Embed Size (px)
Citation preview
A TIME SYNCHRONIZED HYBRID VANET TO
IMPROVE ROAD SAFETY
A THESIS
Submitted by
DAHLIA SAM
in partial fulfillment for the award of the degree
of
DOCTOR OF PHILOSOPHY
Department of Computer Science and Engineering
FACULTY OF ENGINEERING AND TECHNOLOGY
Dr. M.G.R.
EDUCATIONAL AND RESEARCH INSTITUTE
UNIVERSITY (Decl. u/s 3 of the UGC Act 1956)
CHENNAI 600095
AUGUST 2015
DECLARATION BY THE CANDIDATE
I declare that the synopsis of the thesis entitled, “A TIME
SYNCHRONIZED HYBRID VANET TO IMPROVE ROAD SAFETY”, submitted by
me for the degree of Doctor of Philosophy is a bonafide record of work carried out by me
during the period from July 2010 to Dec 2014 under the guidance of Dr. V. Cyril Raj,
Professor, Dean [E & T], Dr. M. G.R. Educational and Research Institute, University and
has not formed the basis for the award of any degree, diploma, associate-ship, fellowship,
titles in this or any other University or other similar institution of higher learning and
without any plagiarism.
I have also published my papers in International Journals (Scopus rated) as
per list of publications in the Annexure.
Signature of Research Scholar
BONAFIDE CERTIFICATE
Certified that the synopsis of the thesis entitled “A TIME
SYNCHRONIZED HYBRID VANET TO IMPROVE ROAD SAFETY” is the
bonafide work of Ms. DAHLIA SAM (Register No. CS10D002) who had carried out the
research under my supervision and without any plagiarism to the best of my knowledge.
Certified further, that to the best of my knowledge, the work reported herein does not form
part of any other thesis or dissertation on the basis of which a degree or diploma was
conferred on an earlier occasion on this or any other scholar.
Dr. V. Cyril Raj
Supervisor
Professor & Dean [E & T]
Dr. M.G.R. Educational and Research Institute University
ABSTRACT
VANET is a type of mobile ad hoc network in which the moving vehicles act as
nodes. It can be simply defined as ‘computer network on wheels’. There has been lot of
research done over the last few years for using VANETs in many applications including
accident prevention, real time safety alerts, improved navigation system, media and
entertainment in vehicles etc.
In this work a hybrid VANET has been developed to improve road safety for both
the drivers and the pedestrians on the road. The hybrid VANET is comprised of VANET
together with roadside sensors and/or pedestrian body unit. These communicate wirelessly
with each other to form an Intelligent Transport System (ITS). The conventional VANET
has only vehicular nodes and will be subject to frequent network disconnections especially
in low traffic areas. Due to this some events in the road may go undetected. It may also
happen that the events detected may not be transmitted to all the vehicles due to lack of
connectivity. In such situations, static roadside sensors could play an important role to keep
the network connected while constantly detecting the happenings in the road. In the first
method proposed here, these wireless roadside sensors were deployed at fixed distances
such that they can communicate with each other. This will make the network constantly
connected and more efficient. It also proves to be a cost effective and feasible option. This
is because sensor technology is well developed and is cheaper to deploy. The Hybrid
VANET with roadside sensors proved to be a promising solution as per the simulation
results.
Another method used to detect human presence in the path of the vehicle was using
a body unit. Though the roadside sensors are good enough to detect most of the road events,
its detection range for humans is lesser than the detection range for vehicles. To ensure that
the presence of the pedestrians will not go undetected at any time, an additional pedestrian
body unit (PDU) is included in the Hybrid VANET. This gives an extra level of protection
to vulnerable pedestrians like kids, handicapped or simply distracted individuals. The
ii
Hybrid VANET system will make sure that any pedestrians on the road or any incident
happening on the road is detected on time and communicated to the vehicles. The alert
message can be communicated to the drivers to give them sufficient time to take appropriate
decision. It can also be connected to the vehicle control system, which automatically
activates the brake control or the throttle control.
For the above-mentioned VANET application, the vehicular nodes have to
constantly communicate with each other as well as with the roadside equipment or the
pedestrian body unit. The messages communicated are very time critical as even a delay of
seconds could lead to fatal accidents. In order for the messages communicated between the
nodes of the hybrid VANET to be meaningful and beneficial, it is important that the clock
time of all the nodes in the network is synchronized. This includes the clocks of the
vehicular nodes, the roadside sensors and the pedestrian body units. To address this issue a
Hybrid Clock Synchronization (HCS) algorithm was developed to synchronize the clock
times in the highly dynamic H-VANET environment. The proposed HCS algorithm
accommodates the frequent topology changes of the VANET. It also takes place within the
few seconds that the vehicles and the other nodes stay in the communication range of each
other.
The above-proposed time synchronized Hybrid network was simulated using
VANET simulators and supports the above requirements. The Black Spot scenario was used
for the experiments. The vehicle used in the experiments was equipped with automatic
braking control wherein the drivers can override and take manual control anytime. The first
case considered for the experiments was when the driver takes manual control. This
situation was analyzed using game theory. It was seen that when both the vehicle and
pedestrian nodes cooperate, accidents could be completely avoided in Black Spots. In the
second case, the time synchronized hybrid vehicle control system that was developed takes
control. The information broadcasted by the pedestrians’ body unit is given as an input to
the vehicle control system which constantly checks for the presence of pedestrians in its
range. Whenever it senses the presence of humans, it compares its location information and
other sensor information to find if the pedestrian is in the bad set i.e. the current position
will eventually lead to a crash. In such cases, it sends a control signal to the advanced
braking system. This will reduce the chances of crash due to human error caused by his
perception-reaction. This was proved mathematically. Some field experiments were also
iii
done. The results showed that accidents could be avoided to high speeds of up to 95 km/h
using the vehicle control system, which without the system would have led to a fatal
accident. On the whole, the vehicle control system using the Hybrid VANET is a very
reliable and efficient solution than any of the existing methods to prevent road accidents.
iv
ACKNOWLEDGEMENTS
First and foremost I should thank Almighty God for helping me to successfully
complete my research.
I convey my gratitude to Thiru. A. C. Shanmugam, our Founder Chancellor, Er.
A. C. S. Arun Kumar, our dynamic President for giving me the opportunity to carry out
my research in Dr. M.G.R Educational and Research Institute University. I also extend my
gratitude to Dr. Meer Mustafa Hussain, Vice Chancellor, Dr. C. B. Palanivelu,
Registrar, Dr. P. Aravindan, Principal, Director R & D, and Dr. A. Thirunavukkarasu,
Dean-Research.
I would like to express my sincere gratitude to my Supervisor, Dr. V. Cyril Raj,
Professor, Dean [E&T], for accepting me as his Research Scholar and providing valuable
guidance all through my research. My research work would not have been possible without
his bounteous efforts which made me strive for excellence. I am also indebted to my
Doctoral Committee members Dr. S. P. Rajagopalan, Professor Emeritus & Dr. T.
Bhuvaneshwari, Asst. Professor, Government Arts College, Ponneri, for their thoughtful
guidance.
My sincere thanks to the Dr. Sumathy Easwaran, HOD [CSE], the other staff of
the Dept. of CSE, fellow research scholars and staff members of the Department of
Research for their continuous support. Finally I would like to thank my husband, my
children, my parents and my other family members for the moral support and
encouragement throughout my research work.
DAHLIA SAM
v
TABLE OF CONTENTS
Chapter
no.
Title Page
no.
Abstract ii
List of Tables xiii
List of Figures xiv
List of Abbreviations xvi
1 Introduction
1.1 Road Accidents 1
1.2 Causes of Human Error 2
1.3 Motivation 3
1.4 Research Objective 5
1.5 Problem Statement 6
1.6 Organization of the thesis 6
2 Literature Survey
2.1 Overview 7
2.2 Work done in VANETs 7
2.3 Contemporary work on clock synchronization 8
2.4 Limitations Observed 12
3 Methodology
vi
Chapter
no.
Title Page
no.
3.1 Introduction 14
3.2 Vehicular Communication 14
3.2.1 Hybrid VANET with roadside sensors 16
3.2.2 Hybrid VANET with pedestrian body unit 17
3.3 Clock Synchronization in H-VANETs 18
4 Vehicular Ad hoc Networks (VANETS): The Platform for an
Intelligent Road Safety System
4.1 Introduction 20
4.2 Characteristics of VANETs 21
4.3 Applications of VANETs 23
4.3.1 Safe smart driving 23
4.3.2 Road services 23
4.3.3 Media and Comfort applications 23
4.3.4 Post-accident investigation 24
4.4 Standards and Protocols 24
4.4.1 Physical Layer 24
4.4.2 MAC Layer 24
4.4.3 Network Layer 25
4.5 VANET Simulators 25
4.5.1 MOVE 27
4.5.2 TraNS 28
4.5.3 VanetMobiSim 28
vii
Chapter
no.
Title Page
no.
4.5.4 NCTUns 28
4.5.5 GrooveNet 29
4.5.6 MobiREAL 29
4.6 Summary 31
5 Hybrid VANET (H-VANET): A Practical Approach to Improve
Road Safety
5.1 Introduction 32
5.2 Proposed Hybrid Vehicular Ad hoc Network 33
5.3 Importance of a Hybrid VANET 33
5.3.1 Road Factors 33
5.3.2 Environmental Factors 34
5.3.3 Human Factors 34
5.3.4 Animal Factors 34
5.4 Model of H-VANET 36
5.5 Experimental Results 37
5.5.1 Field Tests 37
5.5.2 Simulation Results 39
5.6 Summary 42
6 Comparison of Existing Clock Synchronization Protocols
6.1 Computer Clocks 43
6.1.1 Hardware Clock 44
6.1.2 Software Clock 44
viii
Chapter
no.
Title Page
no.
6.1.3 Logical Clock 45
6.1.4 Physical Clock 45
6.2 Clock Synchronization in Distributed Systems 45
6.3 Importance of Clock Synchronization in Computational Systems 48
6.4 Clock Synchronization Terminologies 49
6.5 Classification of Clock Synchronization Protocols 50
6.5.1 Internal vs. External Synchronization 50
6.5.2 Master-Slave vs. Mutual Synchronization 51
6.5.3 Probabilistic vs. Deterministic Synchronization 51
6.5.4 Clock Correction vs. Clock Assumption 52
6.5.5 Pair wise vs. Global Synchronization 52
6.5.6 Sender-Receiver vs. Receiver-Receiver Synchronization 52
6.5.7 Level based vs. Diffusion based 53
6.6 Synchronization Algorithms for Wired Networks 54
6.6.1 Christians Algorithm 54
6.6.2 Network Time Protocol (NTP) 54
6.6.3 Coupled Oscillator Phenomenon 55
6.7 Clock Synchronization in Wireless Mobile Ad hoc Networks 56
6.7.1 Reference Broadcast Synchronization (RBS) 56
6.7.2 DTSR 57
6.7.3 CS-MNS 57
ix
Chapter
no.
Title Page
no.
6.8 Clock Synchronization in Wireless Sensor Networks 59
6.8.1 All-node-based method 59
6.8.2 Cluster-based method 60
6.8.3 Fully Localized Diffusion based method 60
6.8.4 Fault Tolerant Diffusion based method 61
6.8.5 Secure Clock Synchronization 62
6.9 Comparison of Protocols 63
6.10 Summary 66
7 Hybrid Clock Synchronization (HCS) Algorithm for the H-
VANET
7.1 Introduction 67
7.2 Related Work 68
7.3 Hybrid Clock Synchronization (HCS) Algorithm 68
7.4 Results 71
7.4.1 Theoretical Results 71
7.4.2 Simulations 72
7.5 Summary 74
8 Hybrid VANET with Pedestrian Body Unit to Improve Safety in
Black Spots
8.1 Introduction 76
8.2 Background 77
8.3 Hybrid VANET with Pedestrian Body Unit 78
x
Chapter
no.
Title Page
no.
8.3.1 Integrating Body Unit with H-VANET 78
8.3.2 Laboratory Test 80
8.4 Black Spot Management 82
8.4.1 The Game Structure 85
8.4.2 Payoff Calculations 85
8.5 Simulations 88
8.6 Summary 90
9 VANET Based Vehicle Control System to Avoid Human Error
9.1 Introduction 91
9.2 Vehicle Stopping Distance and Time 91
9.2.1 Driver Reaction Time 92
9.2.2 Mechanical Response Time 93
9.2.3 Stopping Distance 93
9.3 Vehicle Control System 95
9.4 Mathematical Evaluation 100
9.5 Simulations 105
9.6 Experiments 107
9.7 Summary 110
10 Conclusion And Future Work
10.1 Summary of research work 112
10.2 Conclusion 114
xi
Chapter
no.
Title Page
no.
10.3 Future Work 116
References 117
Web References 124
List of Publication 125
xii
LIST OF TABLES
Table No. Title Page No.
4.1 Network and Traffic Simulators 26
4.2 Comparison of VANET Simulators 30
5.1 Advantages of H-VANET 35
5.2 Prototype testing platform 38
5.3 Average time taken for an alert message to reach all the nodes
in a group
39
5.4 Simulation parameters 40
6.1 Comparison and classification of the different protocols 53
6.2 Evaluation of the parameters supported by different protocols 65
8.1 Payoff table for the game 87
8.2 Simulation Parameters 88
8.3 Decision making and the occurrence of crash 90
9.1 Vehicle stopping distances 95
9.2 Simulation Parameters 106
9.3 Maximum safe car speeds for different reaction times 107
xiii
LIST OF FIGURES
Figure
No.
Title Page No.
1.1 Stopping distances under different conditions 4
1.2 A massive chain reaction crash in UK involving more than 100
vehicles
5
3.1 Communication in VANET 15
4.1 VANET Model 21
4.2 Types of simulators that support unidirectional communication 26
4.3 Types of simulators that support bidirectional communication 27
5.1 Model of the hybrid VANET 36
5.2 Packets transmitted between the vehicle nodes and RSU 41
5.3 Number of messages delivered within the acceptable time
window
42
6.1 Fast, Slow and Perfect clock with respect to UTC 44
6.2 Example of a Distributed System 46
7.1 Reply message format 70
7.2 Time sequence for one synchronization cycle 71
7.3 Performance of HCS when no vehicle with time difference is
added
73
7.4 Stability of the algorithm when new vehicles with time difference 74
xiv
Figure
No.
Title Page No.
enter the group
8.1 System Model of a Hybrid VANET with PBU 79
8.2 Alerting the vehicle 80
8.3 GPS position of pedestrian and vehicle 81
8.4 Tracing out the paths 81
8.5 Warning message 82
8.6 An example of a black spot 83
8.7 Schematic representation of the vehicle communication system 84
8.8 Time delay for the vehicle 87
8.9 Reduced accident possibility with the H-VANET alert system 89
9.1 Framework of the vehicle control system using H-VANET 97
9.2 Block diagram of the vehicle control system 99
9.3 Process flow diagram of the vehicle control system 100
9.4 Collision Scenario 101
9.5 A sample trajectory of a single vehicle 103
9.6 Bad set within the lower and upper bounds of displacements 104
xv
LIST OF ABBREVIATIONS
ATSP : Attack-tolerant Time-Synchronization Protocol
ASP : Automatic Self-time-correcting Procedure
CBS : Clapping and Broadcasting Synchronization
CSA : Clock Synchronization Algorithm
CS-MNS : Clock Sampling Mutual Network Synchronization
DAS : Driver Assistance System
DSRC : Dedicated Short Range Communication
DTN : Delay Tolerant Networks
DTSR : Distributed Time Synchronization
FCC : Federal Communication Commission
GPS : Global Positioning System
GTSP : Gradient Time Synchronization
HCS : Hybrid Clock Synchronization
H-VANET : Hybrid Vehicular Ad hoc Network
InVANET : Intelligent Vehicular Ad hoc Network
iMANETs : Internet based MANET
xvi
ITS : Intelligent Transport Systems
IVC : Inter-vehicle communication
LAN : Local Area Network
MAC : Medium Access Control
MANET : Mobile Ad hoc Network
MOVE : MObility model generator for VEhicular networks
NCTUns : National Chiao Tung University Network Simulator
NTP : Network Time Protocol
PBU : Pedestrian Body Unit
RBS : Reference Broadcast Synchronization
RTT : Round Trip Time
RSU : Road Side Unit
RVC : Roadside-to-vehicle communication
TraNS : Traffic and Network Simulator Environment
UAN : Underwater Acoustic Network
UTC : Universal Coordinated Time
VANET : Vehicular Ad hoc Network
VNTA : Vehicular Networks & Telematics Applications
WAN : Wide Area Networks
WSN : Wireless Sensor Networks
WBAN : Wireless Body Area Networks
WMN : Wireless MESH netwo
EXAMINER - I
LIST OF CORRECTIONS SUGGESTED BY THE EXAMINERS
AND INCORPORATIONS ARE DETAILED BELOW
S. No. Examiners Comment Corrected Sentence Correction
Page No.
1 Presentation style needs to be
improved
Made appropriate changes in
improving the language standard
Entire thesis
2 Chapter 1 - discuss about
research problems,
motivations of research etc.
Added the necessary details in
section 1.1, 1.2 and 1.3
1-5
3 References insufficient in
Chapter 2
Added more references to literature
review in section 2.2
7,8
4 Chapter 3 - discuss different
methodologies
Added a brief description about the
methodologies used as the new
chapter 3
14-20
5 Is there a Pseudo code or
algorithmic specifications for
H-VANET?
No Not
applicable
in our
context
6 How to detect the human
nodes?
Pedestrians are assumed to have GPS
equipped body unit like smart phone
as explained in section 8.3.1
78-79
7 How the system detects traffic
police?
The traffic police will not be within
the bad set and hence no control will
be sent to the braking system ,
discussed in Section 9.3
95-100
8 Will accident be avoided at
high speed like 120 km/h
Accidents cannot be avoided for
speeds above 95 km/h as per
experimental results in section 9.6
108
9 How the system works in high
traffic volume?
Out of scope of current work. It is
mentioned as part of future work in
section 10.3.
116
10 No Problem definition in
chapter 1
Given in section 1.5 6
11 Conclusion should highlight
the achieved results
Conclusion edited section 10.2 114, 115
12 Reference section should be
organized using a standard
referencing style
Harvard standard is used for
organizing References as per our
University norms
117-124
13 Implementation does not
answer situational questions
Some additional experimental results
have been included in section 9.6
107-110
Signature of the Research Scholar Signature of the Supervisor
(DAHLIA SAM) (Dr. V. CYRIL RAJ)
EXAMINER - II
LIST OF CORRECTIONS SUGGESTED BY THE EXAMINERS
AND INCORPORATIONS ARE DETAILED BELOW
S. No. Examiners Comment Corrected Sentence Correction
Page No.
1 Abstract
Please structure the abstract
Abstract has been structured and
more details included as suggested
ii - iv
2 Chapter 1
Include citations wherever
applicable
Rewrite the research objective
Specify the source of Fig 1.2
Appropriate citations and image
source have been given
Research objective is written more
clearly
4,5
3 Chapter 2
Discuss how the limitations
observed were concluded?
Rewrite the problem statement
Limitations observed is discussed in
section 2.4
Problem statement rewritten and
given in Chapter 1 section 1.5
6, 12, 13
4 Chapter 3
Contents should be related to
the title of the chapter
Please avoid poor clarity
images like fig. 3.2 in the thesis
Mathematical equations must be
numbered
Specify the source of figures
Chapter 3 has been rewritten,
reorganized and made as the new
chapter 6
Original Fig 3.2 is removed
Mathematical equations have been
numbered
Figures in this chapter are our own
43-66
5 Chapter 4 Self-drawn figures 20-31
Specify the source of figures
6 Chapter 5
Are the advantages presented in
table 5.1 general or the work
carried by you?
Review the graphs
They are the advantages of H-
VANET system developed as part of
this work
Co-ordinates metrics have been
added in the graphs
35, 41,42
7 Chapter 6
How did you arrive at the
conclusion that no existing
algorithm fully supports
VANET?
Review graphs
Chapter 6 is revised and given as
chapter 7 here. The existing
algorithms have been discussed in
detail in the current chapter 6.
Co-ordinates have been added to
graphs
67-75
8 Chapter 7
Specify the source of figures
Improve clarity of images
Review graphs presented
Chapter 7 is given here as chapter 8.
Images are our own.
Graphs have been revised.
76-90
9 Chapter 8
Give statistics as on 2015
Mathematical equation must be
numbered and citations given
Specify source of figures
Chapter 8 is now given as Chapter 9
Statistics given in section 1.1 page 1
Mathematical equations have been
numbered and citations given
Figures are our own
91-111
10 Chapter 9
Convey the shortcomings of the
work
It is given as part of future work in
section 10.3
116
Signature of the Research Scholar Signature of the Supervisor
(DAHLIA SAM) (Dr. V. CYRIL RAJ)
CHAPTER 1
INTRODUCTION
1.1 Road Accidents
The process of rapid urbanization has resulted in an unparalleled revolution in the
growth of motor vehicles worldwide. The increase in vehicles in the last decade has put lot
of pressure on the existing highways thereby increasing the number of road accidents. Auto
accidents have been the leading cause of preventable deaths in many countries over the last
several years. It is the 9th leading cause of death around the world and is predicted to
become the 5th leading cause by 2030. Globally, nearly 1.3 million deaths and 50 million
injuries occur annually due to road crashes. This accounts to 2.2% of all the types of deaths
and 3,287 deaths in a day. In India alone, over 2,30,000 deaths occur each year due to auto
accidents. India has earned a distinction of being the leading country in the number of road
accidents. This alarming increase in morbidity and mortality has become a matter of great
concern around the world. Traffic accidents cost losses of about 518$ billion USD
globally. In India, since 2001 there is an increase of 202 % in the number of two wheelers
and 286 % in the number of four wheelers. India is the second largest motorcycle and
fourth largest commercial vehicle manufacturer in the world. In 2015, 2.03 million
passenger cars were sold in India (Indian Motor Vehicles n.d). However, there has hardly
been any road expansion to accommodate this increase. Pedestrian fatalities have also gone
up by 15% in the last 10 years.
Among the total number of road accidents, 57% were caused solely by human error
and it has also been the contributing factor in more than 90% of the accidents. The other
causes include environmental factors, mechanical faults etc. The main reason for this
human error is the limitations in the information processing abilities in human beings. In
critical accident scenarios, the human limitations are exceeded by the situation that leads to
an accident. The reason for the accident in such cases is listed as “Human Error” or
2
“Driver Error”. The factors that contribute to driver error include speeding, drunken
driving, driver fatigue, and distractions like cell phones, eating, smoking, listening to
music or talking to fellow passengers. All these reduce the time the driver gets to observe,
process the incident on the road and to react in a timely manner. The outcome is most
likely an auto accident leading to death, injuries or property damage.
1.2 Causes of Human Error
The vehicle drivers have to process a continuous flow of information that comes as a
visual input. This includes the highway, traffic signals, pedestrians, other cars,
surroundings etc. In addition, the driver will have loads of thoughts going on in his mind
like trying to remember the tasks of the day, remember directions, worrying about
something etc. All these add ups becomes the internal inputs. There are also high chances
that the driver is exposed to other auditory input like music, mobile phone, chatting with
fellow passengers etc. As a result, the human brain processing capacity is drained out.
However, under normal circumstances, the driver manages to process and respond to all
these inputs.
In precarious situations, more attention is needed for e.g.: during low visibility
evening time or night times, high traffic in highways, snowfall or when many pedestrians
are walking on the road. This causes the driver to react to only a subset of the available
inputs. The brain does not process the rest of the information. In such cases, the driver is
most likely to respond in a wrong way. Research on road accidents caused by human error
points out to three main types of errors made by drivers (Road Accidents n.d., Stopping
Distance n.d., Stopping Sight Distance n.d.).
Perceptual Error: In some situations, like glare, low lights or when the driver is
tired, sleepy or drunk, the driver is unable to see some crucial details. This also
happens in Accident Black Spots that include sharp corners in straight road, steep
slopes, a hidden junction or concealed warning signs.
Distraction Error: When the drivers’ mind is concentrating on something else, the
driver often fails to notice a clearly visible pedestrian or car. This is referred to as
3
“Blindness due to inattention”. This will cause a delay in his reaction or an error in
reaction (Khairunnisa & Syah 2014).
Response Error: Even though the driver gets the information correctly, his response
may sometimes be a wrong one. E.g.: Hitting the accelerator instead of the brakes,
making a sharp turn or hard braking to avoid one accident that could lead to
another.
1.3 Motivation
Let us consider an example of a pedestrian who carelessly walks in front of the car.
Suppose the car driver is travelling at 30 mph that is a very reasonable speed. If he notices
the pedestrian at least 45 feet away, the driver has enough time to process the situation in
his brain and apply brakes. The vehicle comfortably stops without hitting the pedestrian. If
however the driver was travelling at a higher speed or if he notices the pedestrian when he
is at a lesser distance, the situation changes dramatically. The severity of accidents at
different vehicle speeds is shown in Fig. 1.1. Even if the driver observes, processes the
situation and applies brake with the same speed as in the first case, the car will hit the
pedestrian. The car that hits at a minimum of 15 mph can seriously injure or kill the driver
or the pedestrian.
4
Fig 1.1: Stopping distances under different conditions
The same scenario can occur in the presence of wandering animals on the roads or a
block due to bad weather condition. The deadliest and disastrous form of such driver error
accidents that are very common in most developed countries is the multi-vehicle collision.
In freeways due to high traffic speed, if one car suddenly halts due to some reason, those
behind it cannot stop in time. This leads to a chain-reaction crash, something similar to the
scenario in Fig 1.2 (Huge chain-reaction crash in UK, Europe 2013). It is clear that that has
to be something more than just the human intervention and reaction to prevent such
accidents caused due to lack of adequate driver reaction time. There has to be a system to
assist the drivers in order to react faster.
5
Fig 1.2: A massive chain reaction crash in UK involving more than 100 vehicles
1.4 Research Objective
The objective of this research is to reduce the time critical accidents due to human
error. Most of the human error is caused due to insufficient time to process the situation
and react appropriately. If these types of accidents could be reduced, it would
automatically improve road safety. To achieve this, a system to improve road safety is
developed atop VANET that collects, processes and shares real time road information
wirelessly. VANET (Vehicular Ad Hoc Network) is a type of network in which the
vehicles on the road and roadside beacons form the mobile nodes of the network.
In road safety and other time critical applications, the network has strict delay
constraints. It is very crucial that the messages communicated via the VANET reaches the
participating nodes on time. For that, having a synchronized clock between communicating
nodes – vehicles, roadside units and pedestrian units, is very important. To ensure this,
VANET system developed should also be time synchronized. The system should be
capable of giving timely warnings about road conditions. This way road safety can be
improved by reducing the major cause of accidents.
6
1.5 Problem Statement
The main problem considered in this work is to develop a Hybrid VANET
consisting of roadside units and pedestrian body units in order to enable vehicle and
pedestrian entities to exchange time synchronized information related to their location and
enabling further process so that time critical accidents can be avoided.
1.6 Organization of the thesis
The report is organized in the following way. Chapter 1 gives the introduction about
the research problem, motivation behind the research and the objectives of this research.
The exact research problem that was taken is defined at the end of this chapter. The
detailed literature survey of the work done in this area has been described in Chapter 2. It
also briefs the limitations observed. Chapter 3 gives a gist of the different methodologies
that are used to meet the research objective. Chapter 4 explains in detail about the work
done so far in VANETs. The different simulators that can be used for VANET simulation
were studied and a suitable one was identified. In Chapter 5, a Hybrid VANET is proposed
that improves the road safety by alerting the drivers about road incidents as quickly as
possible. It is followed by Chapter 6, in which a detailed view on the different existing
clock synchronization algorithms were analyzed. The algorithms needed for different types
of network are compared to find out if any of the exiting algorithms can be applied for the
proposed VANET based system. Chapter 7 describes the proposed Hybrid clock
synchronization algorithm (HCS) for Hybrid VANETs. The advantages of the algorithm
and the simulation results are presented. An extension of the work is proposed in Chapter 8
in which pedestrian body nodes are included as part of the Hybrid VANET. This gives an
extra level of protection to human road users. The final part of the work is connecting the
Hybrid VANET to the vehicle control system, which is given in Chapter 9. This aids to
further reduce the possibility of accidents due to human error. The mathematical
verification of the proposed control system and the field test results are also given. Chapter
10 concludes the thesis highlighting the major research contributions. The scope for future
work in this area is also briefed.
7
CHAPTER 2
LITERATURE SURVEY
2.1 Overview
In order to meet the objective of developing a VANET based system to prevent road
accidents, first the current work related to these needs to be studied. In this chapter, a
detailed review of the current literature in the areas of clock synchronization in different
networks is given. A study of vehicular networks and its recent developments in the area of
road safety is also presented.
2.2 Work done in VANETs
There has been a lot of advancement in hardware, software and communication
technologies over the last few years. This has also led to the development and design of
different types of networks that are deployed in varied environments. One such interesting
field in which networks and communication have crept into is the vehicles on road. It is an
area with tremendous potential for growth. Communication between vehicles using
modern communication technologies (e.g. cellular networks, Bluetooth etc) has become
very common. However, direct communication between two vehicles has been under
research in the past decade.
The first form of vehicular communication that was proposed, used optical laser or
infrared laser. In this, each vehicle can communicate with the vehicle directly in front of it
and the one directly behind it in the same lane. This system has the drawback that each
vehicle can communicate with only two vehicles. The communication is also very sensitive
to the alignment of the vehicles and weather conditions like rain, fog or snow. Another
method proposed was communication using Radio Frequency (RF). Here the vehicle can
broadcast to all the vehicles in its range. Reservation ALOHA (R-ALOHA) protocol is used
for medium access. Later in 1999, the Federal Communications Commission (FCC)
allocated 75 MHz of spectrum at 5.850-5.925GHz for Dedicated Short Range
8
Communications (DSRC). The allotted frequency spectrum enabled wireless
communication between vehicle-vehicle and vehicle-roadside beacons without central
access point. This led to the development of VANETs and its related services.
VANET can be defined as “computer network on wheels”. It is a network with the
moving cars as the mobile nodes (Yousefi, Mousavi & Fathy 2006). These nodes
communicate with each other as well as with the roadside equipments which are within
ranges of 100 to 300 m based on IEEE 802.11p standard. There have been numerous
applications that have been proposed on top of this VANET that collects and processes real
time road information. These include accident prevention, real time alerts about road
conditions, collision warning, smart navigation, merge assistance, media and entertainment
(Sun, Bebis & Miller 2004; Zhao & Cao 2008; Suriyapaiboonwattana, Pornavalai &
Chakraborty 2009; Srinivetha & Gopi 2014; Chandramohan & Kamalakkannan 2014).
Vehicular ad hoc networks are expected to implement wireless technologies such as
dedicated short-range communications (DSRC), which is a type of Wi-Fi (Chandramohan
& Kamalakkannan 2014). The DSRC was established in 1999 by the Federal
Communication Commission (FCC) and allocated a frequency for wireless communication
between vehicles and roadside beacons. Other candidate wireless technologies are cellular,
satellite and WiMAX.
2.3 Contemporary work on clock synchronization
Clock Synchronization is a fundamental requirement in most distributed applications.
The topic has been under research and has been widely studied for many years now. There
have been many algorithms proposed which address different scales ranging from wired
LANs to wireless sensor networks. The Network Time Protocol (NTP) has achieved a
dominant position as the standard algorithm for both LAN and WAN networks. With the
emerging distributed infrastructures and applications like cloud computing, delay tolerant
networking, under water networking, vehicular networking etc., the problem of clock
synchronization is yet to be fully solved. Most of these systems operate ad hoc with mobile
or fixed nodes and without any centralized control.
9
The Network Time Protocol (NTP) is the current standard protocol designed to
distribute accurate and reliable time information for synchronizing on the Internet. It works
in diverse and widely distributed environments where the clocks are synchronized in a self-
organizing hierarchical configuration to the UTC (Mills 1994).
The above-mentioned NTP relies on a hierarchy of time servers and assumes that the
root servers have access to a standard real time source. A non-hierarchical peer-to-peer
approach for synchronizing clocks, referred as Classless Time Protocol has been proposed
by Gurewitz et al. (2006). This reduces the offset errors in order to minimize the global
network wide cost function.
Rui Fan proposed a solution for clock synchronization in highly decentralized
networks where the nearby nodes cooperate to perform some tasks and the far away nodes
interact rarely. In such cases, the Gradient Clock Synchronization algorithm works, in
which the nearby nodes are closely synchronized and the far away nodes are loosely
synchronized (Fan & Lynch 2006)
As large scale infrastructures and services started emerging, the problem of
scalability and node churn i.e. node failures, nodes joining or leaving the system became
an issue in distributed systems. There have been many fault tolerant protocols proposed
(Lamport 1985; Dolev et al. 1986; Ramanathan, Shin & Butler 1990; Lee et al. 2005) in
order to tolerate failures and achieve ultra-reliable assurance levels. A very novel
algorithm was proposed by Baldoni et al. (2010) that combine the gossip-based paradigm
with the nature inspired coupled oscillators phenomenon.
Over the last couple of decades, wireless networks are becoming an important
medium for distributed computations. Wireless applications are growing more diverse and
sophisticated. Time synchronization continues to be a critical requirement to maintain the
correctness and/or performance of many MANET applications, for power saving, network
throughput and efficiency of many protocols. In an IEEE 802.11-based mobile ad hoc
network, many new issues have entered the picture. These include developing an algorithm
that supports wireless, infrastructure-less network, which is more energy-efficient and
tolerates failures as well as dynamic network behavior.
10
RBS is an efficient internal clock synchronization algorithm for wireless networks
(Elson, Girod & Estrin 2002; Kuhn & Oshman 2009). Fan (2005) proposed an algorithm
for wireless networks with no infrastructure support. The algorithm performs both internal
and external synchronization.
In sparse ad hoc networks, the classical algorithms cannot be applied because of
limited communication range of wireless technology. This leads to frequent
reconfiguration of the network topology. Romer (2001) proposed a time synchronization
algorithm specifically for sparse ad hoc networks. Here the computer clocks are not
synchronized. Instead, devices generate time stamps using its local clocks and pass to other
devices. The receiving device adjusts its local time based on this. However this algorithm
shows inaccuracies due to the age of the time stamp and the number of hops a timestamp
has passed.
For multihop environments, an automatic self-time-correcting procedure (ASP) was
proposed by Sheu et al. (2006) to achieve clock synchronization. In the previous method,
each mobile host is responsible for exchanging timing information. As the number of hosts
increases, the scalability problem occurs due to up rise in transmission contentions. The
ASP has two features - First, a faster host gets a higher beacon transmission priority to
send its timing information out than a slower one. This increases the successful
transmission probability for faster hosts. Second, after collecting enough timing
information to accomplish synchronization by itself, a slower host can synchronize to the
faster one by self-correcting its timer periodically. Now this host gets its beacon
transmission priority increased so as to spread the timing information very fast throughout
the whole network.
Another high performance clock synchronization protocol for multihop MANET has
been proposed by Zhou & Lai (2007). This protocol was able to limit the clock offset to 50
µs with long-term stability. However this protocol was slow to get stabilized.
The Distributed Time Synchronization (DTSR) algorithm for MANETs proposed by
Lee et al. (2005) is very robust to traffic load variations and network overload caused by
the two-way message exchanges. However the method has high complexity and overhead.
11
A clock synchronization approach using weight coefficient in ad hoc networks is
proposed by (Wang et al. 2009). Here different weight coefficients are assigned to
corresponding nodes based on time difference between nodes. Later each node adjusts its
clock to achieve synchronization. This method eliminates the interference of stationary
nodes and nodes with unexpected time difference by making their weight coefficients as
zero.
Global clock synchronization protocols for Wireless Sensor Networks were
elaborately described by Li & Rus (2006). They have discussed four methods: node-based
approach, hierarchical cluster-based method, diffusion-based method and fault-tolerant
diffusion-based method.
In the work of Swain & Hansdah (2010), they have introduced the concept of
adaptive clock synchronization based on the application needs and the resource constraints
in the sensor networks. They have described a probabilistic method for clock
synchronization that uses the receiver-to-receiver synchronization described in Reference
Broadcast Synchronization (RBS) protocol. This deterministic protocol is extended to
provide a probabilistic bound on the accuracy of the clock synchronization. This protocol
is also fault tolerant.
Another work done by Hu, Park & Shin (2008) proposes an Attack-tolerant Time-
Synchronization Protocol (ATSP) in which sensor nodes cooperate and safeguard the time
synchronization protocol against malicious attacks.
The Clapping and Broadcasting Synchronization (CBS) for sensor networks (Qian et
al. 2010; Shen et al. 2011) is designed for large scale sensor networks which gives high
synchronization accuracy with low communication overhead by utilizing “broadcaster-
receiver” communication model. Most of the other protocols adopt the basic pairwise
communication model and as a result the synchronization overhead is not well controlled.
A major issue with WSN is that it has very strict energy-constraints. This is because
the batteries have limited capacities and cannot be replaced often. Such networks need
energy efficient protocols with decreased communication overhead and enhanced
performance. Many protocols designed specifically to meet this requirement are proposed.
12
Mirabella et al. (2008) proposed that the nodes can be allowed to sleep for a longer time
and exchange rare synchronization packets with a good quality common clock. Chin &
Tzen (2009) has adopted the flooding time synchronization based on one-way timing
messages in order to conserve energy.
The pair wise broadcast clock synchronization method (Noh & Serpedin 2007)
synchronizes a sunset of sensor nodes by over hearing the timing message exchanges of a
pair of nodes. As a result a group of nodes gets synchronized with no additional messages
thus making this method energy-efficient. Similarly, in the passive cluster based method
(Mamun-Or-Rashid, Hong & Chi-Hyung 2005), cluster of nodes are created using passive
clustering and then asynchronous averaging algorithm is used for clock synchronization.
This way the number of rounds and operations for converging time is reduced which in
turn reduces energy required.
The problem with the above-mentioned methods of enabling sensor on-off mode to
save energy is that a long-term synchronization error is caused due to instability and
nonlinearity. Chen et al. (2010) proposed a feedback based synchronization (FBS) scheme
to compensate the clock drift caused by internal and external perturbations.
The gradient time synchronization (GTSP) (Sommer & Wattenhofer 2009) is
designed to provide accurate synchronization between neighbouring nodes. The nodes
calibrate the logical clocks based on broadcast messages received from direct neighbours
thereby giving accurate results.
2.4 Limitations observed
VANETs have some unique characteristics that pose many challenges in designing a
VANET based system as well as choosing the appropriate clock synchronization
algorithm. Since VANETs face frequent disconnections, the communication between
entities would be unreliable. This issue needs to be addressed first.
The second issue of maintaining a synchronized clock is also challenging with
VANETS. There have been many protocols developed for maintaining the clocks of
different types of networks. There are a number of clock synchronization algorithms for ad
13
hoc networks, sensor networks and wired networks. However, there has not been any work
done so far to synchronize clocks in vehicular networks. The clock synchronization
algorithm (CSA) should cope with unreliable network transmission and massage latencies.
The clock synchronization algorithm should also be able to cope up with the rapidly
changing topology and high mobility. It is also important that the clock synchronization
overhead should not degrade the overall system performance. The already existing logical
and physical clock synchronization algorithms that stand good for wired networks cannot
be applied in ad hoc networks. Most of the algorithms described rely on a special master
node, which is either fixed one or dynamically selected one. Vehicular network has a
constant node churn with nodes moving in and out of the network frequently. This makes it
difficult to have a master node. The existing algorithms for wireless networks like RBS can
be applied for ad hoc situations. However, it fails in the VANET scenario where the
topology and the network participants keep constantly changing. So a whole new algorithm
taking into consideration all these issues has to be developed.
14
CHAPTER 3
METHODOLOGY
3.1 Introduction
The main aim of the work is to reduce accidents involving pedestrians in the road
as far as possible. Since the solution to the problem is based on VANET, it is important to
address certain issues before getting into the details of the system design. First the method
of detecting the pedestrian on the road and communicating to the vehicle has to be
finalized upon. Next it is important that the nodes participating in the VANET
communication needs to be time synchronized so that the messages passed between them
is valid. For this a suitable time synchronization algorithm needs to be implemented. The
issues addressed and the way the system is to be designed is briefly given below.
3.2 Vehicular Communication
VANET is a mobile network, in which the moving cars act as nodes. These nodes
communicate with each other as well as with the roadside equipment’s forming an
Intelligent Transport System (ITS). The communication can take place between vehicles or
between the vehicle and roadside units, known as RSUs within short ranges of 100 to 300
meters. Fixed RSUs connected to the backbone network must be distributed in the
highways to facilitate communication. At any point, the vehicles may or may not have
wireless access to the roadside units. As vehicles fall out of the signal range, other cars
may join in, connecting vehicles to one another to form a mobile inter network. (Forian,
2005; Zeadally et. al. 2012; VANETs n.d., Chandrasekaran 2009)
15
In 1999, the Federal Communications Commission (FCC) allocated 75 MHz of
spectrum at 5.850-5.925GHz for DSRC (Dedicated Short Range Communications). The
allotted frequency spectrum enabled wireless communication between the entities of the
VANET without central access point. In intelligent transportation systems, each vehicle
acts as a sender, receiver and a router to broadcast information. For VRC communication
to happen between vehicles and RSUs, vehicles must be equipped with a Global
Positioning System (GPS) and an on board Unit (OBU) that enables short-range wireless
ad hoc networks to be formed. Automotive companies like General Motors, Toyota,
Nissan, Chrysler, BMW and Ford have already started manufacturing cars equipped with
such devices which will pave the way for developing smarter communication technologies.
As promoted in ITS, vehicles communicate with each other via inter-vehicle
communication (IVC) as well as with roadside base stations via roadside-to-vehicle
communication (RVC). The communication in VANETS can be of mainly three types as
given below. These are represented in Fig 3.1.
Vehicle-to-Vehicle communication
Vehicle-to-Roadside communication
Roadside-to-Roadside communication
16
Fig 3.1: Communication in VANET
The first type of communication is between the vehicular nodes. The second type is
the communication of messages from the vehicular nodes to the roadside units. The final
type is the communication between any two roadside infrastructures. This could be RSU to
RSU communication or message transfer between RSU and base station for
communication with the internet. The main challenge in the communication between
vehicles is the connectivity problem. The vehicular node movement pattern is constrained
but they move with different velocities. Due to this the connections between the vehicles
may become weak or get lost.
3.2.1 Hybrid VANET with roadside sensors
VANET is currently under research and a lot of effort is needed before it can be
practically deployed. For all of the VANET based applications to work in reality, there has
to be a minimum market penetration of at least 10% of VANET equipped vehicles. An
equipped vehicle should have an on-board computer with embedded WiFi card and also
meet some other requirements. For reaching this 10% in a period of 3 years, at least 50%
of the newly produced cars should be VANET enabled i.e., it should support V2V and V2I
communications. This value added vehicles are on their way to being introduced in the
market as the technology has been theoretically proven to be effective and efficient.
While VANET equipped cars are under production in the recent years, there is
another major requirement for the system to work. It is also necessary to install a minimum
number of roadside access points, starting with national highways. The Road Side Units
(RSUs) should also be well equipped, installed and maintained properly. The wide scale
purchase, deployment and maintenance of the required infrastructure for such a system can
be a humongous task as well as an expensive affair. It may not be practically feasible over
the next few years.
Some researchers say that a pure VANET may or may not include roadside access
points. Even if roadside access points are deployed, it is only feasible to deploy them in
17
some important highways due to its high cost. Hence any two consecutive RSUs may not
be in the direct communication range of each other. Whenever there are enough vehicles
on the road a network is formed between the vehicles. However, if the vehicles in the road
are sparse, they may not be in the communication range of each other. This may commonly
happen in remote roads or during low traffic hours. There may not be any vehicles on the
road to sense an event. The vehicle will detect the event only when it is in the close
vicinity, often when it is too late to take any decisions. Even if one vehicle detects the
event and if there is no proper connectivity between the nodes, the collected information
cannot be shared with the other vehicles. The alert message may not get passed on to the
approaching vehicles. The whole basis of all the VANET related applications are,
successful and timely communication.
One solution to this problem is to combine VANETs with the already well developed
wireless sensor networks (WSNs), which are less expensive and has better chance to be
implemented faster than installing RSUs. The wireless sensor nodes can be deployed in the
roads between two access points which could constantly detect the road events and
obstacles. It keeps the network connected all the time and passes on the messages to the
vehicles entering the network. The new network which includes the roadside sensor nodes
and the vehicular nodes can be referred to as the Hybrid Vehicular Ad hoc Network (H-
VANET).
3.2.2 Hybrid VANET with pedestrian body unit
Due to increase in population, usage of vehicles is ever increasing. Similarly,
accidents occurring due to vehicles are also growing day by day. As a result, road traffic
death has risen up to become the fifth leading cause of human death in the society.
According to the Global Status Report on Road Safety, death rate has increased to 1.24
million per annum. Among these pedestrian accidents, highway hit and run death count
comprises of about 22% of the total number. These kinds of road users include
handicapped people, careless kids, drunken persons etc. A high concentration of pedestrian
death is seen in low and medium income countries that are becoming motorized.
18
The Hybrid VANET with roadside sensors is a suitable solution to detect any
obstacles or pedestrians on the road. In order to be 100 % sure that the pedestrians will
never go undetected, the Hybrid VANET is expanded to include pedestrian body nodes.
These days’ smart devices equipped with GPS is becoming very common. Almost every
pedestrian has either a smart phone or other smart device with them all the time. Hence it
can be used to communicate the pedestrians exact location information to the vehicle. This
gives an extra level of protection to the humans on the road.
3.3 Clock Synchronization in H-VANETs
The term ‘synchronization’ is very broadly used in many application areas and refers
to the coordination of events operating in unison. In nature, synchronization is observed
among bird colonies, ant colonies and animal colonies. In real world, synchronization is
important in fields like music to maintain the rhythm, in digital telephony, photography,
telecommunication, multimedia, cryptography, neuroscience and physics. In the field of
computer science, synchronization refers to synchronization of processes and
synchronization of data. Clock synchronization is a part of process synchronization and has
been under research for many years now. In the 18th century, synchronization of clocks had
been a critical problem in long distance ocean navigation until the invention of marine
chronometer. Accurate time keeping and astronomical observations were required to
determine the distance travelled by a vessel and the direction. Similarly, in the 19th century
when railways were becoming popular, the difference in local time between the adjacent
towns started becoming noticeable. It was handled by synchronizing all the stations time to
the headquarters known as the ‘standard railroad time’. In 1971, Joseph C Hafele and
Richard E Keating studied the theory of relativity of time i.e. the time difference was
observed between two events as measured by observers moving relative to each other.
There is also a difference in time when the events are situated in different gravitational
masses. E.g.The clocks on a space shuttle runs slightly slower while the clocks on GPS and
Galileo satellites run slightly faster than the reference clocks on the earth. The laws of
nature are such that the space time will drift due to differences in either gravity or velocity.
This time dilation could affect planned meetings for astronauts. It was also observed that
the time dilation due to gravity measured by using atomic clocks on airplanes was found to
19
be slightly faster with respect to the clocks on the ground. This shows a significant effect
in GPS’s artificial satellites. Time dilations have also been observed due to height
differences of less than 1 meter. The need for strict time synchronization in the above and
other similar applications have motivated many researchers and scientists to come up with
some standards to maintain time.
The idea of Clock Synchronization can be explained as ‘concurrent processes trying
to bring their clock time close together by communicating with each other’. Whenever a
system works with multiple CPUs each having its own clock, the issue of time
synchronization comes into the picture. All the internal clocks are initially set to the same
value and the frequency at which the crystal oscillator of individual processors runs is
fairly stable. It is still impossible to guarantee that the crystals of different machines will
all run at exactly the same frequency. As time passes, due to the deviation in the frequency
of clocks, the time value of each will differ by some amount. A clock drift results due to
different clocks counting time at a slightly different rate. Practically, if a system has n
processors, all n clocks will run at slightly different rates, causing the clocks to gradually
get out of sync with each other. The result of this is a difference in clock values when read
out, which is referred to as clock skew or time skew. There is a need for some clock
synchronization protocol to overcome the consequences of this clock skew. Clock
synchronization (or Time synchronization) is simply adjusting two or more clocks in a
network to run at the same frequency as well as show the same time at a particular epoch.
The clocks can be in the same local system or distributed in the network. It is important so
that multiple unrelated processes running on different machines have the same
understanding about the temporal ordering of events. Only then will the communications
between the processes happen in a correct logical sequence.
The messages that are communicated between the entities of the H-VANET
proposed in this work, to avoid an accident is very time critical. So it is very important that
the nodes are time synchronized with each other. For this a Hybrid Clock Synchronization
(HCS) algorithm is developed to synchronize the clock times of the nodes in a highly
dynamic VANET environment.
20
CHAPTER 4
VEHICULAR AD HOC NETWORKS (VANETS): THE PLATFORM FOR AN
INTELLIGENT ROAD SAFETY SYSTEM
4.1 Introduction
A vehicular ad hoc network (VANET) uses cars as mobile nodes in a MANET. A
VANET turns every participating car into a wireless router or node, allowing cars
approximately 100 to 300 meters of each other to connect and create a network with a wide
range. As cars fall out of the signal range and drop out of the network, other cars can join
in, connecting vehicles to one another so that a mobile inter network is created. VANETs
are a sophisticated technology that integrates ad hoc network, wireless LAN, cellular
technology and sensor networks to achieve advanced intelligent communications between
vehicles, roadside sensors and infrastructure. It is anticipated that the first systems that will
integrate this technology are police and fire vehicles to communicate with each other for
safety purposes. Automotive companies like General Motors, Toyota, Nissan, Chrysler,
BMW and Ford have already started promoting this term. GPS and navigation systems
might benefit as they could be integrated with traffic reports to provide the fastest route to
work. The free VoIP services such as GoogleTalk or Skype can be used between
employees.
Thus VANETs are self-organizing and decentralized systems. These days’ cars are
equipped with devices to sense the surrounding environment. Researchers are working
hard to develop Intelligent Transport Systems (ITS) in which the vehicles can
communicate with each other as well as with roadside infrastructure. This will pave the
21
way for development of smarter communication technologies that will be helpful for many
applications that are currently under research.
Within the IEEE Communications Society, there is a Technical Subcommittee on
Vehicular Networks & Telematics Applications (VNTA). The charter of this committee is
to actively promote technical activities in the field of vehicular networks, V2V, V2R and
V2I communications, standards, communications-enabled road and vehicle safety, real-
time traffic monitoring, intersection management technologies, future telematics
applications, and ITS-based services. Intelligent vehicular ad-hoc network (InVANET) is
another term for promoting vehicular networking. InVANET integrates multiple
networking technologies such as Wi-Fi IEEE 802.11p, WAVE IEEE 1609, WiMAX IEEE
802.16, Bluetooth, IRA and ZigBee. A basic VANET model is shown in Fig 4.1.
Fig 4.1: VANET Model
4.2 Characteristics of VANETs
There are some characteristics that distinguish VANETs from other mobile ad hoc
networks that are:
All the nodes in a VANET (vehicle nodes and the roadside units) act as both
transmitters and receivers.
Topology: The mobility of vehicles is continuous and very fast especially on highways.
Thus the communication links between each vehicle is just for several seconds. The
links are established and broken fast. The result is a rapidly changing topology.
22
Predictable Mobility: Vehicles run on pre-built highways and roads. Hence the motion
pattern of the vehicles can be predicted based on the road topology and layout.
However there could be some uncertainty in the movement of vehicles depending upon
the layout of the road, the traffic density, structure of lane and of course the behaviour
of the drivers.
High Speed: The nodes in a VANET move at a very high average speed compared to
MANETs.
Variable Node Density: The number of nodes in a VANET can be very high in busy
highways and very sparse in remote highways. Similarly in a particular place, the
traffic may be at peak during busy office hours and minimum during midnight hours.
Hence any protocol designed should take into consideration both scenarios.
Frequent Disconnections: Since vehicles are constantly moving, the communication
links between them are constantly established and broken. In remote highways where
the vehicle density is low, existing links can break before the new links are formed.
This may lead to temporary disconnections of the network.
No energy constraints: Since the nodes in a VANET are vehicles, they have constantly
recharging batteries. Due to this abundant resource, vehicles can be equipped with GPS
or other devices.
No infrastructure: The communication between nodes in VANET is direct and does not
rely on any underlying infrastructure. However, it can be connected with the
infrastructure too.
Unbounded network size: VANETs are highly scalable as it can span through regions
of one city or several cities. One of the main challenges of VANETs is operability,
both in very loosely overloaded and highly overloaded networks. VANET must work
in all the situations such as very small density of road traffic and in situations with a
very high traffic density, in other words in the area of traffic jams and major
intersections of roads. The number of active nodes (vehicles) and protocol design has a
great impact on scalability.
Hard delay constraints: In VANETs applications such as the collision warning, pre-
crash sensing etc., the network does not require high data rates but has hard delay
constraints such as bandwidth routing time. Even the maximum delay will be crucial.
23
Better security: VANET nodes are more secure than nodes of other wireless networks.
4.3 Applications of VANETs
The wireless technology has become cheaper and permeating in the last decade that
promises many innovative vehicular applications in the future. Many possible applications
of VANETs have been discussed by Barba et al. (2012) and Cheng, Shan & Zhuang
(2011).
4.3.1 Safe smart driving
These applications focus on giving timely alerts to the drivers about collisions, poor
road conditions, traffic jams, accident and road construction warning systems etc. They
also include providing enhanced navigation and real time guidance to drivers while
merging, driving uphill/downhill or in curvy roads. These safety systems require transfer of
messages in the vehicular network to all vehicles in the range to alert the drivers to prevent
accidents. Another group of applications require the network protocols to forward the
messages from a sender to only the relevant receivers. Also, safety applications are time
sensitive and should be given priority over non-safety applications.
4.3.2 Road services
VANETs can also be helpful in finding the closest fuel station, restaurant or travel
lodge, automated toll payment and providing access to the internet. Another kind of
application focuses on connecting the vehicles to the internet using roadside beacons and
in inter-vehicle communications. These applications provide on-the-road games, media
streaming, digital billboards for advertisements, business mails etc.
4.3.3 Media and Comfort applications
This category of infotainment application includes web browsing, accessing emails,
video streaming etc. The time that would otherwise be wasted in travel, traffic jams,
tollgate queues can be used productively for personal or official work if connected to the
internet. With the help of VANETs, one can check business mails, use skype, browse or
watch a movie while on the road. Some of the VANET supported products to be used in
24
vehicles include remote keyless entry devices, personal digital assistants (PDAs), laptops
and mobile telephones.
4.3.4 Post-accident investigation
The roadside devices can store information about accidents that can be used later.
This will be helpful for investigators in forensic reconstruction and for insurance
companies.
4.4 Standards and Protocols
The standards and protocols used in the different network layers for VANET is
discussed below.
4.4.1 Physical Layer
The standard used for wireless communication is IEEE 802.11 in 5 GHz and 2.4
GHz spectrum band (Wi-Fi). The Federal Communications Commission (FCC) in US has
allotted 75 MHz of frequency spectrum at 5.850-5.925 GHz for Dedicated Short Range
Communication (DSRC). DSRC uses this for many private and public applications like
safety, real time traffic management, real time road information, in car entertainment,
email access, voice chat etc. In Europe, the band allotted for CAR 2 CAR communication
is between 5.885-5.905 GHz.
IEEE 802.11p is an approved amendment to the IEEE 802.11 standard to add
wireless access in vehicular environments (WAVE). This is also a cost-efficient solution
that can be applied in VANETs for both ON board units (OBUs) and Road side units
(RSUs). It uses the licensed ITS band of 5.9 GHz (5.85-5.925 GHz). The basic data rate for
this standard is 3 Mbps for a 10 MHz channel.
4.4.2 MAC Layer
Developing a reliable and efficient medium access control protocol is one of the
current research areas in VANETs. Medium sharing is particularly challenging in
VANETs due to high mobility and fast topology changes. The two approaches developed
for the C2C-CC radio system are IEEE 802.11p and IEEE P1609.4. The MAC algorithm
25
adopted for this is CSMA/CA (Carrier Sense Multiple Access with Collision Avoidance).
Other protocols proposed are VMESH MAC, ADHOC MAC, directional antenna based
MAC (DMAC), RMAC and Clustering based MAC (CMAC).
4.4.3 Network Layer
Vehicular densities in VANETs can be dense or sparse depending on the location.
The network layer protocols provide algorithms for wireless multi hop communication,
routing, congestion control and movement dissemination. Many applications of VANETs
rely on routing. The routing protocols used for ad hoc networks have been modified to
meet the needs of VANETs. Broadcasting is most commonly used for delivering caution
messages in safety related applications. The different approaches used for broadcasting in
VANETs are flooding, probabilistic broadcast and cluster based broadcast.
4.5 VANET Simulators
An ideal VANET simulator should support two different types of simulations:
simulating the mobility of the vehicles and simulating the wireless communication
between them. There are many existing high quality network and traffic simulators. A few
of the commonly used ones are listed in Table 4.1.
Most of the VANET simulators do not allow feedback to be communicated from the
network simulator to the traffic simulators. This is sufficient for infotainment applications
like checking emails in vehicles, media applications etc. In these cases pre-generated traces
can be used and dynamic mobility information is not required. However, when it comes to
safety related applications, two way communication between the traffic and network
simulator is essential. The traffic simulator has to feed dynamic information like the
vehicle position, speed, acceleration, direction etc to the network simulator. The VANET
application that runs on top of the network simulator uses this information along with the
surrounding vehicles’ information to give back a warning about possible collision or
congestion. This information is used to take appropriate decisions and is fed back to the
traffic simulator.
26
Table 4.1: Network and Traffic Simulators
Network Simulators Traffic Simulators
NS-2 SUMO
QualNet MOVE
GloMoSim VanetMobiSim
OPNET FreeSim
SWANS Paramics
GTNetS Corsim
SNS GrooveSim
CityMob
Netstream
STRAW
Unidirectional communication is straightforward and is achieved by combining a
traffic simulator with a network simulator. The trace from the traffic simulator is fed to the
network simulator. Some of the existing simulators that support unidirectional
communication are classified in Fig 4.2.
Fig 4.2: Types of simulators that support unidirectional communication
27
Bidirectional communication is little more complex as it is challenging to couple the
traffic and network simulator. The interface TraCl was developed to couple SUMO with
ns-2 or QualNet. The traffic simulator VanetMobiSim was extended as CanuMobiSim by
incorporating IDM. Similarly SWANS was extended as ASH (Application aware SWANS
with Highway mobility). GrooveSim and NCTUns are integrated simulators with tightly
coupled network and traffic simulators. The types are represented in Fig 4.3.
Fig 4.3: Types of simulators that support bidirectional communication
In this section a brief description is given about the simulators that are used by
researchers to test all the proposed VANET applications. Most of the simulators are open
source.
4.5.1 MOVE
MOVE (MObility model generator for VEhicular networks) is a tool developed to
generate realistic models for VANET simulations. It is a Java-based application built on
top of an open source micro-traffic simulator SUMO. MOVE provides GUI facility that
makes it easy for the user to generate simulation scenarios without writing scripts. The user
does not have to worry about learning the details and scripting of the simulator. It
generates mobility traces from the TIGER database. MOVE has a Map editor and a
Vehicular Movement editor. The Map editor creates maps for network scenario. The
vehicular movement editor generates movement patterns. MOVE generates a mobility
trace file as its output that can be used by network simulators like NS-2 or QualNet.
28
4.5.2 TraNS
Traffic and Network Simulator Environment (TraNS) is a simulator that integrates
both mobility generator SUMO and network simulator NS-2. It is an open source project
written in Java and C++. The main features of TraNS are 802.11p support, automated
generation of networks from TIGER, generation of mobility trace for ns-2 and ability to
simulate road events like accidents. A lighter version called TraNS Lite is developed for
mere mobility modeling without network simulations. The downside of TraNS is lack of
real time results. The output from NS-2 cannot be passed back to SUMO and hence it does
not produce results like real life.
4.5.3 VanetMobiSim
This is an extension of CanuMobiSim (Communication in Ad hoc Networks for
Ubiquitous Computing Mobility Simulator). CanuMobiSim cannot generate random
graphs and produce high levels of details in specific scenarios. VanetMobiSim produces
more realistic details at both macroscopic and microscopic levels. At macroscopic level,
VanetMobiSim supports multi-lane roads, separate directional flows, traffic lights and
human mobility dynamics. At microscopic level VanetMobiSim supports car-car and car-
infrastructure communication. It has a parser to extract road topologies from TIGER and
GDF (Geographical Data Files), which are passed on to network simulators like NS-2,
GloMoSim, QualNet and NET. The downside of this simulator is that it lacks feedback
mechanism. The traces from the network simulator cannot be fed back to VanetMobiSim.
4.5.4 NCTUns
NCTUns (National Chiao Tung University Network Simulator) is a simulator and
emulator written in C++. It can simulate various protocols used in both wired and wireless
networks. NCTUns included ITS support in its 4th version. It provides vehicular simulation
environment and includes both traffic and network simulator in a single module. It also has
a powerful feedback support. It provides a professional GUI that helps the users to draw
network topologies, configure protocol modules, specify the moving path of the nodes and
plot the network performance easily. The drawbacks of NCTUns are: a) It can support a
maximum of 4096 nodes in a single simulation. b) NCTUns also allows only a single
29
instance of TCP/IP version unlike other network simulators that support multiple TCP/IP
versions. c) It requires Fedora to be installed that limits its usage considerably.
4.5.5 GrooveNet
GrooveNet is a hybrid simulator that uses TIGER database and enables
communication between simulated vehicles and real vehicles. It incorporates modeling
with real street map based topology and mobility over a variety of communication models.
It provides multiple network interfaces and also supports simulations based on real
vehicles on-board computer (like GPS). GrooveNet supports 3 types of nodes in its
simulations - vehicular nodes, fixed roadside infrastructure nodes and mobile gateways that
is capable of V2V and V2I communication. GrooveNet supports hybrid simulations in
which the real vehicles can communicate with the simulated vehicles within its
transmission range.
4.5.6 MobiREAL
MobiREAL is a simulator that is able to simulate realistic mobility of humans and
vehicles. It is a rule based simulator that can be used in the cognitive modeling of human
behavior. It is used in MANET simulations by using the mobility support in the Georgia
Tech Network Simulator (GTNetS). A mixture of mobility models can be simulated
concurrently. For vehicular mobility it uses a traffic simulator called NETSTREAM
developed by TOYOTA.
The comparison of different VANET simulators is given in Table 4.2.
30
Table 4.2: Comparison of VANET Simulators
MOVE TraNS VanetMo
biSim
NCTUns Groove
Net
Mobi
REAL
Mobility generator
SUMO SUMO VanetMobiSim
NCTUns Groove Net
GTNetS
Network Simulator
NS-2, QualNet
NS-2 NS-2, GloMoSim, QualNet, NET
NCTUns Groove Net
Graphs TIGER database and user defined
TIGER database
TIGER database and GDF
Bitmap image
TIGER database
NET STREAM
Topology Any Any Any User defined
Any Any
Traffic lights at intersec-tions
Stoch Turns
Stoch Turns
Manually Defined
Automati-cally generated at intersec-tions
Manually Defined
Manually Defined
GUI support
Moderate Good Moderate Moderate Good Moderate
Mobility Models
Random Random and Manual Routes
Random Random and Manual Routes
Random Rule based
Ease of setup
Moderate Moderate Moderate Hard Moderate Easy
Ease of use
Hard Moderate Moderate Hard Hard Hard
31
4.6 Summary
With increasing use of ad hoc networks in different applications, the focus on
VANET has also gained a lot of attention. In this chapter, a detailed survey about
VANETs, its architecture, characteristics, prospective applications and the different
supporting simulators have been discussed. Although there are quite a few challenges for
which there are no known solutions till now, the fast growth and developments in this area
assures us that VANET will soon become part of the global wireless network. VANET not
only provides safety related applications but also improves the navigation system and
vehicular entertainment. VANET is indeed a promising approach for all future vehicular
applications.
As in all types of distributed systems, it is very important to have a synchronized
time even in VANETs. Depending on the type of application, the precision may vary. A
synchronized clock is essential in contexts like combining messages from various vehicles
to provide another useful information, collision alert etc. Maintaining synchronized clocks
among the nodes in VANET is crucial for the road safety application that is discussed in
this work. In the following chapters, the new proposed VANET based system and a clock
synchronization algorithm suitable for the system has been explained in detail.
32
CHAPTER 5
HYBRID VANET (H-VANET): A PRACTICAL APPROACH TO IMPROVE
ROAD SAFETY
5.1 Introduction
As seen VANET is an effective way to communicate between vehicles but there are
few issues as discussed in Chapter 3. The first issue is the connectivity problem in VANET
that is due to different velocities of the vehicles and the chances of no vehicles being
present in the network. Communication links can frequently get disconnected. The second
issue is the heavy cost involved in the practical implementation of VANETs. Though
companies have started working towards equipping vehicles for VANET communication,
it is not an easy task. For VANET communication to be continuous, there has to be well
equipped road side units that need to be installed. Deploying RSUs is an expensive project
that may take years to be achieved.
One of the solutions proposed for the above mentioned problem is discussed by
Fathima & Wahidabanu (2011). They have suggested the use of Delay Tolerant Networks
(DTN) that operates on the principle of store-carry and forward routing. The nodes store
the messages until the next node hop is available for forwarding. Anggoro et al. (2013),
have proposed combining probabilistic relay with AODV and AOMDV protocols. In a
situation if the vehicle, due to its dynamic nature, moves out of the range of its next hop
then obviously the transmission fails. Anggoro et al. (2013), have suggested that the
adjacent vehicles can probabilistically relay unsuccessful transmissions. However, the
trade-offs in both the above proposals is the message delivery delay. V2V message
communications may not be feasible when the vehicles on the road are sparse. The
messages may not reach the destination on time to prevent the accident, which is a very
33
crucial factor. Another method proposed is the use of Ariel remote sensing for highway
incident detection (Mostafa, Kahaki & Nordin 2011). However, this method has been only
80% successful and is also expensive to implement.
5.2 Proposed Hybrid Vehicular Ad Hoc Network
These major barriers for market penetration and connectivity issues can be overcome
by combining the low cost WSN with the VANETs (Khan et al. 2012). It provides a
complementary cost effective solution to overcome the constraints in conventional
VANETs. The new network which includes the roadside sensor nodes and the vehicular
nodes can be referred to as the Hybrid Vehicular Ad hoc Network (H-VANET). In H-
VANET, the VANET is integrated with the low cost Wireless Sensor Nodes (WSN) that
are deployed in between two access points. The Hybrid VANET is more efficient in
detecting the events ahead of time the vehicle reaches the spot, using the static roadside
sensors. Thus H-VANETs provide a much reliable and cheaper solution. The sensor node
can be deployed in curvy roads, tunnels and bridges easily. They can also be used to sense
physical data like temperature, humidity, light or motion. The events detected by the WSN
are very precise and reliable as it is within the road environment. Since it is stationary, the
information is persistent. The sensor nodes are battery powered and run for many months
with a pair of AA batteries. Due to its ease of deployment and low cost, it can easily cover
a wide geographic area. In the next section, the advantages of the Hybrid VANET are
discussed more elaborately.
5.3 Importance of a Hybrid VANET
In this section, some practical examples that can happen in everyday life have been
listed. In real life, accidents can happen due to any of the understated factors. In all of these
examples, it can be clearly seen that the presence of roadside sensors could make VANETs
more effective.
5.3.1. Road Factors
The roads can become slippery as a result of rain or snow. The vehicle that has
passed through the slippery route can send a message to the approaching vehicles. This
34
way the approaching vehicles can take precautionary steps or take an alternate route. A
curvy or steep road ahead can be cautioned to the following vehicles by the front vehicle.
The message reaches the other vehicles through the roadside sensors even if they are not in
the direct communication range of each other.
There can also be a wide range of unexpected road blocks like an accident in the road
or a fallen tree. The roadside sensors can prevent chain accidents by informing the
situation ahead of time helping the driver take timely decisions.
5.3.2 Environmental Factors
In some places fogs cover the roads affecting visibility. The visibility can be reduced
to 10-20 meters. Visibility is also reduced during night time and during rain. A pedestrian
walking in the highway may not be visible to the driver. If the roadside sensors can sense a
human in the road and pass the information to the approaching vehicles, pedestrian
accidents could be avoided.
5.3.3 Human Factors
In practical life one may come across many other emergency situations. Kids playing
in the backyard could accidently run into the roads. Similarly, old age or handicapped
persons trying to cross the road may not be able to see the approaching vehicles or make it
to the other side quickly. In such cases if a roadside sensor could detect their presence and
warn the vehicles beforehand, the drivers will have enough time to process the scenario
and apply the brakes.
The roadside sensor nodes also continuously detect the happenings on the road and
store it within the sensor network. This may be useful in post-accident investigations
especially in hit and run cases.
5.3.4 Animal Factors
It is quite common for animals to keep roaming on the roads that can cause
accidents. In 2000 out of 6.1 million collisions in the US, 247,000 crashes were deer-
vehicle collisions. In India, one can often find cows, buffalos and dogs wandering in the
35
streets. A sensor node in the H-VANET could immediately detect an animal roaming in the
road and pass the information to the approaching vehicles. The driver can slow down and
drive cautiously.
The advantages of a Hybrid VANET over a conventional VANET can be
summarized in Table 5.1.
Table 5.1: Advantages of H-VANET
Scenario VANETs H-VANETs
Reliability Sometimes there may not be any
vehicles on the road to detect a
particular event.
The roadside sensors will never miss
an event.
Deployment Poor network connectivity in
tunnels, remote roads, hills and
bridges.
Sensor nodes can be easily deployed
in any geographical locations.
Network
stability
Network can get disconnected
frequently when the vehicles are
sparse.
The sensor nodes help to keep the
network connected all the time.
Design
flexibility
The network exists only when
vehicles are present on the road.
The events on the road may go
unnoticed.
Here the design is very flexible. A
cloud of sensor nodes can be deployed
in places that are more prone to
dangerous events. Similarly, in safe
roads where there is no need of
constant monitoring, nodes neews not
be deployed.
Feasibility VANET is still under research and
requires high investment cost to
become a reality
Sensor node technology is less
expensive and well developed,
making H-VANET a feasible
alternative.
36
5.4 Model of H-VANET
The proposed hybrid VANET system is shown in Fig. 5.1.
Fig. 5.1: Model of the hybrid VANET
It is designed in the following way. The network is comprised of Vehicle nodes,
Road Side Units (RSUs) and Sensor nodes. Wireless communication is conducted between
these nodes. A device is fixed within every vehicle that can communicate with the devices
in the other vehicles on the road as well as with roadside stations. This device is developed
to collect, share, process and deliver real-time information about road conditions that could
affect safe driving. The sensor node stores all the information collected about any event
that happens in the road along with a time stamp. The roadside wireless sensor nodes are
divided into groups and each group is managed by a RSU. The RSU collects all sensor
information and transmits the aggregated data to the other RSUs. It also maintains the data
in its local database and transfers it to the vehicle nodes when a vehicle comes in its
communication range. Once a vehicle receives the data, it distributes the data to the other
vehicles in a geographical location by the Geocast Protocol. The message is communicated
to the drivers using some Driver Assistance System (DAS) (Singh, 2010). Maintaining the
security of the communication messages is also important and the same is beyond the
37
scope of this study. ANET security protocols have been discussed by Chen et al., (2013);
Pattnaik and Pattanayak, (2014).
The device (or on board unit) in the vehicle will have two interfaces: Embedded
WiFi card (IEEE 802.11) that is used for communication with the other vehicles and a
IEEE 802.15.4 (ZigBee) interface for communication with the RSUs. The sensor nodes
communicate with each other and with the vehicle nodes using the IEEE 802.15.4
(ZigBee) communication interface. Similarly the RSUs also have two communication
interfaces. RSUs and sensor nodes are deployed on both the sides of the road in a two way
highway. There are fewer RSUs that are deployed at fixed distances. The sensor nodes are
deployed in between two adjacent RSUs. The sensor nodes can sense and relay messages
to the RSU whereas the RSUs have the ability to also communicate with the vehicles. The
optimal placement of the RSUs and sensor nodes have been discussed by Rebai et al.,
(2012). IEEE 802.15.4 costs less, is more energy efficient and communicates over a small
geographical area. Hence it is used in the sensor nodes. On the other hand, IEEE 802.11
used in the vehicle node is more expensive but it can transfer data over medium distances
via multi hop communication.
5.5 Experimental Results
5.5.1 Field Tests
A set of experiments were conducted in a large parking lot to test how efficiently the
message is being delivered to all the nodes. The system that was implemented had 3
components-the Road Side Unit (RSU), normal sensor nodes and vehicular nodes. The
vehicle nodes are implemented by fixing a laptop in the vehicle with an attached telosb
mote.The regular sensors and the access points are implemented as Telosb motes with
mounted sensors. The sensors that is used here are long range WiEye Passive Infrared
(PIR) sensors. It has a wide detection cone of 90-100°, a detection range of 20-30 feet for
human presence and 50-150 feet detection range for vehicles depending on the size.
The WiEye has a visual light sensor and acoustic sensor that improves the detecting
ability of the PIR sensor. The WiEye sensor is directly plugged in to the TelosB motes. For
this experiment, 20 TelosB motes were deployed along one side of the road. The distance
38
between the motes was set as 40 m. Every10th mote was set as a RSU. The test lasted for
30 min. Vehicles were driven by volunteers at different velocities from one end to another.
The detailed system specifications are listed in Table 5.2.
Table 5.2: Prototype testing platform
Vehicle node Sensor node
Processor 64bits MIPS, 266 MHz 16 bits MCU, 8 MHz
Memory 512 MB 10 KB RAM
External memory 16MB flash 48 KB flash
Power supply 5.4-22 VDC @ 400mA 3 VDC @ 25 mA
Transceiver 250 kbit/s 2.4 GHz IEEE 802.15.4 -
chipcon wireless transceiver
Network interface IEEE 802.11p IEEE 802.15.4
Connectors UART, USB, MOST, VICS UART, SPI, I2C
Antenna External, Omni-directional Directional or omni-directional
Operating system Linux 2.6 TinyOS
Whenever a vehicle spots an obstacle it immediately informs the nearby RSU and the
vehicles in its range. For the roadside sensors, every object that enters its transmission
range will be detected as an event. This may include a vehicle itself. In order to avoid this
the following assumption were made. A normal vehicle on the road would travel at a
minimum speed of 15 km/h. In this case it will take about 7.2 seconds for the vehicle to
pass the transmission range of the RSU. So the sensors will wait for 7.2 seconds after it
detects an obstacle. If the obstacle still exists in the communication range after 7.2
seconds, an alert message is communicated to the neighboring RSU.
The test was conducted in a parking lot and the maximum speed of the test vehicles
was set as 25 km/hr for safety reasons. First a set of 5 volunteers were asked to drive
through the parking lot. An event was generated at a random time by throwing a dummy
39
doll in the parking lot. The time taken for the sensors to detect the event and communicate
it with the vehicles in the study area was recorded.
The results obtained show that the message gets delivered to all the vehicles within
few seconds, thus enabling the drivers to take decisions accordingly. The times taken for
the message to be communicated in different scenarios are noted. The values are tabulated
below in Table 5.3. When the number of volunteers (or vehicles) were increased, the
average message delivery time also increased. This may be accounted to increased number
of message delivery destinations. There is also more packet loss due to higher interference
and therefore more number of retransmissions.
Table 5.3: Average time taken for an alert message to reach all the nodes in a group
Number of
Vehicles
Velocity
(km/h)
Average message
delivery time (ms)
5 15 660
25 720
10 15 850
25 910
15 15 960
25 1030
20 15 1120
25 1250
5.5.2 Simulation Results
The proposed H-VANET system was simulated using the GrooveNet simulator. In
the simulation model, the vehicles were assumed to be running in a 3-lane highway. The
average flow of vehicles is 400 v/h/l under low traffic conditions and 1200 v/h/l under
heavy traffic conditions. In this simulation the incoming traffic flow was considered as
3000 vehicles per hour. The transmission range of the road side nodes is usually between
40
30-100 m. For experimental purpose, it was set as 80 m. The other parameters that were
fixed for the simulation are shown in Table 5.4.
Table 5.4: Simulation parameters
Highway length 1890020m
Number of sensor nodes 200
Distance between two sensors 100 m
Transmission range of sensor node 80 m
Transmission range of vehicle nodes 250 m
Average packet loss ratio 15%
Average vehicle speed 100 km/h
Synchronization Interval 600 ms
Time between two events 5-7 min
Simulation time 60 min
One of the important things which has to be verified is the massage transmission
between the vehicular nodes and the sensor nodes. It is crucial in all the Hybrid VANET
related applications. Suppose the average vehicle speed as 100 km/h. In this case, the
vehicles will be in the range of the gateways or RSUs for less than a second. A simulation
environment is created in which one car transmits a packet. The scenario was simulated to
compare the number of packets transmitted sucessfully with the theoretical upper bound of
the number of packets that can be received when the nodes are in communication range.
The results are shown in Fig. 5.2. For the experiments the effects of interference that could
affect the transmission when more number of cars are present on the road were ignored.
41
Fig 5.2: Packets transmitted between the vehicle nodes and RSU
The H-VANET was compared with a normal VANET. The systems were compared
considering some random low traffic scenarios. This is because in VANETs, low traffic
scenarios face frequent network disconnections. Some of the typical situations when there
are very few vehicles on the road include remote highways, tunnels, hilly roads and night
time. The message passed between the vehicles will be useful and meaningful only if the
message is delivered early enough for the driver to take an appropriate decision. The
interval between the earliest and the latest time that a message could be delivered such that
the driver is able to perceive and react to the message is referred to as the “Acceptable time
window”. The message delivered before or after this window becomes useless. The
number of messages delivered within this acceptable time window for H-VANET and the
conventional VANET under different traffic conditions was analyzed. It can be seen that
the conventional VANET with RBS fails to deliver the message when the number of
vehicles on the road reduces. The H-VANET however is consistent and obviously more
reliable as seen in Fig. 5.3.
42
Fig. 5.3: Number of messages delivered within the acceptable time window
5.6 Summary
In this work, a novel idea to make the proposed concept of VANETs more reliable is
discussed. All the foreseen applications of VANETs require the detection of real time
events as well as timely communication of the detected events to the vehicles. Due to the
unpredictable number of nodes and the fast changing topology of VANETs, it is sometimes
impossible to detect and communicate the events on time. The new H-VANET architecture
that is proposed integrates sensor nodes with the vehicular nodes to form a hybrid network.
The sensor networking technology is well developed, very cost effective and efficient in
detecting real time events in the roads. Integrating WSN with the VANET leverages the
overall system. The static sensors of the H-VANET that are deployed in the roadside,
assure that none of the events on the roads go undetected. It also assures constant
connectivity of the network irrespective of the number of vehicles present on the road. In
the next chapter, a clock synchronization algorithm to synchronize the clocks of the H-
VANET is discussed. This is one important step for all the H-VANET applications to work
properly.
43
CHAPTER 6
COMPARISON OF EXISTING CLOCK SYNCHRONIZATION PROTOCOLS
6.1 Computer Clocks
The system time of a computer represents a computer system's notion of the passing
of time. A computer clock measures the system time. Computer clock can also be defined
as an ensemble of hardware and software components used to provide an accurate, stable
and reliable time to the operating system. The computer clock is implemented as a simple
count of the number of ticks that have transpired since some arbitrary starting date, called
the epoch. For example, UNIX and POSIX-compliant systems encode system time as the
number of seconds elapsed since the start of the UNIX epoch at 1 January 1970 00:00:00
UT.
The computer clock is a programmable interval timer that counts the oscillations in a
quartz crystal that oscillates at a particular nominal frequency. The timer is associated with
a counter register and a constant register. For each oscillation in the quartz crystal, the
counter register is decremented by one. When the counter register becomes zero, an
interrupt is generated to the CPU. This interrupt is called a clock tick. The counter is then
loaded with the value from the constant register. The clock can be programmed to generate
an interrupt 60 times a minute by setting an appropriate value in the constant register. This
way the computer clock can run synchronized with the global clock.
In practice, the quartz crystal in each computer will run at a slightly different
frequency. The frequency of the oscillator may drift over time depending on the quality of
the oscillator, operating environment like temperature, gravity, electromagnetic
44
interference, aging etc. This leads to the clock values changing over time. This drifting of
the clock values is called the clock skew. The maximum drift rate of the oscillator is
defined by a permitted drift window around the nominal frequency of the oscillator. The
clock drift is represented in Fig 6.1.
Fig 6.1: Fast, Slow and Perfect clock with respect to UTC
6.1.1 Hardware Clock
Hardware clock is a clock that runs independently of any control program running in
the CPU and even when the machine is powered off. The hardware clock can be described
by the equation 3.1.
CH(t) = ft+C0 (6.1)
where (1-ρ) ≤ f ≤ (1+ ρ)
ρ → maximum absolute value for oscillator drift
C0 → counting register
6.1.2 Software Clock
This is the time kept by a clock inside the kernel and driven by the timer interrupt. It
gets incremented by one for every interrupt generated (i.e. a clock tick). It is meaningful
45
only when the operating system is running on the machine. The software clock can be
described by equation 3.2.
Cs(t) = fSt + C0 + A(t) (6.2)
where (1-ρ) ≤ fS≤ (1+ ρ) and A(t) → Adjustment factor
6.1.3 Logical Clock
In many applications it is sufficient to provide a logical ordering of events. It is not
necessary that the time be synchronized to the exact real-world time. This type of
synchronization is the easiest and cheapest to achieve. These clocks that provide relative
synchrony are called logical clocks.
6.1.4 Physical Clock
When the computer clocks are synchronized with each other and with an accurate
real time standard like universal coordinated time (UTC) they are referred to as physical
clocks.
6.2 Clock synchronization in distributed systems
A distributed computer system consists of multiple software components on multiple
computers, but which runs as a single system. A distributed system can consist of any
number of possible configurations, such as mainframes, personal computers, workstations,
minicomputers, and so on. The computers that are in a distributed system can be physically
closer and connected by a local network or they can be geographically separated and
connected by a wide area network (WAN). The components located on networked
computers communicate and coordinate their actions by passing messages. The goal of the
distributed system is to make the network of computers work as a single computer to
achieve a common goal, such as solving a large computational problem. Alternatively,
each computer may have its own user with individual needs and the purpose of the
distributed system is to coordinate the use of shared resources or provide communication
services to the users. The structure of the system (network topology, network latency,
number of computers) is not known in advance, the system may consist of different kinds
46
of computers and network links, and the system may change during the execution of a
distributed program. Each computer has only a limited, incomplete view of the system and
may know only one part of the input. These days distributed systems are found
everywhere. As distributed computing is becoming ubiquitous, centralized operating
systems are gradually giving way to distributed ones. The World Wide Web (WWW) is
one of the biggest examples of distributed system. An example of distributed system is
shown in Fig 6.2.
Fig 6.2: Example of a Distributed System
There are many challenges that arise in distributed applications using distributed
networks. Some of the associated issues include language support, tolerance to partial
system failure, maintaining a consistent view of the overall system, consistency of files and
data available, way of providing system services to user processes, naming of resources,
managing them etc. Resource and process management in distributed systems is very much
an open research subject until now. One of the stringent requirements in this management
47
is maintaining a synchronized time. The multiple processes running on different machines
should be in agreement with the ordering of events in a system.
In any independent system with a single clock, time synchronization is not an issue.
This is because whenever a process needs the time it issues a system call to the kernel.
Another process that requests the time from the kernel will get a higher value of time.
Hence the order of events is not ambiguous. However, when a computer has multiple
processors or when it is connected to any network, time synchronization between the
clocks becomes an important factor. Each processor has its own hardware clock that ticks
at its own rate. These are driven by oscillators, which differ in its quality and price. There
are several factors that influence the oscillator frequency that includes temperature
variations, gravity, vibration, electromagnetic interference and aging. Even if the clock
rates of all the clocks are initially synchronized when it started and the frequency at which
the crystal oscillator runs is fairly stable, it may not remain that way. The clocks
experience a certain variable clock drift of few microseconds per day. This accumulates
over time and the differences between the clocks become significant. It is also impossible
to guarantee that the crystals in different computers run at exactly the same frequency. This
causes each hardware clock to have its own variable clock drift.
In order to ensure that the nodes maintain the correct value of the current time, the
nodes exchange messages about their current state of hardware clocks and try to
synchronize. This process is referred to as clock synchronization. A proven truth is that it
is impossible to synchronize the clocks strictly with respect to both frequency and time.
This is because of a few unavoidable factors. First, when the nodes communicate, it may
not be perpetual. All messages arrive after an unpredictable and variable delay. Secondly,
a node cannot determine on exactly how much another node has progressed since the last
message that was sent. With the above limitations, the main objective of any clock
synchronization algorithm would be to minimize the clock skew and not to correct the
clock skew perfectly. This minimization of global clock skew should occur irrespective of
the distance between the nodes in the network.
48
6.3 Importance of Clock Synchronization in Computational Systems
For almost every application in distributed system, maintaining a synchronized clock
is crucial. Clock synchronization is useful in establishing the time of messages from the
senders and receivers, knowing the time of an event, time interval between two events, for
serializing concurrent access to shared objects, controlling joint activities or relative
ordering of events. It is also important for many applications in the internet as well. In this
section, some of the areas that emphasize the importance of synchronized clocks are briefly
described.
Distributed Databases
The order in which the process has to perform the update and commit the transactions
in a database is important to maintain the consistency and order of the database. This
cannot be done unless the events are properly ordered based on a perfectly synchronized
clock between all the cooperating processes. Also, the computer systems may not be up all
the time. The transfer of data between systems must be in such a way that each system
continues to work even if the one or more of the other systems are down. The data transfer
is synchronized when the systems are up.
Distributed Internet Applications
These include e-commerce, on-line banking operations, on-line reservation systems,
health care system, online education system, scientific research etc. There might exist a
precedence relationship among tasks in different processors. Suppose the task A in one
processor can start execution only after the completion of task B in another processor. The
scheduling algorithms that schedule the tasks will guarantee a correct precedence only if
there is good clock synchronization among processors. Monetary transactions, legal issues,
and other issues require a precise timestamp corresponding to the global clock. The client
and server cannot run on different timings, which could result in duplicate transactions or
fraud.
49
Multimedia Applications
Here clock synchronization is required to support Quality of Service (QoS). There are
three types of internet applications: asynchronous, synchronous and interactive
synchronous. Asynchronous applications are the ones that do not require simultaneous
transfer of two media streams e.g.: - web browsing. In synchronous applications, audio and
video has to be transferred simultaneously e.g.:- video streaming. Finally, the interactive
synchronous applications include real-time applications which are interactive e.g.: - video
conferencing. The latest developments in medical science like remote surgery also comes
under this. In the first and second type of applications, a synchronized time is not a must.
However, in interactive applications like networked gaming, teleoperation and applications
using “super media” a perfectly synchronized time is necessary to maintain the
performance.
Distributed Algorithms
Most of the algorithms and protocols that run on distributed systems also depend on
how well the time is synchronized. Let us consider a sensor network with few nodes (alias
processors) handling the sensor data acquisition. Now, algorithms that run on another set
of processors will process this sensor data. All these processors must always maintain a
fixed timing relationship with the algorithms. This can be maintained only if the processors
participating in the execution of the algorithm are time synchronized. Clock
synchronization is also important in areas like security systems, scheduled operations, fault
diagnosis, failure recovery etc.
6.4 Clock Synchronization Terminologies
Clock Stability: This is defined by how well the physical clock oscillator is able to
maintain a constant frequency.
Reference Clock: A dedicated clock used as a standard for a set of clocks’ state and
rate.
Clock Rate: The clock rate is the frequency of the crystal oscillator in a computer
clock.
Clock State: The state of the clock is the time shown by the computer clock.
50
Clock Accuracy: The accuracy of a clock is the correctness of the time maintained
by the clock with respect to a standard time. It refers to the deviation of the state of
any clock with respect to the state of the reference clock during a period of interest.
Clock Drift Rate: The drift rate of the clock indicates the time (in microseconds)
that the hardware clock drifts from the standard time per second. It is the deviation
of the frequency of the clock under consideration to the reference clock.
Clock Offset: The offset of two clocks is the time difference between the two
clocks.
Precision: It is the maximum deviation in the state (time) of any two clocks during
a period of interest.
Clock Skew: The frequency difference between two clocks is also referred to as
clock skew or time skew.
Clock Synchronization: This refers to adjusting two or more clocks in a network to
run at the same frequency as well as show the same time at a particular epoch.
However, achieving strict synchronization with respect to both frequency and time
is quite difficult.
6.5 Classification of Clock Synchronization Protocols
Different protocols are used to achieve clock synchronization to ensure that all the
processors have a common notion of time. In centralized systems, the centralized server
will dictate the system time. In distributed systems however, the clock synchronization is
much more complex. Before seeing the details of the different protocols, first a brief
description of the different classifications is necessary. The following are the different
classifications of clock synchronization protocols.
6.5.1 Internal vs. External Synchronization
While external synchronization focuses on synchronizing the clocks with an external
time source (real time), internal clock synchronization aims at having a common clock
among a set of nodes in a system. In external synchronization a standard time source such
as UTC that gives the real-world time is provided. The local clocks of all the nodes are
adjusted to this reference clock E.g. NTP. In internal synchronization, the main goal is to
51
minimize the difference between the clocks of all the nodes connected in the network.
These internally synchronized clocks may or may not be synchronized with a standard
external time. This relative clock between the nodes is sufficient for most of the
applications.
6.5.2 Master-Slave vs. Mutual Synchronization
The time is distributed among the nodes in the network using one of the two
methods: master-slave approach and mutual network approach. The master-slave approach
is an open-loop hierarchical approach while the mutual network approach is a closed-loop
distributed approach. In master-slave approach, the groups of clocks align their time to a
reference or master clock. The protocol assigns one node as the master node and all the
other nodes are considered as the slave nodes. The clock reading of the master node is
considered as the reference time. The master node must have high CPU resources. In the
mutual network approach (also called the Peer-Peer Synchronization) all the clocks try to
align their time to one another. Here any node can communicate with any node in the
network. It is a completely decentralized approach with no overhead associated with the
discovery of reference clock. It also improves reliability by eliminating the risk of master
node failure. e.g. RBS protocol, diffusion protocols etc.
6.5.3 Probabilistic vs. Deterministic Synchronization
The deterministic clock synchronization algorithm guarantee strict bounds on the
accuracy of the synchronization. An example of deterministic algorithm is hardware
algorithms in which there is a set of dedicated communication links for the clock to be
broadcasted in the system. This method is expensive but the communication delay is
deterministic and so is the accuracy of synchronization. The probabilistic algorithm on the
other hand provides a probabilistic guarantee on the maximum clock skew. A guarantee is
said to be probabilistic if it fails to hold sometimes. The failure probability however is
bounded or determined in probabilistic algorithms. The advantage of probabilistic
algorithms is that it reduces some extra message transfer and processing (Arvind 1994;
Deng & Zhang 2006). An example of probabilistic algorithm is network algorithm in
52
which the clock synchronization algorithm shares the communication links with the rest of
the system.
6.5.4 Clock Correction vs. Clock Assumption
Most of the clock synchronization protocols correct the local clock to achieve
synchronization. This correction is done instantaneously or gradually depending on the
situation. However, some algorithms do not adjust the clock e.g. RBS. Here a table is
maintained that relate the local clock of each node with the local clock of every other node
in the network. The local timestamps are compared and a common notion of the global
time is achieved. This way the energy spent in adjusting the clock is saved.
6.5.5 Pair wise vs. Global Synchronization
Pairwise synchronization tries to achieve clock synchronization between pairs of
nodes in a system whereas global clock synchronization protocol aims at achieving
network wide synchronization. The pairwise clock synchronization could be single hop
pairwise synchronization or multi hop synchronization. The single hop clock
synchronization discovers and adjusts the clocks between two neighbouring nodes that can
communicate with each other directly. The multi hop pairwise clock synchronization
protocols establish multihop paths in the network so that all the nodes can synchronize
their clocks to the source node. Further two approaches can be followed for single hop
pairwise clock synchronization that is given next.
6.5.6 Sender-Receiver vs. Receiver-Receiver Synchronization
In receiver-receiver synchronization, the reference node broadcasts a reference
packet and pairs of receivers try to identify the clock differences based on the reference
packet e.g. RBS. The receivers communicate with each other about the time at which it
receives the reference packet (instead of with the sender). In sender-receiver
synchronization, each sender communicates with the receiver to estimate the clock
difference. This is the traditional approach in which the sender periodically sends a
message with its local time to the receiver. The receiver then synchronizes its time with the
sender’s time.
53
6.5.7 Level based vs. Diffusion based
The level based algorithm builds a level hierarchy in the network and then
synchronizes the nodes level by level. The diffusion-based algorithm allows each node to
diffuse its clock to its neighbour nodes after it has synchronized to the source node. The
level based approach is suitable for static networks while the diffusion-based approach is
suitable for dynamic networks. The level based approach is more complex but has a high
precision. The diffusion-based approach on the other hand has wide coverage.
Table 6.1 gives a comparison of the different protocols based on the above
classification.
Table 6.1: Comparison and classification of the different protocols.
Protocol Internal
vs.
External
Master-Slave
vs. Mutual
Probabilistic
vs.
Deterministic
Sender-
Receiver vs.
Receiver-
Receiver
Clock
Correction
NTP Both Master Slave Deterministic Sender-
Receiver Yes
RBS Both Mutual Deterministic Receiver-
Receiver No
Coupled
Oscillator Internal N/A Deterministic N/A Yes
Level
Based Both Master Slave Deterministic
Sender-
Receiver No
Diffusion Internal Mutual Deterministic Receiver-
Receiver No
54
6.6 Synchronization Algorithms for Wired Networks
Clock Synchronization is becoming very important in LANs and WANs with the
increase in distributed applications like cloud computing, distributed services, size of
networks, infrastructure etc. The main factors to be considered while synchronizing clocks
in wired networks are the size of the network, dynamic nature of the networks, network
traffic and the convergence time. The algorithm should be able to tolerate node churn i.e.
the algorithm should be stable even in dynamic scenarios with node constantly joining and
leaving the system. There have been many algorithms proposed by different researchers
like Veitch, Babu & Pasztor (2004); Zhao et al. (2008); Scipioni (2009); Bo et al. (2010).
The very commonly used ones are discussed in detail below.
6.6.1 Christians Algorithm
In this method, clock synchronization is achieved by message exchanges between a
process P and a time server S. The time server is connected to the UTC (Universal
Coordinated Time). The algorithm is as follows: -
Process P requests time from the server S.
S responds by appending the time T from its own clock.
P then adjusts its time as T+ RTT/2
The round trip time (RTT) is non-deterministic due to the unpredictable network
traffic and message routing. Hence the algorithm is probabilistic and achieves
synchronization only when the RTT is short compared to the required accuracy.
6.6.2 Network Time Protocol (NTP)
The Network Time Protocol (NTP) (Mills 1994) has been a standard protocol for
external clock synchronization for packet switched networks. The NTP design involves a
hierarchical tree of clock sources. Each level of the hierarchy is termed as a stratum. The
primary server at the root (stratum 0) synchronizes with the UTC. The next level contains
the secondary servers, which act as a backup to the primary servers. The lowest stratum
contains the client nodes that need to be synchronized. The time is synchronized using the
55
offset delay estimation method. If T1, T2, T3 and T4 are the recent timestamps exchanged
between A and B. Let a = T1-T3 and b = T2-T4. The clock offset θ and round trip delay ∂
is given by: θ = a+b/2 and ∂ = a-b.
6.6.3 Coupled Oscillator Phenomenon
Synchronization in clock oscillators would be a phase lock between the two
oscillators. When the clock generates a timestamp based on the counter, clock
synchronization algorithm has to adjust the rate at which the counter increments. The
coupled oscillator phenomenon is applied to synchronize the clocks in a network. The
phenomenon states that enormous systems of oscillators are able to lock to a common
phase with a certain coupling strength despite the differences in the frequencies of the
individual oscillators. For distributed systems, discrete linear coupling equation is
considered.
(6.3)
where Ki= ΦiΔT
i= 1…N(t)
l є Ni
Φi→ measure of how much current clock rate should be influenced.
ΔT → time interval between successive interactions.
The clock difference (Cj(l ΔT) - Ci(l ΔT)) are estimated by means of request-reply
message pattern.
The algorithm at any process Pi does the following steps in every synchronization
round:
Select │Vi│ neighbors to synchronize.
Estimate the difference with every neighbouring clock and itself.
56
Sum the differences and multiply by clock correction factor.
Update the value of Ci with the new computed adjustment factor A (t) computed by
clock correction factor and the value of Ki.
Increment the value of Agei→ the number of successive synchronizations already
performed.
The algorithm was shown to synchronize clocks in very small to very large systems
with very small synchronization errors. The algorithm is also resilient to node churn and
network delays.
6.7 Clock Synchronization in Wireless Mobile Ad hoc Networks
Wireless Ad-hoc networks are decentralized networks that do not rely on any
preexisting infrastructure such as routers or access points. The network uses routing or
flooding to forward data from one node to another and is determined dynamically based on
network connectivity. Clock Synchronization in MANETs is very important to maintain
good quality of service. However, many classical synchronization algorithms used for
wired networks like NTP cannot be used for MANETs due to multihop environment, the
absence of network infrastructure and free movement of nodes. Some of the algorithms
that can be applied for ad hoc networks is discussed by Lai & Science 2004; Rentel &
Kunz 2005; Verma 2005; Hanzlik & Ademaj 2006; Zhou, Reid 2009; Choi & Liang 2012.
A few of the commonly used ones are explained in detail below.
6.7.1 Reference Broadcast Synchronization (RBS)
The reference broadcast synchronization (RBS) is often used in wireless networks in
which an initiator broadcasts a reference message (beacon). The arrival time of the beacon
is used as a reference point by the nodes to compare the clocks (Elson, Girod & Estrin
2002; Kuhn & Oshman 2009). The receivers within the listening distance of the initiator
will receive the same message at approximately the same time. The neighbouring nodes
exchange this arrival time and estimate the relative clock offsets. They are then able to
adjust their clocks. In multi hop networks, nodes are synchronized by the exchange of
messages via intermediate nodes acting as proxies. The non-deterministic transmission
57
delays that affect the accuracy of the protocol are: a. Send time: The time spent by the
sender to transmit the message. b. Access time: The time spent by the sender to get access
of the transmission channel. c. Propagation time: The time taken for the message to reach
the receiver. d. Receive Time: The time spent by the receiver to process the message. The
protocol can give highly accurate results if the receivers can record its local clock values as
soon as the message is received.
6.7.2 DTSR
The DTSR time synchronization procedure synchronizes time by maintaining clock
offset (CF) and clock drift (CDR) of one hop neighboring nodes. Every node in the
network broadcasts its local clock by piggybacking it on Hello messages. When a node
receives the Hello message, it first checks to see if the message is from a new neighbour. If
it is a new neighbour, then there is no CF and CDR stored in the time synchronization
table. The node calculates the CF relative to the sending node using the sent time and the
received time in the message. The clock rate is also broadcasted in the message. The CF
and CDR are stored in the time synchronization table maintained by each node. This
process is called the initial time synchronization. After this, the values are periodically
updated during every time synchronization update period (TSUP). This method is scalable,
precise and robust to varying traffic load conditions and achieves synchronization with a
bounded error.
6.7.3 CS-MNS
The Clock Sampling Mutual Network Synchronization (CS-MNS) makes use of a
virtual time and time stamp packets. The time and frequency adjustment is discrete and
multiplicative, thus reducing the bandwidth and energy requirements. The time process of
an accurate clock is modeled as:
T(t) = β.t + (t) + T(0) (6.4)
where T → time process of the clock
t → real time
58
β → skew of the clock with respect to the real time
→ random process that models time jittering and noisy effects
T(0) → initial time of the clock
For a network with N nodes, each with a clock that has a different skew and initial
time, for the ith node,
Ti(t) = βi .t + Ti(0) , i = {1,2,...,N} (6.5)
All the clocks in the network are synchronized such that || Ti ( t>tc ) - Tj( t >tc ) || ≤ Δ
T , for i ≠ j
Where tc → convergence time of the algorithm
Δ T → tolerable time error
It is necessary to adjust both the time offset and frequency offsets of the clocks to
achieve better performance. Hence the CS-MNS method uses an automatic controller to
adjust both time and frequency offsets.
The time process of the clock is multiplied by a correction factor Si (t). The
controlled time stamp obtained by multiplying the real time stamp by the correction factor
Si (t) is given by Ti (t).
Ti(nT) = Si (nT) βinT + Si (nT) Ti(0) , I = {1,2,...., N} (6.6)
This is updated in every node independently. The multiplicative factor Si (t) is
updated recursively by:
(6.7)
This method is compatible with IEEE 802.11 and IEEE 802.15.4 standards and
achieves synchronization with accuracy in the order of few microseconds. It is scalable and
is applicable for single-hop or multihop ad hoc networks.
59
6.8 Clock Synchronization in Wireless Sensor Networks
A wireless sensor network is a type of ad-hoc network that consists of spatially
distributed autonomous sensors. The sensors are used for a wide range of applications
ranging from target tracking, environment monitoring such as temperature, sound,
vibration, pressure, motion or pollutants, vehicle tracking, monitoring of critical
infrastructures, scientific exploration in dangerous environments etc. The data monitored
by the sensors is cooperatively passed through the network to a main location.
There are several challenges in designing a clock synchronization algorithm for
sensor networks. Sensor networks have spatial constraints. The nodes can communicate
directly only with their neighbours unlike the wired systems in which every node in the
system can communicate with every other node. The nodes in the sensor networks have
limited information about the other nodes. They do not have access to the clock values of
all the other members of the system. This is mainly because sensor nodes are resource
constrained and communicate with each other through short range wireless links. The
communication delay may also be significant. The sensor nodes also have very limited
computational capability. Synchronization algorithms for sensor networks have been
discussed in many works by Blum, Meier & Thiele (2004); Bo (2009); Chaudhari,
Serpedin & Shapoury (2007); Chaudhari, Serpedin & Qaraqe (2010); Chen et al. (2010);
Mirabella et al. (2008); PalChaudhari, Saha & Johnson (2004); Rahman & El—Khatib
(2010); Sun, Ning & Wang (2006); Sundararaman, Buy & Kshemkalyani (2005) . There
are four global clock synchronization methods for sensor networks that are commonly used
which are given below (Rus 2006).
6.8.1 All-node-based method
The all-node-based-synchronization is effective for smaller sensor networks. It
assumes that the clock cycle of each node, the message transmission time over each link
and the handling time on each node is the same. It also assumes that the clock tick time is
much longer than the packet transmission time. The algorithm first establishes a route that
passes through every node in the network. A synchronization message is passed along this
path. The initiating node records its local starting time and ending time of the message.
60
Each node records its local time and its order (number of hops the packet has travelled) in
the path. In the second phase, a clock correction message is sent along the same path that
informs each node of the starting and ending time for the initiating node and the total
number of hops. Each node then computes its clock adjustment and corrects its time. The
maximum error obtained after running the algorithm was ∆3.
6.8.2 Cluster-based method
The cluster-based-synchronization method (Mamun-Or-Rashid, Hong & Chi-Hyung
2005) is used when the whole network can be organized into clusters. The algorithm used
is the same as the previous method. The message path is established with all the cluster
heads and the synchronization message is passed. After the cluster heads are synchronized,
the nodes in each cluster can be synchronized with the head. The clustering can be done
based on the nodes within a transmission range or the nodes within a geographical area.
This method increases the flexibility of the algorithm and every node is not required to
participate in a single synchronization session.
6.8.3 Fully Localized Diffusion based method
The diffusion synchronization methods are fully distributed methods. There is no
need for any initiator. It achieves synchronization by spreading the local information to the
entire network. The algorithm chooses a global consensus value and all the nodes agree to
change the clock to that value. The consensus value can be obtained by taking an average
value of the clock readings. The neighbouring nodes first compare their clock values and
then change their clocks as needed. The clock comparison can be done using different
methods like RBS etc. The method can be synchronous or asynchronous. In synchronous
rate based algorithm, the neighbouring nodes exchange the clock values in a set order
whereas in asynchronous algorithm, the nodes can synchronize with its neighbours in any
order. Hence asynchronous algorithms accommodate node mobility and node failure.
The synchronous diffusion algorithm first compares the clock values of each sensor
with all its neighbours. Suppose ni is a sensor with clock value Ci at time t and nj is its
neighbor with clock value Cj. If ni and nj are within their transmission range and Ci>Cj, then
the value of Ci is increased and Cj is decreased. Thus for a diffusion value proportional to
61
Ci- Cj and diffusion rate rij> 0, nj's time is changed to Cj+ rij (Ci –Cj) and ni's time is
changed to Ci - Σallni 's neighbours njrij (Ci –Cj). In asynchronous diffusion algorithm, all the
nodes can perform the node operations in any order as long as every node is involved in
the operations with nonzero probability. Every node ni asks its neighbours the clock value.
It then computes the average value of the readings and sends out the average value to all its
neighbours. The average operation is assumed to be atomic. The convergence speeds of
these algorithms are slower compared to the algorithms with an initiator.
The asynchronous diffusion algorithm was simulated with randomly generated
network topologies and the following results were observed:
The error rate was seen to decrease exponentially with the increase in the number
of rounds. Here a round is defined as the time taken for a node to finish the average
operation once.
The convergence speed i.e. the number of rounds needed for the network to reach
some error threshold decreased with the increase in nodes. This is because the
number of neighbours for each node increased thereby speeding up diffusion.
The convergence was seen to increase with the increased transmission range since
each average operation covers more number of neighbours.
6.8.4 Fault Tolerant Diffusion based method
The fault tolerant algorithm is designed for synchronization in networks in the
presence of Byzantine faults. The method is mainly useful in hierarchical sensor networks.
Here some tamper-proof nodes are introduced in the network that destroys itself if it is
compromised. These nodes are called N nodes and are more expensive than the normal M
nodes. Each normal node is assumed to be part of a cluster whose head is a N node. It is
also assumed that the route of communication of the two N nodes has no two consecutive
M nodes and at most one third of the neighbors of N nodes can be compromised. The
algorithm has 2 steps: The first step is clock initialization in which the N nodes broadcast
its clock value to all its M neighbours. In the second step, each N node collects the clock
readings from all its neighbors, averages the value and broadcasts it again. This protocol
has four operations: the neighbour discovery, beacon broadcast, collection operation and
62
broadcast of the average value. In the neighbor discovery operation, each N node finds all
its neighbors shared with other N nodes. In the beacon broadcast operation, the N node
broadcasts a synchronization message to the neighbours. The neighbours then record its
current clock value and send it back to the N node in the collect operation. Finally the N
node averages the clock readings and broadcasts the value to the neighbouring nodes.
6.8.5 Secure Clock Synchronization
Generally, to address fault tolerance, redundant ways are used for clock
synchronization. Sometimes two schemes for synchronizing the clocks are assumed: level-
based method and diffusion-based method. This redundancy helps to tolerate partially
missing or false synchronization information provided by compromised neighbouring
nodes. It is assumed that every network has a trusted source node S that is synchronized to
an external clock. All the other nodes are synchronized to the source node. Each node i
maintains its own local clock Ci. The local clock of the trusted source node, Cs is the global
clock. Each node maintains a single hop pairwise clock difference for each neighbour node
j given by i,j = Cj– Ci. Every node also maintains a source clock difference i,s between its
local clock and the source node's clock.
i,s = i, j + j,s
Here the source clock difference is obtained through neighbour node j. Each node i
then estimates the global clock Cs by calculating the difference between its local clock Ci
and the source clock difference i,s.
Cs = Ci + i,s
In order to tolerate up to t malicious nodes, every node must compute at least 2t+1
source clock differences through different neighbouring nodes.
The level-based clock synchronization is mainly used in static networks where the
topology does not change frequently. It has two phases: level discovery phase and the
synchronization phase. In the level discovery phase the nodes are organized into a
hierarchy rooted at the source node S. In the synchronization phase, all the sensor nodes
obtain the source clock differences through their parent nodes and then help their children
63
nodes to synchronize their clocks. The diffusion-based algorithm allows the nodes to
obtain the clock differences through any neighbour nodes without requiring any level
hierarchy. The source node S initiates the synchronization process periodically and
unicasts the synchronization messages to its neighbour nodes. The neighbour nodes of S
update their source clock difference and then unicasts synchronization messages to their
neighbour nodes except S.
6.9 Comparison of Protocols
There are various methods proposed for the clock synchronization in different types
of networks. These different protocols can be compared based on different parameters,
some of which are explained below.
Precision
The precision in synchronization can be of two types: 1. Absolute Precision - This is
the maximum deviation of the clock with respect to the UTC. 2. Relative Precision - This
is defined as the maximum deviation of the logical clock values of all the nodes in the
network relative to each other. For most applications relative precision is sufficient. In
general, higher synchronization precision requires higher computational cost.
Reliability
Fault tolerance is an important factor in wireless sensor networks and other networks,
which are highly prone to failures. Fault tolerance is addressed by using redundant ways to
achieve synchronization. It is assumed that the compromised nodes will self-destroy
themselves. This way its impact on the synchronization protocol will be eliminated thereby
improving the protocol reliability.
Network Traffic
Network traffic is an important issue that affects the performance of the algorithms.
There are many messages that are exchanged between the nodes before the clocks are
synchronized. An increase in the traffic will lead to communication delay thereby affecting
the performance of the algorithm. One of the solutions proposed is piggybacking. The
64
acknowledgement messages are piggybacked with the messages that take the
synchronization data or the clock information is piggybacked with the neighbour
information (hello) messages.
Scalability
Nowadays networks are becoming larger and larger due to cost reduction. Some
applications may demand for the protocol to support a highly scalable environment. To
address this, the nodes are divided into clusters and there is a cluster head. The cluster
heads are first synchronized and then synchronization is performed for the remaining nodes
in each cluster. This is efficient for large networks as it reduces the communication
messages.
Complexity
The complexity of a synchronization protocol can be due to two factors: - 1.
Computational Complexity – The algorithms that require high memory usage during the
synchronization process. 2. Message Complexity – Some algorithms require lots of
message exchanges during the synchronization process.
Efficiency
Efficiency of the algorithm is measured with respect to the algorithm functioning even
with limited resources. The protocols used in sensor networks have to be energy efficient.
Convergence Time
Convergence time is the total time required to synchronize all the nodes in the network.
This is also related to the message complexity. It is also important for a good
synchronization algorithm to have low convergence time. Some algorithms divide the
nodes into clusters and synchronize each cluster separately to improve convergence time.
Node Mobility and Churn
The synchronization protocol should be able to accommodate node churn i.e. nodes
constantly leaving and joining the system. In mobile ad hoc networks like MANETs and
65
VANETs the nodes are constantly moving and do not have a fixed topology. There are
some algorithms implemented to support this.
Power Consumption
This is very important in sensor networks which run on limited resources. If the
algorithm is able to run only when the application demands it, the energy can be
conserved. Some of the algorithms support this feature.
Table 6.2 gives a comparison of five popularly used clock synchronization protocols.
It can be seen from the table that none of the existing algorithms fully support a highly
dynamic environment with limited connectivity as seen in the case of Vehicular ad hoc
networks (VANETs).
Table 6.2: Evaluation of the parameters supported by different protocols
Protocol No of
nodes
Prec-
ision
Conver
-gence
Time
Scala-
bility
Node
Mobility
Dynamic
Traffic
Support
Fault
Tolerance
NTP Many Un-
known
High Good No No No
RBS 2-20 1.85±1
.28µs
N/A Good Yes Limited
support
No
Coupled
Oscillator
{8, 16,
..64K}
10 µs Low Good No No No
Diffusion
based
200-
400
Un-
known
High N/A Yes No Yes
Level
based
200 Un-
known
High Good Yes No Yes
66
6.10 Summary
A detailed study about the different clock synchronization algorithms was done.
Some of the important details of the protocols and the methods used to address the
synchronization issues in the different algorithms have been analyzed. For wired networks,
NTP is the standard synchronization algorithm used worldwide. In wireless networks,
clock synchronization becomes much more important and crucial. Wireless networks have
a wide range of applications. These networks have many limitations with respect to their
energy availability, hardware size, bandwidth restriction and unstable connectivity. There
are also specific challenges related to advanced networks like VANETS, DTNs (Delay
Tolerant Networks), UANs (Ultrasound Acoustic Networks) etc.
A comparative study between the different protocols, with respect to the different
evaluation parameters has been tabulated. However, none of the algorithms can be applied
to achieve synchronization in highly dynamic traffic scenarios. The different factors to be
considered while designing a new algorithm have also been discussed. Based on the above
detailed study, a new algorithm has to be designed to suit VANET specific requirements.
In contrast to the traditional fixed and wireless networks, a VANET is both highly dynamic
and ad hoc. Due to constant changes in traffic, there is also an influence in the network
topology, radio propagation, connectivity of the wireless links and network partitions. All
of these factors pose challenges while synchronizing the time in VANETs. In the next
chapter, the characteristics of VANET, the different application areas and the necessity of
clock synchronization is explained in detail. This is the first step before developing a clock
synchronization algorithm specifically for vehicular networks.
67
CHAPTER 7
HYBRID CLOCK SYNCHRONIZATION (HCS) ALGORITHM FOR THE H-
VANET
7.1 Introduction
In all of the VANET based applications mentioned in the previous chapter, the
communication between vehicles is mandatory. The most important factor in this is that the
clock times of the different nodes have to be synchronized. Suppose a caution message is
sent by one vehicle at time 10:00 a.m. The message is delivered to the vehicle directly
following it. Suppose the time of message delivery in the destination vehicle is 9:59 a.m.
The destination vehicle will not be able to take any decision based on the message. The
caution message becomes meaningless because the time in both the vehicles is not
synchronized. A perfectly synchronized time is also necessary for taking decisions based
on messages sent by multiple vehicles and to deploy a secure communication on a wide
scale. In many similar situations, varying degree of clock precision is required based on the
application. In this work, the VANET is integrated with the road side sensor nodes. Proper
time maintenance is necessary for monitoring the physical world and intra-network
coordination
In Chapter 3, different clock synchronization algorithms have been discussed and
evaluated based on few parameters. There is no existing algorithm that fully supports a
highly dynamic environment like H-VANET that has frequent topology changes, network
partitions, disconnections in wireless links and interference in radio propagation. In the
next section, a clock synchronization algorithm specifically designed for the Hybrid
Vehicular ad hoc network (H-VANET) is discussed.
68
7.2 Related Work
Communication in the Hybrid VANET is dependent on 802.11, which provides some
built-in functions for time synchronization. Beacons that carry the relative time is
generated every 100 ms. In the worst case, when a node enters a new network partition,
the packets transmitted in the first 100 ms are not synchronized.
The vehicular nodes may be GPS equipped which is continuously synchronized.
However, the roadside sensor nodes in the Hybrid VANET are statically deployed and
cannot be connected to the GPS all the time due to its energy constraints. There have been
several synchronization algorithms proposed for sensor networks. However, those cannot
be applied to VANETs because of the dynamic environment.
Synchronization in ad hoc networks can be done in two ways: Centralized
synchronization and decentralized synchronization (Sourour and Nakagawa, 1999; Reid
2009). The centralized approach makes use of a GPS to synchronize with the global time.
This method is used in the TimeRemap protocol discussed by Ben-el-kezadri & Angeles
(2010). This is not well suited for the Hybrid VANET due to the energy constraints of the
roadside sensors. In the decentralized approach any node can initiate the synchronization
process. The different decentralized approaches that have been proposed are: a. Time
signal method: Here, every node transmits a timing signal continuously. The phase offset is
calculated by comparing with the received signals. b. Pulse based method: Here every node
periodically transmits a pulse. Each node corrects its own clock based on the incoming
pulse. e.g.:- Mutual synchronization, slot synchronization 3. Clock offset method: Here
every node transmits its clock time with its neighbours. The nodes calculate the clock
offset by comparing its local clock with the neighbouring nodes clocks. e.g.:- Reference
Broadcast Synchronization (RBS). This approach is the simplest to implement and is used
in the protocol proposed below.
7.3 Hybrid Clock Synchronization (HCS) Algorithm
The proposed new Hybrid Clock Synchronization (HCS) protocol for time
synchronization is a very robust algorithm which follows the decentralized approach and
does not require any knowledge about the network topology. It is scalable and is not
69
affected by a dynamic network which is a characteristic of VANET. It aims to synchronize
the RSU with the sensor nodes and the vehicle nodes within its coverage area. Each vehicle
has its own unique ID, a list of nodes that it is synchronized with and a list of neighbouring
nodes within its coverage area. The neighbors will include the vehicle nodes, RSUs and
sensor nodes. The vehicle nodes maintain its neighbor list by periodically broadcasting its
unique ID. The size of the synchronized members is called the synch scale.
The synchronization process takes place in the following steps:
Step 1: Watch for Initialization
In the Hybrid-VANET system, the RSUs or any vehicle node could initiate the
synchronization process. The RSUs can initiate the synchronization process at fixed
intervals. This interval of time is referred to as a Synchronization Interval (SI). In places
where there are no RSUs deployed or in cases when the RSU is down, any vehicle can
randomly initiate the synchronization process. In either case, the synchronization process
can be initiated only if there hasn’t been an initialization message in one full
synchronization interval i.e., no other node has initiated the process already. This will
prevent multiple synchronization attempts by different nodes.
Step 2: Synch initialization.
If a node has already initiated the synchronization then the other nodes cooperate and
pass on the required information. The initiator, either the RSU or any vehicle will now
broadcast a Collection Message (CM) to all the neighbors in its transmission range. The
collection message contains a collection request, all the neighbors IDs and a reply
sequence for all the neighbors to avoid reply collisions. As soon as the other nodes receive
the Collection εessage it will know that it doesn’t have to initiate the synchronization in
that cycle.
Step 3: Send reply message.
On receiving the collection message, the node will check the reply sequence and find
its time slot. It will then set a timer. When the time expires it will send a Reply Message
(RM) to the initiator. The reply message contains the synch scale, unique ID and the time
70
difference of the node. The time difference is the deviation of the nodes clock with respect
to a standard clock e.g., GMT. The format of the reply message is shown in Fig. 7.1.
Fig. 7.1: Reply message format
Step 4: Reply collection.
The initiator receives the Reply Messages from all the neighbors. The initiator waits
for a time period, Treply to get the reply message from all its neighbors.
replyT N 1 *R (7.1)
Where:
N The number of neighbors
R The duration of one reply message
Step 5: Selecting the synchronizer.
The initiator will compare the synch scales of all the neighbors with its own synch
scale. If any vehicle node has a higher scale than its own synch scale, then that becomes
the synchronizer. The initiator will then send a message to that node informing that it is the
new synchronizer. It will also send a list of all the vehicles’ IDs.On the other hand, if the
initiator itself is the node with the highest synch scale then it will continue and take up the
role as synchronizer.
Step 6: Synchronization.
The synchronizer will edit its synch scale by updating the list of synchronized group
members. It will then send a Clock Adjustment Message to all its group members. The
message consists of the synchronizers’ time difference and all receiver IDs.
71
Step 7: Clock adjustment.
Finally the individual nodes will adjust their own clock and also its synch scale.
7.4 Results
7.4.1 Theoretical Results
An important performance metric for a clock synchronization algorithm is the total
time taken for the synchronization process. This is because the transmission time from the
VANET to the WSN is very critical. The transmission range of the RSUs is between 30-80
m. Suppose the transmission range of the sensor is 30 m and the vehicle is assumed to
travel at an average speed of 70 km/h. Under these conditions the vehicle will be in the
transmission range of the RSUs for less than 1.5 seconds. All the communications for the
synchronization process have to take place within this time. The relative timing scale for
each operation of HCS is given in Fig. 7.2.
Fig. 7.2: Time sequence for one synchronization cycle
Where SI Synchronization Interval
CH Computer Handling
MT Message Transmission
N No of neighbors
R Reply time slot
The total time, Ttotal, taken for one synchronization cycle of the HCS is given by
Equation 6.2:
totalT SI 4*CH 3*MT N 1 *R (7.2)
72
Let us assume SI as 100 ms, CH between 5-30 ms, MT between 10-100 ms and R as
100 ms. Substituting N = 10 in the above equation, the total time taken for one
synchronization cycle will be, Ttotal between 1250 ms and 1620 ms. So if the number of
vehicles in the road is less, the probability of a vehicle to be in the range of the RSU
reduces. But with the above calculation it can be seen that for any number of vehicles
below 10, there is enough time for the vehicles in the transmission range of the RSU.
Suppose if the number of vehicles is increased as N = 25 to consider traffic jam condition.
In this case the total time taken Ttotal is between 2750 and 3120 ms.
During high traffic conditions, there is a higher probability that at least one vehicle in
a group to be synchronized is in the range of the RSU. This guarantees enough time for
communication of synchronization messages between the vehicle nodes and the sensor
nodes.
The inequality between the expected number of retransmissions and the packet loss
ratio is given by Equation 6.3:
in
ii 1
pr
1 p
(7.3)
Where: n Number of sensors
Pi Packet loss ratio
r Number of retransmissions
7.4.2 Simulations
A simulation environment was set up just like the one described in the previous
chapter. The HCS algorithm was evaluated under different scenarios and the results were
compared with that of thr RBS algorithm.
First, the performance of the HCS and RBS protocol is compared under the condition
when no new vehicles that enter have a higher time difference. The results are shown in
Fig 7.3. The average number of vehicles in the highway at any given time is between 40 to
73
60. With the RBS algorithm, initially the synchronization process is very fast. This is
because RBS requires fewer number of transmissions to get completely synchronized. The
HCS algorithm take more number of transmissions before it is fully synchronized.
Fig 7.3: Performance of HCS when no vehicle with time difference is added
Next, the two algorithms are compared under heavy traffic conditons with constant
movement of vehicles. The HCS algorithm is very stable even with a frequent entry of a
new vehicle with time difference. With the RBS algorithm, suppose if intially all the
vehicles are synchronized with a time difference of 3 μs. When a new vehicle arrives with
a higher time difference, say 15 μs, the synchronization process restarts and all the vehicles
get synchronized to this time. All the vehicles which were previously synchronized also
gets synchronized again to a time difference of 15 μs. Now, if another vehicle approaches
with a time difference of 18 μs. All of the vehicles will get synchronized to this vehicle
with a faster clock. This repeated synchronization of the vehicles over and over again
makes the algorithm very unstable. It is also a waste of time and power.
However, in the HCS protocol, if initially all the vehicles are synchronized with a
time difference of 3 μs. Any new vehicle that enters the highway will be synchronized to
74
this clock value. This makes the HCS protocol more efficient achieving a wider
synchronization area than RBS. The algorithm is also more stable as it does not have to
resynchronize even if a new vehicle enters with a faster clock.
Fig. 7.4: Stability of the algorithm when new vehicles with time difference enter the
group
The results (Fig 7.4) show that whenever new vehicles enter the group they are
synchronized to the existing group. The number of vehicles that have synchronized clock
steadily increases as new vehicles enter the group. In contrast the existing algorithms, e.g.:
RBS have to restart the synchronization process whenever a vehicle with a different clock
time enters the group.
7.5 Summary
In this part of the work, a Hybrid Clock Synchronization (HCS) algorithm to
synchronize the clocks of the sensor nodes, roadside access points and the vehicular nodes
is proposed. This is very important as the communicated messages are meaningful only if
the clocks are time synchronized. The HCS algorithm has been simulated using a very
75
reliable simulation platform and its performance has been tested under various conditions.
The results show that HCS is a very stable protocol under both high node mobility and
under low traffic conditions. It can be concluded that the H-VANET system together with
the HCS proves to be a very attractive, cost efficient and reliable networking infrastructure
for supporting all future vehicular applications.
76
CHAPTER 8
HYBRID VANET WITH PEDESTRIAN BODY UNIT TO IMPROVE SAFETY IN
BLACK SPOTS
8.1 Introduction
In the previous chapters, a hybrid VANET was discussed, which along with an
appropriate Hybrid Clock Synchronization (HCS) protocol proves to be an effective
solution to road safety for vehicles. However with the rapid explosion of population, along
with the motor vehicle usage, the road usage by pedestrians is also increasing. Road
pedestrian injuries and death rate is rising day by day. The research on traffic accidents
shows a clear concentration of crashes happening in Black Spots. Accident black spots
include steep slopes, a hidden junction, sharp corners in straight road, curvy roads
concealed warning signs or situations where the oncoming traffic is not visible. The
histories of traffic accident in the recent years have established a dependency between the
accident and the reaction time of the involved persons. This is mainly due to tensed
reactions of the drivers or pedestrians that comes spontaneously while facing a real time
potentially dangerous scenario. In this work, a solution proposed is a Hybrid VANET
based driver alert system. An alert given to the drivers ahead of time gives a better chance
for the drivers to react in a way as to avoid accidents. The system is designed by
integrating a pedestrian body unit along with the vehicular nodes in the H-VANET. The
signal sent by the pedestrian body unit is received by the vehicular nodes in the H-VANET
and are given as input to the alert system. This alert will notify the drivers about
surrounding pedestrians, which in turn gives him more reaction time.
Pedestrian fatalities have increased by 15% in the past decade whereas the other
causes of motor vehicle deaths have decreased by 3%. Pedestrian deaths have mainly
77
remained an urban phenomenon occurring frequently during early evenings, late nights,
foggy times, curves and other low visibility situations. Crashes typically have involved
people who have the habit of alcohol consumption, kids and people older than 70 years.
Some of the possible explanations for the increase in pedestrian deaths in the recent years
include the encouragement of regular walking by doctors for health and environmental
benefits, the effects of the economic recession coupled with the increase in fuel costs that
resulted in many walkers, favourable weather conditions, increase in the number of car
owners driving and the growth in the number of vulnerable population like senior citizens
and immigrants. There have been many measures taken by the Governments and other
organizations to create safe walking conditions like Walking Trails for pedestrians. These
were focussed on saving lives but nothing has come out very successful so far.
8.2 Background
Shinar (2012) has discussed about the safety and mobility of vulnerable pedestrians.
In order to reduce the accidents, a system named pre-crashing detection system using laser
and radar sensors was proposed by Bhumkar, Deotare & Babar (2012). They suggested
using sensing technologies to help detect the obstacles on the road. This operation
enhances the communication but the computation costs still remain high. There has been
some work done to study the pedestrian vehicle interaction behavior (Sun, Benekohal, &
Waller 2003; Markowski 2008; Agarwal 2011; Waizman, 2012; Waizman & Aviv 2015).
None of the proposed systems are fully developed and is yet to be tried out with H-
VANET. The low cost reliable solution proposed in the previous chapter integrates
wireless roadside sensors with the VANET to constantly detect events in the road and
communicate to the vehicles. The system uses a long-range passive infrared (PIR) sensor
which has a 90-100° wide detection cone. It can detect the presence of humans within a 20-
30 feet detection range and detect vehicles within 50-150 feet detection range depending
on the size. The sensor has a rectified acoustic envelope output. It also has a time constant
and low-pass filter that can be adjusted based on its applications. The sensor is directly
plugged in to the IEEE 802.15.4 / Zigbee compatible TelosB motes through the external
connectors. The problem with the above system is that it has a small probability that the
78
presence of a human may go unnoticed by the sensor. This could be a threat to careless
pedestrians.
8.3 Hybrid VANET with Pedestrian Body Unit
To overcome this problem, human body sensors are integrated with the Hybrid
VANET. This will ensure that the presence of any humans will never go undetected. An
integrated system is developed with the help of a Global Positioning System, which aids in
controlling the pedestrian accident rate to a great extent
8.3.1 Integrating Body Unit with H-VANET
The concept of wireless body area network (WBAN) was first proposed in 2001. It is
a Wireless Body Network formed by wireless sensors connected in the body (Hanson,
Powell & Barth 2009; Latre 2010; Cao et al. 2009; Jovanov 2005). These body sensors are
usually compact, flexible and consume less power. These sensors can generate various data
that can be applied in the various applications such as gaming, sport, lifestyle, remote
health monitoring, emergency situations, medical applications, defence, and consumer
electronics. These sensor data are transmitted through high power gateway and generate
control signals.
The proposed hybrid VANET system is represented in Fig. 8.1. It is an integrated
system of H-VANET and a body unit. This integrated system could improve human
detection by vehicles and thereby prevent accidents. The system can alert the motorists
under circumstances like: pedestrians walking in the middle of the road unexpectedly, kids
playing in the backyard accidently running into the road, old age or handicapped people
with limited vision or slow mobility trying to cross the road, pedestrians walking in with
reduced visibility areas or drunken pedestrians.
The system is designed in the following way. The VANET is comprised of Vehicle
nodes and Road Side Units (RSUs). A body unit attached with the pedestrian could be
designed to transmit the pedestrians’ exact location information to the nearby RSU (Road
Side Unit) or to the approaching vehicles via communication interface. This pedestrian
body unit communicates with the RSU or the vehicular node of the H-VANET and will
79
help alert motorists about the pedestrians before they approach the spot. Each vehicle node,
RSU and body sensor communicates with each other using the Wi-Fi (IEEE 802.11)
communication interface.
Fig: 8.1: System Model of a Hybrid VANET with PBU
With the development of low cost wearable computers like smart watches, smart
coats, glasses etc. in the recent past, this system is surely a feasible solution to reduce
pedestrian deaths due to road accidents. In this work, GPS (global positioning system) is
used as the wireless body unit. GPS is now becoming part of everyone’s life. They are very
commonly available as small portable devices and also in smart phones, tablets, laptops
etc. The information from the GPS is exchanged in the structured format within the
neighbouring vehicles and/or RSUs in its communication range. The above data is then
transferred within the H-VANET. The driver is cautioned by this alert message and can
take appropriate measures to prevent forthcoming accidents, such as slowing down,
changing the lane or applying brakes. This communication is demonstrated in Fig 8.2.
80
Fig 8.2: Alerting the vehicle
8.3.2 Laboratory Test
The hybrid VANET comprising of the vehicular node and the pedestrian body unit
was set up in a laboratory. Three laptops were used to emulate the vehicles on the road.
Three smart phones were used to emulate the pedestrians. A GPS application was installed
in the laptops and the mobile phones. With this GPS (Global Positioning System), the
position of the pedestrians and vehicles can be traced out. If the path of the both comes in
the close proximity of each other, a warning message is displayed in the mobile and laptop.
An android application was created for this “alert message” and installed in the mobile. On
seeing this caution message, the driver as well as the pedestrian will get cautious. The
sample result screenshots are shown in Fig 8.3, Fig 8.4 and Fig 8.5. Figure 8.3 shows the
position of the vehicles and pedestrians with respect to each other. The next Figure 8.4
traces the path of movement of the vehicle and the pedestrian. When the vehicle and
81
pedestrian come closer to each other a warning message is displayed which is shown in
Figure 8.5.
Fig 8.3: GPS position of pedestrian and vehicle
Fig 8.4: Tracing out the paths
82
Fig 8.5: Warning message
8.4 Black Spot Management
In this section, the hybrid VANET described above is used to manage and prevent
accidents at “Black Spots”. The statistical reports on traffic accidents show a regular and
continuous occurrence of crashes in black spots. Accident black spots may be caused due
to sharp corners in straight road, steep slopes, a hidden junction, concealed warning signs
or situations where the oncoming traffic is concealed. An example of a Black Spot is given
in Fig 8.6.
83
Fig 8.6: An example of a black spot
A pedestrian is trying to cross an uncontrolled intersection. A large truck parked in
the side of the highway conceals the pedestrian’s view of the oncoming vehicles. Suppose
a vehicle approaches as shown in the figure, it creates a “black spot”. There are high
chances of the occurrence of an accident. Similar case occurs during severe snowfall when
a pile of snow is deposited in the sidewalk and obscures the pedestrians’ view. It is the
same case for the drivers who come and cannot see the oncoming vehicle due to the
obstacle. Black Spots are also common in mountainous roads that are very curvy. The
situation is considered for a single lane road where the vehicles cross the intersection one
at a time. The approaching vehicle should have a complete set of information about the
pedestrian present in the vicinity. Using the hybrid VANET, a system can be implemented
that can detect pedestrians approaching the crosswalk and alert the drivers. A framework is
proposed as shown in Fig 8.7.
84
Fig 8.7: Schematic representation of the vehicle communication system
As can be seen, the vehicle controller gathers it own GPS data and the GPS data of
the pedestrian. Having this information, the vehicle controller notifies the vehicle through
the speaker. The alerted driver can take an appropriate decision to avoid a collision.
Research shows that the pedestrian reaction time is considered as 0.28±0.07 sec and the
driver reaction time is considered as 0.75 sec. The brain reaction time includes the mental
processing time and the body movement time. The mental processing time is the time to
perceive a scenario and the body movement time is the time taken to react to the visual
stimulus. In this case, the body movement time for the pedestrian is the time taken to stop
or move backwards. For the driver, it is the time taken to lift his foot from the accelerator
and apply the brakes. When the brain perceives the situation, it has to take a decision on
how to react. The decision taken plays a major role in avoiding or causing a collision as
well as the severity of collision. This strategic decision-making can be represented as a
game in normal-form game theory, which is explained next.
85
8.4.1 The Game Structure
The players here can be defined as player 1, the vehicle and player 2 as the
pedestrian. In situations where there are two vehicles at the intersection, the players will be
vehicle A and Vehicle B. The actions of the players can be given as below. The vehicle has
the option to neglect the alert and continue to go with the same speed in which case it is
not cooperating. On the other hand it can reduce its speed and gradually come to a stop at
the crosswalk giving way for the pedestrian. This is the case of cooperation. The pedestrian
also has few options. He can continue walking without bothering about the vehicle i.e. not
cooperate or wait for the vehicle to stop and then accept the offer to cross the road
(cooperating). The pedestrian will also not cooperate if he rejects the vehicles offer and
does not cross the road. The strategies adopted here are, Player 1 has to reduce its speed
and come to a stop to offer the crosswalk for player 2. If it does not slow down, it will not
cooperate. Similarly, player 2 has to accept the right of way and cooperate in order to cross
the road without any risk. If he does not accept the offer, he is not cooperating. The
information shared between the two players is the exact time and the location on a real
time basis, the actions of the other player, the strategies and the payoff functions. Payoff
can be defined by the time delay by the actions that directly relates to the occurrence of
collision.
Here, the vehicle and the pedestrian become the players. The prevention of collision
depends on the strategy of these two players. In order for the game to be successful, the
vehicle controller unit must gather the GPS information from the other vehicles and
pedestrians in close vicinity. Being connected to the H-VANET, this information can be
shared easily. It is notified to the driver via the speaker and subsequently the game play is
started. Similarly, the pedestrian unit controller (in this case the handheld mobile phone)
does the same.
8.4.2 Payoff Calculations
In all traffic related scenarios, time is the most crucial factor that has to be taken into
account. When both the vehicle and the pedestrian adopt a cooperative strategy, it
86
consumes some time but it ultimately prevents deadly collisions. Suppose a vehicle
travelling at 50 km/h gets an alert about a pedestrian presence before 100 m. Then for the
player 1, the actual time to cross the intersection from the time it receives the alert will be
about 8 sec. Similarly the average walking speed of a pedestrian is 5 km/h. If the road
width of the single lane road is taken as 3.75 m, then the actual time to cross the
intersection after receiving the alert will be 10 sec. The following calculations show the
time delay for both the players under both actions.
For player 1 (The vehicle)
Actual time, T = 8 sec
Action1: Cooperation (Slows and stops)
Stopping time = Actual time – (Deceleration time + Stopping time)
= 8-(4+2) = 2 sec
Action 2: Non cooperation (Does not stop)
Running time = Actual time – Actual time
= 8-8 = 0 sec
For player 2 (The pedestrian)
Actual time, T = 10 sec
Action1: Cooperation (Stops and accepts the right of way)
Stopping time = Actual time – (Stopping time + Walking time)
= 10-(2+3) = 5 sec
Action 2: Non cooperation (Does not stop)
Running time = Actual time – Actual time
= 10-10= 0 sec
87
The payoff table is shown in Table 8.1. It can be seen from the table that a delay is
caused in the situation where both the vehicle and the pedestrian cooperates. However, this
delay can eliminate the possibility of a collision and hence is the most ideal solution. In
situations where either one of the player cooperates, the delay is not as much. In these
cases the chance of accident is reduced considerably but not completely eliminated.
Finally, in situation where both the players do not cooperate, there is no extra time
consumed at all but it is the worst decision that will lead to a fatal crash.
Table 8.1: Payoff table for the game
Player 1/Player 2 Pedestrian stops and
accepts the path
Pedestrian does not
stop
Vehicle stops 2,5 2,0
Vehicle does not stop 0,5 0,0
In Fig 8.8, the time delay for the vehicle is shown. The straight line shows the
trajectory of the vehicle when it does not stop. The curved line shows the time delay
caused due to the vehicle decelerating and coming to a stop.
Fig 8.8: Time delay for the vehicle
88
8.5 Simulations
The hybrid VANET system with pedestrian unit was simulated using the GrooveNet
simulator. In the simulation model, the vehicles were assumed to be running in a 3-lane
highway. The incoming traffic flow was considered as 3000 vehicles per hour i.e
approximately 1000 v/h/l. The distance between the crosswalk and the obstacle in the setup
was assumed to be 0.75 m. The other parameters that were fixed for the simulation are
shown in Table 8.2.
Table 8.2: Simulation parameters
Length of the Highway 1890020m
Number of sensor nodes 200
Distance between two sensors 80 m
Transmission range of sensor node 100 m
Transmission range of vehicle nodes 250 m
Average packet loss ratio 15%
Average speed of vehicles 100 km/h
Simulation time 60 min
The proposed system was designed and the possibility of accident was analysed for
two different situations. As seen in the previous section, it is important that all the nodes
are time synchronized in order for the system to work properly. The Hybrid Clock
Synchronization was used for this network too. The first time no alert is given and the
driver react normally to the situation. It can be seen that the pedestrians are in danger
when the oncoming vehicle speed is as low as 11 km/h. On the other hand, when the
drivers are alerted using the hybrid VANET-body sensor system, the situation improves
drastically. The pedestrians are in danger only when the vehicle speed is above 36 km/h.
The results are shown in Fig 8.9. As it can seen from the figure, let us take a particular case
when the pedestrian speed is 5 km/h. It can be seen that with no alert system, the accident
89
takes place even when the car is coming at a speed of 14 km/h. However, when an alert is
given the accident occurrence can be prevented for speeds up to 40 km/h.
Fig 8.9: Reduced accident possibility with the H-VANET alert system
This shows that the proposed system of sensing the situation and giving an alert
ahead of time to the drivers can surely help the drivers to react in a smoother way. An alert
given before few seconds can improve the scenario drastically by changing the way the
driver reacts to the same situation.
In the next experiment, the game strategy followed by the driver and pedestrian is
analysed. As seen in the game theory, the decisions taken by the driver and the pedestrian
determine the occurrence of crash. There are four cases that can happen depending on
whether the driver or the pedestrian cooperate. In the first two cases, the driver does not
cooperate which is represented as DN. When the pedestrian cooperates, it is given as PC
and for non-cooperation it is given as PN. It can be seen that when the driver cooperates
(DC), as in the third and fourth cases, accidents can be avoided to a greater extent. The
simulation results of this evaluation are shown in the Table 8.3 below.
90
Table 8.3: Decision making and the occurrence of crash
Vehicle Speed
(km/h)
DNPN DNPC DCPN DCPC
40 No Crash No Crash No Crash No Crash
45 No Crash No Crash No Crash No Crash
50 No Crash No Crash No Crash No Crash
55 Crash No Crash No Crash No Crash
60 Crash No Crash No Crash No Crash
65 Crash Crash Close Stop No Crash
70 Crash Crash Crash No Crash
75 Crash Crash Crash Crash
8.6 Summary
In this part of the work, a pedestrian body unit with GPS was integrated with the
hybrid vehicular ad hoc network (H-VANET. The idea is to make sure that the presence of
pedestrians on the road is known to the drivers to alert him. The proposed system was
simulated considering a specific situation that occurs in accident black spots, where the
vision of the drivers is obstructed. Results show that the system will drastically bring down
the overall accident rate of the people on the road. This new network formed by the
pedestrian body unit and vehicular nodes gives an extra level of safety to the careless
pedestrians who might accidentally get hit by drivers. With the development of low cost
wearable computers in the recent past, this system is surely a feasible solution to reduce
pedestrian deaths due to road accidents in the near future. The situation of Black Spots is a
perfect one where the proposed system will reduce the possibility of accidents. The
approach used by the driver and pedestrian when alerted is analysed using the game theory.
The cooperation strategy by the players could avoid a crash completely.
91
CHAPTER 9
VANET BASED VEHICLE CONTROL SYSTEM TO AVOID HUMAN
ERROR
9.1 Introduction
In the previous chapters, a detailed description of the Hybrid VANET that includes
the pedestrian body unit was given. The system helps by giving timely alerts about the
presence of humans on the road. The distracted drivers get alerted ahead of time and are
assumed to take the right decision. What if the driver is so much diverted that he overlooks
the alert? Or what if he panics on receiving the alert and is not able to think properly?
Since the decision making still depends on a human (driver), there are still chances of
human error to occur. Accidents could be avoided to some extent but depends on the
quickness of the driver in taking the right decision. On facing an exactly same situation one
driver may be able to stop the vehicle without crashing while another inexperienced or
stressful driver may not able to do so. Every human brain acts at different speed and
depends on several factors which is elaborated below. So it will be better if this act of
decision making is automated. The presence of pedestrian and its locations is passed as a
message to the vehicle. A controller in the vehicular unit compares its own location with
the pedestrians’ location along with other sensor inputs like speed, distance, etc. It then
checks to see the possibility of an accident and sends a control signal to the advanced
emergency braking system when needed. The machine intervention makes the whole
system more reliable.
9.2 Vehicle Stopping Distance and Time
When a driver sees some obstacle on the road like a pedestrian in the crosswalk,
another stopped vehicle, a wandering animal or any road debris, his immediate response
92
would be to apply the brakes. The stopping of the vehicle is usually split into two phases:
The Driver Reaction time and The Mechanical Response Time
9.2.1 Driver Reaction Time
The human reaction time can be defined as a measure of the time taken to respond to
any stimulus. There are many factors that could affect the reaction time, some of which are
age, gender, personal characteristics, distraction, sickness, tiredness, intoxication level etc.
In general, a human reaction time is composed of the brain reaction time and the body
movement time. The reaction time of drivers (referred as Driver reaction time) has been
widely studied due to the wide possible consequences. This is because a slower reaction
time can lead to fatal accidents. While on the road, the driver comes across multiple stimuli
and he has a choice to make from multiple possible responses. The reaction time of the
drivers can be split up into the following components.
(a) Mental Processing Time
This is the time taken by the driver to perceive an incident on the road and to come
up with a response. First the sensory organs sense the input which may be auditory ones or
visual ones, the nerve impulses then pass from the receptor to the brain where the brain
processes and recognizes with the help of stored information from memory, interprets the
situation, decides on a response to be taken and finally programs the body movement
mentally. For example, if a driver comes across a pedestrian, he first sees it and recognizes
the situation. If his brain realizes that his driving speed and distance from the pedestrian
could lead to a collision, the brain will select an appropriate response like turning the
steering or applying the brakes. The average mental processing time ranges from 0.5
seconds to 2 seconds. Some studies also show brain reaction times as high as 7 seconds.
The standard time adopted in United States is 1.5 seconds and in Australia it is 1.5
seconds. This time accommodates more than 90% of the different types of drivers who
face simple and moderate level situations. In really complex and unexpected situations, this
time is definitely higher.
93
(b) Body Movement Time
Once the brain decides on an appropriate manoeuver, the body muscles react
accordingly. For the above scenario, the response would be to apply the brakes. This time
taken for the driver’s body to carry out the selected action is called body movement time.
In the above case, the driver has to lift his foot off the accelerator and depress the brake
pedal. There is also a small time required for the brakes to engage. This time is highly
variable again and depends on the braking style, urgency, vehicle condition etc. It varies
from 0.3 seconds to 1 second.
9.2.2 Mechanical Response Time
Finally, there is a time taken by the mechanical device to respond and perform the
manoeuver after the driver has acted. Even after the driver has applied the brakes, the
vehicle takes some time complete the manoeuver i.e. to come to a complete stop. This time
is referred to as the mechanical response time or the manoeuver time. This time varies for
every vehicle and depends on numerous factors like the size of the vehicle, its type,
gravity, road surface, weather conditions, average deceleration of the car, the condition of
the car, its braking capacity, weight of the car, condition of the tyres, the incline of the
road, the available traction etc. A standard deceleration rate adopted is 3.4 m/s2 (11.2 ft/
s2).
9.2.3 Stopping Distance
With an idea about the vehicle stopping time, the stopping distance can be calculated.
Stopping distance is the distance travelled by the vehicle during the two phases of vehicle
stopping. It can be said as the approximate distance before which a driver needs to see an
obstacle in order for him to stop the vehicle without colliding.
In general, given the acceleration and the velocity of a vehicle, the time taken for a
driver to stop the vehicle is given by
Ts = V/a. (9.1)
Where V Velocity of the vehicle (mph)
94
a deceleration rate (ft/s2)
The distance travelled is calculated by s = ½ a t2. The stopping distance of a vehicle
can thus be given by
D = ½ aTs2 = V2/2a = ½ *V*Ts. (9.2)
According to the results of many tests and experiments conducted by the American
Association of State Highway and Transportation Officials (AASHTO), the driver thinking
distance is calculated by a standard equation given by
DT = 1.47 Vt (9.3)
Where t brake reaction time (seconds)
As a result of many studies conducted in many places, 2.5 seconds has been adopted
as a standard for driver reaction time. This includes 1.5 seconds for his mental processing
and 1 second for the body movement.
Similarly, the braking distance is calculated by
DB = 1.075 V2/ a (9.4)
As mentioned above, the deceleration rate of 11.2 ft/ s2 is used by many as a standard
to calculate the stopping distance. However, in more than 90% of the cases, the drivers are
able to decelerate at a higher rate.
Finally, the stopping distance is the sum of the driver’s thinking distance and the
braking distance given by
SD = DT + DB (9.5)
= 1.47 Vt + 1.075 V2/ a
Suppose a person is driving a car or a light truck at a speed of 60 km/h. If the road
surface is dry and the vehicle is well maintained with tires in good condition, an average
alert driver can safely decelerate at the rate of about 3.4 m/s2. The friction can be assumed
to be 0.75. In this case, it will take almost 5 s for the vehicle to come to a complete halt. If
95
1.5 s delay for driver’s thinking time is included, then the driver will be able to stop the
vehicle only after 6.5 s. By then the vehicle would have travelled a distance of more than
45 m. However, the braking skills of the driver improves with experience. The time taken
and the distance taken to stop the vehicle can be drastically reduced with more practise.
Practically, all drivers are not so experienced. The road and vehicle condition also are not
always the best.
The above example clearly illustrates the severity of the situation. The average
vehicle stopping distance assuming a well maintained car with an alert driver on a dry road
for different car speeds has been tabulated in Table 9.1. It shows the stopping distance for
drivers with 1.5 seconds reaction time and 2.5 seconds reaction time.
Table 9.1: Vehicle stopping distances
Vehicle
Speed
Reaction
Distance
1.5 sec
Reaction
Distance
2.5 sec
Braking
Distance
Total
Stopping
Distance
1.5 sec
Total
Stopping
Distance
2.5 sec
Crashing
Speed
Km/h m m m m m Km/h
40 17 28 9 26 37 No crash
45 19 31 12 31 43 No crash
50 21 35 15 36 49 No crash
55 23 38 18 41 56 No crash
60 25 42 21 46 63 2
65 27 45 25 52 70 32
70 29 49 29 58 78 46
75 31 52 33 64 85 57
80 33 56 38 71 93 66
85 35 59 42 78 101 73
90 37 62 48 85 110 79
9.3 Vehicle Control System
As seen in the previous section, a few milliseconds could be the margin of safety
that could save many lives. A minute delay in the perception time of drivers could lead to
fatal results. Today, intelligent automated systems are becoming part of everyday life. If
the most complicated task of perception-reaction while driving could be controlled with
96
technology, it could reduce the chances of accident to a great extent. Continuous research
has been going on to automate cars since 1920. After the first autonomous car was
developed in 1980’s, major automobile companies like εercedes-Benz, Nissan, Ford,
General Motors, Toyota, Renault, Hyundai, Volvo etc. have started developing prototypes
of autonomous vehicles. Even research organizations and universities are working in this
area. The self-driving car developed by Google is one such autonomous car which is under
experimentation (Driverless Car n.d., Autonomous Car n.d.).
The autonomous vehicles are classified into 5 levels where level 0 vehicles have no
automated system. Level 1 and 2 vehicles have limited automation and the drivers can take
control any time. In level 3 and 4 vehicles, full automation can be given to the vehicle in
limited safe environments like highways, good weather etc. The level 5 vehicles are fully
automated and require no human intervention at all. The anticipated advantages of having
an automated vehicle control system are obvious and are listed below:
Smooth and safe traffic flow in highways.
Better driving ability for the physically challenged.
Tireless and stress free driving experience even during long drives.
Accident avoidance.
The proposed framework of the system is shown in Fig 9.1. It can be seen in the
diagram that the system has a vehicular unit and a pedestrian unit that communicates via
the VANET. Each unit has a controller that takes care of the major tasks. The vehicle
controller is implemented using an Arduino microcontroller that acts as the brain behind
the whole system and controls all the actions.
97
Fig 9.1: Framework of the vehicle control system using H-VANET
In the pedestrian side, the communication controller frequently broadcasts its
position to all the oncoming vehicles that are in its communication range via the wireless
transmitter. Concurrently, the wireless receiver on the vehicular side keeps listening for the
broadcast messages. When a message is received, the vehicle control system obtains its
current location from its GPS. It then compares both the GPS coordinates with the
pedestrian GPS coordinates that comes in the message. Along with the vehicle speed,
acceleration, steering angle, pedestrians’ movement direction and distance information, the
controller would check if an accident is bound to happen. If there is an accident possibility
in the black spot, then the vehicle and pedestrian are said to be in the Bad Set boundary.
Within the Bad Set, there is a Trap Set where the occurrence of collision cannot be
prevented by any human initiated controls. The Bad Set boundary is pre-programmed into
the microcontroller and depends on factors like vehicle speed, acceleration, steering angle,
size, road conditions, pedestrians’ movement direction and distance. The microcontroller
then sends a control signal through its logic circuits, to the advanced emergency braking
system (AEBS). The controller also simultaneously sends an alert message to the
communication controller in the pedestrian unit. When an alert is received by the receiver,
the communication controller sends an audio alert to the pedestrian via a speaker.
98
AEBS is also known as the autonomous emergency braking system and is a road
vehicle safety system that is beginning to get mandatory for all new vehicles, starting with
heavy vehicles, as per the United Nations Economic Commission for Europe’s (UNECE)
announcement. AEBS automatically applies the emergency brakes in situations of
impending collision. In most cases these may prevent crashes fully while in other
unavoidable cases the speed of the vehicle is reduced after applying the brakes. The
resulting low speed crash may prevent fatality and property damage caused by the
collision. Some of these systems activate independent of any driver input while others
provide braking assistance to the driver. In the proposed system, the controller is connected
with the AEBS, which in turn is connected to the vehicle braking unit.
The block diagram of the vehicle control system which was implemented using an
Arduino microcontroller board is shown below in Fig. 9.2. The other microcontroller that
could be used is the PIC 8051. The microcontroller constantly waits to receive the
broadcast messages from the pedestrians. The vehicle sensors measure the speed,
acceleration and steering angle of the vehicle. The Arduino board gets this as input from
the vehicle and its position information from its own GPS, which is the NEO6MV2 GPS
module. It then calculates to find if the pedestrian is within the Bad Set boundry as
mentioned above. If yes, a control signal is immediately sent to the AEBS which in turn
activates the braking unit.
99
Fig 9.2: Block diagram of the vehicle control system
The working of the vehicle control system is represented in the form of a flowchart
in Fig 9.3.
100
Fig 9.3: Process flow diagram of the vehicle control system
9.4 Mathematical Evaluation
In this section, a mathematical formalism to describe the vehicle pedestrian scenario
is described. The main idea is that controlling the velocity and displacement of the vehicle
101
can prevent the vehicle-pedestrian crash. Also, an alert message is given to the pedestrian
instructing him to stop or continue cautiously via the controller. All H-VANET equipped
vehicles will have vehicle sensors for measuring the state of the vehicle which gives the
position, velocity, acceleration, brake torque and pedal position of the vehicle. With the
development of Intelligent Transportation System (ITS), it can be assumed that all vehicles
will have the ability to automatically actuate the brake and throttle. In most of the
automatically controlled vehicles, the driver has the ability to override the automatic
control by manually pushing the brakes or pedal. In such cases the vehicle control system
cannot guarantee collision avoidance.
In highways, the vehicles usually follow the predefined route and thus the trajectory
of the vehicle can be defined. The collision scenario considered in this work is given in Fig
9.4. The vehicle emerges at the cross road where a pedestrian is coming from the other
direction. Due to limited visibility there is a possibility of collision. The location of the
potential crash near the intersection is marked in the figure.
Fig 9.4: Collision scenario
To model the dynamic state of a single vehicle, a state space X × V is used, where
102
X the set of all possible longitudinal displacements
V the set of all possible longitudinal velocities
The state of the vehicle is denoted by the vector (x, v) X×V, where
x the longitudinal displacement of the vehicle center of mass
v the tangential velocity of the vehicle center of mass
The control space U: = [ Lu , Hu ] is used to represent the scalar combination of all
possible pedal and brake torque inputs, where
Hu the maximum throttle torque command
Lu the maximum brake torque command
The set of control signals for the vehicle is denoted as S (U). So u S (U) is called a
control signal, whereU R , the set of real numbers. S (U) is the set of all functions
: 0f R U such that f (t) = 0 for all t < 0. Similarly, the control space for the
pedestrian is given by U: = [ Lu , Hu ] which represents the combination of all possible brain
speed or stop commands. Here is the brain stop command and is the brain speed
command.
The flow of the system is defined as the evolution of the vehicle state as time
proceeds. It is given by the function ϕ: R+ ×X×V× S (U) X×V, where R+ non-
negative real numbers. This flow is generated by a controlled dynamic system f: X×V×U
→ X×V. For the initial state (x,v) X×V, the control signal u ϵ S(U) and time t ≥ 0, the
state of the vehicle at time t is given as ϕ (t, (x,v), u) X×V. Here ϕ1 (t, (x, v), u) X is
the longitudinal displacement and ϕ2 (t, (x, v), u) V is the longitudinal velocity. The flow
of the vehicle over an interval of time can be given by ϕ ([0, t], (x, v), u) X×V. This will
represent the trajectory of the vehicle and is shown in Fig 9.5.
103
Fig 9.5: A sample trajectory of a single vehicle
Next, the modeling formalism is developed for the vehicle-pedestrian system using
parallel composition. Here the state vector can be defined for the entire system as
(x,v) := (x1, x2, v1, v2) X×V := X1 × X2 × V1 × V2 , (9.6)
where
(x1, v1) X1 × V1 is the state of the vehicle
(x2, v2) X2 × V2 is the state of the pedestrian
The control signal for the vehicle-pedestrian system can be given as u: = ( )
where is the control signal of the vehicle and
is the brain control signal of the pedestrian. The flow of the entire system is
given by . The vector of displacements is represented by
and the vector of velocities is represented by The
flow of the vehicle is represented by and the flow of the
pedestrian is represented by .
104
The problem of collision avoidance can be formalized as avoiding a bad set of
vehicle-pedestrian states, B X where B represents the set of displacements in the path
that leads to a collision. The bad set can be defined with path geometry as
B: = x X x1 L1, H1 and x2 L2, H2 (9.7)
Here, L1 and H1 are the lower and upper displacements along the vehicle path.
Similarly, L2, H2 are the lower and upper displacements along the path of the pedestrian.
The interval notation ] L, H [ to represent an open interval i.e. x if and only if L <
x < H. If the vehicle and the pedestrian are within these bounds, then there will be a
collision. This bad set is shown in Fig 9.6.
Fig 9.6: Bad Set within the lower and upper bounds of displacements
The solution to this problem will be to design a controller to control the flow of the
system and prevent it from entering the bad set B. For this, first a Trap Set is constructed
given by T (v) X. The velocity of the vehicle-pedestrian system parameterizes this Trap
Set. The trap set corresponds to the set of vehicle and pedestrian displacements x X such
that with the current velocities of the vehicle and the pedestrian v V, there cannot be a
105
control signal u S (U) that will be able to prevent the accident eventually. This can
mathematically be expressed as
T (v): = (9.8)
This trap set is computed online and a control is applied if either the vehicle or the
pedestrian enters the trap set boundary. Suppose the vehicle enters the boundary of the trap
set before the pedestrian, the control signal u* = is applied. This implies that the
vehicle accelerates and the pedestrian stops. Thus collision can be prevented. If on the
other hand, the pedestrian enters the boundary of the trap set, then the control signal u* =
is applied to avoid the crash i.e. the vehicle brakes and the pedestrian speeds. As a
result, the control prevents the flow from entering the bad set.
9.5 Simulations
The above scenario was experimentally tested using the simulator. In this
experiment again the Hybrid Clock Synchronization algorithm was used to synchronize the
clocks. The signals given to the driver and pedestrian were evaluated separately. First, a
case where the pedestrian attempts to cross the road in a non-regulated intersection is
considered. The pedestrian is distracted and does not look around before crossing or there
is an obstacle blocking his vision. In the vehicle control system described above, the
control to the pedestrian comes in the form of an audio input. The pedestrian has to react to
the audio input and salvage the situation. The average reaction time according to different
studies is set as 0.33 seconds for a visual input while for an audio input it is reduced to
0.28 seconds (Shelton & Kumar 2010). There is always a ± difference between the fast
reactors and the slow reactors. So for this work, the estimates used were 0.33+5= 0.38
seconds and 0.33-5=0.28 as the reaction time of fast and slow pedestrian to visual stimuli.
Similarly, 0.28+5= 0.33 and 0.28-5=0.23 are the reaction times for the fast and slow
pedestrians to audio input. The scenario was evaluated for different vehicle speeds. Four
types of situations were considered i.e. the pedestrian reacting slowly to his visual stimuli,
pedestrian reacting fast to visual stimuli, pedestrian reacting slowly to audio alert and the
pedestrians reacting fast to audio input. With the vehicle control system, the audio alert
106
comes to the pedestrian at least 2 seconds before the vehicle reaches the crosswalk. This is
because the nearest roadside sensor will be at a maximum distance of 80 m. In order to
simulate these, the time for the pedestrian to stop after receiving the signal was taken as
0.38 sec, 0.28 sec, 0.33 sec and 0.23 sec respectively. For this a single lane highway is
assumed. The other parameters considered is as given below in Table 9.2.
Table 9.2: Simulation parameters
Length of the Highway 1890020m
Number of sensor nodes 200
Distance between two sensors 80 m
Transmission range of sensor node 100 m
Transmission range of vehicle nodes 250 m
Average packet loss ratio 15%
Average speed of vehicles 100 km/h
Simulation time 60 min
The results of simulation are shown in Table 9.3. The maximal safe speeds for the
car in the different cases of fast and slow reacting pedestrians with/without the alert is
given. It can be seen that giving an auditory alert to the pedestrian drastically reduces the
possibility of accident as opposed to not giving an alert.
107
Table 9.3: Maximum safe car speeds for different reaction times
Safe Car Speed (km/h)
Pedestrian
Speed
(km/h)
Slow Pedestrian
without alert
Fast Pedestrian
without alert
Slow Pedestrian
with alert
Fast Pedestrian
with alert
2 No crash No crash No crash No crash
3 23 No crash 25 No crash
4 19 40 22 No crash
5 14 27 17 42
6 11 20 13 36
9.6 Experiments
A set of real time experiments were conducted in a highway using a test vehicle.
Table 9.4 lists the detailed system specification of the vehicle control system that was used
for the experiments.
Table 9.4. Specifications
For experimental purpose, the pedestrian was assumed to be in a static position.
Vehicles were driven by volunteers at different velocities in a stretch of highway. The
distance considered was 100 m. A dummy doll was placed at one end which represented
the pedestrian trying to cross the road. The roads were in dry condition. The vehicles were
driven at different speeds above 60 km/h. This is because it is above this speed that there is
Processor 64bits MIPS, 266 MHz
Memory 16KB
External Memory 32KB flash
Microcontroller Arduino
Power Supply Breadboard5V
Transceiver 250 kbit/s 2.4 GHz IEEE 802.15.4
Network Interface ESP 8266 WiFi Module
Operating System µC/OS-II
108
a probability of accident even when the driver cooperates, as seen in the simulation results
in section 6. The vehicle receives a message about the obstacle when it is at a distance of
100 m from the dummy doll.
The distance covered for the vehicle to come to a full stop was measured. With
manual control, after the driver sees the pedestrian, he applies the brakes manually. With
the vehicle control system, the vehicle receives information about the pedestrian even
before the driver can actually see. The brakes are automatically applied. Since the vehicle
control system is connected with the AEBS, it applies the maximum power boost. Thus the
stopping distance of the car is reduced as compared to the driver operated cars. The values
are compared with the stopping distance of the manually operated cars. Table 9.5 shows
the results observed.
Table 9.5 : Accident possibility with different vehicle speeds
Manual Control Vehicle Control System
VehicleSpeed
(km/h) Stopping
Time
(s)
Stopping
Distance
(m)
Stopping
Time
(s)
Stopping
distance
(m)
60 6.3 Exactly touches 4.8 -51
65 6.7 6 5.2 -43
70 7.1 12 5.6 -35
75 7.5 18 6 -26
80 7.9 24 6.4 -17
85 8.3 30 6.9 7
90 8.7 36 7.2 Exactly touches
95 9.0 41 7.7 15
100 9.5 47 8 27
The first column shows that with manual control at 60 km/h the vehicle comes to a
stop exactly touching the obstacle. In all the other cases, it hits the obstacle and comes to a
stop beyond the obstacle. In real time, accident could occur anytime above the speed of 60
km/h whereas according to the simulation results, accidents could be avoided for even
higher speeds when the driver cooperates. It is purely dependent on the different responses
of the driver and pedestrian. The distance beyond the obstacle at which the vehicle comes
to a stop is given in the table. Using the vehicle control system, the accident occurs for
109
speeds above 90 km/h. For all speeds below that, accidents can be surely prevented. The
values shown in the second column shows the distance at which the vehicle comes to a full
stop before the obstacle.
The stopping distance of the vehicle, both under manual control and the vehicle
control system are shown in Fig. 9.7 and Fig. 9.8. The point where the vehicle comes to a
stop without hitting the pedestrian under manual control and the autonomous control is
shown in the figures. With the vehicle control system, for all vehicle speeds below 90
km/h, the vehicle is able to come to a stop before hitting the pedestrian. In all other cases,
where the speed of the vehicle is higher, the vehicle stops beyond the pedestrian after
hitting the pedestrian.
Fig 9.7: Stopping distance of the vehicle with manual control
110
Fig 9.8: Stopping distance with Vehicle Control System
When hitting the pedestrian, the speed of the vehicle after deceleration determines
the impact of the accident. The time taken for the pedestrian to react after receiving the
audio alert could not be tested due to safety reasons.
9.7 Summary
In this work, the H-VANET was integrated with the advanced emergency braking
system using a vehicle control system. The automated vehicle control system aims at
reducing a major cause of road crashes i.e. human error. The drivers travelling on the road
may be in different situations – some inexperienced, tired, sick, physically challenged,
aged which will increase the time taken by them to perceive the scenario on the road and
react to it. Even a small delay in reaction could be the cause of fatal accidents. The
proposed vehicle control system replaces this human thinking time by automatically
sensing the road scenario and activating the braking system. The system checks the
pedestrian location, vehicle location, speed and distance to calculate if an accident in
bound to occur. If so a control signal is sent to the advanced emergency braking system.
The proposed system was proven mathematically as well as by simulation. Some field tests
were also done. The results were quite impressive and prove that the collisions can be
prevented using this vehicle control system to speeds of up to 95 km/h. The real time
111
situation is expected to be even better. There are many automobile companies coming up
with vehicles with automated controls. So this development paired up with the fast
growing wireless vehicular communication makes this a very attractive solution to reduce
traffic deaths due to human errors.
112
CHAPTER 10
CONCLUSION AND FUTURE WORK
10.1 Summary of Research Work
In this work, a very novel idea of a time synchronized Hybrid VANET system has
been proposed to improve road safety. The system on the whole prevents accidents caused
due to lack of time for the driver to react. Although there are a few challenges in VANET
for which there are no solutions today, the automobile industry is making rapid
developments in this area which assures us that VANET will soon become part of the
global wireless network.
Chapter 1 starts with an introduction to clock synchronization, the history of
synchronization issues and its importance in distributed systems. The main research
objective and its motivations have been discussed in detail. The main objective of this
work is to improve road safety using a time synchronized Vehicular Ad hoc Network
(VANET). The basic necessity of the road safety application is that the clock times of the
participating nodes have to be time synchronized. Although clock synchronization issues
have been part of every type of distributed systems, one of its application areas, which
have not been much studied upon, is VANETs.
In chapter 2, a literature survey has been done about work done in the area of
synchronizing clocks in distributed networks.
In chapter 3, a comparison of five different clock synchronization algorithms have
been done based on different parameters. The main idea was to check if any of the existing
algorithms could be modified to the vehicular environment. NTP is a standard protocol
used today but it cannot be used in ad hoc networks. Ad hoc networks have many
limitations, hardware size, bandwidth restriction and unstable connectivity. There are also
113
specific challenges related to advanced networks like VANETS, DTNs, and UANs. The
evaluation showed that none of the existing algorithms fully supports a VANET
environment. The comparison provided a base before developing a new algorithm for
VANETs.
Chapter 4 makes a detailed study of VANETs, its properties, application areas and
the simulators used for VANET applications. In the recent years, ad hoc networks are
penetrating into a wide range of applications. One of the popular types of ad hoc network
for all vehicular applications is the VANET. The high deployment cost of VANET and the
connectivity issues are the main limitations of VANET that is addressed in the next
chapter.
VANET is the most promising solution available for improving the road safety.
However, it is quite expensive to be deployed in the near future. A solution to address this
issue known as the Hybrid VANET has been discussed in chapter 5. This newly proposed
H-VANET architecture integrates sensor nodes with the vehicular nodes to form a hybrid
network. Due to the unpredictable number of nodes and the fast changing topology of
VANETs, it is sometimes impossible to detect and communicate the events on time. The
static roadside sensor nodes does this job very efficiently. It also proves to be very cost
effective solution as the sensor nodes are much cheaper than the Road Side Unit.
Integrating WSN with the VANET leverages the overall system by assuring that no events
on the road go undetected. It also keeps the VANET constantly connected which is not
always the case with a conventional VANET.
Chapter 6 gets back to addressing the clock synchronization issue in this Hybrid
VANET. It is the basic requirement for the road safety application to run successfully. A
Hybrid Clock Synchronization (HCS) algorithm was proposed to synchronize the clocks in
the H-VANET. The HCS algorithm has been simulated using a very reliable simulation
platform and its performance has been tested under various conditions. The results show
that HCS is a very stable protocol under both high node mobility and under low traffic
conditions.
114
In chapter 7, the hybrid VANET is improved by integrating pedestrian body sensors
with the vehicular ad hoc network (VANET). This new network formed by the body
sensors and vehicular nodes gives an extra level of safety to the careless pedestrians who
might accidentally get hit by drivers. With the development of low cost wearable
computers in the recent past, this system is surely a feasible solution to reduce pedestrian
deaths due to road accidents in the near future. The situation of Black Spots is a perfect one
where the proposed system will reduce the possibility of accidents. The game theoretic
approach was used to analyse the situation. The cooperation strategy by the players could
avoid a crash. The system was also simulated and the results prove that the chance of
occurrence of accidents is drastically reduced.
The final part of the work is given in chapter 8, in which the alerts from the roadside
sensors and the body sensors are given as inputs to the vehicle control system. In spite of
giving timely alerts about road situations, the system has a possibility of failing if the
driver doesn’t respond in a timely manner. There may be some extreme situations where
the time window between the alert message and the drivers’ reaction may not be sufficient
to prevent an accident. Studies show that even a small delay in reaction time could lead to
fatal accidents. The proposed vehicle control system replaces this human reaction time by
automatically sensing the road scenario and activating the braking system and
simultaneously sending a signal to the pedestrian. The proposed system was proven using
mathematical evaluation as well as simulation. The results show that the road accidents can
be eliminated.
10.2 Conclusion
This is a generation where every activity is being controlled by machines or
automated systems. One of the common everyday task that is still dependent on human
control is driving. Even though driving has become part of human survival, driving still
depends completely on human action and reaction. Unfortunately, the same human action
and reaction is the reason for road accidents. The drivers travelling on the road may be in
different situations - inexperienced, tired, sick, physically challenged, aged - which will
increase the time taken by them to perceive the scenario on the road and react to it. The
main objective of this work was to improve road safety using VANETs in order to reduce
115
this major preventable cause of human death around the globe. A road crash can be an
outcome of carelessness for just a fraction of a second by even one driver using the road.
However capable and experienced the drivers may be, there are many situations that go
beyond their reacting capabilities. There needs to be a non-human support to handle such
extreme situations. In this work, VANET has been shown to provide the best driver
assistance support to avoid road fatalities. VANETs have been under study over the last
few years. The safety on the road depends on how the driver reacts on perceiving a road
scenario. Any human could err when he is distracted or when he comes across an
unexpected situation. To aid the drivers, a Hybrid VANET system was developed that is
able to constantly monitor the road and alert the drivers of any happenings in the road that
requires immediate attention. The Hybrid VANET is cheaper to deploy than the
conventional VANETs and at the same time gives a more reliable, efficient solution. Once
the Hybrid VANET detects an obstacle on the road and alerts the vehicle, appropriate
action can be taken. Another major requirement for this application to work properly is that
the clock time of the nodes has to be synchronized with respect to each other. To achieve
this the Hybrid Clock Synchronization algorithm was developed. This HCS algorithm is
able to synchronize all the clocks under the dynamic vehicular environment. The
pedestrians in the road are given additional safety level by making their presence known to
the vehicles through body sensors. A body sensor could be anything like a GPS connected
phone or smart watch that could communicate with the roadside sensors about the location
of the pedestrian. Once the signal reaches the vehicle about an obstacle on the road, it can
be given as an audio alert to the driver or as an input to the vehicle control system. To
avoid the chances of human error completely, it is appropriate to connect to the automated
vehicle control system instead of depending on the driver’s reaction. The system was
simulated and mathematically proven to work effectively. The situation analysed was that
of Black Spot that often occurs in urban cities. The simulation results prove that the
collisions can be avoided to a great extent when the appropriate control system is given to
the vehicle controller. In the field tests that was done, accidents could be avoided for
vehicle speeds up to 95 km/h. In reality, even better results could be achieved.
With the fast growing wireless vehicular communication technology, the Hybrid
VANET system is surely feasible. It can be concluded without any doubts that the H-
116
VANET system together with the HCS is a very attractive, cost efficient and reliable
networking infrastructure for improving road safety considerably. The improvement in
road safety further reduces property damages, monetary and human loss. This is sure to
lead our society to a better tomorrow with a promising and peaceful, accident free, stress
free road driving experience.
10.3 Future Work
However the idea has been simulated and experimented with the limited assumptions
and implementations, the real traffic scenario has many more factors to be considered. The
traffic regulations and patterns vary from country to country and also from region to
region. Considering all the factors one by one is part of the future work.
The urban scenario is not taken into consideration in this work. In cities, the roads
are broader with multiple lanes and there is usually a heavy flow of traffic. This also
results in higher influence from other vehicles, bandwidth issues, more reflection,
interference and security issues. The effectiveness of vehicular communication when many
vehicles are on the road is yet to be evaluated. A real time field test with multiple test
vehicles could not be done due to lack of testing location.
Also the test vehicle used for the experiments was a level 1 vehicle with only braking
actuation. Steering is still under the drivers control but it doesn’t create a big issue because
vehicles usually follow a predefined path in the highway. So it can be assumed that even if
the steering angle changes, the vehicle stays within the lane. In future, testing can be done
including the steering actuation too. These factors have to be considered in future to further
improve the system.
117
REFERENCES
• Agarwal, N 2011, Estimation of pedestrian safety at intersections using simulation, Dissertation, University of Kentucky Doctoral Dissertations, viewed 20 March 2013
• Anggoro, R, Kitasuka, T, Nakamura, R & Aritsugi, M 2013, ‘Performance evaluation of probabilistic relay in ad hoc on-demand distance vector and ad hoc on-demand multipath distance vector under highly dynamic environments’, Journal
of Computer Science, vol. 9, no. 7, pp. 905-921
• Arvind, K 1994, ‘Probabilistic clock synchronization in distributed systems’, IEEE
Transactions on Parallel and Distributed Systems, vol. 5, no. 5, pp. 474-487
• Baldoni, R, Corsaro, A, Querzoni, L, Scipioni, S & TucciPiergiovanni, S 2010, ‘Coupling-based internal clock synchronization for large-scale dynamic distributed systems’, IEEE Transactions on Parallel and Distributed Systems, vol. 21, no. 5, pp. 607–619
• Barba, C, Mateos, M, Soto, P, Mezher, A & Igartua, M 2012, In Intelligent
Vehicles Symposium, June 3-7, Smart city for VANETs using warning messages,
traffic statistics and intelligent traffic lights, Spain
• Ben-el-kezadri, R & Angeles, δ 2010, ‘TimeRemap: stable and accurate time in vehicular networks’, IEEE Transactions on Communication, vol. 48, no. 12, pp. 52–57
• Bhumkar, SP, Deotare, VV & Babar, RV 2012, ‘Accident Avoidance and Detection on Highways’, International Journal of Engineering Trends and Technology, vol. 3, no. 2, pp. 247-252
• Blum, P, Meier, L & Thiele, L 2004, In Third International Symposium
on Information Processing in Sensor Networks, April 26-27, Improved interval-
based clock synchronization in sensor networks, California
• Bo, G 2009, ‘Research on clock synchronization mechanism in wireless sensor networks’, Information Technology and Applications, vol. 1, pp. 57–60
• Bo, W, Xiao-hui, Y, Hong-bo, L & Yu-ting, Z 2010, ‘A distributed clock synchronization method based on 2-hop neighbors’ time information for ad hoc networks’, 2nd International Conference on Future Computer and Communication, vol. 1, pp.105-108
• Cao, H, δeung, V, Chow, C, & Chan, H 2009, ‘Enabling technologies for wireless body area networks : a survey and outlook’, IEEE Communications Magazine, vol. 47, no. 12, pp. 84-93
118
• Chandramohan, K & Kamalakkannan, P 2014, ‘Traffic controlled-dedicated short range communication: A secure communication using traffic controlled dedicated short range communication model in vehicular AD’, Journal of computer science, vol. 10, no. 8, pp. 1315-1323
• Chandrasekaran, G 2009, Mobile Networking for Vehicular Environments, May 11,
VANETs: The Networking Platform for Future Vehicular Applications, Los Angeles
• Chaudhari, Q, Serpedin, E & Qaraqe, K 2010, ‘Some improved and generalized estimation schemes for clock synchronization of listening nodes in wireless sensor networks’, IEEE Transactions on Communications, vol. 58, no. 1, pp. 63-67
• Chaudhari, M, Serpedin, E & Shapoury, A 2007, ‘Estimation of clock parameters for synchronization in wireless sensor networks’, IEEE International Conference
on Signal Processing and Communications, pp.1391-1394
• Chen, J, Yu, Q, Zhang, Y, Chen, H & Sun, Y 2010, ‘Feedback-based clock synchronization in wireless sensor networks: A control theoretic approach’, IEEE
Transactions on Vehicular Technology, vol. 59, no. 6, pp. 2963–2973
• Chen, δ, Chang, I, Chang, C & Wang, Y 2013, ‘A secure ambulance communication protocol for VANET’, Wireless Personal Communications, vol. 73, no. 3, pp. 1187–1213
• Cheng, T, Shan, H & Zhuang, W 2011, ‘Infotainment and road safety service support in vehicular networking: From a communication perspective’, Mechanical
Systems and Signal Processing, vol. 25, no. 6, pp. 2020–2038
• Chin, W & Tzen, J 2009, ‘Clock synchronization for energy-constrained wireless sensor networks’, IEEE 20th International Symposium on Personal, Indoor and
Mobile Radio Communications, pp. 1371-1375
• Choi, J & δiang, H 2012, ‘DCS : distributed asynchronous clock synchronization in delay tolerant networks’, IEEE Transactions on Parallel and Distributed systems, vol. 23, no. 3, pp. 491-504
• Deng, Y & Zhang, F 2006, ‘A power management for probabilistic clock synchronization in wireless sensor networks’, International Conference on
Communication Technology, pp.1-4
• Dolev, D, Halpern, Y & Strong, R 1986, ‘On the possibility and impossibility of achieving clock synchronization’, Journal of Computer and System Sciences, vol. 32, no. 2, pp. 230-250
• Elson, J, Girod, L & Estrin, D 2002, In Proceedings of the 5th Symposium on
Operating Systems Design and Implementation, December 9-11, Fine-grained
network time synchronization using reference broadcasts, Boston
119
• Fan, R & δynch, N 2006, ‘Gradient clock synchronization’, Distributed
Computing: Special Issue, vol. 18, no. 4, pp. 255-266
• Fan, R 2005, ‘Clock synchronization for wireless networks’, Springer journal, vol. 3544, pp. 400-414
• Fathima, G & Wahidabanu, A 2011, ‘Integrating buffer management with epidemic routing in delay tolerant networks’, Journal of Computer Science, vol. 7, no.7, pp. 1038–1045
• Forian, D 2005, In Proceedings of the 5th international conference on Privacy
Enhancing Technologies, May 31-June 1, Privacy Issues in Vehicular Ad Hoc
Networks, Heidelberg
• Gurewitz, O, Cidon, I, Member, S & Sidi, M 2006, ‘Network classless time protocol based on clock offset optimization’, IEEE Transactions on
Communications, vol. 14, no. 4, pp. 876-888
• Hanson, M, Powell, C & Barth, A 2009, ‘Body area sensor networks: challenges and opportunities’, IEEE Computer Society, vol. 42, no. 1, pp. 58–65
• Hanzlik, A & Ademaj, A 2006, In Proceedings of International Workshop on
Intelligent Solutions in Embedded Systems, June 30, A composable algorithm for
clock synchronization in multi-cluster real-time systems, Intelligent Solutions in Embedded Systems, Vienna
• Hu, X, Park, T & Shin, K 2008, In Proceedings of The 27th Conference on
Computer Communications, April 13-18, Attack-tolerant time-synchronization in
wireless sensor networks, Phoenix
• Jovanov, E 2005, ‘A wireless body area network of intelligent motion sensors for computer assisted physical rehabilitation’, Journal of Neuroengineering and
Rehabilitation, vol. 2, no. 1, p. 6
• Khairunnisa, S & Syah, I 2014, ‘Stress ; the vulnerability and association with driving performance’, International Journal of Cancer, vol. 11, no. 3, pp. 448–454
• Khan, R, Shah, S, Aleem, M, Aleem, M, Bhutto, Z, Shaikh, A & Kumbhar, M 2012, ‘Wireless sensor networks : a solution for smart transportation’, Journal of
Emerging Trends in Computing and Information Sciences, vol. 3, no. 4, pp. 566-571
• Kuhn, F & Oshman, R 2009, In Proceedings of the 13th
International Conference
on Principles of Distributed Systems, December 15-18, Gradient clock
synchronization using reference broadcasts, France
• δamport, δ 1985, ‘Synchronizing clocks in the presence of faults’, Journal of the
Association for Computing Machinery, vol. 32, no. 1, pp. 52-78
120
• Latre, B 2010, ‘A survey on wireless body area networks’, Springer Journal, vol. 17 no. 1, pp. 1-18
• δee, J, Verma, A, δee, H & εa, J 2005, ‘A distributed time synchronization algorithm robust to traffic load for εANETs’, 62nd Vehicular Technology
Conference IEEE, vol. 1, pp. 329-333
• Li, Q & Rus, D 2006, ‘Global clock synchronization in sensor networks’ IEEE
Transactions on Computers, vol. 55, no. 2, pp. 214-226
• Mamun-Or-Rashid, M, Hong C & Chi-Hyung 2005, In Proceedings of Advanced
Industrial Conference on Telecommunications/Service Assurance with Partial and
Intermittent Resources Conference/E-Learning on Telecommunications Workshop July 17-20, Passive cluster based clock synchronization in sensor network, Portugal
• Markowski, MJ 2008, Modeling behavior in vehicular and pedestrian traffic flow
by modeling behavior in vehicular and pedestrian traffic flow, thesis, University of Delaware Doctoral Dissertations, viewed 2 September 2013
• Mills, L 1994, ‘Improved algorithms for synchronizing computer network clocks’, Computer Communication Review, vol.24, no.4, pp. 317–327
• Mirabella, O, Brischetto, M & Raucea, A 2009, In Proceedings of Wireless Days,
December 15-17, Evaluation of clock synchronization protocols for wireless sensor
networks, Paris
• Mirabella, O, Brischetto, M, Raucea, A & Sindoni, P 2008, In Proceedings of 34th
Annual Conference of IEEE Industrial Electronics, November 10-13, Dynamic
continuous clock synchronization for IEEE 802.15.4 based sensor networks, Orlando
• εostafa, S, Kahaki, ε & Nordin, J 2011, ‘Highway traffic incident detection using high- resolution aerial remote sensing imagery’, Journal of Computer Science, vol. 7, no. 6, pp. 949–953
• Noh, K & Serpedin, E 2007, ‘Pairwise broadcast clock synchronization for wireless sensor networks’, World of Wireless, Mobile and Multimedia Networks, vol. 1, no. 6, pp. 18-21
• Palchaudhuri, S, Saha, K & Johnson, B 2004, In Proceedings of Third International
Symposium on Information Processing in Sensor Networks, April 26-27, Adaptive
clock synchronization in sensor networks, California
• Pattnaik, O & Pattanayak, K 2014, ‘Security in vehicular ad hoc network’, American Journal of Applied Sciences, vol. 11, no. 2, pp. 337–346
• Qian, X, Shen, X, Dai, G, Zhang, J & Lv, C 2010, ‘Clapping and broadcasting synchronization in wireless sensor network’, Sixth International Conference on
Mobile Ad-Hoc and Sensor Networks, vol. 16, no. 6, pp. 632-639
121
• Rahman, M & El-Khatib, K 2010, ‘Secure time synchronization for wireless sensor networks based on bilinear pairing functions’, IEEE Transactions on Parallel and
Distributed Systems, no. 99, pp. 1-15
• Ramanathan, P, Shin, KG & Butler, RW 1990, ‘Fault-tolerant clock synchronization in distributed systems’, Computer, vol. 23, no. 10, pp. 33-42
• Rebai, M, Khoukhi, L, Snoussi, H & Hnaien, F 2012, Wireless
Advanced Conference, June 25-27, Optimal placement in hybrid VANETs-sensors
networks, London
• Reid, A 2009, Time synchronization in ad hoc wireless networks, thesis, Lehigh University, viewed 12 June 2012
• Rentel, H & Kunz, T 2005, ‘A clock-sampling mutual network time-synchronization algorithm for wireless ad hoc networks’, IEEE Wireless
Communications and Networking Conference, vol. 1, pp. 638-644
• Römer, K 2001, ‘Time synchronization in ad hoc networks’, Proceedings of the
2nd International Symposium on Mobile Ad Hoc Networking & Computing –
MobiHoc, vol. 1, pp. 173
• Rus, D 2006, ‘Global clock synchronization in sensor networks’, IEEE
Transactions on Computers, vol. 55, no. 2, pp. 214–226
• Scipioni, S 2009, Algorithms and Services for Peer-to-Peer Internal Clock
Synchronization, thesis, Sapienza University of Rome, Italy
• Shelton, J & Kumar, G 2010, ‘Comparison between Auditory and Visual Simple Reaction Times’, Neuroscience & Medicine, vol. 1, no. 1, pp. 30-32
• Shen, X, Quin, X, Zhao, B, Fang, Q & Dai, G 2011, ‘Clapping and broadcasting synchronization in wireless sensor network’, Tsinghua Science & Technology, vol. 16, no. 6, pp. 632-639
• Sheu, J, Chao, C, Hu, W & Sun, C 2006, ‘A clock synchronization algorithm for multihop wireless ad hoc networks’, Wireless Personal Communications, vol. 43, no. 2, pp. 185-200
• Shinar, D 2012, ‘Safety and mobility of vulnerable road users: pedestrians, motorcyclists and bicyclists’, Accident Analysis & Prevention, vol. 44, no. 1, pp. 1-2
• Singh, D 2010, ‘Hybrid auditory based interaction framework for driver assistance system’, Science & Publications, vol. 6, no.12, pp. 1499–1504
• Sommer, P & Wattenhofer, R 2009, International Conference on Information
Processing in Sensor Networks, April 13-16, Gradient clock synchronization in
wireless sensor networks, San Francisco
122
• Sourour, E, & Nakagawa, ε 1999, ‘εutual decentralized synchronization for intervehicle communications’, IEEE Transactions on Vehicular Technology, vol. 48, no. 6, pp. 2015-2027
• Srinivetha, R & Gopi, R 2014, ‘Alert message dissemination protocol for VANET to improve road safety’, International Journal of Emerging technology and
Advanced Engineering, vol. 4, no. 1, pp. 117–120
• Sun, D, Benekohal, R & Waller, S 2003, In Proceedings of 82nd Transportation
Research Board Meeting, January 12-13, Modeling of motorist-pedestrian
interaction at uncontrolled mid-block crosswalks, Washington
• Sun, K, Ning, P & Wang, C 2006, ‘Secure and resilient clock synchronization in wireless sensor networks’, IEEE Journal on Selected Areas in Communications, vol. 24, no. 2, pp. 395 - 408
• Sun, Z, Bebis, G, & Miller, R 2004, In Proceedings of 7th International IEEE
Conference on Intelligent Transportation Systems, October 3-6, On-road vehicle
detection using optical sensors: a review, Washington
• Sundararaman, B, Buy, U & Kshemkalyani, D 2005, ‘Clock synchronization for wireless sensor networks: a survey’, Elsevier journal, vol. 3, no. 3, pp. 281-323
• Suriyapaiboonwattana, K, Pornavalai, C & Chakraborty, G 2009, In Proceedings of
IEEE International Conference on Fuzzy Systems, August 20-24, An adaptive alert
message dissemination protocol for VANET to improve road safety, Jeju Island
• Swain, R, & Hansdah, C 2010, In Proceedings of Second International Conference
on COMmunication Systems and NETworks, January 5-9, An energy efficient and
Fault-Tolerant Clock Synchronization Protocol for wireless sensor networks, Bangalore
• Veitch, D, Babu, S & Pasztor, A 2004, In Proceedings of the 4th Conference on
Internet Measurement, October 25-27, Robust synchronization of software clocks
across the internet, Italy
• Verma, A 2005, ‘A distributed time synchronization algorithm robust to traffic load for εANETs’, IEEE 62nd Vehicular Technology Conference, vol.1, pp.329-333
• Waizman, G 2012, In Proceedings of 11th International Conference on
Autonomous Agents and Multiagent Systems, June 4-8, SAFEPED: Agent-Based
Environment for Estimating Accident Risks at the Road Black Spots, Valencia, Spain
• Waizman, G & Aviv, T 2015, ‘εicro-Simulation Model for Assessing the Risk of Car-Pedestrian Road Accidents’, Journal of Intelligent Transportation Systems:
Technology, Planning, and Operations, vol. 19, no. 1, pp. 63-77
• Wang, B, Ye, X, Xu, K & Cheng, P 2009, In Proceedings of IEEE International
Conference on Communications Technology and Applications, October 16-18, A
123
mutual clock synchronization method using weight coefficient for wireless ad hoc
networks, Beijing
• Yousefi, S, Mousavi, MS & Fathy, M 2006, In Proceedings of 6th International
Conference on ITS Telecommunications, June 21-23, Vehicular Ad Hoc Networks
(VANETs): Challenges and Perspectives, China
• Zeadally, S, Hunt, R, Chen, Y, Irwin, A & Hassan, A 2012, ‘Vehicular ad hoc networks (VANETs): status, results, and challenges’, Telecommunication Systems, vol. 50, no. 4, pp 217-241
• Zhao, J & Cao, G 2008, ‘VADD : vehicle-assisted data delivery’, IEEE
Transactions on Vehicular Technology, vol. 57, no. 3, pp. 1910-1922
• Zhao, Y, Wang, Y, Huang, J & Shi, X 2008, In Proceedings of International
Conference on Network and Parallel Computing, October 18-21, A Stable Clock
Synchronization Based on Clock Drift Rate, Shangai
• Zhou, D, Lai, T & Science, I 2004, In Proceedings of first IEEE international
conference on Mobile and Sensor Systems, October 25-27, Analysis and
implementation of scalable clock synchronization protocols in IEEE 802. 11 ad hoc
networks, Florida
• Zhou, D & δai, T 2007, ‘An accurate and scalable clock synchronization protocol’, IEEE Transactions on Parallel and Distributed Systems, vol. 18, no. 12, pp. 1797-1808
124
WEB REFERENCES
• Autonomous Car n.d., viewed Aug 19 2012,
https://en.wikipedia.org/wiki/Autonomous_car
• Driverless Car n.d., viewed July 12 2012,
<http://en.wikipedia.org/wiki/Google_driverless_car>
• Huge chain-reaction crash in UK, Europe 2013, image, CNN, 5 Sep, viewed 25 Jan
2014, <http://edition.cnn.com/2013/09/05/world/europe/uk-huge-chain-reaction-
crash/>
• Indian Motor Vehicles n.d., viewed Oct 18 2016,
<http://www.knowindia.net/auto.html>
• Road Accidents n.d., viewed Oct 23 2012,
http://www.visualexpert.com/Resources/roadaccidents.html
• Stopping Distance n.d., viewed Jan 15 2013,
<http://www.seas.upenn.edu/~ese302/labcontent/STOPPING_DISTANCE_DOC.p
df>
• Stopping Sight Distance n.d., viewed Nov 2 2012,
<http://en.wikipedia.org/wiki/Stopping_sight_distance>
• VANETs n.d., viewed July 12 2012, < http://en.wikipedia.org/wiki/Vehicular_ad-
hoc_network >
125
LIST OF PUBLICATIONS
Dahlia Sam, Esther Evangelin, T & Cyril Raj, V 2015, ‘Improving Road Safety for
Pedestrians in Black Spots using a Hybrid VANET of Vehicular Sensors and
Pedestrian Body Unit’, ARPN Journal of Engineering and Applied Sciences, vol.
10, no. 10, pp. 4639-4644 (Scopus Indexed)
Dahlia Sam & Cyril Raj, V 2014, ‘A time synchronized Hybrid Vehicular Ad Hoc
Network (H-VANET) of roadside sensors and vehicles for safe driving’, Journal of
Computer Science, vol. 10, no. 10, pp. 1617-1627 (Scopus Indexed)
Esther Evangelin, T, Dahlia Sam & Cyril Raj, V 2014, ‘Wireless Body Area
Networks and its Emerging Technologies in Real Time Applications’, International
Journal of Engineering Sciences and Research Technology, vol. 3, no. 1, pp. 309-
313
Dahlia Sam & Cyril Raj, V 2014, ‘VANETs: A Platform for the future Intelligent
Transport System (ITS)’, Asian Journal of Information Technology, vol. 13, no. 1,
pp. 38-45 (Scopus Indexed)
Dahlia Sam & Cyril Raj, V 2013, ‘A study on clock synchronization protocols in
different networks’, European Journal of Scientific Research, vol. 105, no. 3, pp.
352-363
Dahlia Sam, Esther Evangelin, T & Cyril Raj, V 2015, ‘A novel idea to improve
pedestrian safety in Black Spots using a Hybrid VANET of vehicular and body
sensors’, IEEE International Conference on Information, Innovation in Computing
Technology
