Mobile Phone‐based Vehicle Positioning

and Tracking and Its Application in Urban

Traffic State Estimation

SHA TAO

Licentiate Thesis
KTH Royal Institute of Technology
Stockholm, Sweden 2012
TRITA-ICT/MAP AVH Report 2012:01 KTH School of Information and
ISSN 1653-7610 Communication Technology
ISRN KTH/ ICT/ MAP/AVH-2012:01–SE SE-164 40 Kista
ISBN 978-91-7501-204-9 SWEDEN

Akademisk avhandling som med tillstånd av Kungl Tekniska högskolan framlägges till
offentlig granskning för avläggande av teknologie licentiatexamen i Elektronik och
Datorsystem onsdag den 18 Januari 2012 klockan 14.00 i Sal C2, Electrum 229, Kungl
Tekniska Högskolan, Isafjordsgatan 26, Kista.

© Sha Tao, January 2012

Tryck: Kista Snabbtryck AB
 iii

Abstract

Enabling the positioning and tracking of mobile phones has emerged as a key
facility of existing and future generation mobile communication systems. This
feature provides opportunities for many value added location-based services and
systems. For instance, mobile phones are increasingly employed in traffic infor-
mation systems and present several advantages over traditional sensor-based traffic
systems. However, there are still plenty of aspects that must be investigated and
addressed towards the fully operational deployment. The aim of the research per-
formed in this thesis is to examine and propose solutions to two of the problems in
the deployment of a mobile phone-based smart traffic information system.
The first problem investigated is the mobile phone-based vehicle positioning and
tracking. The investigation starts with a comprehensive study of mobile positioning
with emphasis on existing standardizations. Based on the mobile location methods
standardized in UMTS, possible hybrid solutions are proposed. In addition, a tool
for simulating one of the UMTS mobile positioning methods (i.e., OTDOA) in
vehicular environment is developed. A Kalman filter-based hybrid method, which
can track the mobile phones traveling on-board vehicles, is then implemented. This
method fuses two of the UMTS standard methods (i.e., OTDOA and A-GPS) loca-
tion estimates at the state-vector level. Statistical simulation results demonstrate
that the hybrid method can provide better position and velocity estimations than
each individual method.
The second problem addressed is the mobile phone-based urban traffic state
estimation. A traffic simulation-based framework is proposed to emulate and evalu-
ate the operation of urban traffic state estimation with A-GPS mobile phones as
probes. Based on the emulated mobile phone probe data, algorithms of location data
processing/filtering and average speed estimation are developed and then evaluated
by comparing against “ground truth” data from the traffic simulation. Moreover, the
estimated average speeds are classified to different traffic condition levels, which
are prepared for displaying a traffic map on the mobile phone display. The achieved
simulation results demonstrate the effectiveness of the proposed method, which is
fundamental for the subsequent development of a mobile phone-based smart traffic
information system demonstrator.
iv
 v

Dedicated to
the Ones I Love
and
the Ones Who Love Me
vi
 vii

ACKNOWLEDGEMENTS

E very story has its beginning; mine starts two years ago with a group, once called RaMSiS.
At that time, there were six members in our group, our little ‘family’. Although the six of us
are actually diverse in many ways: different origins, languages, backgrounds, personalities,
etc., when we come together, the feeling is just so right. Maybe this is because family is such
a place where everyone accepts you for who you are.
Therefore, my first and foremost acknowledgment goes to the ones, who spent with me
from the very beginning the past two years of my doctoral study: Assoc. Prof. Ana Rusu,
Prof. Mohammed Ismail Elnaggar, Dr. Saul Rodriguez Duenas, Julian Marcos Garcia, and
Vasileios Manolopoulos. Ana, my main advisor, you introduced me into this family and
encouraged me stepping into a brand new field; you creates a fantastic ‘aura’ around that
makes everyone can easily fit in and teaches me to be open minded to the thing which used
to sounds ‘alien’ to me. You always correct our manuscripts word-by-word with your
“magic red pen” (quoted from Eduard). Your door is always open to us; you are always there
for us listening to our ideas (sometimes immature) and counseling us when we came across
obstacles and felt depressed. You spoil us! However, you are far more than a good ‘profes-
sor’; to me you are a real ‘educator’: you always want us to be not only an excellent student
or researcher but also a better person and an independent thinker. Mohammed, my
co-advisor, though has not been with us very often, you have been always supportive to my
study and trying to give me suggestions. It was such a pleasure talking with you every time
you came. Saul, my co-advisor, who was my master thesis supervisor and actually has been
my supervisor since three years ago when I took a project course; you are the person I
always go to for advices, both academically and daily life related. Thanks a lot for making
my life easier. Julian, our senior PhD student, you are just perfect in many ways: the way
you work (e.g., your attitude and passion), the way you speak/communicate (e.g., when you
presenting and teaching), etc.; I admire you and I wish you all the best in your upcoming
defense. Vasilis, my fellow student, my project teammate, the cool guy I shared the office
with during these two years; of course not only the office, but there were many other things
we have been sharing in our “little dark place” (quoted from Ana): discussions regarding our
viii

study and project, opinions and ideas about people we know and things happened, even the
confusions we had. We have been through phases together in our PhD studies, and at this
point, wish ourselves the best of luck. Once again, thank all of you for making me feel at
home when working in the office; you guys are simply awesome!
My appreciation also goes to our present and former (guest) group members: Maryam
Olyaei, Jinxin Song, Ying Song, Cristina Marghescu, Rocco Luciano Grimaldi, Prof. Edu-
ard Alarcon, Dr. Raul Onet, Dr. Fadi R. Shahroury, Babak Taghavi, and Milad Raz-
zaghpour. I do enjoy the moments we shared, and I am really glad that you guys are/were
around. Babak, you did a fantastic master thesis and you are so wholesome. I had a lot of fun
discussing soccer with you. Milad, you are exceptionally gifted in doing research and you
are born to be a “professor” (as Ana once joked about you). Meanwhile, you are so adorable
and you are delightful to talk with: someone must be insane if he/she doesn’t feel enjoyable
when chatting with you.
Many thanks to professors of the postgraduate courses I took for my Licentiate: Assoc.
Prof. Eilert Berglind, Prof. Gerald Maguire, Assoc. Prof. Lena Wosinska, and Prof. Elena
Dubrova; especially Prof. Gerald Maguire for his valuable feedback in one part of the work
included in this thesis. My thanks also go to Prof. Mikael Östling and Prof. Carl-Mikael
Zetterling (Bellman), for their brilliant managing and organizing stills to keep our depart-
ment running well. In addition, I would like to thank Gunilla Gabrielsson, our department
administrator, for taking care of a variety of administrative issues. I would also like to thank
the ICT IT-Service staffs for the support on computer-related issues.
I gratefully acknowledge the financial support from the Swedish Foundation for Strategic
Research (SSF) for funding my research under the TRAFFIC project. I also acknowledge
the KTH transport platform for organizing meetings & events that provide us opportunities
meeting people around who do related research.
I am thankful to Assoc. Prof. Panos Papadimitratos for reviewing my Licentiate proposal
and for all constructive comments that helped improve this thesis; to Prof. Anders Hallén for
being the internal reviewer of this thesis. I am genuinely grateful to Dr. Wilco Burghout for
accepting the role as opponent at my defense.
Special thanks to my Chinese friends here in Electrum: Xingang Yu, Yu Xiang, Ye Tian,
Tingsu Chen, Dr. Jiantong Li, Dr. Qing Lin, Dr. Jun Luo, Dr. Shun Yu, Yeyu Fang, and
Chen Hu. Tianye and Tingsu, you have just started your PhD study and I wish you nothing
but the best. Yeyu and Huchen, with whom I shared most of my happy lunch hours, I miss
you already. Some of you guys may have left, but I treasure the great time we spent together
during the past two years. My other friends, whether in Stockholm or Beijing, thank you all
for being in my life.
Finally, with all my heart, I would like to express my deepest love to my dearest ones. My
boyfriend, who has always accompanied me during the past years and will be with me the
rest of my life, I feel so blessed to have you. My life would suck without you. My family, in
my sweet home Beijing, especially my mom and dad, my four grandparents, my uncle and
aunt, you are my inspiration all the way and you are the reason who I am. You raise me up.
 ix

CONTENTS

Abstract iii

Acknowledgements vii

Contents ix

Abbreviations and Acronyms xii

1 Introduction 1
1.1 Background ................................................................................................. 2
1.1.1 The Use of Mobile Phones in ITS........................................................ 2
1.1.2 Smart Traffic Information System: An Overview ............................... 2
1.2 Problem Formulation .................................................................................. 4
1.3 Thesis Outline and Contributions ............................................................... 5

2 Vehicle Location Using Mobile Phones: Survey and Prospects 7

2.1 Mobile Positioning and Its Performance Metrics ....................................... 8
2.1.1 Calls for Mobile Positioning ................................................................ 8
2.1.2 Vehicle Location: Requirements and Metrics ...................................... 9
2.2 Legacy Systems and Mobile Positioning .................................................. 10
2.2.1 Legacy Vehicle Positioning Systems ................................................. 10
2.2.2 Mobile Positioning Principles ............................................................ 11
2.2.2.1 Proximity Sensing ........................................................................ 11
2.2.2.2 Circular Lateration ....................................................................... 11
2.2.2.3 Hyperbolic Lateration .................................................................. 12
x

2.3 Mobile Positioning Standardization in Generations of RANs .................. 13

2.3.1 Standard LCS in GERAN .................................................................. 14
2.3.2 Standard LCS in UTRAN .................................................................. 15
2.3.3 Standard LCS in E-UTRAN .............................................................. 16
2.4 Classification and Discussion ................................................................... 16
2.4.1 Major Categories of Wireless Location ............................................. 16
2.4.2 Location Methods in Generations of RANs ....................................... 18
2.4.3 Performance Evaluation of 3G UE Positioning ................................. 19
2.4.4 Further Discussion ............................................................................. 20
2.5 The Proposed Hybrid Positioning Methods .............................................. 21
2.5.1 Combining Network-based with Handset-based Methods ................. 22
2.5.1.1 Hybrid of Cell-ID and A-GPS ..................................................... 22
2.5.1.2 Hybrid of U-TDOA and A-GPS .................................................. 22
2.5.2 Fusion of Two Handset-based Methods............................................. 23
2.5.3 Suggestions for Further Implementation............................................ 23

3 UMTS Mobile Positioning Simulator for Vehicle Location 25

3.1 Sources of Mobile Positioning Error ........................................................ 26
3.2 Radio Network Modeling ......................................................................... 27
3.2.1 Network Topology and MS Mobility ................................................ 27
3.2.2 Radio Propagation Model .................................................................. 28
3.2.2.1 Path Loss...................................................................................... 28
3.2.2.2 Correlated Shadowing ................................................................. 29
3.2.2.3 Rayleigh Fading ........................................................................... 30
3.3 WCDMA Link Simulator ......................................................................... 30
3.3.1 BS Transmitter Model........................................................................ 31
3.3.2 Multipath Channel Model .................................................................. 31
3.3.3 MS Receiver Model ........................................................................... 32
3.4 UMTS Mobile Positioning........................................................................ 32
3.4.1 OTDOA Measurement ....................................................................... 32
3.4.2 Cross-correlation Method .................................................................. 33
3.4.3 MS Location Calculation ................................................................... 33
3.5 Summary ................................................................................................... 34

4 Mobile Phone-based Hybrid Vehicle Positioning and Tracking 35

4.1 The Hybrid Positioning and Tracking Scheme ......................................... 36
4.1.1 The Hybrid Scheme ........................................................................... 36
4.1.1.1 A-GPS and OTDOA Measurements ............................................ 37
4.1.1.2 Location Estimates and Noise Variance ...................................... 37
4.1.1.3 State Estimates after KF .............................................................. 37
 xi

4.1.1.4 Position and Velocity Estimates after Fusion .............................. 37

4.1.2 A-GPS and OTDOA Measurements .................................................. 37
4.1.2.1 A-GPS .......................................................................................... 37
4.1.2.2 OTDOA ....................................................................................... 38
4.2 The Positioning and Tracking Algorithms................................................ 38
4.2.1 Hyperbolic Solution based on OTDOA ............................................. 38
4.2.2 Mobility and Measurement Models ................................................... 41
4.2.2.1 Motion Dynamics ........................................................................ 41
4.2.2.2 Measurement Models................................................................... 42
4.2.3 Kalman Filter for Trajectory Tracking............................................... 43
4.2.4 State-Vector Fusion ........................................................................... 44
4.3 Simulation Results and Evaluation ........................................................... 44
4.4 Summary ................................................................................................... 50

5 Traffic State Estimation with A-GPS Mobile Phones as Probes 51

5.1 Background ............................................................................................... 52
5.2 System Overview ...................................................................................... 53
5.2.1 Simulation-based Framework ............................................................ 53
5.2.2 Urban Traffic Modeling ..................................................................... 55
5.2.3 System Design Parameters ................................................................. 56
5.3 Data Processing and Aggregation ............................................................. 56
5.3.1 A-GPS Probe Data Emulation ........................................................... 56
5.3.2 Filtering of Location Data .................................................................. 58
5.3.2.1 KF-based Tracking ...................................................................... 58
5.3.2.2 Data Screening ............................................................................. 60
5.3.3 Estimation of Aggregated Link Speeds ............................................. 61
5.3.3.1 Allocation of State Estimates....................................................... 61
5.3.3.2 Aggregation of Link Speeds ........................................................ 62
5.4 Simulation Results and Evaluation ........................................................... 62
5.5 Summary ................................................................................................... 69

6 Conclusions 71
6.1 Concluding Remarks ................................................................................ 72
6.2 Ongoing Work: The System Demonstrator .............................................. 72
6.3 Future Work .............................................................................................. 73

Bibliography 75
xii

ABBREVIATIONS AND ACRONYMS

2D Two Dimensional

2G 2nd Generation Cellular Radio System

3G 3rd Generation Cellular Radio System

3GPP Third Generation Partnership Project

4G 4th Generation Cellular Radio System

A-GPS Assisted Global Positioning System

A-GNSS Network-assisted GNSS

AE Absolute Error

API Application Programming Interfaces

AWGN Additive White Gaussian Noise

BTS Base Transceiver Station

BS Base Station

CDF Cumulative Distribution Function

Cell-ID Cell IDentification

CN Core Network
 xiii

CRLB Cramer-Rao Lower Bound

COST European Cooperation in Science and Technology

CVB Cumulative Virtual Blanking

DF Data Fusion

DS Data Screening

EC European Community

FCC Federal Communication Commission

E-CID Enhanced Cell ID

eNode B Evolved UMTS terminology for Base Station

E-OTD Enhanced Observed Time Difference

E-UTRAN Evolved UMTS Terrestrial RAN

FDD Frequency-Division Duplexing

GBA Generic Bootstrapping Architecture

GCC Generalized Cross-Correlation

GDOP Geometric Dilution Of Precision

GERAN GSM EDGE RAN

GNSS Global Navigation Satellite System

GPS Global Positioning System

GSM Global System for Mobile communications

ICT Information and Communication Technology

ILD Inductive Loop Detector

INS Inertial Navigation System

ITS Intelligent Transportation System

IPDL Idle Period DownLink

xiv

JSR Java Specification Request

J2ME Java 2 Platform, Micro Edition

KF Kalman Filter

LCS LoCation Services

LBS location-Based Services

LMU Location Measurement Unite

LOS Line-Of-Sight

LS Least Squares

LTE Long Term Evolution

MC Monte Carlo

ML Maximum Likelihood

MS Mobile Station

MSE Mean Square Error

NLOS Non-Line-Of-Sight

Node B UMTS terminology for Base Station

OBU On-Board-Unit

OSM OpenStreetMap

OTD Observed Time Difference

OTDOA Observed Time Difference Of Arrival

OVSF Orthogonal Variable Spreading Factor

P-CPICH Primary Common Pilot Channel

PN Pseudo-Noise

PV Probe Vehicle

QPSK Quadrature Phase-Shift Keying

 xv

RAN Radio Access Network

RFID Radio Frequency Identification

RMS Root Mean Square

RMSE Root Mean Square Error

RSE Root Square Error

RTD Real Time Difference

RTT Round Trip Time

SFN System Frame Number

SUMO Simulation of Urban MObility

TA Timing Advance

TDOA Time Difference Of Arrival

TIS Traffic Information System

TOA Time Of Arrival

TTFF Time-To-First-Fix

UE User Equipment

UMTS Universal Mobile Telecommunications System

U-TDOA Uplink Time Difference of Arrival

UTM Universal Transverse Mercator

UTRAN UMTS Terrestrial RAN

WCDMA Wide-band Code Division Multiple Access

WGS84 World Geodetic Standard 1984

WLAN Wireless Local Area Network

WPAN Wireless Personal Area Network

XML Extensible Markup Language

CHAPTER 1

INTRODUCTION

Traffic congestion is a universal problem that has enormous personal, business,

safety, and environmental consequences. The delays and inconvenience caused by
traffic jams deteriorate the quality-of-life of people sitting in never ending queues,
lead to money losses to business owners waiting for goods deliveries, prevent first
responders from accessing the scene of an emergency, and have a detrimental effect
on the environment due to the extra emissions of air pollutants. Therefore, if traffic
congestion can be alleviated properly, several social issues can be solved by saving
time and fuel, reducing greenhouse gas emissions and stress.
One way to address this problem is to enable travelers to take fast and informed
decisions through real-time traffic information. The traffic information system
(TIS) is such a system that provides travelers with useful traffic information in order
to assist their route decision making. TISs take advantage of the rapid advances in
sensor, computer, electronic and communication technologies. State-of-the-practice
TISs take two general forms: sensor-based and cellular network monitoring. Sen-
sor-based TIS is expensive to deploy and maintain; it covers only a small fraction of
roadways. Cellular network monitoring TIS can solve the issues of high cost and
limited coverage of the former; however, it suffers large variations in accuracy.
Recently, a new approach has been considered that uses mobile phones (enabled
with positioning features) as traffic data collection devices. This new approach can
provide higher location accuracy, and thus more accurate traffic data. It also takes
advantage of new features of modern mobile phones and offers an opportunity for
development of efficient and sustainable traffic systems, which meet society needs.

1
2 CHAPTER 1. INTRODUCTION

1.1 Background
1.1.1 The Use of Mobile Phones in ITS
Since the 1980s, transportation authorities have moved to advanced information and
communication technologies (ICTs) to make transport safer, more efficient, and
less polluted. This created a new field called intelligent transportation systems
(ITSs), which may take many different forms. ITSs range from simple ra-
dio-frequency identification (RFID) transponders based toll collection systems, to
different variations of traffic monitoring and broadcasting systems, and to more
sophisticated traffic management systems, which control urban traffic and give
automatic route directions to individual drivers with navigation equipment.
For instance, ITS research has been dedicated for finding solutions to the global
problem of traffic congestion. This research focuses on monitoring road traffic and
then disseminating real-time information to drivers. Traffic monitoring systems
deployed until now, use data collected mainly from roadside infrastructures, e.g.,
passive sensors, radars, and video cameras. One challenge to the existing systems is
the lack of accurate, up to date traffic data for the entire road network due to the fact
that deployment of infrastructures at wide scales is cost prohibitive. One alternative
is to employ an ad-hoc network, which includes wireless sensors or GPS-equipped
vehicles. However, these systems require on-board units (OBUs). In addition, the
ad-hoc network may not work properly when the density of vehicles is not adequate
for data to be transmitted between them [1].
By comparison, cellular networks have already been widely deployed and can
provide large population coverage. In addition, the ability to locate mobile phones
inside the network has emerged as a key facility of existing and future generation
mobile communication systems. In this context, ITS based on mobile positioning
technologies (either network-based or handset-based) can provide opportunity to
obtain traffic information in a cost effective manner. In recent years, the increasing
penetration of the positioning-enabled mobile phones makes them more attractive
as vehicle locators and traffic sensors, since an extensive spatial and temporal
coverage is potentially guarantied [2]. Moreover, as indicated in [3], advances in
mobile positioning are now maturing to the point where the positioning accuracy is
sufficient to support the estimation of traffic states on roadways.

1.1.2 Smart Traffic Information System: An Overview

A generic view of the proposed mobile phone-based smart traffic information
system is presented in Figure 1.1. In this proposed system, positioning-enabled
mobile phones are used to locate the vehicles. The mobile phones are also utilized
1.1 BACKGROUND 3

as on-board processing units, which for instance can make appropriate decisions to
protect user’s privacy. In addition, these switched-on mobile phones are employed
as probes to collect traffic data used for real-time urban road traffic state estimation.
The system’s main components include a mobile application (i.e. the MobiTraS [4])
running on each on-board mobile phone and a central server where algorithms, such
as location data processing, traffic state estimation, user privacy protection, etc., are
running in real-time. Additional components of the system are location service
providers, i.e., satellites and the cellular network, used for communication between
different parts and for positioning. The mobile phone's location is calculated using
the embedded GPS receiver and the network-assisted information. Then the mobile
application sends location updates to the server according to an algorithm that
prevents revealing the user's private information. The communication between
server and client is implemented through the cellular network, which the subscriber
is connected to. In order to preserve the security of the system, a secure channel is
established between two parts, which augments the standard generic bootstrapping
architecture (GBA) by anonymous authentication [5]. The server collects locations,
processes them to estimate the traffic and then send to each client personalized and
dynamic information. As shown in Figure 1.1, the information presented on mobile
phone display consists of an area map with the user’s location and colored road
segments showing traffic conditions.

Figure 1.1: Mobile Phone-based Smart Traffic Information System.

4 CHAPTER 1. INTRODUCTION

1.2 Problem Formulation

Although the motivation behind using mobile phones in TIS is obvious, there are
still a range of issues that must be addressed before deploying it to the real world: 1)
the technical issues, e.g., data quality in terms of accuracy and reliability 2) the legal
issues, e.g., safety and privacy of mobile phone users, 3) the institutional issues, and
4) the cost issues.
In this thesis, two major problems in the operational design of the proposed smart
traffic information system (described in Section 1.1.2) are addressed. The first one
concerns positioning and tracking the mobile phone, traveling on board a vehicle,
all around the journey with appropriate accuracy; the second one concerns pro-
cessing the location data collected by mobile phones to estimate in real-time the
traffic state on urban road networks.

Problem 1
The problem of interest consists in analyzing standard mobile positioning methods,
proposing and implementing a solution for the mobile phone-based vehicle location
in our system. Three objectives are identified from this problem.
 Objective 1: review the existing standardized mobile positioning methods
and find possible solutions (e.g., combinations of the standard methods) to
the vehicle location application.
 Objective 2: develop a mobile positioning simulator in order to provide
reasonable estimations on positioning accuracies through modeling the ra-
dio propagation environment in a specific cellular network.
 Objective 3: implement one of the proposed hybrid solutions (achieved in
Objective 1), which can find the optimal location estimates of the mobile
and track its trajectory in an emulated vehicular scenario.

Problem 2
The problem of interest lies in proposing, simulating and evaluating an approach for
mobile phone-based urban traffic state estimation. Two objectives are targeted in
this problem.
 Objective 1: propose a method that makes use of the mobile positioning and
tracking data to perform traffic state estimation on urban road networks.
 Objective 2: show the potential feasibility of the mobile phone-based smart
traffic information system through a simulation testbed.
1.3 THESIS OUTLINE AND CONTRIBUTIONS 5

1.3 Thesis Outline and Contributions

The outline of the thesis is as follows. Chapter 2 presents a survey on vehicle loca-
tion using mobile phones; it serves as an introduction to the following two chapters
by investigating the standard mobile positioning methods and suggesting hybrid
solutions to these methods. Chapter 3 demonstrates the models and methods used to
develop the UMTS mobile positioning simulator, which is targeted for vehicle
location applications. Chapter 4 is devoted to detail implementations of a mobile
phone-based hybrid vehicle positioning and tracking solution, proposed in Chapter
2. Chapter 5 shows how positioning and tracking information from mobile phones
can be used to estimate real-time urban road traffic states. Finally, Chapter 6 con-
cludes the thesis. A summary of each chapter is given next.

Chapter 2
This chapter aims to find practical solutions to vehicle location using wireless wide
area network in the proposed mobile phone-based smart traffic information system.
A thorough investigation has been performed with the focus on standard location
methods specified for different generations of the radio access networks. Based on
the comparative survey, combinations and hybrid solutions to the standard methods,
particularly in UMTS, are suggested in order to improve the location performance.
The chapter is based upon:
 S. Tao, S. Rodriguez, and A. Rusu, “Vehicle Location Using Wireless Wide
Area Network,” in Proceedings of Joint IFIP Wireless and Mobile Net-
working Conference, Oct. 2010, pp. 1-6.

Chapter 3
This chapter presents a tool for simulating UMTS mobile positioning in vehicular
environment. The primary function of this simulator is to locate the UMTS mobile
in realistic propagation environments. This tool first models network configurations
and radio propagations in a vehicular scenario. Based on the system-level radio
network model, it then simulates the pilot signal transmitted by a base station to a
mobile through the standard 3GPP WCDMA FDD downlink. The received pilots at
the mobile are processed to obtain the time-difference-of-arrival estimates, which
are used to construct the hyperbolic equations for mobile position calculation. The
chapter is based upon:
 S. Tao, S. Rodriguez, and A. Rusu, “ UMTS Mobile Positioning Simulator
for Vehicle Location,” in Proceedings of IET International Conference on
Wireless, Mobile and Multimedia Networks, Sept. 2010, pp. 344-347.
6 CHAPTER 1. INTRODUCTION

Chapter 4
In this chapter, a hybrid scheme which combines location estimates from both
A-GPS and OTDOA is proposed for mobile phone-based vehicle positioning and
tracking. In the proposed scheme, the Kalman filtering technique is exploited to
fuse the noisy OTDOA and A-GPS location measurements at the state-vector level.
Moreover, statistical evaluations and field tests are performed to obtain reasonable
noise models for both location measurements. Numerical results demonstrate that
the hybrid scheme can provide better position and velocity estimations than each
individual location method. The chapter is based upon:
 S. Tao, V. Manolopoulos, S. Rodriguez, M. Ismail and A. Rusu, “Hybrid
Vehicle Positioning and Tracking Using Mobile Phones,” in Proceedings
of IEEE International Conference on ITS Telecommunications, Aug. 2011,
pp. 315-320.

Chapter 5
In this chapter, a method of real-time urban traffic state estimation is presented. The
proposed method, taking advantage of the recently booming A-GPS mobile phones,
potentially solves the problems in the current state-of-the-practice traffic systems.
Based on the microscopic traffic simulation and field tests, “realistic” A-GPS mo-
bile probe data is emulated and “ground truth” traffic data is generated. The A-GPS
location samples are firstly processed by Kalman filtering and data screening. The
resultant state estimates are then allocated to road links through simple
map-matching. By aggregating speed estimates on each road link, traffic states are
determined. The achieved simulation results suggest that reliable average link speed
estimations can be generated, which are fundamental for the subsequent develop-
ment of a system demonstrator. The chapter is based upon:
 S. Tao, V. Manolopoulos, S. Rodriguez, and A. Rusu, “Real-time Urban
Traffic State Estimation with A-GPS Mobile Phones as Probes,” accepted
for publication in SCIRP Journal of Transportation Technologies.
CHAPTER 2

VEHICLE LOCATION USING MOBILE PHONES:

SURVEY AND PROSPECTS

Development of vehicle location and navigation systems has a long history, even
though these systems have only recently started to reach the world market. The
current solution to vehicle location and navigation is an inertial navigation system
(INS). Modern INS is normally equipped with (relatively) expensive OBUs. On the
other hand, mobile phones, taking advantage of the development of location ser-
vices in modern communication systems, can be a natural alternative to the posi-
tioning module in INS. Moreover, the communication functionality of mobile
phones can replace the ITS specific communication module. The high resolution
(touch) screen of modern mobile phones can also be an excellent substitute for a
separate human-machine interface module. Economically, it is desirable to develop
vehicle positioning techniques that take advantage of the well-established wireless
networks and the pervasive low cost mobile phones. In this sense, the problem of
positioning a vehicle becomes the problem of accurately locating the mobile phone
traveling on-board. In order to tackle this problem, this chapter first investigates the
standard mobile location methods and exploits this information in ITSs, especially
in vehicle location applications. Hybrid solutions are then proposed based on the
well-established and standardized location methods.

7
8 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

2.1 Mobile Positioning and Its Performance Metrics

2.1.1 Calls for Mobile Positioning
Location services (LCS) refers to services that utilize the position estimate of a
mobile station [6]. As specified in [7], there are four categories of location services:
commercial LCS, internal LCS, emergency LCS, and lawful intercept LCS. The
demand to locate mobile phones in the case of emergency calls is commonly ac-
cepted as the main driving force for LCS. This demand was initiated by the U.S.
FCC (Federal Communication Commission), which ruled that the calling party of
all emergency calls (911) in the United States should be located with a defined
degree of accuracy [8], as specified in Table 2.1. In Europe, the EC (European
Community) defines positioning performance requirements for their E-112 location
systems [9], as specified in Table 2.2. The above requirements are for emergency
location. In contrast, vehicle positioning requires far finer resolution and greater
accuracy. The mobile application in our proposed traffic system is a commercial
LCS. Such an application provides a value-added service to subscriber. In [10],
descriptions of possible location-based services are given. Categories B3.2 “Traffic
Monitoring” and B5.1 “Navigation” fit best to the goals of our proposed system.
Location technologies are increasingly designed to meet the requirements for cer-
tain location-based services (LBS), rather than simply to meet the mandatory reg-
ulations (as per Tables 2.1 and 2.2). Table 2.3 lists some specific LBS and their
expected accuracy ranges [10].
Table 2.1: Accuracy required from FCC.
Solutions 67% of Calls 95% of Calls
Handset-based 50 meters 150 meters
Network-based 100 meters 300 meters

Table 2.2: E-112 accuracy guidelines.

Urban Suburban Rural Crossroads
50 meters 50 meters 100 meters < 100 meters

Table 2.3: Typical LBS position accuracy.

Service Types Accuracy Range
Fleet Management 125m - Cell Radius
Network Planning 10m - Cell Radius
Person Tracking 10m - 125m
Navigation and Route Guidance 10m - 125m
Traffic Congestion Reporting 10m - 40m
2.1 MOBILE POSITIONING AND ITS PERFORMANCE METRICS 9

2.1.2 Vehicle Location: Requirements and Metrics

In order to accomplish vehicle location and navigation functions, the positioning
system should have the ability to determine its current position within 20 meters of
its actual location over 90% of its travel time [11]. This 20 meters resolution is
required to differentiate between closely spaced parallel road segments, and
consequently minimizes the map-matching errors. In addition, as a supplement to
accuracy, the 90% of time coverage is demanded.
Here we consider the accuracy as how closely the location measurements agree
with the actual location of the targets. The positioning accuracy can be defined by
root mean square (RMS) or standard deviation. Specifically, in the 2D case, the
RMS is composed of the standard deviations of two 1D axes [12], as illustrated in
Figure 2.1. The standard deviation can be calculated as:

s2D = 2sN 2 = 2sE 2 (2.1)

In addition, the root mean square error (RMSE) can be defined as an accuracy
evaluation of multiple location measurements [13]:

1 N 2
RMSE =
N
åk =1 éë xmeasured (k ) - xtrue ùû (2.2)

Figure 2.1: The confidence circle.

10 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

More generally, accuracy, geometric dilution of precision (GDOP) and coverage

are three important performance measures for positioning systems. A common
measure of accuracy is the comparison of the mean-square-error (MSE) of the
location estimate with the theoretical MSE based on the Cramer-Rao lower bound
(CRLB) [14]. The MSE in two dimensions is given by:

MSE = e = E [(x - xˆ)2 + (y - yˆ)2 ] (2.3)

where ( x ,y ) are the coordinates of the target and ( xˆ,yˆ ) is the estimated position.
In addition, the GDOP provides a measure of the effect of the base stations’ con-
figuration on the location estimate [14]. It is defined as the ratio of the RMS position
error to the RMS ranging error, which is given by:

E[(x - xˆ)2 + (y - yˆ)2 ]

GDOP = (2.4)
sr

where sr denotes the fundamental ranging error. Another evaluation metrics of

positioning system is coverage. It is defined as the proportion of an area of interest
that is provided with an acceptable level of service by the positioning system [15].

2.2 Legacy Systems and Mobile Positioning

2.2.1 Legacy Vehicle Positioning Systems
Before digging into the location technologies based on wireless wide area networks,
let us first quickly review the legacy positioning systems. Historically, vehicle
positioning systems can be divided into three basic classes [15]: signpost systems,
wave-based systems, and dead reckoning systems.
Signpost systems measure position based on the fact that the vehicle is located
close to a reference point, i.e., a signpost. The “signpost” can be a person standing
by the road or some dispersed roadside beacons. Two important elements of this
beacon system are a vehicle-mounted “tag” and the roadside unit.
Wave-based systems use propagation properties of waves to determine position,
e.g., a radar system. The position of a vehicle is measured relative to reference sites.
For GPS, each satellite is a reference site. The on-board GPS receiver receives
signals from satellites and uses time-of-arrival information to calculate its position.
Dead reckoning systems rely on sensing the vehicle’s acceleration or velocity,
which are used to determine the track of the vehicle. In dead reckoning systems, a
compass and odometer or a gyroscope and accelerometer can be used as sensors.
2.2 LEGACY SYSTEMS AND MOBILE POSITIONING 11

All of the positioning systems described above are inevitably linked with com-
munication systems. In other words, the OBUs in vehicles are not only inertial
sensors and GPS receivers, but also dedicated mobile communication equipments.
Hence, it would be desirable to develop vehicle positioning techniques that take
advantage of the well-established wireless wide area networks and the pervasive
low cost mobile phones.

2.2.2 Mobile Positioning Principles

This section explains the principles of mobile positioning methods. Among the
basic positioning methods [12], three of them are of particular interest for mobile
location estimation in wireless wide area networks. These three approaches (i.e.,
proximity sensing, circular lateration and hyperbolic lateration) will later be em-
ployed by location technologies in the following sections.

2.2.2.1 Proximity Sensing

Proximity sensing is an intuitive way to estimate the rough position of a mobile
terminal. Geolocation of the mobile terminal can be derived from the coordinates of
a base station (BS), which can sense that mobile terminal (or which the mobile
terminal senses). As shown in Figure 2.2, the known position (X1,Y1) of the BS1
that either sends or receives a pilot signal is simply assumed to be (near) the position
of the mobile terminal [12]. In this proximity sensing case, the expected deviation
error is limited by the radius of this sensing area.

Figure 2.2: Illustration of proximity sensing.

2.2.2.2 Circular Lateration

For circular lateration, assume that the distances di between the mobile terminal
and a set of BSs are known. Given three base stations with known coordinates
(Xi ,Yi ) , we can then uniquely determine the mobile terminal’s position ( x ,y ) .
12 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

As illustrated in Figure 2.3 (a), the mobile terminal can be located by the intersec-
tion of the three circles, with distance di as radius. The mobile terminal’s position
( x ,y ) can be analytically determined from the following equations:
ìï d = (X - x )2 + (Y - y )2
ïï 1 1 1
ïï 2 2
íd2 = (X2 - x ) + (Y2 - y ) (2.5)
ïï
ïïd = (X - x )2 + (Y - y )2
ïî 3 3 3

In the realistic case, the measured distance m i may have a derivation error  i
from the actual distance di . Then, the three circles will not intersect at a point
( x ,y ) , but the intersection becomes an error margin, as shown in Figure 2.3 (b).

2

3
1

Figure 2.3: Illustration of circular lateration: (a) the ideal intersection, and (b) the
measurement error margin.

2.2.2.3 Hyperbolic Lateration

For hyperbolic lateration, the mobile terminal’s position is determined by the dis-
tance differences (di - d j ) instead of the absolute distance di . A hyperbola is
defined to be the set of all points for which the difference from two fixed points is
constant [12]. Then two BSs with known locations determine a hyperbolic curve.
The intersection of two hyperbolic curves can determine a unique point. Therefore,
three BSs are required to locate a mobile terminal, as shown in Figure 2.4 (a). The
mobile terminal’s position ( x ,y ) can be derived from:
2.3 MOBILE POSITIONING STANDARDIZATION IN GENERATIONS OF 13
RANS

ìïd - d = (X 2 - x )2 + (Y2 - y )2 - (X1 - x )2 + (Y1 - y )2

ïï 2 1
í (2.6)
ïïd - d = (X 3 - x )2 + (Y3 - y )2 - (X1 - x )2 + (Y1 - y )2
ïî 3 1

As in the circular case, hyperbolic lateration has a derivation error Dd , which is

i -d j

due to the inaccurate measurement of distance differences. The two hyperbolas

intersect in an error margin area, as illustrated in Figure 2.4 (b).

Dd
2 -d 1

Dd
3 -d 1

Figure 2.4: Illustration of hyperbolic lateration: (a) the ideal intersection, and (b) the
measurement error margin.

2.3 Mobile Positioning Standardization in

Generations of RANs
In the past decade, the telecommunication standards organization 3GPP (Third
Generation Partnership Project) has been endeavoring to incorporate location
techniques (introduced in Section 2.2.2) into their standards. LCS (Location Ser-
vices) defined by 3GPP are not exclusively for “3G” networks but rather cover
several generations of RAN (radio access network): GERAN (GSM EDGE RAN),
UTRAN (UMTS Terrestrial RAN) and E-UTRAN (Evolved UMTS Terrestrial
RAN). EDGE stands for “Enhanced Data rate for Global Evolution”. The evolution
is with respect to the GSM (Global System for Mobile communication), i.e., a “2G”
system. UMTS stands for “Universal Mobile Telecommunications System”, which
commonly refers to WCDMA (Wide-band Code Division Multiple Access), i.e., a
“3G” system. E-UTRAN refers to LTE (Long Term Evolution) i.e., a “4G”system
(candidate)1.

1
LTE-Advanced meets the ITU-T's requirements for a 4G system.
14 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

In specification TS 22.071 [10], 3GPP provides an overall description of location

services and requirements (i.e., LCS stage 1). In addition, a description of the
function model, the LCS architecture and the message flow is given in specification
TS 23.271 [7] (i.e., LCS stage 2). Stage 1 is described from the service subscriber’s
points of view and does not include the details of how the service is realized. Stage
2 provides methods to support mobile location services for operators, subscribers,
and third party service providers. The following subsections describe the standard-
ized positioning methods supported in each RAN.

2.3.1 Standard LCS in GERAN

3GPP TS 43.059 (the functional description of LCS in GERAN) [16] specifies the
location methods supported in GSM. There are four positioning methods supported
in the GSM/EDGE RAN: Timing Advance (TA), Enhanced Observed Time Dif-
ference (E-OTD), Global Navigation Satellite System (GNSS) based positioning
(A-GNSS), and Uplink Time Difference Of Arrival (U-TDOA).
In GSM, the TA (Timing Advance) value corresponds to the time a signal takes
from the MS (Mobile Station) to the BTS (Base Transceiver Station). In a proximity
positioning method, the use of TA is a way to limit location uncertainty in cell-ID
(cell identification) technology. The cell-ID of the corresponding BTS is returned
along with the TA value.
The E-OTD method is based on the measurement of relative time of arrival of the
signals (from nearby pairs of BTSs). The position of the MS is determined by the
hyperbolic lateration (trilateration) illustrated in Section 2.2.2. To obtain accurate
trilateration, at least three BTSs with good geometry are required to be within the
range of the MS. This is basically a TDOA (Time Difference of Arrival) method.
For it to work properly, additional location determination software and hardware
must be added to the network [17].
A Global Navigation Satellite System (GNSS) refers to a satellite system that is
used to pinpoint the geolocation of a user's receiver. GNSS uses the same principle
as in Time Of Arrival (TOA) system which determines the MS position by the
circular lateration illustrated in Section 2.2.2. In order to resolve the latency and
sensitivity issues in GNSS, network-assisted is GNSS standardized, which aug-
ments the weak signal received from the satellites with network-assisted data.
The U-TDOA positioning method is based on the measurement of time of arrival
of the signal sent from the MS and received at BTSs. With known BTS coordinates,
the MS’s position can be calculated via hyperbolic lateration. Although based on a
similar principle, the position calculation in U-TDOA is done by the network in-
frastructure, while in E-OTD the position is calculated by the MS.
2.3 MOBILE POSITIONING STANDARDIZATION IN GENERATIONS OF 15
RANS

2.3.2 Standard LCS in UTRAN

3GPP TS 25.305 (the functional description of UE positioning in UTRAN) [18]
provides means to support the calculation of the user equipment (UE2) position.
There are four standard positioning methods supported in UMTS Terrestrial RAN:
Cell-ID (CID), Observed Time Difference Of Arrival (OTDOA), Network-assisted
GPS (A-GPS), and Uplink-Time Difference Of Arrival (U-TDOA).
In UMTS, the Cell-ID based method estimates the position of a UE by knowing
its serving Node B3. The estimated position of the UE is the position of the serving
Node B. Similar to using TA in GSM, additional techniques can be employed in
UMTS to improve accuracy of the CID method. One of them is measurement of
signal round trip delay, i.e., round-trip-time (RTT). Due to the larger bandwidth and
shorter chip duration in UMTS, the accuracy of RTT is significantly higher than the
resolution of the TA-based techniques in GSM [19].
OTDOA is similar to E-OTD since it is a UE-based technology using the TDOA
positioning method. The UE's position is determined by the intersection of hyper-
bolas for at least two pairs of Node Bs, using the same principle demonstrated in
Figure 2.4 (a). In some conditions, UEs may have difficulty hearing a sufficient
number of cells needed for TDOA calculations. One solution, specified by 3GPP for
this problem, is an idle period downlink (IPDL) where the serving Node B intro-
duces idle periods in the downlink to improve the hearability of weaker neighboring
Node Bs.
As in GSM, a network-assisted GNSS method is also standardized. In particular,
network-assisted GPS (A-GPS) is specified in detail as a separated clause in the
standard. Two types of A-GPS methods are specified in UTRAN, namely UE-based
and UE-assisted. Computation of the position can either be performed in UTRAN
for UE-assisted or in the UE for UE-based. In the former, the UE employs GPS
receiver with reduced complexity, whereas in the later, a full GPS receiver is re-
quired in the UE.
The network-based U-TDOA positioning method in UMTS is basically the same
as in GSM. Likewise, the difference between U-TDOA and OTDOA lies in whether
the measurement is done by the network or by the UE. In UTRAN, the U-TDOA
shows significant advantage over its handset-based counterpart [20]: no IPDL in
OTDOA is needed; more complicated processing becomes affordable and therefore
the positioning accuracy can be improved; no need to modify the UE nor the node
B; network management becomes easier, as it is less complex to upgrade the soft-
ware in network entities than it is to upgrade millions of terminals.

2
User Equipment is a UMTS terminology for Mobile Station.
3
Node B is a UMTS terminology for Base Station.
16 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

2.3.3 Standard LCS in E-UTRAN

3GPP TS 36.305 (the functional description of UE positioning in E-UTRAN) [21]
defines the E-UTRAN (Evolved UTRAN) UE positioning architecture, functional
entities, and operations to support positioning methods. The E-UTRAN positioning
capabilities are still under discussion; and the aim is to be compatible to other types
of access and position methods. In the latest release (Release 10), three standard
positioning methods are supported for E-UTRAN: Enhanced Cell ID (E-CID),
Downlink Positioning (Downlink), and Network-assisted GNSS (A-GNSS).
In E-UTRAN, Enhanced Cell ID (E-CID) method estimates the UE with the
knowledge of its serving Evolved Node B (eNode B4) and cell. This E-CID method
uses additional measurements, for instance the UE and/or E-UTRAN resources to
improve the UE location estimate.
In the downlink positioning method, the UE position is estimated based on
measurements (taken at the UE) of pilot signals from several eNode Bs. It can be
compared to OTDOA in UTRAN, since this downlink method is based on measured
time differences observed by the UE. It is worth noting that this downlink posi-
tioning method is still under discussion and cannot be considered definitive yet.
Similar as in UTRAN, when GNSS is designed to cooperate with the E-UTRAN,
the network assists the UE GNSS receiver to improve the performance. Two as-
sisted modes are supported here, namely UE-based and UE-assisted.

2.4 Classification and Discussion

As stated in Section 2.3, various location technologies have been developed for use
in wireless wide area networks. Despite their numerous variations, these standard-
ized methods can be systematically classified. In this section, comparisons and
discussions will focus on several interesting aspects of these technologies.

2.4.1 Major Categories of Wireless Location

Mobile positioning in wireless networks can be divided into two major categories:
network-based and handset-based methods. In the former, the handset position is
determined purely from measurements taken at network infrastructures. Thus, no
modification is required in the handset. In the latter, a position is calculated based
upon measurements in handset. The handset either reports the measurement results
to the network or does the calculation itself. Hence, the handset needs to be
equipped with additional software and hardware. Table 2.4 lists the standard mobile
positioning methods described in Section 2.3, together with their impact on both
network infrastructure and handset.
4
eNode B is an Evolved UMTS terminology for Base Station.
2.4 CLASSIFICATION AND DISCUSSION 17

Table 2.4: Categories of standard mobile positioning.

Standard Positioning Methods Modification Requirements
RANs
Network-based Handset-based Network Handset
CID + TA - Software No

GSM - E-OTD Hardware Software

(2G) - A-GNSS Hardware Both
U-TDOA - Hardware No
CID + RTT - Software No

UMTS - OTDOA-IPDL Hardware Software

(3G) - A-GPS Hardware Both
U-TDOA - Hardware No
E-CID - Software No
LTE
- A-GNSS Hardware Both
(4Ga)
- Downlink Hardware Software
a. LTE is branded as “4G”; although it is not been fully comply with 4G requirements.

A network-based method has the obvious advantage that it can be directly applied
to any legacy network. For operators, this feature is favored since they want to serve
as many subscribers as possible to make greater profit. This method is also benefi-
cial to the subscribers. Firstly, they do not need to invest in new devices and sec-
ondly the network-based measurement significantly reduces the power consump-
tion of their devices. However, the network-based location not only utilizes a cer-
tain amount of network resources, but also involves insignificant changes in net-
work infrastructures. On the other hand, although the handset-based method de-
mands expensive terminals with high power consumption, it is still worth the price:
this approach reduces utilization of network resources; and positioning is less de-
pendent on the network. Hence, the handset-based location is applicable even to
networks which have low capacity and are not specially designed for location
services. Moreover, since the handset is not limited by the network, the location
accuracy can be improved, as needed, by making more measurements. In addition,
in some situations, this method is more secure than its network-based counterpart
since the position and tracking information are not available in the network [22].
18 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

2.4.2 Location Methods in Generations of RANs

The standard mobile positioning methods, whether in GSM, UMTS, or LTE, fall
into four classes: cell coverage based positioning, downlink observed time differ-
ence, network assisted GNSS, and uplink TDOA. Table 2.5 gives a summary of
different techniques specified for each generation of RANs.
Table 2.5: Summary of location methods in generations of RANs.
Location Principle Standard Wireless Estimated
Methods Illustration Techniques Networks Accuracy

CID + TA GSM 80-450m

Cell
Coverage CID + RTT UMTS 180-550m
Based
E-CID LTE -
E-OTD GSM 150-500m
Downlink
Observed Time OTDOA-IPDL UMTS > 30m
Difference
Downlink LTE -
GSM
Network > 10ma
Assisted A-GPS UMTS
GNSS
LTE -
GSM
Uplink Time U-TDOA 40-120m
Difference of UMTS
Arrival
- LTE -
a. This accuracy cannot be achieved in some environments, e.g. urban canyon and dense building.

As discussed above, the second and earlier generation cellular networks were
not dedicated to LCS. The limited bandwidth makes them less attractive to incor-
porate positioning functionalities, since the bandwidth of cellular signal determines
the precision that can be attained in timing measurement [22]. On the contrary, the
third and higher generation networks have protocol provisions for LCS and their
wider bandwidths enhance the positioning accuracy. The standards for E-UTRAN,
however, are not yet finalized. Additionally, LTE is actually defined as a transi-
tional generation, called pre-4G or 3.5G, rather than 4G. The 3G, on the other hand,
has been deployed pervasively worldwide and welcomed its 500 millionth customer
in Jan. 2010, according to the UMTS Forum [23]. Nevertheless, LCS will not be
limited to the 3G networks, and should be forward compatible to the next generation
2.4 CLASSIFICATION AND DISCUSSION 19

wireless wide area network. Despite this trend, further discussions and the proposed
solution in the coming sections will focus on UMTS (a 3G network), because it is
well established, standardized and widely implemented.

2.4.3 Performance Evaluation of 3G UE Positioning

It is challenging to compare these 3G mobile positioning methods since there are so
many performance metrics to evaluate. In [24], several performance criteria are
defined to evaluate different positioning technologies: accuracy, availability, reli-
ability, latency, and applicability.
Among them, accuracy and availability are of major importance. Accuracy refers
to how closely the location measurements agree with the actual location of the
target. Availability is related to environmental situations where positioning is pos-
sible. Availability is characterized by the following scenarios: remote, rural, sub-
urban, urban, indoor, and underground. For example, in E-112 guidelines, the
accuracy is defined according to these availability classifications. Table 2.6 gives
the comparison of 3G positioning accuracy in rural, suburban, and urban areas,
which are valid for our vehicle location application [22], [25]. As it can be seen in
the table, only A-GPS meets the 20 m requirement.
In addition, reliability is defined as the ratio of successful positioning attempts
out of all attempts made. For instance, in the FCC-E911 requirements, 67% or 95%
of successful calls is the reliability constraint. Latency refers to the time from
power-up to the first location measurement being obtained. This is better known as
TTFF (Time-To-First-Fix) in GPS terminology. Applicability is more related to
economic than technical issues. It covers aspects of power consumption, processing
and signaling load, hardware and software size, deployment and operational cost.
Inferred from [26], a comparison of 3G positioning methods in terms of the above
three metrics is given in Table 2.7.
Table 2.6: Comparison of 3G positioning methods: accuracy and availability.

Positioning Accuracy vs. Availability

Methods Rural Suburban Urban
CID + RTT 250-35km 250-2500m 50-550m
A-GPS 10m 10-20m 10-100m
OTDOA-IPDL 50-150m 50-250m 50-300m
U-TDOA 50-120m 40-50m 40-50m
20 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

Table 2.7: Comparison of 3G positioning methods: reliability, latency and

applicability.
Positioning Methods Reliability Latency Applicability
CID + RTT high 1-5s high
a
A-GPS medium 1-10s medium
OTDOA-IPDL medium <10s medium
U-TDOA medium <10s low
a. This latency is limited to the outdoor environment.

2.4.4 Further Discussion

Not surprisingly, the cell-based method shows the poorest accuracy which can be
easily explained due to its dependency on the cell size. However, its reliability and
attainability is outstanding, since this method only requires one base station for
positioning and does not require any change to either the network or the handset.
In contrast, a GPS-based method provides excellent performance in terms of
accuracy, especially in open-space. In stand-alone mode, which means the network
assistance functionality is off, the handset is able to determine its location even
outside the coverage of a UMTS network. However, GPS does not operate well in a
dense urban area or underground. This may be due to not enough satellites being
visible and GPS signal penetration being blocked by walls. Another problem for
stand-alone GPS is the long latency, e.g. signal acquisition time. Network-assisted
mode overcomes these problems by assisting the handset through providing the
correction data and rough UE position along with an accurate reference time. For
A-GPS, the best accuracy is achieved in rural area; while in an urban area it also
suffers from shadowing effects near high buildings, but much less than the pure
GPS. The reliability of A-GPS method is descent since the assistance data from the
network helps to improve the sensitivity of the GPS receiver.
OTDOA seems to offer a good compromise between purely network-based
method and purely satellite-based method. Its latency is acceptable, requiring less
than 10 seconds to calculate the position. Additionally, it shows much better ac-
curacy than cell-based methods, with tens of meters. Compared with A-GPS, which
requires a GPS receiver in handset, the cost of this method in the handset is rela-
tively modest. It only requires minor software updates in the handset. However, the
network modification is needed: Location Measurement Units (LMUs) should be
located in all cells and networked to the Core Network (CN). The reliability of this
method is not high, since it involves at least three Node Bs and the LMUs corre-
spondingly. This requirement cannot be always fulfilled in the real world.
2.5 THE PROPOSED HYBRID POSITIONING METHODS 21

Despite the existence of various OTDOA enhancements, e.g., the standard

OTDOA-IPDL and the CVB (Cumulative Virtual Blanking) [27], the uplink TDOA
methods have been developed and standardized to overcome the problems of
“hearability” in the downlink measurement. Since it is a completely network-based
method, no additional hardware or software needs to be included in the handset.
Thus, one advantage for this uplink method is that it can be applied to all the sub-
scribers. This method works quite well in urban and suburban environment, suf-
fering in some rural settings where the cell sites are not appropriately arranged.
However, U-TDOA requires more complicated measurement units in the network
in addition to the LMUs required by the downlink OTDOA. Thus the applicability
of this purely network-based method is low comparing to others. Moreover, since it
employs location estimation process through hyperbolic lateration, its reliability is
as high, for the same reason, as for OTDOA.

2.5 The Proposed Hybrid Positioning Methods

The mobile positioning techniques, especially the four standard LCS methods in
UMTS, have been presented and evaluated. None of them is clearly the “magic
bullet”, i.e., is perfectly suitable for vehicle location. Generally, network-assisted
GPS seems to be an attractive solution for our application. Moreover, A-GPS
phones are increasingly available in the global mobile phone market. According to
Canalys [28], the annual GPS-enabled smart phone sales are estimated to be above
90 million units. Nevertheless, as discussed in the last section, A-GPS has major
drawbacks that prevent its availability in certain scenarios, such as dense urban area
surrounded by high buildings and underground. This is not catastrophic, but indeed
a serious problem for vehicle navigation services. Hence, it is worthwhile to find an
inexpensive way to supplement the A-GPS method.
Inspired by the previous discussion of the standard location methods, it is
straightforward to expect that combinations can complement the downside of one
method with the advantages of another. In the past decade, attempts have been made
to pursue hybrid mobile location and tracking [29-32]. Nevertheless, they still leave
the doors open for further investigations. Firstly, most of them are based on the old
generations of location techniques and only a few of them use the 3G standard
location methods. Secondly, they aim to complement GPS performance not the
more recently evolved A-GPS. Thirdly, they barely target vehicle location and
navigation applications, which will make difference in performance requirements.
In the following subsections, three hybrid positioning methods are proposed. They
fall into two sets of combing considerations.
22 CHAPTER 2. VEHICLE LOCATION USING MOBILE PHONES: SURVEY
AND PROSPECTS

2.5.1 Combining Network-based with Handset-based

Methods
According to 3GPP active work program description [33], hybrid location has been
defined as the concurrent use of a single network-based location technology in
combination with a single mobile-based location technology in response to a single
location request. This combination helps to optimize location accuracy, increase
location yield, and minimize the latency of location delivery. UTRAN currently
allows for a hybrid location solution for technologies that use separate resources.
Based on this principle, two hybrid network-based/handset-based location methods
are presented below.

2.5.1.1 Hybrid of Cell-ID and A-GPS

Recall that the simplest and most reliable method for UE positioning is Cell-ID
based. However, its accuracy cannot meet the requirements of vehicle location and
navigation services. A wide range of enhancements for the basic Cell-ID have been
developed and standardized, as illustrated in Section 2.3. In the UMTS network, for
instance, the most applicable method is Cell-ID+RTT. It has been claimed to be
theoretically capable of estimating the distance between UE to BS with an accuracy
of 36 meters [19]. Moreover, there is a trend towards greater widespread of WLAN
(wireless local area network) and WPAN (wireless personal area networks). Con-
sequently, these Micro- and Pico- cells can be pervasively used in the future to
enable better estimation accuracy of position. A combination of Cell-ID and A-GPS
can then offer adequate positioning for vehicle location.

2.5.1.2 Hybrid of U-TDOA and A-GPS

An alternative to the network-based/handset-based hybrid solution is the use of
uplink TDOA instead of the Cell-ID method. As mentioned before, the U-TDOA
method does not require any modification to the handset, and offers better accuracy
than Cell-ID+RTT and OTDOA. There are two different ways to combine multiple
location technologies in GSM, suggested by [34]. One way is to combine the two
methods in a fallback arrangement, in which the system calls on one method when
performance of the other has deteriorated due to environmental conditions. The
other is to utilize the two simultaneously, mathematically combining the results to
provide enhanced results. Ideally, these two combinations can be applied to the
A-GPS and U-TDOA hybrid positioning in UMTS. For instance, in a “fallback
arrangement”, A-GPS would provide positioning functions when it is available, and
when it is not, the more robust network-based U-TDOA would take over.
2.5 THE PROPOSED HYBRID POSITIONING METHODS 23

2.5.2 Fusion of Two Handset-based Methods

Two handset-based methods may also be combined together to produce higher
accuracy and availability than a single technology. In UMTS, location estimation
can be based upon fusion from both the A-GPS and the OTDOA measurements.
The driving force for this fusion is obvious, similar to the one behind hybrid of
A-GPS and U-TDOA. Generally, OTDOA is preferable in dense urban areas where
more base stations are available; whereas, A-GPS provides better accuracy on
highways where three satellites are easily visible.
Integration of A-GPS and OTDOA measurements can also be seen as data fusion,
which is a method for merging the information obtained from different sources.
Different sources, however, are subject to different errors that would contribute
unequally to global estimation errors. To address this issue, adaptive data fusion
algorithms have been employed to better integrate different types of position in-
formation, as presented in [35].

2.5.3 Suggestions for Further Implementation

Based on what we have discussed so far, there are majorly three aspects to be further
investigated for implementing the mobile phone-based vehicle location. One is to
develop simulators to emulate the cell coverage based (Cell-ID) or the time based
(OTDOA and U-TDOA) positioning methods, since the network parameters re-
quired by these methods are not accessible by us5. Another is to find a smart fusion
of the above position measurements and the location data acquired by the A-GPS
method. The third one, which can be seen as the continuation to the problems of
vehicle geolocation estimation, is to track the vehicle. For this purpose, not only the
position coordinates, but also the velocity of the vehicle (in our scenario the mobile)
should be estimated. This effort commonly resorts to mobility tracking algorithms
in wireless networks.

5
The network parameters are commonly kept confidential by mobile operators.
CHAPTER 3

UMTS MOBILE POSITIONING SIMULATOR FOR

VEHICLE LOCATION

For vehicle location, the performance of mobile positioning will be affected by

real-life conditions in different outdoor environments. Thus, in order to evaluate the
performance, a reliable way is to perform field trials. However, it is expensive to
evaluate a system in this manner, in all environments [36]. One alternative to the
actual field test measurements is to develop a predictive simulation tool. The posi-
tioning simulator should be capable of providing reasonable estimations on posi-
tioning accuracies under the realistic conditions [37]. This chapter presents such a
positioning simulator that can help to analyse the OTDOA positioning supported by
the UMTS. This UMTS mobile positioning simulator consists of three parts. The
first part models the overall radio network at system level. It includes the network
topology, mobile mobility and radio propagation model. The second part is a link
level WCDMA FDD simulator. It covers the transmission code generation blocks, a
Rayleigh fading channel model, and the correlation reception blocks. The third part
is the position estimation composed of timing measurement and location calculation
algorithm. The system level model and the location algorithms are developed in
MATLAB. The link level simulator is implemented in Simulink, with extensive use
of the Communications Toolbox & Blockset, and the WCDMA End-to-End Phys-
ical Layer demo.

25
26 CHAPTER 3. UMTS MOBILE POSITIONING SIMULATOR FOR VEHICLE
LOCATION

3.1 Sources of Mobile Positioning Error

All positioning techniques are exposed to various environmental limitations due to
diverse and dynamic radio propagation characteristics [38]. In cellular scenarios,
several factors can introduce errors in location estimates. Two sources of error are
common to all cellular radio systems: multipath propagation and non-line-of-sight
(NLOS) propagation. For CDMA system, interference from mobile users in other
cells would also contribute to the error because of the near-far effect [14].
NLOS propagation describes the scenario where the direct path, i.e., the
line-of-sight (LOS) between mobile station (MS) and base station (BS) is blocked
by structures such as buildings in dense urban area [14]. With NLOS propagation,
the signal arriving at the MS is reflected, as shown in Figure 3.1. The NLOS path
can be longer than the LOS path which results in erroneous timing measurement.
Multipath propagation refers to the reception of multiple signals at a MS or BS. It
appears due to the existence of various obstacles in the radio environment that are
source of reflections, diffractions, and scattering of the signal [38]. Even if there is
an existing LOS path, multipath propagation may still lead to erroneous timing
estimates. Moreover, multipath channels result in signal components spread over
time and frequency [38]. It causes interference when selected spreading codes for
separating physical channels in the downlink are not fully orthogonal [39].
All cellular systems suffer from co-channel interference. In cellular CDMA, MSs
share the same frequency band with different spreading codes [14]. Therefore, one
challenge facing UMTS location is the existence of in band interference. The wide
bandwidth of UMTS makes resolution of timing estimation higher than earlier
generations of systems. However, the high interference in WCDMA may work
against this benefit by making it difficult to detect the arriving signal [40].

Figure 3.1: Source of mobile positioning error: NLOS propagation.

3.2 RADIO NETWORK MODELING 27

3.2 Radio Network Modeling

The radio network model is the basis of the positioning simulator. In this section,
the cellular topology and the mobile generation are firstly described. Then, the
modeling of path loss, shadowing, and fading are presented.

3.2.1 Network Topology and MS Mobility

The cellular network is designed as a regular hexagonal pattern. Base stations are
placed on a hexagonal grid with distance of 1000 m; the cell radius is then equal to
577 m. Each BS, equipped with omni-directional antenna, is located in the middle
of the cell. In simulations, the network modeling is limited to a certain number of
cells with strict boundaries. BSs and MSs in the cell boundaries have different
interference than others inside the cellular network. In addition, MSs at the bound-
ary may leave the network and lose their connectivity. To avoid these two problems,
a wrap-around technique is applied in network modeling and simulation [41]. It is
demonstrated in [42] that performing system-level simulation in a 7-cell topology
may underestimate the total interference in the system whereas 19 is considered a
suitable number of cells when wrap around technique is used. A mobility pattern is
defined for updating the MS position. The inputs of the mobility model are the MS’s
initial position and velocity. The MS updates its position after every time step. The
new position would be the product of complex speed and time step added to the old
complex position. An example of MS movement in hexagonal cells with wrapped
boundary is illustrated in Figure 3.2. The MS (in a vehicle) is moving with speed of
50 km/h in a straight line through a city. The simulation area is limited to 19 cells in
the center of the city. When the MS is about to get out of the simulation boundary, it
will be wrapped around to the opposite side of the simulation area.

Figure 3.2: Mobile movement in a wrapped 19-cell network topology.

28 CHAPTER 3. UMTS MOBILE POSITIONING SIMULATOR FOR VEHICLE
LOCATION

3.2.2 Radio Propagation Model

There are numerous possible radio environments where mobile radio systems are
expected to operate. Three generic radio environments have been classified [43].
Among them, the vehicular radio environment is of interest. It is characterized by
macro cell, large transmitting power, and high mobile speed. A three-stage propaga-
tion model is used to describe the characteristics of this environment [44], as illus-
trated in Figure 3.3. The first effect is path loss, which models how the received
signal strength decreases as the MS to BS distance increases. The second effect is
shadowing (referred also as slow or log-normal fading), which models how the
transmitted signal is attenuated through absorption, reflection, scattering and dif-
fraction. The third effect is fast fading (also known as Rayleigh fading or multipath
fading), which is due to multipath reflections of a transmitted signal by obstacles.
Out of all these characteristics, path loss and shadowing are modeled for the system
level analysis, whereas fast fading is included within the link level simulation.

Figure 3.3: Radio propagation effects: path loss, shadowing and fast fading.

3.2.2.1 Path Loss

Two path loss models are implemented in the simulator to emulate two different
situations in urban macro cells. The COST231-Hata-Model calculates the path loss
in absence of environmental information while the COST231-Wallfisch-Ikegami-
Model takes into account the geometry of the buildings and streets. Both empirical
models for outdoor applications in urban areas are developed by the European
COST Action 231, based on extensive measurements in European cities [45].
According to the COST231-Hata-Model, the path loss in an urban area is:
3.2 RADIO NETWORK MODELING 29

PL (dB) = 46.3 + 33.9 log10 ( fC ) - 13.82 log10 (hBS )

(3.1)
-a(hMS ) + (44.9 - 6.55 log10 (hBS ))log10 (d ) + C M
where fC is the carrier frequency; hBS is the height of the BS antenna; hMS is the
height of the MS antenna; d is the distance between MS and BS. In addition,
a(hMS ) is a correction factor for the MS antenna height based on the size of cov-
erage area; C M is 0 dB for medium sized cities and suburbs and 3 dB for metropol-
itan areas.
COST231-Wallfisch-Ikegami-Model allows for improved path loss estimation
by considering additional characteristics of the urban environment. This model
distinguishes between LOS and NLOS situations. For the LOS case, a simple
propagation loss formula is applied:
PL (dB) = 42.6 + 26 log10 (d ) + 20 log10( fC ) (3.2)

For the NLOS case, the path loss is composed of three terms: the free space loss L0 ,
the multiple screen diffraction loss Lmsd , and the roof-top-to-street diffraction and
scatter loss Lrst :
PL (dB) = L0 + Lmst + Lrst (3.3)

where
ìï L = 32.4 + 20 log (d ) + 20 log ( f )
ïï 0 10 10 C
ïï
ïí L = Lbsh + ka + kd log10 (d ) + k f log10 ( fC ) - 9 log10 (b ) (3.4)
ïï mst
ïï
ïïî Lrst = -16.9 - 10 log10 (w ) + 10 log10 ( fC ) + 20 log10 (Dhm ) + Lori

Environmental parameters such as heights of buildings, width of roads, building

separation, and road orientation with respect to the direct radio path [45] are in-
cluded in this model.

3.2.2.2 Correlated Shadowing

Inferred from its alternative name “log-normal fading”, statistical distribution of
shadowing can be approximated by a lognormal law with a zero mean and an en-
vironment-specific standard deviation:

PSH = 10X /10 (3.5)

30 CHAPTER 3. UMTS MOBILE POSITIONING SIMULATOR FOR VEHICLE
LOCATION

where X is a zezo-mean Gaussian random variable with variance s . In addition, a

more realistic model that takes into account the autocorrelation property is proposed
in [46]. A correlated shadowing model is implemented to make the fading process
correlated between BSs. An uncorrelated process W is firstly defined with normal
distribution N( 0, s ) . As introduced in [46], a correlation factor A is defined as:

A = e-vT / D (3.6)
where v is the speed of MS, T is the sampling period, and D is the correlation
distance. This correlation is an exponentially decreasing function over distance. In
this sense, the desired correlated random process X is generated by filtering the
uncorrelated random process W using an autoregressive filter:
X (t + 1) = a ⋅ X (t ) + b ⋅ W (t ) (3.7)
where a and b are filter parameters, which can be solved by correlation as:
ì
ï -vT / D
ïa = A = e
í (3.8)
ï
ï b = s ⋅ 1 - (e -vT / D )2
ï
î

3.2.2.3 Rayleigh Fading

As discussed previously, the transmitted signal may reach the receiver over multiple
paths. Each path may thus result in different attenuations and delays of the received
signal. The mathematical model of the multipath can be presented using the method
of impulse response:
N
h(t, t ) = å ci (t )d(t - ti ) (3.9)
i =1

which is a well-known tapped-delay line model [47]. The delay values ti give
different received times of the single transmitted impulse from multiple paths. The
path amplitudes ci (t ) are time varying, which is described by Rayleigh distribution.

3.3 WCDMA Link Simulator

Based on the system-level radio network model presented in the previous section,
the communication link between a MS and a BS is modeled. Timing measurements
obtained from such BS-MS links will be then used for UMTS mobile positioning.
The BS-MS link-level simulator includes a BS transmitter, a communication chan-
nel, and a MS receiver. The block diagram of the simulator implemented in Sim-
ulink is shown in Figure 3.4.
3.3 WCDMA LINK SIMULATOR 31

Figure 3.4: WCDMA FDD downlink simulator block diagram.

3.3.1 BS Transmitter Model

The BS transmitter model includes spreading, scrambling, and filtering of the pilot
signal. The generated pilot (P-CPICH) is first spread by a channelization code, and
then scrambled by a complex scrambling code [48]. For channelization, Orthogonal
Variable Spreading Factor (OVSF) code is chosen, so that the channelization code
used in each cell is unique. For scrambling, complex Gold code is adopted, which is
composed of a pair of binary PN (Pseudo-Noise) sequences. The chip rate of PN
code is 3.84 Mchip/s, causing a bandwidth of 5 MHz. The resulted signal is Quadra-
ture Phase-Shift Keying (QPSK) modulated. This QPSK signal is then filtered with
a root raised cosine shaping filter with a roll-off factor of 0.22 [48].

3.3.2 Multipath Channel Model

The Rayleigh multipath and Additive White Gaussian Noise (AWGN) channel
model emulates the effects of the multipath propagation of a Rayleigh fading dis-
cussed in Section 3.2. Here, the AWGN represents the co-channel interference. In a
Rayleigh fading model, the number of paths, power attenuation and delay time are
specified for a particular scenario. In this work, channel models from [47] are
adopted for simulation.
32 CHAPTER 3. UMTS MOBILE POSITIONING SIMULATOR FOR VEHICLE
LOCATION

The Doppler frequency is chosen as 92.6 Hz, corresponding to a mobile speed of

50 km/h. The delay vector and the gain vector are the values specified by the typical
urban channel model (Tux) in 3GPP TR 25.943 [47]. The channel models that are
standardized in [47] are chosen as simplifications of the COST 259 model [49].
When applying the Tux model in the simulator, the model can be further simplified
for the UMTS WCDMA application to allow for more efficient and less complex
simulation. The simplification is done with a specific time resolution D T . For
WCDMA FDD, D T is 1/2 of the chip time which gives 130.2 ns. The simplified
WCDMA FDD model is then obtained by sampling the original channel profiles at
the following delays: { 0, D T, 2D T, 3D T ¼} .

3.3.3 MS Receiver Model

The MS receiver model includes filtering, dispreading, and correlation for detection
of the pilot signal. The main challenge at the receiver side is to detect the received
signal that is attenuated and delayed. Here, a Rake receiver is implemented to
compensate for the multipath effect [39]. The Rake receiver consists of four rake
fingers, each associated with a different received component. Code generators and
correlator perform despreading and channel estimation. The delay is compensated
for the difference in the arrival times of the symbols in each finger. Finally, the
Rake combiner sums the output of different rake fingers to recover the energy
across different delays.

3.4 UMTS Mobile Positioning

Location finding in a WCDMA network is challenging since the network is asyn-
chronous. In this sense, special measurements and techniques should be employed
to locate a MS. In UMTS, MS performs signal arrival-time measurements of the
pilot signal (P-CPICH) from different BSs [50]. As introduced in Chapter 2, for
OTDOA method, signals from at least three BSs should be received by MS for
location calculation. The MS position is then determined by solving a nonlinear
system of hyperbolic equations.

3.4.1 OTDOA Measurement

Defined by 3GPP standard, the physical channel accessed for positioning is the
Common Pilot Channel (CPICH), which is in the downlink [48]. It is used for
observing the SFN-SFN (System Frame Number) observed time difference between
neighboring BSs.
3.4 UMTS MOBILE POSITIONING 33

In such an asynchronous system, BSs transmit pilot signals at different instants.

The OTDOA measurements suffer from time offset which can be compensated by
the relative timing difference (RTD). The RTD is obtained by computing the
SFN-SFN observed time difference, which measures the time difference between
the clocks of the BSs [50]. In addition, the MS also measures the SFN-SFN ob-
served time difference. This time difference observed at MS is called observed time
difference (OTD). The OTD is obtained by computing the time difference between
TCPICH and TCPICH . TCPICH is the time when MS receives one Primary
Rxj Rxi Rxj

CPICH (P-CPICH) slot from cell j , and TCPICH is the time when MS receives
Rxi

the P-CPICH slot from cell i [50].

3.4.2 Cross-correlation Method

The detection of TDOA of received pilots from a pair of base stations is based on
Generalized Cross-Correlation (GCC) method [51]. The cross-correlation function
of received pilot rCPICH and rCPICH is given by:
Rx 1 Rx 2

1 T
C 1,2(t ) = ò0 rCPICH (t )rCPICH (t )dt (3.10)
T Rx 1 Rx 2

The TDOA estimate is the value t̂ that maximizes C 1,2(t ) . This GCC method is
implemented using the Find Delay block provided in Simulink.

3.4.3 MS Location Calculation

Given the known coordinates of the i th BS as (x i , yi ) , the known time offset be-
tween two BSs as RTDij , and the observed time difference through correlation as
OTDij , we can estimate the coordinates of the MS that are denoted as (ˆ,
x yˆ) . In
absence of measurement errors, the hyperbolic difference equation for MS location
can be written as:

c(OTDij - RTDij ) = (xˆ - x i )2 + (yˆ - yi )2 - (xˆ - x j )2 + (yˆ - y j )2 (3.11)

where c denotes the speed of light. Thus, we have a set of hyperbolic equations for
three pairs of BSs. This equation set is solved by the classic Chan’s algorithm [52].
The detail deviation and solution to these nonlinear equations can be found later in
Chapter 4 (specifically in Section 4.2.1).
34 CHAPTER 3. UMTS MOBILE POSITIONING SIMULATOR FOR VEHICLE
LOCATION

3.5 Summary
This chapter has presented the models and methods used to develop the UMTS
mobile positioning simulator, which responses to the first investigation aspect
suggested by the end of Chapter 2. The simulation tool is composed of three parts
specifically, the radio network, the WCDMA downlink and the mobile positioning.
This simulator, which is developed in accordance with the 3GPP standards for
OTDOA positioning, can be used to supplement the real field tests, and will be later
(partially) utilized to study the statistics of error in OTDOA location estimation.
CHAPTER 4

MOBILE PHONE‐BASED HYBRID VEHICLE

POSITIONING AND TRACKING

The ability to locate the position of a mobile phone has emerged as a key facility of
existing and future generation mobile systems. Many value added location-based
services have been enabled by this feature. For instance, vehicle positioning and
tracking by locating mobile phones traveling on-board vehicles has become feasi-
ble. However, no single standard positioning method can provide decent trade-off
between accuracy and coverage. Hence, there is a challenge on tracking the position
and velocity of the mobile phone along the journey with appropriate accuracy. To
address this issue, in this chapter, we implement one of the hybrid mobile
phone-based vehicle location solutions proposed at the end of Chapter 2: a hybrid
scheme that combines location estimates from both A-GPS and OTDOA position-
ing methods in UMTS. This chapter starts with an illustration of the proposed
hybrid scheme. It is then followed by detailed descriptions of the position estima-
tion, the trajectory tracking, and the data fusion algorithms. After that, the simula-
tion setups are presented and finally the numerical results are provided.

35
36 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

4.1 The Hybrid Positioning and Tracking Scheme

As introduced in Chapter 2, mobile positioning in wireless networks can be widely
divided into two categories: network-based and handset-based methods. Location
estimates from two handset-based methods (A-GPS and OTDOA) are combined
mathematically to provide enhanced results. In this section, the proposed hybrid
positioning and tracking scheme is firstly illustrated. Then, details about two types
of measurements are further described.

4.1.1 The Hybrid Scheme

The hybrid data fusion scheme is inspired by the classical target tracking problem
using noisy measurements from a variety of sensors [53]. Generally, there are two
approaches for the multi-sensor data fusion: the measurement fusion and the
state-vector fusion. The measurement level fusion is less attractive in real-time
applications since it involves large volume of data transmission. On the contrary,
the state-vector level fusion is preferable in terms of reduced overloads [54]. Each
sensor uses an estimator, which provides an estimate of the state vector and its
associated covariance matrices; then only these state estimates are transmitted to the
fusion center where the fused state vector is obtained. Due to the real-time nature of
this vehicle positioning and tracking application, the concept of state-vector level
fusion is applied. Figure 4.1 shows the diagram of the proposed hybrid positioning
and tracking scheme.

(m p , e p ) ( m t , nt )

X AGPS R1 X OTDOA R2

X̂ AGPS P1 X̂ OTDOA P2

X̂ DF

Figure 4.1: The hybrid positioning and tracking scheme.

4.1 THE HYBRID POSITIONING AND TRACKING SCHEME 37

4.1.1.1 A-GPS and OTDOA Measurements

The A-GPS mobile phone provides two-dimensional (2D) geographical coordinates
m p with horizontal accuracy ep . Timing measurements mt are obtained from the
parameters in WCDMA network, while OTDOA noise nt is determined by many
factors (detailed description can be found later in Section 4.1.2.2).

4.1.1.2 Location Estimates and Noise Variance

The geographical coordinates m p = [ plat plon ]T are translated to the Cartesian
coordinates X AGPS = [xlat xlon ]T . The given accuracy ep is related to the vari-
ance R1 of measurement noise. Hyperbolic equations are solved, which result in
location estimates XOTDOA = [xOTDOA yOTDOA ]T . Variance R2 contains the
uncertainty information in this estimation step.

4.1.1.3 State Estimates after KF

In order to track the mobile, Kalman filter (KF) is used for each set of measurements
separately. After KF, state estimates and their associated error covariance are then
generated: XˆAGPS = [xˆAGPS yˆAGPS xˆAGPS yˆAGPS ]T and P1 for the A-GPS;

XˆOTDOA = [xˆOTDOA yˆOTDOA xˆOTDOA yˆOTDOA ]T and P2 for the OTDOA.

4.1.1.4 Position and Velocity Estimates after Fusion

State-vector fusion is performed by fusing the filtered state estimates X̂AGPS and
X̂OTDOA into a new state estimate XˆDF = [xˆDF yˆDF xˆDF yˆDF ]T . The fusion
results are weighted sums of the two state estimates; while the weight factor is
expressed by the two error covariance P1 and P2 .

4.1.2 A-GPS and OTDOA Measurements

4.1.2.1 A-GPS
For handset-based A-GPS location method, the mobile is the primary position
calculating entity. Thus, the mobile locations are sent to the central server to update
the traffic system with real-time position information. The Java Specification Re-
quest 179 (JSR179) Location Application Programming Interfaces (API) [55] is
38 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

utilized to periodically request location updates from the A-GPS mobile. JSR179 is
the only standardized J2METM “location-related” API that works across different
mobile phones. In this work, we conducted several field tests on Sony Ericsson and
Nokia A-GPS mobile phones. The location updates include latitude and longitude
coordinates, their accuracy, and the timestamp. Horizontal accuracy is the root
mean square (RMS) of easting error (in meters, 1-sigma standard deviation) and
northing error (in meters, 1-sigma standard deviation). Assuming a Gaussian dis-
tribution, this implies that the actual location is within the circle defined by the
returned point and radius at a probability of about 68%.

4.1.2.2 OTDOA
For OTDOA method, signals from at least three BSs should be received for location
calculation. Time differences are obtained from physical layer measurements which
have been defined by 3GPP standard [50]. As illustrated in Section 3.4, the OTDOA
measurements suffer from time offset, which can be compensated by the RTD. The
RTD is observed by the measurement element “SFN-SFN observed time difference
(type 2)”. In addition, the element “SFN-SFN observed time difference (type 1)”
measures the timing difference between two cells, which is called the OTD. In
absence of errors, actual time difference between the mobile and one pair of BSs
can be calculated as OTD - RTD . In reality, measurement errors are present and
can be modeled as zero-mean Gaussian random variables. The specified accuracy
for OTD and RTD are 1 chip and 0.5 chip respectively [56]. In UMTS, the chip rate
is 3.84 Mchips/s, which corresponds to 260.42 ns of chip period DT . In conse-
quence, OTDOA range uncertainty is limited to c ´ DT= 78.126 m .

4.2 The Positioning and Tracking Algorithms

This section presents the mathematical description of the hyperbolic positioning,
the mobile trajectory tracking, and the state-vector fusion which are three major
steps in the hybrid scheme.

4.2.1 Hyperbolic Solution based on OTDOA

Given known coordinates of BSi ( xi ,yi ) and BS j ( x j ,y j ) , the measured time
difference mt = OTDi, j - RTDi, j , and the modeled timing measurement noise
nt , we can estimate the mobile position XOTDOA = [xOTDOA yOTDOA ]T . The
hyperbolic difference equation for mobile location can be written as:
4.2 THE POSITIONING AND TRACKING ALGORITHMS 39

Ri , j = c ´ (mt + nt ) = (xOTDOA - x i )2 + (yOTDOA - yi )2

(4.1)
- (xOTDOA - x j )2 + (yOTDOA - y j )2

where c is the signal speed (ca. 3 ´ 108 m / s ) and Ri, j denotes the distance
difference from the mobile to BS i and to BS j . Consequently, mobile location can
be estimated by solving a set of nonlinear hyperbolic equations. Many methods
have been proposed to solve these nonlinear equations. Compared to iterative
methods that use the linearization techniques such as the Taylor-series expansion,
closed-form methods are faster and are developed for real-time implementation
[57]. Among them, the Chan’s algorithm, proposed in [52], is an approximation of
the maximum likelihood (ML) estimator, which consists of two-step least squares
(LS). The Chan’s algorithm is employed to locate the mobile using OTDOA
measurements.

Suppose that BS j is the serving base station, which is considered as the refer-
ence point; BSi (i = 2, 3, ¼,N ) are neighboring sites that are involved in position
calculation. In this sense, (4.1) can be transformed into the following form:

R2i, j + 2Ri, j Rj = (xi 2 + yi 2 ) - 2(xi - x j )xOTDOA

(4.2)
-2(yi - y j )yOTDOA - (x j 2 + y j 2 )

where Rj = (xOTDOA - x j )2 + (yOTDOA - y j )2 is the distance between the

mobile and the BS j . It is a set of linear equations with unknown vector
za = [xOTDOA yOTDOA Ri ]T to be solved. If we take into account the OTDOA
measurement noise, the OTDOA covariance matrix can be written as:

Q = diag { s2, j 2, s3, j 2 ,..., sN , j 2 } (4.3)

where s N , j 2 are variances of OTDOA timing errors. With this noise, the error
vector derived from (4.2) can be expressed as:

y = h - Gaza0 (4.4)

where
40 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

é R 2 - (x 2 + y 2 ) + (x 2 + y 2 ) ù
ê 2, j 2 2 j j ú
1 ê R3,2 j - (x 32 + y 32 ) + (x j 2 + y j 2 ) ú
h = ê ú
2ê  ú
ê R 2 - (x 2 + y 2 ) + (x 2 + y 2 ) ú
êë N , j N N j j úû (4.5)
é x 2 - x j y 2 - y j R2, j ù
ê ú
êx -x y 3 - y j R3, j ú
Ga = - ê 3  j   úú
ê
ê x N - x j yN - y j RN , j ú
ë û
and y is assumed to be normally distributed. The first-step LS solution of (4.4)
gives a ML estimate of za as:

za = (GaT Y-1Ga )-1GaT Y-1h (4.6)

where Y is assumed to be normally distributed and the covariance matrix of Y can

be written as:
é R2 ù é R2 ù
ê ú ê ú
ê R ú ê R ú
Y = E éê yy ùú = c êê
T 2 3 úQ ê
ú ê
3 ú
ú (4.7)
ë û  
ê ú ê ú
ê ú
RN ú êê RN úú
êë û ë û
The solution given by (4.6) assumes that xOTDOA , yOTDOA and Ri are inde-
pendent. However, in reality they are related. Hence, a second-step LS is applied
which gives the following results:

ìé
ï ùT é1 0ùï ü-1 é ùT éz2 ù
ï
ïï ê 1 0 ú ê ú ïïï ê 1 0 ú ê a ,1 ú
ï ê ú -1 ê ú ï ê ú -1 ê 2 ú
za ' = í ê 0 1 ú ( Y ') ê 0 1 ú ý ê 0 1 ú ( Y ') ê za ,2 ú (4.8)
ï
ï ê1 1ú ê1 1 úï ï ê1 1 ú ê 2 ú
ï
ï ê
ë ú
û êë úû ï ï êë úû ê za ,3 ú
îï ïþ ë û
where
éz ù éz ù
ê a,1 ú ê a,1 ú
Y ' = E éê yy ùú = 4 êê
T
za,2 ú cov(z ) ê
ú ê z ú
ú (4.9)
ë û a a ,2
ê ú
za,3 ú ê za,3 úú
êë û ê
ë û

The resulting za' = [xOTDOA yOTDOA Ri ]T gives the mobile position estimation.
4.2 THE POSITIONING AND TRACKING ALGORITHMS 41

4.2.2 Mobility and Measurement Models

This subsection presents the mathematical description of the vehicle motion dynam-
ics in terms of transition equations, and measurement models that relate location
observations to the dynamic states. It is modeled as a sample-data system [58]. Its
dynamics is described by a continuous-time equation, but measurements are ob-
tained at discrete-time.

4.2.2.1 Motion Dynamics

For computational simplicity, the moving vehicle/mobile is modeled as a dynamic
linear system driven by a random acceleration. As a result, the transition equation is
firstly derived for continuous-time movement, and then after discretization is ex-
pressed in discrete-time. Suppose that a vehicle, equipped with a mobile, moves in a
2D Cartesian coordinate system. In order to track this moving mobile, its state can
be expressed as a dynamic state vector:
T
X (t ) = éê x (t ) y(t ) x(t ) y(t ) ùú (4.10)
ë û
where x (t ) and y(t ) are positions; their first derivatives x(t ) and y(t ) are veloc-
ities. In this way, the motion dynamics can be described by a continuous white noise
acceleration model [59]. In this model, the velocity undergoes perturbations which
are modeled by zero-mean Gaussian white acceleration v(t ) :
v(t ) = x(t ) = y(t ) (4.11)
with the variance

E éë v(t )v(t ) ùû = qc d(t - t ) (4.12)

where qc is the power spectrum density. The continuous-time transition equation is

then written as:
X (t ) = AX (t ) + V (t ) (4.13)
where
é0 0 1 0ù é 0 ù
ê ú ê ú
ê0 0 0 1ú ê 0 ú
A= ê ú V (t ) = ê v(t ) ú (4.14)
ê0 0 0 0ú ê ú
ê0 0 0 0ú ê v(t ) ú
êë úû êë úû
Let the sampling period in this sample-data system to be DT . After discretization,
the discrete-time transition equation is:
42 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

X (k + 1) = FX (k ) + V (k ) (4.15)
with the transition matrix
é1 0 DT 0 ù
ê ú
ê0 1 0 DT úú
F = e ADT = êê (4.16)
ê0 0 1 0 úú
ê0 0 0 1 úú
êë û
T
and the process noise vector V (k ) = éê 0 0 v ( k ) v ( k ) ùú , which models dis-
ë û
turbance in driving velocity; its covariance matrix is:
é1 1 ù
ê DT 3 0 DT 2 0 ú
ê3 2 ú
ê 1 1 ú
ê 0 DT 3 0 DT 2 úú
é Tù ê 3 2
Q = E êV (k )V (k ) ú = ê ú qc (4.17)
ë û ê1 ú
ê DT 2 0 DT 0 ú
ê2 ú
ê 1 ú
ê 0 DT 2 0 DT úú
êë 2 û

4.2.2.2 Measurement Models

T
The discretized version of the state vector X (k ) = éê x ( k ) y ( k ) x ( k ) y ( k ) ùú
ë û
is related to the noisy location observations Y (k ) by the measurement equation:

Y (k ) = HX (k ) + W (k ) (4.18)

é1 0 0 0ù
with the measurement matrix H = ê 0 1 0 0 ú , which takes only the position
êë úû
observations, and the measurement noise vector W (k ) . W (k ) can be modeled as:
é w(k ) ù æ é 0 ù é R 0 ù ö÷
W (k ) = ê ú ~ N çç ê ú , ê ú ÷ , where R = sx 2 = sx 2 is the measurement
êë w(k ) úû èç êë 0 úû êë 0 R úû ÷÷ø
error variance. It is assumed that the variances in both x and y directions are the
same and independent.
4.2 THE POSITIONING AND TRACKING ALGORITHMS 43

4.2.3 Kalman Filter for Trajectory Tracking

In order to track the vehicle/mobile in real-time, a discrete-time Kalman filter (KF)
is applied for each set of noisy location measurements. The KF gives a recursive
solution for the state estimation of a linear system, which is described by (4.15) and
(4.18). The transition and measurement equations can be rewritten in a more com-
pact way as:
ì Xk = FXk -1 + Vk -1
ï
ï
í (4.19)
ï
ïV ~ N (0,Q )
î k -1
ìYk = HXk + Wk
ï
ï
í (4.20)
ï
ïW ~ N (0, R)
î k
where Q and R are covariance matrices for the process error vector Vk -1 and the
measurement error vector Wk , respectively. Then, the optimal estimations (in terms
of minimizing variances) are obtained by the following iterated steps (the deviation
can be found in many state estimation texts, e.g. in [58], [59]):
 Before the measurements are available at tk , priori estimates of the state
mean X̂k- and covariance Pk- are obtained by the time update equations:

Xˆk- = FXˆk -1 (4.21)

Pk- = FXˆk -1FT + Q (4.22)

 Kalman gain K k is then computed to set an appropriate correction term for

the next propagation step, so as to minimize the mean square estimation
(MSE) error:

Kk = Pk-H T (HPk-H T + R)-1 (4.23)

 After processing the noisy measurements Yk , posteriori state mean estimate

X̂ k is obtained by updating X̂k- with a corrected version of measurement
residual:

Xˆk = Xˆk- + Kk (Yk - HXˆk-) (4.24)

Then the covariance matrix Pk associated with X̂ k can be updated from

the priori estimate as:

Pk = (I - Kk H )Pk- (4.25)
44 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

Provided that matrices F , Q , H , R are defined beforehand, the calculation of

Pk- , K k , and Pk will be independent of any measurements. In other words (4.22),
(4.23), and (4.25) can be solved off-line before the tracking starts. This feature is
favored in such a real-time tracking application.

4.2.4 State-Vector Fusion

Following the discussion in Section 4.1, the state-vector fusion, which was targeting
on two-sensor fused-track, is applied in this work. Firstly presented in [53], the
fusion method combines two filtered state vectors xki and xkj into a new state vector.
The fused estimate x̂k and its covariance matrix Pk are given by:

xˆk = xki + (Pki - Pkij )(Pki + Pkj - Pkij - Pkji )-1(xki - xkj ) (4.26)

Pk = Pki - (Pki - Pkij )(Pki + Pkj - Pkij - Pkji )-1(Pki - Pkji ) (4.27)

where Pki and Pkj are covariance matrices for xki and xkj ; Pkij and Pkji are cross
covariance matrices between the two estimates.
In this work, the location estimation errors from OTDOA and A-GPS are as-
sumed to be independent, i.e., the cross covariance items should be eliminated.
Thus, the fused state estimate X̂kDF and covariance matrix PkDF are expressed as:

XˆkDF = XˆkOTDOA + PkOTDOA(PkOTDOA + PkAGPS )-1(XˆkOTDOA - XˆkAGPS ) (4.28)

PkDF = PkOTDOA - PkOTDOA(PkOTDOA + PkAGPS )-1 PkOTDOA (4.29)

where X̂kOTDOA and X̂kAGPS are the estimation results of the two Kalman filters
applied on the OTDOA and the A-GPS measurements; PkOTDOA and PkAGPS are
the associated covariance matrices.

4.3 Simulation Results and Evaluation

To evaluate the feasibility of the proposed hybrid vehicle positioning and tracking
scheme, simulations of typical vehicle motion in urban area are performed. The
simulation first generates a random vehicle trajectory using the dynamic model
defined by (4.13). This mobility model is practical since observations of real vehicle
speeds by traffic engineers show that the distribution of car speed can be modeled as
Gaussian process [60]. According to the typical vehicle motion in urban scenario,
average speed of 50 km/h (i.e., 15 m/s) and mean acceleration of 1 m/s2 are applied.
The acceleration magnitude is Rayleigh distributed with variance s 2 and expecta-
4.3 SIMULATION RESULTS AND EVALUATION 45

tion s p / 2 . Hence, qc in (4.12) can be calculated as 12 ´ (2 / p) . Noisy loca-

tion observations from both OTDOA and A-GPS are then generated along this
simulated trajectory.
The OTDOA location estimation noise is assumed as unbiased Gaussian with 2D
standard deviation s2D . In order to obtain statistics of the location error, Monte
Carlo trials are performed for 10000 runs. In each trial, the true mobile position is
generated uniformly in a hexagonal cell, which is surrounded by six neighboring
cells with radius of 1 km, as depicted in Figure 4.2 (a). The figure of merit used to
evaluate the location estimation is the Root Mean Square Error (RMSE):

1 MC OTDOA T
RMSE = å ( Xi - Xitrue ) ( XiOTDOA - Xitrue ) (4.30)
MC i =1
where MC is the total number of performed Monte Carlo trials; and
XiOTDOA = [xOTDOA yOTDOA ]T is the position estimate in the i th Monte Carlo
run. Different numbers of BSs (range from 4 to 7) are involved in the position
calculation, as shown in Figure 4.2 (b). RMSE of OTDOA location estimation is
obtained with range noises of: 39.063 m, 78.126 m, and 117.189 m, which corre-
spond to 0.5, 1, and 1.5 chip periods in UMTS. As it is marked in the figure, a
RMSE of 82.14 m (where 4 BSs are hearable, and range noise is 78.126 m) is
selected due to two facts: 1) measurements in [61] demonstrate that there is a 90%
probability of hearing more than 3 BSs in urban area, and 2) previous discussion
shows that the OTDOA timing measurement accuracy is limited to 1 chip period.

2000
3

1000
5 4

0
1

-1000 6 2

-2000 7

-2000 -1000 0 1000 2000

Figure 4.2: OTDOA location estimation: (a) Network topology and mobile
distribution, and (b) RMSE of the OTDOA location estimation.
46 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

The position estimation error of A-GPS can be directly obtained from the field
test, as described in Section 4.1.2. Figure 4.3 shows the A-GPS samples collected
during a 15-min trip in urban area of Stockholm. Totally 300 valid location updates
were measured with a sampling period of 3 s. The provided horizontal accuracy of
each location sample is the RMS composed from the standard deviations s N and
s E of the northing and easting axes respectively. Assuming the circular case
( sN = sE ) , A-GPS location error is generated as Gaussian noise with the variance
(RMSkAGPS 2)2 , where RMSkAGPS denotes the recorded horizontal accuracy at
tk . Figure 4.4 shows a sample mobile trajectory, along with the OTDOA and
A-GPS location measurements generated by the described models. As shown in the
zoomed figure, the location accuracy of OTDOA is lower than A-GPS.
A good selection of initial conditions for the filters is essential for the tracking
algorithm to operate well [60]. It is important that the initial state Xˆ0 together with
its covariance P0 reflects realistically its accuracy. In this simulation, the KF is
initialized as follows. The position estimates are initialized by the first location
measurements; the velocity estimates set to zero due to the absence of position
information at t0 . Initial variance of OTDOA position estimate is given by the
Cramér-Rao lower bound (CRLB) which is derived in [52]. Variance of A-GPS
position estimate is initiated by the statistical mean of the given accuracy in the field
test data. The initial variance of the velocity estimates is decided based on the fact
that the initial velocity is uniformly distributed between ±15 m / s .

Figure 4.3: Vehicle trajectory samples collected by the A-GPS mobile phone.
4.3 SIMULATION RESULTS AND EVALUATION 47

6000
Real trajectory
OTDOA estimations
5000 AGPS mesurements
northing in meters

4000

3000

2000

1000

0
0 500 1000 1500 2000 2500
easting in meters

Figure 4.4: Real mobile track vs. the generated OTDOA and A-GPS locations.
A sample trajectory of a mobile after the filtering and fusion processes is shown
in Figure 4.5, along with the estimated trajectories, i.e., the position estimation
results. The trajectory of the hybrid method (labeled as DF) demonstrates a better
match to the real track. Figure 4.6 illustrates the real and estimated velocities over
filter steps in both easting and northing directions. Although being processed at
each filter step tk , the position and velocity estimation results are illustrated con-
tinuously, i.e., the values are connected by solid lines.
The performance metrics used for comparing the estimation accuracies of the
three methods (specifically, OTDOA, A-GPS, and Hybrid) are defined as:
ìï 1 L é
ïï RMSE
ï position = å êë (xk - xˆk )2 + (yk - yˆk )2 ùúû
ï
ï L k =1
í (4.31)
ï
ï 1 L
ï
ï
ï
RMSEvelocity = å éêë (xk - xˆk )2 + (yk - yˆk )2 ùúû
ï
î L k =1
T
where L is the total number of filter steps; Xˆk = éê xˆk yˆk xˆk yˆk ùú denotes the
ë û
state estimates at filter step k .
Table 4.1 shows the RMSE sample mean (mRMSE ) and standard deviation
(sRMSE ) of position and velocity estimations calculated using 1000 independently
generated sample trajectories. The cumulative distribution function (CDF) of the
position and velocity estimation errors for the OTDOA, the AGPS and the Hybrid
(DF) methods, are plotted in Figure 4.7 respectively.
48 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

Position estimation(OTDOA)
6000
northing (m)

4000

2000 Real trajectory

KF Tracking(OTDOA)
0
0 500 1000 1500 2000 2500
easting (m)
Position estimation(AGPS)
6000
northing (m)

4000

2000 Real trajectory

KF Tracking(AGPS)
0
0 500 1000 1500 2000 2500
easting (m)
Position estimation(DF)
6000
northing (m)

4000

2000 Real trajectory

KF Tracking(DF)
0
0 500 1000 1500 2000 2500
easting (m)

Figure 4.5: KF tracking of a sample trajectory based on the OTDOA, the AGPS and
the Hybrid (DF) methods.
4.3 SIMULATION RESULTS AND EVALUATION 49

Easting Velocity Estimation (OTDOA) Northing Velocity Estimation (OTDOA)

20 20
Actual velocity
Estimated velocity
velocity in m/s

velocity in m/s
10
0
0
-20
-10 Actual velocity
Estimated velocity
-20 -40
0 100 200 300 0 100 200 300
filter steps filter steps
Easting Velocity Estimation (AGPS) Northing Velocity Estimation (AGPS)
20 20
Actual velocity
Estimated velocity
velocity in m/s
10
velocity in m/s

0
0
-20
-10 Actual velocity
Estimated velocity
-20 -40
0 100 200 300 0 100 200 300
filter steps filter steps
Easting Velocity Estimation (DF) Northing Velocity Estimation (DF)
20 20
Actual velocity
Estimated velocity
velocity in m/s

velocity in m/s

10
0
0
-20
-10 Actual velocity
Estimated velocity
-20 -40
0 100 200 300 0 100 200 300
filter steps filter steps

Figure 4.6: KF tracking using the OTDOA, the AGPS and the Hybrid (DF) methods:
the actual and estimated velocities (m/s).
50 CHAPTER 4. MOBILE PHONE-BASED HYBRID VEHICLE POSITIONING
AND TRACKING

CDF of Position Estimation RMSE CDF of Velocity Estimation RMSE

100 100
DF
OTDOA
80 80
AGPS

60 60
CDF [%]

CDF [%]
40 40

20 20 DF
OTDOA
AGPS
0 0
0 20 40 60 3 4 5 6
RMSE [m] RMSE [m/s]

Figure 4.7: CDF of estimation RMSE for the OTDOA, AGPS and DF methods.

Table 4.1: Evaluation of Position and Velocity Estimations.

RMSE statistics of Position and Velocity Trackinga

Total number of
generated sample OTDOA A-GPS Hybrid (DF)
trajectories: n=1000
mRMSE sRMSE mRMSE sRMSE mRMSE sRMSE

Position RMSE (m) 45.54 2.65 25.19 1.80 23.71 1.48

Velocity RMSE (m/s) 4.66 0.26 3.97 0.18 3.94 0.17

a. Results of each sample trajectory correspond to 300 filter steps (15 minutes).

4.4 Summary
A hybrid mobile phone-based vehicle positioning and tracking solution has been
implemented and simulated, in response to the last two investigation aspects sug-
gested at the end of Chapter 2. In this hybrid solution, the Kalman filtering tech-
nique is exploited to track the vehicle and fuse the noisy OTDOA and A-GPS
location measurements at the state-vector level. Moreover, statistical evaluations
(by the simulation and the small-scale field tests) are performed to obtain reasona-
ble noise models for both location measurements. Numerical results demonstrate
that the hybrid scheme can provide better position and velocity estimations than
each individual location method. The dynamic model of vehicle motion and the
KF-based location and velocity tracking algorithm will be reused in the following
chapter for real-time urban traffic state estimation.
CHAPTER 5

TRAFFIC STATE ESTIMATION WITH A‐GPS

MOBILE PHONES AS PROBES

Real-time traffic information is essential for supporting the development of many

ITS applications: incident detection, vehicle navigation, traffic signal control,
traffic monitoring, etc. For instance, in the smart traffic information system (intro-
duced in Chapter 1), the up-to-date information on traffic condition is the major
feedback sent to the subscribers. In this chapter, we show how location data col-
lected by A-GPS mobile phones can be used to estimate urban traffic states. The
urban road traffic is generated by microscopic simulation, while small-scale field
tests are used to emulate the A-GPS mobile location measurements. This chapter
begins with the background and motivation for the traffic simulation-based urban
arterials traffic state estimation using A-GPS mobile probes. It is followed by a brief
overview of the simulation-based framework. Then, three data processing steps are
presented: emulation of A-GPS location measurements, filtering of individual
probe data, and estimation of average road link speeds. Simulation setups, results of
every steps, and performance evaluations are given in the end.

51
52 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

5.1 Background
The current state-of-the-practice traffic data collection relies on road-side sensors,
e.g., inductive loop detectors (ILDs), to gather information about traffic flow at
fixed points on the road network [62-64]. Although fixed sensors are a proven
technology, they are not deployed at wide scale mostly because of the high cost.
Moreover, with fixed sensors, it is only possible to measure the spot speed, which is
one inherent deficiency in comprehensive reflection of speed over the entire road
link. Additionally, this type of model is link and detector location specific, which
requires careful calibration [65].
An alternative to these road-side infrastructures is to employ dedicated vehicles
as floating traffic probes [66-68]. These dedicated vehicle probes (PVs) are typi-
cally equipped with a GPS receiver and a dedicated communication link. A large
number of vehicles should be so equipped to have an adequate probe size. Insuffi-
cient number of probes limits the ability of generating information for large num-
bers of streets and the accuracy of the results [69]. Given the trend that
GPS-equipped vehicles are expected to increase in the future, the capacity and cost
of dedicated communication links between in-vehicle equipment and traffic man-
agement center will still limit the sample size of PVs [65]. Moreover, since PVs are
chosen from a particular category of vehicles, e.g., taxis or buses, the traffic in-
formation could be biased and not representative for the whole population [3].
Recently, with the advance of the mobile communication technology, mobile
phones are increasingly utilized for traffic data collection. Economically, this ap-
proach avoids installation and maintenance costs, in vehicles or along roads. In
addition, using mobile phones as traffic probes overcomes the coverage limitation
in road-side sensors and insufficient probes in dedicated PVs. By the end of 2010,
there had been 5.3 billion mobile subscriptions [70], which is equivalent to 77% of
the world population. Ideally, any mobile phone that is switched on, even if not in
use, can act as a probe. Thus, potentially there is a large sample size available to this
type of probe system. In addition, industry statistics indicated that sales of
smartphones (most of which are equipped with A-GPS chipsets) showed strong
growth worldwide in 2010 [71]: total shipments in 2010 were 292.9 million units,
which had been increased by 67.6% from 2009; this makes smartphones 21.5% of
all handsets shipped.
In the past decade, field trials have been conducted to show the feasibility of
using mobile phones as traffic probes [3], [72-75]. Nevertheless, most of these trials
aimed to estimate traffic states on freeways and only few deployments attempted to
monitor urban arterial roads. It has been suggested in [3] that future research efforts
should be focused on obtaining traffic data for arterials where no data is currently
5.2 SYSTEM OVERVIEW 53

available rather than obtaining data from freeways where fixed traffic sensors are
already deployed. However, traffic estimation on arterials is more challenging than
on freeways due to the following facts [67], [73], [76]: 1) arterials have lower traffic
volume, 2) arterials have more variability in speeds, and 3) arterials are controlled
by traffic signals at intersections.
Among the previous deployments, the majority of them employed network-based
probe methods that make use of network signaling information, e.g., the handoff
information or the time/angle (difference) of arrivals. Only very few of them were
handset-based (using GPS-enabled phones), for example, a pioneer field trial held
by Globis Data in 2004 [77] and the Mobile Century field experiment conducted in
2008 [78]. Evaluation results from field trials indicated that the network-based
probe systems cannot provide sufficiently accurate traffic data for arterials. Since
arterials tend to introduce additional complexities, the more accurate handset-based
A-GPS mobile probe is expected to be a better solution for urban arterial roads;
however, this has not yet been verified (both [77] and [78] provided only successful
traffic estimations on highways). Two practical issues were identified as obstacles
to the success of A-GPS mobile phones as traffic probes [79]: 1) additional
communications costs and 2) slow uptake of GPS-enabled phones. These two issues
are no longer problems under the current circumstances: 1) modern mobile
communication networks have wide bandwidths, and 2) A-GPS mobile phones are
increasingly available in the global market.
When evaluating the results from field tests, there is hardly any available ground
traffic data to compare with, especially for arterial roads. Additionally, the field test
data is not suitable for statistical analysis due to variations in different tests and
limited number of observations. In simulation-based studies, on the other hand,
individual vehicle tracks and aggregated traffic states can be extracted as "ground
truth" [74], [80]. In addition, traffic simulations can generate traffic data under a
variety of traffic conditions, featured by different volumes and road networks [81].
Although these simulation studies do not replicate the actual conditions precisely,
they may still provide valuable indication of the potential performance of a
probe-based traffic information system.

5.2 System Overview

5.2.1 Simulation-based Framework
A simulation-based framework is developed to emulate the A-GPS mobile
phone-based urban traffic estimation, as shown in Figure 5.1. The framework con-
sists of three parts: 1) microscopic traffic simulation, 2) location data processing
and speed aggregation, and 3) performance evaluation and results presentation.
54 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

The microscopic traffic simulation is used to simulate the urban road networks
and traffic conditions. It generates “actual” location tracks for each vehicle/mobile,
and prepares “ground truth” traffic states for evaluation. The generated locations are
firstly pre-processed by degradation steps to emulate “realistic” measurements that
A-GPS mobile phones would provide in real-world situations. This pre-processing
defines the percentage of vehicles that are equipped with A-GPS mobile phones
(according to a pre-defined penetration rate), and introduces statistical errors into
the location updates (according to several field tests). These “realistic” location data
are then post-processed by a 2-step filtering process. The Kalman Filter (KF) is
implemented to track each vehicle/mobile and the simple data screening is em-
ployed to filter out undesired position and speed estimates. The next steps are
allocating individual speed estimates to road links, and aggregating them at a
pre-defined time interval. For performance assessment, the accuracy of average
speed estimation is evaluated by comparing it with the corresponding “ground
truth” average speeds. The coverage is examined by finding the fraction of road
links, which has available estimations in that time interval. Finally, with a simple
threshold technique, estimated link speeds are classified into several traffic condi-
tion levels, which can be presented as colored road segments on the service sub-
scribers’ mobile displays.

Road Network Traffic Demand A-GPS Mobile Phone Field Tests

actual locations of
Microscopic Traffic vehicles / mobiles
Simulation on Net-state dumps Pre-processing Location Data
Urban Road Networks Defining the mobile probe penetration
Aggregated edge states Introducing A-GPS positioning errors
“ground truth” the “realistic”
estimations of
average speeds A-GPS data
Performance average speeds
Post-processing Location Data
Evaluation
accuracy & accuracy & coverage Positioning & tracking using KF
coverage acceptable not acceptable Filtering out undesired data
individual location
Adjustment of & speed estimates
Threshold-based Classification
color-coded
System Parameters Estimating Average Link Speeds
road segments Probe penetration Allocating speed estimates to links
Presenting on
Sampling frequency Aggregating at predefined intervals
Mobile Displays

Figure 5.1: Simulation-based Framework.

5.2 SYSTEM OVERVIEW 55

5.2.2 Urban Traffic Modeling

A wide variety of traffic simulators is available nowadays, and selecting a certain
tool depends very much on the system’s requirements. Despite their variations,
traffic simulation tools fall into four classes: macroscopic (e.g., [82]), mesoscopic
(e.g., [83]), microscopic (e.g., [84-86]) and sub-microscopic (e.g., [87]). These four
classes of traffic simulation tools are distinguished according to the modeling level
of detail. Among them, microscopic traffic simulation is based on the emulation of
traffic flow from the dynamics of individual vehicles [88]. Therefore, this is an
attractive approach for our system, which is based on data collected from individual
mobile phone probes.
In this work, the microscopic road traffic simulation package “Simulation of
Urban MObility” (SUMO) [89] is employed to model the urban traffic on arterial
roads. SUMO is selected for our research mainly considering the following features
which meet our needs: 1) it is a microscopic model which can record trajectories of
vehicles, individually; 2) it is open source, highly portable, and can be further
extended; 3) it is capable of making real-time simulations with a large amount of
simulated vehicles in large area; 4) it supports the importing of foreign map formats,
route generation, traffic light systems and several useful output formats; and 5) it
has many extensions, which can enhance the traffic modeling and data processing.
In addition, the credibility of the SUMO tool has been demonstrated in various
projects, e.g., [90-92].
A typical traffic simulation requires two inputs: road network and traffic demand.
In this work, two applications featured in the SUMO package are used to generate
the road network and vehicle routes: NETCONVERT imports digital road networks
from different sources and converts them into the SUMO-format; and DFROUTER
generates random routes and emits vehicles into networks. In addition, eWorld [93],
a SUMO extension, is employed to facilitate the processes of importing the
OpenStreetMap (OSM) data [94], then editing, enriching it and finally exporting the
data files of networks and routes for SUMO simulation.
As a result of the SUMO simulation, two useful datasets can be generated for
further analysis. One is the aggregated speed information for each road link/edge
called “aggregated edge states”. It includes information such as road edge IDs, time
intervals, mean speeds, etc. These aggregated speeds can be used to determine the
“ground truth” of traffic flow. The other is the location information of every vehicle
called “net-state dumps”. It records, at every timestamp, the location of every ve-
hicle in the simulated road network. Each record consists of a vehicle ID, a
timestamp, and the vehicle’s coordinates. This data file is used as the basis for the
simulation of the mobile probe-based traffic information system.
56 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

5.2.3 System Design Parameters

For such a probe-based monitoring system that makes use of collected location
samples, the sample size and the sampling frequency are two key design parameters
that would influence the system performance. These sampling related issues have
been discussed extensively in many previous studies of probe-based traffic systems.
In those studies (as illustrated below), both the experimental figures and the ana-
lytical models were presented, which could be very good references for this work.
As it has been discussed in the introduction, the main concern of probe-based
monitoring is the determination of the probe penetration rate (i.e., the percentage of
vehicles/mobiles that serve as traffic probes) to ensure an acceptable quality. Simi-
lar conclusions were drawn from previous field tests and simulation studies [67],
[81], [95-97]: probe-based system can be expected to work well for freeways with
penetration rates range from 3% to 5%. However, as indicated in [67], [81], [98],
urban arterials may require a penetration rate greater than 7% to provide reliable
speed estimates.
Another essential issue in traffic systems using GPS equipped probes is the data
sampling/reporting intervals. Typical GPS receivers receive location updates at
every 1-3 seconds. This frequency of data collection from a large number of probes
may cause network congestion. To avoid this issue, a temporal sampling method is
usually applied in which probes report their data at a prescribed time. In addition,
[99] claimed that using longer sampling intervals allows gather information over
longer distances, hence reducing the chance of capturing a non-representative
speed. However, this sampling interval cannot be too long, since it affects the
timeliness of data and the system coverage on short road links. A sampling interval
of 10-20 seconds can be used in practice, as indicated by the previous studies [67],
[68], [81], [100], [101].

5.3 Data Processing and Aggregation

Once the urban traffic is modeled and simulated as described in Section 5.2, the
remaining tasks would focus on using the generated data samples to estimate aggre-
gated traffic states on urban road links. In this section, the processes, methods, and
algorithms involved in the right branch of the simulation-based framework are
presented in detail. There are mainly three tasks: emulation of A-GPS location
updates, filtering of individual estimates, and estimation of average link speeds.

5.3.1 A-GPS Probe Data Emulation

Due to technological and practical limitations, location updates collected by A-GPS
mobile phones are not perfectly precise and limited in sample size. Therefore, both
5.3 DATA PROCESSING AND AGGREGATION 57

quality and quantity of the location data generated from traffic simulation should be
reduced in order to emulate the realistic field condition. As described in Section
5.2.1, location data should be degraded in two ways: 1) setting a specified per-
centage of simulated vehicles/mobiles to act as traffic probes, and 2) introducing
statistical positioning errors of A-GPS mobile phones.
The SUMO simulation output file “net-state dumps” may grow extremely large
since it contains detailed information of each vehicle/mobile. Hence, there is a need
of converting this location data into a more compressed one. In addition, positions
in SUMO are expressed in Cartesian coordinates instead of using the WGS84 [102].
In this work, a Java API, called SUMOPlayer [103], is employed to “play” the
SUMO network-dump files in real-time to WGS84 coordinates for each probe. The
SUMOPlayer is also customized by defining parameters such as fraction of tracked
vehicles (i.e., the probe penetration rate).
In order to get practical error statistics of A-GPS mobile location updates, field
tests were conducted on A-GPS mobile phones. The location updates include the
latitude-longitude coordinates, their accuracies, and the corresponding timestamps.
Horizontal accuracy (in meters) is the root mean square (RMS) of the north and east
accuracy (1-sigma standard deviation). Under the Gaussian assumption (can be
motivated by the central limit theorem according to [104]), this implies that the
actual location is within the circle defined by the returned point and radius at a
probability of about 68%. Figure 5.2 shows the A-GPS location samples collected
continuously at a 10-second interval for 1-hour in an urban area of Stockholm. The
median of RMS errors was found to be 8.83 m; and 90% of the errors were found
below 18.53 m, according to Cumulative Distribution Function (CDF). Although
these A-GPS location measurements are less accurate than those from regular GPS
units, they still appear sufficient for traffic state estimation (a location technology
within 20-meter accuracy can produce quantitative travel information [105]).

Figure 5.2: Vehicle trajectory samples collected by the A-GPS mobile phone.
58 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

5.3.2 Filtering of Location Data

After the pre-processing stage, the A-GPS mobile phone-based data collection is
emulated for every traffic probe. These “realistic” location samples cannot be
directly utilized in traffic estimation since they are inherently erroneous. A two-step
post-processing is therefore applied: 1) the Kalman filtering (KF) to transform
A-GPS measurements into dynamical state estimates of position and velocity; and
2) the simple data screening to eliminate some undesired data.

5.3.2.1 KF-based Tracking

Similar as in Section 4.2, the KF is exploited to track the vehicle/mobile probes.
Suppose that a vehicle, equipped with a mobile, moves in a 2D Cartesian coordinate
system (SUMO works with Cartesian coordinates only; and road networks specified
with WGS84 are converted by NETCONVERT using UTM [106]). In order to track
this moving mobile probe, its state can be expressed as a dynamic state vector:
T
X (t ) = éê x (t ) y(t ) x(t ) y(t ) ùú (5.1)
ë û
where x (t ) and y(t ) are the positions and their first derivatives x(t ) and y(t ) are
velocities. In this way, the motion dynamics can be described by a continuous white
noise acceleration model [59], in which the velocity is modeled as a Wiener process
(i.e., the integral of white noise). In this model, the velocity undergoes perturbations
which are modeled by zero-mean white Gaussian noise v(t ) :
v(t ) = x(t ) = y(t ) (5.2)
with the variance

E éë v(t )v(t ) ùû = qc d(t - t ) (5.3)

where qc is the power spectrum density. The continuous-time transition equation

is written as:
X (t ) = AX (t ) + V (t ) (5.4)
where
é0 0 1 0ù é 0 ù
ê ú ê ú
ê0 0 0 1ú ú ê 0 ú
A = êê V (t ) = êê ú (5.5)
ê0 0 0 0 úú ê v (t ) ú
ú
ê0 0 0 0 úú ê v(t ) ú
êë û êë úû
5.3 DATA PROCESSING AND AGGREGATION 59

Let the sampling interval in this system to be D T , then after discretization, the
discrete-time transition equation can be written as:
X (k + 1) = FX (k ) + V (k ) (5.6)
with the transition matrix
é1 0 DT 0 ù
ê ú
ê0 1 0 DT úú
F = e ADT = êê (5.7)
ê0 0 1 0 úú
ê0 0 0 1 úú
êë û
T
and the process noise vector V (k ) = éê 0 0 v(k ) v(k ) ùú , which models the
ë û
disturbance in velocity. The covariance matrix of the process noise vector is:

Q = E êéV (k )V (k )T ùú
ë û
é1 1 ù
ê DT 3
0 DT 2
0 ú
ê3 2 ú
ê 1 1 ú
ê 0 D T 3
0 D T 2ú
ê ú (5.8)
= ê 3 2 ú qc
ê1 2 ú
ê DT 0 DT 0 ú
ê2 ú
ê 1 ú
ê 0 DT 2
0 DT úú
êë 2 û
T
The state vector X (k ) = éê x (k ) y(k ) x(k ) y(k ) ùú is then related to the noisy
ë û
location observations Y (k ) by the measurement equation:

Y (k ) = HX (k ) + W (k ) (5.9)

é1 0 0 0ù
with the measurement matrix H = ê 0 1 0 0 ú , and the measurement noise
êë úû
é w(k ) ù é0ù éR 0 ù
W (k ) = ê ú ~ N (ê ú , ê ú) , where R is the measurement error variance.
êë w (k ) úû êë 0 úû êë 0 R úû

Then, the transition and measurement equations can be rewritten in a more

compact way as:
60 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

ì Xk = FXk -1 + Vk -1
ï
ï
í (5.10)
ï
ïV ~ N (0,Q )
î k -1
ìYk = HXk + Wk
ï
ï
í (5.11)
ï
ïW ~ N (0, R)
î k
where Q and R are the covariance matrices of the process error Vk -1 and meas-
urement error W k , respectively. Then, the optimal estimations are obtained by the
following iterated steps:
 Before the measurements are available at tk , priori estimates of the state
mean X̂k- and covariance Pk- are obtained by the time update equations:

Xˆk- = FXˆk -1 (5.12)

Pk- = FXˆk -1FT + Q (5.13)

 Kalman gain K k is then computed to set an appropriate correction term for

the next propagation step, so as to minimize the mean square estimation
(MSE) error:

Kk = Pk-H T (HPk-H T + R)-1 (5.14)

 After processing the noisy measurements Yk , posteriori state mean estimate

X̂ k is obtained by updating X̂k- with a corrected version of measurement
residual:

Xˆk = Xˆk- + Kk (Yk - HXˆk-) (5.15)

Then the covariance matrix Pk associated with X̂ k can be updated from

the priori estimate as:

Pk = (I - Kk H )Pk- (5.16)

5.3.2.2 Data Screening

From the last subsection, the recursively updated position and velocity estimates of
each probe are obtained. Before they can be aggregated to provide the estimation of
average link speeds, simple data screening process needs to be applied to filter out
some undesired data. One challenge in the network-based mobile probe systems is
the need to distinguish non-valid probes (e.g., mobile users in buildings, on sub-
5.3 DATA PROCESSING AND AGGREGATION 61

ways or pedestrians) from mobile phones travelling on-board vehicles. As stated in

[107], since the outliers’ influence is severe, especially for dense urban areas, they
should be identified and filtered out. In our system, the validity of traffic probes is
not a big issue any more. Unlike the network-based probe method that randomly
monitors mobile users within a wireless network, probe data in this system come
from our service subscribers, and we can assume that the service subscribers would
only start the traffic application when they are in vehicles. Provided that in this
system our mobile application users are unlikely to be non-valid probes, following
criteria are considered for the data screening process:
 Speed estimates that are greater than 120% of speed limits should be elimi-
nated, since those very large speed estimates most probably source from
positioning errors (speed limits in Stockholm downtown areas vary from 30
km/h to 50 km/h [108]).
 Location estimate with a distance to the nearest road link larger than 20 m
should be eliminated. This criteria helps to solve the problem of mapping
estimates between two nearly parallel links (according to [11], location ac-
curacy of at least 20 m is expected to differentiate between closely spaced
parallel urban roads).

5.3.3 Estimation of Aggregated Link Speeds

Generally, in this work the real-time traffic states are characterized by the average
link speeds of the simulated urban arterials along with rolling time periods. Within
each period (i.e., the aggregation interval), speed estimates from individual probes
are firstly allocated to road links and then aggregated to provide estimates of the
average conditions for the road links.
5.3.3.1 Allocation of State Estimates
After applying the filtering steps illustrated in the previous section, the estimated
vehicle/mobile tracks are still deviated from their actual trajectories due to the
introduction of location errors. Therefore, these position/velocity estimates need to
be projected onto road links so as to get traffic information for each link (known as
map matching). Map matching algorithms can be divided into three major catego-
ries: geometric approach, topological approach and statistical approach. In this
work, the simple point-to-curve geometric map matching technique [109] is ap-
plied, because of its effectiveness and given the real-time requirement of this sys-
tem. Projection distances are calculated from a position estimate to each road link
candidate. The road link, which gives the smallest distance, is identified as the
associated link to that position/velocity estimate. As a result, at every sampling step,
speed estimations can be allocated to specific links in the simulated network.
62 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

From the estimation result of 5.3.2, we can easily derive the position and speed of
the i th probe at tk as:

ì
ï T
ïï pˆi (k ) = éê xˆi (k ) yˆi (k ) ùú
ï ë û (5.17)
í
ï
ï vˆi (k ) = ( xˆ i (k ) ) + ( yˆ i (k ) )2
2
ï
ï
î

After map matching, at certain timestamps, we can obtain vˆij (k ) as the speed
estimate of link j from probe i , where i Î {1, 2,..., n p } ( n p is the total number
of probes) and j Î {1, 2,..., nl } ( nl is the number of monitored road links).

5.3.3.2 Aggregation of Link Speeds

In Section 5.3.3.1, the mapped speed estimates are recorded for each link along with
sampling time stamps. Since traffic conditions are characterized by average travel
speed on road links, the rest of the problem is to aggregate the speed estimates over
a specific time interval. Previous works on traffic estimation [3], [67], [110]
indicate that a 10-minute aggregation time appears a reasonable choice taking into
consideration both the real-time requirement and data availability. As a result of the
aggregation, the average speed along track j during the interval of interest is:
tk +DT
1
j
Vave (tk , tk +DT ) = å vˆj (k ) (5.18)
ntj ,t k =tk
k k +DT

where vˆj (k ) is the available speed estimate on link j during interval éë tk , tk +DT ùû ,
and n j is the total number of available estimates.
tk ,tk +DT

5.4 Simulation Results and Evaluation

As shown in Figure 5.3 (a), the road network for parts of downtown area in
Stockholm has been chosen as a simulation case study. The OpenStreetMap (OSM)
XML file is firstly edited in Java OpenStreetMap Editor (JOSM) [111] in order to
remove the road edges which cannot be used by vehicles, such as railway, roadways
for motorcycle, bicycle, and pedestrian, etc. In addition, all the edges are set as
one-way for simplicity. The simplified version of network, which consists of 14
nodes, 22 links as well as 4 traffic lights, is imported into the eWorld, as shown in
5.4 SIMULATION RESULTS AND EVALUATION 63

Figure 5.3 (b). In eWorld, we can further check and edit other properties such as
street name, maximum speed limit, and phase assignment of traffic light. All of this
information can be accessed from the OSM data. Then, with the export feature of
eWorld, the network file, required by SUMO simulation, can be generated using
NETCONVERT. Since OSM data works with the WGS84 instead of the Cartesian
coordinate system used by SUMO, projection is applied during the conversion.
Traffic light logics, speed limits and priorities are also encoded in the network files.

Figure 5.3: Simulated road network: (a) image from the OpenStreetMap, and (b)
simplified diagram in the eWorld.
Once the road network is ready, the next step is to generate traffic on the road
edges/links. In the eWorld, the properties of vehicles, such as acceleration, maxi-
mum allowed speed, and speed variation, are firstly defined. Then, random routes of
vehicles traveling on the road network are generated for a specific time interval
using the DUAROUTER. Since the limited simulation network makes the traveling
vehicles leave the network fast (typically no more than 250 s), in each time step a
given number of vehicles are emitted to the network in order to achieve traffic
equilibrium. These randomly generated vehicles and their routes are also exported
as the SUMO file. As a result of a 3600-second-interval, 575 vehicles have been
generated with random trips.
The SUMO simulation is conducted for 1-hour without incidents. The real-time
traffic data, i.e., the aggregated link/edge state is collected together with the net-
work state dump. As introduced in Section 5.2.2, the aggregated state file is used to
establish the “ground truth” link speeds, and the network state dump file is used as
input to the SUMOPlayer to generate the mobile probe data. Suppose that the travel
64 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

speeds of interest are those of links 1, 4, 5, 6, 9, 12, 15, 18, 19, and 22. The links’
length ranges from 173 m to 236 m. Other links are not included mainly because
their aggregated density and occupancy are generally small, which makes the
estimation less necessary, and potentially it results in small number of probes.
Figure 5.4 shows the mean traveling speeds of selected links recorded in the
“aggregated edge states”. Since only Link22 has a speed limit of 50 km/h and all the
others have speed limits of 30 km/h, average link speed of Link22 is considerably
higher. In addition, Link6 and Link15 have lower traveling speeds, which are
expected to be detected as congestions in the estimation.
15
Link1
Average link speeds (m/s)

Link4
12
Link5
9
Link6
Link9
6 Link12
Link15
3 Link18
Link19
0 Link22
10 20 30 40 50 60
Timestamps (min)

Figure 5.4: Aggregated link speeds - the “ground truth” data.

The SUMO output “net-state dumps” is then used as input for the SUMOPlayer
to generate a large amount of simulated mobile probes in real-time. The
SUMOPlayer first reads the corresponding network file and then chooses randomly
vehicles from the network state dump, according to the probe penetration rate (10%
is specified in this case). As a result, it writes location updates (longitudes and
latitudes) every second for the vehicles that are selected as traffic probes. In order to
be compatible with the SUMO network format, those coordinates are projected
back to zero-origin Cartesian system with the same offset used by the
NETCONVERT. There are in total 56 probes and the time they spend in the system
ranges from 49 s to 241 s. Figure 5.5 (a) plots location data collected from the
chosen probes, which are aggregated at every 10-minute. As illustrated previously,
these locations should be further processed. A sampling interval of 10-second is
used to achieve a balance among communication load, data effectiveness, timeli-
ness and availability. A normal distributed positioning error is introduced to the
position coordinates with 1-sigma standard deviation of 8.83 m in both northing and
easting directions. The emulated A-GPS location samples after this pre-processing
are plotted in Figure 5.5 (b).
5.4 SIMULATION RESULTS AND EVALUATION 65

Aggregated probe data from SUMOPlayer "Realistic" location samples after pre-processing

60
60
Aggregation timestamps (min)

Aggregation timestamps (min)

50
50
40
40
30 800
800 30
20 600
600 20

10 400
400 10

200 0 200
0
0 200 0 200
400 600 0 400 600 0
800 800

Figure 5.5: Probe location aggregated every 10 min: (a) individual probe locations
from SUMOPlayer and (b) emulated “realistic” A-GPS location samples.
As stated in Section 5.3.2, emulated A-GPS location samples are post-processed
in two steps: Kalman filtering (KF) and data screening (DS). The KF step results in
dynamically updated state estimates of position and velocity. Figure 5.6 shows the
A-GPS location updates of all the mobile probes from 1-hour simulation as well as
all the position estimates after the KF. In the DS step, the position/velocity esti-
mates, which fail to meet the criteria (specified in Section 5.3.3.2), are discarded.

Figure 5.6: A-GPS measurements vs. position estimates after Kalman filtering.
66 CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
PHONES AS PROBES

After the post-processing, accuracies of position and speed estimations are evalu-
ated statistically (the actual position and speed of each probe is available in the
SUMOPlayer output). The figure of merits used to evaluate the nth position and
speed estimates (out of totally 701 estimates) are the root square error (RSE) and the
absolute error (AE), respectively:
n
RSE position = (x n - xˆn )2 + (yn - yˆn )2 (5.19)

n
AEspeed = xˆn 2 + yˆn 2 - vn (5.20)

where (x n , yn ) and vn are actual positions and speeds; (xˆn , yˆn ) and xˆn 2 + yˆn 2
are estimated positions and speeds. Table 5.1 lists the statistics (mean, median, and
standard deviation) of the position and speed estimation errors. It is worth noticing
that more accurate speed estimates are obtained from relatively inaccurate position
estimates. This is mainly due to the fact that the speed estimate is derived from two
position estimates and the errors in position estimates are relatively small compared
to the distance traveled by the probe between successive estimates [112].
Table 5.1: Statistics of position and speed estimation errors after data screening.

Total Number of Estimates: Statistics

n = 701 Mean Median Std. deviation
RSE of position estimates (m) 7.7082 7.2769 4.0637
AE of speed estimates (m/s) 1.7985 1.0296 1.9984

Each state estimate after KF and DS is then allocated to a specific road link
through the geometric map matching. Generally, the purpose of map-matching is
twofold: 1) identification of the correct road link, and 2) determination of the vehi-
cle position on that link. Accordingly, two metrics are commonly used for perfor-
mance assessment: 1) the percentage of correct link identification (CLI), and 2)
horizontal accuracy in meters. For example, in [113], several map-matching algo-
rithms are compared according to these two measures. In this work, map-matching
is not used for location estimation since the performance derived from it can be
misleading if the location is projected to an incorrect link. Therefore, here only the
problem of road link identification is addressed, and the percentage of CLI is
evaluated. Specifically, throughout the simulation, totally 63 vehicles are selected
as traffic probes, and each of them has its own route and a set of location/speed
estimates resulting from the KF+DS step. After the map-matching, each of its
5.4 SIMULATION RESULTS AND EVALUATION 67

estimates is allocated to a road link in the network. For each probe, the percentage
of correct matches on its route is firstly calculated, which is followed by a statistical
analysis of the CLI rates from all the probes. Table 5.2 lists the statistics of the
correct link identification (CLI) rates from 63 probe routes. As shown in the table,
in average 84.92% of the estimates are mapped correctly to the road links. Sources
of error in the link identification are largely due to the fact that vehicles traveling on
urban road networks tend to stay stationary for a while (typically 30-60 seconds) in
response to a traffic signal at the cross. Since the mapped speed estimates are used
only to calculate the average speed, this level of CLI is not as problematic as in the
vehicle navigation and road-pricing applications, which rely on very accurate
vehicle positioning.
Table 5.2: Evaluation of the link allocation (map-matching).

Number of Statistics of the Correct Link Identification Rate

Tracks: 63 Mean Median Std. deviation
CLI Rate 84.92% 87.87% 13.06%

For each road link, the mapped speed estimates are accumulated every 10
minutes. They are then aggregated to estimate the average link speed during that
10-min interval. The resultant average link speed estimations for selected links are
shown in Table 5.3, being compared with the “ground truth” average link speeds
recorded from the traffic simulation. As shown in the table, two performance
metrics are considered: 1) the estimation accuracy evaluated by the mean absolute
error, and 2) the system coverage evaluated by the speed estimation availability.
The mean absolute error is defined as the absolute difference between the true speed
on the link and the estimated speed. The speed estimation availability is the fraction
of links that have speed estimates available in the time interval.
In addition, the estimated road traffic speeds are classified into three traffic
condition levels, i.e., green, red, and yellow: 1) green level (smooth traffic) if link
speed is above 7 m/s; 2) red level (congested traffic) if link speed is below 4 m/s;
and 3) yellow level (medium traffic) if link speed is between 4 m/s and 7 m/s. These
two speed thresholds are determined considering the state-of-the-art traffic speed
classification in urban area [68], [114]. As it can be seen in the table, the congested
traffics on link 6 and 15 have been detected. These estimated traffic conditions will
be color-coded on the road network and presented on the service subscriber’s mo-
bile display.
 68

Link# (Street Name) 0-10min 10-20min 20-30min 30-40min 40-50min 50-60min

actual│estimated average link speeds (m/s) every 10min interval
Link1 (Olofsgatan) 6.95│NA 6.86│6.29 7.25│NA 7.03│7.82 7.14│6.73 6.90│NA
Link4 (Tegnérgatan) 6.35│6.33 6.32│4.66 6.02│7.23 6.21│7.13 6.13│6.57 6.46│6.73
Link5 (Drottninggatan) 7.16│NA 7.11│7.36 7.19│NA 7.01│7.52 7.16│7.70 7.18│7.52
Link6 (A. Fredriks Kyrkogata) 1.49│1.61 2.31│2.94 2.66│0.84 2.11│1.55 1.80│2.68 2.14│2.34
Link9 (Kammakargatan) 7.46│8.00 7.43│8.42 6.68│NA 7.26│7.99 7.19│7.33 7.10│6.09
Link12 (Luntmakargatan) 7.37│7.80 7.40│7.76 7.39│NA 7.40│7.86 7.29│7.44 7.45│8.10
Link15 (Sveavägen) 2.67│3.74 3.13│4.66 2.84│3.03 2.80│3.05 2.55│3.66 2.62│3.53
Link18 (Holländargatan) 7.21│8.31 7.28│7.58 7.34│7.47 7.33│7.51 7.07│5.78 7.24│5.95
Link19 (Apelbergsgatan) 7.64│6.80 7.62│7.50 7.66│8.05 7.29│8.05 7.59│7.64 7.47│8.69
Link22 (Olof Palmes Gata) 11.38│NA 11.59│NA 11.39│11.69 11.34│11.77 11.53│11.66 11.49│10.83
Mean Absolute Error (m/s) 0.59 0.71 0.67 0.49 0.61 0.73
Speed Estimate Availability 70 % 90 % 60 % 100 % 100 % 90 %
NA: the speed estimate is not available green: smooth yellow: medium red: congested
PHONES AS PROBES
CHAPTER 5. TRAFFIC STATE ESTIMATION WITH A-GPS MOBILE
5.5 SUMMARY 69

5.5 Summary
In this chapter, a method of real-time urban traffic state estimation has been pre-
sented. The proposed method, taking advantage of the recently booming A-GPS
mobile phones, potentially solves the problems (e.g., cost and urban coverage) in
the current state-of-the-practice traffic systems. Based on the microscopic traffic
simulation and field tests, “realistic” A-GPS mobile probe data is emulated and
“ground truth” traffic data is generated. The A-GPS location samples are firstly
processed by Kalman filtering and data screening. The resultant position/speed
estimates are then allocated to nearest road links through simple map-matching. By
aggregating the speed estimates on each road link, traffic states (i.e., average link
speeds) are determined every 10 minutes for 1 hour. The achieved simulation results
suggest that reliable average link speed estimations can be generated, which are
used for indicating the real-time urban road traffic condition. Future work targets a
smart traffic information system demonstrator that employs the proposed urban
traffic state estimation method.
CHAPTER 6

CONCLUSIONS

Traffic congestion in large cities is an issue that affects millions of people all over
the world in their daily lives. Over the decades, traffic systems, which rely mainly
on data collected from road-side detectors and probe vehicles, have been deployed.
This data is essential for monitoring the actual traffic state on road networks and for
supporting the development of management strategies that address traffic problems.
Since the traditional ways for traffic data collection are costly and have limited
coverage, on-board mobile phone tracking has emerged with promising solutions to
collect vehicle location data and generate traffic information in a cost-effective way.
However, early deployments of mobile phone-based traffic probe systems, most of
them network-based solutions, have not been entirely successful in generating
accurate and reliable traffic information. For a fully operational system to be de-
ployed there are various aspects that must be investigated. This thesis deals with
two major aspects in the deployment of a mobile phone-based smart traffic infor-
mation system: mobile positioning and tracking, and traffic state estimation. In this
concluding chapter, a summary of the contributions of this thesis is firstly presented;
our ongoing work of the system demonstrator is then briefly described; possible
directions for future work are finally outlined.

71
72 CHAPTER 6. CONCLUSIONS

6.1 Concluding Remarks

In the first part of the thesis, the problem of mobile location tracking with velocity
estimation was addressed. The investigation began with a comparative study of
various mobile positioning methods and their existing standardizations. Towards a
good compromise between accuracy and availability of vehicle location, possible
hybrid solutions to standard methods were proposed. Moreover, in order to evaluate
the OTDOA positioning error in absence of network measurement, a UMTS mobile
positioning simulator that targets vehicle location application was developed. A
hybrid mobile positioning and tracking algorithm, which can find location estimates
of the mobiles and track their trajectories in a simulated vehicular scenario, was
implemented. Although there are several hybrid mobile location and tracking
solutions, our approach is distinguished by that the 3G standard OTDOA and
A-GPS location measurements were fused at the state-vector level.
The focus of the second part of the thesis was on the problem of real-time urban
traffic estimation. A traffic simulation-based framework was proposed to emulate
and evaluate the operation of urban traffic estimation using A-GPS mobile phone
probes. Based on the emulated mobile probe data, algorithms of location data pro-
cessing/filtering and average speed estimation were developed and then evaluated
by comparing them against “ground truth” data from the traffic simulation.
Although there are considerable research activities and operational tests studying
the feasibility of mobile phone-based traffic information system, our contribution,
based on traffic simulation, lies in estimating the traffic speed on urban arterials
using only A-GPS location data. Additionally, the real-time traffic condition would
be color-coded and showed on the mobile phone display.

6.2 Ongoing Work: The System Demonstrator

The mobile phone-based smart traffic information system demonstrator is under
development. The system demonstrator aims to integrate different developed
components from several project work packages. For instance, the system demon-
strator employs the traffic state estimation method proposed in Chapter 5 to indicate
the real-time urban road traffic condition.
The system demonstrator integrates the following main components: 1) a mobile
client that periodically sends A-GPS location updates to the main server via UMTS;
2) the real-time urban traffic emulation that generates a large amount of simulated
A-GPS mobile clients; 3) the traffic server implementing a secure authentication
protocol to preserve communication security and user privacy; 4) the server logic
including road network generation, map-matching and traffic state estimation; and
6.3 FUTURE WORK 73

5) a basic user interface implemented on Android devices that presents traffic

information to subscribers using the Google Maps API.

6.3 Future Work

The achieved results demonstrate the effectiveness of the mobile phone-based
methods for vehicle location and traffic estimation. However, this research is just a
small step out of the tremendous amount of work that should be done towards the
fully operational mobile phone-based ITS deployment.
The work presented in this thesis opens several directions for further research and
development. Since the UMTS network parameters are kept confidential by mobile
operators, in this thesis the OTDOA positioning was entirely simulated using
statistical models. It would be interesting to evaluate the performance of the hybrid
positioning and tracking under a realistic cellular environment. The vehicle motion
dynamics (in Chapter 4 and 5) were characterized by the continuous white noise
acceleration model. Although the tracking capability of this mobility model was
verified by simulation, more complex models might be needed in order to take into
account the human control impact. The demonstrator described in the previous
section would be based on the emulated real-time traffic on urban road networks. A
field experiment with large amount of our mobile appilcation users driving on
actual road networks should be a more powerful demonstration.
BIBLIOGRAPHY

[1] Q. T. Minh and E. Kamioka, “Pinpoint: An Efficient Approach to Traffic State

Estimation System Using Mobile Probes,” in International Conference on Wireless
Communications Networking and Mobile Computing, 2010, pp. 1–5.

[2] J. C. Herrera, “Assessment of GPS-enabled smartphone data and its use in traffic
state estimation for highways,” University of California, Berkeley, 2009.

[3] M. A. Bacchus, B. Hellinga, and M. P. Izadpanah, “An Opportunity Assessment of

Wireless Monitoring of Network-Wide Road Traffic Conditions,” 2007. [Online].
Available:
http://www.civil.uwaterloo.ca/bhellinga/publications/Publications/MTO-2007
Wireless Traffic Monitoring Final Report.pdf.

[4] V. Manolopoulos, S. Tao, S. Rodriguez, M. Ismail, and A. Rusu, “MobiTraS: A

Mobile Application for a Smart Traffic System,” in IEEE International NEWCAS
Conference, 2010, pp. 365-368.

[5] V. Manolopoulos, P. Papadimitratos, S. Tao, and A. Rusu, “Securing Smartphone

Based ITS,” in IEEE International Conference on ITS Telecommunications, 2011,
pp. 201-206.

[6] J. Korhonen, Introduction to 3G Mobile Communications. Artech House, 2003.

[7] 3GPP, “3GPP TS 23.271, Functional stage 2 description of Location Services

(LCS) (Release 10),” 2011. [Online]. Available:
http://www.3gpp.org/ftp/Specs/html-info/23271.htm.

[8] FCC, “FCC acts to promote competition and public safety in enhanced wireless
911 services,” 1999. [Online]. Available:
http://transition.fcc.gov/Bureaus/Wireless/News_Releases/1999/nrwl9040.html.

75
76 BIBLIOGRAPHY

[9] A.-K. V. Rongen, “The Necessity for an EU-wide 112 Emergency Wireless Loca-
tion System,” European Journal of Navigation, vol. 3, no. 4, pp. 32-35, 2005.

[10] 3GPP, “3GPP TS 22.071, Location Services (LCS); Service description; Stage 1
(Release 10),” 2011. [Online]. Available:
http://www.3gpp.org/ftp/Specs/html-info/22071.htm.

[11] Y. Zhao, Vehicle Location and Navigation Systems. Artech House, 1997.

[12] A. Küpper, Location-Based Services: Fundamentals and Operation. John Wiley &
Sons, 2005.

[13] “GPS Standard Positioning Service Signal Specification,” 1995. [Online]. Availa-
ble: http://www.pnt.gov/public/docs/1995/signalspec1995.pdf.

[14] J. J. Caffery, Wireless Location in CDMA Cellular Radio Systems. Kluwer Aca-
demic Publishers, 2000.

[15] C. Drane and C. Rizos, Positioning Systems in Intelligent Transportation Systems.

Artech House, 1998.

[16] 3GPP, “3GPP TS 43.059, Functional stage 2 description of Location Services

(LCS) in GERAN (Release 10),” 2011. [Online]. Available:
http://www.3gpp.org/ftp/Specs/html-info/43059.htm.

[17] Y. Zhao, “Mobile Phone Location Determination and its Impact on Intelligent
Transportation Systems,” IEEE Transactions on Intelligent Transportation Sys-
tems, vol. 1, no. 1, pp. 55-64, Mar. 2000.

[18] 3GPP, “3GPP TS 25.305, Stage 2 functional specification of User Equipment (UE)
positioning in UTRAN (Release 10),” 2010. [Online]. Available:
http://www.3gpp.org/ftp/Specs/html-info/25305.htm.

[19] J. Borkowski and J. Lempiäinen, “Practical Network-based Techniques for Mobile

Positioning in UMTS,” EURASIP Journal on Applied Signal Processing, vol.
2006, pp. 1-15, 2006.

[20] Y. J. Guo, Advances in Mobile Radio Access Networks. Artech House, 2004.

[21] 3GPP, “3GPP TS 36.305, Stage 2 functional specification of User Equipment (UE)
positioning in E-UTRAN (Release 10),” 2011. [Online]. Available:
http://www.3gpp.org/ftp/Specs/html-info/36305.htm.

[22] A. Bensky, Wireless Positioning Technologies and Applications. Artech House,

2008.
BIBLIOGRAPHY 77

[23] “3G/UMTS family welcomes its 500 millionth customer,” 2010. [Online]. Avail-
able: http://www.umts-forum.org/content/view/3230/303/.

[24] I. K. Adusei, K. Kyamakya, and K. Jobmann, “Mobile positioning technologies in

cellular networks: an evaluation of their performance metrics,” in IEEE MILCOM,
2002, pp. 1239-1244.

[25] T. Kos, M. Grgic, and G. Sisul, “Mobile User Positioning in GSM/UMTS Cellular
Networks,” International Symposium ELMAR, pp. 185–188, Jun. 2006.

[26] J. Syrjärinne, “Studies of Modern Techniques for Personal Positioning,” Tampere

University of Technology, 2001.

[27] P. J. Duffett-Smith, “Precise UE positioning in UMTS using cumulative virtual

blanking,” in International Conference on 3G Mobile Communication Technolo-
gies, 2002, vol. 2002, no. 489, pp. 355-359.

[28] “Canalys.” [Online]. Available: http://www.canalys.com/index.html.

[29] D. Drakoulis, S. Kyriazakos, and G. Karetsos, “Improving subscriber position

location using a hybrid satellite-assisted and network-based technique,” in IEEE
Vehicular Technology Conference, 2000, pp. 1887-1893.

[30] H. C. I. Son, J. G. Lee, and G. I. Jee, “Mobile station location using hybrid GPS and
a wireless network,” in IEEE Vehicular Technology Conference, 2003, vol. 4, pp.
2716-2720.

[31] Q. Huang and J. Liu, “Mobile Tracking Using Hybrid Measurements in NLOS
Environments,” in International Conference on ITS Telecommunications, 2006,
pp. 1075–1078.

[32] P.-hsuan Tseng and K.-ten Feng, “Hybrid Network/Satellite-Based Location

Estimation and Tracking Systems for Wireless Networks,” IEEE Transactions on
Vehicular Technology, vol. 58, no. 9, pp. 5174-5189, 2009.

[33] 3GPP, “3GPP Work Item Description: GP-091075.” [Online]. Available:

http://www.3gpp.org/ftp /Specs/html-info/FeatureOrStudyItemFile-440003.htm.

[34] “TruePosition.” [Online]. Available: http://www.trueposition.com.

[35] G. Sun, J. Chen, W. Guo, and K. J. R. Liu, “Signal Processing Techniques in

Network-Aided Positioning: A survey of state-of-the-art positioning designs,”
IEEE Signal Processing Magazine, vol. 22, no. 4, pp. 12–23, Jul. 2005.
78 BIBLIOGRAPHY

[36] S. Fischer, H. Koorapaty, E. Larsson, and A. Kangas, “System performance

evaluation of mobile positioning methods,” in IEEE Vehicular Technology Confer-
ence, 1999, vol. 3, pp. 1962–1966.

[37] M. Wylie-Green and P. Wang, “GSM mobile positioning simulator,” in IEEE

Emerging Technologies Symposium: Broadband, Wireless Internet Access, 2000,
pp. 874–878.

[38] J. M. Borkowski, “On applicable cellular positioning for UMTS,” Tampere Uni-
versity of Technology, 2008.

[39] H. Holma and A. Toskala, WCDMA for UMTS. John Wiley & Sons, 2000.

[40] B. Ludden and L. Lopes, “Cellular based location technologies for UMTS: a
comparison between IPDL and TA-IPDL,” in IEEE Vehicular Technology Con-
ference, 2000, pp. 1348-1353.

[41] F. Sallabi, A. Lakas, K. Shuaib, and M. Boulmalf, “WCDMA downlink simulator

with efficient wrap-around technique,” in IFIP International Conference on Wire-
less and Optical Communications Networks, 2005, pp. 472–476.

[42] 3GPP, “3GPP TR 25.942, Radio Frequency (RF) system scenarios (Release 10),”
2011. [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/25942.htm.

[43] T. Ojanpera and R. Prasad, Wideband CDMA for Third Generation Mobile
Communications. Artech House, 1998.

[44] T. S. Rappaport, Wireless Communications: Principles and Practice. Pren-

tice-Hall, 2002.

[45] E. Damosso, “Digital Mobile Radio towards Future Generation Systems,” 1999.
[Online]. Available: http://kom.aau.dk/antprop/pub/cost231.html.

[46] M. Gudmundson, “Correlation model for shadow fading in mobile radio systems,”
Electronics letters, vol. 27, no. 23, pp. 2145–2146, 1991.

[47] 3GPP, “3GPP TR 25.943, Deployment aspects (Release 10),” 2011. [Online].
Available: http://www.3gpp.org/ftp/Specs/html-info/25943.htm.

[48] 3GPP, “3GPP TS 25.213, Spreading and modulation (FDD)(Release 10),” 2010.
[Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/25213.htm.

[49] L. M. Correia, Wireless Flexible Personalized Communications – COST 259:

European Co-operation in Mobile Radio Research. John Wiley & Sons, 2001.
BIBLIOGRAPHY 79

[50] 3GPP, “3GPP TS 25.215, Physical layer; Measurements (FDD) (Release 10),”
2011. [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/25215.htm.

[51] C. Knapp and G. Carter, “The generalized correlation method for estimation of
time delay,” IEEE Transactions on Acoustics, Speech and Signal Processing, vol.
24, no. 4, pp. 320–327, 1976.

[52] Y. T. Chan and K. C. Ho, “A simple and efficient estimator for hyperbolic loca-
tion,” IEEE Transactions on Signal Processing, vol. 42, no. 8, pp. 1905-1915,
1994.

[53] Y. Bar-Shalom and L. Campo, “The effect of the common process noise on the
two-sensor fused-track covariance,” IEEE Transactions on Aerospace Electronic
Systems, vol. 22, no. 10, pp. 803–805, 1986.

[54] J. R. Raol, Multi-Sensor Data Fusion with MATLAB. CRC Press, 2010.

[55] JCP, “Java Specification Request (JSR) 179: Location API for J2METM (Final
Release 3),” 2011. [Online]. Available: http://jcp.org/en/jsr/detail?id=179.

[56] Y. Zhao, “Standardization of mobile phone positioning for 3G systems,” IEEE

Communications Magazine, vol. 40, no. 7, Institute of Electrical and Electronics
Engineers, Inc, 445 Hoes Ln, Piscataway, NJ, 08854-1331, USA,, pp. 108–116,
Jul-2002.

[57] J. Y. Huang, Q. Wan, and P. Wang, “Minimum mean square error estimator for
mobile location using time-difference-of-arrival measurements,” IET Radar, Sonar
and Navigation, vol. 5, no. 2, pp. 137-143, 2011.

[58] D. Simon, Optimal State Estimation: Kalman, H Infinity and Nonlinear Ap-
proaches. John Wiley & Sons, 2006.

[59] Y. Bar-Shalom, X. R. Li, and T. Kirubarajan, Estimation with Applications to

Tracking and Navigation. John Wiley & Sons, 2001.

[60] M. McGuire and K. N. Plataniotis, “Dynamic model-based filtering for mobile

terminal location estimation,” IEEE Transactions on Vehicular Technology, vol.
52, no. 4, pp. 1012-1031, Jul. 2003.

[61] J. Vidal, A. Pagès-Zamora, O. Muñoz, M. Cabrera, M. Nájar, and M. Lagunas,

“D231: Overview of UMTS User Location Possibilities and Limitations,” 2000.
[Online]. Available: http://gps-tsc.upc.es/comm/saturn/pdf/D231_sum.html.

[62] Y. Wang, M. Papageorgiou, and A. Messmer, “Real-time Freeway Traffic State

Estimation Based on Extended Kalman Filter: Adaptive Capabilities and Real Data
Testing,” Transportation Research A, vol. 42, pp. 1340-1358, 2008.
80 BIBLIOGRAPHY

[63] J. Guo, J. Xia, and B. L. Smith, “Kalman Filter Approach to Speed Estimation
Using Single Loop Detector Measurements under Congested Conditions,” Journal
of Transportation Engineering, vol. 135, no. 12, pp. 927-934, 2009.

[64] X. J. Ban, Y. Li, A. Skabardonis, and J. D. Margulici, “Performance Evaluation of

Travel-Time Estimation Methods for Real-Time Traffic Applications,” Journal of
Intelligent Transportation Systems, vol. 14, no. 2, pp. 54-67, Apr. 2010.

[65] R. Long Cheu, C. Xie, and D. H. Lee, “Probe Vehicle Population and Sample Size
for Arterial Speed Estimation,” Computer-Aided Civil and Infrastructure Engi-
neering, vol. 17, no. 1, pp. 53–60, 2002.

[66] C. Nanthawichit, T. Nakatsuji, and H. Suzuki, “Application of Probe-Vehicle Data

for Real-Time Traffic-State Estimation and Short-Term Travel-Time Prediction on
a Freeway,” Transportation Research Board Annual Meeting, vol. 1855, no. 1, pp.
49-59, Jan. 2003.

[67] M. Ferman, D. Blumenfeld, and X. Dai, “An Analytical Evaluation of a Real-Time

Traffic Information System Using Probe Vehicles,” Journal of Intelligent Trans-
portation Systems, vol. 9, no. 1, pp. 23-34, Mar. 2005.

[68] Y. Chen, L. Gao, Z. Li, and Y. Liu, “A New Method For Urban Traffic State
Estimation Based On Vehicle Tracking Algorithm,” in IEEE Intelligent Trans-
portation Systems Conference, 2007, pp. 1097–1101.

[69] R. Cayford and T. Johnson, “Operational parameters affecting the use of anony-
mous cell phone tracking for generating traffic information,” in Transportation
Research Board Annual Meeting, 2003, vol. 1, no. 3, pp. 03–3865.

[70] ITU, “Key Global Telecom Indicators for the World Telecommunication Service
Sector.” [Online]. Available:
http://www.itu.int/ITU-D/ict/statistics/at_glance/KeyTelecom.html.

[71] MobiThinking, “Global Mobile Statistics 2011.” [Online]. Available:

http://mobithinking.com/stats-corner/global-mobile-statistics-2011-all-quality-mo
bile-marketing-research-mobile-web-stats-su.

[72] Y. Yim, “The State of Cellular Probes,” 2003. [Online]. Available:

http://escholarship.org/uc/item/8g90p0vw.pdf.

[73] M. Fontaine, B. Smith, A. Hendricks, and W. Scherer, “Wireless Location Tech-

nology-Based Traffic Monitoring: Preliminary Recommendations to Transporta-
tion Agencies Based on Synthesis of Experience and Simulation Results,”
Transportation Research Record, vol. 1993, no. 1, pp. 51-58, Jan. 2007.
BIBLIOGRAPHY 81

[74] H. Bar-Gera, “Evaluation of a Cellular Phone-based System for Measurements of

Traffic Speeds and Travel Times: A Case Study from Israel,” in Transportation
Research Board Annual Meeting, 2007, vol. 15, no. 6, pp. 380-391.

[75] J. C. Herrera, D. B. Work, R. Herring, X. (Jeff) Ban, Q. Jacobson, and A. M.

Bayen, “Evaluation of Traffic Data Obtained via GPS-enabled Mobile Phones: The
Mobile Century Field Experiment,” Transportation Research Part C: Emerging
Technologies, vol. 18, no. 4, pp. 568-583, Aug. 2010.

[76] B. Hellinga, “Reducing bias in probe-based arterial link travel time estimates,”
Transportation Research Part C, vol. 10, no. 4, pp. 257-273, Aug. 2002.

[77] TransportCanada, “Development and Demonstration of a System for Using Cell

Phones as Traffic Probes.” [Online]. Available:
http://www.tc.gc.ca/eng/innovation/tdc-summary-14300-14359e-1430.htm.

[78] MobileMillennium, “The Mobile Century Field Test.” [Online]. Available:

http://traffic.berkeley.edu/project/mobilecentury.

[79] G. Rose, “Mobile Phones as Traffic Probes: Practices, Prospects and Issues,”
Transport Reviews, vol. 26, no. 3, pp. 275-291, May 2006.

[80] Z. Qiu and P. Cheng, “State of the Art and Practice: Cellular Probe Technology
Applied in Advanced Traveler Information System,” in Transportation Research
Board Annual Meeting, 2007, pp. 1-19.

[81] X. Dai, M. Ferman, and R. Roesser, “A simulation evaluation of a real-time traffic

information system using probe vehicles,” in IEEE Intelligent Transportation
Systems, 2003, vol. 1, pp. 475-480.

[82] “VISUM.” [Online]. Available: http://www.ptvamerica.com/index.php?id=1481.

[83] “Mezzo.” [Online]. Available: http://www.ctr.kth.se/mezzo.php.

[84] “VISSIM.” [Online]. Available: http://www.vissim.de/index.php?id=1801.

[85] “PARAMICS.” [Online]. Available: http://www.paramics-online.com/.

[86] “MITSIM.” [Online]. Available: http://mit.edu/its/mitsimlab.html.

[87] “PELOPS.” [Online]. Available: http://www.pelops.de/UK/index.html.

[88] J. Barceló, Fundamentals of Traffic Simulation, vol. 145. Springer, 2010.

[89] “SUMO.” [Online]. Available: http://sumo.sourceforge.net/.

[90] “CityMobil.” [Online]. Available: http://www.citymobil-project.eu/.

82 BIBLIOGRAPHY

[91] “iTetris.” [Online]. Available: http://www.ict-itetris.eu/.

[92] “TraNS.” [Online]. Available: http://lca.epfl.ch/projects/trans.

[93] “eWorld.” [Online]. Available: http://eworld.sourceforge.net/.

[94] “OpenStreetMap.” [Online]. Available: http://www.openstreetmap.org/.

[95] J.-luc Ygnace, C. Drane, Y. B. Yim, and R. D. Lacvivier, “Travel Time Estimation
on the San Francisco Bay Area Network Using Cellular Phones as Probes,” 2000.
[Online]. Available: http://escholarship.org/uc/item/8xn8m01v.

[96] K. K. Srinivasan and P. P. Jovanis, “Determination of Number of Probe Vehicles

Required for Reliable Travel Time Measurement in Urban Network,” Transporta-
tion Research Record: Journal of the Transportation Research Board, vol. 1537,
pp. 15-22, 1996.

[97] M. Chen and S. Chien, “Determining the Number of Probe Vehicles for Freeway
Travel Time Estimation Using Microscopic Simulation By Determining the Num-
ber of Probe Vehicles for Freeway Travel Time Estimation Using Microscopic
Simulation,” Transportation Research Record: Journal of the Transportation Re-
search Board, vol. 1719, pp. 61-68, 2000.

[98] S. M. Turner and D. J. Holdener, “Probe vehicle sample sizes for real-time infor-
mation: the Houston experience,” in Vehicle Navigation and Information Systems
Conference, 1995, pp. 3-10.

[99] M. Fontaine and B. L. Smith, “Probe-Based Traffic Monitoring Systems with

Wireless Location Technology: An Investigation of the Relationship Between
System Design and Effectiveness,” Transportation Research Record: Journal of
the Transportation Research Board, pp. 3-11, 2005.

[100] J. Yoon, B. Noble, and M. Liu, “Surface Street Traffic Estimation,” in Interna-
tional conference on Mobile systems, applications and services, 2007, pp. 220-232.

[101] W. Shi and Y. Liu, “Real-time Urban Traffic Monitoring with Global Positioning
System-equipped Vehicles,” IET Intelligent Transport Systems, vol. 4, no. 2, p.
113, 2010.

[102] “World Geodetic System 1984.” [Online]. Available:

http://en.wikipedia.org/wiki/World_Geodetic_System.

[103] “SUMOPlayer.” [Online]. Available:

http://sourceforge.net/apps/mediawiki/sumo/index.php?title=SUMOPlayer.
BIBLIOGRAPHY 83

[104] F. Gustafsson and F. Gunnarsson, “Mobile Positioning Using Wireless Networks,”

IEEE Signal Processing Magazine, vol. 4, no. 2005, pp. 41-53, 2005.

[105] Y. Yim and R. Cayford, “Investigation of Vehicles as Probes Using Global Posi-
tioning System and Cellular Phone Tracking: Field Operational Test,” 2001.
[Online]. Available: http://escholarship.org/uc/item/0378c1wc.pdf.

[106] “Universal Transverse Mercator coordinate system.” [Online]. Available:

http://en.wikipedia.org/wiki/Universal_Transverse_Mercator_coordinate_system.

[107] N. Caceres, J. Wideberg, and F. Benitez, “Review of traffic data estimations

extracted from cellular networks,” IET Intelligent Transport Systems, vol. 2, no. 3,
pp. 179–192, 2008.

[108] “Driving and Speed Limits in Stockholm.” [Online]. Available:

http://carrentalscout.com/driving-speed-limits-stockholm.

[109] D. Bernstein and A. Kornhauser, “An Introduction to Map Matching for Personal
Navigation Assistants,” 1996.

[110] M. Westerman, R. Litjens, and J. P. Linnartz, “Integration Of Probe Vehicle And

Induction Loop Data: Estimation Of Travel Times And Automatic Incident De-
tection,” 1996. [Online]. Available: http://escholarship.org/uc/item/8mh629c3.pdf.

[111] “Java OSM Editor.” [Online]. Available: http://josm.openstreetmap.de/.

[112] D. J. Lovell, “Accuracy of Speed Measurements from Cellular Phone Vehicle

Location Systems,” Journal of Intelligent Transportation Systems, vol. 6, no. 4, pp.
303–325, 2001.

[113] M. A. Quddus, W. Y. Ochieng, and R. B. Noland, “Current Map-matching Algo-

rithms for Transport Applications: State-of-the art and Future Research Direc-
tions,” Transportation Research Part C, vol. 15, no. 5, pp. 312–328, 2007.

[114] P. Lovisolo, D. Parata, and C. Ratti, “Real-Time Urban Monitoring Using Cellular
Phones: a Case-Study in Rome,” 2007.