
is an element of link matrix Z whose value is 1 or 0 if ZC j is connected to
BBUCi or otherwise, respectively.
2. Sum of link capacity demands of ZCs which are connected to the BBU con-
troller BBUC i, must be less than or equal to the available link capacity of
BBUC i.
nZC

j=1
ZijCZCj
≤ CBBUCi
(5.8)
∀j
{1, 2, ..., nZC} and ∀i
{1, 2, ..., nBBUC} . CZCj
is jth
element with a value
equal to the capacity demand for ZC j in capacity demand row vector CZC.
3. BBUC i should be serving if atleast one FC-ZC is connected to it.
Zij ≤ Yi (5.9)
∀j
{1, 2, ..., nZC} and ∀i
{1, 2, ..., nBBUC}.

Chapter 5 101
4. BBUC i should not be serving if no FC-ZCs are connected to it.
nZC

j=1
Zij ≥ Yi (5.10)
∀j
{1, 2, ..., nZC} and ∀i
{1, 2, ..., nBBUC}.
5.4.5 Constant Traffic Load (CTL)
CTL algorithm aims to force a constant traffic profile in every BBUC Pool, with
the objective to avoid traffic peaks, taking into consideration the three types of FC-
ZCs: Residential FC-ZCs, Commercial FC-ZCs and Mixed FC-ZCs as illustrated in
Fig. 5.5. Each FC-ZC has different traffic behavior throughout the day. To obtain
multiplexing gains and energy efficiency in a Cloud-RAN approach, as compared to
traditional RAN, an ideal BBUC Pool traffic profile should have a constant traffic
load throughout the day. This algorithm takes the selected hours of the day as
input parameters (i.e. a vector of integers ranging between 1 and 24). Since the
information on the traffic profile of each BBUC Pool is always available and updated
by the SDN controller. Before evaluating the decisions, the controller evaluates the
possible connections, by using a vector of hours. The controller establishes the
connection with BBUC Pool by considering the traffic load. This problem addresses
a time-series type of problem where the capacity demand of all the FC-ZCs are
given with respect to time (demand vs time (duration in hours)). Using CTL, it is
very effective to turn off some idle BBUC’s without any adverse effect on the overall
system thus consuming energy efficiently. Given as input data to the problem is the
capacity row vector for BBUCs, time-series capacity demand matrix for FC-ZCs, a
binary link matrix indicating allowable connections between BBUCs and FC-ZCs
and cost associated with each link.
Objective function
CCTL =
nBBUC

i=1
BTDi (5.11)
where Pi = [Zi×nZC
, Zi×nZC
, ...]T
for i
{1, 2, ..., nBBUC}; Qi = Pi × CZCH for i
{1, 2, ..., nBBUC};
Ri =
nZC
j=1 qj for i
{1, 2, ..., nBBUC} and qj is the jth
column vector in Qi; and

102 Chapter 5
BBUC Pool 1
BBUC Pool 2
SDN Controllers
FC-ZC1
FC-ZC3
FC-ZC4
FC-ZC6
FC-ZCn
FC-ZC5
FC-ZC2
Costant Traffic profile per BBUC
Figure 5.5: Constant traffic load per BBUC
BTDi = standard deviation of Ri for i
{1, 2, ..., nBBUC}. Ri column vector essen-
tially contains the total demand in the BBUCi with each row corresponding to a
duration interval. BTDi is the ith
value indicating the standard deviation of the
variation of total demand in BBUCi during the course of its operation in BBUC
traffic deviation vector BBU-Traffic-Dev BTD.
Constraints
1. ZC j should at least be connected to one BBUC.
nBBUC

i=1
Zij = 1 (5.12)
where j
{1, 2, ..., nZC} and (i, j)
Links. Zij is an element of link matrix Z
whose value is 1 or 0 if ZCj is connected to BBUCi or otherwise, respectively.
2. Sum of capacity demands of FC-ZCs connected to BBUC i must be less than
or equal to the available capacity of that BBUC i.
nZC

j=1
ZijCZCj
(Z, CZCH) ≤ CBBUCi (5.13)

Chapter 5 103
where j
{1, 2, ..., nZC} and i
{1, 2, ..., nBBUC}. CZCj
is jth
element with a value
equal to the capacity demand for ZCj in capacity demand row vector CZC of
size 1 × nZC. CZCH is also a function of the link matrix Z and the capacity
demand matrix for FC-ZCs for all the durations. The capacity demands cor-
responding to each FC-ZC in the CZC are selected by the SDN controller as
follows;
• Find the Ri for each BBUC i. Note the FC-ZC connected to that BBUC
i.
• Find the position of the peak demand in Ri.
• Use the same position to choose the CZC value for the connected FC-
Zcs to BBUCi from the CZCH matrix. Repeat for all BBUCs till all the
elements in CZC capacity demand vector are obtained.
3. BBUC i should be serving if at least one FC-ZC is connected to it.
Zij ≤ Yi (5.14)
where j
{1, 2, ..., nZC} and i
{1, 2, ..., nBBUC}.
4. BBUC i should not be serving if no FC-ZCs are connected to it.
nRRH

j=1
Zij ≥ Yi (5.15)
where j
{1, 2, ..., nZc} and i
{1, 2, ..., nBBUC}
5.4.6 Multi-Objective Optimization
The multi-objective function is mathematically the algebraic sum of the post-operated
individual objective function with the possible lower limit of the value 0 and an up-
per limit of the value of 1. For the purpose of creating an impartial multi-objective
function, each of the individual objective functions described in previous sections is
normalized with respect to a factor (so as to obtain a minimum and maximum value
of 0 and 1, respectively) and a weight of unity is assigned to each individual objec-
tive function to demonstrate equal importance to each of the objectives. The inputs
of the problem include link capacity of BBUCs, time-series demands of FC-ZCs, a

104 Chapter 5
binary link matrix indicating allowable connections between BBUCs and FC-ZCs
and the cost associated with each link. The Objective functions from the previous
sections are given as
C1 =
nBBUC

i=1
CBBUCi Yi
C2 =
nBBUC

i=1
nZC

j=1
CostijZij
C3 =
1
nBBUC

nBBUC

i=1
(D − Di)2
C4 =
1
nBBUC

nBBUC

i=1
(N − Ni)2
C5 =
nBBUC

i=1
BTDi
(5.16)
Hence, the multi-objective function can be formulated as the algebraic sum of nor-
malized individual objectives functions with a weightage factor. It is given by
min Cobj (5.17)
where
Cobj =ω1
C1 − C1,min
C1,max − C1,min
+ ω2
C2 − C2,min
C2,max − C2,min
+
ω3
C3 − C3,min
C3,max − C3,min
+ ω4
C4 − C4,min
C4,max − C4,min
+
ω5
C5 − C5,min
C5,max − C5,min
(5.18)
where C1, C2, C3 and C4 are the Objective functions for the objectives laid out in
Sections 5.4.1, 5.4.2, 5.4.3, 5.4.4 and 5.4.5, respectively. In Equation 5.18, the max
and min subscripts indicate the maximum and minimum values, respectively, for
the objective functions when optimized individually. It is necessary to normalize
the cost function values for each individual objective function between 0 and 1.
ω1, ω2, ω3, ω4 and w5 are the weightage factors for each individual objective func-
tions. Importance to any objectives can be increased or decreased by the service
providers, by altering these values, depending on the dynamic customer needs and
capacity demand ﬂuctuations of users in VANETs. We use Fuzzy Logic to ﬁnd these
weights.

Chapter 5 105
5.5 Hybrid-Fuzzy Logic guided Genetic Algorithm
(H-FLGA)
We propose a Hybrid-Fuzzy Logic guided Genetic Algorithm (H-FLGA) approach
for SDN controller which solves a multi-objective optimization problem. Different
objectives are combined to assign most accurate practical connection arrangement
between BBUCs and FC-ZCs. Depending on the Type of Service (ToS) requirements
of customers, different options are weighted and optimized using Fuzzy Inference
System (FIS) and then used by GA to provide optimal solution.
Fuzzy Inference System (FIS)
Fig. 5.6 illustrates the flow chart of proposed algorithm. The fuzzy system has
two inputs; Type of service (ToS) and value (V alue). ToS is the requirement of
customers based on three parameters i.e., Throughout, Delay and Cost. The out-
puts of FIS is priorities coefficients ωi for optimized weights of the multi-objectives.
Hence, ω = f(Tos, V alue) where ToS = {Delay(D), Throughput(T), Cost(C)} and
V alue = {0, 1}. The outputs are in the range in [0, 1]. We choose the Gaussian mem-
bership function for the inputs and outputs variables. Tables 5.1 and 5.2 show the
possible ToS values and ToS vs. Priority of ωi. To define the rules for each output,
we follow the information in tables 5.1 and 5.2.
Rules
The rules for ωi are as follows;
r1 =
if ToS is D and V alue is zero then ω1 is zero
;
r2 =
if ToS is D and V alue is not zero then ω1 is zero
;
r3 =
if ToS is T and V alue is zero then ω1 is medium
;
r4 =
if ToS is T and V alue is not zero then ω1 is high
;
r5 =
if ToS is C and V alue is zero then ω1 is medium
;
r6 =
if ToS is C and V alue is not zero then ω1 is high
;

106 Chapter 5
Table 5.1: Possible Type of Service (ToS) Values
Type of Service (ToS) Delay (D) Throughput (T) Cost (C)
Normal 0 0 0
Low 1 - 1
High - 1 -
Table 5.2: Type of Service (ToS) Vs. Priority ω for Fuzzy Inference System
ToS / Priority ω Delay D Throughput T Cost C
Min-BBUC(ω1) 0/Zero 0/Medium 0/Medium
1/Zero 1/High 1/High
Cap-LB (ω2) 0/Low 0/Medium 0/Low
1/Low 1/Medium 1/Low
Min-Delay (ω3) 0/High 0/Low 0/Low
1/High 1/Low 1/Zero
FC-ZC-per-BBUC-Bal (ω4) 0/Low 0/Medium 0/Low
1/Low 1/Medium 1/Low
CTL (ω5) 0/Low 0/Medium 0/Medium
1/Low 1/Medium 1/High
In the same manner, we define the rules for ω2, ω3, ω4 and ω5.
After the application of the Fuzzy Inference System, the re-evaluation part of
GA is executed where the fitness of the population is computed using multi objective
functions with the new weights. Details about different objectives are discussed in
section 5.4.

Chapter 5 107
Re-evaluation
Stopping
Criterion
Optimal Solution
Fuzzy Inference
System(Compute Z)
Create Initial Population
Evaluation (Compute the Fitness)
Selection
Fuzz
Fu
Crossover, Mutation
Output(Optimized Weights Z )
De-Fuzzification
Input Type of
Service(ToS)
0 Normal
1 Low
Rule 1
e 1
1
1
1
1
1
1
1 Rule 2 Rule n
Aggregartion
ow
Lo
Lo
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1 L
L
L
Fuzzification
NO
f
)
o
o
0
0 No
No
f
0
0
0
0 N
No
N
0 No
Delay
m
m
r
rm
m
r
y
m
m
rm
y Throughput
ut
t Cost
al
al
l
l
al
al 0 Normal
1 High
l
l 0 Normal
1 Low
If Satisfied
ti l S
yes
)
Fuzzy Inference System
for Optimizing weights
Figure 5.6: Flow Chart of Hybrid-Fuzzy Logic Guided Genetic Algorithm (H-FLGA)

108 Chapter 5
Algorithm 1 : H-FLGA
Input: link capacity of BBUCs’, demands of FC-ZCs, threshold of critical
demands, Type of Service (ToS) and priority V alue
Output: Optimized weights ω1, ω2, ω3, ω4, ω5, Optimized multi objectives
(C1, C2, C3, C4, C5)
Methods: evalfis(), H-FLGA()
H-FLGA:
(1) choose a max generation number GMAX and the cycle size C, set t = 0,
(2) Initialize randomly the population
(3) Compute the fitness for each individual in the population,
(4) Extract and save the current best individual in the population,
(5) if t ≤ GMAX Stop,
else set t = t + 1 and continue to step (5)
(6) Apply selection, crossover and mutation,
(7) if mod(t, C) = 0 back to step (2)
else Call Fuzzy Inference System (FIS) to compute new weights
(8) end
Complexity Analysis
We discuss the complexity of proposed algorithm with respect to algorithm com-
plexity of the proposed H-FLGA including the signaling overhead on the controller.
The complexity of the proposed approach is a function of GA and FIS denoted by
f(O(GA) O(FIS)). O(GA) depends on the operations: generate the first gener-
ation, selection, crossover, mutation and find the best individual. In addition, it
also depends on the number of generations GMAX. In the case of integer optimiza-
tion these operators are simpler to implement. Therefore, O(GA) mainly depends
on the complexity of the multi-objective cost function O(Cobj) and the number of
maximum generations GMAX. Furthermore, O(FIS)) depends on the number of
cycles nc of FIS. Hence, the combined complexity of the proposed algorithm can
be expressed as O(ncGMAXO(Cobj)). It is seen from results in Fig. 5.7 that the
objective is optimized between first 20 generations. Hence, we conclude that the
proposed algorithm is efficient and lightweight.

Chapter 5 109
Signaling Overhead
To analyze the signaling overhead on the controller, we consider a graph G :=
(nBBUC, nZC) comprising a set nZC of vertices together with a set nBBUC ⊂ nZC ×
nZC of edges. There are a total of nnZC
BBUC possible configurations of signalling.
Let the average number of BBUC-Controller control packets be represented by S.
Then the total number of subsets of S is given by 2S
. Equating the total possible
configurations of signaling to the total number of subsets of S, we get
2S
= nnZC
BBUC
S ≈ nZC log2(nBBUC) (5.19)
Hence, the signaling overhead on controller is O(nZC log2(nBBUC)). This shows
that the control overhead on the controller will increase logarithmically, as nBBUC
increases. Hence we conclude that, the complexity of signaling overhead on the
controller will grow very slowly.
5.6 Simulation Results and Discussions
The results for the proposed algorithm are simulated using MATLAB 2017b. We
used evalfis() to implement the Fuzzy Logic rules. The details of the FIS rules and
simulation parameters are provided in Table 5.2 and 5.3 respectively. We perform
a comparison of multi-objective optimization using the GA and the proposed H-
FLGA approach. The main metric for assessing the proposed algorithm is the value
of the multi-objective cost function that should lie between 0 and 5 as seen from
equation 5.18 using the proposed hybrid H-FLGA approach. We test our results
by optimizing the weights of different objectives in equation 5.18. These objectives
indirectly relate to five different resource optimization scenarios focusing on different
network aspects discussed in detail in section 5.4. First, we run the GA to solve
multi-objective optimization problem in Equation 5.18 with fixed values of weights
ωi where we assume that all of them are equal to one, by considering all cost func-
tions C1 to C5 as of equal importance. Fig. 5.8a to Fig. 5.8c show the results of
Min-BBUC. Our results show when ω1 is set to 1 by keeping capacity constraints
under consideration, the utilization of capacity is optimized and consequently the

110 Chapter 5
0 10 20 30 40 50 60
No. of Generations
0
2
4
6
8
10
Multiobjective
Cost
Function
FLGA
GA
Figure 5.7: Variation of multi-objective function value for different numbers of gen-
erations
number of BBUs are minimized. Fig. 5.9a to Fig. 5.9c show the results of Cap-LB,
when ω3 is set to 1 considering the constraints defined for Cap-LB. Fig. 5.10a to
Fig. 5.10c show the results of FC-ZC-per-BBUC-Bal, when ω4 is set to 1 considering
the constraints defined for FC-ZC-per-BBUC-Bal. Similarly Fig. 5.11a to Fig. 5.11c
show the results of CTL, when ω5 is set to 1 considering the constraints defined for
CTL. Fig. 5.7 shows the variation of the objective function for the most optimized
parameters using the H-FLGA approach. In the current study, a population size
of 1000 was used for each generation and the number of generations were increased
from 10 to 50 with an increment of 10. It is observed, increasing the number of gen-
erations optimizes the objective function value however, beyond 10 generations it is
observed that the function value remains more or less the same. Also, a statistical
study was carried out to determine the 95% confidence in the obtained solution with
the algorithm. It is observed that with lower generations, there is likely to be more
variation in the result, however, beyond 10 generations the confidence of obtaining
the optimum value is very high as there is no varying interval on those data points.
The value of cost function using GA is optimized to 10.32 and as the best score
of the multi-objective function. Whereas, the best score of multi-objective function
should be between 0 and 5 as seen from equation 5.18. Therefore, GA could not

Chapter 5 111
optimize the value of multi-objective cost function. Hence, to improve the results
of multi-objective cost function, we run our proposed hybrid H-FLGA approach
as a tool to optimize the weights in the multi-objective function. The results in
Fig. 5.7 shows the value of multi-objective cost function is minimized and is reduced
to an optimized value of 2.2 using the proposed H-FLGA. The results in Figs. 5.12a
to 5.12c show how different objectives are optimized using H-FLGA. Therefore, our
results in Fig. 5.7 prove that our proposed H-FLGA approach performs better when
compared with GA. This is because, the value of multi-objective cost function is re-
duced and minimized from 10.32 to 2.2 when we applied our proposed FIS rules. It
is worth mentioning that the value of optimized weights may vary depending on the
TOS values defined in FIS rules. Hence, depending on ToS requirements of different
customers, service providers can implement different FIS rules for different objec-
tives and assign different priorities of Low, Medium and High to TOS parameters
and get optimized weights. Hence, we conclude that our proposed hybrid H-FLGA
approach performs better than GA and is flexible to set weights of multi-objectives,
depending on QoS demands and requirements of different users.
Fig. 5.13 shows the variation of the end-to-end delay for the vehicles within a max-
imum front-haul distance of 40km using three schemes- H-FLGA, GA and [121].
The number of vehicles counts is increased from 50 to 300 with varying speeds. The
delay is the highest using [121] while lowest using the proposed H-FLGA. For each
data point in the graph, vertical markers are used to indicate the confidence of an
interval of 95%.However, [121] has the widest confidence interval compared to the
GA or the H-FLGA, which indicates, these is likely to be more variation in the delay
estimated using the [121] and the GA. Our results show that when ω2 is set to 1 and
solved only with GA, the maximum value of delay is 0.113s and the maximum value
of delay when calculated using [121] is 0.171s. However, the value of delay is low-
ered and improved to 0.062s when delay is computed using our proposed H-FLGA
approach.

112 Chapter 5
0
0.5
5
Active
[1]
/
Inactive
[0]
Optimum Connections Graph
1
1
4
BBUCs
FC-ZCs
2 3
3 2
1
(a) Optimum number of Connections using Min-
BBUC
1 2 3
BBUC
0
20
40
60
80
100
Total
Capacity
of
BBUC
Total Capacity of BBUC
Utilized Capacity of BBUC
(b) Capacity Utilization of BBUC using Min-BBUC
0
10
5
Capcity
Demands
of
FC-ZCs
20
1
30
4
BBUC
2
FC-ZCs
3
3 2
1
(c) Capacity Demands of FC-ZCs using Min-BBUC
Figure 5.8

Chapter 5 113
0
0.2
0.4
0.6
5
Active
[1]
/
Inactive
[0]
1
0.8
1
4
BBUC
2
FC-ZCs
3
3 2
1
(a) Optimum number of Connections using Cap-LB
1 2 3
BBUC
0
10
20
30
40
50
60
70
80
90
100
Total
Capacity
of
BBUC
(b) Capacity Utilization of BBUC using Cap-LB
0
10
5
Capacity
Demands
of
FC-ZCs
20
1
30
4
BBUCs
2
FC-ZCs
3
3 2
1
(c) Capacity Demands of FC-ZCs using Cap-LB
Figure 5.9

114 Chapter 5
0
0.2
0.4
0.6
5
Active
[1]
/
Inactive
[0]
1
0.8
1
4
FC-ZCs
BBUCs
2 3
3 2
1
(a) Optimum number of Connections using FC-ZC-
per-BBUC-Bal
1 2 3
BBUC
0
10
20
30
40
50
60
70
80
90
100
Total
Capacity
of
BBUC
(b) Capacity Utilization of BBUC using FC-ZC-
per-BBUC-Bal
0
10
5
Capcity
Demands
of
FC-ZCs
20
1
30
4
BBUCs
2
FC-ZCs
3
3 2
1
(c) Capacity Demands of FC-ZCs using FC-ZC-per-
BBUC-Bal
Figure 5.10
5.7 Conclusion
In this chapter, a hybrid Fuzzy Logic guided Genetic Algorithm (H-FLGA) approach
is proposed for the SDN controller, for an optimum resource allocation for our pro-
posed 5G-driven VANET architecture in Chapter 3.
A multi-objective optimization problem is solved, where diﬀerent objectives are com-
bined and proposed FIS is used to optimize the weights of multiple objectives. These
optimized weights are then used by the Genetic Algorithm to optimize connections
between BBUCs and FC-ZCs. This work will help service providers to improve their
spectral eﬃciency, where the network can automatically adapt to dynamic customer

Chapter 5 115
0
6
0.5
5
Active
[1]
/
Inactive
[0]
1
1
4
FC-ZCs
BBUCs
2 3
3 2
1
(a) Optimum number of Connections using CTL
0
6
50
5
Capcity
Demands
of
FC-ZCs
1
FC-ZCs
100
4
BBUCs
2 3
3 2
1
(b) Capacity Demands of FC-ZCs using CTL
1 2 3
BBUC
0
10
20
30
40
50
60
70
80
90
100
Total
Capacity
of
BBUC
(c) Capacity Utilization of BBUC using CTL
Figure 5.11

116 Chapter 5
1 2 3
BBUC
0
10
20
30
40
50
60
70
80
90
100
Total
Capacity
of
BBUC
(a) Capacity Utilization of BBUC using H-FLGA
0
0.2
0.4
0.6
5
Active
[1]
/
Inactive
[0]
Optimum Connection Graph
1
0.8
1
4
BBUC
2
FC-ZCs
3
3 2
1
(b) Optimum number of connections using H-FLGA
0
5
10
15
5
1
20
Capacity
Demands
of
FC-ZCs
25
4
FC-ZCs
BBUC
2
3
3 2
1
(c) Capacity Demands of FC-ZCs using H-FLGA
Figure 5.12: Multi objective Optimization using Optimized weights

Chapter 5 117
0 50 100 150 200 250 300 350
No. of Vehicles
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
End-to-End
(E2E)
delay
(s)
Ref [9]
GA
FLGA
Figure 5.13: End-to-End Delay
Table 5.3: Simulation Parameters
Simulation Parameter Value
Maximum capacity of BBUC pool 100MHz
Maximum Fronthaul distance 40KM
Number of Vehicles 50 to 300
Transmission range of vehicles up to 300m
Speed of Vehicles between 10m/s and 30m/s
MAC protocol IEEE 802.11(11Mbps)
Mobility Model Manhattan grid (2500m × 2500m)
Packet size 512bytes
Population size 2000
Tolerance for objective function 1e − 8
Crossover Operator single (or multi) point
needs and capacity demand ﬂuctuations of users in VANETs, being cost-eﬀective to
operators.
It is concluded from the simulation results that the value of the multi-objective cost
function is minimized using H-FLGA when compared with GA. It is also observed
that the proposed H-FLGA approach minimizes End-to-End delay as compared to

118 Chapter 5
the GA and the 5G driven VANET architecture.

Chapter 6
An End-to-End (E2E) Network
Slicing Framework for 5G
Vehicular Ad-hoc Networks
6.1 Introduction
T o accommodate high volumes of mission critical traffic, reserving radio resources
may lead to over-provisioning of resources [15]. There is a need to define resource
allocation strategies based on on-demand and instant allocations of resources. [16].
Network slicing is considered to be the key enabler to achieve high utilization of
both communication and computing resources and minimize the infrastructure de-
ployment cost [17]. Network slicing is expected to emerge as a promising solution for
end-to-end resource management and orchestration together with Software Defined
Networking (SDN) and Network Function Virtualization (NFV) technologies.
In this chapter, an E2E network slicing framework is presented with the consider-
ation of both the radio access network resources and the core network resources in
5G-driven VANETs.
The following are major contributions of this chapter.
1. A comprehensive network slicing framework is presented to achieve end-to-end
(E2E) QoS provisioning among customized services in 5G-driven VANETs,
119

120 Chapter 6
with the consideration of managing the cooperation of both RAN and Core
Network (CN) using SDN, NFV and Edge Computing technologies.
2. Furthermore, a dynamic radio resource slice optimization scheme is formu-
lated mathematically, which handles a mixture of both best-effort traffic and
mission-critical traffic, by keeping in view resource elasticity requirements
3. The solution adjusts the optimal bandwidth slicing and dynamically adapts
to instantaneous network load conditions such that a targeted performance is
guaranteed.
4. The problem is solved using a Genetic Algorithm (GA) and results are com-
pared with our previously proposed 5G VANET architecture in chapter 3.
Simulations results reveal that the proposed slicing framework is able to opti-
mize resources and deliver the targeted KPIs of mission critical demands.
The remainder of the chapter is organized as follows: Section 6.2 provides some
background on network slicing in 5G architecture. Section 6.3 describes network
slicing framework section 6.4 explains problem formulation. Section 6.5 provides
results and discussions and finally, Section 6.6 concludes the chapter.
The era of the fifth generation (5G) cellular networks is rapidly evolving. 5G net-
works are anticipated to support a number of vertical industries that are character-
ized by diversified use cases and applications. Some of the most popular use-cases are
Intelligent Transportation Systems, Mobile Broadband, Massive number of Internet-
of-Things (IoT), Mission-critical IoT, and e-health care systems, all require diverging
features and performance requirements in terms of latency, reliability, security and
policy control [143], [144]. A key emerging use case of intelligent transportation is
handling mission-critical traffic in 5G vehicular networks. Considering the features
of mission-critical traffic, not only the precise and timely delivery of information is
required, but also other stringent key performance Indicators (KPIs) such as ultra-
reliability and low latency communications need to be achieved [145]. The concept
of network slicing has been thoroughly studied in different domains [146]- [147]. In

Chapter 6 121
the SDN-enabled environments using NFV, virtual network functions (VNFs) (de-
coupled from the physical network infrastructure) are instantiated and are placed
on NFV nodes. In the core network, physical resources are abstracted and are al-
located to different Virtual Machines (VMs) hosting VNFs thorough virtualization
layer. These VNFs are programmable commodity servers running VMs and are
flexibility orchestrated to provide differentiated E2E services. Furthermore, to en-
sure reliable and timely delivery of mission-critical traffic over RAN alone is not
sufficient to guarantee the required end to-end reliability, since the Core Network
(CN) part has to also be considered. Particularly, the transport network plays an
important role in the successful delivery of KPIs, such as reliability. Both mission-
critical and best-effort traffic will have to travel through a complex network while
competing for resources during transmission, buffering, and computing. Moreover,
Mobile Edge Computing (MEC) technology is also introduced by the ETSI ISG on
MEC [148], that offers cloud-computing capabilities within the Radio access net-
work and an information technology service environment closer to end user devices
or edge. MEC environment offers benefits of ultra low latency, high bandwidth
utilization, and real-time access to RAN information that can be used by differ-
entiated services or applications and QoS optimization platforms. To achieve E2E
QoS isolation among different traffic flows, the set of network resources including
computing resources on NFV nodes and bandwidth resources on transmission links
must be carefully isolated [17]. This can be achieved by using the concept of net-
work slicing. A network slice is a collection of Core Network (CN) and Radio Access
Network (RAN) functionalities, configured to meet the diverse service requirements
in terms of functionalities (i.e., mobility, support and security) and delivery per-
formance (i.e., throughput, reliability and latency) [149]. CN Network slicing is
interpreted as bi-resource slicing [150]. Whereas, RAN slicing mainly deals with
how to slice the overall radio resources. In RAN slicing, the radio access functions
on each Base Station are softwarized and are centrally managed by the SDN en-
abled virtualization controller. The SDN controller determines the amount of radio
resources to be allocated to each BS to enhance the overall spectrum utilization.
Existing studies presents network slicing solutions for 5G network domains across
Vehicle-to-everything V2X [146], RAN segments [151]- [152] and CN segments [153].

122 Chapter 6
Most of the studied issues related to network slicing are currently under investiga-
tion and particularly limited research focuses on determining the sets of resources for
customized services, to achieve a desired trade-off between high resource utilization
and end-to-end QoS isolation by considering heterogeneous resources. However, to
the best of our knowledge, this is the first time the concept of network slicing is
studied to achieve end-to-end QoS provisioning among customised services in 5G
Vehicular Ad-hoc Networks.
6.2.1 Network Slicing in 5G Architecture
With the evolution of 5G technology, network slicing has emerged as a major new
networking paradigm, to support a wide range of verticals with a diverse set of per-
formance and service requirements. Different network operators and vendors are all
recognizing it as an ideal network architecture for the upcoming 5G era [18]. Net-
work slicing allows network operators to create multiple virtual end-to-end (E2E)
logical networks running on a common underlying physical or virtual network infras-
tructure [143], [144], [154], [155]. Each slice is logically isolated including network
device, radio access, transport and Core Network (CN), and dedicated for different
types of services with different characteristics and requirements. These slices can
be created on demand with independent control and management [156]. Technolo-
gies like Software Defined Networking (SDN) and Network Function Virtualization
(NFV) are used to tailor the network for a given use case, and create multiple logical
networks on the top of a common physical or virtual network infrastructure [153].
For each network slice, dedicated resources such as bandwidth storage, processing
or Traffic are sliced and virtualized network functions (VNFs) are placed in different
locations (i.e., Edge Cloud or CN Cloud) for each slice depending on different ser-
vice requirements. Also different slices are isolated, which means an error or a fault
occurred in one slice does not cause any service interruption on overall communica-
tion in other slices. Depending on the type of service requirements of different use
cases (e.g., Mobile Broadband, Massive IoT, and Mission-critical IoT), each slice is
isolated to meet certain QoS acquirements, such as Low latency and ultra reliability.
QoS isolation guarantees that the minimum level of QoS experienced by end users or
devices belonging to one type of service is not despoiled on change in network state

Chapter 6 123
such as, mobility and traffic load fluctuation, at another service type. Network slic-
ing can provide more cost effective solutions for Intelligent transportation systems
(ITS) by offering multiple logical networks over a single physical network, instead of
creating dedicated networks for each single service such as, one for autonomous driv-
ing, one for road safety, another for 5G mission-critical applications. In 5G, latency
critical services demand an E2E delay between 1 ms to 100 ms [157]. The network
slicing solution involves the partition of the Core Network (CN) and the Radio Ac-
cess Network (RAN) resources including the configuration of the end-device vehicle
functionality, to support different use cases [146]. Cloud-RAN (C-RAN) technology
plays an important role to attain the on-demand deployment of RAN functionalities.
Using C-RAN the radio and the baseband processing functionalities are segregated.
While the base band functionalities are migrated towards the cloud and form a BBU
pool and are controlled centrally. The centralized processing of the BBU pool func-
tionalities saves time (i.e., processing and signalling time) for handovers as compared
to a distributed processing at each eNodeB. By leveraging the concept of virtual-
ization technologies, C-RAN resources in the pool can be dynamically allocated to
eNode Base stations according to the load on network. This ensures adaptability
to the non-uniform vehicular traffic scenarios during off-peak/rush hours, in urban
or rural environments. In the core network, SDN and NFV have been introduced
to support large capacity and low latency and massive connectivity with seamless
procedures. Since these technologies are not part of the legacy LTE system, exten-
sive research is being carried out for the development in the context of 5G networks.
The main challenge of SDN/NFV-based core network design is the management
and orchestration of heterogeneous resources [158]. Implementation of effective re-
source allocation strategies and functions in heterogeneous environment while also
maintaining low latency is an area of emerging research. Network slicing in wireless
domain (RAN slicing) mainly deals with how to slice the overall radio resources
for different device groups to ensure QoS isolation. The radio access functions on
each BS are softwareized and are centrally controlled and managed by the SDN
enabled virtualization controller. Whereas, in the core network, network slicing is
interpreted as bi-resource slicing. The SDN controller can determine the amount
of radio resources allocated to each Base station to improve the overall spectrum

124 Chapter 6
utilization. As the 5G architecture is still evolving, the existing research present
new architectures for network slicing in different domains like wireless or a core net-
work [151], [152], [153]. However, limited research focus on presenting radio resource
slice optimization schemes in 5G-driven VANETs.
In this study an E2E network slicing framework is constructed that handles a mix-
ture of best-effort traffic from regular users and mission-critical traffic from the
prioritized user in 5G -driven VANETs. Moreover, a dynamic radio resource slice
optimization scheme is presented in the context of E2E reliability of critical traffic
at the softwareized 5G CN level. The network is configured to always guarantee
requested data rate for the mission-critical traffic, even if it leads to deteriorating
the QoS of the best-effort sessions for regular users. Both the best-effort and the
mission-critical traffic will have to travel through a complex network including RAN
and CN to compete for radio resources during transmission.
6.3 End-to-End Network Slicing framework in 5G-
driven VANETs
A detailed practical framework is constructed for the end-to-end network slicing in
5G-driven VANETs, by leveraging the concepts of NFV, SDN, C-RAN and Edge
Computing technologies as shown in Figure. 6.1. The proposed solution handles a
mixture of best-effort traffic from regular users and mission-critical traffic from the
prioritized user and, performs slicing by keeping in view both the RAN and the CN
as shown in Figure. 6.2.
6.3.1 Hierarchy/levels of slicing for proposed E2E slicing
framework:
To have a closer look at how E2E network slices are actually implemented, we
discuss slicing at different levels or hierarchies such as Edge Cloud (EC) and
CN Cloud with respect to proposed framework shown in Figure. 6.2. Dedicated
slices are created for services with different requirements and Virtualized Network
Functions (VNFs) are placed in different locations (i.e., EC or CN cloud) for each

Chapter 6 125
MEC Server
BBU pool
Edge Cloud
(with Distributed Core)
Connected Vehicle
Front-haul links
SDN enabled Virtualization
(CN/RAN) Controller
(C
V-RAN
Core Network (CN)
MEC
server
Connected Vehicle
Figure 6.1: An End-to-End (E2E) Network Slicing Framework for 5G-driven
VANETs

126 Chapter 6
v-RAN-runs
RAN, Core and
MEC
v-RAN
Edge Cloud (EC) (NFV)
Hypervisor
VM/VNF v-RAN
Commercial Server
Hypervisor
VM/VNF CN
Commercial Server
Low bandwidth
CN Cloud (NFV)
Mission Critical
Applications
Slice
Ultra Reliability,
Lowest delay
SDN enabled
Virtualisation Controller
(Network Connections
between VMs)
Missio
ion
n Cr
C
C
Cr
Cri
it
it
it
it
it
iti
i
i
ic
ic
ic
ic
ic
ic l
l
l
al
al
al
al
al
al
al
A li ti
L b d idth
Non- Critical
Applications
Slice
MEC
Server
MEC
Server
Fronthaul
Idle/
off
Idle/off Idle/off
CN
(UP)
CN
(CP)
CN
(UP)
CN
(CP)
CN
(UP)
EC
(CP)
EC
(UP)
EC
(CP)
EC
(UP)
Figure 6.2: E2E mission critical slicing including CN and RAN in 5G-VANETs
slice depending on services. Moreover, some network functions, like policy control,
charging, and etc., required in one slice may not be compulsory in other slices. The
proposed framework will allow the network operators to customize slices based on
different service requirements, in the most cost-effective way by placing VNFs at
different locations using NFV, and by having a separate Control plane (CP) and
user plane (UP) using SDN. The proposed framework is hybrid and flexible where
processing is centralized for some services and distributed for others. Let’s have a
closer look at the levels of slicing and how slices are implemented.
6.3.2 Edge Cloud (EC) and CN Cloud:
EC distributes 5G Core Network (CN) close to cell sites or end-users or edge. Edge
may refer to the base stations/Remote Radio Heads (RRHs), Connected vehicles
and data centres close to the radio network (e.g., located at aggregation points).
Virtualized RAN (v-RAN) runs RAN, Core and MEC operations. Operators
can operate v-RAN by deploying innovative applications and services for different
enterprises and verticals, flexibly and rapidly. Since, there are different slice isolation

Chapter 6 127
requirements that consider specific resource management means to meet various
KPIs. Each slice may need a different control plane (CP) and User plane (UP)
functional split, and a distinct VNF placement (at either EC or CN cloud) to ensure
an optimal performance.
How to implement network slices on EC and CN Cloud
To implement network slices, Network Function Virtualization (NFV) is a basic
requirement. Using NFV, Virtual Network Functions (VNFs) (e.g., MME, S/P-GW
and PCRF in Packet Core, and DU in RAN) are installed on to Virtual Machines
(VMs). Such VMs are deployed on a virtualized commercial server usually known as
Commercial off-the-shelf (COTS), instead of installing on to their dedicated network
equipment individually. According to Figure. 6.2, applications dedicated for each
service (i.e., mission critical and non-critical) are virtualized and installed in each
slice. The VNFs are placed at different locations i.e., EC or CN Cloud depending
on the service requirements. In short, slices can be configured as follows.
Mission-critical slice To meet Key Performance Indicators (KPIs) for mission
critical application slice (i.e., ultra reliably and lowest latency communication),
the VMs of 5G CN (User plane (UP) and Control Pane (CP)) and associated
servers (e.g., V2X server, MEC server and etc.,) are all down in Edge Cloud (EC).
Slice allocation is performed on demand to meet ultra reliability and minimized
transmission delay.
Non-critical slice On the other hand, non-critical application slice is based
on best-effort allocation of resources. The VMs of 5G CN (User plane (UP) and
Control Plane (CP)) will remain on CN cloud. However, some functionalities that
are required to be processed by end user devices or vehicles, will be handled by VMs
of User plane (UP) and Control Pane (CP) placed at EC.
Network slicing between EC and CN cloud
The SDN-enabled virtualization controller performs VM (VNF) Creation and
Control at EC and CN Cloud and provides network connectivity be-
tween VMs in Edge and Core clouds. Traffic Flows for different types of

128 Chapter 6
services, (aggregated through back-haul links) are automated by SDN controller. In
order to maintain the priority of the mission critical flows with respect to other flows
(e.g., best-effort traffic), the SDN controller maps the flows onto a priority queue.
The SDN controller takes the chain of services and apply them to different traffic
flows depending on the source, destination or type of traffic. This Service function
chaining (SFC) capability of SDN controller creates a service chain of connected
network services (such as, firewalls, DNS, network address translation (NAT), In-
trusion Detection System (IDS)), and connect them in a virtual chain [150]. Using
SFC embedding, SFCs can be placed on VNF nodes at various locations along the
paths from CN to EC. SDN Controller also performs provisioning of the virtualized
server (built-in vRouter/vSwitch running in Hypervisor of the server).
The complete E2E process of slicing between CN and EC cloud can be illustrated
from Figure. 6.3. 1. SDN Controller receives two incoming traffic flows from end
user application requests (i.e., FL1 for Mission Critical (MC Flow) and FL2 for
non-Mission Critical (non-MC Flow). Each flow requires different VNFs and logic
SFCs. Using SFC embedding, these logic SFCs will traverse one embedded underlay
path to fulfil the required KPI’s. 2. For each flow, Controller creates the VNF on
demand (i.e., VFL1 for MC slice and VFL2 for non-CR slice). The Packets of flow FL1
and FL2 will go through the VNFs (FFL1 and FFL2 ) on NFV nodes (VEC and VCN )
for processing. These packets will then be transmitted by a set of outgoing underlay
transmission links {L0, L1, L2, ..., Lm} and network routers {R1, R2, ..., Rl} before
arriving at destination. 3. For each slice, an overlay tunnel is created. Connecting
slicing from Edge cloud, to IP/MPLS backbone, and all the way to Core Cloud. The
SDN Controller performs mapping between these tunnels and MPLS L3 VPN (e.g.,
MC slice VPN and non-CR slice VPN). This process will be implemented using
current available technologies and standards (e.g., ( L2/L3 VPN, VXLAN, OTV,
LISP and etc.).
6.4 Problem Formulation
A multi-objective solution, that handles a mixture of best-effort traffic and mission-
critical traffic, by considering capacity allocation per slice and minimising delay is

Chapter 6 129
FL
2
FL
1
VEC
VCN
Router
NFV Node Over Lay Tunnel Underlay Path
R1
Rl
L0
RK
L1
Lp
Lm
Non-MC Flow
FL2
MC Flow
SDN Enabled
Virtualization Controller
Connectivity
between VMs
FFL1
FFL2
FFL1
FFL2
F
End User
Application
VFL2
VFL2
VFL1
VFL1
Figure 6.3: E2E Network Slicing Between EC and CN Cloud

130 Chapter 6
Mission Critical Resource (MCR) block
T1 T2 TN
Incoming Critical Applications ∑ TN
(MCR)=f(Cap, MinDelay)
Figure 6.4: Mission Critical Resource Block
proposed. The overall bandwidth resources are sliced in C-RAN, for mission critical
applications and non-critical applications, by keeping in view resource elasticity re-
quirements. To implement the concept of slicing for mission critical and non-critical
traﬃc, we formulate the problem as follows.
Let RRH = {RRH1, RRH2, ..., RRHn} with cardinality |RRH| = nRRH represents
the set of Remote Radio Heads or Base stations that are distributed in an area and
nBBUC represents number of BBU Controllers (BBUC).
Let BBUC = {BBUC1, BBUC2, ..., BBUCn} with cardinality |BBUC| = nBBUC
represent the set of BBUCs, such that nBBUC ≤ nRRH. Let Links = {BBUCi, RRHj}
represents the set of possible link pairs between BBUCs and RRHs.
Variables
Zij =
⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
1, if RRHj is served by BBUCi
0, otherwise
where(i, j) ∈ Links
Yi =
⎧
⎨
⎩
1, if BBUCi is Chosen
0, otherwise

Chapter 6 131
Run genetic algorithm (GA) for only critical
Flows to perform Mission Critical Slicing
Are all critical demands Served?
Increase BBUC capacity and re-run
No
Operation Prior to Optimize Non-Critical demands:
1. non_loc_critical_RRH (find index location of non-critical
demands from CRRH vector)
2. Partition the following matrices to include only non_critical
demands: CZC_NC, Distance_NC, Link_NC, CBBUC_NC,
Map_NC
Calculate remaining
available capacity of
BBUCs
Yes
Are all non-criticaO
demands?
Display queued RRHs
No
Display optimization summary and
graphs
Stop
Yes
Start
Input Data:
Controller maintains Priority¬Queue¬for Critical Flows
Figure 6.5: Program Flow Chart of Proposed E2E slicing Scheme

132 Chapter 6
6.4.1 Objective function
To deal with mission-critical application, there is a need to meet certain QoS re-
quirements such as ultra-reliability and low Latency. Hence, the KPIs for mission
critical slice are ultra-reliability (guaranteed service) and low latency (minimised
delay) as shown in Figure. 6.4. The main objective function for mission Critical
Resource Block CMCR
is given as
CMCR
=

mindelay
2
+ Caputl
2
(6.1)
The individual objective functions for capacity utilisation and minimised delay are
discussed as follows.
Capacity Utilisation(Cap-utl)
To provide guaranteed delivery for mission critical demands, we allocate resources
by balancing the load on each BBUC. For this we propose the Cap-utl algorithm
that aims to balance the critical demand load in every BBUC Pool. The information
of critical demand load of each BBUC Pool is always available and updated by the
SDN controller. Before evaluating the decision, the controller has the information
of traﬃc load of the possible BBUC Pool connections, thus, the controller will check
not only the maximum capacity limit of the BBUC Pool, but also maintain the
priority queue for mission critical demands. The objective function is given by
Minimize Caputl =
1
nBBUC

nBBUC

i=1
(D − Di)2 (6.2)
where Di =
nRRH
j=1 ZijCRRHj
is ith
element indicating the total critical demand in
BBUCi in the total load demand vector D and D = 1
nBBUC
nBBUC
i=1 Di is the average
demands across all BBUCs. The idea behind the objective function is to reduce the
standard deviations of the total load demand vector D. Under ideal conditions, if
the capacity demand is the same in all BBUCs, the positive objective function value
must be equal to zero.
Minimise Delay
The objective of this problem is to minimize the front-haul delay by connecting
RRHs closer to the possible BBUC Pool location. The SDN controller has all the

Chapter 6 133
possible locations of BBUC Pools, thus, knowing all the distances between possi-
ble link connections between RRHS and BBUCs. Since, delay is considered to be
directly proportional to the front-haul distance between RRHs and BBUCs which
in turn is related to the cost associated with linking BBUC and RRHs. Hence the
objective function is given by
minDelay =
nBBUC

i=1
nRRH

j=1
Costi,jZij (6.3)
where nBBUC is the number of BBUCs in the pool, Costi,j is the front-haul link cost
for linking RRHs j and BBUCi in the cost matrix Cost. Equation 6.2 and 6.3 are
subject to the following constraints;
1. Sum of demands of RRHs connected to BBUC i must be less than or equal to
available link capacity of that BBUC i.
nRRH

j=1
ZijCRRHj
≤ CBBUCi
(6.4)
where j
{1, 2, ..., nRRH} and i
{1, 2, ..., nBBUC}. CRRHj
is jth
element with a
value equal to the critical/non-critical capacity demands of RRH j in capacity
demand row vector CRRH.
2. BBUC i should be serving if atleast one RRH is connected to it.
Zij ≤ Yi (6.5)
where j
{1, 2, ..., nRRH} and i
{1, 2, ..., nBBUC}.
3. BBUC i should not be serving if no RRHs are connected to it.
nRRH

j=1
Zij ≥ Yi (6.6)
where j
{1, 2, ..., nRRH} and i
{1, 2, ..., nBBUC}
6.5 Simulation results and Discussions
In this study, simulations are conducted to evaluate the performance of network
slicing for proposed framework. MATLABR

is used to run the optimization of the

134 Chapter 6
Resource Block - Mission Critical
21.0
29.0
25.0
26.0
23.0
21.0 145.0
Critical Demand Served Capacity
0
50
100
150
Usage
in
Units
of
Capacity
Figure 6.6: Optimum Utilization of resource for Mission Critical slice
objective function. Genetic algorithm (GA) is used for conducting optimization.
The following settings for the GA parameters are chosen: population size of 2000,
generation size of 100, elite rate of 5%, tolerance value of 1e-8. Figure. 6.5 shows the
program flow of how it is executed. In the present problem, the GA is run two times:
(1) for critical demands and (2) for non-critical demands on the remaining capacities
available in different BBUCs. The maximum front-haul distances dfronthaul (that
determines the radius of coverage of every BBU Pool) are set according to the
two transmission technologies. Fibre link – maximum distance of 40 km and link
propagation speed υfronthaul of 200 km/ms. Microwave link – maximum distance of
1.5 km and link propagation speed υfronthaul of 300 km/ms as taken in [159]. For
the core network, we consider two flows, FL1 and FL2 representing two logic SFCs
traversing one embedded overlay network path. We set different packet sizes for
flow FL1 and FL2 at VEC that depends on the service requirements (i.e. mission
critical or non-mission critical). The proposed framework allows FL1 and FL2
that require different levels of QoS, to dynamically control their respective KPIs
depending on the effective demands by each of these flows. To maintain the priority
of the mission critical flows with respect to other non-mission critical flows (i.e. best-
effort traffic), the SDN controller maps the flows onto a priority queue. In mission

Chapter 6 135
Resource Block - Non-Critical
13.0
15.0
17.0
13.0
15.0
16.0
12.0
16.0
19.0
16.0
12.0
19.0
10.0
15.0
176.0
Non-Critical Demand Served Capacity
0
50
100
150
200
250
Usage
in
Units
of
Capacity
non-Critical Demands
in queue
Figure 6.7: Optimum Utilization of resource for non-Critical slice
Critical and Non-Critical Usage (in %)
54.7 %
29.5 %
15.8 %
46.7 %
26.7 %
26.7 %
27.8 %
66.7 %
5.6 %
38.2 %
47.3 %
14.5 %
32.5 %
62.5 %
5.0 %
1 2 3 4 5
BBUCs
0
10
20
30
40
50
60
70
80
90
100
%age
of
Resource
share
of
CR
and
non-CR
Flow
Critical Usage
Non-Critical Usage
Remaining Resource
Figure 6.8: Combined resource utilization for both Critical and non-Critical slices

136 Chapter 6
Figure 6.9: Optimization summary as output from MATLAB
critical communication, (ultra reliable low latency communication URLLC), both
the latency and reliability issues are addressed. For investigating ultra reliability
in mission critical communication, the objective function in equation 6.1 is solved
using GA. All RRHs are allowed to connect with any of the BBUCs and its distance
to each BBUC is calculated using equation 6.3. After running the program, the
optimization summary is shown in Figure 6.9. Figure.6.10 shows the graph obtained
from the GA run process. It shows the mean and best values in each generation. It
is observed that as the optimization progresses the mean and best converges. This
indicates that towards the end of optimization process, all the eligible solutions
become closely the same. we see how to optimize slicing of radio resources at BBUC
pool for mission critical traﬃc, using proposed objective functions in equations 6.2
and 6.3. To exploit resource multiplexing gain, the amount of resources of each slice
is dynamically adjusted according to changes of network conditions. Our results in
Figure. 6.6 illustrate that the mission critical resource block is dynamically adjusted
as per the arrival of mission critical demands from end users/vehicles. The results

Chapter 6 137
Current Best Individual
0 10 20 30 40 50 60 70
Number of variables (75)
0
0.5
1
Current
best
individual
0 20 40 60 80 100
Generation
0
5
10
15
20
Penalty
value
Best: 0.0307752 Mean: 0.0880651
Best penalty value
Mean penalty value
Pause
Stop
Figure 6.10: Optimization using GA
are tested on varying demands of mission critical requests and it is observed all the
mission critical demand are served. The served capacity for mission critical slice is
ﬂexible and is dynamically adjusted with reliability as high as 99.99%. Hence our
results prove that the proposed objective meet ultra reliability for mission critical
communications. On the other hand, non-mission critical demands are served on
best eﬀort delivery and are not prioritized, hence few non-critical applications will
be in queue as illustrated in Figure. 6.7. This is because the controller maintains a
priority queue of critical applications. Figure. 6.8 shows resource utilization for each
BBUC after the allocation for both critical and non-critical resource slices. Each
bar represents various types of usages in each BBUC. We observed that BBUC 1
has the maximum critical allocation with 54.7% while BBUC 3 has the maximum
non-critical allocation at 66.7%. It also displays the remaining capacities. Figure.
6.11 shows the distance of each RRH to the actively connected BBUCs. Ideally,
these distances will be least possible distances to connect after the optimization.
It is observed that RRH 19 and 20 are the farthest RRHs being served with 8
km of front-haul distance. Figure 6.12 shows the optimum active connections of

138 Chapter 6
RRH Distance to Connected BBUC
5.0
3.0
4.0
5.0
7.0
3.0
2.0
3.0
2.0
7.0
2.0 2.0
3.0
6.0
1.0 1.0
8.0 8.0
0 5 10 15 20 25
RRH No.
0
1
2
3
4
5
6
7
8
9
front-haul
distance
d
fronthaul
Figure 6.11: Optimized Front-haul distances of RRHs with BBUCs using equation
6.3
RRH with BBUCs. 1 and 0 indicates active and inactive connections. To analyse
E2E latency of mission critical slice, we first determine the length of front-haul link
between RRH and BBUC which is calculated using equation 6.3. It is important to
note that latency is represented by the measure of Round Trip Time (RTT), which
has a more meaningful impact on Quality of Experience (QoE) than One-Way Delay
(OWD). For simplicity, the RTT δRTT is calculated using 2∗τOWD, where τOWD(ms)
is one way delay [159]. The length of the front-haul link dfronhaul(km) is described
by transmission speed υfronthaul(km/ms) of front-haul link and by the front-haul one
way delay τOWD. We compare results of E2E latency at different link propagation
velocities with previously proposed 5G VANET architecture [9]. Figure 6.13 shows
the maximum E2E front-haul latency at each RRH for mission critical slice. This is
based on taking the different propagation velocities both in fibre and in microwave
links into account. As expected from the previous discussion regarding distance,
the E2E delay will be maximum for RRH 19 and 20 due to maximum front-haul
distance for our proposed slicing framework. Figure 6.13 illustrates that maximum
E2E latency on RRH 19 and 20 (that are at farthest distance from BBUCs) is

Chapter 6 139
20
18
16
14
Optimum Front-haul Connections
12
RRH
0
10
0.5
8
Active
[1]
/
Inactive
[0]
1
1
2 6
BBUC
3
4
4
5 2
Figure 6.12: Optmized Front-haul connections of RRHs with BBUCs using equation
6.1
0 5 10 15 20 25
RRH No.
0
0.5
1
1.5
2
2.5
3
3.5
4
Maximum
Front-haul
Latency
(ms)
10
5
Link tranmission speed (Proposed Slicing) = 300 km/ms, Max dfronthaul
=40km
Link tranmission speed (Proposed Slicing) = 200 km/ms, Max dfronthaul
=1.5 km
Link tranmission speed [9] = 300 km/ms, Max dfronthaul
=40km
Link tranmission speed [9] = 200 km/ms, Max dfronthaul
=1.5 km
Figure 6.13: Comparison of E2E Latency of proposed scheme with 5G VANET
architecture [9]

140 Chapter 6
1.2ms for Fibre link and 1.9ms for microwave link. However maximum E2E latency
of previously proposed 5G VANET architecture [9] is 3ms (for fibre link) and 3.3ms
(for microwave link). The promising results of proposed slicing framework affirm
that E2E latency of critical services in 5G can be supported with reliability as high
as 99.99%. Therefore, the proposed slicing solution meets both KPIs (i.e., Ultra
reliability and low latency) for mission critical services.
6.6 Conclusion
In this chapter, an E2E network slicing framework is presented to achieve desired
level of QoS provisioning among customized services in 5G-driven VANETs, by con-
sidering both RAN and core network, which is a key challenge of 5G networks.
Through SDN-enabled NFV technology, the proposed framework distributes some
services of 5G core close to cell sites using Mobile Edge Computing (MEC) technol-
ogy and, keep other services with centralized processing, to meet desired levels of
KPIs. Furthermore, a dynamic radio resource slice optimization scheme is formu-
lated mathematically, to implement network slicing for mission-critical and and best
effort traffic in 5G-driven VANETs. The solution is solved using GA by keeping in
view the resource elasticity requirements.
It is concluded from simulation results that the proposed slicing scheme achieves the
desired levels of end-to-end reliability and timely delivery of mission-critical traffic.

Chapter 7
Conclusion and Future Directions
I n this chapter we conclude the summary of the contributions and results, and a
number of interesting future directions are listed.
7.1 Conclusion
Next generation Vehicular Ad-hoc Networks (VANETs) will be dominated by expo-
nential growth of heterogeneous data traffic, including additional massive diffusion
of Internet of Things (IoT) traffic. The exponential growth of heterogeneous data
traffic and diversified quality of service (QoS) demands poses significant challenges
for current Vehicular Ad-hoc Networks (VANETs) and is one of the primary reasons
for the evolution of next generation 5G-driven VANETs. To meet these challenges,
network resources need more flexible and optimized resource allocation strategies.
According to our vision, this evolution can be achieved by transforming VANETs
into a more flexible and programmable fabric with a globalized view. This objective
can be acquired by jointly utilising 5G-driven technologies (such as Cloud-RAN,
SDN and NFV) by providing a multitude of diverse services and resource sharing
over a common underlying VANET infrastructure. Besides these technologies, Edge
Computing and Fog Computing technologies are also playing important roles in
offering ultra low latency, high resource utilization, and real-time access to radio
access. These 5G-driven technologies are expected to substantially improve commu-
nication and resource sharing over a common underlying physical infrastructure of
VANETs.
141

142 Chapter 7
Researchers have explored multiple solutions for optimized communication in VANETs
but no work has yet been proposed which advise different frameworks and optimiza-
tion approaches addressing resource allocation with the 5G perspective.
In this thesis, we aim to propose efficient algorithms and approaches to provide
optimized communication and resource allocation in 5G driven VANETs. In this
chapter we provide the summary of the results and contributions.
7.1.1 Literature Review
A comprehensive literature review of heterogeneous Vehicular Ad-hoc Networks and
5G-driven technologies such as Software Defined Networking (SDN), Cloud-Radio
Access Network (C-RAN), Network Function Virtualization (NFV) and Edge Com-
puting along-with their implementation in VANETs is presented. The following
conclusions are drawn from the literature review;
• Current heterogeneous VANETs using different wireless access technologies
such as 4G, LTE, LTE with D2D and 3GPP cannot be easily well cooperated
under the traditional VANET architectures. Consequently, a large number of
of wireless network infrastructures and spectrum resources are wasted, thereby
leading to the low quality of experience (QoE) of vehicle users. There is a
need to rethink current VANET architecture, to turn it into a more flexible
and programmable fabric enabling a globalized view of all resources.
• The management and control of vehicular networks on a large scale becomes
a major challenge due to ever increasing vehicular network size and highly
evolved physical layer technology [3].
• The frequent handoffs between different cellular infrastructures due to high
mobility and rapidly changing topology of VANETs becomes another major
challenge.
• To provide consistency in services with the frequent topology changes and
varying QoS demands, the heterogeneous substrate must have a global view
of all service requests, to provide network functionalities more efficiently on
large scale.

Chapter 7 143
• Integrating 5G enabling technologies such as Software Defined Networking
(SDN), Network Function Virtualization (NFV), Cloud-RAN (C-RAN) and
Fog/Edge Computing, VANETs are expected to provide solutions for these
challenges.
7.1.2 5G Next generation VANETs using SDN and Fog Com-
puting Framework
A detailed description of high level design of the proposed 5G-driven architecture
which includes; description of physical topology and logical structure of architec-
ture and, the roles of each component contributed in the architecture are discussed.
A new Fog Computing (FC) framework is also discussed. Some benefits of the
proposed architecture associating its feasibility in HetVANETs are also discussed.
Simulation is performed to investigate the performance of architecture by comparing
the transmission delay, throughput and control overhead on controller with other
architectures. From the simulation and results, it can be seen that the through-
put is improved, and transmission delay and control overhead on controllers is also
minimized in comparison with previously proposed architectures.
7.1.3 An Evolutionary Game Theoretic Approach for Stable
and Optimized Clustering in VANETs
An innovative Evolutionary Game Theoretic (EGT) approach is proposed to auto-
mate the clustering of nodes and nominations of cluster heads, to achieve cluster
stability in VANETs. The equilibrium point is proven analytically and the exis-
tence of evolutionary equilibrium is also verified using the Lyapunov function. The
proposed game is tested and analyzed with different number of clusters for different
populations of vehicles and cost functions. An optimal cost is suggested that defines
an optimum clustering.
It is concluded from the simulation and results that the proposed approach is
lightweight and semi-distributed, and allows faster convergence. Furthermore, the
signalling overhead and complexity of proposed approach is minimized and the
switching rate of cluster heads is also reduced in comparison to ALM clustering [8].

144 Chapter 7
It is also analyzed through simulation and results that the proposed framework is
able to maintain cluster stability, as the clusters evolve towards balanced sizes and
system is converged with an average total throughput of clusters.
7.1.4 A Hybrid-Fuzzy Logic Guided Genetic Algorithm (H-
FLGA) Approach for Resource Optimization in 5G
VANETs
A hybrid Fuzzy Logic guided Genetic Algorithm (H-FLGA) approach is proposed
for the SDN controller, to support optimum resource allocation over our proposed
5G-driven VANET architecture in Chapter 3. The proposed approach facilitates the
network service providers to implement a more customer-centric network infrastruc-
ture thus improving their spectral efficiency. A multi-objective resource optimization
problem is formulated, where five different objectives of resource provisioning are
combined and solved using a hybrid approach. The proposed Fuzzy Inference Sys-
tem (FIS) is used to optimise weights of multi-objectives depending on the Type of
Service (ToS) requirements of customers.
It is concluded from the simulation and results that the value of multi-objective cost
function of proposed hybrid H-FLGA approach is minimized as compared to GA.
The proposed approach is flexible to set weights of multi-objectives, depending on
QoS demands and requirements of different users.
7.1.5 An End-to-End (E2E) Network Slicing Framework for
5G Vehicular Ad-hoc Networks
An E2E network slicing framework is proposed with the consideration of both the ra-
dio access network resources and the core network resources in 5G-driven VANETs.
Moreover, a dynamic radio resource slice optimization scheme is also formulated,
that handles a mixture of both best-effort traffic and mission-critical traffic, by
keeping in view resource elasticity requirements.
The problem is solved using Genetic Algorithm (GA) and the results are compared
with our previously proposed 5G VANET architecture in chapter 3. It is analyzed
from the simulation and results that the bandwidth slicing ratios are optimality ad-

Chapter 7 145
justed and the network dynamically adapts to instantaneous network load conditions
in a way that a targeted KPIs are guaranteed.
7.2 Future Directions
In this section, based on the assumptions, results and observations discussed in this
thesis, a number of interesting future research directions are listed below;
• In Chapter 3, the performance of proposed 5G-driven VANET architecture is
investigated through simulation, by comparing the transmission delay, through-
put and control overhead on controller with other architectures. It would be
very useful to investigate the performance of proposed architecture through a
real test bed for SDN based system and comparing the result with the simula-
tion result. The test-beds can give more realistic and accurate results towards
the real-life scenarios.
Future challenges include, optimizing route selection, designing protocols at
SDN controller for load balancing, improving service efficiency provision, due
to massive traffic increase for 5G-driven VANETs.
• The proposed Evolutionary Game Theoretic (EGT) approach in chapter 4
paves a way towards stable and optimized clustering in VANETs. The pro-
posed EGT approach can further be extended by analyzing the efficiency of
the proposed game on the overall protocol stack using a network simulator.
Furthermore, investigating it over different VANET routing protocols such as
AODV, DSR, DSDV, OLSR, using a real test bed or network simulator can
quantify the efficiency of proposed approach.
• The proposed H-FLGA approach in chapter 5 can further be extended by con-
sidering multiple resource sharing scenarios by using OpenFlow and Mininet.
The performance of proposed scheme can further be quantified through exten-
sive simulations, by taking energy consumption into account. Future direc-
tions include the possibility of implementing the proposed method in future
ultra-dense networks, especially its implementation in computation offloading
and resource allocation in ultra-dense networks. Furthermore, the proposed

146 Chapter 7
approach may be tested using OpenFlow and Mininet.
• Our proposed E2E network slicing scheme in chapter 6 can further be em-
ployed to analyse customised 5G VANET scenarios with different KPIs, which
involve transporting different flows on a complex network, including RAN and
CN to compete for resources during transmission, buffering, and computing.
However, network slicing for the 5G era is still shaping up, with most of the
concerns and issues remaining unsolved.

References
[1] Elias C Eze, Si-Jing Zhang, En-Jie Liu, and Joy C Eze. Advances in vehicular
ad-hoc networks (vanets): Challenges and road-map for future development.
International Journal of Automation and Computing, 13(1):1–18, 2016.
[2] Songlin Sun, Michel Kadoch, Liang Gong, and Bo Rong. Integrating network
function virtualization with sdr and sdn for 4g/5g networks. IEEE Network,
29(3):54–59, 2015.
[3] He Li, Mianxiong Dong, and Kaoru Ota. Radio access network virtualization
for the social Internet of Things. IEEE Cloud Computing, 2(6):42–50, 2015.
[4] Diego Kreutz, Fernando MV Ramos, P Esteves Verissimo, C Esteve Rothen-
berg, Siamak Azodolmolky, and Steve Uhlig. Software-deﬁned networking: A
comprehensive survey. Proceedings of the IEEE, 103(1):14–76, 2015.
[5] Jun Wu, Zhifeng Zhang, Yu Hong, and Yonggang Wen. Cloud radio access
network (c-ran): a primer. IEEE Network, 29(1):35–41, 2015.
[6] Aleksandra Checko, Henrik L Christiansen, Ying Yan, Lara Scolari, Georgios
Kardaras, Michael S Berger, and Lars Dittmann. Cloud ran for mobile net-
works—a technology overview. IEEE Communications surveys tutorials,
17(1):405–426, 2015.
[7] Rashid Mijumbi, Joan Serrat, Juan-Luis Gorricho, Niels Bouten, Filip
De Turck, and Raouf Boutaba. Network function virtualization: State-of-
the-art and research challenges. IEEE Communications Surveys Tutorials,
18(1):236–262, 2015.
147

[8] Evandro Souza, Ioanis Nikolaidis, and Pawel Gburzynski. A new aggregate
local mobility (ALM) clustering algorithm for VANETs. In Communications
(ICC), 2010 IEEE International Conference on, pages 1–5. IEEE, 2010.
[9] Ammara Anjum Khan, Mehran Abolhasan, and Wei Ni. 5G next generation
VANETs using SDN and fog computing framework. In Consumer Commu-
nications Networking Conference (CCNC), 2018 15th IEEE Annual, pages
1–6. IEEE, 2018.
[10] Kan Zheng, Lu Hou, Hanlin Meng, Qiang Zheng, Ning Lu, and Lei Lei. Soft-
defined heterogeneous vehicular network: architecture and challenges. arXiv
preprint arXiv:1510.06579, 2015.
[11] Bin Cao, Yun Li, Chonggang Wang, Gang Feng, Shuang Qin, and Yafeng
Zhou. Resource allocation in software defined wireless networks. IEEE NET-
WORK, 31(1):44–51, 2017.
[12] China Mobile. C-ran: the road towards green ran. White Paper, ver, 2, 2011.
[13] CHEN Jiacheng, ZHOU Haibo, ZHANG Ning, YANG Peng, GUI Lin, and
SHEN Xuemin. Software defined internet of vehicles: architecture, chal-
lenges and solutions. Journal of Communications and Information Networks,
1(1):14–26, 2016.
[14] Maede Zolanvari. SDN for 5G, http://www.cs.wustl.edu/ jain/cse570-
15/ftp/sdnfor5g.pdf. 2015/10.
[15] Xiaohu Ge, Song Tu, Guoqiang Mao, Cheng-Xiang Wang, and Tao Han. 5G
ultra-dense cellular networks. IEEE Wireless Communications, 23(1):72–79,
2016.
[16] Vitaly Petrov, Andrey Samuylov, Vyacheslav Begishev, Dmitri Moltchanov,
Sergey Andreev, Konstantin Samouylov, and Yevgeni Koucheryavy. Vehicle-
based relay assistance for opportunistic crowdsensing over narrowband IoT
(NB-IoT). IEEE Internet of Things journal, 2017.
148

[17] Qian Li, Geng Wu, Apostolos Papathanassiou, and Udayan Mukherjee. An
end-to-end network slicing framework for 5g wireless communication systems.
arXiv preprint arXiv:1608.00572, 2016.
[18] 5G for mission critical communication Nokia, Espoo, Finland, White Paper,
2016. [Online]. http://www.hit.bme.hu/ jakab/edu/litr/5G/Nokia 5G for Mis-
sion Critical Communication White Paper.pdf.
[19] Ammara Anjum Khan, Mehran Abolhasan, and Wei Ni. An Evolutionary
Game Theoretic Approach for Stable and Optimized Clustering in VANETs.
IEEE Transactions on Vehicular Technology, 67(5):4501–4513, 2018.
[20] Panos Papadimitratos, Arnaud De La Fortelle, Knut Evenssen, Roberto Brig-
nolo, and Stefano Cosenza. Vehicular communication systems: Enabling tech-
nologies, applications, and future outlook on intelligent transportation. IEEE
Communications Magazine, 47(11):84–95, 2009.
[21] Mehran Abolhasan, Justin Lipman, Wei Ni, and Brett Hagelstein. Software-
defined wireless networking: centralized, distributed, or hybrid? Network,
IEEE, 29(4):32–38, 2015.
[22] Xu Li, Petar Djukic, and Hang Zhang. Zoning for hierarchical network opti-
mization in software defined networks. In Network Operations and Manage-
ment Symposium (NOMS), 2014 IEEE, pages 1–8. IEEE, 2014.
[23] Yang Fangchun, Wang Shangguang, Li Jinglin, Liu Zhihan, and Sun Qibo. An
overview of internet of vehicles. China Communications, 11(10):1–15, 2014.
[24] Bruno AA Nunes, Manoel Mendonca, Xuan-Nam Nguyen, Katia Obraczka,
and Thierry Turletti. A survey of software-defined networking: Past, present,
and future of programmable networks. Communications Surveys Tutorials,
IEEE, 16(3):1617–1634, 2014.
[25] Ming Zhu, Jiannong Cao, Deming Pang, Zongjian He, and Ming Xu. Sdn-
based routing for efficient message propagation in vanet. In Wireless Algo-
rithms, Systems, and Applications, pages 788–797. Springer, 2015.
149

[26] Dan Levin, Andreas Wundsam, Brandon Heller, Nikhil Handigol, and Anja
Feldmann. Logically centralized?: state distribution trade-offs in software
defined networks. In Proceedings of the first workshop on Hot topics in software
defined networks, pages 1–6. ACM, 2012.
[27] Kan Zheng, Qiang Zheng, Periklis Chatzimisios, Wei Xiang, and Yiqing Zhou.
Heterogeneous vehicular networking: A survey on architecture, challenges,
and solutions. IEEE Communications Surveys Tutorials, 17(4):2377–2396,
2015.
[28] Zongjian He, Daqiang Zhang, and Junbin Liang. Cost-efficient heterogeneous
data transmission in software defined vehicular networks. In High Performance
Computing and Communications (HPCC), 2015 IEEE 7th International Sym-
posium on Cyberspace Safety and Security (CSS), 2015 IEEE 12th Interna-
tional Conferen on Embedded Software and Systems (ICESS), 2015 IEEE 17th
International Conference on, pages 666–671. IEEE, 2015.
[29] Kan Zheng, Lin Zhang, Wei Xiang, and Wenbo Wang. Heterogeneous Vehic-
ular Networks. Springer, 2016.
[30] John B Kenney. Dedicated short-range communications (dsrc) standards in
the united states. Proceedings of the IEEE, 99(7):1162–1182, 2011.
[31] Jiacheng Chen, Bo Liu, Haibo Zhou, Lin Gui, Ning Liu, and Yiyan Wu. Pro-
viding vehicular infotainment service using vhf/uhf tv bands via spatial spec-
trum reuse. IEEE Transactions on Broadcasting, 61(2):279–289, 2015.
[32] Giuseppe Araniti, Claudia Campolo, Massimo Condoluci, Antonio Iera, and
Antonella Molinaro. Lte for vehicular networking: a survey. IEEE Communi-
cations Magazine, 51(5):148–157, 2013.
[33] Georgios Karagiannis, Onur Altintas, Eylem Ekici, Geert Heijenk, Boangoat
Jarupan, Kenneth Lin, and Timothy Weil. Vehicular networking: A survey and
tutorial on requirements, architectures, challenges, standards and solutions.
IEEE communications surveys tutorials, 13(4):584–616, 2011.
150

[34] Haojin Zhu, Xiaodong Lin, Rongxing Lu, Yanfei Fan, and Xuemin Shen.
Smart: A secure multilayer credit-based incentive scheme for delay-tolerant
networks. IEEE Transactions on Vehicular Technology, 58(8):4628–4639, 2009.
[35] Nguyen B Truong, Gyu Myoung Lee, and Yacine Ghamri-Doudane. Software
defined networking-based vehicular adhoc network with fog computing. In
Integrated Network Management (IM), 2015 IFIP/IEEE International Sym-
posium on, pages 1202–1207. IEEE, 2015.
[36] Mohammad Ali Salahuddin, Ala Al-Fuqaha, and Mohsen Guizani. Software-
defined networking for rsu clouds in support of the internet of vehicles. IEEE
Internet of Things Journal, 2(2):133–144, 2015.
[37] Ian Ku, You Lu, Mario Gerla, Francesco Ongaro, Rafael L Gomes, and
Eduardo Cerqueira. Towards software-defined vanet: Architecture and ser-
vices. In Ad Hoc Networking Workshop (MED-HOC-NET), 2014 13th Annual
Mediterranean, pages 103–110. IEEE, 2014.
[38] Kai Liu, Joseph KY Ng, Victor CS Lee, Sang H Son, and Ivan Stojmen-
ovic. Cooperative data scheduling in hybrid vehicular ad hoc networks:
Vanet as a software defined network. IEEE/ACM transactions on networking,
24(3):1759–1773, 2016.
[39] Open Networking Fundation. Software-defined networking: The new norm for
networks. ONF White Paper, 2012.
[40] Lalith Suresh, Julius Schulz-Zander, Ruben Merz, Anja Feldmann, and Teresa
Vazao. Towards programmable enterprise wlans with odin. In Proceedings of
the first workshop on Hot topics in software defined networks, pages 115–120.
ACM, 2012.
[41] Huawei Huang, Peng Li, Song Guo, and Weihua Zhuang. Software-defined
wireless mesh networks: architecture and traffic orchestration. IEEE Network,
29(4):24–30, 2015.
151

[42] Deze Zeng, Peng Li, Song Guo, Toshiaki Miyazaki, Jiankun Hu, and Yong
Xiang. Energy minimization in multi-task software-defined sensor networks.
IEEE Transactions on Computers, 64(11):3128–3139, 2015.
[43] Abbas Bradai, Kamal Singh, Toufik Ahmed, and Tinku Rasheed. Cellular
software defined networking: a framework. IEEE Communications Magazine,
53(6):36–43, 2015.
[44] Jiaqiang Liu, Yong Li, Min Chen, Wenxia Dong, and Depeng Jin. Software-
defined internet of things for smart urban sensing. IEEE Communications
Magazine, 53(9):55–63, 2015.
[45] Sushant Jain, Alok Kumar, Subhasree Mandal, Joon Ong, Leon Poutievski,
Arjun Singh, Subbaiah Venkata, Jim Wanderer, Junlan Zhou, Min Zhu, et al.
B4: Experience with a globally-deployed software defined wan. In ACM
SIGCOMM Computer Communication Review, volume 43, pages 3–14. ACM,
2013.
[46] Steve Shattil. Cloud Radio Access Network, May 12 2015. US Patent App.
14/709,936.
[47] Georgios Kardaras and Christian Lanzani. Advanced multimode radio for
wireless mobile broadband communication. In Wireless Technology Confer-
ence, 2009. EuWIT 2009. European, pages 132–135. IEEE, 2009.
[48] AB Ericsson et al. Common public radio interface (cpri), interface specification
v6. 0, 2013.
[49] Open Base Station Architecture Initiative et al. Bts sys-
tem reference document, version 2.0. URL: http://www. obsai.
com/specs/OBSAI System Spec V2. 0. pdf, 2006.
[50] GSNFV ETSI. 001. Open Radio Equipment Interface (ORI), 2013.
[51] Bo Han, Vijay Gopalakrishnan, Lusheng Ji, and Seungjoon Lee. Network
function virtualization: Challenges and opportunities for innovations. IEEE
Communications Magazine, 53(2):90–97, 2015.
152

[52] Gianluca Rizzo, Maria Rita Palattella, Torsten Braun, and Thomas Engel.
Content and context aware strategies for qos support in vanets. In Proc. of Int.
Conf. on Advanced Information Networking and Applications (AINA-2016),
2016.
[53] Gartner, ”connected cars from a major element of internet of things”
http://www.gartner.com/newsroom/id/2970017, 2015.
[54] Shucong Jia, Sizhe Hao, Xinyu Gu, and Lin Zhang. Analyzing and relieving
the impact of fcd traﬃc in lte-vanet heterogeneous network. In Telecommu-
nications (ICT), 2014 21st International Conference on, pages 88–92. IEEE,
2014.
[55] Mamta Agiwal, Abhishek Roy, and Navrati Saxena. Next generation 5g wire-
less networks: A comprehensive survey.
[56] Asvin Gohil, Hitesh Modi, and Shital K Patel. 5g technology of mobile com-
munication: A survey. In Intelligent Systems and Signal Processing (ISSP),
2013 International Conference on, pages 288–292. IEEE, 2013.
[57] Zheng Ma, ZhengQuan Zhang, ZhiGuo Ding, PingZhi Fan, and HengChao Li.
Key techniques for 5g wireless communications: network architecture, phys-
ical layer, and mac layer perspectives. Science China Information Sciences,
58(4):1–20, 2015.
[58] Shanzhi Chen, Fei Qin, Bo Hu, Xi Li, and Zhonglin Chen. User-centric ultra-
dense networks for 5g: challenges, methodologies, and directions. IEEE Wire-
less Communications, 23(2):78–85, 2016.
[59] Michelle X Gong, Robert Stacey, Dmitry Akhmetov, and Shiwen Mao. A
directional csma/ca protocol for mmwave wireless pans. In Wireless Commu-
nications and Networking Conference (WCNC), 2010 IEEE, pages 1–6. IEEE,
2010.
[60] Amr El-Keyi, Tamer ElBatt, Fan Bai, and Cem Saraydar. Mimo vanets: Re-
search challenges and opportunities. In Computing, Networking and Commu-
153

nications (ICNC), 2012 International Conference on, pages 670–676. IEEE,
2012.
[61] Rupendra Nath Mitra and Dharma P Agrawal. 5g mobile technology: A
survey. ICT Express, 1(3):132–137, 2015.
[62] Ning Zhang, Shan Zhang, Peng Yang, Omar Alhussein, Weihua Zhuang, et al.
Software defined space-air-ground integrated vehicular networks: Challenges
and solutions. arXiv preprint arXiv:1703.02664, 2017.
[63] Tarik Taleb and Khaled Ben Letaief. A cooperative diversity based handoff
management scheme. IEEE Transactions on Wireless Communications, 9(4),
2010.
[64] Mehran Abolhasan, Tadeusz Wysocki, and Eryk Dutkiewicz. A review of
routing protocols for mobile ad hoc networks. Ad hoc networks, 2(1):1–22,
2004.
[65] Mingjie Feng, Shiwen Mao, and Tao Jiang. Enhancing the performance of
futurewireless networks with software-defined networking. Frontiers of Infor-
mation Technology Electronic Engineering, 17(7):606–619, 2016.
[66] Marc Mendonca, Katia Obraczka, and Thierry Turletti. The case for software-
defined networking in heterogeneous networked environments. In Proceedings
of the 2012 ACM conference on CoNEXT student workshop, pages 59–60.
ACM, 2012.
[67] Xiaohu Ge, Zipeng Li, and Shikuan Li. 5G Software Defined Vehicular Net-
works. arXiv preprint arXiv:1702.03675, 2017.
[68] Fan Li and Yu Wang. Routing in vehicular ad hoc networks: A survey. Ve-
hicular Technology Magazine, IEEE, 2(2):12–22, 2007.
[69] Marwa Altayeb and Imad Mahgoub. A survey of vehicular ad hoc networks
routing protocols. International Journal of Innovation and Applied Studies,
3(3):829–846, 2013.
154

[70] Unai Hernandez-Jayo, Aboobeker Sidhik Koyamparambil Mammu, and Idoia
De-la Iglesia. Reliable communication in cooperative ad hoc networks. 2014.
[71] Sadaf Momeni and Mahmood Fathy. Clustering In VANETs. Springer, 2010.
[72] Yun-Wei Lin, Yuh-Shyan Chen, and Sing-Ling Lee. Routing protocols in ve-
hicular ad hoc networks: A survey and future perspectives. J. Inf. Sci. Eng.,
26(3):913–932, 2010.
[73] P Sheu and C Wang. A stable clustering algorithm based on battery power
for mobile ad hoc networks. Tamkang Journal of Science and Engineering,
9(3):233, 2006.
[74] Alan Mainwaring, David Culler, Joseph Polastre, Robert Szewczyk, and John
Anderson. Wireless sensor networks for habitat monitoring. In Proceedings of
the 1st ACM international workshop on Wireless sensor networks and appli-
cations, pages 88–97. Acm, 2002.
[75] Mohammad S Almalag and Michele C Weigle. Using traﬃc ﬂow for cluster
formation in vehicular ad-hoc networks. In Local Computer Networks (LCN),
2010 IEEE 35th Conference on, pages 631–636. IEEE, 2010.
[76] Wai Chen and Shengwei Cai. Ad hoc peer-to-peer network architecture for
vehicle safety communications. Communications Magazine, IEEE, 43(4):100–
107, 2005.
[77] Yu Wang and Fan Li. Vehicular ad hoc networks. In Guide to wireless ad hoc
networks, pages 503–525. Springer, 2009.
[78] Thomas DC Little and Abhishek Agarwal. An information propagation scheme
for vanets. In Intelligent Transportation Systems, 2005. Proceedings. 2005
IEEE, pages 155–160. IEEE, 2005.
[79] Ram Ramanathan and Martha Steenstrup. Hierarchically-organized, multihop
mobile wireless networks for quality-of-service support. Mobile networks and
applications, 3(1):101–119, 1998.
155

[80] Jean Sebastien Bedo, Salah Eddine El Ayoubi, Miltiadis Filippou, Anasta-
sius Gavras, Domenico Giustiniano, Paola Iovanna, Antonio Manzalini, Olav
Queseth, Theodoros Rokkas, Michael Surridge, et al. 5G Innovations for New
Business Opportunities. 2017.
[81] Peng Fan, James G Haran, John Dillenburg, and Peter C Nelson. Cluster-
based framework in vehicular ad-hoc networks. In Ad-hoc, mobile, and wireless
networks, pages 32–42. Springer, 2005.
[82] Mohammad S Almalag and Michele C Weigle. Using traffic flow for cluster
formation in vehicular ad-hoc networks. In Local Computer Networks (LCN),
2010 IEEE 35th Conference on, pages 631–636. IEEE, 2010.
[83] Samo Vodopivec, Janez Bešter, and Andrej Kos. A survey on clustering al-
gorithms for vehicular ad-hoc networks. In Telecommunications and Signal
Processing (TSP), 2012 35th International Conference on, pages 52–56. IEEE,
2012.
[84] Prithwish Basu, Naved Khan, and Thomas DC Little. A mobility based metric
for clustering in mobile ad hoc networks. In Distributed computing systems
workshop, 2001 international conference on, pages 413–418. IEEE, 2001.
[85] Christine Shea, Behnam Hassanabadi, and Shahrokh Valaee. Mobility-based
clustering in vanets using affinity propagation. In Global Telecommunications
Conference, 2009. GLOBECOM 2009. IEEE, pages 1–6. IEEE, 2009.
[86] Stefano Basagni. Distributed clustering for ad hoc networks. In Parallel Ar-
chitectures, Algorithms, and Networks, 1999.(I-SPAN’99) Proceedings. Fourth
InternationalSymposium on, pages 310–315. IEEE, 1999.
[87] Grzegorz Wolny. Modified dmac clustering algorithm for vanets. In Systems
and Networks Communications, 2008. ICSNC’08. 3rd International Confer-
ence on, pages 268–273. IEEE, 2008.
[88] Mildred M Caballeros Morales, Choong Seon Hong, and Young-Cheol Bang.
An adaptable mobility-aware clustering algorithm in vehicular networks. In
156

Network Operations and Management Symposium (APNOMS), 2011 13th
Asia-Pacific, pages 1–6. IEEE, 2011.
[89] Slawomir Kukliński and Grzegorz Wolny. Density based clustering algorithm
for vanets. In Testbeds and Research Infrastructures for the Development of
Networks Communities and Workshops, 2009. TridentCom 2009. 5th Inter-
national Conference on, pages 1–6. IEEE, 2009.
[90] Nitin Maslekar, Mounir Boussedjra, Joseph Mouzna, and Labiod Houda. Di-
rection based clustering algorithm for data dissemination in vehicular net-
works. In Vehicular Networking Conference (VNC), 2009 IEEE, pages 1–6.
IEEE, 2009.
[91] Zhenxia Zhang, Azzedine Boukerche, and Richard Pazzi. A novel multi-hop
clustering scheme for vehicular ad-hoc networks. In Proceedings of the 9th
ACM international symposium on Mobility management and wireless access,
pages 19–26. ACM, 2011.
[92] Ameneh Daeinabi, Akbar Ghaffar Pour Rahbar, and Ahmad Khademzadeh.
Vwca: An efficient clustering algorithm in vehicular ad hoc networks. Journal
of Network and Computer Applications, 34(1):207–222, 2011.
[93] Agop Koulakezian. Aspire: Adaptive service provider infrastructure for vanets.
PhD thesis, University of Toronto, 2011.
[94] Efi Dror, Chen Avin, and Zvi Lotker. Fast randomized algorithm for hierarchi-
cal clustering in vehicular ad-hoc networks. In Ad Hoc Networking Workshop
(Med-Hoc-Net), 2011 The 10th IFIP Annual Mediterranean, pages 1–8. IEEE,
2011.
[95] Zaydoun Y Rawashdeh and Syed Masud Mahmud. A novel algorithm to form
stable clusters in vehicular ad hoc networks on highways. EURASIP Journal
on Wireless Communications and Networking, 2012(1):1–13, 2012.
[96] Zaydoun Y Rawshdeh and Syed Masud Mahmud. Toward strongley connected
clustering structure in vehicular ad hoc networks. In Vehicular Technology
Conference Fall (VTC 2009-Fall), 2009 IEEE 70th, pages 1–5. IEEE, 2009.
157

[97] Dov Monderer and Lloyd S Shapley. Potential games. Games and economic
behavior, 14(1):124–143, 1996.
[98] Zhu Han. Game theory in wireless and communication networks: theory, mod-
els, and applications. Cambridge University Press, 2012.
[99] Simon Fischer and Berthold Vöcking. Evolutionary game theory with appli-
cations to adaptive routing. In European Conference on Complex Systems
(ECCS), page 104, 2005.
[100] Oriol Sallent, Jordi Pérez-Romero, Ramón Agusti, Lorenza Giupponi, Clemens
Kloeck, Ihan Martoyo, Stefan Klett, and Jijun Luo. Resource auctioning
mechanisms in heterogeneous wireless access networks. In Vehicular Technol-
ogy Conference, 2006. VTC 2006-Spring. IEEE 63rd, volume 1, pages 52–56.
IEEE, 2006.
[101] Wenhui Zhang. Bearer service allocation and pricing in heterogeneous wireless
networks. In Communications, 2005. ICC 2005. 2005 IEEE International
Conference on, volume 2, pages 1367–1371. IEEE, 2005.
[102] Ho Chan, Pingyi Fan, and Zhigang Cao. A utility-based network selection
scheme for multiple services in heterogeneous networks. In Wireless Networks,
Communications and Mobile Computing, 2005 International Conference on,
volume 2, pages 1175–1180. IEEE, 2005.
[103] Giuseppe Bianchi. Performance analysis of the ieee 802.11 distributed co-
ordination function. IEEE Journal on selected areas in communications,
18(3):535–547, 2000.
[104] Kenneth Sorle Nwizege, Michael MacMammah, and Godson Ivbuobe Ik-
hazuangbe. Performance evaluation of path loss exponents on rate algorithms
in vehicular networks. International Journal of Emerging Science and Engi-
neering (IJESE), 1:103–108, 2013.
[105] Dusit Niyato and Ekram Hossain. Dynamics of network selection in heteroge-
neous wireless networks: an evolutionary game approach. Vehicular Technol-
ogy, IEEE Transactions on, 58(4):2008–2017, 2009.
158

Optimized Communication in 5G-Driven

More Related Content

What's hot (17)

Similar to Optimized Communication in 5G-Driven (20)

Recently uploaded (20)

Optimized Communication in 5G-Driven