GNSS Time Synchronisation in
Co-operative Vehicular Networks
A THESIS SUBMITTED TO
THE SCIENCE AND ENGINEERING FACULTY
OF QUEENSLAND UNIVERSITY OF TECHNOLOGY
IN FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Khondokar Fida Hasan
M.Sc. ; B.Sc. (Hons)
School of Electrical Engineering and Computer Science
Science and Engineering Faculty
Queensland University of Technology
2018
GNSS Time Synchronisation in Co-operative Vehicular
Networks
A THESIS SUBMITTED TO
THE SCIENCE AND ENGINEERING FACULTY
OF QUEENSLAND UNIVERSITY OF TECHNOLOGY
IN FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Khondokar Fida Hasan
M.Sc. ; B.Sc. (Hons)
Supervisor: Professor Yanming Feng and Professor Yu-Chu Tian
School of Electrical Engineering and Computer Science
Science and Engineering Faculty
Queensland University of Technology
2018
ii
Copyright in Relation to This Thesis
c© Copyright 2018 by Khondokar Fida Hasan
M.Sc. ; B.Sc. (Hons). All rights reserved.
iii
QUT Verified Signature
Abstract
Time synchronisation is a prerequisite for the successful operation of every distributed
network. It provides a common time frame among all nodes, thus supporting var-
ious network functions such as message transmission, channel scheduling and re-
source sharing in real-time and correct order. Time synchronisation also enables
secure connectivity, data consistency and process coordination among the nodes of
the distributed network system. Accurate time is especially important in network
applications with high mobility such as in vehicular ad-hoc networks (VANETs). In
practice, VANETs differ from other mobile networks by their ad-hoc architecture,
high mobility, and time-sensitive applications. In vehicle-to-vehicle and vehicle-to-
infrastructure communications, the network nodes are moving rapidly and establish-
ing communications by forming networks in an ad-hoc manner. Such ad-hoc links are
usually short lived, as the relative speed of the nodes can be as high as 200 km/h (56
m/s). Therefore, maintaining network-wide accurate time becomes critical. Since the
success of vehicular communication networks impacts on peoples lives and resources
on the road, VANET safety requirements (e.g., delay, reliability, scalability, fairness,
and timeliness of vehicle-to-everything (V2X) communications) are stringent. For
instance, the update rate of vehicle location information can be as high as 10-100Hz.
The requirements for absolute and relative vehicle positioning accuracy can be as high
as 10 centimetres. However, the requirements for time synchronisation in V2X com-
munications is not well understood. While many synchronisation techniques have
been developed for general networks, they are not particularly suited for VANETs.
v
That is, it is critical to understand the efficacy of existing time synchronisation tech-
niques in VANET applications. GNSS has been used for providing precise timing
information in many distributed systems. Since VANET is mostly outdoor-based, the
integration of GNSS-based time synchronisation in-vehicle networks offers promising
technological solutions for better coordination of emerging automated and connected
vehicles.
This thesis identifies some important application scenarios for precise timing and
relative time synchronisation for V2X communications. It compares various proposed
synchronisation methods, and discusses the performance benefits of GNSS timing
techniques for synchronising vehicular networks. Extensive experiments show that
GNSS time synchronisation methods can replace existing time synchronisation func-
tion (TSF) based synchronisation in VANETs. The limitations due to GNSS time
solution outages in urban streets and tunnels are also analysed using vehicle GNSS
tracking data recorded in the Brisbane CBD. The described analyses indicate that
VANET time drift solution outages can be successfully handled and strategies can
be untaken for mitigation. The results of experiments are discussed, that evaluate
the timing accuracy possible using multi-layer GNSS time synchronisation, which
demonstrate its compatibility and feasibility in a number of real Vehicle to Vehicle
(V2V) application scenarios.
vi
Keywords
V2X, Intelligent Transportation System, C-ITS, VANET, GNSS, Time Synchronisation,
Timing Advertisement, 1PPS, Security, MAC
vii
viii
Acknowledgments
It is my utmost pleasure to acknowledge the roles and contributions of several indi-
viduals who were instrumental for completion of my PhD research.
In the very first place, I would like to sincerely thank my principal supervisor Profes-
sor Yanming Feng for his support and invaluable guidance throughout the journey of
my PhD degree. I also would like to extend my special appreciation to Professor Glen
Tian, my associate supervisor, for his invaluable input and significant support during
this research.
I thank my fellow lab-mates for their enthusiastic support, for their caring and also
for all the fun we have had together to make a mindful place. Dr Keyvan Ansari, Dr
Charles Wang and Dr Lei Wang are few names need to be mentioned for their time
and contribution with my research work in various ways.
I also would like to express my thank to all the academic and professional staff in the
School of Electrical Engineering and Computer Science of the Queensland University
of Technology for supporting me in various ways during this project and also for
providing me with a stimulating, supportive research environment.
Last but not the least my most profound appreciation, acknowledgment and my
thankfulness to all of my family members for their generous support throughout the
journey of my PhD degree.
ix
x
Table of Contents
Abstract v
Keywords vii
Acknowledgments ix
List of Figures xiii
List of Tables xv
1 Introduction 1
1.1 Research Background and Motivation . . . . . . . . . . . . . . . . . . . 1
1.1.1 Intelligent transportation Systems and Vehicular Networks . . 2
1.1.2 Co-operative Intelligent Transportation System . . . . . . . . . 5
1.1.3 Importance of Time and Time Synchronisation in Distributed
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 GNSS Time Synchronisation and Challenges . . . . . . . . . . . . . . . 7
1.2.1 GNSS Time Synchronisation Prospects in Vehicular Networks . 7
1.2.2 Synchronisation Challenges and Research Questions . . . . . . 8
1.3 The Objective and Scope of the Research . . . . . . . . . . . . . . . . . . 10
1.4 Research Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
xi
1.5 Dissertation Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 List of Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Fundamental Theories and Related Works 17
2.1 Cooperative Intelligent Transportation System (C-ITS) Fundamentals . 18
2.2 Enabling Architectures, Technologies and Standardisation . . . . . . . 19
2.2.1 Vehicular Ad-Hoc Network (VANET) . . . . . . . . . . . . . . . 19
2.2.2 Dedicated Short Range Communication (DSRC) . . . . . . . . . 21
2.2.3 Wireless Access in Vehicular Network (WAVE) . . . . . . . . . . 21
2.2.4 Overview of IEEE 802.11P . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Basic Models and Techniques of Time Keeping and Time Synchronization 25
2.3.1 Hardware Clocks . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.2 Software Clocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.3 Clock Accuracy and Precision . . . . . . . . . . . . . . . . . . . . 27
2.3.4 Clock Offset, Skew and Drift . . . . . . . . . . . . . . . . . . . . 27
2.3.5 Main Limiting Factors in Time Synchronization . . . . . . . . . 28
2.3.6 Basic Techniques for Time Synchronization in a Decentralized
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.7 Types of Clock Synchronization . . . . . . . . . . . . . . . . . . . 31
2.4 Approaches to the Time Synchronization in Wireless Media . . . . . . . 32
2.4.1 Revisiting Time Synchronization in Wireless Sensor Networks . 33
2.4.2 Why is Time Synchronization an Issue in VANETs . . . . . . . . 35
2.5 Existing Recommendation for VANET Time Synchronization . . . . . . 37
2.6 GNSS Approaches for VANET Time Synchronization . . . . . . . . . . 41
2.6.1 Motivation of GNSS-driven Time Synchronization in VANETs . 42
xii
2.6.2 GNSS Time Synchronization Models for VANET . . . . . . . . . 43
2.6.3 Challenges and Solutions in Absence of GNSS Signals . . . . . 47
2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3 Significance and Requirement Analysis for Time Synchronisation in VANET 51
3.1 The Need for Time Synchronisation in VANETs . . . . . . . . . . . . . . 52
3.2 Time Synchronisation Requirements in VANET . . . . . . . . . . . . . . 54
3.2.1 Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.2 Precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2.3 Availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2.4 Requirements for Different VANET Applications . . . . . . . . 57
3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4 GNSS Time Synchronisation In VANET 67
4.1 Motivation of GNSS-driven Time Synchronisation . . . . . . . . . . . . 68
4.2 Feasibility of GNSS Synchronisation in VANET . . . . . . . . . . . . . . 69
4.2.1 Justification of the Feasibility . . . . . . . . . . . . . . . . . . . . 69
4.2.2 GNSS Timing Information . . . . . . . . . . . . . . . . . . . . . . 72
4.2.3 Errors of the Receiver Timing . . . . . . . . . . . . . . . . . . . . 74
4.3 Availability of GNSS Time Solutions . . . . . . . . . . . . . . . . . . . . 77
4.4 Synchronisation Accuracy of 1 PPS Signals . . . . . . . . . . . . . . . . 78
4.4.1 Characteristics of 1PPS . . . . . . . . . . . . . . . . . . . . . . . . 78
4.4.2 Clock Accuracy of Low Cost GNSS Receivers . . . . . . . . . . . 79
4.4.3 Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.4.4 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
xiii
4.4.5 Result Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5 GNSS Synchronisation with On-board Devices 93
5.1 Problem Definition and Solution Approach . . . . . . . . . . . . . . . . 94
5.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.3 Testbed Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.4 Result Analysis and Performance Evaluation . . . . . . . . . . . . . . . 103
5.4.1 Node Clock Synchronisation With GNSS Receiver . . . . . . . . 106
5.4.2 Field test of GNSS Time Synchronisation . . . . . . . . . . . . . 109
5.4.3 Network Synchronisation with GNSS . . . . . . . . . . . . . . . 110
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6 Synchronisation in Occasional loss of GNSS Signals 117
6.1 Understanding Service Availability . . . . . . . . . . . . . . . . . . . . . 118
6.2 Fall Back Solutions During Signal Outage . . . . . . . . . . . . . . . . . 123
6.2.1 Number of Satellites 1 to 3 (0<NSAT< 4) . . . . . . . . . . . . . 123
6.2.2 Number of Satellite is Zero (No visible satellites . . . . . . . . . 124
6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7 Conclusions and Recommendations 129
7.1 Summary of the Research . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.2 Summary of the Contribution . . . . . . . . . . . . . . . . . . . . . . . . 131
7.3 Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Literature Cited 149
xiv
List of Figures
1.1 The Concept of Intelligent Transportation System. . . . . . . . . . . . . 3
1.2 Vehicle to Everything (V2X) Communication. . . . . . . . . . . . . . . . 4
1.3 Challenges with Existing Time Synchronisation Solution Recommended
in VANET. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 Basic Interactions of Cooperative ITS. . . . . . . . . . . . . . . . . . . . 18
2.2 VANET Communication Architecture. . . . . . . . . . . . . . . . . . . . 20
2.3 DSRC Channel Arrangement. . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 WAVE Protocol Stack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Unsynchronsed Clocks in a network. . . . . . . . . . . . . . . . . . . . . 23
2.6 Synchronsed Clocks in a network. . . . . . . . . . . . . . . . . . . . . . 24
2.7 Message Exchanges between Two Nodes. . . . . . . . . . . . . . . . . . 30
2.8 Reference Broadcasting Synchronisation (RBS) . . . . . . . . . . . . . . 30
2.9 Two Modes of communications in 802.11 Standard Family. . . . . . . . 37
2.10 Beacon Generation Window. . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.11 GPS Time Transfer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.1 WAVE Spectrum: (a) Frequency and Channel Allocation; and (b) Chan-
nel Synchronisation and Guard Interval . . . . . . . . . . . . . . . . . . 59
3.2 Guard Interval Requirements . . . . . . . . . . . . . . . . . . . . . . . . 60
xv
3.3 Examples of Security Issue. . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.4 Concept Tier fo the Requirements of Time Synchronisation Accuracy . 65
4.1 Broad View of Time Synchronisation (a) In-band Time Synchronisation (b)
Out-of-Band External Time Synchronisation. . . . . . . . . . . . . . . . . . . 70
4.2 Time Offsets Among Different Atomic Scale Standards. . . . . . . . . . . . . 72
4.3 Time Transfer Through GNSS . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4 (a) (b) Nodes N1 and N2 are Individually Synchronised with GNSS and Up-
dated with UTC. (c) Effectively, Two Nodes are Synchronised with Each Other
via GNSS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.5 Ideal (GPS signal) and Practical (Produced Signal by GPS receiver) 1PPS Sig-
nal Pulses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.6 Standard Deviation Rating. . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.7 Schematic Diagram of Experimental Setup for Synchronisation Assessment of
GPS Receivers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.8 Pulses Showing Time Difference Between Two Waveforms of Identical Re-
ceivers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.10 Gaussian Distribution of Dataset 5 Minutes that Represents STD of 12.67 ns
for 1σ.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.9 Relative Time Offsets Between Two Receivers of Identical Models. . . 86
4.11 Time offsets between receivers of same model. . . . . . . . . . . . . . . 89
4.12 Time Offset Between Receivers of Different Models over a Long Period. . . . 90
4.13 Time Offset Between Receivers of Different Models Over a Long Period. . . . 90
4.14 Time Offset Between Receivers of Different Models Over a Long Period. . . . 91
5.1 VANET Communication: Vehicle to Everything Scenario. . . . . . . . . . . . 94
xvi
5.2 VANET Communication: Vehicle to Infrastructure (V2I), and Vehicle to Vehi-
cle (v2v) Communication. . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.3 Existing Problem with Time-Advertisement-based Time Synchronisation. 96
5.4 Undefined Situation Using TA Mechanism in Pure Ad-hoc Communication. . 97
5.5 Wireless Access for Vehicular Environment (WAVE) Layers. . . . . . . . . . 99
5.6 Proposed Solution. GNSS Time Synchronisation to TSF Register Through
Application Layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.7 Communicating Node Synchronisation using a GNSS Receiver . . . . 102
5.8 Schematic Diagram of the Experimental Setup on Moving Node to Collect
Data on Real-time Vehicular Environment Including Urban, Suburb, High-
way, etc. Route. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.9 System Clock Synchronised with GPS Data Signal. . . . . . . . . . . . . . . 104
5.10 System Clock Synchronised with PPS Ticks. . . . . . . . . . . . . . . . . . . 107
5.11 System Clock Synchronised with Both GPS PPS Signal. . . . . . . . . . . . . 108
5.12 Statistical Distribution of Measured Time Difference. . . . . . . . . . . . . . 109
5.13 Illustration of Clock Stability and Noises using Allan Deviation (Log-Log scale).110
5.14 GPS-PPS Enabled Clock in Different Road Scenarios. . . . . . . . . . . . . . 111
5.15 Schematic Diagram of the Experimental Setup to Measure the Time
Offsets between Two GNSS Synchronised Computing nodes. . . . . . . 112
5.16 Box-plot of the Time Offsets Between Two GNSS Synchronised Node
Developed on Rpi. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.17 Offsets Between Individual Packets and their Moving Average from the Dataset
of 300 Packets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.1 GPS Challenged Environment in Urban Concept -1. . . . . . . . . . . . 120
6.2 GPS Challenged Environment in Urban Concept-2. . . . . . . . . . . . 121
xvii
6.3 Vehicle Tracks of of GPS, BDS and GPS+BDS on High Rising Roads. . . 122
6.4 The Number of Satellites Under the Signal Coverage of BDS and GPS. 122
6.5 Node Clock Integrated with GNSS. . . . . . . . . . . . . . . . . . . . . . 124
6.6 Schematic Diagram of the Experimental Set-up Between Three Nodes. 125
6.7 Plot of the Clock Drift Recorded over 4 hours. . . . . . . . . . . . . . . . . . 125
6.8 Plot of the Clock Drift Recorded over 4 hours. . . . . . . . . . . . . . . . . . 126
xviii
List of Tables
1.1 Overview of Multi GNSS Environment (Current Status 2018) . . . . . . 8
2.1 Slot Time with Beacon Generation Window. . . . . . . . . . . . . . . . . 38
3.1 List of Timing Accuracy Requirements for Different Applications in the Basis
of Essentialness of VANET. . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.1 The Number of Satellites Available with Different GNSS Services Constellation. 78
4.2 Relative Time offsets of Different Datasets in the Time range of Nanosecond
Between Receivers of Same Model (Ublox-Ublox). . . . . . . . . . . . . . . . 87
4.3 Relative Time Offsets of Different Datasets in the Time Range of Nanosecond
Between Receivers of Same Model (Furuno-Furuno). . . . . . . . . . . . . . 88
4.4 Relative Time Offsets From Different Datasets in the Time Range of Nanosec-
ond Between Receivers of Different Model (Ublox-Furuno) . . . . . . . . . . 88
4.5 Relative Offset in ns Between Receivers of Different Models. . . . . . . . . . 90
4.6 Relative Offset in ns Between Receivers of Different Models. . . . . . . . . . 91
6.1 Number of Satellites Available with Different GNSS Service Constella-
tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.2 GDOP with Different GNSS Services . . . . . . . . . . . . . . . . . . . . 122
xix
xx
Chapter 1
Introduction
1.1 Research Background and Motivation
Transportation is an indispensable part of modern civilisation. It is inseparable from
society and exerts a powerful influence on the lives of individuals and the develop-
ment of nations. According to the US Department of Transportation, their travelling
time is about 500 million hours in vehicles per week, which signifies the importance
of transportation in everyday life in modern societies. Among different forms of
transportation, including air, water and land, the most common and popular is road-
based land transportation. In reference to the Australian Bureau of Statistics [2012],
around 78% of Australians travel to their workplace by car BITRE2 [2012]. Since road
transportation is so popular, it has a record of significantly large rate of causalities.
According to the UN Tackles Road Safety Report (2015), road transport accidents
kill about 1.3 million people every year, which equates to nearly 3400 road fatalities
per day. Road accidents injure 50 million people annually. The report also predicts
that the road crashes will be the fifth leading cause of death by 2030 [WHO, 2015]. It is
understandable that some countries possessing minimal technological facilities have
higher per-capita accident rates. However, according to the Australian Department
of Infrastructure, there was a total of 129 road deaths during December 2017 [BITRE,
2017]. Since the population of Australia is less compared to many other countries, this
1
2 CHAPTER 1. INTRODUCTION
number is alarming. BITRE [2017] also reported that the metropolitan congestion cost
is more than $16.5 billion in the year 2015 and would rise to $30 billion by 2030, which
undoubtedly would have a significant impact on the economy, society and nation as
a whole. To reduce deaths, injuries, and economic losses resulting from motor vehicle
crashes, road transportation networks demand evolution and advancement. As a part
of this endeavour, Intelligent Transportation System (ITS) technologies are emerging
as a tool to alleviate the situation.
Time is considered as one of the important and recognised parameters for success-
ful operation in existing technology-based communication systems, such as computer
networks, cellular network and sensor network. However, today’s transportation sys-
tems do not depend on precise and accurate time in their operation. In order to alle-
viate the road fatalities including death, injuries and economic losses, transportation
system researchers are evolving the concept of an ’Intelligent Transportation System
(ITS)’, where vehicles can collect data within their surrounding area to support inde-
pendent decision making. An ITS would also enable vehicles to communicate with
each other and surrounding road elements through wireless networks. Such tech-
nologies would employ existing or amended sensor, computer and communication-
based technologies together with their attendant protocols. Therefore, it can be as-
sumed that the ITS would support time-sensitive operations and applications, similar
to other existing technology-based sensor operations and communication networks.
Thus, it is essential to understand the timing accuracy and precision required by
emerging ITS-based technologies for road networks, and what would be the best way
to achieve them.
1.1.1 Intelligent transportation Systems and Vehicular Networks
In a broad view of ITS, various sensors and assisting systems are integrated into
vehicles to monitor their local environment, so that they can gain accurate knowledge
1.1. RESEARCH BACKGROUND AND MOTIVATION 3
of their surrounds Fagnant and Kockelman [2015]. Along with this, wireless commu-
nications also support exchange of information between vehicles and the other road
elements. Thus, as a whole, ITS support ranges from Autonomous Vehicle (AV) to
Connected Vehicles (CV) as shown in the Figure 1.1.
GPS
Lidar
Radar
Camera
Ultrasonic
OdometryCentral Computer
Inertial Sensors
V2V
I2I
V2I
V2P
(a) (b)
Figure 1.1: The Concept of Intelligent Transportation System. (a) AutonomousVehicle (b) Connected Vehicles Conde et al. [2015], Luettel et al. [2012].
In autonomous vehicles, a good range of sensors is integrated into the vehicle to
collect the data about the surrounding environment to enable them to operate inde-
pendently. In the case of Connected Vehicles (CV), drivers have access to wireless con-
nections to other neighbouring vehicles, infrastructure, pedestrians and other devices
within its proximity. The principal advantage of having wireless communications is
the ability to access information that may otherwise be beyond the driver’s immediate
awareness. Such wireless communications would help to prevent possible collisions
by exchanging status information (such as the location, speed and direction of travel
of nearby vehicles etc.), and event-driven safety messages (such as lane changing and
collision warnings etc.). Along with these safety warning messages, such communi-
cation would also support sharing traffic information, weather updates and Internet-
based infotainments.
4 CHAPTER 1. INTRODUCTION
This multifarious communication on the road is popularly termed as Vehicle to
Everything or V2X communication. A V2X communication concept is shown in the
Figure 1.2.
V2V
V2P
V2D
V2
I
V2I (V2H, V2G )
Figure 1.2: Vehicle to Everything (V2X) Communication.
Dedicated short-range communication (DSRC) and Long Term Evolution (LTE)
are two widely-used schemes for Vehicle to Everything (V2X) communications, which
enable Connected Vehicle (CV) applications. The implementation of V2X communi-
cations on LTE networks is known as LTE-Vehicle (LTE-V). This supports V2V com-
munications (also known as side-link) using LTE’s direct interface, called PC5. LTE-
V comprises two radio interfaces, PC5 for V2V and the other interface, called Uu,
supports V2I communications [Molina-Masegosa and Gozalvez, 2017]. In compari-
son, IEEE 802.11p-based DSRC communications use the 5.9 GHz band link to connect
both vehicles and infrastructure. Since IEEE802.11p is considered to be the de facto
standard by the vehicular transportation community and authority for V2X commu-
nication [Araniti et al., 2013]; the scope of study herein is limited to DSRC-based
vehicular communications. However, it is acknowledged that GNSS timing systems
also are equally applicable in LTE-V-enabled environments.
1.1. RESEARCH BACKGROUND AND MOTIVATION 5
1.1.2 Co-operative Intelligent Transportation System
Developing wireless communications between vehicles and other road components
leads to the concept of the cooperative environments. DSRC-enabled road commu-
nication technology is often termed a Cooperative Intelligent Transportation System
(C-ITS), which is increasingly considered to be an important tool for managing road
safety and road efficiency. Enabling wireless connectivity on the road is the primary
aim of CITS. It creates an environment of information sharing and active cooperation
for increased road safety and system operation efficiency. A key component of the C-
ITS technology is a Vehicular Ad-hoc NETwork (VANET). Generally, a typical VANET
consists of a large number of network nodes that pass each other at high speed. There-
fore, it is considered to be a highly dynamic and decentralised network. VANETs have
unique characteristics in compared with other types of wireless networks. A C-ITS
crucially relies on VANETs for real-time, wireless communication of data generated
from in-vehicle and roadside sensors. Therefore, the functionality and performance
of VANETs determine how well a C-ITS can support various applications on the road
[Ghosh et al., 2011, Golestan et al., 2012a, Sundararaman et al., 2005].
Due to the decentralized nature of VANET, efficient timing and coordination ca-
pabilities are considered to be important prerequisites for ensuring vehicle network
management and hence safety on roads. Therefore, the possibility of the integration
of GNSS time synchronisation over VANET is considered to be important for realising
C-ITS.
1.1.3 Importance of Time and Time Synchronisation in Distributed Systems
Successfully supporting communications for different services in vehicular environ-
ments is highly reliant on the accuracy of timing information at the network nodes.
The timely delivery of various messages in a correct and precise order is crucial for
effective and efficient VANET services. Some VANET applications have a time-offset
6 CHAPTER 1. INTRODUCTION
tolerance requirement below 100 ms. This may be achievable when all network nodes
operate on the same and commonly agreed clock time.
Time synchronisation is a technique for maintaining clock times at all network
nodes.1 In general, every physical clock drifts away from the actual daytime by 1 µs
to 100 µs per second. This implies a range of deviation about 5 to 15 seconds per day
[Lombardi, 2000]. As VANETs are distributed and decentralised networks in which
the nodes are physically detached from each other. This means that differences in
VANET node times become unavoidable. Maintenance of an exact, network-wide
clock time is, therefore, impossible for VANETs. Therefore, time synchronisation
services and applications need to be developed for all network nodes. The time
synchronisation is required to align the (drifting) clocks of all network nodes with a
global standard time or with each other. In this way, every node in the network could
operate with the same notion of time. This would support reliable and precise time
synchronisation in various VANET services such as coordination, communication,
security, and time-sensitive applications.
In wireless communications, time synchronisation is essential for coordination
and consistent operation of various network elements. It is also necessary for accurate
message sequencing and real-time control tasks.
While VANET communications are considered to be asynchronous in nature, time
synchronisation among vehicles is essential for many applications [Skog and Han-
del, 2011]. This is similar to the Internet, which is embedded with various time-
synchronisation mechanisms. Some VANET applications are highly time-sensitive. In
these applications, maintaining a real-time and also precise timing between commu-
nicating nodes is critical. Therefore, as in other synchronous wireless distributed sys-
tems, many VANET applications depend on synchronous communications to provide
1In contemporary terminology, adjusting merely clock frequency of the clocks in a network refers tosynchronisation of frequency or syntonization. In contrast, synchronising time means setting the clockto agree upon a particular epoch, with respect to a standard time format such as Coordinated UniversalTime (UTC). Synchronising a clock also refers to the synchronisation of both frequency and time. Inthis thesis, the terms time synchronisation and clock synchronisation are used synonymously to refer tomaintaining clocks to the same time and run at the same frequency.
1.2. GNSS TIME SYNCHRONISATION AND CHALLENGES 7
common C-ITS services. Examples includes, the coordination of activities [Cozzetti
and Scopigno, 2011, Sjoberg et al., 2011, Sjoberg Bilstrup, 2009], relative vehicle po-
sitioning, data communications, and security services [Shizhun Wang et al., 2010].
Moreover, in order to record event information over a network, a VANET needs
to maintain accurate physical time. This also demands maintenance of an accurate
standard time through time synchronisation in an asynchronous manner.
1.2 GNSS Time Synchronisation Challenges
1.2.1 GNSS Time Synchronisation Prospects in Vehicular Networks
The Global Navigation Satellite System (GNSS) is a well-known satellite-based com-
munication system that provides users with positioning, navigation and timing (PNT)
services. Initially, the USA launched the Global Positioning System (GPS) satellites,
which now provide complete coverage of the earth. Subsequently, Russia’s GLONASS,
Chinese’s BeiDou and the European Union’s Galileo joined in providing indepen-
dent satellite-based positioning, navigation, timing and communication services. The
number of satellites in current [2018] GNSS systems along with plans for future con-
stellations are given in Table 1.1. As a single satellite system, GPS currently has the
most number of satellites in orbit, namely 31. However, in the multi-GNSS concept,
this number increases to 99 and is expected to reach 134 by the year 2020. This obvi-
ously increases the probability of detecting satellite signals always from a number of
satellites that would help to support accurate PNT services.
The timing service provided by GNSS is considered to be the reference source
worldwide. It is also considered as the most accurate UTC sources because GNSS-
based satellite time is maintained with highly-precise atomic clocks, and satellite time
transfer is more reliable than other ground-based time-transfer techniques.
There are some fundamental reasons for that, first of all, the accuracy of any time
transfer system suffers from the uncertainty due to the variability of propagation path
8 CHAPTER 1. INTRODUCTION
Table 1.1: Overview of Multi GNSS Environment (Current Status 2018)
GNSS SystemNo. of Currently
OperationalSatellite #
Planned# of Satellites
Remarks
GPS 31 3236 orbit planes
Ref: [Uscg.gov, 2018]
GLONASS 24 243 orbitplanes
Ref: [Glonass iac.ru, 2018]
BEIDOU 2235
(by 2020)
6 GEO, 8 IGSO,8 MEO (in 2 planesRef: [I-GNSS, 2018])
GALILEO 2230
(by 2020)
2 under testing and4 under commissioning
Ref: [E-GNSS, 2018]Overall 99 134
delays and its . In comparison with ground-based radio communication systems,
satellite-based GNSS systems have path delays which are straightforward to measure
and calibrate. This is because the variation of path delays are small, as the paths
between satellite and receiver are mostly unobstructed. In addition, signal interfer-
ences caused by the influence of weather is usually less of a problem since they can
be mathematically modelled and compensated. Therefore, the time steering by the
GNSS systems are accurate and ubiquitous under an unobstructed sky.
Meanwhile, modern vehicles are already integrated with a GPS (Global Position-
ing System) or multi-GNSS (Global Navigation Satellite System) receiver for naviga-
tion. Accurate time support from GNSS is, therefore, plausible without any additional
costs. While GPS-based time synchronisation is used in many networks, limited
publications have been found about GPS or GNSS time-synchronisation performance.
1.2. GNSS TIME SYNCHRONISATION AND CHALLENGES 9
1.2.2 Synchronisation Challenges and Research Questions
Despite the recent modernisation of GPS receivers, it is still not well-understood
whether GNSS time synchronisation is beneficial and feasible for vehicular networks.
GNSS receivers are available in different manufacturing grades according to dif-
ferent type of applications and cost. A consumer-grade receiver commonly employs
a quartz oscillator and basic circuitry to make the system inexpensive for use in
everyday navigation applications. Most of the navigational GNSS devices built into
cars are in the consumer grade category. Since they use inexpensive circuitry, it
is crucial to understand what time synchronisation accuracy they may offer. The
IEEE 802.11p based WAVE (Wireless Access for Vehicular Environment) structure has
a synchronisation mechanism incorporated within an IEEE 1609.04 protocol stack,
known as Timing Advertisement (TA), which relies on message transfers between
nodes. A TSF (Time Synchronisation Function) register located in the MAC is respon-
sible for coordinating the synchronisation process. Figure 1.3 shows a diagrammatic
representation of the TA scheme, where the TSF of the Access Point (AP) is able to
maintain time which also sends out time through the communication channel and
moving station receives and updates according to the received time. Due to the
highly dynamic nature message-transfer-based time synchronisation is challenging
the vehicular environments, since with 200 km/h relative speeds, the nodes may be
in their communication range for only a fraction of a second. Although TA-based
schemes have challenges, it is important to investigate the prospects of GNSS time
synchronisation over such systems.
In addition, GNSS signals are prone to being obstructed by buildings, trees and
other barriers, which can lead to service disruptions. Vehicular networks are mostly
outdoors based, where the communications occur under the open sky, thereafter, in
urban areas, tall buildings, trees and tunnels pose challenges for delivering GNSS
services.
10 CHAPTER 1. INTRODUCTION
Application
TSF Register
Application
TSF Register
RSU/AP OBU/STATA
Frame
Ref clock
Data 12:00 am Hardware Medium
TSF
Data 12:00 am Hardware
TSF
Applications
Figure 1.3: Challenges with Existing Time Synchronisation Solution Recommendedin VANET Mahmood et al. [2017].
Based on the above-mentioned challenges, the primary research questions of in-
terest are as follows:
Q 1. Can consumer-grade GNSS receivers provide the time synchronisation accuracy
and precision required for VANETs?
Q 2. In vehicular environments, is GNSS time synchronisation better than existing
decentralised time synchronisation methods that were originally developed for
other wireless networks?
Q 3. To what extent are GNSS services available to provide time synchronisation
support to vehicular networks, and what possible measures can be employed
during signal outages?
1.3 The objective and scope of the research
The primary aims of this research are twofold. Firstly, investigating the feasibility and
prospects for GNSS time synchronisation in vehicular networks. This includes study-
ing the timing accuracy and precision requirements for different vehicular network
applications. The investigations involve examining the achievable accuracy offered
1.4. RESEARCH CONTRIBUTION 11
by the low-cost consumer-grade GNSS receivers. The objectives also include deter-
mining the achievable accuracy of GNSS receivers integrated with on-board devices
in laboratory-based and field experiments. Secondly, in respect of GNSS availability,
understanding the extent and impact of signal outages, and measures for maintaining
timing synchronisation within vehicular networks. Based on the objectives, goals and
challenges discussed above, the main objectives of the research are concisely restated
below.
1. To contribute an understanding of time synchronisation within different vehicle
network applications and also to identify their requirements.
2. To explore the potential of GNSS time synchronisation in vehicular networks
by carrying out comparative analyses of new and existing synchronisation sys-
tems, and conducting experimental validations using integrated on-board GNSS
receiver systems.
3. To identify GNSS timing service availability in vehicular environments and to
suggest potential timing solutions to manage GNSS signal and service outages.
1.4 Research Contribution
The pursuit of time synchronisation methods for distributed networks is a well-studied
problem which has a rich history of developments. GNSS-supported solutions have
also been proposed for various applications in different outdoor based networks, and
for the root server in many indoor-based networks. However, the outdoor nature of
vehicular networks, their unique architectures, along with recent advances of GNSS
systems, has prompted the development of a complete GNSS time synchronisation
solution for vehicular networks. The specific research contributions of this thesis are
described below.
1. One significant contribution to knowledge, is the identification of time-sensitive
12 CHAPTER 1. INTRODUCTION
applications in vehicular networks, through an extensive analysis and literature
survey. The findings of the analysis pertain to the timing accuracy, precision
and availability in vehicular environments. After identifying the time-sensitive
applications and their requirements, they are prioritised with respect to impor-
tance and categorised as essential and desirable, and discussed in the context of
the available technology. This is an important step towards the understanding
of vehicle network timing synchronisation problems.
2. Time synchronisation solutions using GNSS timing services are proposed for
attaining accurate and precise time in distributed vehicular networks. This
involved conducting a thorough feasibility analysis of GNSS time synchroni-
sation. A major research contribution includes the experimental verification of a
UTC-based-1PPS-timing-accuracy method that can be implemented on consumer-
grade GNSS receivers. It also provides an insight into time-synchronisation
performance due to the variation of receivers from different vendors.
3. The proposed GNSS time synchronisation is validated by integrating a GNSS
receiver with on-board communicating devices and conducting experiments in
urban environments. The contributions include a demonstration of GNSS time
synchronisation with a working on-board node and a comparative study with
an existing TA-based synchronisation scheme.
4. GNSS availability has improved with recent advances of multi-GNSS constel-
lations and improvement of receiver technologies. Although the presence of
multiple satellite constellations increases the probability of having visible satel-
lites available, outages do occur. Time-synchronisation performance limitations
during GNSS time solution outages were analysed using data collected from ve-
hicles travelling along high-rise streets in the Brisbane CBD. This contributes to
knowledge of GNSS-based timing availability in signal-impaired environments
such as high-rise urban areas and tunnels.
1.5. DISSERTATION OUTLINE 13
1.5 Dissertation Outline
This dissertation consists of seven chapters. Among them, four chapters describe the
core contributions of this research; three chapters are set aside for the introduction,
background, and conclusion. The content of each chapter is summarised in the fol-
lowing.
Chapter 1. Introduction
In this chapter, the research domain, motivation for the research, and the research
objectives are discussed. The research contributions are briefly summarised, and the
organisation of this thesis is described.
Chapter 2. Fundamental Theories and Related Works
This chapter reviews the applicable literature and explains the fundamental princi-
ples associated with relevant GNSS services. In particular, it surveys the underlying
communication and time synchronisation techniques that exist in different wireless
communication technologies. Subsequently, the basic model and principles of GNSS
time synchronisation for vehicular networks are explained.
Chapter 3. Significance and Requirement Analysis of Time Synchronisation in
VANET
It is important to understand the needs for accurate time, and thus the accuracy
and precision of synchronisation requirements in vehicular networks. Therefore, this
chapter is dedicated to analysing the needs of time synchronisation in V2X environ-
ments and identifying the requirements specific to different vehicular communication
applications.
Chapter 4. GNSS Time Synchronisation in VANET
This chapter analyses the feasibility of GNSS time synchronisation within vehicular
networks. It also analyses the compatibility of GNSS signal in the vehicular envi-
ronment. Some experiments are described which determine the achievable timing
accuracy and signal availability using consumer-grade low-cost GNSS receivers in
14 CHAPTER 1. INTRODUCTION
road environments.
Chapter 5. GNSS Synchronisation with On-board Devices
A GNSS receiver integrated with an on-board device is tested in a series of labora-
tory and field experiments. The ensuing results are presented and analysed in this
chapter. This involved a comparison of an existing synchronisation scheme with the
developed GNSS-based solution.
Chapter 6. Synchronisation in Occasional loss of GNSS Signals
Satellite signals are weak in the earth surfaces and challenges to access GNSS signal
in certain areas that are under covers is one of the major issues of having GNSS
services to be used. In this chapter, some GNSS outage scenarios such as the possible
GNSS signal blocked road scenario where signals are impaired, or only available
intermittently are analysed with the aid of laboratory and field tests, for which some
timing solutions are proposed.
Chapter 7. Conclusion and Future Work
This chapter concludes by summarising the contributions to time synchronisation
in vehicular networks. Finally, some future-work directions along with potential
opportunities for improvement are suggested.
1.6. LIST OF PUBLICATIONS 15
1.6 List of Publications
(Journals)
1. K. F. Hasan, Y. Feng, and Y.-C. Tian ”GNSS Time Synchronisation in Vehicular
Ad-hoc Networks: Benefits and Feasibility,” in IEEE Transaction on Intelligent
Transportation System, 2018 DOI: 10.1109/TITS.2017.2789291.
2. K. F. Hasan, C. Wang, Y. Feng, and Y.-C. Tian ”Time Synchronisation in Vehicu-
lar Ad-hoc Networks: A Survey in Theory and Practice,” Manuscript Submitted
in the Journal of Vehicular Communication, Elsevier Publication, February 2018.
3. K. F. Hasan, Y. Feng, and Y.-C. Tian GNSS-driven Accurate Time Synchroniza-
tion for VANET, In Preparation to be submitted in IEEE Transaction on Vehicu-
lar Technology, April 2018.
(Conferences)
4. K. F. Hasan and Y. Feng, ”A Study on Consumer Grade GNSS Receiver for the
Time Synchronisation in VANET,” presented at the 23rd ITS World Congress,
Melbourne, Australia, 1014 October 2016.
5. K. F. Hasan, Y. Feng, and Y.-C. Tian ”Exploring the Potential and Feasibility of
Time Synchronisation using GNSS Receivers in Vehicle-to-Vehicle Communica-
tions,” In Proceedings of ITM 2018, Reston, VA, Jan 29-2 Feb, 2018.
6. K. F. Hasan and Y. Feng and Y.-C, Feasibility Studies of Time Synchronization
Using GNSS Receivers in Vehicle-to-Vehicle Communications, International Global
Navigation Satellite System (IGNSS)-2018, Sydney, Australia, 7-9 Feb 2018.
16 CHAPTER 1. INTRODUCTION
Chapter 2
Fundamental Theories and Related Works
Time synchronisation in networked systems aims to equalise local times of all net-
work nodes. It provides a common time frame among all nodes, thus supporting
various network functions such as message transmission, channel scheduling and
resource sharing in real-time and in the correct order. Time synchronisation is impor-
tant in network applications with high mobility such as in vehicular ad-hoc networks
(VANETs), in which there are unique requirements for delivery of critical warning
and service messages between nodes. VANETs differ from other mobile networks by
their ad-hoc architecture, high mobility, and time-sensitive applications. While many
synchronisation techniques have been developed for general networks, it is necessary
to understand the applicability of existing time synchronisation techniques in VANET
applications.
This chapter surveys the theory and practice of time synchronisation in VANETs.
Through a systematic approach, insights are developed into existing and emerging
protocols for time synchronisation in VANETs. This addresses the problem and prospects
of time synchronisation in vehicular networks.
17
18 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
Some parts of the content from this chapter has contributed to the following pub-
lications:
1. K. F. Hasan, C. Wang, Y. Feng, and Y.-C. Tian ”Time Synchronisation in Vehicu-
lar Ad-hoc Networks: A Survey in Theory and Practice,” Manuscript Submitted
in the Journal of Vehicular Communication, Elsevier Publication, February 2018.
2.1 Cooperative Intelligent Transportation System (C-ITS) Fundamentals
Cooperative Intelligent Transportation System (C-ITS) is an advanced wireless com-
munication technology and is a branch of Intelligent Transportation System (ITS).
In vehicular environments, the basic interactions of cooperative ITS are known as
Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), Infrastructure-to-Vehicle (I2V),
Infrastructure-to-Infrastructure (I2I) and Vehicle-to-Pedestrian (V2P) communication
as shown in Figure 2.1.
Vehicular communications and networks have significant potential to support
various applications associated with C-ITS goals [Karagiannis et al., 2011]. A wireless
V2V
V2I
V2P
I2I
Figure 2.1: Basic Interactions of Cooperative ITS.
2.2. ENABLING ARCHITECTURES, TECHNOLOGIES AND STANDARDISATION19
version of such network, named Vehicular Ad-Hoc Network (VANET), is a compo-
nent of C-ITS, which provides communications between vehicles and nearby road-
side equipment [Misra et al., 2009]. The IEEE has developed system architectures
in order to provide wireless access on roads, which are known as Wireless Access
in Vehicular Environments (WAVE). Collectively, the IEEE 802.11p and IEEE 1609.x
standards are called WAVE [Uzcategui and Acosta Marum, 2009]. The Dedicated
Short Range Communications (DSRC) for WAVE is sometimes referred as Cooper-
ative Intelligent Transportation System [Alexander et al., 2011]. Along with DSRC,
other candidate wireless technologies for VANET are cellular, satellite and WiMAX.
The Technical Committee of International Organisation for Standardisation (ISO) (204
Working Group) leads the development of the frameworks of C-ITS and is also known
as Communications Access for Land Mobiles (CALM) [Ansari, 2014, Emmelmann
et al., 2010].
2.2 Enabling Architectures, Technologies and Standardisation
2.2.1 Vehicular Ad-Hoc Network (VANET)
Vehicular Ad Hoc Networks (VANET) are a special kind of Mobile Ad-Hoc Network
(MANET). Communications in VANETs are governed by the protocol stack WAVE,
which conform to a series of IEEE standards. The WAVE controls the wireless medium
for VANET through a bundle of IEEE protocols, such as IEEE 1609 (e.g., 1609.2, 1609.3,
1609.4) and IEEE 802.11p. IEEE 802.11p is an amendment to IEEE 802.11 for regulation
of data link and physical layers.
The basic communication architecture in VANETs consists of two blocks: On Board
Units (OBU) and Road Side Units (RSU). An OBU is a vehicle, while an RSU refers to
the road side infrastructure for communications [Hartenstein and Laberteaux, 2010,
Joe and Ramakrishnan, 2016, Sharef et al., 2014]. This is illustrated in Figure 2.2.
20 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
RSU
RSU
Network Layer
Transport Layer
Application Layer
Data Link Layer
Physical Layer
RSU
OBU
Figure 2.2: VANET Communication Architecture.
Vehicle-to-Vehicle (V2V) communications are typical ad-hoc network communica-
tions. In V2V, mobile nodes communicate directly with each other. In such com-
munications, all types of packet deliveries, such as unicast, multicast and broadcast,
take place between vehicles without the intervention or support of any other network
components.
Vehicle-to-Infrastructure (V2I) communications are implemented through wireless
interactions between OBUs and RSUs. They enable real-time services, such as traffic
information and weather updates. V2I also provides support for secure sparse and
long-distance communications [Sharma and Tomar, 2016].
In addition, Internet infrastructure, private infrastructure and in-vehicle communica-
tions also support some VANET services and applications. This can assist with remote
2.2. ENABLING ARCHITECTURES, TECHNOLOGIES AND STANDARDISATION21
identification of vehicle’s performance and monitoring drivers’ conditions such as
fatigue and drowsiness. In-vehicle communications are considered as an architectural
part of the latest definition of VANET communications. It is a significant component
of safety and other applications in VANETs Alexander et al. [2011], Emmelmann et al.
[2010].
2.2.2 Dedicated Short Range Communication (DSRC)
The DSRC spectrum is specifically reserved for the radio operation of VANET. It
operates in the band of 5.9 GHz to support VANET. The Federal Communications
Commission (FCC) allocates the spectrum from 5.850 to 5.925 GHz for this operation.
This spectrum is divided into seven 10 MHz DSRC spectrum channels [Jiang et al.,
2006, Miao et al., 2012] as shown in following Figure 2.3. On-going research is con-
cerned with improving its efficiency and complying with safety requirements [Alam
et al., 2009, Bai and Krishnan, 2006, Bilstrup et al., 2009, Kenney, 2011, Miao et al.,
2012, Tang and Yip, 2010, Yi Qian et al., 2008, Yu and Biswas, 2007].
Figure 2.3: DSRC Channel Arrangement, Based on [Ansari, 2014]
2.2.3 Wireless Access in Vehicular Network (WAVE)
Wireless Access in Vehicular Networks (WAVE) is a system architecture that was de-
veloped by the IEEE by combining IEEE 802.11p and IEEE 1609 standards [Uzcategui
and Acosta Marum, 2009]. IEEE 802.11p is an amendment of IEEE 802.11 [Standards
Association et al., 2001] standard which is basically a set of media access control
(MAC) and physical layer (PHY) specifications for implementing wireless local area
22 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
network (WLAN) computer communications. The major targeted technological de-
velopment of this research work takes place in this architecture. The WAVE protocol
architecture with its components is shown in Figure 2.4.
WAVE SecurityServices
WAVEManagement
Entity
TCP/IP
IPv6
WSMF
MLME Extension
MLME
PLME
LLC
WAVE MAC Channel Coordinaiton
MAC
PHY
Management Plane Data Plane
Scope of IEEE 1609.2 WAVE
Scope of IEEE 16.09.4 WAVE
Scope of IEEE 1609.3 WAVE
Scope of IEEE 802.11 WAVE
Figure 2.4: WAVE Protocol Stack.
2.2.4 Overview of IEEE 802.11P
IEEE 802.11p employs the contention-based channel access also known as EDCA
(Enhanced Distributed Channel Access) as the MAC method, which is the basically
extended model of the basic Distributed Coordination Function (DCF) from 802.11
protocol stack. EDCA uses Carrier Sense Multiple Access (CSMA) with Collision
Avoidance (CSMA/CA). In this method, a node that is expected to transmit checks
the medium at the beginning of the communication, and it will able to proceed if it
received a free AIFS (Arbitration Inter-frame Space), otherwise, the node will defer
the transmission by a random back-off time[Cozzetti et al., 2009, Jiang and Delgrossi,
2008, Miao et al., 2012]. This is one of the specific points of interest of this research
2.2. ENABLING ARCHITECTURES, TECHNOLOGIES AND STANDARDISATION23
and a significantly alternative idea, called time slotted protocol will be proposed.
sectionFundamentals of Time Synchronization in Communication Networks
Clocks used in communication networks may be grouped into hardware and
software clocks. A physical clock made up of an oscillator to generate a pulse train
and a counter to count and store them is called a hardware clock. Hardware clocks
can be constructed from different materials ranging from most precise and expensive
caesium (i.e. atomic clocks) to inexpensive quartz-powered clocks. A software or log-
ical clock is a software-enabled programmable device that uses counting algorithms
to track a local time value and maintain the time base of the system. Essentially
in a standalone system, logical clocks follow a systems hardware clock, thus clock
accuracy depends on the performance of the hardware clock.
The quality of hardware clock, however, mainly depends on the stability of the os-
cillator as well as of the counting device. The stability is subject to changes of various
parameters such as the nominal frequency of the oscillator, temperature, and other
environmental factors. Such influences create a deviation in the device clock from the
actual time, which may be known as clock drift. Figure 2.5 shows the frequency of
an ideal clock that is theoretically considered as a constant over time. However, in
practice, the frequency changes due to both internal and external influences and drift
from its theoretical value. As a consequence, each local clock system deviates from
a more precise clock time and also from each other. The difference is known as time
offset. Therefore, in a communication network as shown in Figure 2.5 (b), all the node
clocks may report different times.
Operating communication networks requires alignment of node clocks to a ref-
erence clock, or synchronization of network time to a reference time. Fundamen-
tal operations may include successful communication, channel scheduling, real-time
control messages. Alignment refers to reducing the effects of clock offset and drift
between nodes to an acceptable level. A straightforward solution is to use an accurate
24 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
Ideal Clock Practical Clock
8:15:25
8:15:26 8:15:24 8:15:27
8:15:28Communication Network
Drift
Offset
(a) (b)
Figure 2.5: Clock in a communication network. (a) Ideal and Practical physical clockand their frequency. (b) State of clocks in an unsynchronised communication network.
source of time such as an atomic clock in every device of the network, which is
expensive and not realistic in most real network scenarios. Communication network
nodes are usually equipped with inexpensive quartz clocks. A method called clock
synchronization is used to equip all node clocks with the same time. The basic idea
is to minimize clock drifts and offsets resulting from various errors and inaccuracies.
This is achieved by communicating messages that help to transfer time from one node
to another. Figure 2.6 shows the fundamental concept of message transmission that
synchronize clocks node by node. Ideally, such messages can be transmitted from
the sender node to receiver node or back and forth between sender and receiver to
attain a common agreed time Freris et al. [2011], Kadowaki and Ishii [2015]. The
accuracy and precision of clock synchronization, therefore, depends on the accurate
transmission and reception of the messages. A number of synchronization protocols
Ideal Clock Practical Clock
8:15:25
8:15:25 8:15:25 8:15:25
8:15:25Synchronized Network
(a) (b)
One Way
Two Way
Time Sync
Figure 2.6: Clock in a communication network. (a) Communication between ideal(reference) clock and practical node clock. (b) State of clocks in an Synchronisedcommunication network.
2.3. BASIC MODELS AND TECHNIQUES OF TIME KEEPING AND TIMESYNCHRONIZATION 25
have been evolved targeting both wired and wireless networks over the time. In
all cases, they deal with the fundamental problems of measuring the variation in
sending and receiving time of messages, including access and propagation time over
the medium by comparing the timing information received from the nodes Lenzen
et al. [2015], Mills [2016]. The efficiency of synchronization protocols hence lies on
the ability to accurately predict and eliminate message transmission-related delays
by comparing their clocks.
Based on the above fundamental concepts, the next section canvasses technical details
and principles of time keeping and clock synchronization.
2.3 Basic Models and Techniques of Time Keeping and Time Synchroniza-
tion
This section presents general clock models and error sources that limit accurate time
keeping towards achieving a common notion of time in a communication network.
The levels of clock accuracy are also discussed together with basic techniques for
time synchronization in decentralized communication network systems.
2.3.1 Hardware Clocks
A hardware clock consists of a counter to count time ticks, which are ideally of a fixed
length. A hardware oscillator updates the counter at a constant rate, i.e., frequency.
The quality of the clock thus depends on the stability of the oscillator. Let the reading
of a clock counter t be denoted by C(t), the rate f (t) at time t is C(t), we have
f (t) = dC(t)/dt (2.1)
For an ideal clock, the rate is 1. However, a real clock fluctuates over time due to
the fact that the rate changes because of various limiting factors. In a typical node p
26 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
with a quartz crystal, whose nominal frequency is defined as f 0p , the relative frequency
deviation is
ρp(t) = fp(t)/ f 0p − 1 (2.2)
According to [Gaderer et al., 2006], a model for a real clock is expressed as
fp(t) = f 0p · [1 + ρi
p(t) + ρap(t) + ρn
p(t) + ρep(t)] (2.3)
where, ρip(t) is the initial frequency deviation at start-time, ρa
p(t) considers the aging
effect, ρnp(t) denotes the jitter due to short-term noise, and ρe
p(t) represents the jitter
due to environment changes.
The environmental jitter ρep(t) is a major factor influencing the quartz clock drift, in
which variation in temperature typically contribute the most. According to [Packard,
1997], a jitter in order of 10−6 to 10−5 could be introduced by temperature changes.
Other environmental influences in supply voltage and mechanical effects such as
shock and vibration, can cause fluctuations. The short-term noise is typically in the
order of 10−8 to 10−12. The aging effect ρap(t) is in an order of 10−7 per month. Overall,
the systematic deviation for the initial frequency while restarting an oscillator can
grow at an order of 10−5 [Armengaud et al., 2007].
2.3.2 Software Clocks
Software clocks or logical clocks are algorithms residing in programmable devices.
They take local clock value C(t) as input and convert it to time S(C(t), which all
programs use for time-dependent applications. This time S(C(t))) is the consequence
of time synchronization. The mathematical model of such a typical software clock is
S(C(t)) = t0 + C(t)− C(t0) (2.4)
2.3. BASIC MODELS AND TECHNIQUES OF TIME KEEPING AND TIMESYNCHRONIZATION 27
where t0 is the (correct) real time. Such a software clock runs with the speed of the
hardware clock.
2.3.3 Clock Accuracy and Precision
Clock accuracy and Clock Precision are two related yet different concepts. Clock ac-
curacy, denoted by α, refers to the degree of correctness of the clock time. In com-
parison, clock precision refers to the consistence of the clock time with some other
and/or standard clock. In synchronization nomenclature, the accuracy is the largest
or maximum acceptable clock offset between the node clock and the reference clock.
It is determined by measuring the mean of the error between the node and external
reference clock and usually represents as synchronization bias Mahmood et al. [2017].
The clock of a Node p can run with the accuracy α if the clock value Cp(t) is in an open
α-neighbourhood around the standard absolute time t in an observable period of T
Fan et al. [2004], Gaderer et al. [2006], Horauer and Holler [2002]. Thus,
|Cp(t)− t| ≤ α, ∀taT (2.5)
The clock precision β, however, is the measure of the standard deviation of the mean
clock error and quantifies the synchronization jitter. It is often also called instantaneous
precision, which is represents the boundary of the difference between two clocks p and
q, i.e.,
|Cp(t)− Cq(t)| ≤ β, ∀taT (2.6)
In internal synchronization environments, the clock precision is the maximum dif-
ference between two clocks. For external synchronization with a standard time, this
difference is the accuracy as expressed in Equation (2.5).
28 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
2.3.4 Clock Offset, Skew and Drift
A free-running clock is influenced by a number of factors as described in Equation 2.3.
It fluctuates and deviates from the actual time due to the clock drift. Theoretically,
software clocks are similar to hardware clocks, as software clock algorithms follow
the system clocks, which depend on hardware clocks. Thus, the accuracy of software
clocks relies on the accuracy of hardware clocks.
The accuracy of a clock as defined above pertains to the overall degree of clock
uncertainties relative to a reference standard time. Clock uncertainties can be further
described through Offset, Skew and Drift [Sundararaman et al., 2005]. Clock Offset is
defined as the time differences between a clock time and the standard true time. It is
|Cp(t) − t| for node Np. It is seen from Equation (2.5) that the clock accuracy is the
absolute value of the clock offset. The relative clock offset between two nodes Np and
Nq at time t is expressed as
Clock Offset = Cp(t)− Cq(t) (2.7)
Clock Skew is defined as the difference of the clock frequencies between a system clock
and a perfect clock. It is the first derivative of the clock offset with respect to the real
time t. The clock skew of a clock Cp relative to Cq at time t can be expressed as
Clock Skew = C′p(t)− C
′q(t) (2.8)
Clock Drift of a clock Cp is defined as the second derivative of the clock value with
respect to the real time t, i.e., C′′(t). Therefore, the relative clock drift between two
nodes Np and Nq is represented by
Clock Drift = C′′p(t)− C
′′q (t) (2.9)
Overall, the above three terms are frequently use to characterize the performances
2.3. BASIC MODELS AND TECHNIQUES OF TIME KEEPING AND TIMESYNCHRONIZATION 29
of a typical clock in a communication system.
2.3.5 Main Limiting Factors in Time Synchronization
The performance of time synchronization methods are affected by two main factors:
the inherent performance of the clock oscillators and how effectively a chosen syn-
chronization technique works between them. The systematic and random errors of
clock oscillators accumulate over time [Exel and Ring, 2014], which impact on the
synchronization accuracy.
Several issues in synchronization techniques can affect the clock synchronization.
The first issue is the capability of the technique to deal with the uncertainty of mes-
sage delay during radio communication. Other issues include Clock Adjustment Prin-
ciple and Timestamping Accuracy. The estimation of various latencies during Sending
time, Accessing time, Propagation time and Receiving time is crucial for adjusting clocks
precisely over a network. The clock adjustment performance is highly dependent on
the method and quality of the synchronization algorithm. Timestamping is a method
of adding time into the packet during the transmission and reception of a message.
In packet-based synchronization techniques, precise time-stamping is crucial. By
calculating the egress and ingress timestamps, the propagation delay is measured. It
is known that the accuracy of clock synchronization varies from one protocol layer to
another [Chen et al., 2015, Cooklev et al., 2007, Weibel and Bechaz, 2004]. This is due
to the uncertainty of inter-layer delays. Physical layer time-stamping is considered to
be the most accurate way so far. After receiving the timestamps, a node needs to run
an operation to adjust the clock. The performance of this adjustment operation also
determines the accuracy and quality of the synchronisation technique.
30 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
2.3.6 Basic Techniques for Time Synchronization in a Decentralized System
VANETs are decentralized systems. To explain the time synchronization mechanism
in a distributed or decentralized network, we consider a model involving two nodes
Np and Nq, as shown in Figure 2.7. When Node Np sends a message with its local time
stamp tpx to Node Nq (Figure 2.7(b)), Node Nq receives the signal at tq
y and updates its
time accordingly. This is known as Unidirectional Synchronization. In unidirectional
synchronization, the transmission delay is not considered. It suffers from a large
synchronization error. Therefore, a more complicated Round-Trip Synchronization tech-
nique is more acceptable. In this technique, Node Np sends message at tpx to node Nq
to ask for the timestamp tqy. After getting the response from node Nq, node Np per-
forms calculation to determine the round-trip time d = tpz − tp
x . This round-trip time
is basically the time interval of two-way message transmissions as shown in Figure
2.7(c). Then, it is used to improve the precision of the time synchronization between
the two nodes. The drawback of this synchronization method is the introduction of
message exchange overheads.
Np
qN
Np
qN
xpt z
d d‘ ‘’
pt d
t yq
pt x
pt y
Np
qNqt y
qt x
d
(a) (b) (c)
Figure 2.7: Message Exchanges between Two Nodes.
Another effective method, namely, packet-based clock synchronization is Reference
Broadcasting Synchronization (RBS). Its operation is shown in Figure 2.8. In RBS, a
beacon sends a synchronising message to all nodes. For example, in Figure 2.8, node
Nb is the beacon node. It sends beacon a message to nodes Np and Nq. The delay d′
for Np and delay d′′
for Nq are almost the same. After receiving the beacon signal,
Nq sends its time stamp tqx to node Np. Then, node Np calculates the time interval
2.3. BASIC MODELS AND TECHNIQUES OF TIME KEEPING AND TIMESYNCHRONIZATION 31
d = tpy − tp
x . The result is a measure of the time difference between nodes Np and Nq.
Np
qN
Nb
qN
pt y
(a) (b)
Nb
d
pNd‘
d ‘’ xpt
qt x
Figure 2.8: Reference Broadcasting Synchronisation (RBS).
2.3.7 Types of Clock Synchronization
Several parameters such as the source of the reference clock, the required accuracy of
the synchronization, the communication medium between nodes and the supported
applications can all impact on the method of clock synchronization used. Therefore,
depending on the variation of methods and their applications, clock synchronization
may classify differently.
For example, when any system maintains synchronization with a standard ref-
erence clock time, it is known as absolute clock synchronization. When nodes in a
network are synchronized with respect to each other’s time, the method is known
as relative clock synchronization Tian et al. [2008]. Again, based on the variation of
the synchronization protocols we can classify time synchronization differently. Some
developed protocols for time synchronization commonly differs from each other in
some aspects again sometimes resemble each other in some other aspects Sundarara-
man et al. [2005]. For example, consider deterministic and probabilistic clock syn-
chronization. Deterministic protocols stipulate a strict upper bound on the offset
error certainty compared to probabilistic synchronization where it uses less message
transfers and, therefore, less processing overhead Arvind [1994], PalChaudhuri et al.
[2003].
32 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
However, the most popular way of classifying clock synchronization methods is
based on the time references system that is used. According to the time scale, clock
synchronization in distributed network can be classified into two main types: syn-
chronization with internal time-scale and synchronization with external time-scale.
In ad-hoc like networks such as WSN, VANET, etc., time synchronization can be
implemented locally with an internally consistent time-scale. However, VANETs are
outdoor wireless ad-hoc networks, therefore, it is also possible to deploy global time
synchronization with an external time scale.
Synchronization methods with an internal time-scale is realized through a set of
operations and message exchanges between nodes. This requires estimating both
offset and skew of the local clocks relative to each other. Hence, synchronization with
internal time-scale maintains a relative time notion with respect to each other. Such
relative synchronization is the basis of most indoor networks such as indoor wireless
sensor networks and Wifi.
Synchronization methods with an external time-scale method are implemented
with respect to an absolute or external reference time standard, such as Coordinated
Universal Time (UTC). Such an external reference time-scale is usually transmitted
and distributed by using a global radio system. Typical global radio systems include
satellite-based Global Navigation Satellite System (GNSS) and the short-wave WWVB
station [Beehler, 1981, Groves, 2013].
The next sections examine the synchronization techniques currently practiced in
different wireless networks.
2.4 Approaches to the Time Synchronization in Wireless Media
Most of the time synchronization protocols for communication networks are applica-
ble in both wired and wireless media. For example, the well adopted NTP, which is
considered as the backbone of wired computer communication networks, appears to
2.4. APPROACHES TO THE TIME SYNCHRONIZATION IN WIRELESS MEDIA 33
be implementable in wireless media with certain accuracy Elson and Romer [2003],
Ganeriwal et al. [2003]. However, although the fundamental principles are similar,
performance improvements are expected in further technology evolutions.
This section explores the prominent synchronization techniques over Wireless
Sensor Network (WSN), which provides a basis for synchronization in vehicular wire-
less networks. This is followed by presentation of the challenging issues and require-
ments of time synchronization in vehicular networks.
2.4.1 Revisiting Time Synchronization in Wireless Sensor Networks
As a VANET is a special type of mobile wireless networks, it is worth examining
the existing synchronization techniques for other types of mobile ad-hoc networks.
The focus is on wireless sensor networks (WSNs), for which considerable research
effort has been directed to time synchronization. Five main WSN synchronization
techniques are to be discussed below: time-stamp synchronization (TSS), reference-
broadcast synchronization (RBS), lightweight time synchronization (LTS), a timing-
sync protocol for sensor networks (TPSN), and flooding time synchronization proto-
col (FTSP).
Time-stamp Synchronization (TSS) is a WSN time synchronization method based
on internal synchronization on demand [Romer, 2001]. TSS does not use specific
synchronization messages for time synchronization. Instead, it uses timestamps em-
bedded in other packets to perform synchronization post-facto. The time offset is
estimated through calculation of the round-trip delay between the transmitters and
receivers. For single-hop WSNs, the average uncertainty of TSS is recorded as 200 µs.
In multi-hop networks, the maximum uncertainty of 3 ms is achieved in 5 hops.
Reference-Broadcast Synchronization (RBS) uses beacon broadcast for time synchro-
nization. In RBS, any nodes in a basic single-hop network can send a beacon to broad-
cast its time reference. A node compares its local reference time with the reference
34 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
times received from other neighbor nodes and adjusts its clock accordingly. RBS
performs both offset and rate corrections when updating the clock. Making use of
physical layer broadcasts, it does not carry explicit time-stamps. This synchronization
is enacted for the whole network.
In a multi-hop network, all network nodes are grouped into clusters. In each
cluster, a single beacon is used to synchronize all nodes in the cluster. A gateway node
is used to transfer time-stamps from one cluster to another. This helps maintain the
same reference time to compute offset and rate corrections. RBS uses the last minute
time-stamps in order to reduce random hardware delay and access delay. Its average
uncertainty is measured as 11 µs in laboratory experiments with 30 broadcasts. For
multi-hop networks with n hops, the average error grows in O(√
n). While RBS
provides comparatively high accuracy, it is subject to excessive protocol overheads.
Lightweight Time Synchronization (LTS) aims to reduce the complexity of synchro-
nization overhead [van Greunen and Rabaey, 2003]. Therefore, unlike other synchro-
nization methods, it provides with a specified precision. As a centralized algorithm,
it begins with the construction of a spanning tree for the network with n nodes. Next,
a pair-wise synchronization is performed along the n− 1 edges of the spanning tree.
The root of the spanning tree works as the reference node. It initiates all on-demand
resynchronization operations. The average synchronization error in LTS is recorded
as 0.4 s. The maximum error can reach as high as 0.5 s.
Timing-Sync Protocol for Sensor Networks (TPSN) is a network-wide synchroniza-
tion protocol based on a hierarchical approach [Ganeriwal et al., 2003]. It follows the
classical approach of sender-receiver synchronization to create a hierarchical topol-
ogy. The hierarchy maintains multiple levels in order to distinguish nodes to perform
actions. TPSN performs time synchronization through two phases. In the first phase,
a node at level 0 acts as the root node. It initiates a ‘level discovery’ broadcast message
with its identity and level in the hierarchy. Its immediate neighbors receive this mes-
sage and assign themselves level 1 below the root node. After that, each node at level
2.4. APPROACHES TO THE TIME SYNCHRONIZATION IN WIRELESS MEDIA 35
1 broadcasts a ‘level discovery’ message, which will be received by other neighbor
nodes at lower levels. This process continues until all nodes are reached by such
‘level discovery’ messages. In the second phase, all nodes synchronize their clocks to
their root or parent nodes in the tree by using a round-trip synchronization operation.
This round-trip synchronization is conducted at the MAC layer. Therefore, message-
delay uncertainties are largely eliminated. The accuracy of TPSN is considerably
high. Experimental results show that TPSN synchronization of two Berkeley motes
has reached an accuracy of 17 µs. A drawback of TPSN is the significant message
exchange overhead particularly for a large number of nodes.
Flooding Time-Synchronization Protocol (FTSP) is a hybrid time synchronization pro-
tocol built upon RBS and TPSN [Maroti et al., 2004]. In FTSP, a node with the lowest
node identity becomes the root node, which works as the reference time sender. If
this node fails, a node with the next lowest identity becomes the root node. The
root periodically floods the network with synchronization message with the reference
time. In this way, the whole network becomes synchronized. FTSP is a self-organized
algorithm. It constructs a hierarchy to perform low-level time stamping and local
clock correction. A FTSP experiment with an eight-by-eight grid of Berkley motes
shows an average error of 1.7 µs with the maximum of 38 µs per hop.
2.4.2 Why is Time Synchronization an Issue in VANETs
Time synchronization in distributed network systems is a well-recognized problem.
In wireless communication networks, time synchronization is considered as a key
element for consistent data traffic and also for accurate real-time control of message
exchanges Ghosh et al. [2011]. Many network applications require precise clock syn-
chronization among the nodes to ensure correctly ordered operations. Otherwise,
the performance of these applications and hence the network operations could be
disrupted. Over the years, the issue of time synchronization has been extensively in-
vestigated in the context of computer and telecommunication networks Bregni [2002],
36 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
Johannessen et al. [2001], Mills [1997]. Many protocols have been proposed and im-
plemented to perform time synchronization over computer and telecommunication
networks. Those protocols vary in terms of the required precision of timing and also
according to the services and types of networks. For example, in a routed network,
physical time is not a critical issue. Thus, protocols based on a routed network,
i.e., Packet-over-SONET/SDH links (POS), requires synchronization to ensure the se-
quence of the order. One the other hand, some networks require synchronization with
high time accuracy Cozzetti et al. [2011], ETSI [1 12], Scopigno and Cozzetti [2009].
For example, in pure synchronous optical networking (SONET) and synchronous
digital hierarchy (SDH) networks, the precision of time along with fixed time-division
multiplexing mechanism is mandatory.
In VANET, physical time plays an important role in many applications, which
cannot be satisfied by logical time or any kind of event ordering models. Most com-
municating interactions for time-based decisions rely upon a time-of-day clock. For
example, VANET enables traffic management on individual levels by providing com-
munication among vehicular nodes and share road information such as vehicle dy-
namics, driving intentions etc Cunha et al. [2016], Dua et al. [2014], Englund et al.
[2015]. The current status of the nodes in a VANET, therefore, needs to be determined
precisely in terms of position, speed and other real-time values. This frequently
scheduled work requires time synchronization to develop accurate and precise time
on node.
VANETs increase road safety by enabling different critical safety applications. For-
ward Collision Warning (FCW), Cooperative Collision Warning (CCW), Emergency
Electronic Brake Lights (EEBL) are a few examples that alert a driver about possible
crash scenarios ahead. In these applications, each vehicle is required to broadcast
their basic safety messages (BSM) that include vehicle location data periodically at 10
Hz. Event trigging warning messages are time sensitive and need to be transmitted
2.5. EXISTING RECOMMENDATION FOR VANET TIME SYNCHRONIZATION 37
and received orderly securing stringent delay requirements (typically 100 ms Harten-
stein and Laberteaux [2010]). If nodes clock in VANET does not have any commonly
agreed accurate time maintained among them, such periodical and event trigging
safety message from the sender may report with a past timestamped information
or with advanced timestamped information with respect to the receiver time and in
either case, those messages may be discarded after reception by the receiver nodes
considering as an outdated message. Under such circumstances, a warning message
would fail to alert drivers, thus leading to a risk of collision and other road casualties.
Time synchronization in VANET, therefore, is essential to achieve accurate and precise
time over the network Hussein et al. [2017].
Physical time is also crucial for proper bandwidth utilization and efficient channel
scheduling. Therefore, it is required that all the nodes in a VANET are able to report
the same time, regardless of the errors of their clocks or the network latency the
network nodes may have.
Furthermore, certain security measures in VANETs, such as duplication detection
and identification of session hijacking and jamming, require absolute time synchro-
nization Engoulou et al. [2014], Loo et al. [2016]. Time plays a critical role in deter-
mination of two distinct real-world events to develop traceable communication for
reconstruction of packet sequence on the channel Ben-El Kezadri and Pau [2010].
2.5 Existing Recommendation for VANET Time Synchronization
Time synchronization for VANET has been solely based on protocol IEEE 802.11p.
IEEE 802.11p is an amendment to Wireless Local Area Network (WLAN) protocol
IEEE 802.11. Therefore, the synchronization technique from IEEE 802.11 family is
naturally applicable to VANETs. In IEEE 802.11 standard family, a station (STA) can
be attached to an Access Point (AP) in a centralised mode called Basic Service Set
(BSS). It can also communicate with other STAs in decentralised ad-hoc mode called
38 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
Independent BSS (IBSS). These two modes are shown in Figure 2.9.
STA
STA
STA
STA
STASTASTA
AP
AP
STA
STA
(a) (b)
Figure 2.9: Two Modes of communications in 802.11 Standard Family.
In 802.11 networks, time synchronization is predominantly required for frequency
hopping and scheduling of sleep phases. The standard requirement of time syn-
chronization is 274 µs, which is also the threshold of out-of-synchronization [PIS-
CATAWAY, 1996]. Time synchronization within 802.11 systems rely on a Timing
Synchronization Function (TSF) timer. The TSF timer is a 64-bit hardware counter
with a resolution of 1 µs and thus is capable of performing 264 modulus counting.
It employs a local clock oscillator built on WLAN chipset with a frequency accuracy
of ±0.01%. The adjustment of the timer and hence the accuracy of synchronization
depends on the operation mode, e.g., BSS mode in centralised communications or
IBSS mode in a decentralised network. In BSS, the AP transmits beacons with TSF
timer values, and an STA sets its own TSF timer with delay usually corrected by offset
adjustment without rate correction.
In the infrastructure-based BSS mode, an AP acts as a master clock. It broadcasts
the reference time for all STAs to be time synchronized. When beacon transmits,
other data exchange operations are suspended so that the master can broadcast TSF
synchronization values to all attached STAs. The period of beacon transmission de-
pends on the network resource sharing mode as shown in the Table 2.1. In this mode,
the receiving station only accepts TSF values and updates its clock.
In the ad-hoc IBSS mode, all STAs adopt a common value, aBeaconPeriod, which
2.5. EXISTING RECOMMENDATION FOR VANET TIME SYNCHRONIZATION 39
Table 2.1: Slot Time with Beacon Generation Window.
FHSS DSSS OFDM
aCWmin 15 31 15aSlotTime (µs) 50 20 9Speed (Mbps) 1 2 1 2 5.5 11 6 12 24 54Beacon length 13 8 34 22 14 12 12 8 5 4
characterises the length of a beacon interval. At the beginning of the beacon interval,
a beacon generation window forms. It consists of ω + 1 as shown in Figure 2.10. For
the station that initiates IBSS, this interval also defines Target Beacon Transmission
Times (TBTTs) in aBeaconPeriod times apart. A time zero is defined to be a TBTT. At
the TBTT event, all STAs perform the following process:
1. At TBTT, suspend the backoff timer for any pending non-beacon transmission.
The STA calculates a random delay distributed in the range [0,ω), where ω =
2*aCWmin*aSlotTime.
2. All STAs wait for the period of the random delay.
3. If the beacon is received before the expiration of the random delay timer, cancel
the remaining random delay.
4. When the random delay timer expires, STA sends beacon using its TSF timer
value as a timestamp.
5. When a station receives the beacon, it updates its TSF timer following the times-
tamp of the beacon if the beacon value is later than the station’s TSF timer.
Therefore, the TSF synchronizes timers with the fastest STA in IBSS.
The above synchronization procedure is designed for single-hop networks. Such
a procedure with TSF suffers from poor scalability and inability to handle conges-
tions. When the number of nodes increases, the node with the fastest clock faces
difficulties in successfully sending out beacon frames. As a result, its clock gradually
40 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
Beacon Interval
Beacon Generation Window(W+1 slots)
Figure 2.10: Beacon Generation Window.
drifts away from the clocks of other nodes. This problem is known as fastest node
asynchronism [Huang and Lai, 2002].
In a multi-hop network employing the native IEEE 802.11 clock synchronization
mechanism, the whole network is partitioned into multiple disjoint clock islands. If
every island is out of synchronization with one another, the time partitioning problem
appears [So and Vaidya, 2004].
Improved techniques have been proposed to address the so-called fastest node
asynchronism problem and the time partitioning problem. The basic ideas are to
enhance scalability and mitigate congestion. Two well-recognised improvements are
Adaptive TSF (ATSF) and Multi-hop TSF (MTSF).
ATSF modifies the basic 802.11 TSF. It adds a priority scheme to overcome the
fastest node asynchronism problem. This method involves maintaining and adjusting
the transmission frequency of the beacon [Huang and Lai, 2002]. When a node re-
ceives a beacon message with a larger timestamp, it reduces its beacon transmission
frequency. It keeps updating the beacon transmission frequency until it reaches the
maximum allowed value. This allows the fastest node to have a higher probability of
transmitting beacon messages.
In MTSF, each node maintains a path to the fastest node. The beacon is transmit-
ted from the fastest node to all other nodes without being suppressed anywhere in
the middle of the network. MTSF consists of two phases: a beacon window phase
and a synchronization phase [Chen et al., 2006]. In the beacon window phase, all
2.5. EXISTING RECOMMENDATION FOR VANET TIME SYNCHRONIZATION 41
neighbour nodes construct a synchronization group and identify the fastest node as
the root node of the group. In the synchronization phase, root nodes are synchronized
with each other. In this way, the fastest node asynchronism problem can be avoided
together with the partitioning problem.
The average maximum clock drift with TSF is 124.5 µs for 20 nodes. It increases
to 500.2 µs when the number of nodes is 60. In comparison, MTSP performs much
better. Experimental measurements show that the average clock accuracy of MTSF is
22.4 µs for 20 nodes and 39.1 µs for 60 nodes, respectively [Cheng et al., 2006].
Such TSF-based synchronization lacks support from timing standards such as
UTC, TAI etc. The 2012 amendment of IEEE 802.11 proposes two techniques, i.e.,
Timing Advertisement (TA) and Timing Measurement (TM) mechanisms, to obtain
the support of global time [IEEE, 2011]. In TA, the external reference clocks are
attached to Access Points (APs). In TM, the frames use physical layer timestamps
to perform synchronization between AP and STA, thus reduces multi-hop errors.
However, the TA mechanism architecture requires a cascade of four clocks, which
does not direct how the external clock will be synchronized to the AP and perform
accurate time-stamping. Reference [Mahmood et al., 2015] and [Mahmood et al.,
2017] have discussed details on that issue and proposed some measures in WLAN
scenarios. VANET networks are more ad-hoc in nature compared to WLAN, where a
large portion of it relies on STA to STA but STA to AP communication. Therefore, the
feasibility of employing such a mechanism in VANET requires an extensive investi-
gation.
The next sections highlight the feasibility of GNSS-based time synchronisation in
VANET.
42 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
2.6 GNSS Approaches for VANET Time Synchronization
Global Navigation Satellite Systems (GNSS) are a well-established international util-
ity for positioning, navigation and timing (PNT). The generic term ”GNSS” refers to
the USA’s Global Positioning System (GPS), Russia’s GLONASS, Europe’s GALILEO
and China’s BEIDOU navigation satellite system (BDS). It is noted that GPS, GLONASS,
Galileo and BDS use different reference time systems creating time offsets between
them. However, the offsets can be determined at the system level or user level. Any
one or more constellations can offer the same global standard UTC time. With their
worldwide coverage, continuous service, GNSS has become one of the most efficient
and standard systems for time dissemination in many applications. Many industries
such as energy, meteorology and telecommunications rely on GNSS for accurate time
synchronization in their systems and devices. The accuracy achieved by GNSS-based
time synchronization using GPS is better than 40 ns 95% of time ?. This can meet the
most restrict requirements for VANET time synchronization.
This section begins with discussing the motivation for using GNSS for VANET
time synchronization. This is followed by descriptions of GNSS models and methods
for time transfer. The challenges and solutions due to absence of GNSS signals in
vehicular environments are then summarised.
2.6.1 Motivation of GNSS-driven Time Synchronization in VANETs
Most of the earth-based time transfer techniques suffer from path delay measurement
uncertainties. In contrast, the satellite-based GNSS time transfer systems possess
measurable constant path delays. This arises because the variation of path delays
are small and due to clear, unobstructed paths to receivers. Therefore, the delay
measurements are straightforward and can be more easily calibrated compared to
any ground-based systems. In addition, the radio interferences due to weather or any
other ground-based noise have less impact in satellite-based GNSS systems.
2.6. GNSS APPROACHES FOR VANET TIME SYNCHRONIZATION 43
In telecommunication networks, GNSS is used to synchronize some major nodes
called root or server nodes outdoors. Through these root nodes, other nodes in
the system are synchronized by using other synchronization techniques, which are
mostly based on message transfer between nodes.
In contrast to telecommunication networks, VANETs are outdoor-based networks.
Except in some tunnels and blocked roads, nodes in VANETs on the road are mostly
under the coverage of GNSS signals. It is a straightforward choice for VANETs to
use GNSS for synchronization. GNSS receivers have already been used for vehicle
navigation and positioning. Nowadays, multi-GNSS constellations, more precise
GNSS services, such as space-based argumentation systems (SBAS), differential GNSS
(DGNSS) services and precise point positioning (PPP), are available for VANET de-
ployments. The GNSS-based time synchronization is indeed plausible in VANETs.
It is therefore prudent to understand how GNSS time solutions provide synchro-
nization in VANETs and what the possible solutions are when GNSS services are
absent, such as when vehicles travel in tunnels. The feasibility and accuracy of GNSS
time solutions are referred to a recent work by Hasan et al in Hasan et al. [2018a,b].
2.6.2 GNSS Time Synchronization Models for VANET
Different GNSS systems follow the same estimation principle for position, velocity
and time computing. Without loss of generality, this section discusses the theory of
GPS time, time transfer from GPS, and time propagation. It explains time synchro-
nization model by using GPS data. It also outlines possible support of synchroniza-
tion in the absence of GPS signals.
2.6.2.1 GPS Time
GPS time is one of the standard times related to UTC. It is a continuous time generated
from a precise atomic clock and maintained by some control segments. GPS time is
44 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
related to UTC by leap seconds. At present, GPS time is 18 s ahead of UTC time.
This means that the leap seconds between GPS time and UTC time are 18s. This is
indicated in USNO navy’s website tycho.usno.navy.mil/leapsec.html.
2.6.2.2 GPS Time Receiver and Time Transfer
There are a variety of GPS receivers which differ within applications, technologies
and manufacturers. Most consumer-grade GPS receivers receive single-frequency
C/A code. The clocks in GPS receivers are mostly quartz clocks. They are synchro-
nized by GPS signals. The GPS receiver clock solution is obtained from the pseudo-
range measurements. For the purpose of this work, the pseudo-range measurement
can be written as [Petovello, 2011]:
Pu = ρ + cdt + ε (2.10)
where Pu is the Pseudo-range measurement, ρ is the geometric distance between the
receiver and the satellite, c is the speed of light, dt is the receiver time offset or bias
with respect to the receive clock time tag, and ε is the sum of all errors.
It is clear from the Equation 2.10 that dt can be directly obtained from the observed-
computed (O-C) difference (Pu- ρ) if the distance is known. Taking average or weighted
average over all the O-C differences would improve the accuracy of the dt solutions.
This is the static mode for time transfer. However in a VANET, the vehicle nodes are
moving. The distances ρ are computed with the approximate vehicle states X0. The
coordinate biases of X0 can affect the accuracy of dt solution if ignored. Considering
the coordinate biases, Equation 2.10 can be rewritten as:
Pu − ρ(X0) =∂ρ0
∂X0dX + cdt + ε (2.11)
where the partial derivatives of the geometric distance ρ0 are computed with respect
2.6. GNSS APPROACHES FOR VANET TIME SYNCHRONIZATION 45
to the 3-dimensional approximate coordinate vector X0; dX is the 3-dimensional po-
sition deviation with respect to the approximate states X0. They can be estimated
along with the clock bias dt. The least square or weighted least square procedures are
usually applied to solve the estimation problem with four or more satellites in view
[Misra and Enge, 2006].
Several techniques have been developed to transfer GPS time based on the above
equations. The simplest method is ‘time dissemination’, which is also known as ‘One
Way’ method. It predominantly aims to synchronize an on-time pulse, or to calibrate a
clock frequency source. Figure 2.11(a) illustrates the one-way concept where the clock
bias is determined by the difference between observed range Pu and computed range
ρ, namely (O-C). With more satellites in view as shown in 2.11(b), the clock bias can
be estimated from the average of all the (O-C)s. The user-position biases will affect
the clock bias dt solutions. As long as four or more satellites in view the coordinate
bias vector dX and clock bias dt can be determined with the linear Equation 2.11. A
typical clock solution accuracy with GPS-only signals is 40 ns [Hasan et al., 2018b].
A more accurate and elegant technique for GPS time transfer is ‘Common View’. As
shown in Figure 2.11(c) , it measures the clock bias difference between two receiver
oscillators using the difference of the (O-C)s between two receivers. This differencing
leads to the cancellation of satellite orbit and clock error and local ionosphere and
troposphere delays, thus providing a higher accuracy for the clock offset, saying in
the level of 10 ns. Similarly when multiple satellites are in view as shown in 2.11(d),
the common view method is equivalent to differential GPS, and determines the 3D
coordinate offsets and clock offset between two receivers. There is a highly accurate
technique for GPS time transfer called the ’Carrier-Phase’ method. In this method,
both L1 and L2 carrier phase signals are used to calculate time [Parker and Matsakis,
2004]. The timing accuracy achieved from this method is in sub-nanosecond level.
However, dual-frequency phase receivers are more costly and may be not a popular
choice for vehicle users.
46 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
(a) (b)
(c) (d)
A B A B
Figure 2.11: GPS Time Transfer.
2.6.2.3 A Simple Model for GNSS Time Synchronization
Based on the above ”one-way” and ”common-view” modes, the GNSS time synchro-
nization model is outlined as follows. An on-board GPS receiver tracks satellites.
Once the clock bias dt is obtained with a one-way time transfer, the receiver can
determine its UTC time tUTC,
tUTC = tu − dt− dtUTC (2.12)
where tu is the receiver time, which is usually the time tag of a standard time epoch;
dtUTC is the offset between GPS time and UTC time, which includes a integer term
for leap seconds and a fractional correction term calculated from GPS navigation
messages. Both are the same for different receivers at the same time. In the common
view mode, receiver B obtains the clock bias with respect receiver A, i.e., dtBA, the
2.6. GNSS APPROACHES FOR VANET TIME SYNCHRONIZATION 47
receiver B’s UTC time is obtained as follows:
tBUTC = tAUTC − dtBA (2.13)
A typical GNSS receiver has an internal quartz-based oscillator that continuously
runs and follows GPS time. Generally, in the one-way time transfer, the clock update
rate can be the same as the receiver sample rate. A low-end receiver updates its
outputs at 5 to 10 Hz, while a high-end geodetic receivers sample rates can be up
to 50 Hz. However, such quartz clocks still exhibits deviations because the frequency
of each clock is different and tend to diverge from each other. This divergence is
known as clock skew. The clock drifts with respect to time is the derivative of clock
skew [Sundararaman et al., 2005]. Following [Sichitiu and Veerarittiphan, 2003] and
[Levesque and Tipper, 2016], in general, a node clock in a GNSS-synchronized dis-
tributed network over a time interval of minutes to hours can be characterised as as:
Ci(t) = di.t + bi (2.14)
where t is the time corresponding to the UTC time. di is the clock drift due to the
oscillator’s frequency differences and the result of to the environmental changes at
the node, e.g., variations in temperature, pressure and power supply voltage. bi is the
offset between the receiver local clock and the UTC time obtained from one-way time
transfer approach. This reflects the effect of hardware delays of the clock.
Any two such GNSS-synchronized clocks can be represented as:
C1(t) = d1.t + b1
C2(t) = d2.t + b2
(2.15)
48 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
They can also be related as follows:
C1(t) = Θ12C2(t) + β12 (2.16)
where Θ12 is the relative drift between two receivers and b12 is the offsets due to
the bias variations. If the two receivers are the same model, the relative offset can be
small and the drift Θ12 is 1.
To measure the time offset between two GNSS powered nodes, experiments have
been performed at physical layer on 1PPS signals generated by two GPS receivers.
The experimental results are presented in [Hasan and Feng, 2016, Hasan et al., 2018b].
It has been shown that two same-model consumer-grade GNSS receivers are capable
of synchronizing network nodes with nano-second scale timing accuracy.
2.6.3 Challenges and Solutions in Absence of GNSS Signals
Signal transmissions between satellites and receivers solely relies on the principle
of the Line-of-Sight (LOS) wave propagation technique. Drivers often experience
outages of navigations when driving through high-rise streets. This does not neces-
sarily mean loss of time synchronization. First, GNSS time solutions can be obtained
with a single satellite at reduced accuracy. The availability of valid time solutions is
much higher than the availability of valid position solution. For instance, a vehicle
experiment shows that the percentage of valid GPS+Beidou position solutions over
some Brisbane high-rise streets is 99.25%, while the percentage of a minimum of one
satellite is 100% Hasan et al. [2018b].
Secondly, some measures have been proposed as fall back solutions in the block-
ages of GPS signals. One of them is to switch from normal mode to holdover mode
using GPS Disciplined oscillator (GPSDO). GPSDO is a specially made firmware for
holdover mood. It enables the internal oscillator to predict and imitate the original
timing and frequency of the GNSS system. A GPSDO is primarily made of a phase
2.6. GNSS APPROACHES FOR VANET TIME SYNCHRONIZATION 49
detector and voltage control oscillator (VCO). Its fundamental purpose is to acquire
information from the GNSS signal of satellites to control the frequency of local quartz
or rubidium oscillators. When GPS signals are unavailable, GPSDO keeps its oscilla-
tion in a stable frequency using the knowledge of its past performance.
In order to boost the performance of GPSDO, some additional technique have
been developed such as adaptive temperature and aging compensation during the
holdover period Penrod [1996]. The adaptive temperature and aging compensation
are based on a recursive implementation of linear regression. As an additional cir-
cuitry, a simple semiconductor ambient temperature sensor and an A/D converter are
used. The performance of both types improved GPSDOs are the same to some extent.
However, it is not well defined how long the independent free-running GPSDOs are
executed. Nevertheless, experiments have been conducted to test the performance
of GPSDO. An experiment was carried out over a week for holdover on 4 GPSDOs,
in which an oscillator is made of quartz and the other three are made of rubidium
Elson et al. [2002], ETSI [1 12]. After a week , the time offset from the quartz oscillator
was shown to be 82 µs. In comparison, the best time offset performance of less than
3 µs was measured for the three rubidium oscillators. This level of synchronization
accuracy is considered to be acceptable for VANET time synchronization applications.
Finally, the problem of GPS signal blockages can be addressed by incorporating
GPS synchronization with other methods. If some vehicles nodes that can view satel-
lites have GNSS time solutions, they act as root servers for synchronization of other
nodes through a non-GPS time synchronization technique. NTP-GPS is the back-
bone of general computer networks, in which the standard time hosting servers are
synchronized with GPS. For example, time synchronization based on absolute GPS
is employed in Automatic Identification System (AIS) for ships Tetreault [2005]. In
DSRC-based networks, Time advertisement (TA) has been specified in the IEEE1609.4
to provide time solutions to other devices where GNSS signals are not available.
However, to date the performance of TA for VANET is not well understood.
50 CHAPTER 2. FUNDAMENTAL THEORIES AND RELATED WORKS
2.7 Summary
Communications in VANETs involve V2V and V2I communications. They form the
basis of VANETs for network connectivity and various road safety applications. Due
to the highly dynamic and mobile characteristics, precise timing and accurate mea-
surement of transmission delay become critical in VANETs. Time synchronisation
helps establish an agreed time over VANETs. It enables proper coordination and
consistency of various events throughout the networks. It also allows accurate se-
quencing and real-time control of message exchanges over the networks.
This chapter has discussed why time synchronisation is necessary for VANETs.
It has also discussed why most existing synchronisation techniques used for other
types of wireless networks are not directly satisfactory in VANETs. The discussions
are accompanied by detailed evaluations of existing time synchronisation protocols
in various distributed network systems.
Time synchronisation is a challenge in VANETs. Under certain road conditions,
VANETs require a high accuracy in time synchronisation. Some security measures
also need precise time synchronisation, which is currently not achievable in VANET
environments. Synchronisation techniques developed for general WSNs face com-
patibility issues when applied to VANETs. GNSS-driven time synchronisation is a
promising technique for VANETs.
Chapter 3
Significance and Requirement Analysis for Time
Synchronisation in VANET
Time synchronisation ensures that all nodes in a network have the same reference
clock time. The clock of any network node is imperfect thus if all node clocks agree
on a reference time at a given time, they are not capable of keeping the track over
time. Maintaining a common notion of time is necessary for VANETs for various
applications and also for the functioning of many system-level protocols. That is,
synchronising the time throughout the network is needed periodically. So far, there
has been limited work [Cozzetti et al., 2011, Morgan, 2010] on time synchronisation
and stringent timing requirements for VANET. The concepts and techniques of time,
time quality and time synchronisation in VANET have been directly adopted from
Wireless Local Area Network (WLAN) standards following IEEE 802.11p. However, a
WLAN is an infrastructure-centred asynchronous network, in which communications
are implemented among low-mobility wireless nodes through Access Points (APs).
As a WLAN is not time critical, precise time synchronisation is not required nor is it
51
52CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
really achieved. In compared with WLANs, VANETs consist of both infrastructure-
based vehicle-to-infrastructure (V2I) and ad-hoc vehicle-to-vehicle (V2V) communi-
cations. The integration of V2I and V2V make VANET more challenging and time
critical than WLAN.
Time synchronisation in VANET depends not only on the synchronisation ac-
curacy but also on performance, compatibility and feasibility issues. The primary
objective of this chapter is to provide an in-depth analysis of time synchronisation
requirements in VANETs. It also discusses the importance of time synchronisation in
vehicular networking for different applications.
A general description of different time synchronisation parameters and VANET-
specific requirements is introduced below. The importance of time synchronisation in
different VANET applications along with their requirements is also discussed.
Some of the content from this chapter has contributed to the following publica-
tions:
1. K. F. Hasan, Y. Feng, and Y.-C. Tian ”GNSS Time Synchronisation in Vehicular
Ad-hoc Networks: Benefits and Feasibility,” in IEEE Transaction on Intelligent
Transportation System, 2018 DOI: 10.1109/TITS.2017.2789291.
2. K. F. Hasan, C. Wang, Y. Feng, and Y.-C. Tian ”Time Synchronisation in Vehicu-
lar Ad-hoc Networks: A Survey in Theory and Practice,” Manuscript Submitted
in the Journal of Vehicular Communication, Elsevier Publication, February 2018.
3.1 The Need for Time Synchronisation in VANETs
Time synchronisation in distributed network systems is a well-recognised problem.
In wireless communication networks, time synchronisation is considered as a key
3.1. THE NEED FOR TIME SYNCHRONISATION IN VANETS 53
element for consistent data traffic and also for accurate real-time control of message
exchanges [Ghosh et al., 2011]. Many network applications require precise clock
synchronisation among the nodes to ensure correctly ordered operations. Otherwise,
the performance of these networks and applications are subject to disruptions. Over
the years, the issue of time synchronisation has been extensively investigated in the
context of computer and telecom networks [Bregni, 2002, Johannessen et al., 2001,
Mills, 1997]. Many protocols have been proposed and implemented to perform time
synchronisation over computer and telecommunication networks. Those protocols
vary in terms of the required precision of timing and also according to the services and
types of networks. For example, in a routed network, physical time is not a critical is-
sue. Thus, protocols based on a routed network, i.e., Packet-over-SONET/SDH links
(POS), requires synchronisation to ensure correct sequence order. One the other hand,
some networks require synchronisation with high-accuracy timing [Cozzetti et al.,
2011, ETSI, 1 12, Scopigno and Cozzetti, 2009]. For example, in pure SONET/SDH
networks, the precision of time along with fixed time-division multiplexing mecha-
nism is mandatory.
In VANET, physical time plays an important role in many applications, which
cannot be satisfied by logical time or any kind of event ordering models. Most com-
municating interactions for time-based decisions rely upon a time-of-day clock. For
example, VANETs enable traffic management on individual levels by providing com-
munication between vehicular nodes and sharing road information such as vehicle
dynamics, driving intentions etc. [Cunha et al., 2016, Englund et al., 2015]. The
current status of VANET nodes, including position, speed and time, is important
and needs to be determined precisely. That is, accurate frequently scheduled work
requires time synchronisation to develop accurate and precise time on a node.
VANETs can support increased road safety by enabling different safety-critical
applications. For example Forward Collision Warning (FCW), Cooperative Collision
Warning (CCW), Emergency Electronic Brake Lights (EEBL) can alert a driver about
54CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
possible crash scenarios ahead. The underlying communication messages are time
sensitive and need to be transmitted and received securely according to stringent
delay requirements (typically 100 ms [Hartenstein and Laberteaux, 2010]). If the
VANET node clocks do not maintain commonly agreed times, critical safety messages
may be accorded incorrect timestamped information or with advanced timestamped
information with respect to the receiver time and in either case, those messages may
be and erroneously discarded after reception by the receiver nodes considering as an
outdated message. Under such circumstances, a warning message would fail to alert
drivers, which may lead to collisions and casualties. Time synchronisation in VANET
is therefore essential to achieve accurate and precise time over the network [Hussein
et al., 2017].
Physical time is also crucial for proper bandwidth utilisation and efficient channel
scheduling. Therefore, it is required that all the nodes in a VANET are able to report
the same time, regardless of the possible impreciseness of their clocks or any network
latency that may be present.
3.2 Time Synchronisation Requirements in VANET
VANET is a real-time communication technology having hard and strict time bound-
aries for end-to-end transmission latency. Some applications characterise the time-
sensitive nature of VANET. First of all, VANETs are very dynamic with vehicles com-
ing in and moving out, making relative time synchronisation more difficult to achieve
and maintain in compared to other fixed networks. Secondly, for vehicle safety ap-
plications, vehicle location and velocity data, as part of Basic Safety Messages (BSM),
need to be frequently exchanged, e.g., at 10 - 100 Hz rates, between vehicles over a
single hope or multiple hops. Most importantly, the event-driven safety messages to
be transferred over VANETs are highly time-sensitive. For example, WAVE (Wire-
less Access for Vehicular Environment) messages about accidents, stop/slow vehicle
warning, Blind Spot Warning (BSW), and Emergency Electronic Brake Light (EEBL)
3.2. TIME SYNCHRONISATION REQUIREMENTS IN VANET 55
are required to be broadcast to targeted nodes within a fraction of a second. It is
understood that typical end-to-end network latency is up to 100 ms for many VANET
applications [Karagiannis et al., 2011]. Any further offset will expose vehicles in at
risk in the hazardous road environment. Thus, precise clocks that depend on time
synchronisation are required in VANETs.
A number of time synchronisation mechanisms including software protocols and
algorithms have been proposed and implemented over the years in computer and
communication networks. However, most often they are ill-suited and/or incompat-
ible with VANET requirements.
The time synchronisation mechanism for vehicular ad-hoc networks and its ap-
plications needs to address the following three criteria:
i) the degree of accuracy needed, that is, the accuracy with respect to some exter-
nal time reference,
ii) the degree of precision needed, that is, the range of accuracy that can be main-
tained at the node clock, and
iii) the availability and longevity of synchronisation mechanism, that is, whether
the system needs to stay synchronised indefinitely or establishes synchronisa-
tion on-demand.
3.2.1 Accuracy Requirement
The Timing Accuracy parameter refers to the closeness of the node time to the stan-
dard physical time such as universal coordinated time (UTC) or the standard local
time. It is a measurement value resulting from the external synchronisation of the
nodes. Let V(t) ⊆ N be a set of vehicular nodes aligned with the physical time t. In
such a communicating network, let p and q be pair of nodes such that p,q ∈ N. If α is
the required timing accuracy parameter then for each node the system should satisfy
56CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
∀t∀p ∈ V(t) : |Cp(t)− t| ≤ α (3.1)
Where α is an accuracy parameter and Cp(t) is the time maintained by the clock
of node P. VANETs are a real-world networks in which most of the communication
interactions involve time-based decisions. For example, the real-time status messages
of VANETs carry data about the position, velocity etc., which are need to deter-
mined accurately and delivered precisely on time. Communication with the outside
world via the Internet or other means requires maintaining accurate physical time for
meaningful and successful interaction. Therefore, maintaining timing accuracy with
respect to a standard reference time is a prerequisite in VANETs.
3.2.2 Precision Requirement
The term time precision pertains to the agreement and closeness of a set of results. In
a network, it refers to the closeness of the times kept at two or more nodes. It is often
called instantaneous precision and is measured as the degree or boundary of difference
between clocks. The precision requirement concerns the internal synchronisation of
nodes and a precise system should satisfy,
∀t∀p, q ∈ V(t) : |Cp(t)− Cq(t)| ≤ β (3.2)
Where β is the parameter of precision. Cp(t) and Cq(t) are the time of the node
p and q respectively. Precision is the expression of a boundary of accuracy which is
essentially a function of the Mean (a) and Standard Deviation (σ) of a set of times (t).
a = 1N ∑N−1
i=0 xi
σ =√
1N−1 ∑N−1
i=0 (xi − a)2
(3.3)
3.2. TIME SYNCHRONISATION REQUIREMENTS IN VANET 57
Where xi is the instantaneous time difference between nodes. In case of a single
node synchronised with an external standard clock, Equation 3.2 reverts to the ac-
curacy equation 3.1. This implies that a precise system should satisfy the specified
accuracy requirement.
The VANET spectrum is limited (75 MHz) and maintaining accurate time can
result in better spectrum utilisation. To avoid interference and accommodate time
inaccuracies between nodes, a period called Guard Interval (GI), also known as a
Guard Band, is used in synchronous and asynchronous communication channel co-
ordination protocols. A longer GI is undesirable because it generates idle time in the
transmission process, which reduces the data throughput. Thus, maintaining good
timing precision reduces the necessity of a GI or Guard Band
3.2.3 Availability Requirement
Availability is the measure of service performance, which can generally be defined as
the ability to deliver services upon demand. It captures the continuity, quality and
functionality of the service. Along with accuracy and precision, availability of time
synchronisation services is equally important in VANETs. Since communication in
a vehicular network is a progressive process, the underlying time synchronisation
processes need to be continuously updated with time.
The set of requirements for time synchronisation in VANETs can be classified into
two categories: Performance-oriented requirements and Application-oriented requirements.
The former is based on the system-wide objectives including protocol support and
compatibility with the synchronisation method, whereas the latter is based on the
needs of end-user applications.
58CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
3.2.4 Requirements for Different VANET Applications
3.2.4.1 Requirements for System Level Applications
The requirements for time synchronisation in system-level applications are essential
in VANETs and are discussed below.
Network Interoperability and Coordination refers to the ability of networks to
send and receive messages, and communicate information between inter-connected
networks, devices and nodes. It is the capacity for efficient and meaningful coordina-
tion among network nodes and components for information exchange.
Scheduling of Channels is required for efficient use of channel resources. VANET
utilises short-range communications, e.g., 5.9 GHz Dedicated Short-Range Communi-
cation (DSRC) technology, typically within a range of 1 km, to provide high data rate
and low latency. In general, a trade-off exists between the efficiency and reliability
of VANET communications. The efficiency is typically characterised by bandwidth
consumption, channel utilisation, and channel coordination. Therefore, designing an
efficient WAVE Medium Access Control (MAC) protocol is essential for improved
efficiency, enhanced Quality of Service (QoS), and reliable packet transmission. Time
synchronisation plays a crucial role in MAC coordination. Existing IEEE 802.11p
MAC uses Time Synchronisation Function (TSF) to coordinate channels. The underly-
ing medium access mechanism is Carrier Sense Multiple Access/Collision Avoidance
(CSMA/CA), which is asynchronous in nature [Hafeez et al., 2013] and is only appli-
cable to systems in which precise sub-second timing is not required. In dense VANET
scenarios, CSMA/CA does not support highly accurate time synchronisation.
Road safety is a critical objective in VANET. To avoid unpredictable events that
may cause road accidents, VANET network delays need to be small and predictable.
Effective channel scheduling will help reduce the unpredictability of VANET commu-
nications. Compared with CSMA/CA, time-slotted access protocols offer collision-
free communications with predictable dynamics. Slotted protocols, e.g., STDMA,
3.2. TIME SYNCHRONISATION REQUIREMENTS IN VANET 59
MS ALOHA, RR ALOHA and UTRA TDD, offer good scalability, high reliability,
and fair use of channel resources [Bilstrup et al., 2009, Bohm, 2013, Cozzetti and
Scopigno, 2011, Cozzetti et al., 2009, Cozzetti and Scopigno, 2009, Golestan et al.,
2012a, Scopigno and Cozzetti, 2009, Verenzuela et al., 2014]. However, these time
slotted protocols require precise absolute time synchronisation for time slot coordina-
tion [Lim, 2016].
A Guard Interval is used between two-time slots in slotted access protocols. Packet
propagation starts at the beginning of a new slot after a guard interval. This helps
accommodate timing inaccuracy and propagation delay. The guard interval should
be set to be bigger than the worst time synchronisation accuracy for the slotted access
mechanism to work properly. Precise absolute time synchronisation helps reduce the
guard interval greatly, thus increasing the channel slot duration significantly [Ebner
et al., 2002].
For example, the time-slotted access protocol, STDMA, is used in shipping nav-
igation through Automatic Identification System (AIS). A typical frame length of
STDMA in AIS is 2, 016 slots. It is shown [ETSI, 2012] that compressing the guard
interval by 10 µs will accommodate 45 new slots of 496 µs each. This is translated to
a noticeable increase in channel capacity, which implies that more time slots can be
used for packet delivery.
Better spectrum utilisation is considered as one of the key targets in wireless com-
munication. VANETs use a limited spectrum of 75 Mhz. Better use of this spectrum
enables to increased data throughput. In a VANET protocol stack using the “Wireless
Access for Vehicular Environment (WAVE)”, the available bandwidth is divided into
service and control channels, as shown in Figure 3.1 [Gupta et al., 2015, Li, 2010,
Morgan, 2010].
For efficient channel coordination, communicating nodes need to be synchronised.
In practical operation, the clocks of all nodes can delayed for many reasons and tend
to lose synchronisation. To accommodate the time differences among the nodes, a
60CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
Control Channel
172
175
174 176 178180 182
181184
Service Channel Service Channel
GI
5.8
50
5.8
65
5.8
75
5.8
85
5.8
95
5.9
05
5.9
15
5.9
25
5.8
55
Frequency (GHz)
SCH SCHCCH
CCH interval
CC
HS
CH
SCH interval
Guard intervalSynchronization interval
CCH inactive
SCH inactive
Start of UTC second
(a)
(b)
Figure 3.1: WAVE Spectrum: (a) Frequency and Channel Allocation; and (b) ChannelSynchronisation and Guard Interval Morgan [2010].
GP GPtransmit
GP GPtime slot #n time slot #n+1
receive
t j
t i
tij
t
t
Ni
NJ
TGP
Figure 3.2: Guard Interval Requirements Scopigno and Cozzetti [2009].
Guard Interval, which is also known as Guard Band , is used in communication
design. As shown in Figure 1 (b), a guard Interval is a period of time for separation of
two consecutive and distinct data transmissions from different users in a time-slotted
mechanism or from the same users in a frequency slotted mechanism.
The Guard Interval requirements and its relationship with time synchronisation
accuracy in VANETs are demonstrated graphically in Figure 3.2. As shown in Figure
3.2, assume that nodes Ni and Nj have time offsets of ∆ti and ∆tj, respectively, with
reference to the global standard of time. When node Nj sends a burst to node Ni, then
3.2. TIME SYNCHRONISATION REQUIREMENTS IN VANET 61
the observed time offset ∆tij at node Ni is estimated as
∆tij = ∆tj − ∆ti + dij/c (3.4)
where dij is the distance between the two nodes, and c is the speed of light. A
successful reception of data at node Ni from node Nj can be achieved if there is no
any overlap of communications due to time offsets. Therefore, a guard interval (TGP)
is introduced to avoid such an overlap. This requires the guard interval TGP to be
greater than the time offset ∆tij, i.e.,
∆tij < TGP (3.5)
Condition (3.5) is required for time synchronisation in wireless networks. It is seen
from this requirement that the Guard Interval can be reduced if VANETs are better
time synchronised. Moreover, since a Guard Interval is an addition to the commu-
nicating slot length, it consumes spectrum resources and consequently leads to a
longer time to transmit a message. Therefore, a reduced Guard Interval implies better
spectrum utilisation.
The impact of the guard interval on the performance of VANET services varies
with the type of the underlying communication protocol. Asynchronous wireless
communication protocols in IEEE 802.11 networks use CSMA/CA as the channel
coordination mechanism, in which a Guard Interval is used to avoid transmission
disruption due to propagation delays, echoes and data reflections. In 802.11n net-
works, cutting the guard interval by half from 800 ns to 400 ns leads to an increase in
effective data transmission rate by 11% [Perahia, 2008].
In synchronous-slotted protocols, e.g., TDMA and STDMA, a Guard Interval ac-
commodates clock inaccuracies. This enables the avoidance of message collisions
and message losses in time-slotted medium access protocols. For example, the com-
monly used frame length of STDMA in Automatic Identification System (AIS) of ship
62CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
navigation is 2, 016 slots. In such a framing, a reduction of 10 µs in Guard Interval
means that 45 new slots can be accommodated for every 496 µs, thus increasing the
channel capacity by about 9% [ETSI, 2012]. Therefore, precise time synchronisation
contributes to increased communication capacity of wireless networks.
In VANETs, network environments frequently change over time. Consider a dy-
namic scenario in which in the network density changes from a small number of
nodes (< 20) to a large number of nodes (>100). To maintain a QoS level during
this network density change, an efficient, reliable and scalable medium access control
mechanism is required that has precise time synchronisation. In another common
VANET scenario, dynamic changes occur in the location of mobile network nodes
from one geographical region to another. The fastest moving nodes can momentar-
ily connect and disconnect from node clusters. Maintenance of QoS in such highly
mobile networks requires all nodes to follow the same time standard, which can be
achieved through time synchronisation.
3.2.4.2 Requirements for Performance Enhancing Applications
In compared with many other wireless ad-hoc networks, VANETs are noticeably dy-
namic and highly mobile. In VANETs, the relative speed between two nodes can
be as high as 220 km/h. This implies that a high-speed node may only stay within
the transmission range of other nodes for a few seconds. Moreover, some VANET
applications require an extremely small end-to-end delay. For example, the maximum
acceptable end-to-end latency for pre-crash sensing warning messages is 50 ms. For
Lane Change Warning and Forward Collision Warning, it is specified to be 100 ms by
National Highway Traffic Safety Administration of US [Eze et al., 2014, Hartenstein
and Laberteaux, 2010, Rasheed et al., 2017]. Therefore, to meet the requirements of
these safety applications, accurate timing is required for deterministic and reliable
communications in VANETs.
Security is a significant concern in VANET. Session hijacking and jamming are
3.2. TIME SYNCHRONISATION REQUIREMENTS IN VANET 63
Tracking Node
Tracking Node
Tracking Node
Tracking Node
Communication Failure
Cyber Attack
(a) (b)
Figure 3.3: Examples of Security Issues. (a) Cyber Forensic, (b) Cyber Attack (Security). Inboth of the cases time synchronisation is important to log the events accurate and preciselyBen-El Kezadri and Pau [2010].
two communication threats for the forensic security experts and transport regulation
authorities. Precise time synchronisation is a key tool for development of traceable
and reliable communications. This allows reconstruction of packet sequences on
channels, and thus helps overcome security threats.
In cyber security domains, log files are used as sources of evidence. Analysing
communication network log files relies on having precise timestamp records, which
is only possible if the network is time synchronised. The synchronisation accuracy de-
mand, however, depends on the applications requirements. In VANETs, researchers
indicated that a fine-grained analysis of channel activity between concurrent trans-
missions requires stringent timing guarantees of 8µs for DSRC communication [Ben-
El Kezadri and Pau, 2010].
Figure 3.3 (a) (b) shows two scenarios, namely, unwanted communication failure
and cyber-attack, which are examples where a forensic analysis is required for acquir-
ing potential evidence. In both cases it can be assumed that neighbouring nodes that
are not part of the incident can serve as observer or a tracking nodes. Any roadside
unit can also be an observer. In such cases it is preferable that the tracking node is
synchronised independently of the incident node.
64CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
However, in packet-based time synchronisation (i.e. in-band message transfer
based time synchronisation) inter-contacts are required. In these cases, the contacts
should have made shortly before and after the incident. Since VANETs are highly
dynamic networks, such tracking node contacts may be not feasible. Again, if track-
ing nodes are time synchronised by malicious attacking nodes, the ensuing forensic
analyses may not be accurate as the timestamps of attacks may be compromised.
In such circumstances, any time synchronisation solutions that are independent of
explicit signalling will probably be the best candidates.
Wormhole attacks are a routing problem in mobile ad-hoc networks that confuse
routing mechanisms by generating a fake node path that is shorter than the actual
route. This is also a considerable threat for location-based wireless security systems.
A well-known approach to prevent wormhole attack is the so-called packet leash,
which also requires highly accurate clock synchronisation [Isaac et al., 2010]. In a
variety of temporal leashes the time synchronisation accuracy is higher, of the order
of a few microseconds or even hundreds of nanoseconds [Hu et al., 2003].
GPS-based relative vehicle positioning requires time synchronisation. Time-to-
Collision (TTC) on roads depends on relative locations of two vehicles. The most
stringent requirements for relative positioning accuracy is about 10 cm [Caporaletti,
2012]. Vehicle state and time information are exchanged between vehicles to compute
relative vehicle positions for safety decisions. For a vehicle traveling at the speed of
110 km per hour, a timing error of 10 ms will cause a position uncertainty of 30 cm,
which is too high for collision avoidance. To meet the accuracy requirement of 10 cm,
it is required to keep timing errors within 3 ms, giving a relative positioning error of
about 9 cm. This requirement does not seem very high. However, it emphasises that
time synchronisation is essential.
Localisation is a typical application in VANET. It is directly related to the challeng-
ing issue of determination of accurate vehicle positions and ranges to other vehicles.
Terrestrial radio frequency based ranging techniques, e.g., Time of Arrival (TOA)
3.3. SUMMARY 65
and Time Difference of Arrival (TDOA), are effective ways for location determina-
tion [Golestan et al., 2012b, Lee et al., 2009, Yoon et al., 2012]. They essentially involve
distance calculations involving the speed of light. This implies that a timing accuracy
of 10 ns corresponds to a distance measurement accuracy of about 3 m. If time
synchronisation is accurate to 1 ns, the ranging accuracy can be improved to 0.3
m. Therefore, highly accurate absolute time synchronisation will enable the of use
terrestrial radio ranging signals in vehicle location determination.
Overall, Figure 3.4 shows the concept tier of time synchronisation requirements
for different application with respect to the timing accuracy demand.
ns
System Level Application
Performance Enhancing Application
Timing Requirements of
Different VANET
Applications
LocalizationGuard Interval
Slotted MACSecurtiy
SchedulingNetwork Coordination
Non-Real Time/User Experience Connectivity
Figure 3.4: Concept Tier that Illustrate the Requirements of Time Synchronisation Accuracyfor Different Applications in VANET.
3.3 Summary
Time Synchronisation is a critical service in VANETs it is a fundamental require-
ment for many system-wide network applications. From basic network coordination,
channel scheduling to cyber-security, time synchronisation is important for increasing
system capacity and also in location-based services for correct location identification.
The accuracy and precision requirements, however, are different in differing applica-
tions, which can be classified into two categories: coarse grain and fine grain timing
requirements.
66CHAPTER 3. SIGNIFICANCE AND REQUIREMENT ANALYSIS OF TIME
SYNCHRONISATION IN VANET
Table 3.1: List of Timing Accuracy Requirements for Different Applications in the Basis ofEssentialness of VANET.
Applications Timing TypesAccuracy Req./Performance IMP.
EssentialNetwork Coordination Coarse ∼ms
Scheduling of ChannelsNon-SlottedSlotted
CoarseFine
sub-mssub-µs
Relative Vehicle Positioning Coarse 3 msSecurity Fine 8 µsDesirable
Guard IntervalNon-SlottedSlotted
CoarseFine
Rate 11%AIS-10µs@45slots
Localisation Fine10 ns @ 3m1 ns @ 30cm
An overall summary of time synchronisation requirements in the category of es-
sential and desirable services are tabulated in the Table 3.1.
Chapter 4
GNSS Time Synchronisation In VANET
The previous chapter identified the importance of vehicular networks and their time
synchronisation requirements. This chapter analyses the feasibility of GNSS time
synchronisation with respect to vehicular environment compatibility, timing accuracy
achievability and signal availability. The result of a series of experiments are pre-
sented to verify the 1PPS accuracy achievable using consumer-grade low-cost GNSS
receivers.
Some parts of the content from this chapter has contributed to the following pub-
lications:
1. K. F. Hasan, Y. Feng, and Y.-C. Tian ”GNSS Time Synchronisation in Vehicular
Ad-hoc Networks: Benefits and Feasibility,” in IEEE Transaction on Intelligent
Transportation System, 2018 DOI: 10.1109/TITS.2017.2789291.
2. K. F. Hasan and Y. Feng, ”A Study on Consumer Grade GNSS Receiver for the
Time Synchronisation in VANET,” presented at the 23rd ITS World Congress,
Melbourne, Australia, 1014 October 2016.
67
68 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
4.1 Motivation of GNSS-driven Time Synchronisation
Global Navigation Satellite Systems (GNSS) are regarded as an international utility
for positioning, navigation and timing (PNT). The generic term ”GNSS” refers to the
USA’s Global Positioning System (GPS), Russia’s GLONASS, Europe’s GALILEO and
China’s BEIDOU navigation satellite system (BDS). It is noted that GPS, GLONASS,
Galileo and BDS use different reference time systems creating time offsets between
them. However, the offsets can be determined at the system level or user level. Any
one or more constellations can offer the same global standard UTC time. With their
worldwide coverage, continuous service, GNSS has become one of the most efficient
and standard systems for time dissemination in many applications. Many industries
such as energy, meteorology and telecommunications rely on GNSS for accurate time
synchronisation in their systems and devices. The accuracy achieved by GNSS-based
time synchronisation using GPS is better than 40 ns 95% of time.
Most of the earth-based time transfer techniques suffer from path delay mea-
surement uncertainties. In contrast, the satellite-based GNSS time transfer systems
possess measurable constant path delays. This arises because the variation of path
delays are small and due to clear, unobstructed paths to receivers. Therefore, the
delay measurements are straightforward and can be more easily calibrated compared
to any ground-based systems. In addition, the radio interferences due to weather or
any other ground-based noise have less impact in satellite-based GNSS systems.
In telecommunication networks, GNSS is used to synchronise some major nodes
called root or server nodes outdoors. Through these root nodes, other nodes in the
system are synchronised by using other synchronisation techniques, which are mostly
based on message transfer between nodes.
In contrast to telecommunication networks, VANETs are outdoor-based networks.
Except in some tunnels and blocked roads, nodes in VANETs on the road are mostly
under the coverage of GNSS signals. It is a straightforward choice for VANETs to
4.2. FEASIBILITY OF GNSS SYNCHRONISATION IN VANET 69
use GNSS for synchronisation. GNSS receivers have already been used for vehicle
navigation and positioning. Nowadays, multi-GNSS constellations, more precise
GNSS services, such as space-based argumentation systems (SBAS), differential GNSS
(DGNSS) services and precise point positioning (PPP), are available for VANET de-
ployments. The GNSS-based time synchronisation is indeed plausible in VANETs.
It is, therefore prudent to understand how GNSS time solutions provide synchro-
nisation in VANETs and what the possible solutions are when GNSS services are
absent, such as when vehicles travel in tunnels. The feasibility and accuracy of GNSS
time solutions are discussed and examined through experiments in the sections that
follow.
4.2 Feasibility of GNSS Synchronisation in VANET
GNSS is a space-based Positioning, Navigation and Timing (PNT) utility alongside
wireless wide area networks and communications. It has a potential to bring signifi-
cant benefits to centralised and accurate time synchronisation in VANET.
4.2.1 Justification of the Feasibility
First of all, GNSS services are capable of providing absolute time-synchronisation
support over local and decentralised time synchronisation protocols [Shizhun Wang
et al., 2010]. Such external synchronisation techniques are also free from additional
message transfer delays between nodes, as shown in Figure 4.1 (a). Most time-synchronisation
techniques are designed with message exchanges between nodes. Therefore, they
rely on the data communication networks. This is shown in Figure 4.1 (b). The
performance of in-band internal time synchronisation depends on the channel ac-
tivities, the density of the communicating nodes, and the condition of the networks.
However, message delivery in such networks may suffer from a significant latency
and jitter [Chen et al., 2015, Gunther and Hoene, 2005, Loschmidt et al., 2012]. On
70 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
the contrary, GNSS based out-of-band external synchronisation techniques do not
use the communication networks in their operation. They do not use any network
bandwidth resources as all nodes are synchronised with external GNSS signals. Thus,
their performance is independent of the number of network nodes. It is worth men-
tioning that in the absence of GNSS signals, in-band time synchronisation has to be
activated for better time keeping. For instance, Timing Advertisement Frames in
DSRC-based VANET may be used to assist in time synchronisation if GNSS signals
become unavailable.
Network
Delay and
Jitter Network
GPS
Sync Link
Network
Delay and
Jitter Network
Propagation
Time
(a) (b)
Figure 4.1: Broad View of Time Synchronisation (a) In-band Time Synchronisation (b) Out-of-Band External Time Synchronisation.
The feasibility of GNSS time synchronisation is also justified by the fact that a
single visible satellite is able to provide time solutions. A GNSS receiver normally
tracks all satellites in view to obtain pseudo-range and Doppler measurements at
each frequency for Position, Velocity and Time (PVT) computing. The time states
include clock bias and clock rate. There are basically two modes for estimation of
time states: dynamic mode and static mode. The dynamic mode is used in moving
platform applications when the position is unknown. In this case, the receiver can
compute its own position and time by tracking four or more GPS satellites. The static
mode is the preferred mode for applications with a known fixed position. In this case,
4.2. FEASIBILITY OF GNSS SYNCHRONISATION IN VANET 71
the receiver can compute time bias and time rate by tracking one or more satellites.
In the dynamic mode with unknown position, the PVT estimation is performed by
solving a set of linear observation equations with the least-squares approach epoch by
epoch. In other words, the 4D states of the vehicle moving platforms are determined
without assuming knowledge of the dynamics of the receiver. The accuracy of the
state solutions depends on two factors: the user range equivalent error (URE) for
the observation accuracy, and the geometric dilution of precision (GDOP) about the
satellite geometry. In general, the PVT solutions obtained under the GDOP ≤ 6 are
considered to be valid and usable. The position and time errors are almost of the same
order of magnitude.
In the static mode, if the receiver position is known to a certain accuracy through
alternative positioning techniques or predictions, one tracked satellite at a time is still
able to provide timing information with a reasonable accuracy [Lombardi et al., 2001].
When there are fewer satellites than four in view, or the satellite geometry is very
poor, alternative positioning techniques, such as inertial measurement units (IMU),
are normally involved to determine or improve the position and velocity states. As
long as there is one satellite in view, the receivers PVT processor can calculate a
time solution. The time solution outage is always much lower than that of the us-
able PVT solution outages. In fact, most new GNSS receivers are designed to track
multiple GNSS signals, i.e. signals from GPS, GLONASS, Galileo and Beidou (BDS)
constellations. As a result, visible multi-GNSS satellites would double or quadruple
the number of GPS visible satellites. This will reduce the outage of time solutions
significantly, thus making GNSS-based time synchronisation much more feasible and
reliable than with GPS alone.
The feasibility of GNSS time synchronisation is further justified by the fact that
absolute time synchronisation with an external global time standard is particularly
suitable for VANET applications. VANET nodes operate outdoors most of the time.
The density of the nodes vary significantly from time to time or from one location to
72 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
another depending on the traffic conditions. In-band time synchronisation will suffer
from large jitter and variable latency in exchanges of synchronisation messages. This
difficulty can be easily overcome by using out-of-band external GNSS time synchro-
nisation.
Moreover, consumer-grade GNSS receivers are already mounted in most modern
vehicles for positioning and navigation. They are ready to provide GNSS timing
information for time synchronisation without additional hardware cost.
4.2.2 GNSS Timing Information
The timing information provided by GNSS is highly precise and accurate, as it is
generated from atomic clocks and maintained very stringently. In a GNSS system,
there are three scales of time, i.e., GNSS time, satellite time and standard time (such
as UTC). These times are different from each other [Misra and Enge, 2006, Scott and
Demoz, 2009]. In the satellite time transfer method, the offset between GNSS and
UTC time are transmitted to user receivers for correction. As of September 2017, UTC
is ahead of GPS by 18 s whereas international atomic time, also known in French as
temps atomique international (TAI), is lagged by 19 s as shown in Figure 4.2.
The satellite system provides UTC time to a ground receiver and is adjusted using
navigation messages as shown in Figure 4.3. In general, a typical satellite has an
18s
37s
UTC GPS TAI
UTC- Universal Coordinate Time
GPS- Global Positioning System
TAI- International Atomic time
Figure 4.2: Time Offsets Among Different Atomic Scale Standards.
4.2. FEASIBILITY OF GNSS SYNCHRONISATION IN VANET 73
atomic clock to maintain its own time tp. This time is regulated by the earth bound
control segment with GPS time tgps. Along with tgps, the control segment also uploads
navigation messages containing ∆tutc, which is regulated by the United States Naval
Observatory (USNO). A GPS receiver has its clock system with time tr. With GPS
pseudo-range measurements, the receiver can compute clock bias ∆tr with respect to
GPS time, together with the receiver position states. The ∆tr parameter is defined as:
∆tr = tr − tgps. (4.1)
Control
Segment
(tgps)
Satellite time (tp)
Receiver time
(tr)
Figure 4.3: Time Transfer Through GNSS [?].
The receiver clock bias in a receiver hardware system is also known as the instan-
taneous receiver clock offset relative to GPS [Misra and Enge, 2006]. Since this offset
can be large, GPS code measurements are known as pseudo-ranges. The clock bias
is adjusted to the GPS time when the magnitude reaches a certain limit, such as 0.1s.
As a result, a time GPS receiver adjusts the bias and obtains its UTC time using the
following relationship:
tutc = tr − ∆tr − ∆tutc, ∆tutc = tgps − tutc, (4.2)
74 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
where the offset ∆tutc between GPS time and UTC time can be obtained from nav-
igation messages. The ∆tutc parameter contains the leap seconds (currently 18s) as
shown in Figure 4.2 and a fractional part since the last leap second adjustment.
With the above GPS time transfer technique, all nodes of a network are individu-
ally synchronised with the GPS time. Then, GPS times are adjusted with an additional
UTC time offset. As a result, all nodes are synchronised with the UTC time. This is
demonstrated in Figure 4.4. As shown in Figure 4.4 (a) and (b), nodes N1 and N2 are
individually time synchronised with a satellite (tgps), and then are updated to UTC
time. Effectively, this will synchronise Node N1 and Node N2 with each other as
shown in Figure 4.4 (c).
1 1
N2
11
(a) (b)
(c)
tgpstrtutc
tgpstrtutcRTC
2 2
N1
N1 N2
RTC
(c) N N 1 2
(a) (b)
Figure 4.4: (a) (b) Nodes N1 and N2 are Individually Synchronised with GNSS and Updatedwith UTC. (c) Effectively, Two Nodes are Synchronised with Each Other via GNSS.
4.2. FEASIBILITY OF GNSS SYNCHRONISATION IN VANET 75
4.2.3 Errors of the Receiver Timing
The errors of receiver timing are now examined. Only the uncertainty of the GNSS
clock bias ∆tr will affect time synchronisation because ∆tUTC is common to all VANET
nodes. However, the basis for time transfer func intions in GNSS-based products is 1
PPS (one pulse per second) signal generated by the receivers. Such receiver generated
signal is a short logic pulse, where an edge of it is adjusted by the receiver to be
on time corresponding to the one second epoch of UTC or GPS time (GNSS time).
Errors in the time of occurrence of the 1 PPS pulses from the GNSS receiver, there-
fore, consists of three parts: 1) bias or offset due to uncompensated propagation and
hardware delay errors in the receiver/antenna system; 2) drift, which is the variation
in time over an extended period due to changes of satellites tracked over time; and
3) jitter, which is the short-term variation in timing from pulse to pulse. These error
sources are inherent in both GPS system and GPS receiver design/implementation.
The total effect of these errors is typically tens of nanoseconds to a few microseconds,
depending upon the quality of GPS receivers. Receiver manufactures usually cali-
brate the receiver bias well, yielding a timing accuracy of 10 ns or better under ideal
observational conditions. This high level of accuracy is achievable down to the reason
that the timekeeping mechanism maintained within the GPS system is repeatedly
adjusted in the system level to null out the timing errors that may generates.
A typical GNSS receiver has an internal quartz based oscillator that continuously
runs and follows the GPS time. Currently, a low-end GNSS receiver can update
position and time solution at a rate of up to 20 Hz, which corresponds to an interval
of 50 ms.
However, such a quartz-based clock deviates over time. This is because the fre-
quencies of the quartz oscillators are different. As time elapses, a quartz clock tends
to diverge from the perfect clock, i.e., the real time, and also from others in a network.
Ideally, for a perfect clock, the clock rate of change dC/dt is equal to 1. In practice, this
rate may increase or decrease due to the variation of the clock oscillator’s frequency.
76 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
In a quartz clock, this frequency variation is commonplace due to the impact of
the environmental changes at the node, e.g., variations in temperature, pressure,
and power voltage upon the local clock. In clock synchronisation terminology, this
difference in the frequencies of the practical clock and the perfect clock is known as
clock skew. The rate of change of the clock skew is known as clock drift. This means
that clock drift is the derivative of clock skew [Sundararaman et al., 2005].
Though the frequency of clock oscillator depends on ambient environmental con-
ditions and may change over time, for an extended period, e.g., minutes to hours,
the frequency of node clock can be approximated with good accuracy by an oscillator
with a fixed frequency [Levesque and Tipper, 2016, Sichitiu and Veerarittiphan, 2003].
Therefore, the clock of a node can be expressed as:
Ci(t) = di · t + bi, (4.3)
where t is the standard time of the measurement (UTC); di is the clock drift due to the
oscillator’s frequency differences resulting from environmental changes at the node
(e.g., the impact of variations in temperature, pressure, and power voltage upon the
clock); bi is the initial offset in the GNSS synchronisation framework, and can be
correlated to systematic ranging errors and hardware delays. The di and bi can be
different from node to node. However, the clock skew di is different from the drift in
GNSS timing errors offered by GNSS 1PPS outputs. It is also worth mentioning that
the clock skew can actually be estimated from GNSS Doppler measurements along
with the velocity states.
Now, any two such GNSS synchronised clocks (such as Figure 4.2(a) and Fig-
ure 4.2(b)) can be expressed as:
C1(t) = d1 · t + b1,
C2(t) = d2 · t + b2.
(4.4)
4.3. AVAILABILITY OF GNSS TIME SOLUTIONS 77
By following Figure 4.2(c), they can be related as:
C1(t) = Θ12C2(t) + β12, (4.5)
where Θ12 is the relative drift between two receivers and β12 is the offset due to
the bias variations. While the clock bias of two receivers are the same, the offset is
cancelled, implying that β12 = 0.
Following Figure 4.2(c), the overall impact of GNSS enabled synchronisation on
the networking nodes can be practically estimated by an end-to-end timing compar-
ison. The results of a series of extensive experiments which predict the accuracy of
GNSS time synchronisation are presented in Section 4.4.
4.3 Availability of GNSS Time Solutions
Availability may be defined as the capability of a system to provide usable service
within a specified coverage area. GNSS provides well-recognised and accepted tools
for Position, Navigation and Timing (PNT) estimation and its signals should theo-
retically be available from any regions of the world. However, receiving satellite
signals relies on having near Line of Sight (LOS) propagation, therefore, the signal
are restricted by different obstacles, like buildings, trees and other obstructions etc.
Since vehicular environments are mostly outdoor based, it is plausible to cover by
the signal under general circumstances and scenarios. However, there are still certain
places where signal availability can be restricted and thus become a potential issue to
consider. For instance, signal within urban areas and under tunnels.
Chapter 6 presents the results of experiments and investigation of GNSS signal
availability within Brisbane urban high-rise areas. The aim of these experiments was
to understand, to what extent the obstruction of GNSS signals by physical structures
reduces the availability of GNSS position and time solutions. Consumer-grade re-
ceivers were used to collect the data and versatile RTKLIB GNSS data processing
78 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
software was employed to process and analyse the acquired data.
From the obtained experimental results, it was found that position outages occur
due to signal path obstructions, and incorrect position solutions result due to weak
satellite geometry. However, accessing multiple satellite constellations did improve
availability substantially as there was always a visible satellite present. Table 4.1
summaries that visibility of GNSS satellites in three cases: four or more satellites,
1 to 3, and null (< 1). The visibilities for a minimum of 1 satellite in the three same
constellations were 100%, 99.98% and 100%, respectively. This shows that having
access to multiple constellations reduces the impact of signal path obstructions on the
availability of GNSS time services. Thus, this experiment demonstrates the feasibility
of GNSS time synchronisation in the high-rise urban areas, particularly with multi-
GNSS constellations.
Table 4.1: The Number of Satellites Available with Different GNSS Services Constellation.
GNSS System NSAT ≥ 4 NSAT = (1 to 3) NSAT < 1GPS 77.32% 22.68% 0%BDS 82..93% 17.05% 0.02%BDS+GPS 99.25% 0.75% 0%
4.4 Synchronisation Accuracy of 1 PPS Signals
This section presents the an experiment and results that were conducted to assess how
accurately consumer-grade GPS receivers are able to synchronise their clocks with 1
PPS signals in VANET.
4.4.1 Characteristics of 1 PPS GNSS Signal
Acquiring the one pulse per second or 1PPS signal is important for the operation and
application of GPS receivers. This is an electrical signal having a width of less than
4.4. SYNCHRONISATION ACCURACY OF 1 PPS SIGNALS 79
one second that precisely repeats every second. Maintaining precise timing perfor-
mance is vitally important to the operation of GPS receivers. To maintain a precise
pulse rate, consumer grade GPS receivers use a high-precision crystal or rubidium
oscillator that can synchronise itself to the common time stamp generated by the
atomic clock of GPS satellites. In practice, the received 1PPS signal suffers from jitter
with respect to ideal 1PPS signals that are generated by the highly accurate satellite
atomic clocks. Along with GPS ephemeris, signal conditions due to orbit paths,
receiver oscillator drift, cable connections and device interconnections can contribute
to the jitter [Texas Instruments, 2012, Witherspoon and Schuchman, 1978].
A comparison between an ideal GPS 1PPS signal’s rising edge with a typical
receivers 1 PPS signal rising edge is shown in Figure 4.5 [Texas Instruments, 2012].
This theoretical figure reflects the result of mitigating jitter to produce an accurate
timing signal at the receiver’s end.
Rising Edge Threshold
Rising Edge Delayed Rising Edge
Δ t
Time
Amplitude
Figure 4.5: Ideal (GPS signal) and Practical (Produced Signal by GPS receiver) 1PPS SignalPulses.
80 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
4.4.2 Clock Accuracy of Low Cost GPS Receivers
The GPS time, that is sent and controlled by the GPS system, is generated by highly
precise atomic clock. This GPS time is usually not adjusted with the Universal Co-
ordinated Time (UTC), and therefore some offsets always remain between them. Ac-
cording to the data recoded in July 2015, GPS time is ahead of UTC by 18s [QPS, 2015].
Another very accurate timing system, the International Atomic Time (TAI) that uses
200 caesium atomic clocks in over 50 national laboratories worldwide, also differs
from GPS by 19s. This offset usually adjust by adding leap seconds [Behrendt et al.,
2006, Lewandowski et al., 1999, 1993].
For GNSS timing support, GPS uses four atomic on board clocks that are usually
accurate to few nanoseconds (ns) of each other [Behrendt et al., 2006]. GPS sends
two frequencies are known as L1(1575.42 MHz) and L2(1227.6 MHz) [Lombardi et al.,
2001]. L1 frequency carries civilian code has a time accuracy specifications of 340ns in
the scale of 2 standard deviations, and in practice it provides 35ns accuracy [Behrendt
et al., 2006].
Fundamentally, Consumer grade receiver units consist of an antenna unit, an elec-
tronic receiver unit with a correlator in order to lock to the satellite signals, a reference
time oscillator and a counter called time interval counter (TIC) to measure the arrival
time of received signals and the GPS-sent signals. Six-kinds of delays are encoun-
tered: one the usual offset of the on-board satellite clock, two the propagation delay
between the satellite and antenna, three the inherent signal delay in antenna unit, four
the delay from antenna to the receiver input, five the receiver-processing delay and six,
the time offset of the reference clock [De Jong and Lewandowski, 1997]. The first two
kinds of delays rely on the GPS system, whereas the remainder of the delays depend
on the receiver performance. Therefore, both GPS time accuracy and GPS receiver
module accuracy individually depends on various issues that needs to be considered
to demonstrate the overall accuracy of the timing and synchronisation accuracy of
GNSS systems. According to [Behrendt et al., 2006], the statistical probability of the
4.4. SYNCHRONISATION ACCURACY OF 1 PPS SIGNALS 81
accuracy of commercial GPS clocks range from 50 ns to 1 ms. This is a 1 standard
deviation (1σ) rating. A typical comparison between GPS time accuracy (34 ns, 2 σ)
and GPS receiver module accuracy (50 ns, 1 σ) is shown in Figure 4.6.
Figure 4.6: Standard Deviation Rating.
In practice, GPS timing accuracy can vary considerably can be seen from the
above figure and depends upon the manufacturer. Considerable research [Bogovic,
2013, Bullock et al., 1997, King et al., 1994, Mumford, 2003] has been conducted to
study 1PPS signal timing performance. However, their results vary with application,
manufacturer and type GPS receiver module.
In 1997, [Bullock et al., 1997] compared 1PPS signal timing for a GPS receiver with
a caesium atomic clock. Motorola UT Oncore GPS receivers were used in their study.
They observed a maximum offset of 236.5 ns with a standard deviation of 37.1 ns
(average) in position-hold mode, and a maximum spread of 538.6 ns with a standard
deviation of 78.5 ns (average) in position-fix mode.
Instead of comparing timing with a precise atomic clock, the relative offset be-
tween identical receivers (Motorola Oncore UT) has been investigated by [Bogovic,
2013]. They observed standard deviations of <13 ns in navigation-mode and <8 ns
in position-hold mode, respectively.
82 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
Conversely, in 2003, [Mumford, 2003] conducted another research on 1PPS signals
of three inexpensive GPS receivers from different manufacturers. He reported a typ-
ical maximum relative offset of 1450-1500 ns between two different receivers with a
typical jitter of 140 ns.
4.4.3 Experimental Design
The experiments described below employed GPS receivers from U-Blox (MAX-M8Q,
NEO-6T, Evaluation Series-6 EVK-6H-0-001 models) and Furuno (VN-872 model),
which are low-cost and consumer-grade. Both U-Blox and Furuno use Temperature
Compensated Crystal Oscillator (TCXO) that have excellent stability over a broad
range of temperatures. The receivers are equipped with a Pluto+ RPT5032A model
quartz oscillator. The clocks in these GPS receivers are designed with an advanced
temperature compensation circuit. According to their technical specifications, they
offer sub 0.1 ppm frequency stability over an extended temperature range (55◦C to
105◦C).
The experimental setup is shown in Figure 4.7. Two GPS receivers, along with two
identical antennas were connected to a high resolution oscilloscope, 200MHz Agilent
Technology DSO-X-2024A. The oscilloscope was used to measure the offset between
the 1 PPS signals of the receivers. A process to calibrate and record experimental
data was developed using Lab-View software on a computer system as shown in
Figure 4.7. The experiment was conducted on the 13th floor of a building at QUT
under an unobstructed open sky.
4.4.4 Data Acquisition
The experiment measurement and data acquisition method consisted of two phases.
In the First Phase, the timing offset was measured between identical receivers
(i.e., the same model from the same manufacturer). With identical receivers, it was
4.4. SYNCHRONISATION ACCURACY OF 1 PPS SIGNALS 83
GPS
1G
PS 2
Osc
illos
cope
1 PP
S1
PP
S
Syst
em t
o r
eco
rd a
nd
ana
lysi
s
USB
U
SB
Antenna
Tri
late
rati
on
Figu
re4.
7:Sc
hem
atic
Dia
gram
ofEx
peri
men
talS
etup
for
Sync
hron
isat
ion
Ass
essm
ento
fGPS
Rec
eive
rs.
84 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
expected that their clock oscillators and temperature compensation circuitry should
be exactly the same.
In the Second Phase, timing offset measurements were acquired between different
receiver models from different manufacturers. In this phase, we conducted the varia-
tion in two ways to understand the receivers performances, One the timing offsets
were measured between two receivers from entirely different manufacturers; (i.e.
vendors) two, the timing offsets were measured between two receivers of different
models from the same manufacturers.
In each case, the time offset was measured with respect to the single pulse wave-
form provided by the receivers as the 1 PPS signal output. The width and/or the rise
time together with the time delay between the signals were measured and compared
using the oscilloscope. For repetitions of the same experiments, the calibration of
the oscilloscope was controlled by developed LabView software. In particular, a pro-
cess cycle was developed and implemented using LabView to control the oscilloscope
settings and to record the data into a database. This process enabled the width of the
pulses to be measured and recorded every second. The logged data was subsequently
analysed using a MATLAB data analyser. The experiments were carried out under the
temperature (18◦C-25◦C).
4.4.5 Result Analysis
The results of a series of experiments are presented in this section and grouped ac-
cording to the type of GNSS receiver. The standard deviation (STD) and root-mean-
square of the offsets (RMS), for four different types of datasets (according to offset
time durations of 5 minutes, 1 hour, 10 hours and 24 hours) are presented in the
following.
4.4. SYNCHRONISATION ACCURACY OF 1 PPS SIGNALS 85
Figure 4.8: Pulses Showing Time Difference Between TwoWaveforms of Identical Receivers.
4.4.5.1 Time Offset Between Receivers of the Same Model
This sub-section reports the results of a series of experiments conducted with identical
receivers and receivers from two manufacturers, Ublox and Furuno. In the first stage
of the experiment, results from identical Ublox (NEO-6T) receivers are presented and
analysed. Then the results for the Furuno (VN-872) GNSS receiver are presented.
Figure 4.8 shows observations from the oscilloscope of the exact time differences
between two waveforms of identical receivers (Ublox-Ublox). As the main interest is
to measure the relative time differences of the 1PPS signal, this waveform indicates
the exact time differences about the threshold level. This threshold level was calcu-
lated at a trigger level of 800mV at which the value of ∆t was 22ns.
Figure 4.9a shows plots of the time offset and jitter observed over 5 minutes
between two receivers of the same model and from the same vendor. It can be seen
from this figure that the differences between the 1 PPS outputs of the two receivers
vary randomly over time. It is observed from this data set that the peak value of
|offset+jitter| is 30 ns. The mean value of the offset measurement is calculated as 3.6
ns. Figure 4.10 shows the statistical distribution of the observed data. The distribution
86 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
-40
-30 -20
-10 -3.60 10 20 30 40
Time Offset (ns)
01
45
6
Instant OffsetMin
MaxAvg
23
Elapsed Time (Minutes)
(a)Time
OffsetD
istributionofD
atafor
5M
ins.
05
10
15
20
25
30
35
40
45
50
55
Elapsed T
ime (H
ours)
-60
-40
-20 0 20
40
Time Offset (ns)
ub
lox-u
blo
x (same m
od
el) 1ho
ur
Instant offsetM
oving Average
(b)Time
OffsetD
istributionofD
atafor
1hour.
01
23
45
67
89
10
Elap
sed T
ime (H
ou
rs)
-80
-60
-40
-20 0
20
40
60
80
Time Offset (ns)
ub
lox-u
blo
x (s
am
e m
od
el) 1
0 h
ou
rs
Instan
t offset
Mo
vin
g A
verag
e
(c)Time
OffsetD
istributionofD
atafor
10H
ours.
02
46
81
01
21
41
61
82
02
22
4
Elap
sed T
ime (H
ou
rs)
-80
-60
-40
-20 0
20
40
60
80
Time Offset (ns)
ub
lox
-ub
lox
(sa
me
mo
de
l) 24
ho
urs
Instan
t offset
Mo
vin
g A
verag
e
(d)Time
OffsetD
istributionofD
atafor
24H
ours.
Figure4.9:
PlotoftheD
ifferentDatasets
ofRelative
Time
Offsets
Between
Two
IdenticalReceivers
(Same
Manufacturer
Same
Model.
4.4. SYNCHRONISATION ACCURACY OF 1 PPS SIGNALS 87
of measured data corresponds to a Gaussian distribution. The standard deviation is
equal to ± 12.67 ns for 1σ.
-50 -40 -30 -20 -10 0 10 20 30 40 50Offset
0
5
10
15
20
25
30
35
Density
Offset/ErrorNormal distribution
Figure 4.10: Gaussian Distribution of Dataset 5 Minutes thatRepresents STD of 12.67 ns for 1σ..
Table 4.2 summarises the observed of peak, mean, standard deviation (STD), and
root mean square (RMS) of observed time offsets from four independent experimental
sessions of 5 min, 1 hour, 10 hour and 24 hour durations. The results show that the
peak values vary between 30 ns to 90 ns, with standard deviations ranging from 9 ns
to 13 ns. Consistent RMS values between 10 ns and 13 ns were also observed. These
statistics match the accuracy and precision specifications of the GNSS receiver.
Table 4.2: Relative Time offsets of Different Datasets in the Timerange of Nanosecond Between Receivers of Same Model (Ublox-Ublox).
Test Session Peak Mean STD RMS5 min ±30 3.6 12.67 12.671 hr ±90 4.15 12.58 13.210 hrs ±40 3.8 9.73 10.424 hrs ±60 1.75 12.2 12.2
The observed time offset of the 1PPS solutions between two receivers of the same
model over 24 hours are shown in Figure 4.9d. As shown in Table 4.2, in this test case,
the STD, mean value and peak value of the time series are 12.2 ns, 1.75 ns, and ±60
88 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
Table 4.3: Relative Time Offsets of Different Datasets in the TimeRange of Nanosecond Between Receivers of Same Model (Furuno-Furuno).
Test Session Peak Mean STD RMS5 min ±17 8.5 3.5 9.11 hr ±20 5.6 3.8 4
10 hrs ±21 2.12 4.1 4.324 hrs ±21.1 1.67 4.8 5.1
ns, respectively. These results indicate a consistent match between the pulse signals
from the two receivers.
Similar experiments were conducted with the GNSS receiver manufactured by
Furuno. Two identical Furuno VN872 model receivers were used in this experiment.
The Furuno receiver was selected in the experiment because it is widely used in
ship navigation, such as the Automatic Identification System (AIS) in the Brisbane
city ferry network. In AIS, GNSS timing signal is already been used for system
synchronisation.
The time offset results from this experiment are tabulated in Table: 4.3. From the
table, the maximum peak can be identified from the dataset of 24 hrs, whereas, the
mean value is remarkably low at 1.67 ns. From all of this datasets it can be seen that
the performance of Furuno in terms of STD and RMS is slightly better than of Ublox.
However, in all cases, the result indicates, the consistent match between 1PPS signals
from identical receivers.
4.4.5.2 Time Offset Between Receivers of Different Models
Similar experiments were performed with two different model GPS receivers from
two vendors. The observed peak, mean, STD and RMS results of the experiments are
summarised in Table 4.4. It can be seen that tabulated values are much larger than
those in Table 4.2 measured from two GPS receivers of the same model, however, a
maximum offset of under 200 ns was again observed.
4.4. SYNCHRONISATION ACCURACY OF 1 PPS SIGNALS 89
0 60 120 180 240 300
Elapsed Time (Sec)
-5
0
5
10
15
20T
ime
Off
set (
ns)
Furuno-Furuno (same model) 5 mins
Max = 17 ns
Min = 0 ns
Instant offsetMoving Average
(a) Short Term Time Offset Recorded Between Receivers of Same Models (5 mins).
0 2 4 6 8 10 12 14 16 18 20 22 24
Elapsed Time (Hours)
-25
-20
-15
-10
-5
0
5
10
15
20
25
Tim
e O
ffse
t (ns
)
Furuno-Furuno Same Receiver 24h
Max = 21.1 ns
Min = 21.9 ns
Instant offsetMoving Average
(b) Long Term Time Offset Recorded Between Receivers of Same Models (24 hours).
Figure 4.11: Plot of the two datasets (5 minutes and 24 hours long) time offsetsbetween two identical receivers manufactured by Furuno (Model VN 872). (a) 5 minsdata represents a maximum of 17 ns deviation along with 3.5 ns STD and 9.1 ns RMS.(b) 24 hours data represents a maximum of 21.1 ns deviation with 4.8 ns STD and 5.1 nsRMS.
The offsets measurements over 24 hours are depicted in Figure 4.12. The STD
value is measured to be 30 ns. The mean offset is 6.9 ns and peak values are +80 ns
and −180 ns. The mean values over a moving data window of 2 hours show that the
offset is consistently within ±30 ns.
The results of another experiment between different model receivers but same
90 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
Table 4.4: Relative Time Offsets From Different Datasets in theTime Range of Nanosecond Between Receivers of Different Model(Ublox-Furuno)
.
Test Session Peak Mean STD RMS5 min ±55 10.6 9.1 311 hr ±45 8.6 26.4 29
10 hrs ±60 11.8 28 29.324 hrs ±180 6.9 30 31.4
0 2 4 6 8 10 12 14 16 18 20 22 24
Elapsed Time (Hours)
-200
-150
-100
-50
0
50
100
Tim
e O
ffse
t (ns
)
Instant offsetMoving Average
Figure 4.12: Time Offset Between Receivers of Different Models over a Long Period.
manufacturer are now presented; one is a receiver with a PPS output for which the
manufacturer does not claim that it is for timing applications, and the other one is a
receiver with a PPS output for which the manufacturer claims that it is designed to be
used for dedicated timing applications. In this experiment, interestingly, large offsets
between the two receiver PPS signals were observed. One receiver was observed
to always lead the other with a significant and constant offset. Table 4.5 shows the
recorded offset over 24 hours. It can be seen that the maximum peak is 764 ns with
mean value of 610 ns. Considering the mean value as the constant bias, we processed
the data by subtracting 610 ns form the dataset ended up with a peak value of 154 ns
with a mean of .5ns, STD 21.9 and RMS 21ns as shown in the Table 4.6.
This implies a constant timing offset bias was present due to the calibration and/or
4.4. SYNCHRONISATION ACCURACY OF 1 PPS SIGNALS 91
0 2 4 6 8 10 12 14 16 18 20 22 24
Elapsed Time (Hours)
450
500
600
700
800
450
500
600T
ime
Off
set
(ns)
Max = 764 ns
Min = 507 ns
Instant offsetMoving Average
Figure 4.13: Time Offset Between Receivers of Different Models Over a Long Period.
Table 4.5: Relative Offset in ns Between Receivers of Different Models.
Test Session Peak Mean STD RMS24 hrs ±764 610 21.9 610
0 2 4 6 8 10 12 14 16 18 20 22 24
Elapsed Time (Hours)
-200
-150
-100
-50
0
50
100
150
200
Tim
e O
ffse
t (ns
)
Max = 154 ns
Min = -103 ns
Instant offsetMoving Average
Figure 4.14: Time Offset Between Receivers of Different Models Over a Long Period.
algorithm within the two particular receivers. It is expected that such a bias could be
removed by a further calibration process.
92 CHAPTER 4. GNSS TIME SYNCHRONISATION IN VANET
Table 4.6: Relative Offset in ns Between Receivers of Different Models.
Test Session Peak Mean STD RMS24 hrs ±154 .5 21.9 21
4.5 Summary
It is observed the results of experiments that using the same model of GPS receivers
in a network enables more accurate time synchronisation than using different model
GPS receivers. In comparison with the experimental results from GPS receivers of the
same model, time synchronisation errors almost doubled when GPS receivers from
different vendors are used. This is observed from the above-described practical exper-
iments, and is not claimed for all VANET scenarios. It is inferred that GPS receivers
from different vendors may not adopt the same error models and mitigation algo-
rithms in their receiver navigation processors. It is understood that the amplification
of relative PVT errors can be minimised if the interoperability requirement for vehicle
GNSS receivers is addressed appropriately [ARRB-Project-Team, 2013]. Nevertheless,
the observed timing errors of tens of nanoseconds can be accommodated for most
VANET applications with strict time synchronisation accuracy requirements.
Consumer-grade receivers are low-end GNSS devices. Such receivers use C/A
code (Single band (L1) Coarse Accusation code) for PVT solutions [Misra and Enge,
2006]. Clocks in these receivers with inexpensive quartz oscillators are responsible for
receiver clock skews, drifts and noises. In our experiments, we have recorded relative
time offsets between two consumer grade receivers. The maximum time offset in our
tests is 180 ns, which is the combined effect of individual receiver hardware delays,
clock variations, and noises. It is worth mentioning that this relative-receiver time
offset is not the timing parameter to be used for further positioning and velocity
calculation but is proposed as a measure of time synchronisation capacity between
GNSS receivers that support VANET applications and services.
Chapter 5
GNSS Synchronisation with On-board Devices
The previous chapter identified promising aspect of GNSS time synchronisation in
vehicular networks and assessed feasibility through compatibility, availability and
accuracy experiments. In this chapter, the aim is to integrate GNSS time synchronisa-
tion into on-board portable devices and examine achievable network-wide synchroni-
sation accuracy. Therefore, this chapter starts with a review of previously-suggested
time synchronisation techniques. It discusses the challenges faced by the suggested
techniques and proposes an architecture for GNSS services. Subsequently, a series of
experiments and validation is performed to demonstrate the achievable accuracy of
the system.
Some parts of the content from this chapter has contributed to the following pub-
lications:
1. K. F. Hasan, Y. Feng, and Y.-C. Tian ”Exploring the Potential and Feasibility of
Time Synchronisation using GNSS Receivers in Vehicle-to-Vehicle Communica-
tions,” In Proceedings of ITM 2018, Reston, VA, Jan 29-2 Feb, 2018.
93
94 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
5.1 Problem Definition and Solution Approach
The concept of Vehicle to Everything (V2X) communication refers to the message
transfer between vehicles and any other objects in an Intelligent Transportation Sys-
tem (ITS). In connected vehicle technology, communications includes interactions
between vehicles, vehicles with infrastructure including Road Side Units (RSU) to
receive network coordinating messages and Internet services, with smart power grids
to charge vehicles or at home for supplementary charge support. This communication
also extends to communication devices to connect pedestrians, cyclist, etc. as shown
in Figure 5.1.
V2V
V2P
V2D
V2
I
V2I (V2H, V2G )
Figure 5.1: VANET Communication: Vehicle to Everything Scenario.
The basic communication, however, is considered as Vehicle to Vehicle (V2V)
and Vehicle to Infrastructure (V2I) communication. The enabling technology of V2V
and V2I communication is known as Dedicated Short Range Communication (DSRC)
technology. The 75 MHz spectrum in the 5.9 GHz frequency band has been allo-
cated to DSRC communication technology, which offer low latency communication
5.1. PROBLEM DEFINITION AND SOLUTION APPROACH 95
1 4
32
5
6
V2I scenario V2V scenario
Figure 5.2: VANET Communication: Vehicle to Infrastructure (V2I), and Vehicle to Vehicle(v2v) Communication.
links in vehicular environments. In respect to the variation of V2V and V2I, it can
be seen that V2I is infrastructure oriented that resembles a generic WLAN infras-
tructure having access points (AP) and stations (STAs). The stations are connected
to an Access Points to form a Basic Service Set (BSS)1. In contrast, the V2V relies
on ad-hoc communications, where there is no fixed infrastructure as explained in
Figure 5.2. In VANET, the idea of time-synchronisation concepts were adopted from
the existing synchronisation practices in WLANs (i.e., Wi-Fi). The VANET standard
802.11p is an amendment of 802.11, which is the generic WLAN standard. The time-
synchronisation mechanism in 802.11 relies on the Time Synchronisation Ftion unc-
tion (TSF) timer, which is essentially a register in MAC layers capable of performing
216 modulus counting with a resolution of 1 s, that update with the received broadcast
messages (i.e., beacon or probe messages) timing information. Such messages are
transmitted by a network element that recognises the sources of the common clock
for that particular BSS. In infrastructure mode of operation, APs are considered as
the sources of a common clock. In ad-hoc mode, station nodes (STA) send time
frame messages to enable synchronisation each other. A number of TSF-based time-
synchronisation method have been proposed and implemented in WLANs. The TSF
is a register responsible for coordinating the channel access of the wireless medium, it
1Basic Service Set (BSS) is a group of IEEE 802.11 stations coordinated and configured by an AccessPoint (AP) to communicate with each other over using wireless medium [Kenney, 2011]
96 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
Application
TSF Register
Application
TSF Register
RSU/AP OBU/STATA
Frame
Ref clock
1 2
34
TA
Frame
(a)
(b)
Data 12:00 am Hardware Medium
TSF
Data 12:00 am Hardware
TSF
(c)
Applications
Figure 5.3: Existing Problem with Time-Advertisement-based Time Synchronisation. (a)BSS Communication, RSU Sending Beacon Containing TA Frame for Synchronisation. (b)TA frame is Transmitting from RSU to OBU. (c) Time Development (transfer) in TA process.
is, however, not meant to be used by the application layer of the system for any other
system and application purposes. In order to support time synchronisation across
other layers and applications, a 2012 IEEE 802.11 amendment proposed a method
called Timing Advertisement (TA).
In Timing Advertisement (TA) mechanisms, the AP sends beacon messages, which
contain time advertisement frames having both the time of the local clock and the
offset between the local clock and the global standard of time as ’time value’ and
’time error’ fields, respectively. The beacon or the probe messages can also use these
5.1. PROBLEM DEFINITION AND SOLUTION APPROACH 97
1 2
34
TA
Frame
Figure 5.4: Undefined Situation Using TA Mechanism in Pure Ad-hoc Communication.
two parameters to introduce time synchronisation under a single BSS. STA node re-
vives the signal and updates calculating the timing information sent across as shown
in Figure 5.3. However, it is clear that such synchronisation mechanisms depend
entirely on the in-band data communication between AP and STAs, thus relying on
accurate computational performance and also adding overhead to the transferable
payloads. In VANET, a significant part of network nodes are used for ad-hoc V2V
communications. In comparison, infrastructure-based V2I communications resem-
ble AP to STA communications, where the synchronisation signals are provided by
the infrastructure. This make VANET communications significantly different from
WLANs.
Along with the time-transfer primitives TA, the 2012 IEEE 802.11 standard in-
troduces another mechanism, known as the Timing Measurement (TM) frame, that
allows individual STA (i.e., communicating nodes) to measure the relative offset be-
tween STAs and take any action necessary. At this point, it is not clear that how
this relative time measurement would be calculated and what degree of accuracy can
be achieved. The 2016 IEEE 802.11 standard, introduces another frame, named Fine
Timing Measurement (FTM), to maintain estimation of relative-time offsets between
98 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
AP and/or STAs. The standard recommends that this time offset is achieved through
Round-Trip-Time (RTT) calculations by regular transfer of messages between commu-
nicating nodes. In view of communication mode, the timing advertisement works on
forward broadcasting messages, whereas Fine Timing Management (FTM) is based
on back and forth message transmission mechanism between network elements. At
this stage, the conditions under which the above methods would be employed for
network synchronisation remain unclear. In addition, the performance of TA and TM
methods have been difficult to confirm, as work has been scarcely reported by the
scientific research community [Mahmood et al., 2017].
In a vehicular network, the WAVE MAC standard IEEE 802.11p adopts time-
synchronisation principles following IEEE 802.11 family standard, as stated above.
In association with 802.11p, WAVE defines a MAC extension (IEEE 1609.4) as shown
in Figure 5.5 to enable multi-channel operation in a vehicular context. This time-
synchronisation extension accommodates any open technology that can provide a
common time such as GPS [SCC32, 2006]. However, it is not clear how overall system-
level time synchronisation in WAVE devices would be achieved and what level of
accuracy can be maintained. In infrastructure mode, a node can be synchronised
with a RSU, but in ad-hoc based V2V communications, it remains unclear whether
such communications are feasible and can be studied. A number of available of ad-
hoc MANET protocols for WSN are discussed in Chapter 2. But those protocols are
ill-suited to VANET, given its high-mobility, short-range communications, and end-
to-end latency requirements.
Instead of TSF synchronisation, a complete GNSS-time synchronisation method
for VANETs is proposed as shown in Figure 5.6. In this method, individual nodes
obtain synchronisation using external GNSS receivers as their reference clocks. This
enables both applications and TSF registers for multi-channel operations to be syn-
chronised with an external GNSS reference clock.
5.2. RELATED WORK 99
PhysicalPLME
MAC
LLC
MLME Ext
WSMPUDP/TCP
IPv6
Upper Layers
Managment Plane Data Plane
Man
agem
ent E
ntit
y
Multi Channel Operation
802.11p
802.11p
MLME
1609.4
Figure 5.5: Wireless Access for Vehicular Environment (WAVE) Layers.
5.2 Related Work
Previous researchers have envisaged the possibility of GNSS time integration in VANET.
As mentioned in the literature survey, references [Scopigno and Cozzetti, 2009] and
[Cozzetti et al., 2011] have proposed the idea of GNSS time integration into vehic-
ular network scenarios. Reference [Cozzetti et al., 2011], suggested a tight-coupling
model, in which navigational data and timing data from a GNSS system are inte-
grated into on-board communication module of an intelligent transportation system
(ITS ) for road navigation and synchronisation. Although this article argued the
system integration possibility, it did not mention any experiments that demonstrate
its accuracy and feasibility.
Some European researchers have carried out scholarly work on GPS integration
in interconnected vehicle solutions. Under the Fleetnet Project [Enkelmann, 2003],
decentralised time synchronisation for vehicular ad hoc networks have been studied,
in which the concept of GPS time integration with slotted MAC protocols has been
proposed [Franz et al., 2005]. Sjoberg et al. 211 investigated 802.11-based MAC limita-
tions and proposed Self Organised Time Division Multiple Access (STDMA) in place
of Carrier Sense Multiple Access (CSMA) [Bilstrup et al., 2009, Bilstrup et al., 2010,
100 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
Application
TSF Register
OBU/STA
Application
TSF Register
OBU/STA
GPS Ref clock GPS Ref clock
System 12:00 am Hardware Medium
TSF
Data 12:00 amHardware
TSF
GPS GPS
OBU/STA OBU/STA
Figure 5.6: Proposed Solution. GNSS Time Synchronisation to TSF Register ThroughApplication Layer.
Sjoberg et al., 2011]. The efficiency of a TDMA-based access schedule requires precise
and accurate time synchronisation, and therefore, they suggested the use of GPS as
an external time source. [Morgan, 2010] proposed synchronising DSRC Network
Interface Controllers (NICs) with GNSS signals. He suggested that the DSRC NIC
Clocks adopt the phase and frequency of the PPS signal from GPS receivers. But
there were no clear guidelines mentioned about the DSRC layer in which GNSS time
would be integrated.
Network synchronisation experiments using GPS-PPS timing signals have been
conducted by previous researchers. Recently [Guo and Crossley, 2017] described an
experiment that uses GNSS signals for time synchronisation in a power transmission
sub-station. Although the major portion of the electric power station network is based
indoors, they blended GPS and PTP to achieving a complete timing solution. In
contrast, VANETs are mostly outdoor networks, where GNSS signals are expected to
be widely available. [Fraleigh et al., 2001] conducted an experiment in an IP backbone
5.3. TESTBED DEVELOPMENT 101
network where they have used special network card DAG for synchronisation with a
GPS reference clock. [Mangharam et al., 2007] proposed the concept of synchronising
the operating system clock to a GPS PPS signal in vehicular environments. This idea
is in-line with the work presented herein, however, their experiments achieve sub-
200µs local synchronisation accuracy.
5.3 Testbed Development
A testbed was developed using a single board computer (SBC), namely, a Raspberry
pi-3, running a Linux-based operating system, Jessie (kernel 4.4). The SBC was se-
lected because of its simplicity and ease of integration with vehicle communication
systems. It provides a GPIO port, which enables a straightforward way to integrate
an external GPS module with the system. This involved integrating a PPS-enabled
GPS receiver from UBLOX (Model: Ublox MAX-M8Q).
The selected GPS receiver has two distinct outputs: GPS data which contains
NMEA sentences and a PPS pulse signal. The PPS-API driver was installed and
enabled on a node. Its primary function was to timestamp external events with a
high resolution. In this experiment, it was used for connecting GPS receivers with the
node. The kernel of the operating system was configured with the PPS-API patch.
The PPS-API fetches the PPS signal from DCD pin of the serial port, evaluates the
offsets between the system clock and the PPS reference clock and passes it to the time
daemon. In addition, the kernel of the operating system was compiled to fetch the
driver closer to the daemon. A time daemon was used for locking the system clock
with reference clocks. It also monitored various timing statistics. There are several
time daemons that can be used for this experiment. The Network Time Protocol
Daemon (NTPD) was used herein. The NTPD is sensitive to the availability of GPS
connections. When a GPS signal is disrupted, the NTPD exhibits deflections in its
output. Statistical results were recorded by generating log files. Since the PPS pulse
does not contain any information about the absolute time (i.e., second, minute, hour,
102 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
Driv
ers P
PS
AP
IN
ME
A
hw
clo
ck
syste
m c
lock /
swclo
ck
Ref c
lock
start
stopTO
D
GP
S
PPS N
TP
Ap
plic
atio
n
TS
F
12
Devic
e In
terfa
ce
kN
TP
D
Hard
ware
Abstra
ction L
ayers
Є
Syste
m C
lock
TS
F
User Sp
ace
Ke
rne
l Space
Ap
plic
atio
ns
(a)
(b)
NT
PD
MA
C
Ref c
lock
1 2 34
5
Figure5.7:
Com
municating
Node
developmentusing
aG
NSS
Receiver.(a)Schem
aticD
iagramofthe
Processto
Synchronisethe
SystemC
lockofthe
LinuxBased
Nodes
(Raspberry
Pi-3)tothe
GN
SSR
eceiver.(b)LayeredV
iewofthe
Developm
ent.This
FigureShow
sD
ifferentClock
Timing
andalso
DifferentJitter
throughouttheSystem
.
5.4. RESULT ANALYSIS AND PERFORMANCE EVALUATION 103
Rpi Node GPS Receiver
Power
Supply
Antenna
System Setup
Antenna cable
Figure 5.8: Schematic Diagram of the Experimental Setup on Moving Node to CollectData on Real-time Vehicular Environment Including Urban, Suburb, Highway, etc.
Route.
day, month and year). It just produces a tick every second, which marks the beginning
and/or ending of an arbitrary second. A GPS NMEA driver was used to receive
NMEA sentences and obtain the real physical time, which was then locked to the
PPS signal. Another external time source such as the Internet wall clock could be
employed, but NMEA receiver outputs were used here to make the node independent
and standalone.
A systematic development approach is shown in Figure 5.8. In Figure 5.8 (a) the
system development among different clock systems integrated with GPS ref clock
is shown. Equivalent layered development is shown in Figure 5.8 (b) where the
development space both in user space and kernel space is indicated. The figure also
shows different clock timing and different jitter throughout the system.
5.4 Result Analysis and Performance Evaluation
The overall experiment involved three distinct stages. In the first stage, a laboratory
experiment was conducted to measure the timing accuracy exhibited by a GPS re-
ceiver and the integration capabilities of a host node to synchronise it’s clock with
it. The measurements included the accuracy of the time from NMEA data, the PPS
104 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
0 50 100 150 200 250 300 350 400 450 500
No. of Observation in 10h
-0.02
-0.01
0
0.01
0.02
Offse
t (s
)
Time from RS232 messages
Max = 9.2 ms
Min = -8.7 ms
Instant offset/JitterMoving AverageAverage
0 200 400 600 800 1000 1200 1400
No. of Observation in 24h
-0.02
-0.01
0
0.01
0.02
Offse
t (s
)
Time from RS232 messages
Max = 9.2 ms
Min = -20.07 ms
Instant offset/JitterMoving AverageAverage
Figure 5.9: System Clock Synchronised with GPS Data Signal.
5.4. RESULT ANALYSIS AND PERFORMANCE EVALUATION 105
Algorithm 1 GPS Synchronised Node Development on RPi
1: Start2: Initial Environment Setup
Require: picocom, ntp, libcap-devinstall← picocominstall← ntpinstall← libcap-dev
3: Freeing Serial Port from Consolestop← [email protected]← [email protected]
4: Update and Upgradeinstall← apt-getupdate← apt-getdist-upgrade← apt-gethold-upgrade← ntp
5: Configuring the Noderemove← console=ttyAMA0,115200include← pps-gpio (modules)include← dtoverlay=pps-gpio (config)include← gpiopin=18 (config)
6: Enabling NTPD that supports PPSRequire: ntp-4.2.8p10, ntpstat, ntpdate,libssl-dev
install← ntp-4.2.8p10install← ntpstateinstall← ntpdateinstall← libssl-dev
7: NTPD compilation with PPSRequire: bc, ncurses-dev
install← bcinstall← bc ncurses-devgit clone← address: https://github.com/raspberrypi/linuxenable← Old Idle Dyntics configenable← PPS kernel consumer support
8: Configuring NTPDEnable Local Server, stratum-10server← 127.127.1.0fudge← 127.127.1.0Enable PPSserver← 127.127.22.0 mode 17 iburst true preferfudge← 127.127.22.0 flag1 1Enable GPSserver← 127.127.28.0 mode 17 iburst true preferfudge← 127.127.28.0Enable GPS-PPSserver← 127.127.20.0 mode 17 iburst true preferfudge← 127.127.20.0 flag1 1
9: Setting up Ethernetset← IP
10: End
106 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
output and the accuracy achievable by locking them together. In the second stage, the
developed node was integrated into a car and a field test was conducted. This enabled
the assessment of the developed GNSS-based time synchronisation system in terms of
signal availability and environmental parameters including vehicular speed. In the
third stage, laboratory experiments were conducted with three synchronised GNSS
nodes to determine relative time accuracy between them.
5.4.1 Node Clock Synchronisation With GNSS Receiver
To evaluate and analyse the performance of GNSS time integration within the devel-
oped node, long-term laboratory tests and on-road field tests were also conducted.
A series of long-term timing-accuracy measurements were conducted in a laboratory
environment. The overall evaluation process followed two steps. In the first step,
signals were recorded from individual GPS receivers, i.e., the timing information
received from RS232 messages (through a UART port) and PPS signals (from DCD
pins), which are plotted in Figure 5.9 and Figure 5.10, respectively. In the second step,
both GPS message times and the PPS signal are bound together to lock the system
clock. The resulting clock frequency is locked to the PPS ticks and the real physical
time received through the RS232 NMEA messages.
To understand the effect of temperature on the systems reception of the RS232
message stream, experiments were conducted in two different environments. One
experiment was conducted in a temperature-controlled room with 16◦C (top plot of
the Figure 5.9) and the other was conducted at normal room temperature (20◦ −
25◦C) as shown in bottom plot Figure 5.9. It can be seen from the log data that
temperature does impact on the message stream performance. In the low temperature
environment, low average offsets of 75 µs were recorded over 10 hours.
The second graph of the Figure 5.9, shows a 5.2 ms time-offset variation over a
24 hour period. In both cases, significant and random time offsets can be observed.
From the dataset logged at room temperature, the maximum recorded offset was +9.2
5.4. RESULT ANALYSIS AND PERFORMANCE EVALUATION 107
0 200 400 600 800 1000 1200
No. of Observation in 24h
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2O
ffset
(s)
PPS ticks at Node
Max = 1.11 s
Min = -1.82 s
Instant offset/JitterMoving AverageAverage
Mean= -42ns
Figure 5.10: System Clock Synchronised with PPS Ticks.
ms (-20.7 ms in negative scale). The positive and negative offset values show that the
GPS times and system clocks can alternately lead or lag each other. The average value
of 5.2 ms is high enough in practice and therefore does not meet the requirements for
many vehicle network applications.
The later part of the first stage evaluation involved another long-term measure-
ment of PPS signals. This results are plotted in Figure 5.10. Compared with the GPS
RS232 data stream, it can be seen that the PPS clock is highly accurate and exhibits an
average offset of 42 ns. From the 24-hour-measurement duration, a maximum of 1.11
µs (1.82 µs in negative scale) can observed. Note that VANETs are real time networks,
where physical time plays a significant role in many applications. The PPS signal does
not carry any timing information. Therefore, depending exclusively on PPS signals
does not provide a timing solution for vehicular environments. Consequently, a
secondary source is required for accurate estimation of the physical time, i.e., the time
of day (TOD). Some researchers [Ben-El Kezadri and Pau, 2010, Guo and Crossley,
2017, Mangharam et al., 2007] have conducted experiments in which TOD timing
information is sourced from Internet, which requires coverage by another network to
provide Internet access. However, as VANETs are outdoor networks, the provision
108 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
0 200 400 600 800 1000 1200 1400 1600 1800 2000
No. of Observation in 24h
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5O
ffset
(s)
GPS PPS time locked at Node
Max = 2.16 s
Min = -2.06 s
Instant offset/JitterMoving AverageAverage
Mean= 127ns
Figure 5.11: System Clock Synchronised with Both GPS PPS Signal.
of continuous Internet access is not practical on-board and should not be considered
as a part of the solution. Therefore, the experiment described herein combined the
TOD (i.e. UTC) from GPS RS232 data streams with a PPS signal. Figure 5.11 shows
the performance of the developed GPS-PPS-locked clock system over 24 hours. It
can be seen that the Peak-to-Peak maximum offset is 4.22µs (2.16µs+2.06µs). That is,
any two such networks can be synchronised with an accuracy of 4.22 µs (max). The
statistical distribution of the observed offsets is shown in Figure 5.12. It can be seen
that the distribution of the measured data corresponds to Gaussian distribution. The
standard deviation equal to 0.69µs.
Clock stability and noise level can be characterised by the so-called Allan Variance
(AVAR). The Allan variance is a recognised tool for time-domain frequency stability
measurement and can be derived from the expected value < . > of normalised fre-
quency y. It is calculated over a period of time (τ) and can be expressed as [Loschmidt
et al., 2012];
σ2y =
12〈(yn+1 − yn)
2〉 (5.1)
5.4. RESULT ANALYSIS AND PERFORMANCE EVALUATION 109
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Offset (s) 10-6
0
50
100
150
Den
sity
Offsets
Normal Distribution
Figure 5.12: Statistical Distribution of Measured Time Difference.
Equation 5.1 is a characterisation of clock stability. It provides an estimate of sta-
bility due to the inherent noise processes with the clock. The calculated Allan variance
serves as an indicative parameter of clock stability and a low value indicates high
stability. The Allan Deviation (ADEV) is the square root of the AVAR is expressed as:
σy(τ) =√
σ2y (τ) (5.2)
There are different, alternate methods for estimating the ADEV. Overlapped ADEV
methods provide better noise estimation performance than non-overlapped ADEV
calculations. It averages the normalised frequency over blocks of n data samples. For
this experiment, the Overlapped ADEV was estimated in MATLAB and the (σy) is
plotted in Figure 5.13. The noise template (κτµ) slope and sum of tangents (∑ κτµ) are
also plotted. From the slope analysis, white (thermal) noise (µ=-.5), pink (Electronic)
noise (µ=0) and red (Angle rate) noise (µ=.5) are calculated, which indicate that the
developed GPS-PPS-locked clock system exhibits good stability.
110 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
100 101 102 103
[s]
10-9
10-8
10-7
10-6
y
Allan deviation [overlapping]
y( )
k k
k - 5 11.2301 897.0000 2.1288e-07
0 897.0000 897.0000 7.1077e-09 0.5 897.0000 897.0000 2.3732e-10
µ τ min τ max
Noise Identification
Figure 5.13: Illustration of Clock Stability and Noises using Allan Deviation (Log-Log scale).
5.4.2 Field test of GNSS Time Synchronisation
The results of a group of experiments conducted on different road scenarios are plot-
ted in Figure 5.14. It is interesting to see that the clock binding performance on roads
is better than in the afore-mentioned laboratory experiments. The observed maxi-
mum offset between reference GPS and system clocks was 1.62µs, with an average of
533 ns over 30 minutes of logged data within suburban (40-60 km/hr speeds) traffic
scenario. In highway traffic (with 80-100 km/hr speeds), the observed maximum
offset was 2.04µs, with an average of 495ns over 30 minutes of logged data. In a
mixed suburban and city environment (40 to 60 km/hr speeds), a maximum 2.17µs
deflection and a peak-to-peak difference of 4.07µs (max) were observed. It is assumed
that better signal availability (through the GPS RS232 data stream) led to the overall
performance improvement. These results also suggest that higher vehicle speeds (up
to 100 km/hr) do not result in significant performance degradation.
5.4. RESULT ANALYSIS AND PERFORMANCE EVALUATION 111
0 20 40 60 80 100 120 140 160 180
No. of Observation in 30 mins
-2
-1
0
1
2
Offs
et (s
)
10-6 Suburb 40-60 km/hr
Max = 1.62 s
Min = -1.49 s
PPS-GPS Mean
0 20 40 60 80 100 120 140 160
No. of Observation in 30 mins
-3
-2
-1
0
1
2
3
Offs
et (s
)
10-6 Highway 80 km/h
Max = 2.04 s
Min = -2.38 s
PPS-GPS Mean
0 50 100 150 200 250
No. of Observation in 2 hours
-3
-2
-1
0
1
2
3
Offs
et (s
)
10-6 Mixed Environment with suburb and city
Max = 1.90 s
Min = -2.17 s
PPS-GPS Mean
533 ns
495 ns
372 ns
Figure 5.14: GPS-PPS Enabled Clock in Different Road Scenarios.
112 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
C1 C2
C3
Figure 5.15: Schematic Diagram of the Experimental Setup to Measure the Time Offsetsbetween Two GNSS Synchronised Computing nodes. Here, Node C3 is Acting as a Server toSend Messages to the Other Two Clients, C1 and C2.
5.4.3 Network Synchronisation with GNSS
This phase of the experiment investigates the timing accuracy between two GNSS-
synchronised nodes that were developed according to the phase one experiment de-
tailed above. The experimental setup involved one server node and two client nodes.
The nodes were situated sufficiently close to each other to communicate wirelessly, as
shown in the Figure 5.15.
Figure 5.15 depicts a server node C3 sending UDP broadcast messages that are
received by clients C1 and C2, which are both synchronised with identical-model
GNSS receivers. The central idea is that, the server broadcasts messages to identically-
configured receivers, which are assumed to received at the same time. As they receive
the message at the same time, the nodes individual time-stamps would yield their
5.5. SUMMARY 113
clock times. In order to send and receive UDP messages, a simple socket program was
written in python. The tcpdump was used to record the time-stamps and compare the
clock times. The experiment was implemented using the Raspberry Pi-3 integrated
built-in wireless support (802.11n).
Time-stamps were collected from the received packets of individual nodes in three
datasets: 10 packets, 100 packets per second and 300 packets per minute.
To understand the comparative distributional characteristics the results of the
experiments are presented in the box-plot provided in Figure 5.16. For the 10-packet
dataset, it can be seen that there is an inter-quartile spread of approximately 6 µs,
a median offset of 1 µs and a maximum offset recorded here is 8 µs. Similarly, the
100-packet dataset shows a 5 µs inter-quartile spread, a median offset of 1.5 µs and
a maximum offset of 6.5 µs. The 300-packet dataset, yields more variation, namely,
offsets ranging from 7.5 µs to 10.75 µs , although the 50% (inter-quartile) values reside
between 1.5 µs to -3.6 µs. The median value is minimal at 250 ns.
To understand individual differences between packets, the 300 packet dataset is
plotted in Figure 5.17. From the figure, the presence of outliers can be seen over 8 to 10
µs. In this experiment it is assumed that the receiver systems are identical, therefore,
the timing jitter and time-stamping should be same. In practice, small parametric
differences can exist within the receiver components. Thus, the measured offsets are
not exact but only approximate the time offsets between the nodes.
5.5 Summary
This chapter has presented an experimental verification of a GNSS solution for stable
and accurate time synchronisation in vehicular networks. It started with a review of
previously-proposed VANET time-synchronisation mechanisms and explained their
limitations. Subsequently, an application-layer GNSS time integration is proposed,
where the system clock frequency is guided by both the PPS signal and real physical
114 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
0
-5
-10
5
10
15
-15
10 Packets 100 Packets 300 Packets
Mic
rose
co
nds
Figure 5.16: Box-plot of the Time Offsets Between Two GNSS Synchronised Node Developedon Rpi. Three boxes were Generated from 3 sets of data: 10-packet, 100-packet and 300-packetdatasets.
0 50 100 150 200 250 300 350
Number of Packets
-12
-10
-8
-6
-4
-2
0
2
4
6
Tim
e O
ffse
t (m
icro
seco
nd)
Time offset between two GNSS synchronized nodes
Instant offsetMoving Average
Figure 5.17: Offsets Between Individual Packets and their Moving Average from the Datasetof 300 Packets.
5.5. SUMMARY 115
time obtained from the GPS RS232 data stream. The operation of a mobile node that
exploits the developed GPS-PPS-time-synchronisation system is also described. The
results of a series of experiments are then presented which determine the accuracy
and precision of the timing signal.
The experimental results from both laboratory and field tests show that a PPS-
enabled-GPS integration implemented on an on-board unit (OBU) can achieve tight
coupling with UTC at a maximum clock deviation of 2.16 µs. From the experiments
described in the previous chapter it is understood that a GNSS receiver is capable of
providing UTC synchronisation with an accuracy of 40 ns. However, when GNSS
receivers are integrated with communications devices having different physical and
operating layers, timing jitter is generated, which results in synchronisation offsets.
On a Raspberry pi-3 platform, it was possible to provide 2 µs synchronisation accu-
racy with a median value of around 500 ns. Since there is no direct way to accurately
compare the clock of two communication nodes while avoiding jitter, an experiment
was conducted in which two identical systems time-stamped the same events. It was
assumed that the communication paths were symmetric, therefore, messages sent
from one server node are identically received at two client nodes. This experiment
showed that two-GNSS-synchronised nodes exhibit a relative time offset of sub-10
µs. Larger time offsets would probably be experienced in practice due to the presence
of some timing jitter. Overall, it has been demonstrated that the developed GPS-
PPS-time-synchronisation system can better support vehicular networks than existing
systems.
116 CHAPTER 5. GNSS SYNCHRONISATION WITH ON-BOARD DEVICES
Chapter 6
Synchronisation in Occasional loss of GNSS
Signals
The previous chapters addressed the feasibility, integration and performance of GNSS
time synchronisation systems for vehicular communication networks. This chapter
analyses time synchronisation performance limitations during GNSS time solution
outages using data collected from vehicles travelling along high-rise streets in the
Brisbane CBD. The analyses show that indicates that time drift solution can be suc-
cessfully handled and potential strategies can be developed to bridge time-solution
outages. A laboratory experiment also conducted to understand and mitigate the drift
rate of node clocks during the absence of GNSS signals such as the tunnel scenario.
Some content from this chapter has contributed to the following publications:
1. K. F. Hasan, Y. Feng, and Y.-C. Tian ”GNSS Time Synchronisation in Vehicular
Ad-hoc Networks: Benefits and Feasibility,” in IEEE Transaction on Intelligent
Transportation System, 2018 DOI: 10.1109/TITS.2017.2789291.
117
118CHAPTER 6. SYNCHRONISATION IN OCCASIONAL LOSS OF GNSS SIGNALS
6.1 Understanding Service Availability
VANET is a highly dynamic and time-sensitive network. The nodes and network
elements in a VANET periodically send beacon signals to communicate updates about
status information on the road. To maximise communication efficiency, all nodes need
to maintain common time synchronisation. However, GNSS-based time synchroni-
sation relies on receiving signals from overhead satellites, which requires near Line
of Sight (LOS). Poor GNSS signal reception can occur in buildings, tunnels, under
trees and around other obstacles which are frequently encountered in urban road
scenarios. Therefore, it is important to understand the efficacy of GNSS-based time-
synchronisation systems operating in GPS-challenged environments.
GPS is a recognised tool for PNT (Positioning, Navigation, and Timing) solution.
With the standard positioning services (SPS) with pseudo-range measurements, the
precision depends on the number of satellites visible at the receiver also depends
on the geometry. The satellite/user geometry measurement metric is Dilution of
Precision (DOP). The DOP relates the geometric strength of navigational satellites
of GPS systems on the accuracy of a navigation solution.
To understand the issue of availability of a GNSS based time synchronisation
system, an experiment was conducted along a cluttered road in the Brisbane CBD.
In this experiment, a vehicle travelled past high-rise buildings, under tree canopies
and short overpasses. The vehicles paths are shown in Figure 6.1.
Figure 6.2 illustrates the different sorts of obstacles encountered on the vehicle
paths, such as high-rise building in Figure 6.2 (a) and (b), a short overpass in Fig-
ure 6.2 (c) and trees in Figure 6.2 (d). All of them are potential obstacles of barrier to
the GNSS signals.
The primary objective of this study to understand that to what extent the shielding
of GNSS signals by different urban structures on and around the roads impacts on
6.1. UNDERSTANDING SERVICE AVAILABILITY 119
the availability of GNSS time solutions. In this experiment, consumer-grade GNSS-
receivers were used to record data at the rate of 10 Hz. The open source RTKLIB was
used to process the collected data. In this experiment, we particularly focused on two
GNSS constellations, GPS and BDS (Beidou).
Figure 6.3 (a) (b) and (c) show the vehicle tracks reported by GPS, BDS and GPS+BDS,
respectively. It is evident from the individual GNSS vehicle tracks that some discon-
tinuities are present. The discontinuities are the result of position outages due to the
signal blockages and incorrect position solutions due to the weak geometry. Figure 6.4
plots the number of GPS and BDS satellites that were visible throughout the vehicle
trajectories over elevations of 10 degrees or more. It can be seen that there was at
least one satellite in view throughout the vehicles travel. For the constellations of
GPS, BDS and GPS+BDS, respectively, Table 6.1 summaries the percentages of the
satellite visibility in three cases: the number of satellites is more or equal to 4 (NSAT
≥ 4), between 1 and 3 (NSAT = 1 3) and is zero (NSAT < 1).
The Geometric Dilution of Precision (GDOP) values for NSAT ≥ 4 are listed in
Table 6.1 (b). From the table it can be observed that the GDOP ≤ 6, the parameters
of 4D states (for both position and time) in our case of high-rise urban areas are
low, i.e., 49.6%, 33.59% and 80.25% with GPS, BDS and GPS+BDS constellations,
respectively. The visibility for a minimum of 1 visible satellite for the same three
constellations are 100%, 99.98%, and 100% respectively. These results imply that
the availability of GNSS time solutions can be much higher than the availability of
positioning solutions within the high-rise streets, provided that the position states
can be fixed to a certain accuracy with alternative technologies. Overall, the use of
GNSS for time synchronisation in VANET is more feasible in urban areas than for
vehicle positioning in VANET.
120CHAPTER 6. SYNCHRONISATION IN OCCASIONAL LOSS OF GNSS SIGNALS
(a)TopV
iewofthe
Projectedarea
inBrisbane
CBD
.(b)Selected
Route
tocollectD
atain
BrisbaneC
BD.The
Red
Marks
Shows
theD
ataPoints
onthe
Route
Figure6.1:
GPS
Challenged
Environmentin
Urban
Concept.Sun
anEnvironm
enthasD
ifferentSortofSignalBlockage/Lim
itingelem
ents,forexam
ple,High
Rising
Building,Trees,ShortOver
Pass,etc.
6.1. UNDERSTANDING SERVICE AVAILABILITY 121
(a)T
opSt
reet
Vie
ww
ith
Hig
hR
isin
gBu
ildin
gs.
(b)S
tree
tVie
ww
ith
Hig
hR
isin
gBu
ildin
gs.
(c)S
tree
tVie
ww
ith
Shor
tOve
rPa
sson
the
Trac
ked
Rou
te.
(d)
Stre
etvi
eww
ith
Hig
hR
isin
gBu
ildin
gan
dTr
ees
onth
eTr
acke
dR
oute
.
Figu
re6.
2:G
PSC
halle
nged
Envi
ronm
enti
nU
rban
Con
cept
.Diff
eren
tObs
tacl
eson
the
Rou
tear
ePr
esen
ted.
Pict
ures
are
Take
nfr
omG
oogl
eMap
.
122CHAPTER 6. SYNCHRONISATION IN OCCASIONAL LOSS OF GNSS SIGNALS
BDSGPS GPS+BDS
Figure 6.3: Vehicle tracks of of GPS, BDS and GPS+BDS on High Rising Urban roads inBrisbane.
5:16 5:20 5:25 5:30 5:35
2468
10
1214161820
GPS + BDS
No
of S
atel
lites
Elapsed Time
Figure 6.4: The number of satellites under the signal coverage of BDS and GPS.
Table 6.1: The number of satellites available with different GNSS service constellation.
GNSS System NSAT ≥ 4 NSAT = (1 to 3) NSAT < 1GPS 77.32% 22.68% 0%BDS 82..93% 17.05% 0.02%BDS+GPS 99.25% 0.75% 0%
Table 6.2: GDOP with Sifferent GNSS Services.
GNSS System NSAT< 4 GDOP 6 6 GDOP > 6GPS 22.68% 49.61% 27.71%BDS 17.07% 33.59% 49.34%BDS+GPS 0.75% 80.25% 19.00%
6.2. FALL BACK SOLUTIONS DURING SIGNAL OUTAGE 123
6.2 Fall Back Solutions During Signal Outage
Undoubtedly, the GPS-based time-synchronisation can be one of the most accurate
sources for time dissemination in open environments, and particularly road commu-
nication networks. However, there are still the possibility of complete signal outages
that can occur while travelling in tunnels and under bridges. In such challenged
environments, GPS signal may be intermittent or entirely absent for a while. To fos-
ter a better understanding of GPS-challenged environments and to propose solution
measures, two scenarios are considered below:
1. a signal blockage that causes the number of satellites to be less than four, but
there is always at least one (NSAT=1 to 3), and
2. when there are no satellite signals at all (NSAT=0).
6.2.1 Number of Satellites 1 to 3 (0<NSAT< 4)
It was observed that a high percentage of 1 to 3 visible satellites can occur in high-
rise urban areas. Dropping the number of visible satellites can happen from time to
time - depending on the road clutter conditions. To determine the impact of dropped
satellites, an experiment was conducted in a mixed-road environment. This experi-
ment involved developing a mobile node using the Raspberry Pi 3 (RPi3) platform,
and synchronising it with the GPS signal, as illustrated in Figure 6.5. The RPi3 used
the Linux operating system, Jessie. A consumer grade GNSS receiver manufactured
by Ublox was connected to its GPIO port to synchronise the node clocks with the re-
ceiver. The RPi3 was selected for the experiment because of its simplicity like a DSRC
device, (it is programmable and movable). An ntpd daemon was used to monitor
and record the timing performance on the road. The node was mounted in a car that
travelled for over 2 hours in different road conditions within the Brisbane CBD. The
vehicle environment included dense urban, suburban and highway scenarios and the
speed varied from 40 km/hr to 100 km/hr.
124CHAPTER 6. SYNCHRONISATION IN OCCASIONAL LOSS OF GNSS SIGNALS
0 20 40 60 80 100 120 140 160 180
No. of Observation in 30 mins
-2
-1
0
1
2
Offset (s
)
10-6 Suburb 40-60 km/hr
Max = 1.62 s
Min = -1.49 s
PPS-GPS Mean
0 20 40 60 80 100 120 140 160
No. of Observation in 30 mins
-3
-2
-1
0
1
2
3
Offset (s
)
10-6 Highway 80 km/h
Max = 2.04 s
Min = -2.38 s
PPS-GPS Mean
0 50 100 150 200 250
No. of Observation in 2 hours
-3
-2
-1
0
1
2
3O
ffset (s
)10-6 Mixed Environment with suburb and city
Max = 1.90 s
Min = -2.17 s
PPS-GPS Mean
533 ns
495 ns
372 ns
Figure 6.5: Performances of a Node Clock Integrated with GNSS in a Mixed RoadEnvironment Including Urban, Sub-urban, Low speed and High Speed Areas.
As shown in Figure 6.5, the effects of momentarily reverting to 1-3 satellites is
reflected in the peak-to-peak offset variation of the timing solution. Note that, during
momentary periods when only 1 to 3 satellites are available, the vehicle position states
may not be updated. The projection of the position errors on the line-of-sight paths
follows the direction cosine rule. In the worst case, the position error of 300 m leads
to 1s clock bias. The Figure 6.5 shows the clock error variation between - 2.17 and 1.90
s, while the average deviation of the node clock from the GPS clock is -372 ns. Thus,
GPS can provide valid time solutions during periods where only 1 to 3 satellites are
in view, provided that clock errors of about 2s meet VANET timing requirements.
6.2.2 Number of satellite is Zero (No visible satellites)
Consider the second scenario where no satellites are visible (NSAT=0) for an extended
period. A two-step experiment was conducted to understand the consequences of
complete GNSS signal losses. In this experiment, three Rpi3 were individually con-
nected to three Ublox consumer-grade GNSS receivers as shown in Figure 6.6. Ini-
tially, in Step 1, all of the three nodes were synchronised with GNSS signals. The
6.2. FALL BACK SOLUTIONS DURING SIGNAL OUTAGE 125
node with Clock C3 was used for the server of the reference clock for comparing
the time differences among the three GNSS synchronised systems. Later, in Step 2,
the node with Clock C1 was disconnected from the GNSS antenna to mimic a GNSS
signal loss. The Figure 6.7 shows how the node clock C1 is drifts without the benefit
of a received time-synchronisation signal. Figure 6.7 shows the overall result of the
clock drift over a 4-hour duration.
C1 C2
C3
Ste
p 2
Ste
p 1
Figure 6.6: Schematic Diagram of the Experimental Set-up Between Three Nodes. NodeC3 is used to Record the Data where All of the Nodes are Synchronised with Identical GPSreceivers. GPS Lost is Testing by Removing GPS Device from the Node C1.
0 20 40 60 80 100 120 140 160 180 200
No. of Observation in 4h
-600
-500
-400
-300
-200
-100
0
100
200
Offs
et (
s)
Node Clock drift while GPS signal lost
Clock C1 (GPS lost)Clock C1 (Linearized)Clock C2 (GPS Guided)
Figure 6.7: Plot of the Clock Drift Recorded over 4 hours.
126CHAPTER 6. SYNCHRONISATION IN OCCASIONAL LOSS OF GNSS SIGNALS
5 8 10 15 20 25 30 35 40 45 50 55 60(1h) 65 70 75
No. of Observation in 1h
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
5O
ffset
(s)
Node Clock drift while GPS signal lost
Clock C1 (GPS lost)Clock C1 LinearizedClock C2 (GPS Guided)
Figure 6.8: Plot of the Clock Drift Recorded over 4 hours.
Figure 6.8 provides a closer view of the clock drift results over a 1 hour period.
It can be seen that from about 8 minutes that the GNSS signal is absent. Over the
next hour a drift of 80 s arises. The blue line represents the original drift with respect
to elapsed time, and the red line is their moving average. From the moving average
track, it follows that the trend of the drift is close to linear. That is, the clock drift is
predictable over short periods. Thus, a predicted clock drift can bridge the timing
outages when vehicles travel through tunnels where GNSS signals are completely
blocked. This is at least feasible in Australia, as the longest tunnel in Australia is 5.25
km. If the speed limit is 60 km/h, the signal outage could be 5.25 minutes. Figure 6.8
indicates that the clock offsets would increase 7 to 8 s over 5 to 6 minutes. The offsets
can be further reduced if a prediction model is used.
6.3 Summary
This chapter has focused on the availability aspects of GNSS time synchronisation in
vehicular environments. The vehicle-based experimental results have shown that the
availability of a minimum of 1 visible GPS satellite is 100% within Brisbane high-rise
CBD areas, while the availability of valid 4D position and time solutions was as low
6.3. SUMMARY 127
as 49.6%. Hence the availability of GNSS time solutions can be much higher than
the availability of positioning solutions within high-rise locations. The analysis of a
vehicle-tracking experiment indicates that the impact of user position biases due to
GNSS positioning outages on the clock errors is within acceptable range. Indeed, if
the clock errors of about sub 10 µs meet VANET timing requirements, GPS is con-
sidered still offering valid time solutions during periods where 1 to 3 satellites are
in view. If GNSS signals are completely lost, such as within tunnels, the suggested
fall-back solution is to use the predicted clock solution to bridge the outages for time
synchronisation in VANET. This is at least feasible in Australia, where the longest
tunnel is 5.25 km. In this case, signal outages of 5 to 6 minutes can lead to the clock
offsets 7 to 8 µs. The offsets can be further reduced if a prediction model is used.
128CHAPTER 6. SYNCHRONISATION IN OCCASIONAL LOSS OF GNSS SIGNALS
Chapter 7
Conclusions and Recommendations
The research findings and the original contributions of this study presented on this
thesis report are summarised in this chapter. Some additional research issues are also
presented here as the possible future works.
7.1 Summary of the Research
Time synchronisation is a critical and important issue in vehicular networks. Ve-
hicular networks are fundamentally decentralised in nature, therefore, the network
nodes need to use a time synchronisation method to maintain accurate and precise
time. The research described herein addresses the needs and applications of time
synchronisation in vehicular networks and contributes timing techniques that can be
provided by GNSS systems.
The thesis begins with a survey of the theory and practices of time synchronisa-
tion in different decentralised networks. This surveys various time synchronisation
protocols for emerging wireless, ad-hoc networks from a range of perspectives, in-
cluding precision, accuracy, cost and complexity. An analysis of time synchronisation
protocols is presented, which serves as a guide for evaluating and tailoring protocols
specific to applications in various road networks, and understanding the difficulties
and prospects of time synchronisation for vehicular networks. The analysis identifies
129
130 CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS
time-sensitive applications and requirements of vehicular networks. According the
findings, there are numerous applications that are sensitive to accurate and precise
time. The identified time-sensitive applications are subdivided into essential and
desirable categories, in which a stringent timing requirement is defined as sub-10
µs. The performance of existing time synchronisation practices in different networks
is examined and the feasibility of GNSS time synchronisation is specifically investi-
gated.
A feasibility analysis is presented to provide an understanding of the accuracy,
precision and availability offered by consumer-grade, low-cost, GNSS receivers. These
receivers are commonly used in modern vehicles for navigation purposes. It is found
that such receivers can provide around 30 ns timing accuracy in favourable envi-
ronments. An investigation was conducted into timing synchronisation accuracy for
applications involving the integration of GNSS receivers with on-board devices. This
investigation found that sub-10µs accuracy is achievable, which can meet vehicle net-
work application requirements. However, results of experiments reveal that the dif-
ference between achievable accuracy and the capacity of the GNSS receiver is mostly
due to the integration jitter, which occurs and accumulates in the system interfaces
and different communication stack layers, can degrade the available accuracy.
The experiments described herein serve to compare an existing time synchronisa-
tion solution and a developed GNSS time synchronisation solution for vehicular net-
works. The analyses address the limitations of the Time Advertisement (TA) scheme
of IEEE 802.11p in vehicular environments, and address the feasibility of using GNSS
time, which has been proposed as a source for standard UTC time in IEEE 1609.4. In
consideration of the outdoor and dynamic nature of VANET nodes, vehicle-based
experiments were conducted to investigate GNSS signal availability within urban
environments. The experimental results show that the availability for a minimum
of 1 visible GPS satellites is 100% in the Brisbane high rising CBD areas, while the
availability of valid 4D position and time solutions was as low as 49.6%. This implies
7.2. SUMMARY OF THE CONTRIBUTION 131
that the availability of GNSS time solutions can be much higher than the availability
of positioning solutions in the high-rising streets. Clearly the compatibility, accuracy,
compatibility and availability results described herein demonstrate that GNSS time
synchronisation is a promising method for vehicular network applications.
7.2 Summary of the Contribution
The main contribution of this thesis is to develop a time synchronised cooperative
vehicular network using the support of Global Navigation Satellite System (GNSS)
time synchronisation mechanism. A great deal of time synchronisation is being done
with technology-based wired and wireless communication. But they are not always
compatible or ill-suited with highly dynamic network VANET.
To understand the time sensitiveness of emerging cooperative vehicular network,
in the first place, this research work contributes to identifying different applications in
connected vehicular technology that relies on accurate and precise time and presented
them with their requirement threshold. This contribution to the knowledge helps to
recognise the importance of precise time synchronisation in the vehicular networks.
The research work demonstrates the efficacy of GNSS system to provide time syn-
chronisation solution for the vehicular network. In this part, the specific contribution
includes a thorough feasibility analysis of GNSS time synchronisation mechanism by
conducting and validating different setup variations with different real-time experi-
ments.
Another important contribution of this research work is to identify the shortcom-
ings of existing TSF time synchronization solutions proposed in VANET. This con-
tribution also extends to the demonstration of the capability of GNSS time solution
using DSRC like devices.
The other area of significant contribution is the demonstration of the potential
of the multi-GNSS constellation in signal availability for the timing solution. This
132 CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS
research work conducts several real-time experiments with multi-GNSS enabled con-
sumer grade receivers in different obstructed road scenario to prove the improvement
of the availability of signal reception with the help of he advancement of multi-GNSS
constellations. Overall, this thesis proposed a complete solution of time synchronisa-
tion for the vehicular ad-hoc network using GNSS timing support.
7.3 Future Works
This thesis is primarily concerned investigating GNSS time synchronisation and in-
tegration for vehicular networks. However, several unresolved issues remain, which
prompt recommendations for continuing research. This section, describes some areas
of interest for potential future research.
1. Many security applications depend on maintaining accurate time synchronisa-
tion with respect to a standard time-scale such as UTC. Some security issues
have been described for different wired and wireless communication networks,
including vehicle, power, financial and telecommunication networks. It is sug-
gested that further analyses of vehicular network security are undertaken to
manage safety vulnerabilities and possible terrorist attacks. In our work, we
have addressed few of the security issues, but another level of investigation is
important which we are considering as one of the future steps of the research.
2. It has been identified that the GNSS receiver integration within on-board de-
vices creates additional jitter. This results in timing uncertainties of a few mi-
croseconds between the practical timing accuracy offered by GNSS systems
and the achievable on-board accuracy. It is suggested that the jitter within
different integration layers is analysed further in the interest of reducing the
timing uncertainty.
3. In the absence of GNSS signals, the unguided quartz-based clock oscillators
7.3. FUTURE WORKS 133
drift and deviate with respect to the reference time. Although quartz clocks
are influenced by environment factors such as temperature, opportunities may
exist to use prediction models to improve the clock performance. For example,
it may be possible to train a clock system with past corrections to predict the
future corrections.
4. The scope of time synchronisation studied in this work is limited to a DSRC-
centric analyses. An alternative V2X communication interface candidate, LTE-
V, is increasing in popularity. The application of GNSS time synchronisation to
LTE-V and ensuing performance investigations are suggested for future work.
Similarly, the rapid growth of the IOT devices (Internet of Things) and their
applications, represent follow-on opportunities for GNSS time synchronisation
studies.
134 CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS
Literature Cited
Alam, N., Balaei, A. T., and Dempster, A. G. (2009). Range and range-rate
measurements using DSRC: facts and challenges. In IGNSS Symposium.
Alexander, P., Haley, D., and Grant, A. (2011). Cooperative intelligent transport
systems: 5.9-GHz field trials. Proceedings of the IEEE, 99(7):1213–1235.
Ansari, K. (2014). Development of an inter-vehicle communications & positioning platform
for transport safety applications (Doctoral dissertation). PhD thesis, Queensland
University of Technology, Brisbane, Australia.
Araniti, G., Campolo, C., Condoluci, M., Iera, A., and Molinaro, A. (2013). LTE for
Vehicular Networking: a Survey. IEEE communications magazine, 51(5):148–157.
Armengaud, E., Steininger, A., and Hanzlik, A. (2007). The effect of quartz drift
on convergence-average based clock synchronization. In 2007 IEEE Conference on
Emerging Technologies and Factory Automation (EFTA 2007), pages 1123–1130. IEEE.
ARRB-Project-Team (2013). Vehicle positioning for cooperative-its in australia
(background document), 2013. Technical Report Austroads Project No. NT1632,
AustRoads Research Report, Austroads Project No. NT1632.
Arvind, K. (1994). Probabilistic clock synchronization in distributed systems. IEEE
Transactions on Parallel and Distributed Systems, 5(5):474–487.
Bai, F. and Krishnan, H. (2006). Reliability analysis of DSRC wireless communication
for vehicle safety applications. In Intelligent Transportation Systems Conference, 2006.
ITSC’06., pages 355–362. IEEE.
135
136 LITERATURE CITED
Beehler, R. (1981). Time/frequency services of the US National Bureau of Standards
and some alternatives for future improvement. IETE Journal of Research, 27(10):389–
402.
Behrendt, K., Fodero, K., et al. (2006). The perfect time: An examination of
timesynchronization techniques. In DistribuTECH. Citeseer.
Ben-El Kezadri, R. and Pau, G. (2010). TimeRemap: stable and accurate time in
vehicular networks. IEEE Communications Magazine, 48(12):52–57.
Bilstrup, K., Uhlemann, E., Strom, E. G., and Bilstrup, U. (2009). On the ability
of the 802.11p MAC method and STDMA to support real-time vehicle-to-vehicle
communication. EURASIP Journal on Wireless Communication Netwwork, 2009:1–13.
Bilstrup, K. S., Uhlemann, E., and Strom, E. G. (2010). Scalability issues of the MAC
methods STDMA and CSMA of IEEE 802.11 p when used in VANETs. In 2010 IEEE
International Conference on Communications Workshops (ICC), pages 1–5. IEEE.
BITRE (2017). Road Death Australia-Monthly Bulletins.
BITRE2 (2012). Road Death Australia-Monthly Bulletins.
Bogovic, A. (2013). Use of Global Positioning System as Time Reference in Astrophysics
Experiments and Its Accuracy: Diploma Thesis. PhD thesis, A. Bogovic.
Bohm, A. (2013). Delay-sensitive wireless communication for cooperative driving
applications. PhD thesis, Halmstad University.
Bregni, S. (2002). Synchronization of digital telecommunications networks. Wiley New
York.
Bullock, J. B., King, T. M., Kennedy, H. L., Berry, E. D., and Zanfino, G. (1997). Test
results and analysis of a low cost core GPS receiver for time transfer applications. In
Proceedings of the 1997 IEEE International Frequency Control Symposium, 1997, pages
314–322. IEEE.
LITERATURE CITED 137
Caporaletti, G. (2012). The ERSEC project: Enhanced road safety by integrating egnos-
galileo data with on-board control system. Technical report, SAE Technical Paper.
Chen, G.-N., Wang, C.-Y., and Hwang, R.-H. (2006). MTSP: multi-hop time
synchronization protocol for IEEE 802.11 wireless ad hoc network. In International
Conference on Wireless Algorithms, Systems, and Applications, pages 664–675. Springer.
Chen, W., Sun, J., Zhang, L., Liu, X., and Hong, L. (2015). An implementation of IEEE
1588 protocol for IEEE 802.11 WLAN. Wireless Networks, 21(6):2069–2085.
Chen, W., Sun, J., Zhang, L., Liu, X., and Hong, L. (2015). An implementation of IEEE
1588 protocol for IEEE 802.11 WLAN. Wireless networks, 21(6):2069–2085.
Cheng, X., Li, W., and Znati, T. (2006). Wireless Algorithms, Systems, and Applications:
First International Conference, WASA 2006, Xi’an, China, August 15-17, 2006,
Proceedings, volume 4138. Springer.
Conde, L., Chelim, R., and Nunes, U. (2015). Collaborative vehicle self-
localization using multi-gnss receivers and v2v/v2i communications. In Intelligent
Transportation Systems (ITSC), 2015 IEEE 18th International Conference on, pages
2525–2532. IEEE.
Cooklev, T., Eidson, J. C., and Pakdaman, A. (2007). An implementation of IEEE 1588
over IEEE 802.11 b for synchronization of wireless local area network nodes. IEEE
Transactions on Instrumentation and measurement, 56(5):1632–1639.
Cozzetti, A., Scopigno, R., and Lo Presti, L. (2011). Tight coupling benefits of GNSS
with VANETs. Aerospace and Electronic Systems Magazine, IEEE, 26(4):15–23.
Cozzetti, H. and Scopigno, R. (2011). Scalability and QoS in MS-Aloha VANETs:
Forced slot re-use versus pre-emption. In 14th International IEEE Conference on
Intelligent Transportation Systems (ITSC), pages 1759–1766.
Cozzetti, H., Scopigno, R., Casone, L., and Barba, G. (2009). Comparative analysis
138 LITERATURE CITED
of IEEE 802.11p and MS-Aloha in Vanet scenarios. In IEEE Asia-Pacific Services
Computing Conference (APSCC 2009), pages 64–69.
Cozzetti, H. A. and Scopigno, R. (2009). RR-Aloha+: A slotted and distributed MAC
protocol for vehicular communications. In Vehicular Networking Conference (VNC),
pages 1–8. IEEE.
Cunha, F., Villas, L., Boukerche, A., Maia, G., Viana, A., Mini, R. A., and Loureiro,
A. A. (2016). Data communication in vanets: Protocols, applications and challenges.
Ad Hoc Networks, 44:90–103.
De Jong, G. and Lewandowski, W. (1997). GLONASS/GPS time transfer and the
problem of the determination of receiver delays. Technical report, DTIC Document.
Dua, A., Kumar, N., and Bawa, S. (2014). A systematic review on routing protocols
for vehicular ad hoc networks. Vehicular Communications, 1(1):33–52.
E-GNSS (2018). European GNSS Sservice Centre : Constellation Information.
https://www.gsc-europa.eu/system-status/Constellation-Information.
Ebner, A., Rohling, H., Lott, M., and Halfmann, R. (2002). Decentralized slot
synchronization in highly dynamic ad hoc networks. In The 5th Int Symp on Wireless
Personal Multimedia Communications, volume 2, pages 494–498.
Elson, J., Girod, L., and Estrin, D. (2002). Fine-grained network time synchronization
using reference broadcasts. ACM SIGOPS Operating Systems Review, 36(SI):147–163.
Elson, J. and Romer, K. (2003). Wireless sensor networks: A new regime for time
synchronization. ACM SIGCOMM Computer Communication Review, 33(1):149–154.
Emmelmann, M., Bochow, B., and Kellum, C. (2010). Vehicular networking: automotive
applications and beyond, volume 2. John Wiley & Sons.
Englund, C., Chen, L., Vinel, A., and Lin, S. Y. (2015). Future Applications of VANETs,
pages 525–544. Springer International Publishing, Cham.
LITERATURE CITED 139
Engoulou, R. G., Bellaıche, M., Pierre, S., and Quintero, A. (2014). Vanet security
surveys. Computer Communications, 44:1–13.
Enkelmann, W. (2003). Fleetnet-applications for inter-vehicle communication. In
Intelligent Vehicles Symposium, 2003. Proceedings. IEEE, pages 162–167. IEEE.
ETSI (2011-12). Intelligent Transportation Systems (ITS); Performance Evaluation of
Self-Organizing TDMA as Medium Access Control Method Applied to ITS; Access
Layer Part. Technical Report DTR/ITS-0040021, European Telecommunications
Standards Institute.
ETSI (2012). Intelligent Transport Systems (ITS): STDMA recommended parameters
and settings for cooperative ITS; Access Layer Part. Technical report, European
Telecommunications Standards Institute.
Exel, R. and Ring, F. (2014). Direct-digital time-domain oscillator stability
measurement. In Microelectronic Systems Symposium (MESS), 2014, pages 1–5. IEEE.
Eze, E. C., Zhang, S., and Liu, E. (2014). Vehicular ad hoc networks (vanets): Current
state, challenges, potentials and way forward. In 20th International Conference on
Automation and Computing (ICAC), 2014, pages 176–181. IEEE.
Fagnant, D. J. and Kockelman, K. (2015). Preparing a nation for autonomous vehicles:
opportunities, barriers and policy recommendations. Transportation Research Part A:
Policy and Practice, 77:167–181.
Fan, R., Chakraborty, I., and Lynch, N. (2004). Clock synchronization for wireless
networks. In International Conference On Principles Of Distributed Systems, pages
400–414. Springer.
Fraleigh, C., Diot, C., Lyles, B., Moon, S., Owezarski, P., Papagiannaki, D., and
Tobagi, F. (2001). Design and deployment of a passive monitoring infrastructure.
Evolutionary Trends of the Internet, pages 556–575.
140 LITERATURE CITED
Franz, W., Hartenstein, H., and Mauve, M. (2005). Inter-vehicle-communications Based
on Ad Hoc Networking Principles: The FleetNet Project. Universitatsverlag Karlsruhe.
Freris, N. M., Graham, S. R., and Kumar, P. (2011). Fundamental limits on
synchronizing clocks over networks. IEEE Transactions on Automatic Control,
56(6):1352–1364.
Gaderer, G., Loschmidt, P., and Sauter, T. (2006). Quality monitoring in clock
synchronized distributed systems. In 6th IEEE International Workshop on Factory
Communication Systems (WFCS06), pages 13–22. Citeseer.
Ganeriwal, S., Kumar, R., and Srivastava, M. B. (2003). Timing-sync Protocol for
Sensor Networks. In Proc of the 1st Int Conf on Embedded Networked Sensor Systems,
pages 138–149, New York, NY, USA.
Ghosh, S., Kundu, A., and Jana, D. (2011). Implementation challenges of
time synchronization in vehicular networks. In Recent Advances in Intelligent
Computational Systems (RAICS), pages 575–580.
Glonass iac.ru (2018). Information and Analysis Center for Positioning, Navigation
and Timing : GLONASS Constellation Status, 21.02.2018. http://glonass-
iac.ru/en/GLONASS/.
Golestan, K., Jundi, A., Nassar, L., Sattar, F., Karray, F., Kamel, M., and Boumaiza,
S. (2012a). Vehicular ad-hoc networks (VANETs): capabilities, challenges in
information gathering and data fusion. In Autonomous and Intelligent Systems, pages
34–41. Springer.
Golestan, K., Seifzadeh, S., Kamel, M., Karray, F., and Sattar, F. (2012b). Vehicle
Localization in VANETs Using Data Fusion and V2V Communication. In
Proceedings of the Second ACM International Symposium on Design, DIVANet’12, pages
123–130, New York, NY, USA.
LITERATURE CITED 141
Groves, P. D. (2013). Principles of GNSS, inertial, and multisensor integrated navigation
systems. Artech house.
Gunther, A. and Hoene, C. (2005). Measuring round trip times to determine the
distance between wlan nodes. In Networking, volume 3462, pages 768–779. Springer.
Guo, H. and Crossley, P. (2017). Design of a time synchronization system based on
gps and IEEE 1588 for transmission substations. IEEE Transactions on Power Delivery,
32(4):2091–2100.
Gupta, N., Prakash, A., and Tripathi, R. (2015). Medium access control protocols
for safety applications in Vehicular Ad-Hoc Network: A classification and
comprehensive survey. Vehicular Communications, 2(4):223–237.
Hafeez, K. A., Zhao, L., Ma, B., and Mark, J. W. (2013). Performance analysis
and enhancement of the dsrc for vanet’s safety applications. IEEE Transactions on
Vehicular Technology, 62(7):3069–3083.
Hartenstein, H. and Laberteaux, K. (2010). VANET: vehicular applications and inter-
networking technologies, volume 1. Wiley Online Library.
Hasan, K. F. and Feng, Y. (2016). A Study on Consumer Grade GNSS Receiver for
Time Synchronization in VANET. In 23rd ITS World Congress, Melbourne, Australia,
10 - 14 October.
Hasan, K. F., Feng, Y., and Tian, Y.-C. (2018a). Exploring the Potential and
Feasibility of Time Synchronization using GNSS Receivers in Vehicle-to-Vehicle
Communications. In In Proceedings of ITM 2018, Reston, VA, Jan 29-2 Feb.
Hasan, K. F., Feng, Y., and Tian, Y.-C. (2018b). Gnss time synchronization in
vehicular ad-hoc networks: Benefits and feasibility. IEEE Transactions on Intelligent
Transportation Systems.
Horauer, M. and Holler, R. (2002). Integration of high accurate clock synchronization into
ethernet-based distributed systems. na.
142 LITERATURE CITED
Hu, Y.-C., Perrig, A., and Johnson, D. B. (2003). Packet leashes: a defense against
wormhole attacks in wireless networks. In INFOCOM 2003. Twenty-Second Annual
Joint Conference of the IEEE Computer and Communications. IEEE Societies, volume 3,
pages 1976–1986. IEEE.
Huang, L. and Lai, T.-H. (2002). On the scalability of IEEE 802.11 ad hoc networks.
In Proceedings of the 3rd ACM international symposium on Mobile ad hoc networking &
computing, pages 173–182. ACM.
Hussein, S., Krings, A., and Azadmanesh, A. (2017). Vanet clock synchronization for
resilient dsrc safety applications. In Resilience Week (RWS), 2017, pages 57–63. IEEE.
I-GNSS (2018). The Multi-GNSS Experiment and Pilot Project (MGEX).
http://mgex.igs.org/.
IEEE (2011). IEEE std 802.1 as-2011, IEEE standard for local and metropolitan area
networkstiming and synchronization for time-sensitive applications in bridged
local area networks. IEEE Standards Association and Others, 30:292.
Isaac, J. T., Zeadally, S., and Camara, J. S. (2010). Security attacks and solutions for
vehicular ad hoc networks. IET Communications, 4(7):894–903.
Jiang, D. and Delgrossi, L. (2008). IEEE 802.11 p: Towards an international standard
for wireless access in vehicular environments. In Vehicular Technology Conference,
2008. VTC Spring 2008. IEEE, pages 2036–2040. IEEE.
Jiang, D., Taliwal, V., Meier, A., Holfelder, W., and Herrtwich, R. (2006). Design of 5.9
GHz DSRC-based vehicular safety communication. Wireless Communications, IEEE,
13(5):36–43.
Joe, M. M. and Ramakrishnan, B. (2016). Review of vehicular ad hoc network
communication models including wvanet (web vanet) model and wvanet future
research directions. Wireless networks, 22(7):2369–2386.
LITERATURE CITED 143
Johannessen, S., Skeie, T., and Lokstad, T. (2001). Time synchronization in computer
network. US Patent App. 10/465,944.
Kadowaki, Y. and Ishii, H. (2015). Event-based distributed clock synchronization for
wireless sensor networks. IEEE Transactions on Automatic Control, 60(8):2266–2271.
Karagiannis, G., Altintas, O., Ekici, E., Heijenk, G., Jarupan, B., Lin, K., and Weil, T.
(2011). Vehicular networking: A survey and tutorial on requirements, architectures,
challenges, standards and solutions. IEEE Communications Surveys & Tutorials,
13(4):584–616.
Kenney, J. B. (2011). Dedicated short-range communications (DSRC) standards in the
United States. Proceedings of the IEEE, 99(7):1162–1182.
King, M., Busch, D., and Miranian, M. (1994). Test Results and Analysis of a Low
Cost Core GPS Receiver for Time Transfer Application. In Proceedings of the 1994
National Technical Meeting of The Institute of Navigation, pages 99–109.
Lee, E.-K., Yang, S., Oh, S. Y., and Gerla, M. (2009). RF-GPS: RFID assisted localization
in VANETs. In 6th IEEE International Conference on Mobile Adhoc and Sensor Systems
(MASS’09), pages 621–626.
Lenzen, C., Sommer, P., and Wattenhofer, R. (2015). Pulsesync: An efficient and
scalable clock synchronization protocol. IEEE/ACM Transactions on Networking,
23(3):717–727.
Levesque, M. and Tipper, D. (2016). A survey of clock synchronization over packet-
switched networks. IEEE Communications Surveys & Tutorials, 18(4):2926–2947.
Lewandowski, W., Azoubib, J., and Klepczynski, W. J. (1999). GPS: primary tool for
time transfer. Proceedings of the IEEE, 87(1):163–172.
Lewandowski, W., Petit, G., and Thomas, C. (1993). Precision and accuracy of GPS
time transfer. IEEE Transactions on Instrumentation and Measurement, 42(2):474–479.
144 LITERATURE CITED
Li, Y. J. (2010). An overview of the DSRC/WAVE technology. In International
Conference on Heterogeneous Networking for Quality, Reliability, Security and
Robustness, pages 544–558. Springer.
Lim, J.-H. (2016). Understanding STDMA via computer simulation: feasibility to
vehicular safety applications, configurations, and time synchronization errors.
EURASIP Journal on Wireless Communications and Networking, 2016(1):181.
Lombardi, M. (2000). Computer Time Synchronization.
Lombardi, M. A., Nelson, L. M., Novick, A. N., and Zhang, V. S. (2001). Time and
frequency measurements using the global positioning system. Cal Lab: International
Journal of Metrology, 8(3):26–33.
Loo, J., Mauri, J. L., and Ortiz, J. H. (2016). Mobile ad hoc networks: current status and
future trends. CRC Press.
Loschmidt, P., Exel, R., and Gaderer, G. (2012). Highly accurate timestamping
for ethernet-based clock synchronization. Journal of Computer Networks and
Communications, 2012.
Luettel, T., Himmelsbach, M., and Wuensche, H.-J. (2012). Autonomous ground
vehicles-concepts and a path to the future. Proceedings of the IEEE, 100(Centennial-
Issue):1831–1839.
Mahmood, A., Exel, R., and Sauter, T. (2015). A simulation-based comparison of ieee
802.11’s timing advertisement and synctsf for clock synchronization. In 2015 IEEE
World Conference on Factory Communication Systems (WFCS), pages 1–4. IEEE.
Mahmood, A., Exel, R., and Sauter, T. (2017). Performance of ieee 802.11s timing
advertisement against synctsf for wireless clock synchronization. IEEE Transactions
on Industrial Informatics, 13(1):370–379.
Mangharam, R., Rajkumar, R., Hamilton, M., Mudaliget, P., and Bait, F. (2007).
LITERATURE CITED 145
Bounded-latency alerts in vehicular networks. In 2007 Mobile Networking for
Vehicular Environments, pages 55–60. IEEE.
Maroti, M., Kusy, B., Simon, G., and Ledeczi, A. (2004). The Flooding Time
Synchronization Protocol. In Proc. of the 2nd Int Conf on Embedded Networked Sensor
Systems, pages 39–49, New York, NY, USA.
Miao, L., Djouani, K., van Wyk, B. J., and Hamam, Y. (2012). Evaluation and
enhancement of IEEE 802.11 p standard: A survey. Mobile Computing, 1(1).
Mills, D. L. (1997). Computer network time synchronization. In Report Dagstuhl
Seminar on Time Services Schloß Dagstuhl, March 11.–March 15. 1996, volume 12, page
332. Springer.
Mills, D. L. (2016). Computer network time synchronization: the network time protocol on
earth and in space. CRC Press.
Misra, P. and Enge, P. (2006). Global positioning system: signals, measurements and
performance second edition. Massachusetts: Ganga-Jamuna Press.
Misra, S., Woungang, I., and Misra, S. C. (2009). Guide to wireless ad hoc networks.
Springer Science & Business Media.
Molina-Masegosa, R. and Gozalvez, J. (2017). LTE-V for Sidelink 5G V2X Vehicular
Communications: A New 5G Technology for Short-Range Vehicle-to-Everything
Communications. IEEE Vehicular Technology Magazine, 12(4):30–39.
Morgan, Y. L. (2010). Notes on DSRC & WAVE standards suite: Its architecture,
design, and characteristics. IEEE Communications Surveys & Tutorials, 12(4):504–518.
Mumford, P. (2003). Relative timing characteristics of the one pulse per second (1PPS)
output pulse of three GPS receivers. In Proc. 6th International Symposium on Satellite
Navigation Technology Including Mobile Positioning & Location Services, pages 22–25.
Citeseer.
146 LITERATURE CITED
Packard, H. (1997). Fundamentals of quartz oscillators: application note 200-2. New
York.
PalChaudhuri, S., Saha, A., and Johnson, D. B. (2003). Probabilistic clock
synchronization service in sensor networks. IEEE Transactions on Networking,
2(2):177–189.
Parker, T. E. and Matsakis, D. (2004). Time and frequency dissemination: advances in
gps transfer techniques. GPS world.
Penrod, B. M. (1996). Adaptive temperature compensation of GPS disciplined quartz
and rubidium oscillators. In Frequency Control Symposium, 1996. 50th., Proceedings of
the 1996 IEEE International., pages 980–987. IEEE.
Perahia, E. (2008). IEEE 802.11n Development: History, Process, and Technology.
IEEE Communications Magazine, 46(7):48–55.
Petovello, M. (2011). Gnss solutions: Clock offsets in gnss receivers. Inside GNSS,
pages 23–25.
PISCATAWAY, N. (1996). Wireless LAN medium access control (MAC) and physical
layer (PHY) specifications. IEEE P802. 11 D3.
QPS (2015). UTC to GPS Time Correction. http://www.qps.nl/.
Rasheed, A., Gillani, S., Ajmal, S., and Qayyum, A. (2017). Vehicular Ad Hoc Network
(VANET): A Survey, Challenges, and Applications, pages 39–51. Springer Singapore,
Singapore.
Romer, K. (2001). Time synchronization in Ad hoc networks. In Proce of the 2nd ACM
Int Symp on Mobile ad hoc Networking & Computing, pages 173–182.
SCC32, C. (2006). IEEE P1609. 4 standard for wireless access in vehicular
environments (wave)-multi-channel operation. draft standard ed. IEEE Intelligent
Transportation Systems Council.
LITERATURE CITED 147
Scopigno, R. and Cozzetti, H. (2009). GNSS Synchronization in Vanets. In 3rd Int Conf
on New Technologies, Mobility & Security (NTMS), pages 1–5.
Scott, G. and Demoz, G.-E. (2009). GNSS applications and methods. Artech House,
685.
Sharef, B. T., Alsaqour, R. A., and Ismail, M. (2014). Vehicular communication ad hoc
routing protocols: A survey. Journal of network and computer applications, 40:363–396.
Sharma, M. S. P. and Tomar, R. S. (2016). VANET based communication model for
transportation systems. In Symposium on Colossal Data Analysis and Networking
(CDAN), pages 1–6. IEEE.
Shizhun Wang, Pervez, A., and Nekovee, M. (2010). Converging time
synchronization algorithm for highly dynamic vehicular ad hoc networks
(VANETs). In 7th Int Symp on Communication Systems Networks & Digital Signal
Processing (CSNDSP), pages 443–448.
Sichitiu, M. L. and Veerarittiphan, C. (2003). Simple, accurate time synchronization for
wireless sensor networks. In Wireless Communications and Networking, 2003. WCNC
2003. 2003 IEEE, volume 2, pages 1266–1273. IEEE.
Sjoberg, K., Uhlemann, E., and Strom, E. G. (2011). Delay and interference comparison
of CSMA and self-organizing TDMA when used in VANETs. In 7th International
Conference on Wireless Communications and Mobile Computing Conference (IWCMC),
pages 1488–1493.
Sjoberg Bilstrup, K. (2009). Predictable and scalable medium access control for
vehicular ad hoc networks. Technical report, Department of Signals and Systems,
Chalmers University of Technology, Goteborg, Sweden.
Skog, I. and Handel, P. (2011). Time synchronization errors in loosely coupled GPS-
aided inertial navigation systems. IEEE Transactions on Intelligent Transport Systems,
12(4):1014–1023.
148 LITERATURE CITED
So, J. and Vaidya, N. (2004). MTSF: A timing synchronization protocol to support
synchronous operations in multihop wireless networks. University of Illinois at
Urbana-Champaign, Tech. Rep.
Standards Association, I. et al. (2001). IEEE Standard for Information Technology-
Telecommunications and Information Exchange Between Systems-Local and Metropolitan
Area Networks-Specific Requirements: Part 11: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) Specifications. IEEE.
Sundararaman, B., Buy, U., and Kshemkalyani, A. D. (2005). Clock synchronization
for wireless sensor networks: a survey. Ad Hoc Networks, 3(3):281–323.
Tang, A. and Yip, A. (2010). Collision avoidance timing analysis of DSRC-based
vehicles. Accident Analysis & Prevention, 42(1):182–195.
Tetreault, C. B. J. (2005). Use of the Automatic Identification System (AIS) for
maritime domain awareness (MDA). In Proceedings of MTS/IEEE, OCEANS, 2005,
pages 1590–1594. IEEE.
Texas Instruments (2012). TI GPS PPS timing applications note Texas Instruments.
Tian, G.-S., Tian, Y.-C., and Fidge, C. (2008). High-precision relative clock
synchronization using time stamp counters. In Engineering of Complex Computer
Systems, 2008. ICECCS 2008. 13th IEEE International Conference on, pages 69–78.
IEEE.
Uscg.gov (2018). Navigation Center: The Navigation Center of Excellence : GPS
Constellation Status for 02/10/2018.
Uzcategui, R. and Acosta Marum, G. (2009). WAVE: a tutorial. IEEE Communications
Magazine, 47(5):126–133.
van Greunen, J. and Rabaey, J. (2003). Lightweight Time Synchronization for Sensor
Networks. In Proceedings of the 2Nd ACM International Conference on Wireless Sensor
Networks and Applications, pages 11–19, New York, NY, USA.
LITERATURE CITED 149
Verenzuela, D., Liu, C., Wang, L., and Shi, L. (2014). Improving scalability of vehicle-
to-vehicle communication with prediction-based stdma. In 80th IEEE Vehicular
Technology Conference (VTC Fall), pages 1–5.
Weibel, H. and Bechaz, D. (2004). Implementation and performance of time stamping
techniques. In Proceedings of the 2004 Conference on IEEE, volume 1588.
WHO (2015). Global Status Report on Road Safety 2013. http://www.who.int/.
Witherspoon, J. T. and Schuchman, L. (1978). A time transfer unit for GPS. In Proc.
9th Annual PTTI Applications and Planning Meeting, pages 167–176.
Yi Qian, Kejie Lu, and Moayeri, N. (2008). A Secure VANET MAC Protocol for DSRC
Applications. In Global Telecommunications Conference, 2008. IEEE GLOBECOM 2008,
pages 1–5.
Yoon, J.-Y., Kim, J.-W., Lee, W.-Y., and Eom, D.-S. (2012). A TDoA-based localization
using precise time-synchronization. In 14th Int Conf on Advanced Communication
Tech (ICACT), pages 1266–1271.
Yu, F. and Biswas, S. (2007). Self-configuring TDMA protocols for enhancing vehicle
safety with DSRC based vehicle-to-vehicle communications. IEEE Journal on
Selected Areas in Communications, 25(8):1526–1537.
150 LITERATURE CITED