Autonomous Vehicles-Intelligent Transport Systems and Smart Technologies PDF

ENGINEERING TOOLS, TECHNIQUES AND TABLES
AUTONOMOUS VEHICLES
INTELLIGENT TRANSPORT SYSTEMS
AND SMART TECHNOLOGIES
No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.
ENGINEERING TOOLS, TECHNIQUES
AND TABLES
Additional books in this series can be found on Nova‟s website

under the Series tab.
Additional e-books in this series can be found on Nova‟s website

under the e-book tab.
ENGINEERING TOOLS, TECHNIQUES AND TABLES
AUTONOMOUS VEHICLES
INTELLIGENT TRANSPORT SYSTEMS
AND SMART TECHNOLOGIES
NICU BIZON
LUCIAN DASCALESCU
AND
NASER MAHDAVI TABATABAEI
EDITORS
New York
Copyright © 2014 by Nova Science Publishers, Inc.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.
For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com
NOTICE TO THE READER

The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or
implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.
Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.
This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.
Additional color graphics may be available in the e-book version of this book.
Library of Congress Cataloging-in-Publication Data
Autonomous vehicles : intelligent transport systems and smart technologies / editors, Nicu Bizon,
Lucian Dascalescu, and Naser Mahdavi Tabatabaei (Faculty of Electronics, Communication and
Computers, University of Pitesti, Romania).
pages cm
Includes index.
ISBN: (eBook)
1. Autonomous vehicles. I. Bizon, Nicu. II. Dascalescu, Lucian. III. Mahdavi Tabatabaei, Naser.
TL152.8.A88 2014
388.3'12--dc23
2014022215
Published by Nova Science Publishers, Inc. † New York

Dedicated to
all our teachers and colleagues
who enabled us to write this book,
and our family and friends
for supporting us all along.
CONTENTS
Foreword ix
Preface xi
Part I. Intelligent Transport Systems 1
Chapter 1 Radar Sensor Detectors for Vehicle Safety Systems 3
Modar Safir Shbat and Vyacheslav Tuzlukov
Chapter 2 Airborne Ultrasonic Imaging: SONAR Based Image
Generation for Autonomous Vehicles 57
Dorel Aiordachioaie, Laurentiu Frangu, Silviu Epure,
Radu Belea, Marius Mazarel, Constantin Miholca
and Claudiu Chiculita
Chapter 3 Autonomous Underwater Vehicle Dynamics 81
Mustafa Dinc and Chingiz Hajiyev
Chapter 4 Advances in Environment Sensing and Perception Technologies
and Algorithms for Autonomous Ground Vehicles 113
Constantin Ilas, Irina Mocanu and Mariana Ilas
Chapter 5 Intelligent Highway Surveillance and Safety Systems 147
Valentin Alexandru Stan, Radu Serban Timnea
and Razvan Andrei Gheorghiu
Chapter 6 Cyber Security Evaluation of Critical Infrastructures Systems 185
Alecsandru Patrascu and Emil Simion
Chapter 7 Sustainable Information Systems Management
for Spatial Governance 207
Angela Ionita, Marcel Foca and Maria Visan
Chapter 8 Standards of Communications in the Intelligent
Transport Systems (ITS) 235
Maaruf Ali
viii Contents
Part II. Smart Technologies Applied in the Electrical and Hybrid Vehicles 247
Chapter 9 Power Electronics and Drives Used in Automotive Applications 249
Ersan Kabalci
Chapter 10 Power Control of Plug-In Electric Vehicles in Smart Grids 275
Amin Hajizadeh, Amir Hossein Shahirinia
and David C. Yu
Chapter 11 Optimal Short-Term Scheduling of Aggregators
for Plug-In Electric Vehicles 293
Mohammad Moradi-Dalvand
and Behnam Mohammadi-Ivatloo
Chapter 12 Energy Efficiency of the Hybrid Power Source Used
in the Plug-In Fuel Cell Vehicles 313
Nicu Bizon, Marian Raducu, Mihai Oproescu
and Luminita Mirela Constantinescu
Chapter 13 Dynamic Response of Electric Machines for Electric
Vehicles/Hybrid Electric Vehicles (EV/HEV) 353
Robert Beloiu, Mariana Iorgulescu and Corina Savulescu
Chapter 14 Use of the External Magnetic Field for Supervision
and Diagnosis of Electrical Traction Motors 379
Remus Pusca, Raphael Romary and Andrian Ceban
Chapter 15 Batteries for Electrical Vehicles: A Review 409
Bogdan Enache, Emilian Lefter and Costin Cepisca
Chapter 16 Computational Study of Li-Ion Batteries for
Automotive Applications 431
Inigo Aramendia, Unai Fernandez Gamiz,
Jose Antonio Ramos, Javier Sancho
and Ekaitz Zulueta
Chapter 17 Ni-Cd Batteries for Automotive Applications: The ac Electrical
Data of the Ni-Cd Battery in the Havriliak-Negami
Dielectric Representation 453
Mohammad A. Alim, Sudip Bhattacharjee, Sanjida Khanam,
Nasiha Muna and Sabiha Runa
Chapter 18 Advanced Drive Architectures and State of the
Art In-Wheel Motors 485
Blaz Stefe and Gorazd Lampic
Chapter 19 The Future of Electrical Vehicles 509
Wim J. C. Melis
Index 531
FOREWORD
The Electric Vehicles (EVs) use electric power source for transportation capabilities.
Currently electric cars are feasible alternative rather than to the conventional gasoline
powered cars because they are significantly less complex, more reliable/efficient, and
environmentally friendly, and cost less to operate.The success of extensive electric vehicle
carrying depends on the reliability of both the electric grid and the charging equipment. In car
accidents, EVs are considered with standard impact a lot better than most gasoline-powered
vehicles.
The fossil fuel powered vehicles are generally non-friendly regarding to the
environmental aspects. In the past, the vehicles with lower consumption have been presented.
After that, the research direct towards more environmentally friendly vehicles which hybrid
and electrical vehicles have been developed. The hybrid electric vehicles (HEVs) and
developed plug-in hybrid vehicles (PHEVs) are improved for the integration of electrical
vehicles which should allow for the vehicle to be charged by green power.
Electric Vehicle components include electric motor, battery pack, battery charger, motor
controller, contactor, fuse and circuit breaker, vacuum pump, power steering pump, DC/DC
converter, and the instrumentation which each of them has individual technology to obtain the
EV efficiency improvement. Also, many EVs have been specialized with special features
including resealing tires, programmable charging systems, and regenerative breaking, low
maintenance and fairly cost efficient vehicles, with long life batteries.
The current book concerns in development of the advanced autonomous electric vehicles
applied in intelligent transport systems and using the smart technologies. Although electric
tractions are not new subjects, but their application to electrical automotive vehicles and
technological developments for responding to the modern life requirements are discussed.
The above and also advanced aspects related to electric vehicles including the safety
systems, plug-in electric vehicles, autonomous ground and underwater vehicles, intelligent
transport systems, automotive applications, smart grids, hybrid power source, plug-in fuel cell
vehicles, electrical traction motors, and in-wheel motors are illustrated in this book by the
editors and authors, in the following topics: radar sensor detectors; sonar based image
generation; dynamics analysis and responses; advanced sensing technologies and algorithms;
intelligent highway surveillance; cyber security evaluation of critical infrastructures systems;
sustainable information systems management for spatial governance; standards of
communications; power electronics and drives; power control of smart grid; optimal short-
term scheduling of aggregators; energy efficiency of the multiport power converters; external
x Arif M. Hashimov
magnetic field for supervision and diagnosis; batteries for electrical vehicles; computational
study of Li-ion and Ni-Cd batteries; advanced drive architectures; and the future of electrical
vehicles.
The book chapters and materials are very efficient in theoretical and application issues
and are highly recommended for studying and considering in educational and research fields.
February 15, 2014
Academician Arif M. HASHIMOV

Institute of Physics
Azerbaijan National Academy of Sciences
Baku, Azerbaijan
PREFACE
Environmentally friendly, high efficiency, green power charging, renewable sources

storage, low maintenance cost, and hybrid integration of electrical vehicles (EVs) over current
fossil fuel powered vehicles are generally considered for excellent characteristic features and
solutions in transportation development and marketing. The electric vehicles with smart
metering, intelligent control and communication interfaces are developed as environmentally
friendly cars which use the energy efficiently and minimize the pollution problems caused
by burning of the fossil fueled automobiles.
The topics of autonomous ground vehicles, fuel-cell electric vehicles, battery-electric
vehicles, automated driver assistance systems, transportation interface equipment, traffic
management system, infrastructure elements development, traveller information system,
sustainable information systems management, and inter-vehicular communications have
grown during the expanding of the electric vehicle systems.
The autonomous vehicles as the intelligent transport systems have recently become more
attractive options over combustion engine based vehicles, due to the high fuel price and
environmental concerns. The electric vehicles act as vehicle to grid (V2G) systems, which
discharge their batteries and send electricity back to the power grid when they are parked and
connected to an electrical outlet.
The plug-in electric vehicles (PEVs) are defined as electric vehicle conversions of hybrid
electric vehicles (HEVs) and conventional internal combustion engine vehicles which have
battery configuration to obtain the required recharged energy for propulsion over the electric
grid, the conventional internal combustion engine and regenerative braking. It is essential that
the hybrid electric vehicles cannot be recharged from any off-vehicle electric energy sources
that are not categorized in plug-in electric vehicles.
Plug-in hybrid electric vehicles (PHEVs) are more flexible loads comparing to traditional
transportation systems to increase the power system loading which the large amount of
charging power will take place on peak hours, because of stopping at evening and charging in
the morning.
The successful operations of extensive electric vehicles depend on the consistency of the
electric grid and the charging equipment as well as the cost of the batteries which imply the
robust and reliable charger, high power efficiency and minimum impact on the power quality
of smart grid distributions systems. The electric vehicles act as regulated consumers and
production units as well as the distributed sources to restore the variable energies coming
from distributed generation systems.
xii Nicu Bizon, Lucian Dascalescu and Naser Mahdavi Tabatabaei
The hybrid power sources (HPS) combine energy storage devices (batteries,
ultracapacitors, etc.) in a hybrid stack (passive, semi-active or active controlled mode) to be
used in hybrid electric vehicles (HEVs) and fuel cell vehicles (FCVs) to manage the power
flow between the energy sources, and the storage devices and loads based on an energy
management unit.
One of the essential parts of the EV/HEV is electric drives in a variety of configurations
with more specific characteristics with sufficient reliability and autonomy as well as including
faster control of drive torque and improved dynamic performance.
Radar sensor technologies and ultrasound imaging are including high information and
accurate measurements from the target vehicle and speed recognition system as well as
weather conditioning which are widely used in electric vehicle transportation systems. The
technologies are very useful in widespread range of marine systems and unmanned
underwater vehicles.
Batteries as the chemical-electrical energy converters through electrochemical discharge
reactions should be designed to reduce the possibility of any leakage problem. Two types of
available batteries including the primary type discharged after single usage used as the
automotive starting battery, and the secondary type such as Hg-Zn, Hg-Cd and Li-Cd
batteries are designed for frequently usage which are rechargeable by reversing the chemical
reactions used as power of the microelectronic devices sealed to not allow venting of gas to
the atmosphere.
The certain problems related to the vehicle‟s batteries are the long recharging period,
reduced reliability, a limited autonomy, high cost price, high energy density, long life,
rechargeable capability, safe handling, and flexibility in applications which should be noted in
the research and industry phases.
Autonomous Vehicles: Intelligent Transport Systems and Smart Technologies is a book
aimed to highlight the above concepts and challenges by a systematic approach and,
therefore, to provide understanding on how the electrical vehicles influence the modern
transportation systems, and the ways of controlling grounded and underwater navigation
systems by using different research methods.
A large number of specialists are joined as authors of the book chapters to provide their
potentially innovative solutions and research related to advanced researches in the electric
vehicles systems, in order to be useful in developing new ways in transportation systems,
design and operational strategies. Several theoretical researches, case analysis, and practical
implementation processes are put together in this book that aims to act research and design
guides to help the postgraduates and researchers in electrical power engineering and energy
systems. In particular, the specific purposes of the subjects are also including as knowledge
base and applications used by undergraduates regarding to indicating research fields and
innovative solutions as the challenges and opportunities for solving the problems.
The book presents significant results obtained by leading professionals from the industry,
research and academic field, can be useful to the variety of groups in specific areas analyzed
in this book. All works contributed to this book are new, previously-unpublished material or
extended version of published papers in the proceedings of international conferences and
transactions on international journals.
The nineteen chapters present the advanced theories and applications in electrical
vehicles systems and their peripherals subsystems and hybrid conditions considering different
operating cases. The topics in these chapters are including the advanced methods and
Preface xiii
applications in intelligent electrical vehicles and automotive applications in smart grids

related to the fields of protection systems, plug-in electric vehicles, autonomous ground and
underwater vehicles, hybrid power source, fuel cell vehicles, electrical traction and in-wheel
motors.
Chapter 1 introduces the signal detection performance evaluation of detectors employed
by radar sensor systems used in vehicle safety structures and discusses about the generalized
approach to signal proceeding and noise detection performance such as use of alarm rate and
generalized detectors. The chapter includes some important objects related to definition
method of the target vehicle parameters. Besides, the chapter introduces middle range radar
and short range radar applications.
The applications of ultrasonic waves image diagnosing is given in Chapter 2. The various
and important applications in industry, for fault detection and localization, and medicine, for
non-invasive diagnostic techniques, are discussed and compared with each other. The
algorithms in the chapter are related to ultrasonic images and sonar image generation and
processing which the airborne ultrasound images and raw processing methods for navigation
purposes are investigated and interpreted.
Chapter 3 presents detailed explanations on the theory and design of the dynamic
mathematical modeling of autonomous underwater vehicles (AUVs). The chapter also
investigates the simulation model and also implement control and navigation algorithm for
the underwater vehicles. The parameter identification of hydrodynamic coefficients based on
a least square estimation algorithm for the nonlinear mathematical modeling of autonomous
underwater vehicles is also studied in the chapter.
The environment sensing and perception technologies and algorithms used in
autonomous ground vehicles (AGVs) are described in Chapter 4. The environmental sensing
technologies are also presented in the chapter, focusing mostly on LIDAR, video camera and
radar. The chapter shows the increased usage of video cameras, and video camera plus
LIDAR or video camera plus radar employed in AGVs. The chapter also shows the
implementation of the SVM algorithm based on video camera images and compares the
performance with HOG and DSIFT.
The concept and the implementation solution for the integrated highway traffic
surveillance system with complex safety integrated systems are introduced in Chapter 5 in
order to increase safety and security on European highway. The application of traffic
surveillance, traffic re-routing, emergency services announcement, driver informing on
modern highway traffic conditions are introduced as well. The chapter also discusses about
the highway traffic issues related to traffic video surveillance, automatic incident detection,
automatic adaptation of traffic signalization, traffic statistics, speed detection, automatic car
identification, and law enforcement system.
Chapter 6 is entirely highlighted on the issues of cyber security, criminals and attacks in
cyber systems. The chapter denotes to the attack types such as social engineering, port
scanning, packet sniffing, password cracking, denial of the service attacks, machine bugs, and
server bugs. Then the cyber threats on SCADA based critical infrastructure like
transportations systems, automotive factories, telecommunication, water supply and
electricity generation are also discussed. The chapter also covers some attack scenarios
explained in detail for in-depth understanding of the concept involved in cyber security.
The complexities of sustainable information systems management in the framework of
spatial governance are focused in Chapter 7 in the areas of sustainability based on information
xiv Nicu Bizon, Lucian Dascalescu and Naser Mahdavi Tabatabaei
technology and communication (ICT). The chapter shows a reorientation proposal of the
sustainability for positioning spatial governance based on human behavior and the processing
of spatial knowledge. The chapter also describes the requirements for a sustainable
information systems management in spatial governance focusing on intelligent transportation
systems.
Chapters 8 provides the current standards for communication in intelligent transportation
systems (ITS), particularly related to inter-vehicular statement. The chapter presents an
overview of the IEEE protocol standard relevant to ITS communication. The latest
standardizations in the dedicated short range communication, and the IEEE wireless access in
vehicular environments also conferred in the chapter. The infrastructure based and centralized
cellular networks are also demonstrated.
Chapter 9 presents power electronics and drives used in electric vehicles. The chapter is
also describes electric motors selecting views based on construction structure, efficiency,
maintenance requirements, electromagnetic interference, reliability, and speed range
parameters. The power electronics devices and motor drives applications used in electric
vehicles systems are also described.
In Chapter 10, the plug-in electric vehicles (PEVs) with dual energy storage sources are
presented. The chapter shows PEVs dynamic state space modeling containing combined
battery systems with supercapacitors. The Lyapunov based Neuro-Fuzzy control strategy to
stabilize the DC-link power and distribute the power between dual energy storage sources in
plug-in electric vehicles system are analyzed in the chapter. The robust sliding mode control
strategy for voltage source converter based on positive and negative symmetrical components
are also shown.
Chapter 11 deals with market prices modeling for plug-in hybrid electric vehicles
(PHEVs) using stochastic programming technique. The chapter introduces objective of the
PHEV aggregator to maximize its profit by charging the PHEVs on the low price hours. The
chapter also presents electricity market and PHEVs characteristics constraints incorporating
into the proposed method considering an illustrative example.
A mathematical analysis of the energy efficiency for the series and parallel multiport
power converter (MPC) topologies is proposed in Chapter 12 to be used in the plug-in fuel
cell vehicles (PFCVs). The chapter also defines the electrical motors and inverter systems,
and the DC input sources parameters for electrical vehicles, and the strategies for the energy
management unit. The other aspect related to choosing the optimal parameters of the MPC
control for an efficient use of the PFCV is also introduced. The operation regimes of the
energy storage system (ESS) including charge-sustaining, charge-depleting, and charge-
increasing as well as the MPC energy efficiency of the PFCVs in the ESS
charging/discharging period from/to the smart grid have been studied.
Chapter 13 provides the dynamic response and the behavior of three types of electric
motors for using in the electric vehicle/hybrid electric vehicle (EV/HEV) implemented by
different car manufacturers. The chapter shows a testing regime based on the proposed drivers
considering all the encountered functional situations. The chapter illustrates the behavior of
the rapid and slow start-stop regime, and speed change regime of the DC, induction and
brushless machines used in EV/HEV.
The Chapter 14 aims to provide a supervision method for detection of AC electrical
machine faults, used in electrical transportation, considering the external magnetic field
information nearby the machine may cause the failure responses. The chapter studies two
Preface xv
methods based on the analysis of the magnetic field outside the machine which are reliable,
original, inexpensive and simple to implement in electrical automotive vehicle. The
theoretical analysis and experimental validation are resulted regarding to the low frequency
components of the magnetic field which is not influenced by the presence of electronic
converters.
The Chapter 15 introduces a review of electrical vehicles batteries as an important factor
in development of the automotive industries. The chapter also presents new developments and
solutions for creation of batteries applicable to experimental research. The chapter
emphasizes that the environmental care completion regarding to reduce the automobiles
impact on the environment and also on people‟s health which is carried out by accelerating
development process of batteries for the electric and hybrid vehicles.
Chapter 16 discusses about designing of a lithium ion cells batteries with the
performances of consisting an optimum cooling and heat transfer evaluation along the testing
time. The chapter analyzes the battery model based on cold plates based on the geometry and
mesh of the elements. The physical properties and initial conditions as well as the
convergence of waste products, air speed, cells temperature and heat transfer are also
discussed in the chapter.
In Chapter 17 the secondary type of the batteries are discussed with rechargeable
performance by reversing the chemical reactions. The chapter shows the AC small signal
electrical data of the Ni-Cd battery in the impedance plane and discuss about the various
analytical aspects using the Davidson-Cole (D-C) representation. The results of the chapter
also show that the possible meaning of the depression parameters extracted from the breadth
and skewness as well as deviation of the semicircular Debye or Cole-Cole (C-C) dielectric
relaxation loci.
The major mechanical design possibilities of the in-wheel electric motors, including
direct drive motors and gears are studied in Chapter 18. It is noted that in-wheel motors
increase the total unsprung mass of the vehicle, and therefore the theoretical model of the
motors is analyzed in the chapter. The chapter also discusses about the designing topics of in-
wheel motors for different kinds of vehicles considering different mechanical constructions.
Chapter 19 focuses on the future prospective of the electrical vehicles and their
efficiency, comparing with combustion engine (CE) vehicles. The chapter also provides some
alternative fuels and storage mechanisms for improving the electrical vehicles performance.
The use of batteries as well as alternative fuels such as fossil fuels and fuel cells are studied in
the chapter regarding to the environmental impact and also energy storage mechanisms.
The editors recommend book as suitable for dedicated and general audiences that
includes the power system professionals, as well as researchers and developers of energy
sectors and electrical power engineering community. It is expected that readers to be
graduates of energy and power engineering degree programs having a basic mathematical
background.
As a whole, the book covers both theoretical background and application examples in
transportation systems, and special and professional fields of electric vehicle systems in
altogether. Because the subjects‟ area approached in this book is vast in itself, it has been a
little bit hard to balance the theoretical and applicative aspects in each of the chapters, so
efforts of editors have been made to well cover the essential topics of the book. Specific in-
depth further studies are pointed to the dedicated intensive resources of the book subjects for
xvi Nicu Bizon, Lucian Dascalescu and Naser Mahdavi Tabatabaei
interested readers. Meanwhile, the application and study cases are possibly selected with as
much real implications.
Finally, the editors hope that this book will be helpful to undergraduate and graduate
students, researchers and engineers, trying to solve energy and power engineering problems
using modern technical and intelligent systems based on theoretical aspects and application
case studies.
Pitesti, Romania Nicu Bizon

Professor
Faculty of Electronics
Communication and Computers
University of Pitesti
Arges, Romania
E-mails: nicubizon@yahoo.com,
nicu.bizon@upit.ro
Poitiers, France Lucian Dascalescu

Professor
PPRIME Institute, UPR 3346 CNRS
University of Poitiers-ENSMA, France
E-mail: lucian.dascalescu@univ-poitiers.fr
Tabriz, Iran Naser Mahdavi Tabatabaei

Professor
Electrical Engineering Department, Seraj
Higher Education Institute
Tabriz, Iran
E-mail: n.m.tabatabaei @gmail.com
February 20, 2014 Editors

PART I. INTELLIGENT TRANSPORT SYSTEMS
In: Autonomous Vehicles ISBN: 978-1-63321-324-1
Editors: N.Bizon, L. Dascalescu and N.M. Tabatabaei © 2014 Nova Science Publishers, Inc.
Chapter 1
RADAR SENSOR DETECTORS FOR VEHICLE

SAFETY SYSTEMS
Modar Safir Shbat and Vyacheslav Tuzlukov†

School of Electronics Engineering,
College of IT Engineering,
Kyungpook National University,
Sankyuk-dong, Buk-gu, Daegu, South Korea
ABSTRACT
This chapter is devoted to a target return signal detection performance evaluation of
detectors employed by radar sensor systems used in vehicle safety systems, for example,
safety driving applications. Detection performance and characteristics of several types of
the constant false alarm rate (CFAR) detectors and the generalized detector (GD)
constructed based on the generalized approach to signal proceeding in noise are
compared under the same conditions for the cases of presence and absence of interfering
vehicles. The method of definition of the required target vehicle parameters for the
middle range radar (MRR) and short range radar (SRR) applications is also discussed.
Keywords: Radar Sensor System, Vehicle Safety System, Constant False Alarm Rate
(CFAR) Detector, Generalized Detector (GD), Noise Power Estimation, Multilateration,
Attenuation Factors
NOMENCLATURES
ACC Adaptive Cruise Control
ACGO-CFAR Adaptive Censored Greatest of CFAR
AF Additional Filter

Email: modboss80@knu.ac.kr.
†
Email: tuzlukov@ee.knu.ac.kr.
4 Modar Safir Shbat and Vyacheslav Tuzlukov
AWGN Additive White Gaussian Noise

BSD Blind Spot Detection
CA-CFAR Cell Averaging CFAR
CFAR Constant False Alarm Rate
CVD Closing Vehicle Detection
DB Decision Block
FFT Fast Fourier Transform
FMCW Frequency Modulation Continuous Wave
FSK Frequency Shift Keying
GCML Generalized Censored Mean Level
GD Generalized Detector
GO-CFAR Greatest of CFAR
i.i.d. Independent and Identically Distributed
LFMCW Linear FMCW
LRR Long Range Radar
MRR Middle Range Radar
MSG Model Signal Generator
NP detector Neyman-Pearson detector
OS-CFAR Ordered Statistic CFAR
pdf Probability Density Function
PF Preliminary Filter
RCS Radar Cross Section
RLC Resistor, Inductor, and Capacitor oscillatory circuit
SGSA Signal Model Generator Switching Apparatus
SNR Signal-to-Noise Ratio
SO-CFAR Smallest of CFAR
SRR Short Range Radar
THRA Threshold Apparatus
UWB Ultra Wide Band
1. INTRODUCTION
Recent investigations discussed in [1] demonstrate a superiority of implementation of
radar sensor systems for vehicle collision avoidance and safety driving applications in
comparison with other technologies. Radar sensor systems offer a high information content
providing accurate measurements of the target vehicle range and relative velocity in addition
to ability to operate under any weather conditions and harsh environments.
In this chapter, we evaluate the signal detection and signal processing algorithms used in
radar sensor system for vehicle collision avoidance and safety driving applications.
Performance metrics and simulation conditions are defined with the purpose to obtain an
impartial comparison of different detectors. Under comparison we define the detection
performance of the target return signal as the main characteristic.
Detection performance of the target return signal is the function of the signal-to-noise
ratio (SNR) and detection threshold and, consequently, the noise power sensitivity. Detection
Radar Sensor Detectors for Vehicle Safety Systems 5
threshold should be defined based on the observed in real time noise power if the noise
variance is varied. Thus, the adaptive detection threshold is very important parameter to
maintain the constant probability of false alarm PFA . For any radar sensor system with high
duty cycle, for example, the linear frequency modulation continuous wave (LFMCW) radar
sensor system, a sensitivity of the noise power is an important problem. The constant
probability of false alarm PFA is a desirable requirement for most radar sensors applications.
By this reason, there is a need to define the detection threshold of target return signal based
on the noise power or noise variance estimation if the noise power or variance is varied as a
function of time. The noise power estimation is a vital factor allowing us to define adaptively
the detection threshold in radar sensor systems and controller area networks used to design
the radar sensor systems for safety driving applications, collision avoidance systems, and
vehicle tracking systems.
The appropriate signal detection and signal processing algorithms for radar sensor
systems employed in numerous vehicle safety and collision avoidance applications are
investigated for two different schemes. The first scheme is the constant false alarm rate
(CFAR) detector, and the second scheme is the generalized detector (GD) constructed based
on the generalized approach to signal processing in noise. GD is a combination of the
correlation detector that is optimal in the Neyman-Pearson criterion sense (NP detector) for
detection of signals with a priori known parameters and the energy detector that is optimal in
the Neyman-Pearson criterion sense under detection of signals with a priori unknown
parameters. This combination allows us to formulate a decision-making rule about the
presence or absence of target vehicle based on definition of the jointly sufficient statistics of
the likelihood function mean and variance with complete compensation of the noise
component that is specific both for the NP detector and the energy one. The target return
signal processing technique is introduced to define the required target vehicle parameters,
namely, the range, relative velocity, and position or coordinates in order to distinguish the
threaten and non-threaten detected targets.
Several CFAR detectors are discussed, namely, the cell averaging CFAR (CA-CFAR)
detector, ordered statistic CFAR (OS-CFAR) detector, generalized censored mean level
(GCML) detector, and adaptive censored greatest of CFAR (ACGO-CFAR) detector. All
these detectors can be employed in the linear frequency modulation continuous wave
(LFMCW) radar sensor system, which is widely used by the vehicle collision avoidance and
safety driving systems for the middle range radar (MRR) and short range radar (SRR)
applications such as the closing vehicle detection (CVD) and blind spot detection (BSD).
All the mentioned CFAR detectors and GD are compared by detection performance of
the target return signal following the proposed framework and under the same initial
conditions. We consider two main cases. The first case – there are no any interfering target
vehicles. The second case – there are one or more interfering target vehicles. Based on the
theoretical analysis, a simulation is carried out at the conditions closed to practice that allows
us to evaluate the above-mentioned detectors employed by LFMCW radar sensor system in
the case when the interfering targets are present or not.
The implementation of GD in radar sensor systems used for vehicle safety applications
demonstrates the better detection performance in comparison with the CFAR detectors at the
same conditions and promises the high probability of detection. This result is confirmed by
implementation of the noise power estimation for GD based on sliding window procedure
with the same number of reference cells used by CFAR detectors in order to perform the fair
comparison.
2. INTELLIGENT VEHICLE AND TOTAL SAFETY DRIVING APPROACH

A lot of technologies are developed by researches and companies as a consequence of
great growth of automotive industry. Until recent time, the automotive industry has been
dominated by mechanical engineering. Nowadays this industry needs the efforts to signal
processing, automatic control, electronic engineering, communication systems, and computer
hardware and software in order to construct and develop the modern vehicle intelligent
systems, for example, the self-driving and self-parking, vehicle to vehicle communications,
navigation systems, and safety driving applications.
The theoretical, experimental, and operational aspects of electronic engineering and
information technologies are integrated to enrich the vehicle by required ability to drive
safety, avoid accidents, and protect the people inside and outside the vehicle form serious
injuries in the case of accident occurrence [2]. The total safety driving approach means that
the vehicle should be equipped by specific systems to avoid accidents and prevent injuries.
This approach can be achieved by installing different kinds of technologies such as the
cameras with image processing, navigation and information systems, and sensors, for
example, the laser, ultra-sound, infrared, and radar sensors to form a network of active and
passive safety driving applications.
The safety driving applications based on radar sensors can be categorized by three main
parameters: the range, sensor type, and location of radar sensor system around the vehicle
(frond, side, and rear). Radar sensors offer high information content and provide good ranging
ability and relative velocity measurement in addition to other advantages, namely, the
operation under any weather conditions and harsh environment (dust, dirt, light, rain, snow,
and so on). Radar sensor systems are classified into three types based on the range, namely,
the long range radar (LRR), MRR, and SRR [3]. Many safety driving applications are
developed using the radar sensor systems, for example, the adaptive cruise control, BSD,
front and rear collision warning and avoidance, parking aid, lane change assistant and CVD
[4-6]. Figure 1 presents many examples of safety driving applications classified according to
the range and radar sensor location.
Figure 1. Safety driving applications.

Radar sensor systems operate typically in pulse (pulse radar sensor) or continuous wave
mode, namely, the LFMCW radar sensor, the frequency shift keying (FSK) radar sensor, and
the combined LFMCW and FSK radar sensor systems [7]. The coded radar sensor, spread
spectrum radar sensor, and random noise radar sensor systems are not widely used in safety
driving applications. Table 1 present the most popular types of radar sensor systems
employed by various safety driving applications at 24 and 77 GHz operation frequencies [8].
From Table 1, it can be noticeable that LFMCW, FSK, and pulsed radar sensor systems
are the most used technologies implemented in various safety driving applications. The
LFMCW radar sensor system is considered as the best decision, especially for the MRR
applications, such as CVD [9], and SRR applications, such as BSD, owing to comparison of
the main features of the radar sensor systems [1] presented in Table 2.
Table 1. Typical radar sensor systems used in safety driving applications
Range/Velocity/View Sensor Radar Sensor Operation

Application
Requirements Category Type Frequency
0.2~5 m
Parking Aid 0~30 Km/h SRR UWB/Pulsed 24 GHz
Full vehicle width
0.5~10 m LFMCW/
SRR
BSD Reasonable velocity interval Pulsed/ 24 GHz
Two side regions FSK/
1~150 m LFMCW/
Adaptive Cruise LRR
Reasonable velocity Pulsed/ 77 GHz
Control
3 lanes in front of the vehicle FSK/
0.5~150 m/ 0.5~40 m LFMCW/ 77 GHz
LRR/MRR
Stop & Go System Reasonable velocity Pulsed/ 24 GHz
3 lanes in front of the vehicle FSK/
0.5~30 m
CVD Any velocity SRR/MRR LFMCW/FSK 24 GHz
About 45 degree
0.5~150 m/ 0.5~40 m 77 GHz
Collision LRR/MRR LFMCW/FSK
Any velocity 24 GHz
Avoidance
3 lanes in front of the vehicle
Pre-Crash 0.5~10 m/ 0.5~30 m
Reversible Any velocity SRR/MRR LFMCW/FSK 24 GHz
Restraints Full vehicle width
Table 2. Comparison of the main features of radar sensor systems
Criteria Radar Sensor System

Pulsed FSK LFMCW
Range Resolution Good Average Good
Measuring Ability Average Good Average
Fixed Obstacle Detection Good Poor Good
Robustness to Jamming Poor Good Good
Under design of any safety driving systems, many important practical issues related to the
radar sensor network design requirements and controller area network should be taken into
consideration. These practical issues can be summarized as follows:
 The time synchronization of radar sensor networks that is important for target vehicle
state estimation. In many cases, the data transmission rate and delay are critical
factors, especially, for SRR systems.
 The interface for communication between the distributed radar sensor networks.
 The position or location of every radar sensor. For example, the radar sensor location
behind the vehicle bumper affects the performance and should be defined very
accurately to guarantee a precise determination of azimuth angle estimation.
 The possible cross talk and undesired microwave propagation behind the vehicle
bumper must be avoided.
 The computation complexity is increased in the case of radar sensor networks. All
radar sensor signals have to be evaluated and the data association and fusion have to
be performed and examined.
 The optimal allocation of digital signal processing blocks or units in the radar sensor
network is an important problem.
 The structural complexity should be as low as possible to reduce the average number
of failures and the cost.
 The number and size of components have to be small to solve the problem of
integration space.
 The radar sensor systems must possess the same quality; otherwise, mismatches
under signal detection and signal processing should be taken into consideration.
More information and details concerning about these practical aspects and the design of
radar sensor networks can be found in [10]. Appropriate signal detection and signal
processing procedures for FMCW radar sensors are used to design the MRR and SRR safety
driving systems.
3. OVERVIEW OF MRR AND SRR SYSTEMS

There are the following main parameters of the target vehicle, namely, the target vehicle
range R , the relative velocity between the subject vehicle and target vehicle Vr , and the target
vehicle position defined by the coordinates. These parameters are required to be used by
safety driving systems and can be obtained under the signal detection and signal processing
procedures as shown in Figure 2.
The frequency modulated signal, i.e. the transmitted signal, is reflected from the target
vehicles. The target return signal is multiplied by the transmitted signal copy (the transmitted
signal model) at the receiver. The difference in frequency between the transmitted signal and
target return signals can be obtained using the mixer. This difference in frequencies is called
the beat frequency. If the LFMCW radar sensor generates a triangle waveform (see Figure 3),
the up- and down-beat frequencies are distinguished using the fast Fourier transform (FFT) at
the mixer output. If the target vehicle is stationary the beat frequency is a constant value.
Figure 2. Overview of MRR and SRR systems.
Figure 3. LFMCW radar sensor waveform.
However, if the target vehicle is moving there is the Doppler frequency and, as a
consequence, the beat frequency is changed. To detect the target return signal, the CFAR
detector and GD are employed and their detection performances are compared. Using the beat
frequencies, we can determine the target vehicle ranger R and relative velocity Vr between the
subject and target vehicles in the following form [11, 12]:
c( f bu  f bd )TS
R
8B , (1)
c( f bd  f bu )
Vr  , (2)
4 fc
where f bu is the up-beat frequency, f bd is the down-beat frequency, c is the velocity of light, B
is the bandwidth of the LFMCW radar sensor, f c is the centered or operation frequency, and
TS is the period of up-chirp and down-chirp waveforms. The relative velocity and range of the
target vehicle are functions of the up- and down-beat frequencies [13]. After definition of the
target vehicle range the target vehicle position or coordinates can be determined as an
intersection point of circles with the radius equal to the measured range (multilateration
process) using two or more radar sensors and the centers of these circles are the radar sensor
positions.
The LFMCW radar sensor system changes linearly the transmitted waveform frequency
as a function of time. The modulating waveform consists of two sweeps: the up-sweep or up-
chirp and down-sweep or down-chirp. The instantaneous transmitted frequency fT (t ) can be
presented in the following form:
 B
 f c  0.5T t , 0  t  0.5TS

f T (t )  
S
. (3)
 f  B t , 0.5T  t  T


c
0.5TS
S S
The delay time between the transmitted and target return signals is given by
2R

c . (4)
The target return signal delay in the case of moving target with the Doppler shift in
frequency takes one of the following forms:
 2( R  Vr t )
  for approachin g target vehicle;
c
m   (5)
 2( R  Vr t )
 for receding target vehicle.
 c
The LFMCW sinusoidal transmitted signal can be written in the following form:
ST (t )  AT sin[T (t )] , (6)
where AT is the amplitude, and T (t ) is the instantaneous phase defined as

tin
T (t )  2  fT (t )dt . (7)
0
From (3), (4) and (7), the sinusoidal transmitted signal can be rewritten in the following
form:
  B 
ST (t )  AT sin 2  f c t  t 2  . (8)
  TS 
The target return signal coming at the detector input can be written in the following form:

  B 
S R (t )  AR sin[ R (t )]  w(t )  AR sin2  f c (t   )  (t   ) 2    w(t ) , (9)

  TS 
where AR is the target return signal amplitude,  R (t ) is the target return signal phase, and w(t )
is the additive white Gaussian noise (AWGN) with zero mean and variance  n2 .
If the Doppler shift in frequency is taken into consideration (9) takes the following form:
   2V t  B  2V t   
2
S R (t )  AR sin[ R (t )]  w(t )  AR sin2  f c  t    r    t    r     w(t ) .
   c  TS  c  

(10)
In practice, under detection of the target return signal there is a need to estimate the noise
variance or noise power in real time or, in other words, to apply a procedure for definition of
adaptive detection threshold. Noise power estimation techniques are widely employed by
wireless communication, cognitive radio, speech recognition, radar sensor, remote sensing,
and other systems. In the case of radar sensor system, the noise power estimation is needed to
define the adaptive detection threshold owing to expected changes in the noise power,
especially, under the non-homogenous noise.
4. CFAR DETECTOR
Using the Neyman-Pearson criterion, the false alarm probability PFA must be minimal,
while the probability of detection PD must be maximal. The probability of false alarm PFA is
defined as the probability of event that the observed sample of signal sequence at the radar
sensor system output exceeds the detection threshold when the noise only is coming in at the
radar sensor system input. The probability of false alarm PFA under employment of the square
law detector can be expressed as [14]:

 THR 

PFA  exp  2  , (11)
 n 
 
where  n2 is the noise variance and THR is the detection threshold. The detection threshold
THR can be defined based on the probability of false alarm PFA given before in the following
form:
 1 
THR   n2 ln  . (12)
 PFA 
It is important to note that the probability of false alarm PFA is very sensitive to any
variations of the detection threshold value. Figure 4 shows the normalized detection threshold
THR  n2 as a function of the probability of false alarm PFA . The CFAR detection or, or in
other words, the adaptive threshold and automatic detection are a set of signal detection
techniques providing the predictable detection and false alarm rate behavior [15]. Under
CFAR detection, the noise power must be estimated in order to achieve the predictable and
consistent probability of false alarm PFA . By this way, the detection threshold can be adjusted
to maintain the desired and fixed probability of false alarm PFA .
Figure 4. Normalized detection threshold versus the probability of false alarm.
For any CFAR detector the noise power is estimated after processing the number of
reference cells using the sliding window technique. The required threshold is defined by
multiplication between the estimated noise power and scaling factor. This threshold is
compared with the signal power in the test cell or the investigated cells. CFAR detectors are
differed based on the reference cell processing procedure and scaling factor. The threshold of
any CFAR detector can be defined in the following form [16]:
THR   n2 , (13)
where  is the scaling factor. Figure 5 presents a simple functioning diagram for CFAR
detector. CFAR detectors use the estimation technique based on arithmetic averaging. The
square law detector shown in Figure 5 generates the output signal amplitude proportional to
the squared input signal amplitude.
Figure 5. CFAR detector functioning diagram.
We consider the following CFAR detectors: CA-CFAR, OS-CFAR, GCML, and ACGO-
CFAR. All these detectors can be employed by LFMCW radar sensor system. The above-
mentioned detectors are compared by detection performance of the target return signal under
the same initial conditions. We analyze two cases:
 There are no any interfering target vehicles.

 There are one and/or several interfering target vehicles.
4.1. CA-CFAR Detector

The CA-CFAR detector [16] has an optimal performance under the homogeneous noise
conditions when the neighboring reference cells contain the noise samples with the same
statistical characteristics as the test cell. The CA-CFAR detector estimates the noise power by
averaging the data samples in the reference cells of sliding window and after that defines the
detection threshold. Basic CA-CFAR detector structure is shown in Figure 6.
Figure 6. The main structure of the CA-CFAR detector.

The noise in the cells of leading and lagging windows and in the test cell is the
independent identically distributed (i.i.d.) process. If the target return signal reading is in the
test cell, the cells of leading and lagging windows do not contain the target return signal
readings from other targets that can bias the detection threshold. CA-CFAR detector uses the
maximum likelihood estimator with respect to readings in the cells of leading and lagging
windows with the purpose to form the noise power estimation.
The noise power estimation can be presented in the following from [17]:
N
 Xi
1
 n2CA  , (14)
N i 1
where X i , i  1,, N are the square law detector output samples stored by the reference cells.
The signal amplitude at the square law detector output is proportional to the squared input
signal amplitude and N is the number of reference cells. The probability of detection PDCA can
be determined in the following form:

PDCA  1   CA (1  SNR  N
, (15)
where SNR is the average SNR. The scaling factor for CA-CFAR detector is a function of the
number of reference cells N and the probability of false alarm PFA :
 CA  ( PFA ) 1 N  1 . (16)
The CA-CFAR detection threshold is defined as follows
THRCA   CA n2CA

. (17)
4.2. OS-CFAR Detector

In many practical cases, the non-homogeneous noise is a result of spatial and temporal
variations in the noise power or closely spaced target return signals resulting in a bias in the
noise power estimation and, consequently, the detection threshold. The OS-CFAR detector is
employed under the non-homogeneous noise conditions [18]. In the multiple target case, the
detection performance of OS-CFAR detector is better in comparison with the detection
performance of the mean level CFAR detector family including the CA-CFAR detector,
greatest of CFAR (GO-CFAR) detector, and smallest of CFAR (SO-CFAR) detector. The
OS-CFAR detector rearranges the reference cell data samples {X1 ,, X N } to form a new
sequence according to the increasing power {X (1)  X (2)    X (k )    X ( N ) } , where X (k )
is called the kth order statistics selected as the noise power reading (see Figure 7).
This approach is based on the fact that the interfering target signal power is usually
higher than the noise power. Thus, the OS-CFAR detector is able to reject the number N  k
of the target return signal readings caused by the interfering target vehicles, which are stored
in the reference window cells. Minimal OS-CFAR detector losses are achieved if k has a
reasonable value defined by
k  0.75k . (18)
Figure 7. The main structure of the OS-CFAR detector.
The probability of detection PDOS for OS-CFAR algorithm is determined by the following
from [18]:
k 1
N i
PDOS   . (19)
 OS
i 0 N i 
1  SNR
The scaling factor in the case of OS-CFAR procedure is given by
 OS  ( PFA ) 1 k  1 . (20)
Finally, the OS-CFAR detection threshold is determined as:
THROS   OS  n2OS   OS X ( k ) . (21)
4.3. GCML Detector
GCML detector is employed under the non-homogeneous noise [19]. The GCML
detector defines the number and location or coordinates of interfering target vehicles by data
stored in the reference window cells and discards them prior to define the noise power and
threshold. Thus, there is no need to know the number of interfering target vehicles under
noise power estimation (Figure 8).
The reference window is divided on two windows equal by length or number of cells,
namely, the leading window cells with data {X 1 , X 2 ,, X M } , and the lagging window cells
with data {X M 1 , X M 2 ,, X 2M } where M  N 2 . The data stored in the cells of these two
windows are processed independently in a parallel way by censoring processors. The
censoring algorithm ranks the data stored in the cells of these two windows in ascending
order of the amplitude as {X (1)  X (2)    X ( M ) } and {X ( M 1)  X ( M 2)    X ( 2M ) } ,

respectively. The algorithm considers the lowest order X (1) as the noise power sample and
defines the threshold THRX (1)  1 X (1) , where  n is the scaling factor or constant chosen to
obtain a specific probability of false censoring and n  1,, M 1 . The probability of false
censoring can be determined in the following form [19]:
Figure 8. The main structure of the GCML detector.
M  n 1
PFC  . (22)
(1  M  n) M n
The threshold THRX (1) is compared with X ( 2) and if X ( 2) exceeds the threshold THRX (1) the
censoring processor decides the data {X 2 , X 3 ,, X M } are the target return signal samples from
the interfering target vehicles. If X ( 2)  THRX (1) the algorithm decides that X ( 2) is the noise
sample without interference. After that the censoring algorithm generates a sum of two lower
ordered samples X (1,2)  X (1)  X (2) , defines the threshold THRX (1, 2) , and compares it with
X (3) . The censoring process is stopped when the hypothesis H1 is true. The noise power
estimation  n2GCML is scaled by the scaling factor  GCML to achieve the desired probability of
false alarm PFA . The GCML detection threshold is defined as follows:
THRGCML   GCML n2GCML (23)
After definition of the threshold THRGCML , the test cell data are compared with it and a
decision a “yes” or a “no” target vehicle is made.
4.4. ACGO-CFAR Detector
The ACGO-CFAR detector can be considered as a modified version of the GCML

detector. The statistics stored in the cells of the leading and lagging windows are not the same
[20]. The detector suppresses the clutter edge false alarm by calculating the average noise
power in the leading and lagging window cells separately after applying the censoring process
and selects the highest averaged value considering it as the required noise power estimation.
The structure of ACGO-CFAR detector is shown in Figure 9.
Figure 9. The main structure of the ACGO-CFAR detector.
Assume that after applying the censoring process to the leading and lagging window cells
n1 samples are censored from the leading window cells and n2 samples are censored from the
lagging window cells. The remaining samples from the leading window cells m1  M  n1 and
from the lagging window cells m2  M  n2 are used to estimate the noise power as follows
[20]:
 1 1
m
U 


m1 i 1
X (i ) ,
 m
(24)
 1 2
W  
m2 i 1
X (i ) .

The noise power estimation is set to be a maximum of U and W :
 n2ACGO  ˆ n2  max(U ,W ) . (25)
The noise power estimation  n2ACGO is scaled by the scaling factor  ACGO to achieve the
desired probability of false alarm PFA . Then, the output of test cell is compared with the
ACGO-CFAR detector threshold to make a decision whether the test cell contains the target
return signal sample or not. The ACGO-CFAR detector threshold is given by
THRACGO   ACGO n2ACGO . (26)

5. CFAR DETECTION PERFORMANCE

The detection performance of all discussed above CFAR detectors is evaluated under the
same initial conditions for two cases:
 There are no interfering target vehicles;

 There are interfering target vehicles.
Figure 10. Probability density functions of RCS fluctuations described by the Swerling models when
 av  1.5 .
We consider the following scenario. There are the subject vehicle and the only target
vehicle approaching from the rear side of the subject vehicle with relative velocity
Vr  10 m/s and 70 m away, i.e. the target vehicle range R  70 m . The interfering target
vehicles are located at different distances away from the subject target with various relative
velocities. The radar cross section (RCS) fluctuations of all targets are modelled by the
Swerling 2 model [21]. The statistics associated with the Swerling 2 model of RCS
fluctuations is applied to the targets consisting of small scatterers with the comparable RCS,
i.e. the independent fluctuating point scatterers with approximately equal physical
dimensions. The probability density function (pdf) of the target RCS fluctuations described
by the Swerling 1 and 2 models (see Figure 10) is defined as:
1   
f ( )  exp   ,  0 (27)
 av   av 
where  av denotes the average RCS fluctuations of total number of targets.

The target RCS fluctuations described by the Swerling 2 model are rapid in comparison
with the target RCS fluctuations described by the Swerling 1 model. The losses of target RCS
fluctuations described by all Swerling models at the probability of false alarm PFA equal to
104 as a function of the probability of detection PD are presented in Figure 11. In practice, if
the probability of detection PD is greater than 0.5, i.e. PD  0.5 , the losses of RCS fluctuations
described by the Swerling 1 and 2 models are higher than the losses of RCS fluctuations
described by the Swerling 3 and 4 models. For example, when the probability of detection PD
is equal to 0.8, the losses of RCS fluctuations described by the Swerling 3 model are
approximately 2.3 dB and the losses of RCS fluctuations described by the Swerling 2 model
are 5 dB.
Figure 11. Losses of the fluctuations described by the Swerling models.
In practice, the LFMCW radar sensor system parameters are chosen to get the specific
range and velocity resolution that can be defined using the following forms [22]:
c
R 
2B ; (28)

V  , (29)
TS
where R is the range resolution, V is the velocity resolution, and  is the wavelength. If
there is a requirement to achieve the range resolution equal to R  0.75 m, the required
bandwidth must be equal to B  200 MHz . When the velocity resolution is equal to
V  0.2 m/s, the required modulation time TS must be 0.0625 sec for the radar sensor
operation frequency f c equal to 24 GHz.
The simulation for the discussed CFAR detectors is carried out for the ultra-wide band
(UWB) LFMCW radar sensor system with the bandwidth B  600 MHz , operation frequency
f c  24 GHz , modulation time TS  0.0625sec that is equivalent to the up-sweep waveform
period equal to 0.03125sec and the probability of false alarm equal to PFA  10 4 . Under
simulation, the probability of detection PD is defined after K observations. In this case, the
probability of detection PD is determined as the ratio of the number of overshoots K o with
respect to the detection threshold (the detection threshold exceeding) to the total number K of
observations [14]:
Ko
PD 
K . (30)
Table 3 presents the main parameters of the investigated LFMCW radar sensor system.
The LFMCW radar sensor system parameters are chosen for the MRR and SRR sensor
systems implemented in safety driving applications [23-25].
Table 3. Main simulation parameters
Simulation Parameters Values

Operation Frequency 24 GHz
Bandwidth (B) 600 MHz
Modulation Time (Ts) 0.0625 s
Probability of False Alarm (PFA) 10-4
Reference Cells (N) 20
Number of Observations (K) 1000
Signal Power 100 mW
Selected Cell (k) 15 (for OS)
First Target Range (R1) 60 m
First Target Relative Velocity (V1) 10 m/s
Probability of False Censoring (PFC) 10-3
Second Target Range (R2) 63 m
Second Target Relative Velocity (V2) 10 m/s
Third Target Range (R3) 66 m
Third Target Relative Velocity (V3) 10 m/s
Fourth Target Range (R4) 70 m
Fourth Target Relative Velocity (V4) 10 m/s
Fifth Target Range (R5) 73 m
Fifth Target Relative Velocity (V5) 10 m/s
Sixth Target Range (R6) 76 m
Sixth Target Relative Velocity (V6) 10 m/s
No Guide Cells
Subject target vehicle is simulated as Swerling 2 model
Interfering target vehicles are simulated as Swerling 2 model
Figure 12 demonstrates the additive mixture of the up-chirp target return signal and noise
at the mixer output (see Figure 2) after applying FFT to obtain the up-beat frequency. The
CA-CFAR processing can be applied to the mixer output using 20 reference cells in order to
make a decision about the presence or absence of the target return signal. The detection
threshold is set according to the noise power estimation and the system can detect the target
return signal if the detection threshold is exceeded, see Figure 12, SNR  9.21 dB .
Based on simulation, the probability of detection PD is determined after K observations, as

shown in Figure 13 when the number of overshoots K o with respect to the detection threshold
is counted.
Figure 12. Mixer output process: a) after FFT and b) CA-CFAR detection, SNR  9.21 dB .
Figure 13. The number of overshoots for CA-CFAR detection with respect to the threshold: a) counting
the number of overshoots; b) the number of overshoots for the CA-CFAR detector in the up-chirp target
return signal.
Based on simulation, the probability of detection PD is determined after K observations, as

shown in Figure 13 when the number of overshoots K o with respect to the detection threshold
is counted.
Figure 14 presents a comparison between the detection thresholds of the CA-CFAR and
OS-CFAR detectors in the up-chirp signal when the up-beat frequency is equal to 7360 Hz. In
a general case, when the noise is presented only in the reference cells of sliding window, the
OS-CFAR detection threshold is higher than the CA-CFAR detection one.
Theoretical probability of detection as a function of SNR for the Neyman-Pearson, CA-
CFAR, and OS-CFAR detectors is presented in Figure 15. The CA-CFAR detector has the
better detection performance in the case of homogeneous noise in comparison with the OS-
CFAR detector. Detection characteristics of the CA-CFAR and OS-CFAR detectors are very
close to the detection performance of the Neyman-Pearson detector. In practice, the multiple
target scenario, i.e. there are several vehicles in the observation area, is widely used. Assume,
there are the primary target vehicle, i.e., the subject vehicle, with the specific range and
relative velocity, and the interfering target vehicle with the same relative velocity but its
range is different from the primary target vehicle one. The target return signal samples from
the interfering target vehicle are stored in the cells of the reference window where the target
return signal data from the primary target must be stored. For the case of two target vehicles,
namely, the primary target and one interfering target, the up-chirp signals are shown in Figure
16.
Figure 14. Comparison of the CA-CFAR and OS-CFAR detector thresholds: a) enlarged snapshot and
b) comparison of the CA-CFAR and OS-CFAR detector thresholds.
The target masking phenomenon occurs when there are several target vehicles and the
target return signal sample belonging to a specific target vehicle is stored in the test cell of
sliding window, while the target return signal samples from other target vehicles are stored in
the reference cells of sliding window. In the case of the CA-CFAR detector, the target return
signal samples stored in the reference cells can mask the target return signal stored in the test
cell of sliding window. The target masking phenomenon is considered as a disadvantage of
the CA-CFAR detector, since the noise power estimation is based on determination of the
arithmetic mean or averaging the data stored in the reference cells of sliding window.
Figure 17 demonstrates a comparison between the detection thresholds of the CA-CFAR
and OS-CFAR detection procedures for two targets. As shown in Figure 17, the target return
signal of the first target vehicle exceeds the CA-CFAR and OS-CFAR detection thresholds.
The target return signal of the second target vehicle exceeds only the OS-CFAR detection
threshold. Thus, the OS-CFAR detector is able to detect two target vehicles, while the CA-
CFAR detector can detect only the first target vehicle owing to the target masking
phenomenon. In other words, this is the target masking case for the CA-CFAR detector.
Figure 18 presents a detection of the target return signals for two target vehicles by the
OS-CFAR detector at k  15. The noise power estimation procedure under OS-CFAR
detection technique allows the detector to be insensitive with respect to the target vehicles
closely spaced between each other or to be robust relative to the interfering target vehicles.
Figure 15. Theoretical detection performance of the NP, CA-CFAR, and OS-CFAR detectors.
Figure 16. The up-chirp signals in the case of two target vehicles.
Figure 19 demonstrates a comparison of detection performances between the CA-CFAR

and OS-CFAR detectors for three cases:
 There are no any interfering target vehicles;
 There is one interfering target vehicle;
 There are two interfering target vehicles.
We assume that the interfering target vehicles have the same radar cross section (RCS) as
the primary target. If there is no interfering target vehicle, in this case, the CA-CFAR
detection procedure is the best by performance. If we observe one or two interfering target
vehicles, the OS-CFAR detection is the better and demonstrates an advantage in robustness
with respect to the interfering target vehicles. At the same time, the CA-CFAR detection is
very poor by performance owing to the target masking problem.
Figure 17. The target masking phenomenon of the CA-CFAR detector.
Figure 18. OS-CFAR detection of two closely spaced target vehicles.
Figure 19. Detection performance comparison of the CA-CFAR and OS-CFAR detectors.
The GCML detection procedure allows us to estimates the number of interfering target
vehicles. In this case, the target return signals of interfering target vehicles are stored in the
reference cells of sliding window. Using cell by cell technique, the censoring procedure
allows us to accept or reject the data stored in the cells of the reference window. Thus, to
make a decision about detection of the subject vehicle there is no need to know a priori the
number of interfering target vehicles. Under employment of the GCML detector, the scaling
factors  n , n  1,, M  1 are chosen to get the given before probability of false censoring PFC
in (22). For the case when the number N of reference cells equals to 20, i.e. M  N / 2  10, and
the probability of false censoring PFC is equal to 10 3 , the related scaling constant values are
presented in Table 4. Detection performance of the GCML detector is similar to the CA-
CFAR detection performance if there are no interfering target return signals.
Figure 20. False alarm for GCML detector caused by clutter edge transition: a) the ACGO-CFAR and
GCML detectors clutter edge case and b) false alarm for the GCML detector.
Table 4. The scaling constants for censoring procedure and PFC  10 3
nth step 1 2 3 4 5 6 7 8 9
n 1999 53.7723 14.8740 7.4090 4.6968 3.3737 2.6106 2.1209 1.7826
The ACGO-CFAR detector can suppress the clutter edge false alarm by independent
processing the data stored in the cells of leading and lagging windows with the purpose to
obtain two tentative noise power estimations. In the case of GCML detector, there is a false
alarm in the clutter transition region. The ACGO-CFAR detector with the high detection
threshold is less susceptible to false alarms within the limits of the transition regions (see
Figure 20).
Figure 21. Comparison of detection performance of the OS-CFAR, GCML, and ACGO-CFAR
detectors: no interfering target vehicles.
detectors: one interfering target vehicle.
detectors: two interfering target vehicles.
Figure 21 demonstrates a comparison between the OS-CFAR, GCML, and ACGO-CFAR

detection performance when there are no interfering target vehicles. The curves are very close
to each other with slightly vantage to GCML detector. Although the detection performance of
the ACGO-CFAR detector is inferior to the GCML detection performance, the ACGO-CFAR
detector exhibits the robust false alarm control property and has a capability to censor the
undesired interfering target vehicles. Figures 22 and 23 demonstrate the cases when one
interfering target vehicle and two interfering target vehicles are introduced, respectively. The
OS-CFAR, GCML, and ACGO-CFAR detectors demonstrate robustness and the same
detection performance ranking shown in Figure 21.
detectors: five interfering target vehicles.
Figure 24 demonstrates the detection performance of the OS-CFAR, GCML, and ACGO-
CFAR detectors when there are five interfering target vehicles. In this case, all detectors show
degradation in the detection performance and also keep similar ranking as before excluding
the OS-CFAR detector that has the worst detection performance at the high SNR values (more
the 20 dB).
Figure 25. Comparison of detection performance six interfering target vehicles.
Figure 25 indicates the fact that the OS_CFAR detection performance is the worst case
when there are six interfering target vehicles. The OS-CFAR detector presents a good
detection performance if there are five or less interfering target vehicles at N  20 and k  15 .
In this case, the OS-CFAR detector can handle the number of interfering target vehicles that
is not more than N  k  5. Thus, under the condition N  20 five interfering target vehicles are
a cutoff case for the OS-CFAR detector and any additional interfering target causes a serious
degradation in the detection performance.
The simulation confirms robustness of the OS-CFAR, GCML, and ACGO-CFAR
detectors in multiple target scenario. A prior knowledge about the number of interfering target
vehicles is needed in the case of the OS-CFAR detection to maintain a good detection
performance and determine the cutoff interference conditions. The GCML and ACGO-CFAR
detectors are robust with respect to the interfering target vehicles independently of their
number.
6. DEFINITION OF TARGET VEHICLE COORDINATES

There are several procedures to define and estimate the target vehicle position or
coordinates using the radar sensor systems. Some of these procedures are based on the
mechanical and/or electronic scanning radar sensor systems that can be classified as an
expensive solution for large-scale production. The use of multiple radar sensors to form a
radar sensor network system is a practical solution to define the target vehicle position or
coordinates employing the multilateration method based on definition of target vehicle ranges
only. The case when there is a network of two radar sensors is shown in Figure 26. For each
radar sensor system or network, the target vehicle position or coordinates can be determined
as an intersection point of circles with the radius equal to the determined target vehicle range,
the centers of which are the radar sensors on the subject vehicle. Each radar sensor measures
the target vehicle range with error. An iterative approximation algorithm based on the least
square estimation is applied to determine the target vehicle position or coordinates [26].
Figure 26. Definition of the target vehicle coordinates using network of two radar sensors.
Assume the actual position of the target vehicle in the Cartesian coordinates can be
defined as
p  [ xt , yt ] . (30)
The two radar sensor network positions in the Cartesian coordinates are known and given by
s1  [ xs1 , ys1 ] , s 2  [ xs2 , ys2 ] (31)
Based on (30) and (31), the difference in positions or coordinates between the target
vehicle and the first and second radar sensors of the subject vehicle defined by the vectors l1
and l 2 can be presented in the following form, respectively:
l1  s1  p  [ xs1  xt , ys1  yt ] , l 2  s 2  p  [ xs2  xt , ys2  yt ] , (32)
rm1  ( xs1  xt ) 2  ( ys1  yt ) 2  w1 , rm2  ( xs2  xt ) 2  ( ys2  yt ) 2  w2 , (33)
where rm1 , i  1,2 is the target vehicle range defined by the radar sensor, and wi , i  1,2 is the
additive noise.
The equation (33) shows that a relation between the measured range of the target vehicle
and coordinates of the target vehicle is expressed by the non-linear function. This non-linear
function can be presented in the following form:
z  h(x)  w , (34)
where z is the vector of target vehicle range, h(x) is the non-linear function vector, and w is
the additive noise vector. The non-linear equation can be converted into the linear equation by
linearization procedure near the target vehicle coordinates as follows:
rmi rmi
rmi  rmi  ( xt  xt )  ( yt  yt )  F (e) , (35)
xt 
yt 
p p
where rmi , i  1,2 is the distance between the target vehicle coordinates p  and the radar sensor
network, and F (e) is the error term. The target vehicle coordinates are given by
p  [ xt , yt ] . (36)
Neglecting the error term F (e) in (35) we obtain the following equation:
 r  rm1 
 m1 
rm  rm   xt

p
yt   
xt  xt 

 p 
z  H  x    ,
1 1
(37)
rm  rm   r 

rm2   yt  yt 

 2 2  m2
 xt yt  
 p 
p 
where z is the vector of difference between the real target vehicle range and the expected
target vehicle range, x is the vector of errors in determination of the target vehicle
coordinates, and H is the normalized distance between the radar sensors and target vehicle in
the vector form determined as follows
 xs1  xt ys1  yt 
  
 ( xs1  xt ) 2  ( ys1  yt ) 2 ( xs1  xt ) 2  ( ys1  yt ) 2 
H  . (38)
 xs2  xt ys 2  y t 
  
 ( xs2  xt ) 2  ( ys2  yt ) 2 ( xs2  xt ) 2  ( ys2  yt ) 2 
The least square solution of linearized equations is obtained by minimizing the weighted
sum of squared deviations [17]:
J
0, (39)
x
where J is given by
J  [z  Hx ] T Cov 1[z  Hx ] , (40)
where Cov is the covariance matrix of noise given by the following form:
 2 0 
Cov  E[ww T ]   1 . (41)
0  22 
In (40) and (41), T is denotes the matrix transpose. The solution of (39) gives us the error
in determination of the target vehicle coordinates:
x  [HT Cov 1H]1 HT Cov -1z . (42)
When the radar sensor networks operate under the same noise conditions, i.e., the
variance of noise is the same for all radar sensors, (42) can be simplified and rewritten in the
following form:
x  [HT H]1 HT z . (43)
Using the error in determination of the target vehicle coordinates given by (43), the
process of definition of the target vehicle coordinates is stopped when the error becomes
negligible (the error is converged to infinitesimal value after the definite number of
iterations). The convergence of error is usually observed after four or five iterations. All
previous steps are presented in Figure 27.
The steps of target vehicle coordinates definition can be simply generalized by the
following way:
Figure 27. Steps of iteration under determination of the target vehicle coordinates.
 Initiate the iteration procedure assuming the target vehicle coordinates are given;
 Define the matrix H and linearized equations;
 Define the error in determination of the target vehicle coordinates;
 Correct the selected target vehicle coordinates at the first step using the obtained
vector of errors x ;
 Repeat the steps from the starting point if the error is not converged.
The calculation of the target vehicle coordinates using the multilateration technique is
carried out for the radar sensor network consisting of two radar sensors under the effect of the
additive Gaussian noise (see Figure 26). The vectors s1 and s 2 of coordinates of the two radar
sensors in the Cartesian coordinates take the following form: s1  0 ,  1 and s1  0 , 1 . The
vector p of the real target vehicle Cartesian coordinates takes the following form: p  50,5.
The calculation is carried out for the following initial target vehicle coordinates x and y: 1)
x  1, y  1; 2) x  5, y  5; 3) x  10, y  10 .
Figure 28 presents the calculation results using the multilateration technique when the
three initial target vehicle coordinates are assigned before. The calculated coordinates of the
target vehicle are converged to the real coordinates of the target vehicle after five iterations
for all cases. The multilateration technique based on the least square estimation is considered
as a robust estimation procedure for definition of the target vehicle coordinates.
7. GD MAIN FUNCTIONING PRINCIPLES

7.1. General Statement
The simple GD model constructed based on the generalized approach to signal processing
in noise [27] is represented in Figure 29. We use the following notations: MSG is the model
signal generator (local oscillator), the AF is the additional filter (the linear system), and the
PF is the preliminary filter (the linear system). A detailed discussion of the AF and PF can be
found in [27, 28].
Consider briefly the main statements regarding the AF and PF. There are two linear
systems at the GD front end that can be presented, for example, as the band-pass filters,
namely, the PF with the impulse response hPF ( ) and the AF with the impulse response
hAF ( ) . For simplicity of analysis, we think that these filters have the same amplitude-
frequency responses and bandwidths. Moreover, a resonant or centered frequency of the AF is
detuned relative to a resonant or carrier frequency of the PF on such a value that the target
return signal cannot pass through the AF, for example, on a value that is more the signal
bandwidth. Thus, the target return signal and noise can be appeared at the PF output and the
only noise is appeared at the AF output. It is well known, if a value of detuning between the
AF and PF resonant frequencies is more than 4  5f a , where f a is the signal bandwidth, the
processes forming at the AF and PF outputs can be considered as independent and
uncorrelated processes. In practice, the coefficient of correlation is not more than 0.05.
Figure 28. Definition of target vehicle coordinates applying the multilateration technique: a)
x  1, y  1; b) x  5, y  5; c) x  10, y  10 .
Figure 29. Principle flowchart of GD.

In the case of a “no” target return signal in the input process, the statistical parameters at
the AF and PF outputs are the same, because the same noise is coming in at the AF and PF
inputs (the ideal case). We may think that the AF and PF do not change the statistical
parameters of input process, since they are the linear GD front end systems. By this reason,
the AF can be considered as a generator of reference sample with a priori information a “no”
signal obtained in the additional reference noise at the AF output.
There is a need to make some comments regarding the noise forming at the PF and AF
outputs. If the Gaussian noise w(t ) comes in at the AF and PF inputs (the GD linear system
front end), the noise at the AF and PF outputs is Gaussian, too, because the AF and PF are the
linear systems and, in a general case, take the following form:
K K
 i   hPF wi and ηi   hAF wi
i 1 i 1 , (44)
where  i is the sample of observed noise at the PF output, ηi is the sample of observed noise
at the AF output, and K is the sample size (i  1,K ) . If AWGN with zero mean and two-
sided power spectral density 0.5 N 0 is coming in at the AF and PF inputs (the GD linear
system front end), then the noise at the AF and PF outputs is Gaussian with zero mean and
variance given by
N 002
 n2 
8 F
, (45)
where, in the case if AF (or PF) is the RLC oscillatory circuit (resistor, inductor, capacitor),
the AF (or PF) bandwidth  F and resonance frequency 0 are defined in the following manner
R R
 F   , 0  ,  , where R is the resistance, L is the inductance, and C is the
LC 2L
capacitance, respectively.
The main functioning condition of GD is the equality over the whole range of parameters
between the model signal forming at the GD MSG output and the detected target return signal
at the GD input liner system (the PF) output. How we can satisfy this condition in practice is
discussed in detail in [27, 28]. More detailed discussion about a choice of PF and AF and
their impulse responses is given in [29].
7.2. GD Structure
Under conditions closed to practice, we can use the GD flowchart presented in Figure
7.1, see [28], in which the threshold apparatus (THRA) device defines the GD threshold, and
the signal model generator switching apparatus (SGSA) is used to switch on the MSG to
define the unknown parameters of the detected target return signal. The decision block (DB)
with the decision function employs the decision-making rule of the GD. The switch K1 takes
the position “1” to define the GD detection threshold, and takes the position “2” after
definition of the GD threshold and after that a detection procedure of the target return signal
is carried out. The switch K2 works to put the THRA in and out of service.
in
Let Z GD i
be the GD input stochastic process observed within the limits of the time interval
[0, T ] ; ai is the sample of target return signal; wi is the sample of the observed additive
Gaussian noise with zero mean and known variance  n2 . The detection problem of the target
return signal from the target vehicle can be presented in the following form:
 ai  wi  H1 ;
in
Z GD  (46)
i
 wi  H0 ;
where H 1 is the hypothesis about the presence of the target return signal from the target
vehicle; and H 0 is the alternative hypothesis. The signal at the GD PF output in the time
domain can be defined as Yi  ai  i . The GD AF output is the reference noise ηi given by
(44). The model signal is defined as:
ai  ai , (47)
where ai is the sample of generated model signal,  is the factor of the proportionality, and
i  1,K .
The use of the generalized approach to signal processing in noise assumes some
modifications concerning the initial premises of classical and modern signal detection
theories. The signal detection algorithm constructed based on generalized approach to signal
processing in noise can be presented in the following form [27, 28]:
K K K
 2Yi ai  Yi2  i2  H 1 THRGD , (48)
i 1 i 1 i 1 H0
where Yi is the stochastic process sample observed at the PF output and THRGD is the GD
threshold. The first term in (48) corresponds to the correlation receiver, which is optimal in
the Neyman-Pearson criterion sense, when the target return signal parameters are known a
priori, with twice the gain and is considered as the sufficient statistics of the likelihood mean.
The second term in (48) corresponds to the energy detector, which is optimal in the Neyman-
Pearson criterion sense, when the target return signal parameters are unknown a priori, and is
considered as the sufficient statistics of the likelihood variance [27, 28]. The third term in
(48) presents the reference noise power at the AF output generated according to the main
functioning principles of the generalized approach to signal processing in noise [27, 28].
Equation (48) represents the decision-making rule under implementation of the generalized
approach to signal processing in noise in any signal processing system.
In the case of target return signal presence (the hypothesis H 1 ), Yi  ai  i , the left side
of (48) takes the form i1 ai2  i1i2  i1 i2 where i 1 ai2  Ea is the signal energy, and
K K K K
i 1i2  i 1i2 is
K K
the background noise at the GD output. The background noise is a
difference between the noise power at the GD PF and GD AF outputs. In the opposite case,
i.e. the target return signal is absent (the hypothesis H 0 ), Yi   i and the left side of (48) is the
i 1i2  i 1i2
K K
background noise only. In the ideal case as K   the GD background
i 1i2  i 1i2
K K
noise tends to approach zero in the statistical sense.
7.3. GD Threshold for Radar Sensor Systems

According to the main functioning principles of GD, the target return signal must exceed
the GD threshold by absolute value before the target return signal parameters would be
defined. The fixed GD threshold can be used if the noise power or variance at the GD input is
constant. Thus, setting the GD threshold by absolute value more than the background noise
power or variance at the GD output may result in the acceptable probability of detection with
the low false alarm rate. This condition cannot be guaranteed in the case of unexpected
external sources of noise in different radar sensor systems. The main parameters used to
determine the GD threshold are the background noise power or variance at the GD output and
the probability of false alarm PFA . Applying the Neyman-Pearson criterion to GD, the
probability of false alarm PFA is determined in the following form

PFA   f H
Z GD0
( x)dx , (49)
THRGD
where f H0 ( x) is the background noise pdf at the GD output corresponding to the hypothesis
Z GD
H0 .
Under consideration of the hypothesis H 0 in the narrowband noise case with the Rayleigh
amplitude envelope and the phase uniformly distributed within the limits of the interval
[0,2 ] , the pdf f H0 ( x) can be presented in the following form:
Z GD
 1  | x | 
 exp   , x0
f H0 ( x)   4 n
2
 2 n2  (50)
Z GD

0 , x0 .
The background noise pdf given by (50) is defined by the exponential-type law when the
observation time interval [0, T ] is infinitesimal, i.e. T  0 . Based on (49), the probability of
false alarm PFA can be presented in the following form [29]:

1 
 THRGD 

PFA   f H
Z GD0
( x)dx 
2
exp 


2 
2 n 

. (51)
THRGD
According to (51), the GD threshold can be determined in terms of the probability of

false alarm PFA in the following form:
THRGD  2 n2 ln(2PFA ) . (52)
Introducing the GD scaling factor  GD  2 ln(2PFA ) , the GD threshold can be written as
THRGD   GD n2 . (53)
Taking into consideration a definition for the scaling factor, the probability of false alarm
PFA can be presented in the following form:
1   
PFA  exp  GD  . (54)
2  2 
The last equation can be used to determine the probability of false alarm PFA for a given
scaling factor  GD or to determine the required scaling factor  GD for the assigned probability
of false alarm PFA .
7.4. Noise Power Estimation for GD
In order to maintain the constant probability of false alarm PFA required to achieve the
acceptable detection performance stability, the GD threshold should be a function of the noise
power or variance at the GD AF output, i.e. the GD threshold should be adaptive. Based on
the previous discussion, we see that there is a need to determine the noise power in real time.
Thus, there is a need to apply the noise power estimation technique in order to define the GD
threshold. For this purpose, the estimation procedure should be valid even in the case of high
noise power depending on the noise standard deviation. Owing to the fact that the bandwidths
of the expected target return signals of interfering target vehicles are matched with the GD PF
bandwidth, the noise power estimation is not affected by these target return signals of the
interfering target vehicles. In the case when the bandwidth of interference from jammer or
any other sources is matched with the GD AF bandwidth, the noise power estimation
procedure will be affected.
Figure 30. Noise power estimation for GD.

Different noise power estimation procedures can be modified to be implemented in GD

[30, 31]. The procedures based on the sliding windows techniques allowing us to estimate the
noise power by processing a group of reference cells would be suitable solutions in the case
of GD employment in radar sensor systems. For sliding window technique, the noise power
estimation is performed assuming that the noise samples stored in the reference cells are i.i.d.
Cell averaging procedure is a simple and effective method to process the reference cells of
sliding window with the purpose to estimate the noise power using the noise sample 1, , N
(the reference noise) at the GD AF output (see Figure 30).
The noise power estimation using the cell averaging by N reference cells of sliding
window can be presented in the following form:
1 n 2
ˆ n2  i .
N i1
(55)
The GD threshold can be defined adaptively based on the noise power estimation
substituting (55) into (52). The number of reference cells N of sliding window is chosen in
accordance with tradeoff between the SNR losses, processing complexity, and noise power
estimation accuracy.
8. GD PERFORMANCE
The basic GD output signals are presented in Figure 31. The diagram shown in Figure
31a is the background noise at the GD output when there is no target return signal at the GD
input (the statistical mean is equal to zero). The diagram shown in Figure 31b presents the
GD energy detector output when there is the target return signal at the GD input. We can
observe a shift toward the negative region of amplitude of the GD output signal, i.e. the
statistical mean is not equal to zero that is a characteristic of a “yes” target return signal, but
we do not know this target return signal is caused by the target vehicle at the GD input or not.
To answer on this question there is a need to define the parameters of the target return signal.
For this purpose we use the GD correlation receiver. The diagram shown in Figure 31c
demonstrates the amplitude of the GD output signal when a “yes” target return signal at the
GD input and the GD correlation receiver and energy detector operate simultaneously. The
process observed at the GD output is a characteristic of the target return signal power, i.e. we
are able to define some parameters of the target return signal and make the following
decision: this target return signal is caused by the target vehicle at the GD input or not.
The GD detection performance is compared with the CA-CFAR detection one when there
is no any noise power estimation procedure applied to GD under assumption that the noise
power is known. The CA-CFAR detector uses the sliding window technique with the number
of cells in the reference window equal to 20, i.e. N  20 , Figure 32. The CA-CFAR detector
has the better detection performance at SNR  2 dB that is a region of the very low probability
of detection PD which cannot be used in practice for any signal processing application. At
SNR  2 dB, the GD demonstrates the better detection performance. We obtain the critical
SNR equal to 2 dB under the simulation parameters mentioned in Table 3.
Figure 31. Basic signal at the GD output: a) the background noise; b) the GD energy detector output; c)
the GD output under simultaneous operation of the correlation receiver and energy detector
Figure 32. Comparison of the CA-CFAR detector and GD by detection performance: a “no” noise
power estimation for GD.
Figure 33 demonstrates the GD and CA-CFAR detection performance when the noise
power estimation procedure based on the sliding window technique with the cell averaging
and the number of reference cells equal to 20, i.e. N  20 is applied to both detectors. At the
low probability of detection PD , for example, PD  0.1 that is not a practical case for all types
of signal processing systems the CA-CFAR detector overcomes the GD by detection
performance. If PD  0.5 , the acceptable case in practice, the GD demonstrates superiority in
detection performance for wide range of SNR in comparison with the CA-CFAR detection.
For example, at PD  0.8 , the required SNR for CA-CFAR detection is 15 dB, while the
required SNR for the GD is 13 dB. Based on Figures 32 and 33, we can see that the GD
performance is deteriorated using the noise power estimation technique. The GD performance
degradation is predictable since the estimated noise power is not exactly the same by value as
the real noise power owing to noise power estimation errors.
Figure 33. Comparison of the CA-CFAR detector and GD by detection performance: a “yes” noise
power estimation for GD.
According to the conditions of GD PF and AF choice, the reference noise at the GD AF

output is used to estimate the noise power and define the GD threshold. Thus, the target return
signals from the desired and/or interfering target vehicles cannot pass through the GD AF
[27–32]. The signals reflected from the interfering target vehicles (the interference signals)
have the same statistical characteristics as the target return signals from the desired target
vehicles, see (10). The bandwidth of interfering signals is matched with the GD PF
bandwidth.
If there are no interfering target vehicles, the GD output takes the following form:
K
out
Z GDi
  (2Yi ai  Yi2  i2 )
i 1 . (56)
The observed sample at the GD PF output when there is an interfering signal I (t ) can be
Yi  ai  I i  i , (57)
where I i is the interfering signal sample at the GD PF output. If the condition ai  ai is
satisfied, the GD output takes the following form:
K
out
Z GDi
  (ai2  2I ii  I i2  i2  i2 ) , (58)
i 1
where I i2 is the interfering signal energy and I i i is the component caused by the interaction
between the noise and interference.
Additional comparison between the GD and CA-CFAR detector by detection

performance is performed for the following cases:
 There is only one interfering target vehicle;

 There are two interfering target vehicles.
As before, we assume the bandwidth of interfering signals is matched with the GD PF

bandwidth. This comparison is presented in Figure 34. The GD noise power estimation
procedure is based on the sliding window technique with the cell averaging for cells of the
reference window as in the case of CA-CFAR. The GD demonstrates a good robustness with
respect to the interfering signals matched by bandwidth with the GD PF, because these
interfering signals do not effect on the noise power estimation procedure and, consequently,
on the definition of the GD detection threshold THRGD . Thus a superiority of GD in detection
performance over the CA-CFAR detector is evident in the case of one and/or several
interfering target vehicles.
Figure 34. Comparison of the CA-CFAR detector and GD by detection performance for two cases:
there are no the interfering target vehicles and there are the interfering target vehicles. The bandwidth
of interfering signals is matched with the GD PF one.
9. ATTENUATION EFFECT ON THE RADAR SENSOR PERFORMANCE

This section is about how some specific attenuation exerts on the radar sensor detection
performance. In practice, under implementation of any radar sensor system, the empirical
detection performance will be different from the theoretical detection performance owing to
many reasons. One of them is a signal attenuation caused by various sources. The main goal
of this section is to clarify the attenuation effect on the detection performance of the target
return signal from target vehicles. This effect is analyzed theoretically and the simulation is
made for CA-CFAR detection scheme. The obtained results can be generalized to other
CFAR detectors and GD because the detection performance degradation caused by
attenuation is the same regardless of the detector type.
9.1. Signal Attenuation and Radar Range Equation

If the electromagnetic signal is propagated through any environment, b for example, air
or atmosphere, an interaction with molecules of environment absorbs the signal energy. Thus,
the signal power is decreased with increasing in the target vehicle range. Introducing the
attenuation factor  af per meter, the relation between the received signal power Pr and
transmitted signal power Ptr can be written as follows:
Pr  Ptr exp( 2 af R)

, (59)
where R is the target vehicle range. We can express the attenuation factor  af in dB/km
using the basic logarithmic statement given as
logb X
log a X 
logb a
. (60)
It is easy to find that lg X  0.43429 ln X . Thus, the linear attenuation factor  af can be
 dB  4342.9 af dB/km. (61)
From (57) and (59), we can find that:
10 lg Pr  10 lg Ptr  2 dB R , (62)
where R is target vehicle range expressed in km.

The general attenuation factor  afgen is the sum of several attenuation factors representing
different attenuation effects such as the atmosphere, rain, fog, dust, snow, sand, and others.
We consider the following representation of the general attenuation factor:
 afgen   atm   rain   fog   dust   etc , (63)
where  atm is the atmospheric attenuation factor;  rain is the rain attenuation factor;  fog is
the fog attenuation factor;  dust is the dust attenuation factor; and  etc is the attenuation
factor caused by other sources.
9.2. Atmospheric Attenuation
The atmospheric attenuation factor  atm is a function of the operation frequency of the
radar sensor system. Figure 35 demonstrates the atmospheric attenuation factor  atm as a
function of specific atmospheric parameters, namely, the pressure, temperature, and density
of water vapour, i.e. humidity. As we can see from Figure 35, we observe the smallest
atmospheric attenuation factor at the operation frequency of the radar sensor system equal to
24 GHz in comparison with radar sensor systems functioning under the operation frequencies
equal to 60 GHz, 70 GHz, and 120 GHz.
Figure 35. The atmospheric attenuation factor values as a function of the operation frequency.
Figure 36. Specific attenuation due to atmospheric conditions (ideal parameter).
Figure 36 presents a comparison between the total atmospheric attenuation factor and the
dry air and water vapour attenuation factors as a function of the operation frequency of radar
sensor systems under the ideal parameters at the sea level, i.e. the pressure, temperature, and
density of water vapour [33, 34]. We can see that at the operation frequency of radar sensor
system equal to 24 GHz, the total atmospheric attenuation factor is lesser 0.2 dB/km.
As we mentioned before, the atmospheric parameters effect on the attenuation factor. By
this reason, we present in Figures 37 and 38 the same comparison as in Figure 36 for different
Figure 37. Specific attenuation due to atmospheric conditions (temperature =273.15 K).
Figure 38. Specific attenuation due to atmospheric conditions (air pressure = 898.75 hPa).
values of air temperature and pressure, respectively. We see that the total attenuation factor is
independent of air temperature when the weather is cold (the low air temperature).
Figure 38 demonstrates the total atmospheric attenuation factor and the dry air and water
vapour attenuation factors as a function of the operation frequency of radar sensor systems
when the air pressure is changed. In reality, this situation can be happened at high altitudes
above the sea level where the air pressure is low. We observe that the total atmospheric
attenuation factor is not changed considerably when the air pressure is varied in comparison
with the ideal case.
9.3. Rain Attenuation

The rain attenuation effects on the radar sensor detection performance are strongly related
to rain intensity that is called the rainfall rate. For example, that light rain having a uniform
spatial distribution can be predicted by relatively easy way. However, the heavy rain comes
jointly with squall and it is impossible to predict the rain intensity and drop size distribution
[33]. Mie scattering theory [35] is generally applied to frequencies up to the submillimeter
wave band. In this case, the rain attenuation factor can be determined using the following
form:
 rain   L , (64)
where the rain attenuation factor  rain has a dimension of dB/km, L is the rain rate expressed
in mm/hour,  and  are the coefficients depending on the operation frequency of radar
sensor system. Table 5 presents the values of the coefficients  and  at the operation
frequency of radar sensor system equal to 24 GHz.
Figure 39 presents the rain attenuation factor for different kinds of rain (the rain density)
as a function of the operation frequency of radar sensor system.
Figure 39. Rain attenuation factor.
Table 5. Empirical coefficients used to determine the rain attenuation
Coefficient 24 GHz
 0.19
 0.905
The rain attenuation factor at the operation frequency of radar sensor system equal to 24
GHz is presented in Table 6 for different types of the rain, namely, the drizzle rain, the light
rain, the medium rain, the heavy rain, the downpour rain, and the tropical rain.
Table 6. Rain attenuation factor at 24 GHz
Rain Type Attenuation Factor [dB/Km] Rainfall Amount [mm/hr]

Drizzle 0.019 0.25
Light rain 0.089 1.25
0.3 2.5
Medium rain
1.5 12.5
Heavy rain 3.5 25
Downpour 7 50
Tropical 19 100
Table 7. The CA-CFAR detector probability of detection under atmospheric and rain
attenuation
Type Probability of Detection (SNR=9.2 dB)

No Attenuation 0.7123
Atmospheric 0.6871
Atmospheric + Drizzle Rain (0.25 mm/hr) 0.6834
Atmospheric + Light Rain (1.25 mm/hr) 0.6821
Atmospheric + Medium Rain (2.5 mm/hr) 0.6800
Atmospheric + Medium Rain (12.5 mm/hr) 0.6135
Atmospheric + Heavy Rain (25 mm/hr) 0.5452
Atmospheric + Downpour (50 mm/hr) 0.3681
Atmospheric + Tropical (100 mm/hr) 0.0827
Figure 40. Rain attenuation factor action on the CA-CFAR detection performance.
Simulation results given in Table 7 demonstrate the probability of detection for CA-
CFAR detector under various values of the atmospheric attenuation factor and combination
between the atmospheric and rain attenuation factors. Simulation is carried out at the specific
SNR  9.2 dB and the target vehicle range equal to 140 m. From the results presented in
Table 7, we see that the effect of the atmospheric and low (the drizzle, light, and medium rain
when the rainfall amount is less than 12.5 [mm/hr]) rain attenuation does not change
considerably the probability of detection of radar sensor system. Starting from the case of
heavy rain density, a degradation of the detection performance is evident, especially, in the
case of the downpour and tropical rain.
Table 8. The general simulation parameters
Parameter Value
Carrier Frequency 24 GHz
Probability of False Alarm PFA 10-3
Number of Reference Cells N 20
Target Range (one way) 70 m
Temperature 20◦C
Moisture Content 7.5 g/m3
The simulation results concerning the rain attenuation effects on the CA-CFAR detection
performance are shown in Figure 40. These results are obtained based on the simulation
parameters presented in Table 8. As follows from Figure 40, the detection performance of the
CA-CFAR detector suffers from serious degradation in the case of downpour and tropical
rain. Additionally, we can conclude that the attenuation caused by the rain exerts more
influence on the detection performance of radar sensor system in comparison with the
atmospheric attenuation. The curves in Figure 40 corresponding to the attenuation factors for
the case of ideal, atmospheric, atmospheric plus light rain, and atmospheric plus medium rain
are very close to each other and it is very difficult to distinguish a difference between them.
9.4. Fog Attenuation
The fog is created by the atmospheric water vapour condensation into water droplets
suspended in the air. Thus, the fog can be defined as the water droplet density that restricts or
reduces a visibility to less than 1 km where advection of the fog is caused by horizontal
movement of a warm air mass. Another type of the fog called the radiation fog is caused by
air cooling overnight. This type of the fog is often occurs around the rivers, lakes, and
swamps. An empirical equation that relates a visibility of liquid water content in the case of
advection fog is given by [34, 36]
LW  1660 1.43 , (65)
where LW is the liquid water content expressed in g/m3, and is the visibility expressed in
foot, 1 foot is equal to 0.3048 m. The Rayleigh approximation can be used to evaluate the fog
attenuation factor for frequencies that are specific for micro- and millimeter waves [36].
Thus, the fog attenuation factor can be presented in the following form [33], [35]:
 fog  LW
, (66)
where  fog is the fog attenuation factor expressed in dB/km, and  is the attenuation
coefficient expressed in dB/km per g/m3 at a specific operation frequency of radar sensor
system, Figure 41.
Figure 41. Fog attenuation factor as a function of operation frequency.
Figure 42 presents the detection performance of the CA-CFAR detector under the
atmospheric plus fog attenuation factors. As we can see from Figure 42, the heavy and serious
types of fog are the reason of dramatic degradation in the detection performance. Other fog
types have not considerable effects on the detection performance. It is difficult to distinguish
the detection curves for the ideal case, the atmospheric attenuation, and the light, thick, and
dense types of fog. The visibility and attenuation factors are presented in Table 9 for different
types of fog.
Table 9. Fog densities with related visibility and attenuation factors
Fog Type Visibility [m] Attenuation [dB/Km]

Serious Fog 3 – 10 40 – 7.21
Heavy Fog 10 – 40 7.21 – 0.99
Dense Fog 40 – 70 0.99 – 0.45
Thick Fog 70 – 250 0.45 – 0.07
Moderate Fog 250 – 500 0.07 – 0.02
Light Fog 500 – 1000 0.02 – 0.01
Figure 42. Fog attenuation factor action on CA-CFAR detection performance.
9.5. Dust Attenuation
A widely used method to characterize the density of aerosols suspended in the

atmosphere or air is based on visibility of any target vehicle. In this case, the visibility can be
defined as the target vehicle range if a high contrast target vehicle would be observed clear.
Taking into consideration the Rayleigh approximation, the dust attenuation factor of radar
sensor systems can be presented in the following form [34, 37]:
2.317 10 3 1
 dust   [dB/km]
[(   2) 2   im
2
]  1.07
, (67)
where   is the real term of permittivity of the dust particle,  im is the imaginary term of
permittivity of the dust particle, and  is the wavelength. Figure 43 presents the relationship
between the visibility and attenuation factor in the case of coal dust [34].
Figure 44 demonstrates a comparison between the water attenuation factor and coal dust
attenuation factor. From Figure 44, we can see that an action of the dust attenuation factor is
very weak in comparison with the water attenuation one [34].
Figure 45 presents the detection performance of the CA-CFAR detector for the ideal case,
the atmospheric attenuation factor, and the atmospheric plus dust attenuation factors. The dust
attenuation factor is varied in accordance with variations in visibility from 0.5m to 50m. As
follows from Figure 45, the high dust density conditions are the reason of small visibility
ranges (less than 1 m) causing a high deterioration in the detection performance.
Figure 43. Visibility as a function of the attenuation factor for coal dust.
Figure 44. Visibility versus the attenuation factor for coal dust and water droplets.
Figure 45. The dust attenuation factor action on the CA-CFAR detection performance.
In conclusion, we can state that the rain and fog attenuation factors, especially the high
rainfall and heavy or serious fog, should be considered as a serious reason of deterioration in
the detection performance of the CFAR detectors and any other ones. On the other hand, the
atmospheric and the dust attenuation factors have less effect on the detection performance. In
practice, the worst attenuation conditions should be considered. Thus, a whole set of
attenuation factors must be taken into consideration, even the attenuation factors with weak
action, because an accumulation of all weak attenuation factors can generate a considerable
CONCLUSION
Recently, the total safety driving approach has been become very popular. Any safety
driving application can be implemented by installing radar sensors to build a system with
passive and active vehicle safety elements. This system compensates the errors made by
drivers and avoids an occurrence of accidents. Implementation of the controller area network
and radar sensors to construct safety driving applications and collision avoidance systems is
an effective solution in the automotive industry.
The FMCW radar sensor system is chosen to be used for both the MRR application such
as CVD and the SRR application such as BSD to provide the controller area network with the
required data, for example, the range, relative velocity, and target vehicle azimuth angle or
position.
The main problem in any radar sensor system is to detect the target return signal within
the limits of system scanning or observation area. The target return signal detection
techniques depending on detection threshold in radar sensor systems have a major
disadvantage with respect to sensitivity of noise power. The detection threshold should be
defined based on the observed in real time noise power when the noise variance is varied as a
function of time. Thus, the adaptive detection threshold is essential in order to maintain the
constant probability of false alarm and stable detection performance with the predictable
probability of detection. Simulation and detection performance analysis are carried out under
conditions closed to practice with the purpose to evaluate the mentioned CFAR detectors
employed by LFMCW radar sensor system when the interfering target vehicles are absent or
present.
The CA-CFAR detector is the optimal among all other CFAR detectors in the case of
homogeneous noise and absence of interfering target vehicles (no target masking
phenomenon). The OS-CFAR detector suffers a minor degradation in the detection
performance in comparison with CA-CFAR detector. In multiple target scenarios and in the
case of non-homogeneous noise the OS-CFAR detection performance is superior in
comparison with the CA-CFAR detector when we know a priori the number of interfering
target vehicles. The GCML and ACGO-CFAR detectors maintain a good robustness when the
number of interfering target vehicles is known a priori. Additionally, the ACGO-CFAR
detector is capable to suppress the clutter edge false alarm. The GCML detector has the best
detection performance in non-homogenous noise, but the main disadvantage of this detector is
the great complexity and computation cost, as in the case of ACGO-CFAR detector. Thus, the
OS-CFAR detector can be considered as the best tradeoff between the detection performance
and complexity (low computation cost).
The GD is constructed based on the generalized approach to signal processing in noise.
The engineering interpretation of the GD is a combination of the correlation receiver that is
optimal in the Neyman-Pearson sense under detection of signals with a priori known
parameters and the energy detector that is optimal in the Neyman-Pearson sense under
detection of signals with a priori unknown parameters. This combination allows us to take
into consideration a very important statistical parameter such as the likelihood function
variance and formulate a decision-making rule with respect to the presence or absence of the
target return signal based on definition of the jointly sufficient statistics of the mean and
variance of the likelihood function.
Noise power estimation procedure based on the sliding window technique with cell
averaging is presented under the GD implementation in radar sensor systems. In order to
evaluate a feasibility of GD implementation with the proposed noise power estimation
technique, the GD detection performance is compared with the CA-CFAR detector which is
widely used in practice and demonstrates the best detection performance among whole CFAR
detector family in homogeneous noise. GD overcomes the CA-CFAR detector by the
detection performance under both the homogeneous and non-homogeneous noise conditions.
Detection performance of GD is also better than that of other CFAR detectors in non-
homogeneous noise when the bandwidth of interfering signals is matched with GD PF one.
The GD implementation allows us to reduce the SNR losses required to compensate the effect
of inaccurate noise power estimation (the estimation error) and achieve the desired constant
probability of false alarm PFA . The improvement in detection performance demonstrates an
effectiveness of the proposed solution based on the GD implementation. Employment of the
noise power estimation techniques can improve the noise power estimation quality (to reduce
the estimation error) and allows the GD to achieve the best detection performance. The
problem dealing with the interfering signals matched by bandwidth with the GD AF one leads
us to employment of an interference cancellation approach such as the beam forming and

antenna array with or without direction of arrival estimation to eliminate the interference
action on the GD performance.
Under construction of driving safety systems based on radar sensor network a whole set
of attenuation factors must be taken into consideration, even the attenuation factors with weak
action, because an accumulation of all weak attenuation factors can generate a considerable
ACKNOWLEDGMENTS
This research was supported by SL Corporation and Industry Academic Cooperation
Foundation of KNU within the limits of the research project on design of signal detection
algorithms for car applications, Grant No. 20101459000 and Kyungpook National University
Research Grant 2012.
REFERENCES
[1] Shbat, M. S.; Khan, Md. R.; Yi, J. H.; Lee, I. B.; Tuzlukov, V.; Modern Radar Systems
and Signal Detection Algorithms for Car Applications. Fizika, vol. XVII (1-en), pp. 29-
60, 2011.
[2] Hosaka, A.; Mizutani, H.; Improvement of Traffic Safety by Road Vehicle Cooperative
Smart Cruise Systems. IATSS Research, vol. 24, no. 23, pp. 4-42, 2000.
[3] Bloecher, H.; Sailer, A.; Rollmann, G.; Dickmann, J; 79 GHz UWB Automotive Short
Range Radar - Spectrum Allocation and Technology Trends. Advances in Radio
Science, vol. 7, pp. 61-65, 2009.
[4] Mende, R.; Rohling, H.; New Automotive Applications for Smart Radar Systems.
German Radar Symposium (GRS), Bonn, Germany, pp. 35-40, 2002.
[5] Schneider, M.; Automotive Radar - Status and Trends. German Microwave Conference
(GeMic), Ulm, Germany, pp.144-147, 2005.
[6] Australian Communications Authority; A Review of Automotive Radar Systems -
Devices and Regulatory Frameworks. Document: SP 4/01, 2001.
[7] Kulpa, K.; Continuous Wave Radars - Monostatic, Multistatic and Network. NATO
Security through Science Series, Warsaw, Poland, vol. 2, pp. 215-242, 2006.
[8] Wenger, J.; Automotive mm-Wave Radar: Status and Trends in System Design and
Technology. IEE Colloquium on Automotive Radar and Navigation Techniques,
London, England, Ref. No. 1998/230, pp. 1-7, 1998.
[9] Draft International Standard ISO/DIS 17387. Intelligent Transportation Systems - Lane
Change Decision Aid Systems - Performance Requirements and Test Procedures.
International Organization for Standardization, 2006.
[10] Etschberger, K.; Controller Area Network, Technical Report. 2nd Ed., Munich, IXXAT
Automation GmbH, 2011.
[11] Skolnik, M.; Radar Handbook. 3rd Ed., New York, USA, McGraw Hill, 2008.
[12] Skolnik, M.; Introduction to Radar Systems. 3rd Ed., USA, McGraw-Hill Book
Company Wiley-Inter-Science, 2001.
[13] Isakov, V; Microwave Circuit for 24 GHz Automotive Radar in Silicon Based
Technologies. New York, USA, Springer, 2010.
[14] Mahafza, B. R; Elsherbeni, A. Z; Matlab Simulation for Radar Systems Design.
Chapman & Hall/CRC CRC Press LLC, 2004.
[15] Richards, M. A.; Fundamentals of Radar Signal Processing. New York, USA McGraw-
Hill, 2005.
[16] Finn, H.; Johnson, R.; Adaptive Detection Mode with Threshold Control as a Function
of Spatially Sampled Clutter-Level Estimates. RCA Review, pp. 414-464, 1968.
[17] Kay, S.; Fundamentals for Statistical Signal Processing, Vol. 1: Estimation Theory. NJ,
USA, Prentice Hall, Englewood Cliffs, 1993.
[18] Rohling, H.; Radar CFAR Threshold in Clutter and Multiple Target Situations. IEEE
Transactions on Aerospace and Electronic Systems, vol. 19, no. 4, pp. 608-621, 1983.
[19] Himonas, S. D.; Barkat, M.; A Robust Radar CFAR Detector for Multiple Target
Situations. IEEE National Radar Conference, Dallas, Texas, USA, pp. 85-90, 1989.
[20] Himonas, S. D.; Adaptive Censor Greatest-of CFAR Detection. IEE Proceedings-f,
Radar and Signal processing, vol. 139, no. 3, pp. 247-255, 1992.
[21] Swerling, P.; Probability of Detection for Fluctuating Targets. IRE Transactions on
Information Theory, IT-6, pp. 269-308, 1960.
[22] Rohling, H.; Some Radar Topics: Waveform Design, Range CFAR and Target
Recognition. Advances in Sensing with Security Applications NATO Security through
Science Series, vol. 2, pp. 293-322, 2006.
[23] Yi, J. H.; Lee, I. B.; Shbat, M. S.; Tuzlukov, V.; 24 GHz FMCW Radar Sensor
Algorithms for Car Applications. International Radar Symposium (IRS 2011), Leipzig,
Germany, pp. 465-470, 2011.
[24] Shbat, M. S.; Tuzlukov, V.; Generalized Approach to Signal Processing in Noise for
Closing Vehicle Detection Application Using FMCW Radar Sensor System.
International Radar Symposium (IRS 2011), Leipzig, Germany, pp. 459-464, 2011.
[25] Shbat, M. S.; Tuzlukov, V.; Signal Processing in Automotive Controller Area Network
Based on Radar Sensors. 11th International Conference on Control, Automation, and
Systems (ICCAS 2011), Gyeonggi-do, South Korea, pp. 616-620, 2011.
[26] Klotz, M.; An Automotive Short Range High Resolution Pulse Radar Network Aachen.
Germany, Shaker Verlag GmbH, 2002.
[27] Tuzlukov, V.; A New Approach to Signal Detection Theory. Digital Signal Processing:
A Review Journal, vol. 8, no. 3, pp. 166-184, 1998.
[28] Tuzlukov, V.; Signal Detection Theory. Boston, Springer-Verlag, 2001.
[29] Tuzlukov, V.; Signal Processing Noise. Boca Raton, London, New York, Washington
D.C CRC Press, Taylor & Francis Group, 2002.
[30] Shbat, M. S.; Tuzlukov, V.; Noise Power Estimation under Generalized Detector
Employment in Automotive Detection and Tracking Systems. 9th IET Data Fusion and
Target Tracking Conference (DF&TT‟12), London, UK, doi:10.1049/cp.2012.0416,
2012.
[31] Shbat, M. S.; Tuzlukov, V.; Generalized Detector with Adaptive Detection Threshold
for Radar Sensors. 13th International Radar Symposium (IRS 2012), Warsaw, Poland,
pp. 91-94, 2012.
[32] Tuzlukov, V.; Signal Processing in Radar Systems. Boca Raton, London, New York,
Washington D.C CRC Press, Taylor & Francis Group, 2012.
[33] ITU-R P.676-6 Recommendation, Attenuation by Atmospheric Gases, 2012,

http://www.itu.int/rec/R-REC-P.676/en.
[34] Brooker, G.; Sensors and Signals. Australian Centre for Field Robotics, 2006.
[35] Mie, G.; Contributions to the Optics of Turbid Media, Especially Colloidal Metal
Solutions. Annals of Physics, vol. 330, no. 3, pp. 377-445, 1908.
[36] Zhao, Z.; Wu, Z.; Millimeter-Wave Attenuation due to Fog and Clouds. International
Journal of Infrared and Millimeter Waves, vol. 21, no. 10, pp. 1607-1615, 2000.
[37] Goldhirsh, J.; Attenuation and Backscatter from a Derived Two-Dimensional Dust
Storm Model. IEEE Transactions on Antennas and Propagation, vol. 49, no. 12, pp.
1703-1711, 2001.
Chapter 2
AIRBORNE ULTRASONIC IMAGING:

SONAR BASED IMAGE GENERATION FOR
AUTONOMOUS VEHICLES
Dorel Aiordachioaie*, Laurentiu Frangu, Silviu Epure,

Radu Belea, Marius Mazarel, Constantin Miholca
and Claudiu Chiculita
Electronics and Telecommunications Department, “Dunarea de Jos”
University of Galati, Galati, Romania
ABSTRACT
Ultrasound imaging refers to images obtained by ultrasonic waves. An emitted pulse
is reflected back by the environment and the pixel‟s intensity is imposed by the amplitude
of the reflected pulse at the place where the image is constructed.
Ultrasound images have various and important applications both in industry, for fault
detection and localization, and medicine, for non-invasive diagnostic techniques. A new
area is coming now in robotics for artificial vision, for solving some general tasks as
recognition and navigation.
In robotics, navigation based on ultrasonic images is a new field of research,
especially when discussing about airborne ultrasound images, i.e. images obtained by
ultrasound waves propagating in air.
The reason of such delay in using airborne ultrasound images in robotics is explained
by the high absorption of ultrasound frequencies in air, where frequencies over 200 kHz
are not often used. In this front line, new technologies for image generation and
processing are expected and welcomed by the market and users of mobile robots.
We have used simple sonar heads, which have a biomimetic functional behavior with
one emitter and two receivers. The sonar heads have six degrees of freedom and are
named ROVIBAT (ROmanian VIrtual BAT).
This chapter presents some results concerning sonar image generation process by
using a fixed ultrasonic frequency. The chapter is organized in sections. Section 1
*
Corresponding author: dorel.aiordachioaie@ugal.ro.
58 Dorel Aiordachioaie, Laurentiu Frangu, Silviu Epure et al.
presents the rationale to use ultrasonic images in the context of smart vehicles. Section 2
describes the structure of sonar heads. Section 3 presents the electronics of transmitter
and receivers. Section 4 describes the effect of external pinnae mounted on the front of
receivers. Section 5 deals with the image generation and raw processing methods for
navigation purposes, including detection and recognition tasks.
Keywords: Sonar, Ultrasonic Imaging, Navigation, Signal Processing, Image Generation
1. INTRODUCTION
Ultrasound imaging points refers to 2D or 3D images obtained by ultrasonic waves. The
generation technology uses an emitted pulse through target and which is reflected back. The
pixel‟s intensity is imposed by the amplitude of the reflected pulse at the place where the
image is constructed.
Ultrasound images are used both in industry, for fault detection and diagnosis, and
medicine, in diagnosis procedures. All these applications do not use air as main propagation
medium.
Applications that use air as propagation medium have a major difficulty; the absorption
of ultrasounds waves is very high. This makes inefficient frequencies over 200 kHz.
Ultrasound imaging devices uses arrays of ultrasound transducers, in order to control the
beam width and thus to select the details in the explored environment. The ultrasound
transducers are accompanied by powerful signal processors. This is an expensive and
complex solution, and – some time – for simple applications discrete ultrasound transducers
could offer an acceptable and still efficient solution based on classical method of image
generation, based on pixel-by-pixel generation. The price for simplify is the appearance of the
uncertainty in ultrasound imaging and the generation time could be huge for many
applications.
The generation of airborne ultrasonic images is a challenge, the scientific literature and
practical experiments are almost missing. This work uses a simple method for obtaining
airborne ultrasound images, under the objective of target detection and classification. The
objective requires mainly the existence of information about the shape of the objects. The
methodology of using ultrasonic images for navigation is not dealt with here, by considering
that any classical method of image processing from computer vision field could be a valid
solution.
1.1. The Context of Smart Vehicles
In the framework of smart grounded vehicles (or cars), ultrasound imaging could be used
in common applications, e.g. navigation and self-orientation, [1], and special conditions (e.g.
dark or smog environments), i.e. where other well-known and used methods fail. Ultrasound
imaging could also be used in special applications related to pattern recognition and
classification, e.g. texture of surfaces. These are also valid for robots driven inside closed or
open environments, but with a finite set of trajectories, e.g. robots working in warehouses.
Airborne Ultrasonic Imaging 59
Reference [2] presents another point of view, related to the use of imaging as part of the
parking building for parking monitoring. „Detection methods based on cameras and image
processing suffer from a lack of accuracy and can be affected by environmental or weather
circumstances‟, [2]. The ultrasonic image system could improve the performances of camera-
based systems and, in bad environmental conditions, to solve detection problems.
During the past ten years, many research projects on smart cars were considered
worldwide. Most of them approach new technologies to implement automation and real time
software in the components of the car. We here mention two projects: (i) The City Car of
Massachusetts Institute of Technology (MIT) [3], a lightweight electric vehicle. This car
employs fully integrated in-wheel electric motors and suspension systems, which are self-
contained, digitally controlled, and reconfigurable; (ii) The Cybercars project develops road
vehicles with fully automated driving capabilities. [4]. More details could be found in [5] and
[6], which introduce the structure of the smart car and discuss the obtained major results during
projects, worldwide.
For safety reasons mainly, many cars have smart (intelligent) systems in order to assist
driver in driving, e.g. collision avoidance system (CAS), side blind spot warning, lane
departure control [7], automated switching of lights, park assistance systems, radar [8],
pedestrian detection [9], etc. All these systems solve specific tasks, independently or driver
enabled.
A smart car means something more, a car with systems, which can solve, in an automated
way, various common tasks from its life, such as automated choice of the right road to reach an
imposed destination (autopilot), antilock braking system (ABS), automated parking, etc. These
tasks are performed independently and in an automated way.
Autonomous vehicles refer mainly to vehicles without any driver. All necessary actions,
beginning with motor starting and continuing with safely driving on roads, are solved by the
on-board systems of the vehicle.
Finally, it is useful to say that all sensors, inside and outside the vehicles, take into account
both to the behavior of the car and driver, and - more – there it is an interaction with players of
the roads, i.e. pedestrians and other cars. A short presentation of these aspects is made in [10].
Taking into account the various atmospheric phenomena during driving and the necessity
to cover almost all driving conditions imposed by environment, i.e. environment with a very
high intensity of light or – the other way around – with a very low light intensity (dark
environments), special techniques for orientation and navigation must be considered. Airborne
ultrasonic imaging could be one of them, which helps directly the car if the sonar head (which
generates the ultrasonic images) is installed on it. The help can be indirect if the sonar head is
working outside the car, in fixed positions, as in parking buildings.
2. THE SONAR HEAD: STRUCTURE AND BASIC ELECTRONICS

The common sonar heads used in robotics, mobile or fixed, use frequency in the range of
40 to 50 kHz and with ultrasonic transducers in fixed positions. These have relatively poor
performances in the navigation of the robots especially in environment with walls and
complex obstacles.
The biomimetic sonar heads mimic the anatomy of bats: the transmitter corresponds to
mouth and the two ears correspond to the two receivers, left and right. The biomimetism is
double, in the sense the behavior of bats is translated to similar functions of the biomimetic
sonar head.
Biomimetism is used in many navigation and optimization algorithms, particularly in the
navigation of robots as in [11], [12] or [13]. Biomimetic sonar heads are already available, as
in [14], [15], and [16]. The basic hardware contains servomotors for the orientation of the
transducers and signal processing boards, with parallel and synchronously inputs. Classical
solution use external (outside of the sonar head) signal processing and control boards, which
could generate constraints on the quick movement of the robot.
Figure 1 presents the structure of the used biomimetic sonar head. The biomimetism
refers to the level of functions and to the degree of freedom, i.e. one emitter and two
receivers. The investigated sonar head has all the processing boards (signal plus control)
inside the sonar head. The sonar head is called ROVIBAT, an acronym from ROmanian
VIrtual BAT. Figure 2 shows sample pictures of two versions of the sonar head.
Four main blocks can be seen in the picture: (1) the mechanics of the head, i.e. mainly
bearings and servomotors (SONAR head); (2) the position control board. It computes the
driving signals for the orientation of each transducer; (3) the signal processing board. It
makes analogous and digital filtering, sampling and analog to digital conversions; (4). a
standard computer for coordination and supervised control.
Figure 1. The structure of the sonar heads.

2.1. The Position Control Board
The Voltage Regulator protects the equipment from over voltage and provides steady
regulated power. The weight is 1.4 oz.; the working voltage range is 4.8-8.0 V, and the
current up to 15A. Input voltage must be at least 1.0 V greater than the intended output
voltage. The regulator is ACC234 rated for 20 Watts of power dissipation. Other key features
are available on technical sheet [17].
Common pulse proportional servos are used for positioning the shaft. For a range of 90o
motion pulses in the range of 0.9 to 2.1 ms are transmitted, with repetition every 20 ms. a
specialized electronic board (SSC-32), [18], is used for servomotors driving. From
technological point of view, SMT (Surface Mount Components) were used. The electronics of
the receiver (amplifier) is placed under the transducers of the receiver. The gain of the
amplifier is around 50-55 dB and the frequency cut-off around 50-55 kHz.
Figure 2. Overall pictures of the ROVIBAT_01 and ROVIBAT_02 sonar heads.
2.2. The Signal Processing Board
The signal processing board has ADC (Analog to Digital Converter) (AD7266). The
sampling frequency is 2MS/s and 12 bits for resolution were used. The circuit is interfaced to
the Blackfin processor through a SPORT (Synchronous Serial Peripheral Port) port. The
digital signal processing is performed by specialized module based on dual core BF561
processor from Analog Devices. The working frequency is 600 MHz and provides 1200
MMACs (Million Multiply Accumulate Cycles per Second).
Figure 3 shows a block diagram of the board and an image of the signal processing board.
A special module for ultrasound transducers (or ultrasound sensor array) is considered outside
of the signal processing board in order to assure the modularity structure of the sonar head.
Sensor Array
Amplifiers and Filters
5V Emitter Circuitry
AD 7266 AD 7266
linear
Converter Converter
2x2MSPS 2x2MSPS
4 4
SPORT SPORT
3.3V
Emitter
switching
control
BF 561 v2 I/O
Dual Core
600 MHz
USB 64MB RAM GPIOs
UART
PC adapter SPI
PPI
JTAG video
camera
interface
Figure 3. Block diagram and image of the signal processing board.
The boards of signal processing and position control are driven by MATLAB software. It
receives commands from computer regarding the number of emitted pulses for transmission
and the sampling rate for acquisition. The board then drives the emitter and acquires signals
from ultrasonic receivers for a limited duration. Data from ADC circuits is transferred to
SDRAM (Synchronous Dynamic Random Access Memory) using DMA (Direct Memory
Access). Currently, data processing is made by PC with MATLAB software. After the
algorithms are developed and tested, the software can be embedded into the board for real
time processing.
3. ULTRASONIC TRANSDUCERS AND CONTROL CIRCUITS

Wide frequency ranges are more suitable for exploration of the environment, mainly
because of the variable shape of the transmitted pulses, adapted to transmission channels, and
- later on - it is possible to combine (aggregate) discrete ultrasonic images at various
frequencies. We are interested in wide range frequency transducers, namely capacitive or
piezoelectric transducers, which could be used up to 300 kHz.
Based on results reported in literature of wide range ultrasonic sensors, e.g. [19, 20, 21]
for general sensing application and [22, 23, 24] for ultrasonic transducers, the EMFi (Electro
Mechanical Film) material, [25], was considered.
Experiments with sensors based on these new materials show a strong relation between
the frequency range, including the frequency resonance, and the mechanical structure of the
ultrasonic transducer. Constraints are imposed also by the connecting (gluing) technology of
the terminals. Now of working, this technology is not mature to provide reliable ultrasonic
transducers.
By considering the parameters of the EMFi material, [25], as sensitivity of Sq = 25-250
[pC/N] and the specific capacitance of 40 [pF/cm2], and by using small loads for transducer,
i.e. resistors with resistance greater than 1-5 MΩ, the output voltage for a P [N/cm2]
pressure variation is:
V  (1/ C0 )  Sq  P (1)
The minimum necessary pressure at the reception point is
P  V  C0 / Sq
(2)
In terms of radiation, as a model, the simplest type of source is the pulsating sphere.
Assuming that the radius a of the sphere is small compared to the wavelength λ of the
emitted sound wave, the pressure field can be expressed as
p( r , t )  j  0  c  U 0   k  a  e 
a j  t  kr 
r (3)
where: U0 is the speed amplitude of the vibration; an - the average radius of the sphere;  0 -
the mass density of an undistributed fluid element.
A source of particular interest is the rigid piston of radius a mounted flush with the
surface of an infinite baffle, and vibrating with a time harmonic motion. The final expression
for the pressure of a single source at the point placed of distance r and θ angle is:
0  c  U 0  ka 2 j  t kr   2 J1 (ka sin  ) 

p ( r , , t )  j  e    (4)
2r  ka sin  
where J1 represents the Bessel-function of the first kind of order 1.

An ultrasonic transducer with EMFi material was built. Final and intermediary stages of
the transducers are presented in Figure 4. A double-sided PCB is used. Both sides are covered
by EMFi. One side is covered by graphite (Figure4.b). By connecting it to a ground plane, an
electromagnetic shield is obtained. There are some advantages as minimizing the capacitive
coupling between active electrode of the transducers and receiver, and – another advantage –
the user is protected over high voltage of the active electrode.
(a) (b) (c)
Figure 4. Custom ultrasonic transducer: (a) PCB; (b) PCB and EMFi; (c) Final product.
EMFi based transducers with different active areas have been built and tested for
transmitter (4 cm2) and for receiver (1.5 cm2). The main diagrams for these transducers are
presented in Figure 5. A wide frequency range centered on 250 kHz and a low width of the
main beam were obtained. The diagram was measured at 280 kHz.
(a) (b)
Figure 5.(a) Frequency response of the transducer-air-transducer path; (b) Directivity diagram at 280
kHz.
3.1. The Receiver
Reference [26] reported a structure of the receiver designed for a sensibility of 100 V
and composed of two blocks: a preamplifier with fixed gain (1897 amplification) followed by
a variable gain stage.
In this work, the designed receiver is built on three layers, 20x20mm PCB, as presented
in Figure 6. One layer is used as electrode for EMFi foil, one for input preamplifier
components, and the middle one is used as a shield. The resulting receptor is small, compact
and robust to external perturbations.
(a) (b)
(c)
Figure 6. Layout of the receiver: (a) components layout; (b) EMFIT sensor layout; (c) cross-section of
PCB.
An octagon EMFi foil is glued on the PCB, which forms the bottom electrode, and the
top electrode is obtained with a layer of graphite, sprayed over the foil. The top electrode is
connected to the ground so that electrostatic and capacitive perturbations are minimized.
Figure 6 presents some details from the receiver circuit and ultrasonic transducer.
Test version of the receiver circuit, Figure 7, is based on AD8022 dual high speed, low
noise operational amplifier from Analog Devices, which offers a frequency of 130 MHz, at
unity gain. Input bias currents, typical 2 μA, are compensated with R8, R4, and C3, under
classical configuration. Resistor R1 can be used to disconnect EMFi sensor, to separately test
the amplifier. A value of 170 kΩ is obtained as input impedance, by considering R3 = 2.2 M
Ω, R2 = 220 kΩ, R4 = 680 kΩ, and 20 kΩ the differential input impedance for AD8022.
A bigger input differential impedance Zidif will conduct to higher input impedance, and
necessarily to a better sensibility of the receiver amplifier. The preamplifier gain is about 660.
The first Operational Amplifier (OA) circuit amplifies signal by 11, and the second one by
60. This will lower the upper frequency of the bandwidth to 2 MHz. The capacitor C2 is used
to limit this value further, to 500 kHz.
By testing the preamplifier, it results flexibility in gain and bandwidth around the
necessary values and robustness to noise signals. The input impedance could be higher and a
pin-compatible op-amp with JFET inputs will be tested in a further version.
From technological point of view, the electronic circuit is mounted on dual-layer PCB,
one layer used as ground plane. The transducer and the preamplifier of the receiver have the
same size and placed back-to-back.
Figure 7. Preamplifier schematics.

3.2. The Transmitter
The EMFi based transducer needs high voltages to achieve high Sound Pressure Levels
(SPL). For example, in [22], signals of about 300 Vpp were used. Three circuits have been
tested: (1) a hybrid power amplifier with transformer; (2) a bootstrap high voltage amplifier;
(3) an active load A-class H bridge amplifier.
The first one is a circuit that includes an operational amplifier and other active
components connected in such a way that the resulting circuit will exceed one or more op-
amp IC (Integrated Circuit) limits. Largely, the operational amplifier entity refers to a
general-purpose DC voltage amplifier that may be configured by a passive network.
The schematic in Figure 8 consists of IC (Integrated Circuit) and a configuration network
from feedback resistor R2, ground resistor R1 and in-loop compensation condenser C2. The
circuit amplifier can drive a 150 [pF] EMFI transducer using a 1:30 rising voltage
transformer. The IC is a 3 MHz bandwidth op-amp (TL071). The power transistors T1 and T2
are BD439/BD 440. The amplifier output voltage V0 is applied to the load capacity CL via the
rising voltage transformer, which generates a voltage up to 700-750 Vpp. Even with no
problems of stability, a capacity value C2 must be considered by experimenting.
Figure 8. High voltage amplifier with transformer.
The second solution, the bootstrap high voltage amplifier has feedback loops for supply
voltages, in order to increase the range of the output voltage. There have been used high
voltage transistors (ex. STN83003, [27], STP 93003, [28]), faster op-amp (ex. LM7171, [29]),
the supply voltages can be increased at ±180 V and the output signal will vary up to 336 V.
There are many sonar systems with exploring signals as short bursts, i.e. rectangular
pulses. These signals need high quality amplifiers at the output stage. The last circuit
considers a high quality amplifier. The circuit operates between “stand-by” and “burst
emission” modes.
The considered circuit is presented in Figure 9. Four stages could be considered: (1)
voltage-current converter: the op-amp, T2, D2, and R2; (2) active load: T4, R8, and R4; (3)
voltage divisor: RB1, TB and RB2; (4) electronic switch K that connects the resistor R6, in
parallel with R2 when ultrasonic burst is emitted We reduce the power dissipation between
two successive bursts, by imposing amplifier work in two modes: (1) Burst-Emission: when
the contact K is close; (2) Stand-By: when the contact K is open.
(a)
(b)
Figure 9. (a) The high quality amplifier; (b) the output characteristics of T2 transistor.
4. NARROW DIRECTIVITY BY EXTERNAL PINNAE

Biological studies show the effect of the pinnae in environment living of animals, e.g. the
bats. The external shape of the ear has many non-uniformities and the surface is slightly
mobile and flexible. Such a set of features provides a sensibility of hearing and the possibility
to precisely localize various targets on the hunting field. This is an adaptive behavior of bats,
which requires, by mimetism, an adaptive working regime for the biomimetic sonar heads,
related to frequency and shape of the transmitted signals. From the receiving point of view,
adaptation means the possibility of slightly changing the shape and the orientation of the
external ears.
In the context of the autonomous vehicles and sonar-based systems, a first point to do is
to implement adaptation of the beam of directivity to the details of the environment. There are
some examples, where [30] and [31] made simulations and experiments related to the
optimum shape of the pinnae. Simulation based studies show that the best shape of artificial
made pinnae is a kind of paraboloid, in fact a portion of the paraboloid. One of the success
examples was also within CIRCE research project, [14].
This section describes some results from the measurements processes of pinnae.
Ultrasonic transducers working on fixed frequency (40 kHz) were used, by removing the
housing of the transducer (MA40S5), [32]. Data are recorded by an automated ultrasonic
measurement system, as it is described in [33].
Figure 10, [33], shows an example of directivity charts, in horizontal plane for: (1)
original transducer with no pinnae; (2) transducer with no housing or pinnae; (3) transducer
with small pinnae; (4) transducer with big pinnae. Clearly, the bigger the ratio between the
size of the paraboloid and the size of the transducer, the better directivity is.
The explored ears use one-half of the paraboloid. The effect is an asymmetry in vertical
plane, as presented in Figure 11: (1) transducer without parabolic surface; (2) transducer with
small pinnae; (3) transducer with big pinnae. In this last case, a secondary lobe is appearing,
which later will introduce distortions in ultrasonic images. More details are in [33]. The
experimental measurements show good estimations of theoretical computed values. By using
such reflectors, the width of the main lobe of 20° is available. Also, the sensitivity is
improved with an order of 3 dB.
Figure 10. Directivity charts, horizontal plane, various configurations.

Figure 11. Directivity charts, vertical plane, various ear sizes.
5. ULTRASONIC IMAGE GENERATION AND RAW PROCESSING

Airborne ultrasound imaging is difficult to develop because of the high absorption of
ultrasound waves in air. Thus, the generation of such images is a very difficult task and there
is a need for new methodologies and technologies that could solve the above problem.
Accordingly, this section introduces airborne ultrasonic image generation and processing
methodology.
In the field of mobile robots, map building for navigation purposes is solved by using
sonar systems, and implementing signal processing and pattern recognition algorithms, e.g.
[34, 35, 36] for detection and localization of reflectors, or [37, 38] for feature extraction. In
medicine, ultrasound images are the result of echography, providing an efficient method for
diagnosis and right treatments.
Industrial applications are mainly used to detect non-uniformities and fabrication faults
inside materials, e.g. [39, 40]. A quite new application is through-wall sensing, which helps to
see inside structures of buildings, to identify the materials within building, e.g. [41, 42, 43].
The modern way for generation of ultrasound imaging is based on an array of ultrasound
transducers, e.g. [44, 45]. The solution is quite complex, requiring high power signal
processors and memory.
Sometimes, solutions using discrete ultrasonic transducers must be considered, at least

from the point of view of cost and reliability. The approach of generating ultrasound images
is based on pixel-by-pixel basis, each pixel being associated with a pair of signals
(transmitted and received). The main problem is related to the processing time necessary to
obtain the echoes signals. The greater the image‟s size, the greater the generation and
processing time.
The following sub-sections focus on the main functions in order to generate all signals
involved in ultrasonic image generation process.
5.1. Control Functions
The sonar head works by executing specific tasks, which are described by a set of general
functions. These are used to orient the sonar head to target‟s position in the 3D space. The
basic function must orient the emitter and the receiver to the position of estimated or expected
target. Depending on the type of the sonar, active or passive, the orientation function must be
synchronized with other tasks, e.g. for transmitting signals in the case of active sonar. In the
case of passive sonar, the maximum amplitude of the received signal will impose the
orientation of the sonar head.
Let target point be denoted by T ( x, y, z )  T (r , , ) . The pseudo code for target
catching is described below and contains all commands and parameters to properly drive the
servomotors.
Function Orient_To_Target (FOT)

#1: Initial data (arguments):
- Target_Azimuth;
- Target_Elevation;
- Initial positions of servos: // the middle of the range
P0_ini ; P1_ini;
P4_ini ; P5_ini ;
#2: Initializations: //all servos
#0 P P0_ini #1 P P1_ini
#4 P P4_ini #5 P P5_ini
#3: Positioning: // emitter and receivers to target
#0 P<Target_Azimuth> // set the azimuth of transmitter
#1 P<Target_Elevation> // set the elevation of transmitter
#4 P<Target_Azimuth> // set the azimuth of left receiver
#5 P<Target_Azimuth> // set the azimuth of right receiver
End FOT.
In order to use the right notations and usage of the involved signal, Figure 12 introduces
the model of the active sonar system. A digital signal source generates the signal x(t ) . This
is amplified and - at the output of the ultrasonic transducer- the signal s (t ) is obtained. The
medium between target and emitter, considered time-invariant and linear, is described by
impulse response functions hd (t ) and hR (t ) . If n1 (t ) and n2 (t ) are the specific noises, then
the processed signals are s1 (t ) and s2 (t ) :
s1 (t )  s(t )  n1 (t ) (5)
s2 (t )  s(t )  hD (t )  hT (t )  hR (t )  n2 (t ) (6)
Figure 12. The model of the sonar system.
The received signal, s2 (t ) , has two representations obtained by the left and, respectively,
the right ear: s2 L (t ) and s2 R (t ) . The signals n1 (t ) and n2 (t ) are noise or assimilated as
noise signals (e.g. quantization errors). The operator “  ” means convolution and it is
accepted as long as blocks from transmission path are linear.
The basic function involved in signal acquisition is considered. The initial parameters are
related to the transmitted impulses, i.e. the number of pulses. This is proportional with the
distance to target. Common values could start from 4 to 8 pulses for distances up to 1 m, and
might go to 32-50 pulses for longer distances (2-3 m). The acquisition and storage of raw
samples start after a delay interval, D, which depends of distance to target.
Function Get_Data(FGD)
#1: Initial parameters (arguments):
- width TX of the impulse x(t); // pulse wave
- delay D; // depends of distance to target
#2: Generate x(t);
#3: Transmit s(t); // output of transducer
#4: Wait(D);
#5: Read signals s(t), s1(t) and s2(t). // sampling and recording
End FGD.
Data stored locally, in the local memory of the sonar head, is transmitted via USB
(Universal Serial Bus) port to a central computer. Here, the digital signal is arranged in
specialized arrays and processed by the high-level simulation languages, e.g. MATLAB.
Below, the pseudo code for transmitting data from sonar head to central (personal) computer
is presented.
Function Transmit_Data (FTD)

#1: Initial parameters (arguments):
- Speed; // USB port
- Protocol; // USB port
- Data // USB port
#2: Open port;
#3: Make connection;
#4: Data Transmission;
#5: Close port;
EndFTD.
The final function will link the previous described function. This is presented below as a
procedure:
Procedure SONAR
#1: Target_Orientation();
#2: Get_Data();
#3: Transmit_Data();
#4: Process_Data();
End SONAR.
A small remark must be discussed, finally. Depending on the objective of environment

exploring, some different trajectories could be considered. In the context of image generation,
these could be considered as grids. Each grid is defined by points with angular coordinates 
and . Figure 13 presents three of such possible grids used in the exploration of the
environment or image generation task.
Figure 13. Examples of pattern motions, with fixed points, variable points and uniform grid.
5.2. Description of the Experiments
The objective of the experiments is to build airborne ultrasonic images, by considering

various types and sizes of targets. In this chapter, three objects are considered: (1) - a ball of
20 cm diameter; (2) - a box with an edge oriented to the sonar head; (3) - a box (the same as
in case #2) with a face oriented perpendicularly to the sonar head. The considered objects are
presented in Figure 14. The used sonar head is ROVIBAT-02, which has a Pan & Tilt
structure, and commercial Murata ultrasonic transducers. The distance between receivers was
d = 4.5 cm.
Image generation is made pixel-by-pixel. For each explored direction, the echoes are
stored and processed. The values of the main parameters involved in the experiments are:
sampling frequency, 450 kHz; exploring distance, 3 m; target distance, 1m; resolution of the
image, 20 by 20; azimuth range, 200 ; elevation range, 400 . The value of 00 corresponds
to the middle of the target. Each pixel is quantized with 8 bits.
As example, Figure 15 shows some echoes, exy(t), which correspond to directions of
coordinates (2,2), (2,15), (15,2) and (15,15). It is observing that the positions and amplitudes
of received echoes are changing in time.
Figure 14. The investigated objects.
Figure 15. A set of recorded echoes, in the case of box like target.
5.3. Ultrasonic Image Generation
Figure 16, [46], introduces the main computation blocks for ultrasonic image generation.
Each explored direction is characterized by the coordinates pair (x,y), each from the set of {1,
2,…, n=20}.
The computation of the correct values of x and y and the correct positioning are the tasks
of the position control board, which was described already in a previous section.
The image generation method is composed of two structural processing frames, which
means processing in hardware and processing in software, with more details in [46]. All
processing steps assigned to hardware running are related to the sonar head and consists of
orientation, emission of pulse signals, filtering and conversion of the received signals
(echoes). Data coming from sonar head are transmitted to the central (personal) computer for
intensive processing by MATLAB functions. On each path, left and right, the main
considered processing steps are: matched filtering, envelope detection, image construction,
and uncertainty compensation.
Images from left and right side are next fused. By fusion, a set of image transforms is
considered as registration, enhancement, and filtering. The key element of the image
constructor is a maximum value detector and storing, so a pixel of the ultrasonic image has
the maximum values of the envelope from a given exploration time (range) interval. Based on
the content of the images generated by the first method, object detection and size
discrimination algorithms are able to use and develop, e.g. in applications as obstacle
avoidance and object recognition.
Some blocks are represented with solid line. They implement complex processing tasks.
These are optional, in the sense that these could be replaced with more simple processing
functions. Matched filtering can be replaced by cross-correlation and uncertainty processing
could miss. Uncertainty processing block consider the effect of the uncertainty of ultrasonic
transducers on generated images. If the directivity of transducers is narrow, then this block
could be removed.
From the structure of the generation process of images, it could be observed that time is
used as a parameter, so the ultrasonic shape (images) of the object is also parameterized by
time. Once the image set is finished, all classical algorithms of image processing could be
applied, e.g. [47, 48].
Figure 16. Main processing blocks in the ultrasonic image generation process.
The obtained ultrasonic un-processed images are presented in Figure 17, [46]. Because
the ultrasonic transducers are located at a small distance from one another, about 5 cm, the
left and right images do not differ very much. Thus, a very simple fusion scheme was
considered here for the finally generated image, by averaging the left and right images.
Figure 17. Ultrasonic images for the considered cases (second method).
Figure 18, [46], shows an example of the uncertainty processing. The effect of using the
block is to introduce a local average. Mainly, a Gaussian response (weighting) surface is
considered for each element of the image. The variance of the surface is estimated from the
real directivity of the ultrasonic transducer.
CONCLUSION
Preliminary results of airborne ultrasonic image generation and processing were
presented. The objective of the work was to investigate the difficulties of ultrasonic image
generation with some new sonar heads, having portability, simplicity and biomimetism as
specific features.
Figure 18. Ultrasonic images with uncertainty compensation.
Two sonar heads were built, with a small and wide range of working frequency. The
present work considers only the small bandwidth (5-8 kHz) and the fixed working frequency
(40 kHz). The sonar heads were presented from hardware and software points of view. The
structure of the hardware is under portability (among various vehicles) and biomimetism (one
transmitter plus two receivers, and six degrees of freedom). Various solutions for narrow
beam of directivity were considered. A solution based on parabolic reflectors was presented
in this work.
From the imaging point of view, two processes were described: generation and basic
processing. Generation process considers a pixel of the image assigned to a pair of signals
(transmitted pulse plus received echo). Depending on the amplitude of the received echo, the
intensity of the pixel varies from black (zero intensity, corresponding to zero echo) to white
(the highest intensity of the received echo). Ultrasonic images have low resolution, 20x20 or
40x40. In both cases, generation time is high and other strategies should be considered in
order to obtain ultrasonic images in real time. Image processing process considers here fusion
of left and right images.
Ultrasonic images have poor information content, which is explained by the uncertainty
generated by the width of the main beam of ultrasonic transducers. Uncertainty must be
considered also for generation process, depending also on distance to target. The obtained
images are not very close to the images of real objects. This is common to ultrasonic field and
– sometimes – an external interpreter could help with the proper use of the image content.
Future research steps will consider a wide frequency range, from 40 to 200 kHz and the
fusion of ultrasonic images at various exploring frequencies (40, 100, and 200 kHz).
ACKNOWLEDGMENTS
This document contains partial results of ADBIOSONAR research Grant no. 12079/2008
under CNMP (UEFICDI) Romanian Authority.
REFERENCES
[1] Libelium; 50 Sensor Applications for a Smarter World. Get Inspired.
http://www.libelium.com/ top_50_iot_sensor_applications_ranking.
[2] Kianpisheh, A.; Mustaffa, N.; Limtrairut, P.; Keikhosrokiani, P.; Smart Parking System
(SPS) Architecture Using Ultrasonic Detector. International Journal of Software
Engineering and its Applications, vol. 6, no. 3, pp.51-58, 2012.
[3] MIT; City Car, USA, 2012, http://www.media.mit.edu/news/citycar.
[4] Cybercar. 2006,http://www.cybercars.org/.
[5] Sun, J; Wu, Z; Pan, G; Context Aware Smart Car: from Model to Prototype, Journal of
Zhejiang University, ISSN 1673-565X (Print); ISSN 1862-1775 (Online), vol. 10, no. 7,
pp. 1049-1059, 2009.
[6] Forrest, A.; Mustafa, K.; Autonomous Cars and Society. Department of Social Science
and Policy Studies, Worcester Polytechnic Institute, May 1, 2007, http://www.wpi.edu/
Pubs/E-project/Available/E-project-043007-205701/unrestricted/IQPOVP06B1.pdf.
[7] Fleming, B.; Human/Machine Interface Issues Take Precedence. IEEE Vehicular
Technology Magazine, pp. 4-9, March 2009.
[8] Stevenson, R.;Long Distance Car Radar. IEEE Spectrum Magazine, pp. 50-55, October
2011.
[9] Paisitkriangkrai, S.; Shen, C.; Zhang, J.; Fast Pedestrian Detection Using a Cascade of
Boosted Covariance Features. IEEE Transactions on Circuits and Systems for Video
Technology, vol. 18, no. 8,pp. 1140-1151, 2008.
[10] Jones, W.D.; Building Safer Cars. IEEE Spectrum Magazine, pp. 82-85, January 2002.
[11] Barshan, B.; Kuc, R.; A Bat-Like Sonar System for Obstacle Localization. IEEE
Transactions on Systems, Man, and Cybernetics, vol. 22, no. 4, pp. 636-646, 1992.
[12] Kampchen, N.; Evolving Pinna Like Surfaces for a Biomimetic Sonar Head. M.Sc.
Thesis, University of Edinburgh, 2000.
[13] Kuc, R.; Biomimetic Sonar Recognizes Objects Using Binaural Information. JASA,
vol. 102, no. 2, pp. 689-696, 1997.
[14] Peremans, H.; CIRCE Project. http://www.ua.ac.be.
[15] Hallam, J.; ChiROPing Project. http://www.chiroping.org.
[16] Muller, R.; Kuc, R.; Biosonar-Inspired Technology: Goals, Challenges and Insights.
Bioinspiration & Biomimetics, vol. 2, pp. 146-161, 2007.
[17] Maxx Products International, LLC.,http://www.maxxprod.com.
[18] Lynxmotion, Inc., Users ManualSSC-32 Ver 2.0.http://www.lynxmotion.com.

[19] Reinard, L.; Streicher A.; Sutor, A.; Broadband Ultrasonic Transducer. 19th
International Conference on Acoustics, Madrid, Spain, 2007.
[20] Doring, J.; Bovtun V.; Bartusch, J.; Beck, U.; Kreutzbruck, M.; Cellular
Polyproplylene Ferroelectret Films Piezolectric Material for Non-Contact Ultrasonic
Transducers. 17th World Conference on Nondestructive Testing, Shanghai, China,
2008.
[21] Bovtun, V.; Doring, J.; Bartusch, J.; Beck, U.; Erhard,A.; EMFIT Ferroelectret Film
Transducers for Non-Contact Ultrasonic Testing.ECNDT, 2006.
[22] Streicher, A.; et al.; Ultrasonic Transducers for a Biomimetic Sonar System. IEEE
International Ultrasonics, Ferrolectrics and Frequency Control Joint 50th Anniversary
Conference, pp. 1142-1145, 2004.
[23] Jimenez, A.; et al; EMFI Based Ultrasonic Transducer for Robotic Applications.
Sensors and Actuators, A., vol. 148, pp. 342-349, 2008.
[24] Paajanen, M.; Lekkala, J.; Kirjavainen, K.; ElectroMechanical Film (EMFi) -A New
Multipurpose Electret Material. Sensors and Actuators, vol. 84,pp. 95-102, 2000.
[25] EMFIT Film Specifications. 2009, www.emfit.com.
[26] Ealo, et al.; A Fabrication Procedure for Airborne Ultrasonic Phased Arrays Based on
Cellular Electromechanical Film. IEEE Ultrasonics Symposium, IUS 2008, pp. 891-
894, 2008.
[27] http://www.st.com/stonline/products/literature/ds/12325/stn83003.htm.
[28] http://www.st.com/stonline/products/literature/ds/12329/stn93003.htm.
[29] http://www.national.com/mpf/LM/LM7171.html.
[30] Carmena, J.M.; Kampchen, N.; Kim, D.; Hallam, J.C.T.; Artificial Ears for a
Biomimetic Sonar Head: From Multiple Reflectors to Surfaces. Artificial Life, vol. 7,
no. 2, pp. 147-169, 2001.
[31] Carmena, J.M.; Kim, D.; Hallam, J.C.T.; Designing Artificial Ears for Animat
Echolocation: From Animals to Animats. J-A. Meyer et al. (Eds.), MIT Press, vol. 6,
pp. 73-80, 2000.
[32] MA40S5 Product Page. http://search.murata.co.jp.
[33] Epure S.; Aiordachioaie, D.; On Pinnae Design and Fabrication Technology. 17th
IEEE International Symposium for Design and Technology in Electronic Packaging
(SIITME-2011), Timisoara, Romania, pp. 133-136, 2011.
[34] Urena, J.; et al.; Classification of Reflectors with an Ultrasonic Sensor for Mobile
Robot Applications. RAS, vol. 29, pp. 269-279, 1999.
[35] Reijniers, J.; Peremans, H.; Biomimetic Sonar System Performing Spectrum Based
Localization. IEEE Trans. on Robotics, vol. 23, no. 6, pp. 1151-1159, 2007.
[36] Rikoski, R.J.; et al.; On Correlating Sonar Images. Robotics: Science and Systems I,
MIT Press, Cambridge, Massachusetts, 2005.
[37] Sejdic, E.; et al.; Time Frequency Feature Representation Using Energy Concentration:
An Overview of Recent Advances. Digital Signal Processing, vol. 19, pp. 153-183,
2009.
[38] Araujo, E.G.; Grupen, R.A.; Feature Extraction for Autonomous Navigation Using an
Active Sonar Head. IEEE International Conference on Robotics and Automation, San
Francisco, 2000.
[39] Gudra T.; et al.; Airborne Ultrasonic Transducers for Ultrasonic Trans. Tomography in
Gaseous Media. 20th International Congress on Acoustics (ICA-2010), Sydney, pp. 1-
5, 2010.
[40] Gomez F.; Althoefer K.; SeneviratneL.D.; Simulation of Ultrasound Imaging Inside
Fully Charged Pipes. Automation in Construction, vol. 15, pp. 355-364, 2006.
[41] Gomez F.; Althoefer K.; SeneviratneL.D.; Simulation of Ultrasound Imaging Inside
Fully Charged Pipes. Automation in Construction, vol. 15, pp. 355-364, 2006.
[42] Baranoski, E.J.; Through-Wall Imaging: Historical Perspective and Future Directions.
Journal of the Franklin Institute, vol. 345, pp. 556-569, 2008.
[43] Borek, S.E.; An Overview of through the Wall Surveillance for Homeland Security. 34th
Applied Imagery and Pattern Recognition Workshop, vol. 6, pp. 19-21, 2005.
[44] Bernus, L.; et al.; Sampling Phased Array: A New Technique for Signal Processing and
Ultrasonic Imaging. INDTCM, Publisher: NDT.net, vo. 50, no. 3, pp. 153-157, 2008.
[45] Jensen, J.A.; Ultrasound Imaging and its Modeling, Imaging of Complex Media with
Acoustic and Seismic Waves. Springer Verlag, vol. 84, pp. 135-165, 2002.
[46] Aiordachioaie, D; Frangu, L.; Some Results on Low Resolution Airborne Ultrasonic
Image Generation and Processing. IEEE International Symposium on Signal
Processing and Information Technology (ISSPIT-2011), Bilbao, Spain, pp. 434-439,
2011.
[47] Jahne, B.; Digital Image Processing. 5th Edition, Springer, 2002.
[48] Pratt, W.K.; Digital Image Processing. John Wiley & Sons, 2001.
Chapter 3
AUTONOMOUS UNDERWATER VEHICLE DYNAMICS
Mustafa Dinc and Chingiz Hajiyev†

Aeronautics and Astronautics Faculty, Istanbul
Technical University, Maslak, Istanbul, Turkey
ABSTRACT
The research on underwater systems has gained enormous attention during the last
two decades because of applications taking place in many fields. Therefore, the
significant number of Unmanned Underwater Vehicles (UUVs) has been developed for
solving the wide range of scientific and applied tasks of ocean research and development
in the world. Guidance, navigation, and control techniques are key research and
development areas for the success of those sophisticated UUV missions.
One of the main objective of this chapter is to provides detailed explanations on the
theory behind the main concepts that directly influence the design of the dynamic
mathematical model of AUV and then to accomplish dynamic mathematical modeling of
an AUV in MATLAB version 7.5 environment under different swimming conditions.
In order to develop high fidelity AUV simulation model and implement control and
navigation algorithm for the vehicle, we need to know overall AUV modeling.
Another main focus of this chapter is to realize the parameter identification of hydro-
dynamic coefficients based on a Least Square Estimation (LSE) algorithm for a nonlinear
mathematical modeling of AUV. Parameter Identification is very important to have the
estimated values of these coefficients in order to accurately simulate the AUV‟s dynamic
performance. The estimated hydrodynamic coefficients can be used as inputs not only for
a mathematical model to analyze the maneuvering performance but also for the AUV‟s
motion controller.
Keywords: Autonomous Underwater Vehicle, Dynamic Modeling, Parameter Identification,

Least Square Estimation

Mustafa Dinc: Aeronautics and Astronautics Faculty, Istanbul Technical University, Maslak, Istanbul, Turkey. E-
mail: mdinc1972@gmail.com.
†
Chingiz Hajiyev: Aeronautics and Astronautics Faculty, Istanbul Technical University, Maslak, Istanbul, Turkey.
E-mail: cingiz@itu.edu.tr.
82 Mustafa Dinc and Chingiz Hajiyev
1. INTRODUCTION
Unmanned Underwater Vehicle (UUV) research began in the 1960s, with the first proto-
types emerging in the 1980s. Nevertheless, the research on underwater systems has gained
immense interest during the last two decades with applications taking place in multiple fields
of marine systems.
Therefore, the significant numbers of UUVs have been developed for solving a wide
range of scientific and applied tasks of ocean and seabed research and development in the
world [1]. The military, as well as civilian industries can see great potential uses of UUVs in
the underwater environment.
UUVs by definition are small submersible vehicles that contain independent propulsion
systems and are capable of carrying sensors such as side-scan sonar, video cameras, depth
sensor, and other oceanographic measuring devices [1]. UUVs are highly desirable as they
can at least limit the level of human life risk and direct physical human involvement in a
mission.
Typically, UUVs can be classified into two unmanned underwater systems that are
Remotely Operated Vehicles (ROVs) and Autonomous Underwater Vehicles (AUVs): Each
branch has its own pros and cons due to the mission type. ROVs, characterized by direct
human assistance, is remotely operated through the presence of a tether cable; the other group
is known as the AUVs, which is the topic of this chapter, characterized by their autonomous
behavior, having cylindrical geometric shape and absence of a tether cable. This classification
is of course not precise given the varying degrees of autonomy in both groups which can
differ according to the requirements of the designed mission. In addition, the presence of a
tether cable does not necessarily mean that the vehicle cannot perform autonomous tasks. The
basic difference between AUVs and ROVs is that AUVs use “intelligence”, such as sensing
and automatic decision-making. They have predefined plan of operations in its “mind”
allowing them to perform tasks autonomously [1]. ROVs are remotely controlled by a human
with the help of communication links based on tether cable (such as cupper, fiber optic, etc.).
The AUVs have shown efficiency at performing complex search and inspection missions,
and opening a number of new important application areas which include environmental
monitoring, surveillance, underwater inspection of harbors and pipelines, geological and
biological surveys, mine countermeasures, etc. [1].
Due to their versatility, compact size, independence and covertness, AUVs are a highly
valuable asset in these environments. Possible missions for AUVs range from dedicated, rapid
environmental assessment, special operations to reconnaissance, surveillance and intelligence.
This wide range of possible applications and their partly contradictory nature calls for a large
variety of vehicle configurations with regard to speed, range, depth, endurance, launch and
recovery, payload and communication systems.
The nonlinear and multivariable dynamics, accompanied with parametric uncertainties
and also immeasurable disturbances cause position control of an AUV to become an
important issue. Dynamic equations of AUV are strongly nonlinear, multivariable and
uncertain. Sea currents are disturbances for the system. Parameters of equations, which called
hydraulic coefficients, should be estimated. Some of the parameters, such as hydrodynamic
coefficient, do not have accurate value. Computation and stimulation of these parameters is so
complex and so time consuming and costly.
Autonomous Underwater Vehicle Dynamics 83
These parameters can vary due to environmental conditions. On the other hand, the
control of AUV presents a number of unique and formidable challenges. An AUV will be
required to move in adverse environments showing good performance and a great deal of
autonomy under a variety of load conditions and unknown sea conditions. Moreover, AUV
dynamic is highly nonlinear, coupled, time varying and subject to hydrodynamic uncertainties
and external disturbances. For trajectory tracking, the problem statement requires the design
of a depth and head control system for the vehicle in order to achieve precise trajectory [2].
In order to develop an accurate and robust navigation and control system, we need to
know an overall mathematical modeling of AUV, which involves the interdisciplinary study
of kinematics, hydrostatics, and hydrodynamics.
Hydrostatic is concerned with the equilibrium of underwater bodies at rest or moving
with constant velocity, whereas hydrodynamics is concerned with bodies having accelerated
motion [3]. In this chapter, we describe nonlinear dynamic model of an AUV for free
swimming conditions.
Otherwise, in many literature, scientists employ parameter specific nonlinear AUV model
for only one swimming condition.
Another main focus of this chapter is to realize the parameter identification of
hydrodynamic coefficients based on Least Square Estimation (LSE) algorithm for a nonlinear
mathematical modeling of AUV. The hydrodynamic coefficients of AUV can be estimated
based on the calculations of the first principle of dynamics of AUVs and from statistical LSE
of experimental data [4]. The identification of the parameters of the item being tracked can be
approached in many different ways. The parameters can be determined once, from a model of
the expected input signal. It could also be done adaptively utilizing a recursive method to
identify the parameters of the incoming signal.
Another approach that estimates the hydrodynamic coefficients of AUVs is the observer
method, in which a model-based estimation algorithm is used. A representative method
amongst observer methods is the Kalman filter (KF), which has been widely used to estimate
state variables and parameters [4]. Hwang [5] estimated the maneuvering coefficients of a
ship and identified the dynamic system of a maneuvering ship using an extended Kalman
filter (EKF) technique. Additionally, referencing [6], Meng and Veras, they concentrated on
the application and comparison of EKF and iterated EKF for aerodynamic parameter
estimation of a fixed wing UAV.
In another reference [7], Chowdhary compared the performance of three recursive
parameter estimation algorithms for aerodynamic parameter estimation of two aircrafts
derived from real flight data. These algorithms are the EKF, the simplified version of the
Unscented Kalman filter (UKF) and the augmented version of the UKF.
On the other hand, the report [8] compares responses obtained by the KF, the least
squares estimation, and the linear model for the NPS Phoenix AUV. The LSE provided
results similar to those obtained by the KF, but the latter produced a more accurate model.
Referencing [9] Vandersteen, Rolain, Schoukens and Pintelon proposed a robust
estimation algorithm for the estimation of static and nonlinear systems which can be
described as a nonlinear function corrected with a rational form. The errors-in-variables-based
algorithm solves the starting-value problem using an iterative, weighted least-squares
procedure, which constructs the rational form such that the set of normal equations becomes
best conditioned, and uses a maximum-likelihood estimation step to increase the efficiency of
the estimates.
The main focus of this chapter is to provides detailed explanations on the theory behind
the main concepts that directly influence the design of the dynamic mathematical model of
AUV and then to accomplish dynamic mathematical modeling of an AUV in MATLAB
version 7.5 environment under different swimming conditions. Another main focus of this
chapter is to realize the parameter identification of hydrodynamic coefficients based on the
LSE algorithm for a nonlinear mathematical modeling of the AUV. It is important to have the
estimated values of these coefficients in order to accurately simulate the AUV‟s dynamic
performance. The estimated coefficients can be used as inputs not only for a mathematical
model to analyze the maneuvering performance but also for a AUV‟s motion controller.
2. DYNAMIC MATHEMATICAL MODELING OF AN AUV

This section provides detailed explanations on the theory behind the main concepts that
directly influence the design of the dynamic mathematical model of AUV.
In order to achieve robust and accurate navigation, guidance and control of an AUV, we
require overall modeling of AUV, which is fairly complicated, and even when considered as a
rigid body, an exact analysis is only possible by including the underlying infinite dimensional
dynamics of the surrounding fluid [3]. This can be done using partial differential equations,
which are solved by Computational Fluid Dynamics (CFD), but it still involves a formidable
computational burden, which is infeasible for most practical applications.
As a result, the conventional approach has been to use finite-dimensional approximations.
On the other hand, AUV modeling involves the interdisciplinary study of kinematics,
hydrostatics, and hydrodynamics. The study of hydrostatics is concerned with the equilibrium
of underwater bodies at rest or moving with constant velocity, whereas hydrodynamics is
concerned with bodies having accelerated motion [3].
An increased knowledge of hydrodynamic parameters then leads to a better navigational
system design and performance on AUVs.
In this chapter we try to develop nonlinear mathematical modeling of an AUV for free
swimming conditions and then as case study, directly use the mathematical model of REMUS
AUV that is designed to perform hydrographic analysis in the very shallow water.
REMUS is used for missions such as hydrographic surveys, mine counter-measure
operations, harbor security operations, environmental monitoring debris field mapping, search
and rescue operations, fishery operations, and scientific sampling and mapping.
2.1. Modeling Assumptions
2.1.1. Environmental Assumptions

The corresponding assumptions are made about the vehicle with respect to its
environment [10]:
● The AUV is deeply submerged in a homogeneous and unbounded fluid. In other

words, the AUV is located far from the free surface (no surface effects, i.e. no sea
wave or vehicle wave-making loads), walls and bottom.
● The AUV does not experience underwater currents.
2.1.2. Dynamics Assumptions

In dynamic modeling of AUV, the following assumptions are used [11]:
● The AUV behaves as a rigid body of a constant mass.

● The earth‟s rotation is negligible for acceleration components of the vehicle‟s center
of mass
● The primary forces that act on the AUV are inertial and gravitational in the center of
buoyancy and are derived from hydrostatic, propulsion, thruster, and hydrodynamic
lift and drag forces.
● The thruster assumption is that it uses an extremely simple propulsion model, which
treats the vehicle propeller as a source of constant thrust and torque.
2.1.3. 6-DOFRigid-Body Equations of Motion

AUVs move in six degrees of freedom (6-DOF) since six independent coordinates are
necessary to determine the position and orientation angle of a rigid body dynamics. The first
three coordinates and their time derivatives are based off of translational motion along the x, y
and z-axes, while the last three coordinates (  , , ) and their time derivatives are used to
describe orientation angle and rotational motion [12].
Velocity and angular velocity components of the AUV relative to the body axes ( x, y, z )
are denoted by the velocity of surge, sway, heave motion, and angular velocity of roll, pitch,
and yaw motion ( p, q, r ) , respectively. X, Y, Z, K, M and N represent the resultant forces and
moments with respect to the x, y, and z axis. For AUVs, it is common to use the SNAME
notation. In Table 1 below, the six different translational and rotational motion components
are defined as: surge, sway, heave, roll, pitch and yaw, respectively [3].
2.1.4. Coordinate Frames

Typically, three different right-handed and rectangular coordinate frames are used for
defining AUV motion, which is a 6 degrees of freedom (6-DOF). First, the body axes have
their origin at the center of buoyancy (CB) with x directed toward the bow along the hull
centerline axis, y directed to the starboard side, and z toward the keel (Figure 2).
Table 1. AUV Dynamic Components [1]
Motion Forces and Linear and Angular Position and Euler

Components Moments Velocities Angles
Surge X u X
Sway Y V Y
Heave Z w Z
Roll K P 
Pitch M Q θ
Yaw N R ψ
Figure 1. 6-DOFnavigational frame [11].
Figure 2. AUV body-fixed and earth-fixed coordinate system.
The axes fixed in the earth are with the plane in the water surface and ze directed
downward into the ocean. Second, the Earth-fixed coordinates frame is also measured to CB
of AUV. If roll, pitch and yaw orientation angles of the AUV are zero, the axes will be
parallel to the axes, respectively.
The third coordinate frame is fixed in the fluid, which can move with a constant velocity
relative to the earth-fixed frame. The axes are always parallel to the axes, respectively [3].
A set of axes commonly used with the Earth-fixed axis system is shown in 0, where axis
is chosen to point north, axis points east with the orthogonal triad being completed when axis
pointing down.
A transformation matrix containing „Euler‟ angles where is roll, θ is pitch, and ψ is yaw,
must be defined. More specifically, the adopted Euler angles move the reference frame to the
referred body frame, following a Z-Y-X (or so-called 3-2-1) rotation sequence [13]. As a
result, the transformation order from the earth-fixed frame ( e ) to the body-fixed frame
( b ) is given in the following equation:
u
e 3   
 1
u2
  2 
u1
 b
(1)
Transformation matrix or direction cosine matrix is defined in the following equation:
 cos ψ cos θ sin ψ cos θ - sin θ 



T ( , θ , ψ ) =  cos ψ sin θ sin  - sin ψ cos  sin ψ sin θ sin  + cos ψ cos  cos θ sin  
 
 cosψ sinθ cos + sin ψ sin  sin ψ sin θ cos  - cos ψ sin  cos θ cos  
(2)
Transformation from a global velocity vector to the local velocity vector is given by
u  X 
 v  = T  ,θ,ψ .  Y 
    
 w Z 
  (3)
The global angular velocity vector can be transformed into the rates of change of the
Euler angles as given in Equation (4):
  1 sin  tan θ cos  tan θ   p 

  
 θ  = 0 cos  - sin    q 
ψ  0 sin / cosθ cos / cos θ   r 
  (4)
Three dimensional (3D) spatial rotations can be parametrized using both Euler angles and
unit quaternions. Unit quaternion provides a convenient mathematical notation for represent-
ting orientation angles and rotations of vehicle in three dimensions.
Compared to Euler angles, unit quaternions are simpler to compose and avoid the
problem of the gimbal lock. Additionally, compared to rotation matrices, they are more
numerically stable and may be more efficient [14].
Nevertheless, Euler angles are preferred in this chapter for simplification reasons.
After a general model structure for AUV is derived, we look further into the modeling
aspects in terms of environmental disturbance models, in which waves (wind generated),
wind, and sea currents will be considered. In general, these environmental disturbances will
both be additive and multiplicative to dynamic EOM [3].
Transformation order from body-fixed frame ( b ) to sea current frame ( w ) using
orientation angle of α and β is given in following equation
u
b 
u2

 1 
3 
 w
(5)
The transformation matrix from the body-fixed frame ( b ) to the sea current frame
( w ):
Cˆ    R2  -   R3   
b, w
cos  cos  -cos  sin  -sin  

 b, w 
ˆ
C   sin  cos  0 
 sin  cos  -sin  sin  cos  
(6)
The sea current velocity components are assumed constant relative to axes fixed in the
earth. This permits the use of axes fixed in the fluid as an inertial frame of reference. AUV
velocity with respect to the sea is expressed as
Vb w  Vb e  Vw e
(7)
Then, dynamic pressure is calculated as
1
Qd   Vb2 w
2 (8)
In this chapter, however, we ignore the sea currents, which normally effects the AUV
motion. Additionally, wind and wind generated waves phenomena will not be discussed since
the attention is focused on AUVs performing a motion or manipulation task in an underwater
environment.
2.1.5. Newtonian and Lagrangian Mechanics

The EOM of AUVs are highly nonlinear, time-varying and coupled due to hydrodynamic
added mass, lift, drag, coriolis and centripetal forces, which are acting on the vehicle and
generally include uncertainties [3]. Overall 6-DOF nonlinear dynamic EOM can be expressed
in the matrix form as
M (v)v  C (v)v  D(v)v  g ( )   (9)
where, M(v) is inertia matrix (including added mass), C(v) is matrix of Coriolis and
centripetal terms (including added mass), D(v) is damping matrix, is vector of gravitational
forces and moments, is vector of control inputs, and v  [u, v, w, p, q, r ] .
T
The coupled EOM of AUV are derived from two possible modeling approaches; one is a
Lagrangian method and the other is a Newtonian-Euler formulation.
Basically, the Lagrangian approach consists of three main steps: first, to formulate a
suitable expression for the vehicle's kinetic (T) and potential energy (P), second, to compute
the Lagrangian L (L=T-P), and finally to apply the Lagrangian (L) to the Lagrangian formula-
tion [12].
d  L  L
   J T  
dt     (10)
Another modeling approach is the Newtonian-Euler formulation, which is based on

Newton‟s Second Law that relates mass (m), acceleration (a) and force (F). Euler suggested
expressing Newton's Second Law in terms of conservation of both linear and angular
momentum. The forces (F) and moments (M) refers to the body's center of gravity [3].
In this study, the dynamic behavior of an AUV is described through Newton's laws of
linear and angular momentum.
Newton‟s Second Law is expressed as:
d
F {mV }
dt (11)
d
M {H }
dt (12)
where F represents the sum of all externally applied forces, M represents the sum of all
applied torques, and H is the angular momentum.
It is convenient to regard the sums of applied torque (M) and force (F) as consisting of an
equilibrium point and a perturbational component. Thus, assuming constant AUV mass
d
F  F0  F  m {V }
dt (13)
d
M  M 0  M  {H }
dt (14)
The subscript “0” denotes the equilibrium condition, and “  ” the component of
perturbation. Equilibrium of AUV by definition must be an unaccelerated motion along a
straight path; during this motion, the linear velocity vector relative to fixed space is invariant,
and the angular velocity is zero. Thus, both and are zero.
Furthermore, since the axis system being used as an inertial reference system is the Earth-
fixed coordinate system, Equations (15) and (16) can be expressed as
d
F  m {VT }E
dt
(15)
d
M  {H }E
dt (16)
The force equation based on rate of change of velocity V relative to Earth‟s fixed axis
frame is given by
   
F = De mV = Db mV + wb e × mV
(17)
where is the angular velocity of the AUV with respect to the Earth fixed coordinate frame.
The open form of Equation (17) is given by
 Fx = X = m  u + qw - rv 

F   Fy = Y = m  v + ru - pw 

 Fz = Z = m  w+ pv - qu  (18)
After the rearrangement of F, translational accelerations become:
Fx
u= - qw+ rv
m
Fy
v= - ru + pw
m
F
w = z - pv + qu
m (19)
Similarly, after transforming from the body-fixed frame to the Earth-fixed frame the
moment equation and its open form becomes
   
M = De Iˆ ωb e = Db Iˆ ωb e + ωb e × Iˆ ωb e  
  
 M x = K = I x p - I yz q - r - I zx  r + pq  - I xy  q - rp  - I y - I z qr

2 2


 
M   M y = M = I y q - I zx r 2 - p 2 - I xy  p + qr  - I yz  r - pq  -  I z - I x  rp

    
 M z = N = I z r - I xy p 2 - q 2 - I yz  q + rp  - I zx  p - qr  - I x - I y pq
(20)
For a rigid body, angular momentum can be defined as
H  Iω (21)
where the inertia matrix is defined as
 I xx  I xy  I xz 
 
I    I xy I yy  I yz 
 
  I xz  I yz I zz 
(22)
where denotes a moment of inertia, and a product of inertia j  i . In this chapter, we assume
that the AUV is symmetrical along the XY and XZ planes, therefore cross inertia parameters
become
I xy  I yz  I xz  0 I y  Iz
and (23)
As a result of this symmetry, the rotational EOM of AUV in particular, becomes:
M x = Ix p
M y = I y q - Iz - Ix r p

M z = Iz r - Ix - I y p q (24)
After rearrangement, Equation (24) is expressed as:
Mx
p=
Ix
1
q=  M y +  I z - I x  r p 
Iy 
r=
1 
Iz 

M z + Ix - I y p q

(25)
Finally, the derivative of Euler angles is defined as:
q sin φ+ r cos φ
ψ=
cos θ
θ = q cos φ - r sin φ
φ = p +  q sin φ+ r cos φ  tan θ
(26)
2.1.6. Gravitational Forces

Gravitational forces always exist in the AUV. It can be assumed that gravity acts at the
center of gravity (CG) of the vehicle. Since the centers of mass and gravity coincide in an
AUV, there is no external momentum produced by gravity on the CG.
Therefore, for the body axis system, gravity contributes only to the external force vector
F. Three components of the gravitational force in the body frame depend on the AUV‟s
attitude relative to an inertia frame. The gravitational force acting upon an AUV is most
obviously expressed in terms of the Earth‟s axes. With respect to these axes, the gravity
vector mg, is directed along the axis. Figure 3 demonstrates the alignment of the gravity
vector with respect to the body-fixed axes. In Figure 3, θ represents the pitch angle between
the gravity vector and the plane; the angle is positive when the nose of the AUV goes up, Φ
represents the roll angle between axis and the projection of the gravity vector on the plane.
Direct solution of the vector mg into X , and components produces:
g x  mg sin( )  mg sin 

g y  mg cos( )sin   mg cos  sin 
g z  mg cos( ) cos   mg cos  cos 
(27)
In general, the Euler angles (  ,  ,  ) are not simply the integrals of the angular
velocity p, q, and r (Equation (25) ). It is necessary to relate these and their derivatives to the
angular velocities p, q, and r. This depends upon whether the gravitational vertical seen from
the AUV is fixed or whether it rotates relative to inertial space [16].
The manner in which the angular orientation and velocity of the body axis system with
respect to the gravity vector is expressed by depending upon the angular velocity of the body
axes about the vector mg. The external forces acting on the AUV can be expressed as:
X  Fx  g x
Y  Fy  g y
Z  Fz  g z (28)
where are the gravitational terms, and represent the hydrodynamic and thrust forces
respectively.
Figure 3. The orientation of gravity vector with respect to the body axis system [15].
Hence, from the Equations (13) and (15), the force equations become:
Fx  maxCG  m(u  qw  rv  g sin  )

Fy  ma yCG  m(v  ru  pw  g cos sin  )
Fz  mazCG  m( w  pv  qu  g cos cos  )
(29)
The gravitational acceleration forces and moments are represented by the weight minus
buoyancy (W−B) and weight moment terms respectively [4]. In Figure 4, the Simulink model
of gravitational force calculation is shown.
2.1.7. Hydrostatic Forces and Moments

When an AUV is submerged in a fluid under the effect of gravity, two forces act on the
vehicle: the gravitational force, which is mentioned in the previous sub-section and the
buoyancy, which is called “hydrostatic effect”.
The buoyant force acting on the center of buoyancy (CB) is represented in body-fixed
frame (Figure 5). It can be recognized that the difference between gravity and buoyancy (W-
B) only affects the linear force acting on the vehicle. It is also clear that the restoring linear
force is constant in the Earth-fixed frame [11]. On the other hand, the two vectors of the first
moment of inertia W and B affect the momentum acting on the vehicle and are constant in the
body-fixed frame. A solid body submerged in a fluid will have upward buoyant force acting
on it equivalent to the weight of displaced fluid, enabling it to float or at least appear to
become lighter. If the buoyancy exceeds the weight, then the object floats; if the weight
exceeds the buoyancy, the object sinks. If the buoyancy equals the weight, the body has
neutral buoyancy and may remain at its level. Discovery of the principle of buoyancy, which
is a result of the hydrostatic pressure in the fluid, is attributed to Archimedes [17].
After applying hydrostatic force coefficients to Equation 29, translational force equations
become:
Fx  maxCG  m(u  qw  rv)  (W  B )sin  


Fy  ma yCG  m(v  ru  pw)  (W  B )cos  sin  

Fz  mazCG  m( w  pv  qu )  (W  B)cos  cos  
(30)
Figure 4. Gravitational force calculation.
Similarly, after applying the hydrostatic moment coefficients to Equation (26), the
moment equations become:
Figure 5. Hydrostatic forces and moments [15].
Figure 6. Hydrostatic force and moment calculation.
M x = I x p + (z gW - zb B )cos  sin 
M y = I y q -  I z - I x  rp + (z gW - zb B )sin 
 
M z = I z r - I x - I y pq
(31)
In Figure 6, hydrostatic forces and moments are implemented in Simulink.
2.1.8. Hydrodynamic Forces and Moments

In this subsection, the main hydrodynamic effects acting on an AUV moving in a fluid
(seawater) will be briefly discussed. This is the key part of AUV modeling. Standard EOM
contain only stability derivatives for the specific AUV configuration of interest.
Hence, trajectory simulation or prediction using traditional methods requires a priori
knowledge of the hydrodynamic characteristics of the vehicle in the flow regimes, which may
occur during the maneuver [15].
All hydrodynamic parameters are defined uniquely for a given AUV shape by formulae
based on the results of theory and experiment.
There are primarily three methods of determining the hydrodynamic coefficients in the
design process of underwater vehicles: (1) towing tank tests, (2) numerical computations, and
(3) field tests. Among these methods, the most reliable results are obtained from the field
tests, where the whole designed model is tested in a real sea environment [1]. Towing tank
tests are performed with a scaled model and the hydrodynamic forces and moments can
accurately be determined. However, experimental testing of designs is a time consuming and
costly process (construction of the models, instrumentations, test infrastructure, etc.).
Numerical Computations are mainly based on semi-empirical or CFD methods. Semi-
empirical or potential theory-based methods are generally utilized in the preliminary design
process, where it is important to determine the hydrodynamic characteristics in a short period
of time. CFD methods give accurate results and are used in the detailed design process [12].
The theory of fluid dynamics is rather complex and it is difficult to develop a reliable and
robust model for most of the hydrodynamic effects. A rigorous analysis for incompressible
fluids would need to resort to the Navier-Stokes equations, which are the basic governing
equations for a viscous, heat-conducting fluid.
These equations describe how the velocity, pressure, temperature and density of a moving
fluid are related. The hydrodynamic prediction method is coupled with a 6-DOF EOM solver
to predict vehicle trajectories. The predicted motion characteristics of the AUV are also
sensitive to details of the predicted hydrodynamic characteristics of the vehicle.
Small perturbations in the flow field, which cause small; variations in the vehicle forces
and moments, accumulate over the length of the hull, and can produce large perturbation; in
the calculated trajectory.
In this chapter in order to determine hydrodynamic forces and moments, we used
SUBFLO_2, which is an engineering physically based and commercially available software
tool. SUBFLO_2 has four major components which are hull separation vortex method, fin
horseshoe vortex method, propeller/propulsion models, and 6-DOF equations of motion
solver. The hydrodynamic prediction method is coupled with a 6-DOF EOM solver to predict
vehicle trajectories. The predicted motion characteristics of the AUV are also sensitive to
details of the predicted hydrodynamic characteristics of the vehicle. Small perturbations in the
flow field, which cause small; variations in the vehicle forces and moments, accumulate over
the length of the hull, and these can produce large perturbation; in the calculated trajectory
[23].
In order to determine hydrodynamic coefficients utilizing with SUBFLO_2, first we find
three static hydrodynamic force coefficients of CFx, CFy, and CFz, and then three static
hydrodynamic moment coefficients of CMSFx, CMSFy, and CMSFz, for the predefined AUV
geometry.
These coefficients are calculated due to the parameters of δs, δr, α, β, Vb/w, which are
elevator deflection, rudder deflection, angle of attack, side slip angle, and velocity vector with
respect to water, respectively. In Figure 7, the Simulink model of hydrodynamic force and
moment calculation is shown.
Figure 7. Hydrodynamic force and moment calculation.
2.1.9. Added Mass and Inertia

When a rigid body is moving in a fluid, the additional inertia of the fluid surrounding the
body, which is accelerated by the movement of the body, has to be considered. This effect can
be neglected in industrial robotics since the density of the air is much lighter than the density
of a moving mechanical system.
However, in underwater applications the density of the water, ρ ≈ 1000 kg/m3, is
comparable with the density of the vehicles. In particular, at zero degrees, the density of the
fresh water is 1002.68 kg/m3; for sea water with 3.5% of salinity it is ρ = 1028.48 kg/m3.
Since the fluid surrounding the body is accelerated with the body itself, a force is then
necessary to achieve this acceleration (the fluid exerts a reaction force which is equal in
magnitude and opposite in direction).
This reaction force is the added mass contribution [16].
The added mass is not a quantity of fluid to add to the system such that it has an
increased mass. Different properties hold with respect to the six by six inertia matrix of a
rigid body due to the fact that the added mass is a function of the body‟s surface geometry.
As an example, the inertia matrix is not necessarily positive definite. The added mass has
also an added Coriolis and Centripetal contribution.
In this study, “Added Mass and Inertia” effects of water is not taken into consideration.
Added (virtual) mass should be understood as pressure-induced forces and moments due to a
forced harmonic motion of the body, which are proportional to the acceleration of the body
[12].
Consequently, the added mass forces and acceleration will be 180˚ out of phase to the
forced harmonic motion. However, this is not true when AUV is close to surface.
2.1.10. Propeller Effect

The propeller produces the main thrust. Consequently, the reaction of the body to the load
torque of the propeller produces a moment with respect to its rotational axis. The vehicle is a
nonlinear system: all equations of motion of the system include coupled terms [1]. The main
terms of this type are in the longitudinal (X) force and roll moment (K) equations because the
thrust forces and moments act in the direction of the x-axis.
A propeller with a rudder can produce a thrust vector within a range of directions and
magnitudes in the horizontal plane for low speed maneuvering and dynamic positioning. In
our physically based model, we used a simple fixed-trust model due to the AUV velocity.
2.2. Mathematical Model of AUV
Detailed explanations on the theory behind the main concepts that directly influenced the
design of the dynamic mathematical model of AUV are given in the previous sub-section.
Now we arrive at the combined overall non-linear EOM for the AUV in 6-DOF as follows
[18]:
Surge or translational motion along the x-axis:
X  X u u  m  u  zG q  yG r   X uu u u  ( X wq  m)wq  X qq  mxg q 2   
 
  X vr  m  vr  X rr  mxg r 2  mz g pr  W  B  sin   X prop
(32)
Sway or translational motion along the y-axis:
 
Y  Yv v  Yr r  m  v  z g p  xg r   Yuv uv  Ywp  m wp  Yur  m  ur 
   
 mz g qr  Ypq  mxg pq  Yvv v v  Yrr r r  W  B  cos sin   Yuu r u 2 r
(33)
Heave or translational motion along the z-axis:
  
Z  Z w w  Z q q  m   w  xg q  yg p   Zuq  m uq  Z vp  m vp  mz g p 2    
  2
 
 Zuwuw  mz g q  Z rp  mxg rp  Z ww w w  Z qq q q  W  B  cos cos   Zuu e u 2 e
(34)
Roll or rotation about the x-axis:
    
K  mz g u  myg r  I xx  K p p  myg w  I zz  I yy qr  mz g wp    
     
 mz g ur  z gW  zb B cos sin  y gW  yb B cos cos  K pp p p  K prop
(35)
Pitch or rotation about the y-axis:
       
M  mz g u  mxg  M w w  I yy  M q q  M rp  I zz  I xx rp  mz g vr 
  mz g  wq   Muq  mxg  uq  M uwuw   M vp  mxg  vp  M qq q q  M ww w w 
  z gW  zb B  sin    xgW  xb B  cos cos  M uu u 2 e e
(36)
Yaw or rotation about the z-axis:

    
N  myg u  Nv  mxg v   N r  I zz  r  Npq  I xx  I yy pq  N wp  mxg wp 
  Nur  mxg  ur  Nuv uv  N rr r r  Nvv v v   xgW  xb B  cos sin   ygW  yb B  sin   Nuu u 2 r
r
(37)
where X, Y, Z, K, M, and N represent the resultant forces and moments with respect to the
body-fixed coordinates. The parameters of the mathematical model in Equations (32)-(37)
with their definitions can be found in Tables 3-5 at the end of this chapter.
2.3. Numerical Integration of the AUV EOM
The nonlinear differential equations defining the AUV accelerations and the kinematic
equations give us the vehicle accelerations in two different reference frames. Given the
complex and non-linear nature of these equations, we will use numerical integration to solve
for the vehicle speed, position, and attitude in time frame [19]. Consider that at each time
step, we can express nonlinear differential equation as follows:
xn  f  xn , un 
(38)
where x is the AUV state vector:
x  u v w p q r x y z    
T
(39)
and un is the input vector:
x  u v w p q r x y z    
T
(40)
There are two common numerical iteration methods to solve the non-linear differential
equation: one is Euler‟s method and the other is Runge-Kutta method.
2.3.1. Euler's Method First Order

We will first consider Euler's method, a simple numerical approximation that consists of
applying the iterative formula:
xn1  xn  f  xn , un  .t
(41)
where is the modeling time step.

Although the least computationally intensive method, Euler's method is unacceptable as it
can lead to divergent solutions for large time steps [19].
2.3.2. Runge-Kutta Method

In this subsection, we will introduce one of the most powerful predictor-corrector
algorithms- one which is so accurate, that most computer packages designed to find numerical
solutions for differential equations will use it by default - the Runge-Kutta fourth order
method. The Runge-Kutta method further improves the accuracy of the approximation by
averaging the slope at four points [19]. We used this method to solve the nonlinear dynamic
model of the AUV. In this method, numerical approximations that consist of applying the
iterative formula:
k1  xn  f  xn , un 
 t 
k2  f  x  k1 , u 1 
 2 n 
 2
 t 
k3  f  x  k2 , u 1 
 2 n 
 2
k4  f  x  tk3 , un 1 
(42)
where the interpolated input vector is:
1
u
n
1   un  un1 
2
2
(43)
The combination of these two Equations (42)-(43) yields to:
t
xn1  xn   k1  2k2  2k3  k4 
6 (44)
which is simply the x-value of the current point plus a weighted average of four different x-
jump estimates for the interval, with the estimates based on the slope at the midpoint being
weighted twice as heavily as those using the slope at the end-points [10].
For estimations of rigid-body aerodynamic or hydrodynamic models, the second order
Runge-Kutta algorithm is usually adequate and hence recommended during the initial
iterations of the iterative estimation algorithms, switching over the forth-order Runge-Kutta
only during the final iteration [20].
2.4. Dynamic Simulink Model and Control of AUV
The data achieved from the modeling and identification process of the AUV was
implemented into a vehicle which was used in a MATLAB Simulink environment.
This makes it possible to simulate the behaviour of the AUV or programs without the
need of a real sea environment.
The detailed AUV Simulink models are depicted in Figure 8. In this model,
hydrodynamic, hydrostatic, gravitational, and kinematic EOM of AUV and sliding mode
control are implemented in the MATLAB version 7.5 Simulink environment.
The sliding mode control algorithm, which is a robust control, is applied to yaw channel
of the dynamic model. The control input of course is
u        r   
(45)
The sliding mode control law for course control of AUV is
5
u sign  u 
18 (46)
The control strategy used for physically based model of AUV is depicted in Figure 9,
which is the Simulink model of sliding mode control of yaw (ψ).
Figure 8. Simulink model of AUV.
Figure 9. Simulink Model of Sliding Mode Control.

2.5. Simulation Results
The simulation of the dynamic model is developed by using the MATLAB version 7.5
Simulink environment. Numerical simulations are made to show the dynamic model of the
AUV.
In Figure 10, the simulation results of depth change in z axis, dynamic position maneuver
in x and y axis, and yaw angle change are shown. In Figure 16, the simulation result of
deflection angle change of rudder under the sliding mode control and total velocity during 50
sec. For this simulation, the AUV has a maneuver speed of 8 m/s and rudder angle applies to
10 deg. From the start, elevator angle applies to 1 deg, and the simulation time is 50 sec.
Figure 10. Simulation Results of X, Y, Z, Roll, Pitch and Yaw Angles (50 sec).
Figure 11. Simulation Results of V_tot, δr and δe (50 sec).

3. PARAMETER IDENTIFICATION WITH LEAST SQUARES ESTIMATION

This section focuses on the parameter identification of hydrodynamic coefficients of
AUVs based on the Least Square Estimation (LSE) algorithm for a nonlinear mathematical
modeling of AUVs. Hydrodynamic coefficients strongly affect the dynamic performance of
an AUV. Therefore, it is important to have the true values of these coefficients in order to
accurately simulate the AUV‟s dynamic performance. The estimated coefficients can be used
as inputs not only for a mathematical model to analyze the maneuvering performance but also
for a controller model to design AUVs under development.
However, parameter identification of AUV dynamics is complicated because of its non-
linear identification models and the combination of noisy sensor measurements.
3.1. AUV System Identification
System identification is the determination on the basis of an observation of input and

output, of a system within a specified class of systems to which the system under test is
equivalent [20].
When the system identification is applied to AUVs, the equations governing the AUV
dynamic motion are postulated and an experiment is designed to obtain measurements of
input and output variables [21].
The hydrodynamic forces and moments acting on AUV cannot be measured directly.
However, hydrodynamic modeling followed by parameter estimation allows
determination of specific hydrodynamic characteristics (such as lift, drag, and side force
coefficients, and rolling, pitching, and yawing moment coefficients in terms of stability and
control derivatives) from the related measurements such as accelerations, angular rates, etc.
[20].
The selected EOM for system identification are sway and yaw
 
Y  Yv v  Yr r  m  v  z g p  xg r   Yuv uv  Ywp  m wp  Yur  m  ur 
   
 mz g qr  Ypq  mxg pq  Yvv v v  Yrr r r  W  B  cos sin   Yuu r u 2 r
(47)
       
N  myg u  Nv  mxg v   N r  I zz  r  Npq  I xx  I yy pq  N wp  mxg wp  Nur  mxg ur 
   
 Nuv uv  N rr r r  Nvv v v  xgW  xb B cos  sin  ygW  yb B sin   Nuu r u  r
2
(48)
The dynamic Equations (47) and (48) are augmented with output equations that specify
the connection of AUV states and controls to measured outputs, along with measurement
equations describing the measurement process [21]. In Figure 12, Model Postulation is based
on a priori knowledge about the AUV kinematics and hydrodynamics.
The Experiment Design includes selection of an instrumentation system, and
specification of an AUV configuration and maneuvers for system identification.
Data Compatibility Analysis, in practice, measured AUV response data that can contain
systematic errors, even after careful instrumentation and experimental procedures.
To verify data accuracy, data compatibility analysis can be applied to measured AUV
responses. Model Structure Determination in AUV system identification means selecting a
specific form for modeling from a class of models, based on measured data. Parameter and
State Estimation constitutes a principal part of the AUV system identification procedure.
Parameter estimation for linear dynamic systems based on maximum likelihood and the least
squares principles can also be formulated in the frequency domain.
Collinearity diagnostics is in almost all practical applications of linear regression and the
model terms are correlated to some extent. Diagnostic information can aid in deciding what
corrective actions are necessary.
Model Validation is the last step in the identification process [22]. The identified model
must demonstrate that its parameters have physically reasonable values and acceptable
accuracy, and that model has a good prediction capability on comparable maneuvers [21].
3.2. Parameter Estimation for AUV
The parameters can be estimated from calculating the first principles of dynamics of
AUVs, from statistical LSE of experimental data, or a combination of the two. The
identification of the parameters of the item being tracked can be approached in many different
ways. The parameters can be determined once, from a model of the expected input signal.
It also could be done adaptively utilizing a recursive method to identify the parameters of
the incoming signal.
The modern era of system identification is marked by the implementation of the
maximum-likelihood method [20]. Therefore, we used the Least Square model for the
uncertainties in the parameters and measurements in this chapter. Based on this model, is a
vector of unknown constant parameters and v is a random vector of measurement noise. There
are two different models that are Bayesian and Fisher [21]. The LSE algorithm always gives
the best linear fit when the noise is white and Gaussian.
MODEL POSTULATION EXPERIMENT DESIGN
MEASURED DATA
DATA COMPATIBILITY
ANALYSIS
INPUT/OUTPUT DATA
MODEL STRUCTURE
DETERMINATION COLLINEARTIY
& DIAGNOSTICS
PARAMETER AND
STATE ESTIMATION
DIFFERENT SETS MODEL VALIDATION

OF DATA
Figure 12. AUV System Identification Method [21].

Moreover, it is easier to tune the identification process in the latter part with a priori
knowledge of the system [21]. A recursive LSE was chosen in this study for the parameter
estimation since it is easy to apply the system dynamic and allows for the consideration of
modeling and measurement errors.
The parameter estimation process consists of finding values of unknown model parameter
in an assumed model structure, based on noisy measurements z. An estimator is a function of
the random variable z that produces an estimate of the unknown parameters  . Since the
estimator computes based on noisy measurements z, ˆ is a random variable [21]. In order to
estimate the hydrodynamic coefficients of AUVs, the LSE is designed using the observer
model. The estimated parameters yields to
  [YvvYrrYuvYurYwp Nvv Nrr Nuv Nur N wp ]

(49)
The output y is given by:
y  X (50)
where the matrix X is assumed to be known. Then the measurement equation can be
expressed as:
z  X  v (51)
where is the length measurement vector, is the vector of unknown parameters (np=n+1), is the
matrix of vectors of ones and regressors, and is the vector of measurement errors.
In general, there are n0 measured outputs, and a vector of measurements is taken at each
sample i, where i = 1,2,3,…, Nis the number of sampled data points. A single measured
output is assumed, son0=1, and z is a vector composed of N scalar measurements [21].
Neglecting interactions from other parts of the system, the sway and yaw measurement
subsystems can be modeled as
ZYmeas  Y  mwp  (mz g )qr  (mxg ) pq  Yuur u 2 r

(52)
Z Nmeas  N  ( N pq  I xx  I yy ) pq  ( ygW )sin   Nuur u 2 r

(53)
For the least squares model, there are no probability statements regarding  , but v is
assumed to be zero mean and uncorrelated, with a constant variance:
E (v)  0 and E(vvT )   2 I (54)
The unknown parameter vector can be estimated via LSE as follows

ˆ  ( X T X )1 X T z (55)
The np  np matrix matrix is always symmetric. If the regressor vectors that make up the
columns of X are linearly independent, then is positive definite and the eigenvalues of are
positive real numbers, and the associated eigenvectors are mutually orthogonal so the exists
[21].
Finally, the covariance matrix of the parameter estimate also known as the covariance
matrix of the estimation error is simplified to
cov(est )  E[(est   )(est   )T ]   2 ( X T X )1 (56)
3.3. Simulation Results of Parameter Identification
Numerical simulations are made to show the parameter identifications of the AUV. The
hydrodynamic coefficients associated with horizontal and vertical motions are estimated by
simulating the combined yaw and sway motion of the AUV.
The numerical values of the AUV physical parameters, hydrodynamic force coefficients
and hydrodynamic moment coefficients are given in Tables 3-5 respectively.
We will describe how state-of-the art algorithms are used to find a dynamic model that
enables accurate simulation of steering dynamics.
In summary, the system identification task to estimate the parameters (56). The parameter
identification results are given in Table 2.
Numerical results demonstrate that the LSE estimates the hydrodynamic coefficients very
well.
Table 2. Parameter Identification Results
Parameters Real Value Est. Value with Int. Nav. Percent % Difference
Yvv -196,26 -192,650 1,839 3,610
Yrr 8,30 8,580 3,373 0,280
Yuv 38,39 38,277 0,295 0,113
Yur 13,41 13,320 0,674 0,90
Ywp 35,50 36,105 1,704 0,605
Nvv -7,95 -7,733 2,736 0,218
Nrr -24,13 -24,051 0,328 0,079
Nuv -21,89 -21,843 0,216 0,047
Nur 1,93 1,922 0,435 0,008
Nwp -19,30 -19,367 0,347 0,067
The simulation results of the real model and identified model are shown in Figures 13-15
respectively. Figures show that the AUV had a surge velocity changed between 5 and 0 m/s.
The depth changed between -15 m and -13.5 m. The elevator was operated at the maximum
deflection of 15 deg. for most of the time and similarly the rudder deflection operated at the
maximum deflection of -11 deg. The iteration number Nis chosen to be 5,000 and the
sampling time is chosen to be 0.001 sec. As seen from Figures 13, 14 and 15, outputs of real
and identified models are very close. On the one hand, this is obviously seen in Figure 16. As
a result, Simulation results show that the proposed method works very well. For model
verification, the existing in literature methods [10, 20, 22] can be used.
Figure 13. Velocity simulation results of real and identified model.
Figure 14. Position simulation results of real and identified model.

Figure 15. Attitude simulation results of real and identified model.
Figure 16. Attitude simulation results of real and identified model (Zoom in Version).
Table 3. AUV Physical Parameters [10]
Parameter Value SCI unit Definition

ρ 1010 g/m3 Fluid density
g 9.81 m/s2 Gravitational acceleration
d 0.191 m Hull cylindrical radius
l 1.33 m Overall hull length
Sw 0.7981 m2 Submerged Area (πdl)
Ap 0.2540 m2 Body base area (dl)
Af 0.0287 m2 Body cross area (πd2/4)
cds 0.004 - Surface drag coefficient
cdF 0.3 - Drag coefficient (Af)
cda 0.0166 - Total drag coefficient (for Sw)
Table 3. (Continued)

lcp 0.2645 M Distance to center of pressure
tfin 1010 g/m3 Fluid density
hfin 0.0960 m Rudder height
Ar, As 0.0071 m2 Vertical/horizontal rudder surface area
cdf 1.558 - Rudder drag coefficient
W 300 N AUV weight
B 306 N Buoyant force
xb 0 m Buoyancy center about the x-axis
yb 0 m Buoyancy center about the y-axis
zb 0 m Buoyancy center about the z-axis
xg 0 m Gravity center about the x-axis
yg -0.008 m Gravity center about the y-axis
zg 0.0196 m Gravity center about the z-axis
Ixx 0.177 kg m2 Moment of inertia about the x-axis
Iyy 3.45 kg m2 Moment of inertia about the y-axis
Izz 3.45 kg m2 Moment of inertia about the z-axis
R 5.87E-2 m Diameter of propeller
u0 1.5 m/s Design velocity
Table 4. AUV Hydrodynamic Force Coefficients [10]

X u|u| -6.68 kg/m Axial drag
Xu -0.513 kg Added mass
Yv|v| -196.26 kg/m Cross flow drag
Yr |r | 2
8.30 kg m/rad Cross flow drag
Yuv -38.39 kg/m Body and rudder buoyancy
Yv -42.13 kg Added mass
Yr -5.16 kg m/rad Added mass
Yur 13.41 kg/rad Added mass + Rudder buoyancy
Yuu r 16.99 kg/m/rad Buoyant force of rudder
Z w|w| -196.26 kg/m Cross flow drag
Z q| q| -8.30 kg m/rad2 Cross flow drag
Zuw -38.39 kg/m Body and rudder buoyancy
Zw -42.13 kg Added mass
Zq 5.16 kg m/rad Added mass
Zuq -13.41 kg/rad Added mass + Rudder buoyancy
Zuu s -16.99 kg/m/rad Rudder buoyancy
Table 5. AUV Hydrodynamic Moment Coefficients [10]

K p| p| 2 2
-5.03 kg m /rad Roll drag
Kp -0.095 2
kg m /rad 2
Added mass moment
M w|w| 7.95 kg Cross flow drag
M q| q| 2 2
-24.13 kg m /rad Cross flow drag
Added mass + Rudder buoyancy +
M uw 21.89 kg
MunkMomenti
Mw 5.16 kg m Added Mass Inertia
Mq 2
-7.57 kg m /rad Added Mass Inertia
M uq -16.56 kg m/rad Added mass + Rudder buoyancy
M uu s -13.92 kg/rad Rudder Buoyancy Moment
N v|v| -7.95 kg Cross flow drag
2
N r |r | -24.13 kg m /rad Cross flow drag
Added mass + Rudder buoyancy +
Nuv -21.89 kg
MunkMomenti
Nv -5.16 kg m Added Mass Inertia
2
Nr -7.57 kg m /rad Added Mass Inertia
Nuu r -13.92 kg rad Rudder Buoyancy Moment
CONCLUSION
The research on underwater systems has gained immense attention during the last two
decades because of applications taking place in many fields. Therefore, the significant
number of UUVs has been developed for solving the wide range of scientific and applied
tasks of ocean research and development in the world. The military, as well as civilian
industry can see great potential uses of AUVs in the underwater environment.
This chapter is concerned with implementing dynamic modeling of an AUV. This chapter
has also given an insight and theoretical background about AUV kinematics, hydrodynamics
and hydrostatic. Because physically based mathematical modeling of an AUV is an inter-
disciplinary study of kinematics, hydrostatics, and hydrodynamics.
To achieve an accurate hydrodynamic model for any vehicle that is not a standard shape
by purely analytical means is extremely difficult at best.
This chapter also discussed the use of parameter identification methods to predict the
hydrodynamic derivatives of the AUV. LSE is used to tackle the problem of parameter
identification of an AUV. This classical approach seems to have a better performance in cases
where a specific parameter will be identified and the LSE results obtained are satisfactory.
Therefore, numerical results demonstrate that the proposed method works very well, in both
cases of using simulated data. It has been shown that the characteristic functions for LSE can
be obtained in a stable way even in noisy environments.
Numerical simulations are made to show the parameter identifications of the AUV. The
hydrodynamic coefficients associated with horizontal and vertical motions are estimated by
simulating the combined yaw and sway motion of the AUV.
The simulation results indicate that for AUV hydrodynamic parameter identification
purposes, the LSE algorithm remains a feasible tool, which consistently returns quality results
and is the least costly in terms of computational demand.
Autonomous guidance, navigation, and control techniques are key research and
development areas for success of the specific AUV missions.
Moreover, further work is needed for in precision navigation, sensor development and
integration, and improving the reliability and robustness of long term and complex mission
completion.
REFERENCES
[1] Inzartsev, A. V., Underwater Vehicles. In-Tech, Rijeka, 2009.
[2] Narimani, M., Loueipour, M. and Depth Control of Autonomous Underwater Vehicles.
UDT Europe, 2007.
[3] Fossen, I. T., Guidance and Control of Ocean Vehicles. University of Trondheim, John
and Wiley Sons Press, 1994.
[4] Fauske, K. M., Gustafsson, F., Hegrenaes, O., Estimation of AUV Dynamics for Sensor
Fusion. 10th International Conference on Information Fusion, 2007.
[5] Hwang, W. Y., Application of System Identification to Ship Maneuvering. Ph.D.
Dissertation, Massachusetts Institute Technology MIT, Cambridge, US, 1980.
[6] Meng, L., Li, L., Veres, S. M., Aerodynamic Parameter Estimation of an Unmanned
Aerial Vehicle Based on Extended Kalman Filter and Its Higher Order Approach. 978-
1-4244-5848-6, 2010.
[7] Chowdhary, G., Jategaonkar, R., Aerodynamic Parameter Estimation from Flight Data
Applying Extended and Unscented Kalman Filter. Aerospace Science and Technology,
vol. 14, pp. 106-117, 2010.
[8] Marco, D. B., Martins, A., Healey, A. J., Surge Motion Parameter Identification for the
NPS Phoenix AUV. Naval Postgraduate School, Center for AUV Research, Monterey,
CA, 93943-5000, 2005.
[9] Vandersteen, G., Rolain, Y., Schoukens, J., Pintelon, R., On the Use of System
Identification for Accurate Parametric Modeling of Nonlinear Systems Using Noisy
Measurements. IEEE Transactions on Instrumentation and Measurement, vol. 45, no.
2, pp. 605-609, 1996.
[10] Prestero, T., Verification of a Six-Degree of Freedom Simulation Model for the
REMUS Autonomous Underwater Vehicle, M.Sc. Thesis in Ocean and Mechanical
Engineering, MIT, US, 2001.
[11] Healey, A. J., Good, M. R., NPS AUVII Autonomous Underwater Vehicle Testbed:
Design and Experimental Verification. Naval Engineers Journal, vol. 104, pp. 191-202,
1992.
[12] Hemminger, D. L., Vertical Plane Obstacle Avoidance and Control of the REMUS
AUV Using Forward Look Sonar. M.Sc. Thesis, Naval Post Graduate Scholl, US, 2005.
[13] Bekir, E., Introduction to Modern Navigation Systems. World Scientific Publishing Co.
Pte. Ltd., 2007.
[14] Guo, Z., Hao, Y., Sun, F., Gao, W., Design of Seamless Changing Method for AUV
Integrated Navigation System. First International Workshop on Education Technology
and Computer Science, IEEE Computer Society, pp. 160-165, 2009.
[15] Healey, A. J., Dynamics of Marine Vehicles (MA-4823). Class Notes, Naval
Postgraduate School, Monterey, CA, 1995.
[16] Antonelli, G., Fossen, T. I., Yoerger, D. R., Underwater Robotics. US, pp. 989-996,
2007.
[17] http://en.wikipedia.org/wiki/Buoyancy.
[18] Dinc, M., Hajiyev, Ch., Development of Autonomous Underwater Vehicle Navigation
System Based on Kalman Filtering Technique. 18th St. Petersburg International
Conference on Integrated Navigation Systems (ICINS-2011), St. Petersburg, Russia, pp.
238-242, 2011.
[19] Mendel, J. M., Lessons in Estimation Theory for Signal Processing, Communications
and Control, Prentice Hall PTR, Englewood Cliffs, NJ, US, 1995.
[20] Jategaonkar, R. V., Flight Vehicle System Identification: A Time Domain Methodology.
Reston, Va.: AIAA, 2006.
[21] Klein, V., Morelli, E. A., Aircraft System Identification Theory and Practice. Reston,
VA.: AIAA, 2006.
[22] Gadzhiev, Ch. M., The Information Provision of Offshore Platform Supervision and
Control. Elm, Baku, Azerbaijan, 1996 (in Russian).
[23] Yalcin, L., Ilgaz, M., Erdem, B., Hydrodynamic Analysis of DARPASUBOFF Model.
SubSeaTech, 2007.
Chapter 4
ADVANCES IN ENVIRONMENT SENSING AND

PERCEPTION TECHNOLOGIES AND ALGORITHMS
FOR AUTONOMOUS GROUND VEHICLES
Constantin Ilas,*1 Irina Mocanu1 and Mariana Ilas2

1
Department of Computer Science,
University Politehnica of Bucharest,
Bucharest, Romania
2
Department of Electronics, Telecommunications and IT,
University Politehnica of Bucharest, Bucharest, Romania
ABSTRACT
This chapter discusses the main environment sensing and perception technologies
and algorithms used in autonomous ground vehicles (AGVs). For the beginning, a review
of main environment sensing technologies is presented, focusing mostly on LIDAR,
video camera and radar. The chapter shows that in the area of environment sensing
technologies, several main trends exist today, including an increased usage of video
cameras, as well as introduction of integrated solutions, consisting of two complementing
technologies, such as video camera plus LIDAR or video camera plus radar. The
objectives and the generic architecture of the perception subsystem are then presented
and the most used perception algorithms employed in AGVs are discussed, insisting on
classification algorithms. Also, the chapter discusses the most used feature descriptors
employed by these algorithms, i.e. HOG and DSIFT. The chapter includes a review of
several implementations, as presented in the literature in the past years. Thus, the main
trends are revealed: increasing the complexity and the performance of algorithms,
reducing the number of sensors and increasing the usage of video camera. The chapter
presents an implementation of the SVM algorithm based on video camera images and
compares the performance of this method when used with HOG and DSIFT respectively.
Using four different classes, the classifier gives information not only on the global car vs.
non-car decision, but also on the orientation and relative distance of the detected vehicle.
The overall performance of this classifier, in terms of precision and recall rate is
*
E-mail: constantin.ilas@cs.pub.ro.
114 Constantin Ilas, Irina Mocanu and Mariana Ilas
presented. It confirms the expectation that video camera can play a more important role
in the perception sub-systems of AGVs.
Keywords: Autonomous Ground Vehicle, Driverless Car, Environment Sensing

Technologies, Image Classification, Support Vector Machine, Dense Scale-Invariant
Feature Transform, Histogram of Gradients, Video Camera, LIDAR
NOMENCLATURES
AGV Autonomous Ground Vehicles
AEB Autonomous Emergency Braking Systems
DARPA Defense Advanced Research Projects Agency
DSIFT Dense Scale-Invariant Feature Transform
ESC/ESP Electronic Stability Control/ Program
FOV Field of View
HOG Histogram of Oriented Gradients
IMU Inertial Measurement Unit
LIDAR Light Detection and Ranging
SIFT Scale-Invariant Feature Transform
SURF Speeded Up Robust Features
SVM Support Vector Machine
US Ultra-Sonic
1. INTRODUCTION
In recent years, there has been a substantial research activity in the area of autonomous
ground vehicles (AGVs). Today, several prototypes developed by universities as well as
companies exist [1-4]. Many companies are working on such vehicles or related technologies
[5, 6], some of them envisioning first commercially available autonomous cars (with limited
functionality) in the medium future [5]. In parallel, semi-automated driver assistance systems
have started to be introduced in production cars and much more have been announced for the
near future [6]. The fact that AGV is a hot topic is confirmed by the studies related to its
impact and implications [6, 7, 8] as well as by the fact that discussions on this topic have
grown during the past year (mid-2012 to mid-2013) in the US and so has the consumer
receptivity [6]. Washington DC and three US states allow the operation of AGVs on public
road, for testing purposes [6].
Despite this evolution, the technical challenges on the development of fully autonomous
vehicles are very high and sustained research activity is still needed in order to improve the
performance and the robustness of these prototypes. The sensing and the perception
technologies play a critical role in the overall performance of AGVs, and consequently they
are areas of sustained research activity.
Advances in Environment Sensing and Perception Technologies … 115
Figure 1. Control system architecture of an autonomous robot - high level diagram.
Looking at the history of the evolution of AGVs, we can see that important milestones
were the challenges organized by the Defense Advanced Research Projects Agency -DARPA.
DARPA initiated and organized some challenges for driverless vehicles. Thus, the first two,
organized in2004 and 2005, were held in a desert area, while the third, from 2007 was held in
an urban environment [4, 9, 10, 11, 12, 13]. Many prototypes of autonomous vehicles were
designed and developed in order to participate in these challenges. After 2007, research has
continued to evolve, key focus areas being perception algorithms as well as sensing devices.
The main goal of these recent research activities has been to improve the performance and
robustness of the perception and sensing stages.
The AGV evolution is in correlation with the evolution of other types of autonomous
robots, for different applications. These include not only ground robots, but also robots
designed to move and operate in other environments, such as on/ in water or in air [14].
However, despite the fact that all autonomous robots share a similar operating principle of the
control system, their specific challenges can be very different [1, 2, 15]. Thus, in the case of
AGVs the navigation speed should reach high values (similar to normal cars), while the
environment can be unstructured, i.e. complex and rapidly changing, or even extremely
unstructured. Normally, the urban traffic is the highest on the levels of unstructured
environment. This is mainly because of existence of multiple intersections, pedestrian
crossings, traffic signs, as well as of a variety of other traffic participants. These include
vehicles which move in different directions, change their lanes or speed, but also vehicles
which are parked or enter the traffic, as well as pedestrians who cross the street (not always
on the crossing areas), or approach crossings, etc.
As the history of the DARPA Challenges confirms, autonomous ground vehicles have
evolved from first prototypes operating in different types of desert areas (i.e. medium-
structured environments) towards complex urban traffic, which is a very unstructured
environment.
Figure 1 [15] shows the high-level diagram of AGV (and autonomous robots in general)
architecture, with its main subsystems. Thus, the course information on the environment is
captured by sensors and transmitted to the perception subsystem. This extracts the useful
information, and outputs data that can then be used by planning and control subsystem. Based
on this information, as well as on navigation algorithms, this subsystem generates the
references used by the actuators control subsystem. For both perception and planning and
control subsystems, several algorithms can be used, thus different levels of precision and
complexity being possible.
2. MAIN TYPES OF ELECTRONIC SENSING DEVICES FOR

ENVIRONMENT PERCEPTION IN AUTONOMOUS GROUND VEHICLES
By sensing devices for environment perception we understand the devices providing
information on the vehicle surroundings (such as road/terrain, obstacles, traffic signs, etc.).
The data obtained from these sensors is then processed by perception algorithms in order to
extract real-time, accurate information, based on which navigation decisions can be made.
Since the algorithms as well as the performance and robustness of perception depend on the
types and characteristics of the environment perception sensors, it is useful to review the main
types of such sensing devices. As first shown in [15], we will analyze both those used in
AGV prototypes as well as those currently employed in mass production cars.
2.1. Type of Electronic Sensing Devices for Environment Perception Used in

AGV Prototypes
The main types of electronic sensing devices used in AGV prototype cars for
environment perception are:
• LIDAR (Light Detection and Ranging) sensors;

• cameras;
• radars;
Other types of sensors which are less frequently used include ultra-sonic (US) sensors
and contact sensors. It is worth mentioning that many AGVs use GPS for position detection.
GPS accuracy can be increased if data from IMU (Inertial Measurement Unit) devices is used.
Also, IMU data can also be used in some applications to adjust the indications from some of
environment perception sensors.
LIDAR, radar and US are active sensors, and are all based on the principle of sending a
pulse of wave (of certain nature and frequency), receiving it reflected by the obstacle and
measuring the total time of flight.
An important parameter of LIDAR sensors is the range in which they can detect (and
determine the distance to) objects. The range is heavily influenced by the reflectivity of the
object as well as by weather conditions (such as humidity). Therefore, range is sometimes
indicated by either its maximum value, or by a value at which poor reflecting objects
(typically objects with 10% reflectivity) can be detected. For example, the maximum range
can be 80 m, whereas its value for objects with 10% reflectivity is 30 m. LIDAR sensors can
be referred to as long-range (typically with maximum range beyond 50m) or short-range.
By definition, a laser beam is highly concentrated, meaning that in order to scan a region
of the 3D environment, either the beam periodically changes its angle, or multiple beams are
used. In reality both approaches are used simultaneously. Based on the shape of the
environment part they scan, LIDARs can be either 2D (or planar), when they scan in a given
plane, or 3D, when they cover a 3D region of the surrounding space. 2DLIDARs spin and
scan radially (typically in the horizontal plane), covering a sector of a circle, typical values
including 80o or 270o. This is also referred to as field of view (FOV). 3DLIDARs cover, for
each given position in the horizontal plane, a given sector in the vertical plane. A 3DLIDAR
scanning sector is indicated by the horizontal (azimuth) coverage (e.g. 270o, or 360o) as well
as the vertical (elevation) coverage (e.g. 26.8o). To achieve a 3D coverage, several laser
beams are positioned so that together they cover the given part of the vertical plane, whereas
the entire unit spins to cover the horizontal one. Because typically a laser beam covers less
than one degree, a few dozens of laser beams are normally used (e.g. 32 or 64). For each of
these beams, separate emitter and receiver units are needed.
Cameras provide digital image of the covered region of space. They are based on an array
of passive light sensors. Each sensor determines the value of an image pixel. The image can
be either monochrome (grey-scale) or color and the resolution can be in the range of 640x480
to 1328x1048 pixels. Since the image is 2D, distance to particular objects in image can only
be determined using complex processing algorithms.
In automotive applications, there are two particular challenges for a camera: the
sensitivity to low intensity of light as well as the capability of simultaneously capturing very
bright and dark areas (also referred to as dynamic range). The sensitivity to low intensity light
is given by the efficiency of the sensor as well as by the electronic amplification of the signal.
The higher this gain, the more sensitive is the camera, but the amplification has to be applied
only to useful information (not the noise). For night vision, cameras should capture also infra-
red (IR) radiations. Typically, cameras spectrum cover also near-IR, which is reflected by
objects also at night. It is also possible to use an external IR source (IR LED) to illuminate the
objects, thus enhancing the images in the near-IR. Far-IR (thermal) cameras, which measure
the IR radiation emitted by objects (thermal radiation) are not commonly used in automotive.
Considering that the intensity of the light reflected by an object can vary from 10-1 lux
(night) to 105 lux (sunshine), the dynamic range for very good cameras for automotive should
be 120 db (standard cameras offer around 60 db). Radar technology uses electromagnetic
radiation in several frequency bands (e.g. 24 GHz, 77 GHz). It has the advantage of being less
affected by weather conditions, but, in order to achieve a good angular resolution, the emitted
signal has to be restricted to a narrow beam. Also, it is possible to employ several other
techniques to improve the resolution (such as direction of arrival method, spatial power
spectrum measurement, etc.).
Radars are classified as short - sometimes called also mid-range - (up to 100-150m) or
long-range (up to 250 m). Obstacles at short distances (a few meters) can be detected only by
short-range radars. Their FOV is typically 45o (for front radars) and the resolution within a
few degrees. Given the range (typically above 100m), and the resolution, radars can detect
possible obstacles at a high distance, but the resolution normally allows only preliminary
actions to be taken.
An important advantage of radars is that they also can determine the relative speed of the
detected obstacle, which is very important for decision making in AGVs as well as for
applications such as automatic braking or ACC.
Based on the above considerations, as well as by the results in [16], we summarize the
main properties of the main technologies for environment perception in Table 1. Camera is a
particular case, as by itself it cannot detect objects, but it depends on image processing
algorithms to do so. As it can be seen, none of these technologies is ideal and an attractive
idea is to complement two or more to obtain best results.
Table 1. Comparison of main technologies for environment perception
Criteria LIDAR Radar Camera US

Only for short Very good
Very short range (0-1m) detection Poor Ok
range radar
Short range (1-30 m) detection Very good Very good Good Poor
Long range (30-100+ m)
Medium Very good Poor No
detection
Angle < 10 Very good Good Good Poor
Angular resolution Very good Good Good Poor
Velocity measurement No Yes No No
Operation in adverse weather
Poor Very good Poor Good
conditions
Operation at night Very good Very good Limited Very good
2.2. Types of Electronic Sensing Devices for Environment Perception

Currently Used in Mass-Production Cars
When discussing about the sensors for the future AGV, it is interesting to have a brief
review of the types of sensors for environment perception used in current mass-production
cars. As mentioned before, some mass-productions cars have started to offer advanced driver
assistance features, in some cases even limited versions of some of the functions of an AGV.
Consequently, these cars have been equipped with several types of advanced sensors for
environment perception. Most common are radars and cameras, which are replacing more
mature US sensors [17]. More recently, also (short range) LIDARs have been introduced, and
also an integrated camera and short range LIDAR device has been announced [18, 19].
These LIDARs currently operate for short distances (up to around 10 m) and can detect
other cars (and their speed) in a couple of sectors (e.g. front, front-left, front-right), thus
facilitating the decision of braking vs. steering. In the near future, with the announced of the
integrated camera and LIDAR, the reliability of the perception will be improved (by
comparing the data provided by each of the two sensors).
Multi-layered LIDARs (combining usually 4 different planes of view) seem particularly
useful for automotive, allowing for object detection even when, due to humps or holes on the
road, a simple 2DLIDAR beam would not reach the obstacle.
Radar deployment is increasing. A challenge for introducing radars on mid-range
production cars is the relatively large size of the device, because of the size and number of the
antennas and because of the semiconductor technology which offers limited integration
capabilities [20]. However, in recent years there has been a great amount of research activities
for finding new, more compact and cheaper technologies [20, 21, 22]. Also, many current
cars have inertial sensors (in most cases needed for electronic stability control/program -
ESC/ESP) [23].
A general trend is integration of two or more sensors into one single device. Most
noteworthy are the devices combining a camera and a radar [24], and even a camera and a
LIDAR [19]. This is justified by the fact that the two systems complement each other for
better perception and by the need of having fewer and more compact devices.
Generally speaking, the evolution of object detection sensors in mass-production cars
(both long range and short range) is stimulated by the growing importance of Autonomous
Emergency Braking Systems (AEB). Such systems will be included, starting with 2014 in the
Euro NCAP rating scheme [25], therefore they will probably be more widely applied to mid-
range cars also.
As already happening [15, 26] some of the algorithms designed in AGV prototyping for
perception and control, may be adapted in the future to be used in various systems for driver
assistance. It is also interesting the trend [15] of decreasing the total number of sensors in
AGV prototypes. This is possible because of the evolution of the perception algorithms,
through a more optimal design of the architecture of sensing and perception subsystems.
Based on [26], we will discuss the objective and the architecture of the perception subsystem
in AGVs, as well as review the main algorithms used in this subsystem.
3. OBJECTIVES AND ARCHITECTURE OF PERCEPTION SUBSYSTEM

3.1. Objectives of Perception
The perception subsystem [26] has the goal of determining the parameters required by the
motion planning and control subsystems. This is done by processing the information provided
by various sensors. The output parameters required include the coordinates of the road, the
position and speed of obstacles (on the road), the position of pedestrian crossings, the position
of lane markers, etc. It seems natural that the outputs of this subsystem would depend on the
type of environment for which the vehicle is designed to navigate [26]. Of course, the more
unstructured the environment is, the higher is the number and complexity of its outputs
(Table 2).
Thus, in highly unstructured environments, such as urban traffic, the perception
subsystem is required to detect diverse types of objects and to classify them, but also to track
them and to predict their future position. This results in a perception subsystem with a very
complexity. Such a classification is useful to understand the perception subsystems used in
different prototypes, as well as those designed for particular applications. A fully operational
vehicle, designed to navigate in most of these environments, including urban, (as current
manned vehicles do) would clearly need to determine all these parameters.
As the review of literature will reveal (see following paragraphs), many papers discuss
implementation and results of only a limited portion of the perception subsystem. This is due
to the complexity of the algorithms required for perception as well as to the growing maturity
of the area, which allow researcher to reuse existing solutions and to focus on improving
particular blocks and algorithms. Also, some AGV prototypes are designed for particular
applications, and hence they operate only in specific environments such as off-road, or in
simpler conditions, such as vehicle follower. Clearly, in these cases, the perception subsystem
has reduced requirements.
3.2. Generic Architecture of Perception Subsystem
The perception subsystem consists in general of three main blocks: detection,

classification and tracking. Detection is the first step and it deals with separating different
objects from the environment data provided by the sensors. For sensors such as cameras or
LIDAR, the data is a matrix containing information for each point in the area covered by that
sensor.
The number of points depends on the sensor resolution and is in the range of 1M for each
image of a camera and 1M per second in the case of 3DLIDARs [27].
Classification refers to identifying the detected object as belonging to a particular class
(e.g. car, pedestrian, etc.) whereas tracking deals with determining the change in the position
of a given object from one set of data to the other.
Tracking is very useful for determining the speed of the respective object and for
predicting its future movement [28]. Apart from detection results, tracking and classification
use also data from each other, to enhance their performance. A diagram of the perception
subsystem is presented in Figure 2.
Table 2. Perception subsystem outputs for different type of environments
Type of environment
Off-road Highway Country
Urban
Perception output (variable (medium roads
(highly
degree of degree of (low degree
unstructured)
structure) structure) of structure)
Obstacle position Needed Needed Needed Needed
Sometimes
Obstacle speed Needed Needed Needed
needed
Sometimes
Road Needed Needed Needed
needed
Lane markers n/a Needed Needed Needed
Traffic signs n/a Needed Needed Needed
Pedestrian crossing n/a n/a Needed Needed
Sometimes Sometimes
Obstacle classification Not needed Critical
needed needed
Obstacle movement Sometimes Sometimes
Not needed Critical
prediction needed needed
In many implementations presented in the literature, several particular aspects exist. For
instance, in order to limit the total time required for perception, several tasks are done in
parallel, on independent subsystems.
This can be the case of lane markers detection, which may be done with dedicated
sensors (planar LIDARS or camera) and also traffic signs can have a dedicated subsystem,
based on a separate camera [29, 30]. Similarly, radar is used for detecting fast moving cars,
especially outside cities.
Figure 2. Architecture of perception system in AGVs.
4. MAIN ALGORITHMS USED FOR PERCEPTION

In this section we will give some details on the main algorithms used in AGV for object
detection, classification and tracking. Data used for complex perception algorithms comes, in
most AGVs, from camera or 3GLIDARs. In both cases it has the form of a matrix, each point
containing a structure of several data. Because of this, the algorithms used for perception are
usually similar (at least as principle). In the description of the algorithms, we will use the term
of image to describe the matrix output of either cameras or LIDARs.
4.1. Detection Algorithms
The goal of detection is to separate the objects existing into an image frame. Several
approaches exist, including background subtraction and segmentation. Background
subtraction is based on the idea of eliminating the image background to find existing objects.
This is best used for identifying moving objects, when the background is relatively fixed from
one frame to the other. In such case, the difference in the image comes from the moving
objects. Segmentation separates the regions of the image which have “similar” points.
Consequently, a segmentation algorithm uses a certain quantification of similarity. One of the
segmentation algorithms is the mean-shift clustering. It starts with a large number of potential
cluster centers, spread across the image. At each step, the new centers are moved to the mean
of the data contained inside an ellipsoid constructed based on the old centers. The vector
defined by the old and the new cluster center is called mean-shift vector. The algorithm is
repeated until all centers remain fixed. During this process, several potential cluster centers
may merge.
4.2. Tracking Algorithms
Tracking can be formulated as the correspondence of detected objects represented by

points across different frames [28]. Several solutions for this problem exist [28], one of the
most widely used being the Kalman filter. A Kalman filter starts from the basic structure of a
linear observer, but the gain matrix (Kalman gain) is computed as a solution of the Riccati
equation, taking into account the noise and its covariance [28]. Kalman filters are designed
for state variables having a Gaussian distribution. For other type of distributions, a particle
filter may be used [28].
4.3. Classification Algorithms
4.3.1. SVM
Support vector machine (SVM) is a non-probabilistic binary classifier. Using a set of
training data, it determines on which of the two possible classes a certain object belongs.
Because it is a binary classifier, it has to be run several times in order to classify objects
belonging to more than two possible classes. Basically, it starts from a set of training vectors
(xi, yi) of dimension L and based on them, it computes i coefficients (by solving a quadratic
problem [31, 32]), and the support vectors S (of size Ns) are then given by the (x, y) elements
corresponding to i > 0. With these the w and b value can be computed:
L
w   i yi  xi (1)
i 1
1  
b   y j   k yk xk  x j  (2)
Ns jS  kS 
And finally, for any new input vector x , the classification is given by:
y  sign  w  x  b  (3)
As presented, this is a linear problem, but SVM can also be used for non-linear ones, by
performing the so-called kernel trick [32].
The linear SVM full algorithm [33]:
Notation: The training set consisting of L classified examples {(x1, y1), …, (xL, yL)}, with
yi either 1 (meaning vector xi belongs to the specified class, or -1 (if it doesn‟t).
Training (obtaining the SVM):
1. Set up and solve the dual optimization problem (using a quadratic programming
solver):
L
1 L
maximize  i   
 i yi y j xi  x j  j
2 i , j 1
(4)
i 1
subject to: i  0 (5)
L
and  i yi  0 (6)
i 1
2. Determine the weights:
L
w   i yi xi (7)
i 1
3. Determine the set of support vectors x j  S , by finding the indices j so that  j  0 .
 
Note that for any support vector, we have: y j w  x j  b  1 .
4. Determine b coefficient by solving the above relation over the entire set of support
vectors (get an average value):

yj w xj  b 1  (8)
Classification:
For any new x , the classification output y  (1 if it belongs to that class, -1 otherwise) is
determined, based on the above w and b, as:
y  sign  w  x  b  (9)
The value w  x  b is referred to as the classification score and it gives information on

the confidence of the estimation, and can also be used to compute the probability of correct
estimation, i.e. the posterior probability.
Figure 3. Illustration of the mapping principle used in SVM for non-linear classification problems.
In some situations, the vectors xi cannot be perfectly separated by solving this linear
problem. However, it was shown [32] that it is always possible, for a finite set of points, to
find a dimension where separation of the point set can be achieved by a hyper-plane. This
means that the vectors xi should be mapped with a mapping function (x) to a space of
sufficiently high dimension so that they become separable. In this case, the original space of
vectors xi is called input space, whereas the space of the mapped vectors is called the feature
space. In the feature space the linear SVM can be applied, see Figure 3. Thus, in such cases
the above algorithm can be applied, by replacing x with (x). The mapping function (x)
depends on the actual problem and, it was shown [32] that an adequate selection of this
function can also increase the performance of the SVM. This is because if data is not
perfectly separable in one dimension, the SVM will still work, but with a higher error rate.
For instance, in the case of a handwriting recognition problem, linear SVM has around 8.5%
error rate, whereas, with a particular mapping, the error rate is reduced to around 1% [32].
A typical mapping function used in many applications is the polynomial mapping, which for
2 dimension input vectors is:
 : R 2  R3 ,  xi1 xi 2    zi1 zi 2 zi 3 
 
(10)
  xi1 , xi 2   xi21 2 xi1 xi 2 xi22
Interestingly, computation error does not increase massively with mapping, because the
only operation needed in the feature space is the vector product. This can be verified easily by
replacing x with (x). For instance, if we take the last equation (the classification stage):
 L 
y  sign  w  x  b   y   sign  w  ( x)  b   sign   i yi ( xi   ( x)  b) (11)
 i 1 
This operation of computing the vector product in the feature space is also called
computation of the kernel function:
K  x1x2     x1     x2  (12)
It can be shown [32, 33] that the kernel function can be computed directly, which is less
complex than performing the actual vector product in the feature space. For instance, in case
of the above polynomial mapping, the kernel function can be directly computed as:
K  x1 x2    x1  x2 
2
(13)
instead of:
K  x1x2     x1     x2   x11
2
 2 x11x12 2
x12
 x21
2
2 x21 x22 2
x22  (14)
This property of the kernel function gives the name of kernel trick used when referring to
the SVM computations in the feature space. Without getting into details [32, 33], we will note
that mapping the vectors into a higher dimension increases the likelihood that data can be
separated, but does not guarantee this. Hence, it is not surprising to note that finding the
mapping function for a practical problem is not always an easy task [32].
Despite the fact that the output of the SVM classifier is binary, a posterior probability can
be calculated, starting from the SVM classification score. There are several techniques for
this, such as binning, sigmoid fitting or isotonic regression Bayesian framework [34]. Also
there are some approaches extending the SVM formulation to handle multi class
classification. Several ways are defined in order to extend SVMs [35, 36, 38, 39]:
• One-Versus-All SVMs (one-versus-rest SVMs), convert a k-class problem into k

two-class problems [35, 37, 38]: all the training samples are used for training each
SVM. For each SVM, i, i ≤ k, samples are labeled as positive or negative: samples
from the in the ith class are positive, and the other are negative. After that five-fold
cross validation is used for parameter selection. During the test phase, the ith SVM
separates class i from all other classes.
• Pair-Wise SVMs (one-versus-one SVMs), use a SVM for each pair of classes, thus
converting a k-class problem into k*(k-1)/2binary (two-class) problems. Each of
these SVM classifiers is trained for its two classes. Thus, we can denote sij the SVM
classifier which is trained on data from the ith class and jth class, with i< j. The
classifier parameters are automatically selected by a five-fold cross validation.
During the testing phase, the sij SVM divides all data into classes i and j. The two-
class classifiers have to be combined using a voting scheme. Possible voting schemes
include:
• Max-Win Majority Voting (based on the Max Wins strategy described in [40]):
for an input x, if the SVM classifier sij categorizes x in the kth, where k {i, j}
class, then the vote of x for the class k increases by one. Once all classifiers have
voted the class that the largest voting score is assigned to x. In [38] it is chosen
the class with a smaller index as the prediction result when two classes have
identical voting scores.
• Confidence Score Voting: as described in [41], proposes an absolute decision
value distance measure indicating the closeness between a sample and the
optimal separating hyper-plane. If the absolute value of confidence score is
small, the sample is close to the optimal separating hyper-plane; if the
confidence score is large, the sample is far away from the optimal separating
hyper-plane. Decision values are computed by:
L
D( x)   i yi K ( xi , x)  b
i 1 (15)
If +1 and -1 are labels of the two classes, a confidence score is imposed on the decision
value: -1 ≤ d ≤ 1, in order to ensure high generalization ability (if d> 1, d is rounded to 1; if
d< -1, d is truncated to -1). In case of SVM classifier sij categorizes x into the kth class, k {i,
j}, then the vote of x for the class k increases by |d|. Once all SVM classifiers have voted,
sample x is categorized to the class that has the largest voting score.
4.3.2. Boosting Algorithms

Boosting framework is a method of combining the classification given by a set of
multiple, simple classifiers (weak-classifiers) into a single, accurate one (strong-classifier).
Ada Boost, Gentle Boost and Joint Boost are some versions. The weak classifiers are
constructed in several ways. For instance, the weak classifiers can be constructed iteratively,
using different weighting factors for the training data, factors which are computed based on
the error obtained by the previous classifier.
The Adaboost algorithm (Ada stands for Adaptive) has the following steps [42]:
Notation: The training set consisting of L classified examples {(x1, y1), …, (xL, yL)}, with
yi either 1 (meaning vector xi belongs to the specified class, or −1 (if it doesn‟t). M is the
number of weak classifiers used.
Training:
1. Initialize the observation weights
1
wi0  (16)
L
2. For m=1, …, M , (and for all training vectors, i.e. i=1, …, L):
- Construct/select a (weak) classifier hm(x) using the wi weights, so that it minimizes:
 wim1
yi  hm  xi 
(17)
- Compute the error of this classifier:
 wi I  yi  hm  xi  
L
errm  i 1 (18)
 i 1wi
L
 1 yi  hm  xi 
I  yi  hm  xi     (19)
0 otherwise
- Compute
 1  errm  
m  0.5 log   (20)
 errm 
- Set new, increased weights for the points where the (weak) classifier gave a false
output:
1  errm
wim  wim1 exp  m  I  yi  hm  xi    wim1 I  yi  hm  xi   (21)
errm
3. The final (strong) classifier is obtained as:
 
h  x   sign   m hm  x   (22)
 m 
Classification: apply the final classifier for any new vector x‟:
 
y  h  x '  sign   m hm  x '   (23)
 m 
The number of weak classifiers, M, can be set at the beginning of the training stage, or it
can be determined during the iterations, e.g. until the classifier error cannot be further
minimized.
4.3.3. Bayes Filter

Considering the conjoint probabilities:
P  B A -the probability of B if A is true and P  A B  - the probability of A if B is true,
an important Bayes‟s theorem gives the relationship between these two probabilities in the
form [36]:
P  A P  B A 
P  A B  (24)
P  B
In classification problems, we can consider that A is our hypothesis (prediction) based on

an old set of data, and B is a new set of data. Applying Bayes‟s theorem we obtain that: the
updated probability of the hypothesis A based on the new data B equals the product of the
prior probability P(A) and the probability of data under the hypothesis (or likelihood)P(B|A),
divided by the probability of the new data P(B). If P(B)cannot be computed, it can be
regarded just as a normalizing constant. In other words, the posterior (probability of our
hypothesis) is the product between the prior (probability of our hypothesis) and the likelihood
(of the new data to exist, considering our hypothesis), divided by a normalizing term.
In classification problems, we will compute in parallel both the probability of object
belonging to class A and, respectively to class B. The approach can thus be extended to cover
multiple classes, or, if just a binary classification is needed, class B means that object does not
belong to class A. The algorithm starts with an initial probability P(A) and P(B), which can be
based for instance on the statistical differences of class A and B sizes. Then, at each step a
new feature is read and the two probabilities are updated, considering that the probability of
the existence of that feature for objects in class A and B respectively, is known. These
probabilities are determined during the training stage.
The algorithm steps are [43]:

Notation: The training set consisting of L classified examples {(x1,y1), …, (xL, yL)}, with
yi either 1 (meaning vector xi belongs to a specified class A, or −1 (if it doesn‟t belong, or
otherwise said it belongs to class B – of “non A” objects). Consider that the elements of all
training vectors xi(each having size M) belong to a finite set D, of size N.
Training:
Compute the probability of each of the n elements dj of set d to appear in a training vector
x if the object belongs to class a or class b respectively.
 j y1..1N I  xi  D j 
i

P Dj A   (25)
 j 1I ( yi  1)
M
Compute the overall probability of an object to be in class a or class b based on the

training set
 y 1yi  y 1 yi
P  A  i
, P  B  i
(26)
L L
Classification:
Start with the initial probability of the object being in class A or B.

For each new object, described by vector x‟:
1) For each element of x , xi , with xi  D j : read from the training results the
probabilities that the element appears for class A and respectively class B objects: P D j A  

and P D j B . 
2) Compute the updated probabilities (without dividing by the normalizing term, which is
independent on the probabilities updates):
P  Aupdated  P  A P D j A   
, P  B updated  P  B  P D j B  (27)
j j
Optimization: instead of computing the above values (which decrease continuously and
may become very small), it is better to compute:
ln P  Aupdated  ln P  A   ln P D j A   (28)
j
3) Classify:

y  sign ln P  Aupdated  ln P  B updated  (29)
All these algorithms are machine learning algorithms, needing a set of training data.
Training data consists of pairs of object features and associated outputs (classified labels). A
common parameter for the feature descriptors used by classification algorithms is the
histogram of oriented gradients (HOG) [44]. This is an interesting aspect, because several
other approaches for feature descriptors exist, such as SIFT and SURF. The major
characteristic of such approaches is the fact that the descriptor is computed only for key
(interest) points within the image. These are points which allow constructing descriptors
which are invariant to scale and rotation and robust to change of illumination [45].
Figure 4. Orientation assignment principle, as applied for SIFT method.
The main benefits of SIFT method, as indicated by the author [45] are: it is scale
invariant (in other words different sizes of the object do not make a difference) and rotation
invariant (perspectives from different angles do not make a difference). It is also quite robust
to affine distortion, noise and changes in illumination. The image descriptor is a so called
local descriptor, computed not for the entire image, but only for selected key points in the
image. The algorithm consists of two principal parts, key points identification and descriptor
computation. The key points identification process consists on several main steps: determine
the interest points (points in the image which are not affected by different image scales),
outliner rejection to obtain the key points and orientation assignment (selecting the dominant
orientation for each key point).
The descriptor computation involves computing the gradients in a 16x16 region around
each key point, dividing the region in 16 (4x4) blocks and for each compute a 8-bin
histogram. For this, the orientation of the gradients is expressed relative to the key point
orientation. The descriptor around each key point will be a vector with a 16x8=128 size.
The major stages in the algorithm for computing the SIFT descriptor are [45]:
1) Scale-space extrema detection - identifies locations and scales that can be repeatedly
assigned under differing views of the same object.
2) Key point localization - to reject points with low contrast or poorly localized.
3) Orientation assignment - to assign a consistent orientation to each key point based on

local image properties. The descriptor can be represented relative to this orientation and
achieve invariance to image rotation, see Figure 4, reproduced from [45].
4) Key point descriptor - to compute a descriptor for the local image region that is highly
distinctive yet is as invariant as possible to remaining variation.
Experimental results on SIFT descriptors applied to image classification shows that better
classification results are often obtained by computing the SIFT descriptor over dense grids in
the image domain opposite to sparse interest points (practically skipping the first stages in the
algorithm, which is the selection of the key points).A larger set of local image descriptors
computed over a dense grid usually provide more information than corresponding descriptors
evaluated at a much sparser set of image points. Based on these results [46, 47] proposed the
dense SIFT descriptors. DSIFT is usually accompanied by a clustering stage, where the
individual SIFT descriptors are reduced to a smaller vocabulary of visual words, which can
then be combined with a bag-of-words model or related methods [48, 49].
By contrast, HOG descriptors are computed over the entire image (on a so-called dense
grid) and have no orientation alignment. In general, for image classification problems HOG
descriptors are preferred, as their performance is higher. The main idea behind this method, as
revealed in [44], is that the look and shape of an object can be sufficiently characterized by
the local distribution of intensity gradients, even if gradients are calculated exactly. This
method is implemented by dividing the image into small windows called and cells. For each
cell it is computed one dimensional local gradient histogram. After that local histograms are
combined to create a global representation. To solve the problem of lighting and shadows
zonal contrast normalization is applied. This is done by accumulating value of a number of
such cells (a block) and normalizing all the cells in the block. These blocks typically overlap.
Two main types of geometries are:
1. R-HOG: blocks are square grids represented by three parameters: the number of cells
per block, the number of pixels per cell and the number of channels per cell histogram.
2. C-HOG: blocks are circular. It can be described by four parameters: number of angular
and radial bins, radius of the center bin and the expansion factor for the radius of additional
radial bins.
The main steps for obtaining a HOG descriptor for an image (region) are (typical values
are indicated):
1) Divide the image in blocks of size 16x16 pixels, blocks with 50% overlapping
2) Each block will consist on 2x2 cells (each of 8x8 pixels)
3) Compute gradients, on the x and y axis, for each cell
4) Compute gradients orientation, i.e. the angle between the resultant gradient and the x
axis
5) Construct the histograms (9-bin) by grouping the gradients within corresponding
bins, given their orientation (a gradient does not fall within a single bin, but is
proportionally distributed between adjacent bins, according to the distance between
its orientation and the closest centers of two adjacent bins)
6) Weight the gradients (with Gaussian function), in order to down-weight the cells near
the image edges
7) Construct the global image (region) descriptor, by concatenating all HOGs (for all
cells within the image)
For example, considering the case of an image region of 64x128 pixels (this is typically
for person representation), there will be 7x15=105 blocks and 105x4=420 cells, thus the
feature descriptor for the image will have 420x9=3780 elements.
In the case of images from a LIDAR, it is possible to construct similar descriptors.
However, in literature, the number of blocks per image used for obtaining the HOG is
different (e.g. 6 or 10).
5. PERCEPTION ALGORITHMS IN AGV-REVIEW OF

EXISTING SOLUTIONS
In this section we will review several implementations of perception algorithms for
different AGV prototypes [3, 4, 27, 50-62]. This review is based on description of algorithms
and their results, as presented in literature between 2008 and 2012. Consequently, it includes
both some of the prototypes participating in the DARPA Urban Challenge (2007) and results
of research activities carried on after that date. The latter are either full AGV prototypes, or
partial implementations, used for testing and validating new perception algorithms.
In Table 3 we summarize the type of detection, tracking and classification algorithms
used in several relevant studies, as well as the reported results (where available).
The performance of a classification algorithm is indicated by two main parameters:
precision and recall rate. Precision is defined as the ratio of correctly identified objects per
category (NCIO) versus the total number of identified objects in that category (NIO):
p  NCIO / NIO (30)
whereas the classification recall rate, is the ratio of correctly identified objects per category
versus the total number of objects in that category (NO):
r  NCIO / NO , with NO  NCIO  NMO (31)
where NMO is the number of missed objects (or “false negative”).

For example, if at a certain moment there are 10 different objects in the image (6 cars, 3
persons, 1 other) and the perception algorithm identifies 4 cars (3 correctly, 1 which actually
is a different object, wrongly identified or “false positive”), the precision will be: 3/4 and the
recall rate 3/6. Clearly, in most applications, both precision and recall are important, because
both false positive and false negative classifications can generate big issues. It is somehow
intuitive that a classifier with low misses (good recall) may have a higher probability of
generating false positives (lower precision). Therefore, when assessing the performance of a
classifier, a graphic showing the recall rate vs. the number of false positives (or maybe vs.
precision) may be shown. Of course, in practice the classifier is tuned according to the point
on the graphic corresponding to the best compromise between recall and precision, depending
on the application. Generally, the classifiers described in literature have a precision which is
higher than recall (number of false positives smaller than number of false negatives).
Because real-time implementation is a critical condition for perception (and control)
algorithms in AGV, we also included in results information on the computational time (where
available). Also, where available, we included the AGV operation speed, which depends on
several factors, including total computational time.
As it can be seen in the table, only few implementations perform detection, tracking and
classification of several classes of objects. In some cases only detection and either tracking or
classification is performed, whereas in others a simple classification is done, separating for
instance obstacles from road. Also, some implementations focused on achieving real-time
functionality of the vehicle (trade-off between precision and computational time), whereas
others proposed and tested new algorithms which have to be optimized (and possibly
simplified, affecting their performance) in order to become real-time. Despite of these
differences, it can be seen that Kalman filter (extended or linear) is a common choice for part
of the tracking algorithm (being present in all implementations performing tracking), whereas
SVM is similarly common for classification (in 3 out of 4 solutions performing classification
of several classes of objects). Out of these, 2 use HOG as main way of generating the feature
(descriptor) vectors.
Historically, most AGV prototypes used 3DLIDAR data for classification algorithms, but
more recent studies present also results achieved using images from cameras.
Talos [4], a prototype competing in the DARPA Urban Challenge Event (2007) had a
perception system consisting of the following blocks: lane detection (based on camera and
digital map), fast vehicle detector (based on radar), obstacle detector (based on LIDAR) and
hazard detector (based on LIDAR), the latter being used to detect bumps and holes in the
road, implemented on several CPUs [4]. The lane detection is achieved using matched filter to
determine the lane direction and random sample consensus algorithm (RANSAC) to select the
lane centerline. Obstacle detection and tracking combined data from 12 planar and 1
3DLIDARs. First, the ground plane is identified starting from points that are on the same
plane, which are then fit into a non-parametric ground model, based on maximum acceptable
road slopes, to exclude false candidates (e.g. car roofs). Based on knowing the ground plane,
obstacle points are detected considering areas above ground plane which generate a certain
minimum numbers of returns from lasers.
A clustering algorithm groups data from neighboring obstacle points, to generate an
obstacle. For tracking, obstacles from new data frames are associated with previous ones
based on a voting scheme, which considers the region of overlapping between the new and
old data frame obstacles. Speed is estimated using a Kalman filter, thus separating between
stationary, slow and fast moving obstacles. As it can be seen, this approach is based on
relatively simple, but robust algorithms which allows for the perception and control system to
be implemented in real-time (max. vehicle speed was 40 km/h).
In [27] it is described a more complex perception subsystem, focusing mainly on
classification of several types of objects. The algorithms were tested on Stanford Track
Collection database, and operate in real-time, considering just objects on the road and an 8-
core computer. Detection is done by segmentation, using the connected components
algorithm and clustering using flood fill algorithm. Tracking uses a Kalman filter and data
association is achieved based on distance. Classification is based on boosting method, applied
twice, for two types of classifiers: segment and holistic (track) classifiers.
Table 3. Summary of implementation details of perception algorithms for AGVs
Implementation Detection Tracking Classification Results

A. Teichman, Clustering (no Tracking of Boosting + Discrete Augmented DBF
J. Levinson, details) clusters Bayes Filter (DBF) best
S. Thrun, [27] centroids, using variants (augmented, 97.5% precision
2011 linear Kalman normalized, naïve) (overall)
filter (no Using 29 Real time (for
LIDAR sensors details) dimensional segment road objects only)
descriptors
M. Occupancy grid Kalman filter Support Vector Used in vehicle
Himmelsbach, Object hypothesis Machine (SVM) tracking
A. Muller, for segmentation Using 28 Max. error (for
T. Luttel, dimensional feature tracked vehicle):
H.J. Wunsche vector (incl. HOG) 1m
[50] Low speed,
2008 distance 15-20 m
from tracked
LIDAR sensors vehicle
J. Leonard, Clustering of Voting scheme Only for road and 96.7% success on
J. How, S. chunks by (based on obstacle tracking (all
Teller, et al. [4] proximity into overlapping identification existing objects)
2008 groups groups) Identify candidate Velocity
Velocity points (within a estimation:
estimation using plane) - good for objects
Kalman filter Fit non-parametric with high speed
road model to - noisy for
exclude false objects with low
LIDAR sensors candidates speed
Detect obstacles Real-time. Max
based on height from speed 40 km/h.
the ground plane.
J. Choi, J. Lee, Camera Used only for road/
D. Kim, data(dedicated obstacle
G. Soprani, system) for: Based on z vs. x,y
P. Cerri [51] - pedestrian coordinates variation
2012 crossing via IPM Fit model
transform and Risk integration
LIDAR sensors Sobel filter
- speed bumps via
color
N. Suganuma, Occupancy grid Extended - Robust to
T. Uozumi [3] maps for detection Kalman filter weather
2011 Clustering of (to estimate conditions
dynamic objects motion) Real time. AGV
Lane marking: speed= 27.5 km/h
dedicated system
LIDAR sensors (separate sensor)
Table 3. (Continued)
Implementation Detection Tracking Classification Results

D.Z. Wang, Graph-based SVM Classifier: Precision > 95%
I. Posner, clustering using separates foreground Non real time:
P. Newman [53] Euclidean data in 3 classes (car, clustering 3.3 s/
2011 Minimum pedestrian, bicyclist) frame
Spanning Tree Using 120
(EMST) dimensional feature
Separate vector
LIDAR sensors foreground (all
classes of interest)
from background
(others)
A. Ess, Together with Hypothesis-and- SVM using HOG as Recall rate: 60-
K. Schindler, classification verify approach input features; run 70%
B. Leibe, algorithm (output Dynamic independent Prediction
L. Van Gool of classifiers motion using classifiers for each precision: 70-
[52] indicating extended class (car, 80% (next 2
2010 probability of an Kalman filter pedestrian) frames),
image region to be Trajectory Ground plane and 60-70% (next 10
a pedestrian/ car) generation in object estimation frames) – next 1 s
parallel for all using Bayesian Non real time
Video Camera detected objects network (HOG
computation 6s/
frame)
The segment classifiers operate on segment images, i.e. images of a detected object at a
particular moment in time. The holistic (or track) classifiers operate with a set of images of a
particular object, tracked during a time interval, in other words a succession in time of
segments corresponding to the same object.
The boosting method used is a version of Gentle Boost and Joint Boost, creating both a
strong segment classifier, and a strong holistic (track) classifier respectively. Each weak
classifier (used to create the two strong classifiers) is based on a particular descriptor of a
segment or respectively a track. The descriptors used by track classifiers include maximum
and mean speed and acceleration of the object (computed for the entire track duration, relative
to the track centroid) while segment classifiers descriptors include HOG and spin images
representations. Spin images is a technique used particularly in the case of dense collections
of 3D points, such as those generated by 3DLIDARS, to represent and compare a particular
objects with a set of rotated images of the reference object [51]. Of course, for efficiency
purpose, compressed representations of spin images are used [51].
In the final stage, both strong classifiers are combined into a single, final one, using
different versions of discrete Bayes filters. In these versions, the log of odds for the segment
and the holistic classifier are taken into account. As the paper shows, the best results are
obtained for the augmented discrete Bayes filter (ADBF). The precision is better if
classification is done per track (not for each segment, i.e. not for the object at any given
moment, but for the tracked object). In this case, the overall precision considering all three
classes of objects (car, pedestrian, and bicyclist) is 98.5% for the two combined classifiers
and 97.6% if only the segment classifier is used. If on the other hand, the classification is
done for each segment, employing just the segment classifier gives a 93.0% precision. In
order to use the holistic classifier, holistic data have to be approximated based only the
information on that segment. In this case, there is still an important improvement, the overall
precision being 97.5%. [27] Consequently, an important conclusion is that holistic data can
improve the overall precision, even (especially!) for classification done for each segment.
In [52] a complex system, performing detection, classification as well as tracking is
described. The system uses video image processing, with data achieved from a forward-
looking, stereo camera pair. The main focus is object tracking and path prediction, based on
the output of pedestrian and car classifiers. The dynamic models used are constant-velocity
one for pedestrians and Ackermann model for car motion.
The object classification is performed using SVM and HOG descriptors. Both pedestrian
and cars categories are classified. In order to cope with the variance of a car image depending
on the viewing angle, seven separate classifiers are used, one for each relevant angle (e.g.
back, lateral, front-profile, back-profile – where last three are doubled, as per left/ right side).
The output of each classifier is the probability of a true/ false match, rather than a final
decision. There is no clear indication on the method used to convert SVM score in a posterior
probability. Using these probabilities, as well as the stereo depth, the authors use a Bayesian
framework to jointly estimate the objects as well as the ground plane (assuming all objects
reside on a common ground plane). This joint estimation of objects and ground plane has the
advantage, according to the authors, that improves the results. For example, in a scene
crowded with objects, ground plane is difficult to be estimated alone, whereas in scenes with
few objects, ground plane estimation helps in objects detection.
Tracking is done considering separate movement models for pedestrians and respectively
cars. The state vector for the pedestrian movement model at moment tis  xt yt t vt  ,
T
with x and y the coordinates,  the pedestrian orientation angle and v its speed, expressed as
a state element in order to estimate it. The movement model for cars is based on the
Ackermann model, thus the state vector is  xt yt t vt t at  , where the first four
T
components have the same meaning as in the pedestrian model, and  is the car steering
angle, whereas a is its acceleration. Both models are nonlinear and thus Kalman filter cannot
be directly used, Extended Kalman filter EKF being the algorithm used by the authors.
Classification recall rate is in the order of 60-70%, whereas prediction precision is around
70-80% for short prediction horizon (4 frames) and 60-70% for longer (10 frames). The
overall algorithm, implement on a graphical processor (GPU) is not real time, the bottleneck
being the HOG computation, but authors indicate that this time can be reduced, if parallel
implementation of HOG computation is used.
As it can be seen also from the data summarized in Table 3, typically the
implementations increased the complexity of the algorithms over the past years, with the goal
of classifying with good precision objects of at least several classes. The increased
complexity makes real-time implementation a challenge, even when using multiple
processors. In fact, in many of the reviewed papers, the research focused mostly on algorithm
precision improvement, while implementation and complexity optimization is left as a future
work.
6. EXPERIMENTAL RESULTS
For experimental results we created a database with cars viewed from different angles:
rear view (near and far), left and right view. Each of these views is considered a separate
class. The database was created staring from Vehicle Image Database [63], which integrates
images from Caltech Database [64, 65] and also from TU Graz-02 Database [66, 67]. The
database was completed with images from Traffic Signs Dataset [68]. These images are first
segmented in order to identify possible component objects. After segmentation, the
component objects are validated by the user and new images are created by cutting off each
object using a surrounding space. For each of the four classes we have 800 images, and each
image contains one car.
We evaluated image classification using DSIFT and HOG feature descriptors using the
basic measures for evaluating classification methods: precision, recall as well as the
confusion matrix for the test set. Precision and recall rates are computed based on the
definitions (30) and (31), in section 5. Since for AGV navigation it is essential not to miss
existing cars, we will mostly refer to this parameter.
The confusion matrix displays the number of correct and incorrect predictions made by
the model compared with the actual classifications in the test data. The matrix is n-by-n,
where n is the number of classes. It contains information about actual and predicted
classifications done by the classification system. Of course, ideally this matrix should have all
elements outside the main diagonal equal to zero.
Table 4. Optimal number of training images for SVM with DSIFT, for each class
Cars Number of training images

Left Cars 115 - 120
Right Cars 115 - 120
Middle Cars 90
Far Cars 95
Table 5. Optimal number of training images for SVM with HOG, for each class
Cars Number of training images

Left Cars 110
Right Cars 95
Middle Cars 90
Far Cars 95
The database is manually divided in four classes: Left Cars, Right Cars, Middle Cars and
Far Cars (we use the defined classes in the Vehicle Image Database [63]). In this case we
train a SVM for each class (Left, Right, Middle and Far Cars classes). Each SVM was trained
with different numbers of training images in order to obtain the best testing accuracy for each
class separately. The recall of each SVM based on the number of training images is presented
in graph from Figure5 and Figure 6, for DSIFT and HOG methods.
The numbers of training images which leads to best performance for each class - for
DSIFT and respectively HOG descriptors are shown in Tables 4 and 5. For the classification
rates, we use one-versus-all SVM in order to compute the best number of training images.
The overall recall rates are computed using three types of decisions: one-versus-all SVM,
pair-wise SVM with majority voting and pair-wise SVM with confidence score voting.
Figure 5. SVM recall rate for different number of training images for DSIFT method.
Figure 6. SVM recall rate for different number of training images for HOG method.
The recall results for both methods and all four classes are shown in Table 6. It can be
observed that both pair-wise SVM methods outperform the one-versus all SVM method. The
confidence score voting method performs better than the majority voting method. Based on
the overall recall, the class Far class has the lowest performance, for all three types of SVMs.
This is of course understandable and is not essential, since it does not impact immediate
decisions of the control algorithm.
Table 6. Summary of image classification recall tests using DSIFT and respectively
HOG
Recall
One versus all SVM Pair-wise SVM Pair-wise SVM
Cars with majority voting with confidence score
voting
DSIFT HOG DSIFT HOG DSIFT HOG
Far 69% 71% 72% 74% 73% 75%
Left 83% 85% 84% 86% 86% 89%
Right 81% 84% 82% 85% 85% 88%
Middle 75% 79% 78% 81% 80% 73%
Figure 7. Precision/Recall curve for tests on entire image database using DSIFT feature descriptor.
The confusion matrix on the set test of four classes using DSIFT with pair-wise SVM
with confidence score voting is presented in Tables 7 and 8, for DSIFT and HOG methods.
The tables contain the classification accuracy for the individual categories, too. The most
difficult categories are Far Cars and Middle Cars for DSIFT method, respectively Far Cars
for HOG method.
Precision and recall curves are given in Figure 7 and Figure 8 for DSIFT and respectively
HOG method, for each class from image database, using one-versus-all SVM.
Figure 9 shows precision and recall curve for Left Cars class (the class with the highest
overall accuracy) using both methods: DSIFT and HOG. It can be observed that HOG feature
description method outperforms DSIFT method in terms of precision and recall curve.
Figure 8. Precision/Recall curve for tests on entire image database using HOG feature descriptor.
We can see that HOG method performs generally better, on all considered image classes.
Figure 10 shows examples of cars correctly categorized by the SVM with HOG
descriptor, for each of the four classes. As it can be seen, for Far Cars, not only the cars are
smaller, but the quality of the image is also poorer, which also contributes to the lower
performance achieved for this class.
Table 7. Confusion matrix for DSIFT feature descriptor
Predicted values
Far Cars Left Cars Right Cars Middle Cars
Far Cars 73 4 3 11
Left Cars 9 86 9 3
Right Cars 8 8 85 6
Middle Cars 10 2 3 80
Accuracy 73% 86% 85% 80%
Table 8. Confusion matrix for HOG feature descriptor
Predicted values
Far Cars Left Cars Right Cars Middle Cars
Far Cars 75 2 4 9
Left Cars 10 89 4 3
Right Cars 7 6 88 5
Middle Cars 8 3 4 83
Accuracy 75% 89% 88% 83%
Figure 9. Precision/Recall curve for tests on Left Cars using DSIFT and respectively HOG feature
descriptor.
Far Cars
Left Cars
Right Cars
Middle
Cars
Figure10. Example of correct images classified with HOG feature and pair-wise SVM with confidence
score voting.
Figure 11. Example of correctly detected cars from images using HOG method for classification (black
contour for each correctly detected car).
Figure 12. Example of wrongly detected cars from images (white contour for falsely reported cars).
After classification, car detection is performed on test images, in order to separate the
object in that class. A detection result is considered true or false, according to the overlap
ratio between the predicted bounding box Bp and the ground truth (the actual) bounding box
Bgt.

area B p  Bgt   50%
area  B p  Bgt 
Figure 11 shows a set of correct car detection based on the HOG method. For a true
detection we used the above equation. We obtained an overall detection rate of over 90%.
Also, in Figure 12 we present a couple of examples of wrongly detected cars. As it can be
seen, objects which appear small in the image, or portions of other objects were wrongly
classified as cars.
CONCLUSION
The review of existing sensing technologies and perception methods used in AGVs
revealed some important trends, including the more emphasis given recently to video camera
for vehicle detection and classification. Therefore, we investigated in more detail the object
classification based on camera images. For this, we implemented an SVM classifier and we
compared two types of feature detectors: DSIFT and HOG. We used four distinct classes of
cars, depending on their orientation and relative distance to the camera and present the results
for the classification in each of these classes. Obtaining the information on detected cars
orientation and distance is helpful for the AGV control. Of course, the classification
information can also be aggregated, in a car vs. non-car decision. In general, the HOG feature
descriptors lead to a higher performance than DSIFT. The achieved results confirm the
expectation that video cameras can play a much higher role in perception sub-systems for
future AGVs.
ACKNOWLEDGMENTS
In this chapter we used several existing image databases: mainly the Vehicle Image
Database [63], which integrates images from Caltech Database [64, 65] and also from TU
Graz-02 Database [66, 67]. The database was completed with images from Traffic Signs
Dataset [68].
REFERENCES
[1] Folsom, T.C.; Energy and Autonomous Urban Land Vehicles. IEEE Technology and
Society Magazine, pp. 28-38, 2012.
[2] Luettel, T.; Himmelshach, M.; Wuensche, H.J.; Autonomous Ground Vehicles -
Concepts and a Path to the Future. Proceedings of the IEEE, Centennial Issue, 2012.
[3] Suganuma, N.; Uozumi, T.; Development of an Autonomous Vehicle -System
Overview of Test Ride Vehicle in the Tokyo Motor Show 2011. 2012 SICE Annual
Conference, Japan, pp. 215-218, 2012.
[4] Leonard, J.; How, J.; Teller, S.; et al.; A Perception-Driven Autonomous Urban
Vehicle. Journal of Field Robotics, Special Issue: Special Issue on the 2007 DARPA
Urban Challenge, Part III, vol. 25, no. 10,pp. 727-774, 2008.
[5] Bigman, D.; Driverless Cars Coming to Showrooms by 2020, Says Nissan CEO Carlos
Ghosn. Forbes,www.forbes.com/sites/danbigman/2013/01/14/driverless-cars-coming-
to-showrooms-by-2020-says-nissan-ceo-carlos-ghosn/.
[6] KPMG, LLP; Self-Driving Cars: Are We Ready. Report, www.kpmg.com/US/en/Issues
AndInsights/ArticlesPublications/Documents/self-driving-cars-are-we-ready.pdf.
[7] Beiker, S.A.; Legal Aspects of Autonomous Driving. Santa Clara Law Review 1145,
vol. 52, no. 4,pp. 1145-1156, 2012.
[8] KPMG,LLP; CAR (Center for Automotive Research); Self-Driving Cars: The Next
Revolution.Report,www.kpmg.com/US/en/IssuesAndInsights/ArticlesPublications/Doc
uments/self-driving-cars-next-revolution.pdf.
[9] Dominguez, R.; Onieva, E.; et al.; LIDAR Based Perception Solution for Autonomous
Vehicles.11th International Conference on Intelligent Systems Design and
Applications, Spain, pp. 790-795, 2011.
[10] Fernandez, C.; Gavilan, M.; et al.; Free Space and Speed Humps Detection Using
LIDAR and Vision for Urban Autonomous Navigation. Intelligent Vehicles
Symposium, Spain, pp. 698-703, 2012.
[11] Ozguner, U.; Stiller, C.; Redmill, K.; Systems for Safety and Autonomous Behavior in
Cars: the DARPA Grand Challenge Experience. Proceedings of IEEE, vol. 95, no. 2,
pp. 397-412, 2007.
[12] Rouff, C.; Hinchey, M.; Experience from the DARPA urban challenge. Springer-Verlag
London Ltd., 2012.
[13] Cremean, L.B.; Foote, T.B.; et al.; Alice: An Information-Rich Autonomous Vehicle
for High-Speed Desert Navigation. Journal of Field Robotics, Special Issue: Special
Issue on the DARPA Grand Challenge, Part 2, vol. 23, no. 9, pp. 777-810, 2006.
[14] Guizzo, E.; Deyle, T.; Robotics Trends for 2012. IEEE Robotics and Automation
Magazine, pp. 119-123, 2012.
[15] Ilas, C.; Electronic Sensing Technologies for Autonomous Ground Vehicles: A
Review.8th International Symposium on Advanced Topics in Electrical Engineering
(ATEE 2013), Romania, pp. 1-6, 2013.
[16] Jenkins, A.; Remote Sensing Technology for Automotive Safety. Microwave Journal,
Dec. 2007, www.microwavejournal.com/ articles/5689-remote-sensing-technology-for-
automotive-safety.
[17] Conner, M.; Automobile Sensors May Usher in Self-Driving Cars.EDN Magazine, May
26, 2011. www.edn.com/design /automotive/4368069 /Automobile-sensors-may-usher-
in-self-driving-cars.
[18] Continental, A.G.; SRL 1 and SRL1C Infrared Short Range LIDAR Sensor Datasheet.
www.conti-online.com/generator/www/
de/en/continental/industrial_sensors/themes/download/srl1_datasheet_en.pdf.
[19] Continental, A.G.; SRL-CAM400CMOSCamera and Infrared LIDAR in a Compact
Unit. PressRelease, www.conti-online.com/generator/www/com/en/continental/press
portal/themes/press_releases/3_automotive_group/chassis_safety/press_releases/pr_201
2_10_17_srl_cam_en.html.
[20] Stevenson, R.; Long-Distance Car Radar. IEEE Spectrum, Oct. 2011,
http://spectrum.ieee.org/green-tech/advanced-cars/longdistance-car-radar.
[21] Fleming, B.; Recent Advancement in Automotive Radar Systems. IEEE Vehicular
Technology Magazine, vol. 7, no. 1, pp. 4-9, 2012.
[22] Bosch; LRR3- Third Generation Long-Range Radar Sensor Datasheet 2009.
www.bosch-automotivetechnology.cz/media/db_application/downloads/pdf/safety_
1/en_4/lrr3_datenblatt_de_2009.pdf.
[23] Bosch; Combined Inertial Sensor for Vehicle Dynamics Control SMI650 Datasheet.
www.bosch-semiconductors.de/media/pdf_1/einzeldownloads/vehicledynamics_
systems/datenblatt_smi650.pdf.
[24] Delphi Inc.; Integrated Radar and Camera System. RACam Product Info,
http://delphi.com/shared/pdf/ppd/safesec/racam.pdf.
[25] Euro NCAP; Autonomous Emergency Braking -AEB information.
www.euroncap.com/Content-Web-Page/2788ff8c-788d-468e-996b-
7405569412b1/autonomous-emergency-braking---aeb.aspx.
[26] Ilas, C.; Perception in Autonomous Ground Vehicles - A Review. Electronics,
Computers and Artificial Intelligence (ECAI) International Conference, Pitesti,
Romania, pp. 1-6, 2013.
[27] Teichman, A.; Levinson, J.; Thrun, S.; Towards 3DObject Recognition via
Classification of Arbitrary Object Tracks. IEEE International Conference on Robotics
and Automation (ICRA 2011), pp. 4034-4041, 2011.
[28] Yilmaz, A.; Javed, O.; Shah, M., Object Tracking: A Survey. ACM Computing Surveys,
Article 13, vol. 38, no. 4, 2006.
[29] Moegelmose, A.; Manubhai Trivedi, M.; Moeslund, T.B.; Vision-Based Traffic Sign
Detection and Analysis for Intelligent Driver Assistance Systems: Perspectives and
Survey. IEEE Transactions on Intelligent Transportation Systems, vol. 13, no. 4,pp.
1484-1497, 2012.
[30] Ruta, A.; Yongmin, L.; Xiaohui, L.; Real-Time Traffic Sign Recognition from Video
by Class Specific Discriminative Features. Journal of Pattern Recognition, vol. 43, pp.
416-430, 2010.
[31] Burges,C.J.C.; A Tutorial on Support Vector Machines for Pattern Recognition. Data
Mining and Knowledge Discovery, vol. 2, issue 2, pp. 121-167, 1998.
[32] Weston, J.; Support Vector Machines (and Statistical Learning Theory) Tutorial.
www.cs.columbia.edu/~kathy/cs4701/documents/jason_svm_tutorial.pdf‎
[33] Forsyth, D.A.; Ponce, J.; Computer Vision: A Modern Approach. Prentice Hall, 2011.
[34] Zadrozny, B.; Elkan, C.; Transforming Classifier Scores into Accurate Multiclass
Probability Estimates. 8th ACM SIGKDD International Conference on Knowledge
Discovery and Data Mining (ACM 2012),pp. 694-699, 2002.
[35] Abe, S.; Support Vector Machines for Pattern Classification. Springer, New York,
2005.
[36] Schlkopf, B.; Smola, A.J.; Learning with Kernels: Support Vector Machines,
Regularization, Optimization and Beyond. 1stEd., Cambridge, The MIT Press, MA,
2001.
[37] Vapnik,V.N.; The Nature of Statistical Learning Theory. 2nd Ed., Springer, New York,
2000.
[38] Hsu, C.W., Lin, C.J., A Comparison of Methods for Multiclass Support Vector
Machines. IEEE Transactions on Neural Networks, vol. 13, no. 2, pp. 415-425, 2002.
[39] Wang, L.; Fu, X.; Data Mining with Computational Intelligence. Springer, Berlin, 2005.
[40] Friedman, J.H.; Another Approach to Polychotomous Classification. 1996,
http://old.cba.ua.edu/~mhardin/poly.pdf.
[41] Phung, W.; Naghdy, G.; A Multi-Class Image Classification System Using Salient
Features and Support Vector Machines. International Conference on Sensors, Sensor
Networks and Information Processing (ISSNIP 2007), Melbourne, Australia, pp. 431-
436, 2007.
[42] Rojas, R.; AdaBoost and the Super Bowl of Classiers - A Tutorial Introduction to
Adaptive Boosting. Freie University of Berlin, 2009, www.inf.fu-berlin.de/inst/ag-
ki/adaboost4.pdf.
[43] Downey, A.B.; Think Bayes - Bayesian Statistics Made Simple. E-book, Green Tea
Press, 2012, www.greenteapress.com/thinkbayes/.
[44] Dalal, N.; Triggs, B.; Histograms of Oriented Gradients for Human Detection.
International Conference on Computer Vision & Pattern Recognition (CVPR'05).vol.
1,pp. 886-893, 2005.
[45] Lowe, D.G.; Distinctive Image Features from Scale-Invariant Keypoints. International
Journal of Computer Vision, vol. 60, no. 2,pp. 91-110, 2004.
[46] Bosch, A.; Zisserman, A.; Munoz, X.; Image Classification Using Random Forests and
Ferns. 11th International Conference on Computer Vision (ICCV'07), Rio de Janeiro,
Brazil, pp. 1-8, 2007.
[47] Bosch, A.; Zisserman, A.; Munoz, X.; Scene Classification via pLSA. 9th European
Conference on Computer Vision (ECCV'06), pp. 517- 530, 2006.
[48] Csurka, G.; Dance, C.; Fan, L.; Willamowski, J.; Bray, C.; Visual Categorization with
Bags of Keypoints. International Workshop on Statistical Learning in Computer Vision
(ECCV'04), Prague, Czech, pp. 1-22, 2004.
[49] Lazebnik, S.; Schmid, C.; Ponce, J.; Beyond Bags of Features: Spatial Pyramid
Matching for Recognizing Natural Scene Categories. IEEE Conference on Computer
Vision & Pattern Recognition (CVPR'06), New York, USA,pp. 2169-2178, 2006.
[50] Himmelsbach, M.; Muller, A.; Luttel, T.; Wunsche, H.J.; LIDAR-Based 3D Object
Perception. 1st International Workshop on Cognition for Technical Systems, 2008.
[51] Choi, J.; Lee, J.; Kim, D.; Soprani, G.; Cerri, P.; Environment Detection and Mapping
Algorithm for Autonomous Driving in Rural or Off-Road Environment. IEEE
Transactions on Intelligent Transportation Systems, vol. 13, no. 2, pp. 974-982, 2012.
[52] Ess, A.; Schindler, K.; Leibe, B.; Van Gool, L.; Object Detection and Tracking for
Autonomous Navigation in Dynamic Environments. The International Journal of
Robotics Research, vol. 29, no. 14,pp. 1707-1725, 2010.
[53] Wang,D.Z.; Posner, I.; Newman, P.; What Could Move? Finding Cars, Pedestrians and
Bicyclists in 3DLaser Data. IEEE International Conference on Robotics and
Automation (ICRA'2012),pp. 4038-4044, 2012.
[54] Chakraborty, N.; Patel, R.S.; Intelligent Agents and Autonomous Cars: A Case Study.
International Journal of Engineering Research & Technology (IJERT), vol. 2, no. 1,pp.
1-7, 2013.
[55] Esteves, B.A.; Megda, P.T.; Becker, M.; Autonomous Embedded Navigation System-
Detection, Classification and Trajectory Prediction of Obstacles in Urban Environments
Based on Laser Sensor Data and KalmanFilter. ABCM Symposium Series in
Mechatronics, vol. 5, pp. 1032-1041, 2012.
[56] Chen, L.; Li, Q.; Li, M.; Zhang, L.; Mao, Q.; Design of a Multi-Sensor Cooperation
Travel Environment Perception System for Autonomous Vehicle. Open Access Sensors,
vol. 12, pp. 12386-12404, 2012.
[57] Dominguez, R.; Onieva, E.; et al.; LIDAR Based Perception Solution for Autonomous
Vehicles. 11th International Conference on Intelligent Systems Design and
Applications, Spain, pp. 790-795, 2011.
[58] Fernandez, C.; Gavilan, M; et al.; Free Space and Speed Humps Detection Using
LIDAR and Vision for Urban Autonomous Navigation. Intelligent Vehicles Symposium,
Spain, pp. 698-703, 2012.
[59] Urmson, C.; Baker, C.; et al.; Autonomous Driving in Traffic: Boss and the Urban
Challenge.Artificial Intelligence (AI) Magazine, vol. 30, no. 2, pp. 17-29, 2012.
[60] Luber, M.; Arras,K.O.; Plagemann,C.; Burgard, W.; Classifying Dynamic Objects: An
Unsupervised Learning Approach. Autonomous Robots, vol. 26. No. 2-3, pp. 141-151,
2009.
[61] Spinello, L.; Triebel, R.; Siegwar, R.; Multiclass Multimodal Detection and Tracking in
Urban Environments. The International Journal of Robotics Research, vol. 29, no.
12,pp. 1498-1515, 2010.
[62] Johnson, A.; Hebert, M.; Using Spin Images for Efficient Object Recognition in
Cluttered 3DScenes. IEEE Transactions on Pattern Analysis and Machine Intelligence,
vol. 21, no. 5, pp. 433-449, 1999.
[63] Arrospide, J.; Salgado, L.; Nieto, M.; Video Analysis Based Vehicle Detection and
Tracking Using an MCMCSampling Framework. EURASIP Journal on Advances in
Signal Processing, Article ID 2012, vol. 2, 2012.
[64] The Caltech Database. Computational Vision at California Institute of Technology,
Pasadena, www.vision.caltech.edu/html-files/archive.html.
[65] Fergus, R; Perona, P.; Zisserman, A; Object Class Recognition by Unsupervised Scale-
Invariant Learning. IEEE Computer Society Conference on Computer Vision and
Pattern Recognition, Madison, Wisconsin, USA,pp. 264-271, 2003.
[66] The TU Graz-02 Database. Graz University of Technology.www.emt.tugraz.at
/~pinz/data/GRAZ_02.
[67] Opelt, A.; Pinz, A.; Object Localization with Boosting and Weak Supervision for
Generic Object Recognition. Image Analysis Lecture Notes in Computer Science, vol.
3540, pp. 862-871, 2005.
[68] Larsson, F.; Felsberg, M.; Using Fourier Descriptors and Spatial Models for Traffic
Sign Recognition. 17th Scandinavian Conference on Image Analysis (SCIA'2011),
LNCS 6688, pp. 238-249,www.cvl.isy.liu.se/research/traffic-signs-dataset.
Chapter 5
INTELLIGENT HIGHWAY SURVEILLANCE AND

SAFETY SYSTEMS
Valentin Alexandru Stan, Radu Serban Timnea

and Razvan Andrei Gheorghiu†
“Politehnica” University of Bucharest, Bucharest, Romania
ABSTRACT
This chapter presents the concept and the implementation solution for the integrated
highway traffic surveillance system with complex safety integrated systems (speed radar
detectors, vehicle profile automatic identifier, plate recognition system and weight in
motion measurement). This concept has already been implemented for one of the busiest
routes in Europe and the results are very good, increasing safety and security on highway.
The continuously increase of traffic value on roads and highways requires an
improvement of public and vehicles safety. That is possible by increasing the Police force
and Highway administrator system, the best solution being to implement electronic
systems for traffic surveillance, speed radar, incident detector and informing tables on the
highway.
Modern highway surveillance is not just a local circuit television system, in fact it
realizes all the functions which would be done by field personnel (generally police
officers): traffic surveillance, traffic re-routing, emergency services announcement, driver
informing on traffic conditions or restrictions. Moreover, the power computing system
can automatically optimize traffic based on real field conditions and generate alarms in
case of emergency.
An Intelligent Highway Surveillance System combines Control Centre,
Communication and Operating technology to provide the customer with a traffic
management, control, and resolution plan. This concept provides real time active traffic
management according to the main elements in the field and improves control, flow and
safe on the route.

Email: valentin.stan@smart-tc.ro.

Email: radu.timnea@upb.ro.
†
Email: andrei.gheorghiu@upb.ro.
148 Valentin Alexandru Stan, Radu Serban Timnea and Razvan Andrei Gheorghiu
In order to increase traffic safety, there was an absolute need to develop and
implement a system which would ensure the follow facilities:
 Traffic video surveillance which brings real-time images from the field to the
operators and also records those images as evidence of road traffic violations;
 Automatic incident detection which generates alarms in the dispatching center
(and, as an option, to the mobile teams) in case of major incident on the road
(accidents, abnormal traffic, fire, etc.).
 Automatic adaptation of traffic signalization in according with traffic values on
areas of highway, so as to optimize cars fluency in maximum safety conditions.
 Traffic statistics resulted by automatically vehicle count, estimation of vehicle
weight and average speed on segments of road, generally dangerous areas,
accidents, weather conditions and other.
 Speed detection (speed radar) so as to have permanent measurements of vehicle
speed in fixed points or mobile equipment.
 Automatic car identification in fixed points, using high definition cameras and
plate recognition software will be able to generate alarms in case any vehicles
are identified which are recorded in the data-base of the “wanted” (e.g. stolen
cars, cars with no technical or expired inspection clearances or others).
 Law enforcement system with alarm and recording facility (driving over speed
limits, ignoring traffic restrictions, ignoring signs etc.) is a useful instrument for
the Police force.
The technical solution adopted implements a high speed data network, along the
route and connects the technical infrastructure (video surveillance cameras, speed sensors
and others) to this network. Thus, this infrastructure can be unlimitedly extended, with
new cameras or other sensors.
Keywords: Video Surveillance, Highway Surveillance, Road Surveillance, Road Network,

Fibre Optics Network, Logistics, Highway Transport, Highway Monitoring Applications,
Information and Communication Technologies (ICT), Informing Service, Monitoring
Service, Dispatch, Euro-Corridors, Video-Camera, Speed Radar, Weather Sensors,
Weather Monitoring for Highway System, Traffic Video Surveillance, Automatic
Incident Detection, Traffic Signaling, Traffic Adaptation, Speed Detection, Optical
Character Recognizer, Plate Recognizer, Car Identifier, Doom Camera, Long Way
Camera, Radar, Speed Detection, CCTV, Wall Screen, Loop Radar, Traffic Info,
Emergency Detection, Emergency Team
1. INTRODUCTION
Generally, a system can be defined as a set of components or elements that are inter-
correlated, in order to achieve a certain objective. The system‟s components can be objects,
concepts, processes or even people. Any system has certain terminators that establish the
correlation between system components and environment. The environment is composed
from all the external elements and these elements cannot be controlled. The system‟s
components are connected through interfaces, which control the components interactions. An
interurban traffic management system is composed from infrastructure elements,
interconnected in order to achieve system‟s objectives. These components are physical
Intelligent Highway Surveillance and Safety Systems 149
components (cameras, variable message signs, traffic lights, etc.), communication equipment,
a traffic management centre, software and centre operators, the politics and procedures
established for events analysis etc. These components can be grouped in subsystems (i.e.
traveller information subsystem, ramp metering subsystem) that interact between each other
and with the environment.
The system‟s objectives are related with the system operating zone, in order to satisfy the
user needs and to be adapted to the traffic conditions and policies. However, there are some
objectives that are common for all these systems, such as:
 Reducing recurrent congestion and their impact on traffic

 Reducing travel time and of other non-recurrent congestions effects
 Increasing travel safety and efficiency on the interurban road network
 Traveller information, in order to provide a more efficient use of the offered facilities
 Assuring an fast and efficient incident response
 Reducing incident impact on traffic
 Decreasing fuel consumption
 Decreasing air pollution, etc.
According to these objectives, it can be inferred the main functions of a dynamic

interurban traffic management system:
 Traffic surveillance – the main function of any traffic management system; assures
traffic surveillance of the network covered by the system; along with traffic
monitoring, it also includes other functions: emergency vehicle localization, public
transport vehicle localization, meteorological condition monitoring, pavement
monitoring etc.
 Prioritization for public transport or emergency vehicles
 Incident management – fast incident detection, necessary resources assessment, fast
intervention, reduced traffic impact, incident information for other drivers etc.
 Ramp metering – controlled highway access, in order to reduce traffic congestion on
highways entries or to offer alternative routes, in case of highway capacity overflow.
 Driver information – before or during the travel; the data obtained from the traffic
surveillance function are centralized and transmitted to the driver, in order to permit
an informed decision; transmitted data are: estimated travel time for a certain route,
meteorological condition, pavement condition, incident location, congestions, special
events and other information regarding some abnormal traffic conditions.
The main advantage of a dynamic traffic system is its capacity to adapt to real traffic
conditions, obtained from the real time traffic monitoring, and to inform the drivers regarding
these real traffic data. Knowing this information, the drivers can take better decisions for their
routes or itineraries, they can make a better estimation for the travel time and they can travel
in safer conditions, by avoiding congestion, adjusting travel speed according to
meteorological and pavement condition etc.
2. MODERN CONCEPT ABOUT ACTIVE SAFETY ON HIGHWAYS

Modern highway surveillance is not just a local circuit television system, in fact it
realizes all the functions which would be done by field personnel (generally police officers):
traffic surveillance, traffic re-routing, emergency services announcement, driver informing on
traffic conditions or restrictions. Moreover, the power computing system can automatically
optimize traffic based on real field conditions and generate alarms in case of emergency [1].
An Intelligent Highway Surveillance System combines Control Centre, Communication
and Operating technology to provide the customer with a traffic management, control, and
resolution plan. This concept provides real time active traffic management according to the
main elements in the field and improves control, flow and safe on the route.
In order to increase traffic safety, there was an absolute need to develop and implement a
system which would ensure the follow facilities:
 Traffic video surveillance which brings real-time images from the field to the
operators and also records those images as evidence of road traffic violations;
 Automatic incident detection which generates alarms in the dispatching centre (and,
as an option, to the mobile teams) in case of major incident on the road (accidents,
abnormal traffic, fire, etc.).
 Automatic adaptation of traffic signalization in according with traffic values on areas
of highway, so as to optimize cars fluency in maximum safety conditions.
 Traffic statistics resulted by automatically vehicle count, estimation of vehicle
weight and average speed on segments of road, generally dangerous areas, accidents,
weather conditions and other.
 Speed detection (speed radar) so as to have permanent measurements of vehicle
speed in fixed points or mobile equipment.
 Automatic car identification in fixed points, using high definition cameras and plate
recognition software will be able to generate alarms in case any vehicles are
identified which are recorded in the data-base of the “wanted” (e.g. stolen cars, cars
with no technical or expired inspection clearances or others).
 Automatic law breach detection with alarm and recording facility (driving over speed
limits, ignoring traffic restrictions, ignoring signs etc.) is a useful instrument for the
Police force. By an efficient identification of drivers who do not observe traffic laws,
the safety and normal climate on the road will be significantly improved.
3. INTELLIGENT SURVEILLANCE AND SAFETY SYSTEMS FOR ITS

As explained before, it is obvious that active safety on highways can be obtained only by
integrating more systems and their components, by means of unitary intelligent system
architecture.
A dynamic interurban traffic management system implies to implement policies,
strategies and technologies that will lead to an increased level of performance for the road
network. A special attention must be granted to highways traffic. The highways represent
those road segments that permit travelling at great speeds and with increased safety
conditions, normally with a controlled access, in order not to overpass the transport capacity
and to reduce the risk of congestions or traffic jams. In this chapter, we will report mainly to
highways and we will treat the secondary roads as alternative routes, for traffic deviation, if
case, and as origins and destinations for highway traffic.
A special consideration will be held to traffic surveillance function, as the main
component of the integrated system, and by emphasizing its connections with incident
management function and with emergency situation function [2].
The traffic management matter is a very complex one. It cannot be solved only locally, or
by taking into account only one segment of an interurban road network. In order to obtain an
efficient traffic management, we have to take into consideration also the existing connections
between different components of the network, the adjacent roads and their connection
(entrance and exit) with the highway, the roads that connect different highways and the
connection points between interurban and urban networks. Consequently, an efficient
interurban traffic management system must take into account the whole transport corridor [3].
Considering that the main entities of such a logical system are: the highway traffic
management system, the secondary roads traffic management system and the travellers, the
connection and the data exchange between these subsystems is depicted in Figure 1.
TRAVELLER
PC, laptop, mobile

phone, PDA etc.
Data Data
dissemination dissemination
systems for systems for
secondary road Traffic Traffic highways
management management
system for system for
secondary roads Mutual data highways
dissemination
systems
Highway Data
Data Secondary Mutual data traffic collection
collection roads traffic collection control systems for
systems for control system center highways
secondary center
roads
Management system Management system

for secondary roads for highways
Figure 1. Logical components interface for a transport corridor management system.

Traffic surveillance on
Ramp metering for
highways and
highways entries
secondary roads
Incident Integrated Variable speed

management management
Inter-urban Traffic
Management System
Traffic data and route

guidance for travelers
Traffic lanes
Emergency vehicles management
management
Figure 2. Control techniques for an integrated interurban traffic management system
Figure 2 contain a schematic representation of some of the main interurban traffic

management techniques, which can be implemented in an integrated system:
 Traffic surveillance ant control

 Ramp metering (access control)
 Variable speed limit management
 Lane management
 Incident management
 Emergency intervention management
 Traveller information and route guidance
4. SENSORS AND DETECTING SYSTEMS

In a traffic management system, the detection and surveillance component represent the
part of the system that collect data which describe or characterize the highway traffic
conditions. These data are used to offer traffic condition information for other components of
the system or for other similar systems. The control and surveillance system assure the
necessary data for other traffic management functions, such as: traffic flow measurement,
environmental condition analysis, traffic management decisions, information dissemination
for travellers, system performance monitoring and evaluation, incident detection and
verification, emergency intervention management, ramp control management.
The collected data and their accuracy are reliant on the specific application that is used.
The detection and surveillance component is not limited at obtaining data regarding traffic,
but it can also offer information regarding the meteorological conditions, road pavement
condition, visibility, vehicle position data (for emergency vehicle, for example) and others
[4].
The main data assured for interurban traffic management are:
 Traffic volume or traffic flow

 Travel speed
 Occupancy
 Emergency vehicle location or public transport vehicle location [5]
 Queue presence and queue length
 Road pavement condition (ice, snow, wet etc.)
 Meteorological conditions (rain, snowfall, blizzard, humidity, fog etc.)
4.1. Detection Types
The detectors and the detection methods can be classified by many parameters (Figure 3).
For example, some detector types register mainly the moving vehicle, other the stopped
vehicles, and other the vehicles that are moving with at least 3 – 5 km/h.
The magnetic or normal loop detectors can operate in pulse mode or presence mode. The
loop produce a short exist pulse, when detects a vehicle passing over it, but no longer then the
vehicle presence in the detection area, due to the fact that vehicle delay will produce a longer
signal, directly proportional with the vehicle standing in the detection area.
An extended call detector had an extended exit signal, this meaning the holding or the
extension of the call generated by a vehicle for a period of time, previously defined by means
of an integrated timer.
A delayed call detector doesn‟t emit an exit signal when the detection area is occupied,
for a period of time set by means of an incorporated timer.
Figure 3. Detection types.

Actually, hybrid detectors are frequently used for extended and/or delayed call.
It can be also mentioned the detectors that extend the capabilities of regular detectors or
controllers, through hardware configuration that contain timers and that display controlling
circuits.
The software used allow activating or de-activating of some detectors, control the green
time and also activate the extension of the signal for a phase, in order to supplement the
controller program.
An example is the “green colour extension system”, in order to protect the vehicle
situated in a dilemma zone, in a semi-actuated signalling junction. Another type of auxiliary
logic detection is the „speed analysis system”. This system is a hardware component formed
by two loops detectors and an auxiliary logic. The two loops are installed on the same lane, at
a measured distance. Vehicles that pass over the loops induce two commands. In order to
establish if the vehicle speed is higher or lower than a reference speed, it is measured the time
between the two commands [5].
4.2. Functional Description of Detectors
Most detectors are used to identify the movement of the vehicles that pass over a certain
area of the road. This information is transmitted to a traffic controller, to a traffic counter or
to other devices.
If a detector is equipped with elements that contain directional characteristic, it will
register only the vehicles that are passing on a certain direction.
A movement detector will detect a vehicle that pass over it only if its speed is higher than
3-5 km/h. These detectors are used for actuated or semi-actuated controllers. For location
with very low speed movements or where parking is permitted, and where directional
characteristics are not necessary, there are normally used presence detectors.
When the vehicles are forced to stop or to travel with very low speed, in the proximity of
a junction with actuated traffic signals, it is necessary to detect presence, in order to assure the
detection of the vehicles stopped in the detection area.
The presence detectors that don‟t have a directional recognition characteristic cannot
make the difference between the vehicles that enter or exit the detection area. However, the
direction it is not important, because the controller normally doesn‟t memorize the call, once
the vehicle exit the detection area. The call will not be memorised by the passing detector
either, for the vehicle that exit the detection area. The difference is the operation mode of the
controller: locked or unlocked.
For traffic measurements are usually used passing detectors. This type of detector can be
used to count the vehicles for one lane, for more than one lane of a specific direction or for all
lanes, both directions. If the lane number that register the number vehicles with one detector
increase, the registration accuracy decrease, when the detector is simultaneously occupied by
more vehicles.
Passing detectors are also used to measure vehicles flow speed and traffic volumes. The
data can be measured for each lane or for all the lanes of one direction.
In order to obtain correct data for vehicles speed and traffic volumes, the detectors must
be placed at a convenient distance from the junction, so the queues don‟t occupy the detection
areas. If this situation occur, only traffic volume can be registered.
Table 1. The Advantages and disadvantages of vehicle detection systems
Technology Advantages Disadvantages

Flexible design, which allow implementation of many application Need of pavement works for installing
Mature and well known technology Incorrect installing reduce pavement reliability
Can provide the basic traffic parameters (volume, presence, size, speed, Installing and maintenance works impose lane closing
Inductive loop
direction and distance between vehicles) Cable coils can be affected by traffic and temperature efforts
It can be used for vehicle classification, by working at higher frequencies. Usually, it must be installed more than one detector to provide a zone monitoring
Not influenced by meteorological conditions (rain, snow, fog) Detection accuracy can be lower, in case of necessity to identify many vehicle types
Need of pavement works for installing
Magnetometer Less traffic influenced.
Incorrect installing reduce pavement reliability
(flow meter Some models permit data transfer by radio waves.
Installing and maintenance works impose lane closing
with two gates) Not influenced by meteorological conditions (rain, snow, fog)
Some models have reduced detection area
Magnetometer Can be used in places that are inappropriate for inductive loops (bridges)
Need of pavement cutting or tunnel digging, for installation.
(magnetometer Some models can be installed without pavement cutting.
Cannot detect stationary vehicles.
with induction Less traffic influenced.
coil) Not influenced by meteorological conditions (rain, snow, fog)
Not influenced by meteorological conditions
Microwave Antenna radiation lobe and waveform must be adapted to the specific application.
Direct speed measurement.
radar Doppler sensors don‟t detect stationary vehicles.
Multiple lane monitoring.
Active detectors transmit multiple beams in order to accurate detect the
Infrared vehicles position, speed and class. Active detectors operation is affected by fog and snow.
radiation Multizone passive detectors are used for speed measurement. Passive detectors can have a lower sensibility in fog or rain conditions.
Multiple lane monitoring
Performances may be affected by external conditions, such as temperature variation or
air turbulence.
Ultrasound Some models are provided with temperature adjustment modules.
Possibility of detection for vehicles which exceed the height limit.
Increased duration of pulse recurrence can negatively influence the lane occupancy
measurements, when speeds are medium or big
Passive detection.
Low temperature affects measurement precision.
Acoustic Negative influenced by precipitation.
Some models are not recommended for zones with low speed or frequent queues.
Sensibility at any external condition variations, vehicle shadows on adjacent lanes,
Multiple lane monitoring and multiple zones per lane monitoring. day-night transitions, vehicle/load contrast, rain drops, fog etc.
Video image Zones can easy be added or modified. At least 15-21 meters high mounting, for optimal detection of presence and for speed
processing Generous range of data available. measurements.
In case of camera interconnection, it can be provided data for extended areas. Some models can be affected by strong wind.
Cost effective only if multiple zones detection is possible with each camera.
Table 2. Vehicle detection systems: data provided
Data provided Multiple lanes,

Communicati
Technology Vehicle multiple detection
Vehicle counting Speed Occupancy Classification on bandwidth
presence areas
Inductive loop yes yes yes1 yes yes2 small-medium
Magnetometer (flow meter with two gates) yes yes yes1 yes small
Magnetometer (magnetometer with induction coil) yes yes1 yes small
Microwave radar yes yes yes yes yes yes medium
Infrared radiation yes yes yes yes yes yes small-medium
Ultrasound yes yes yes small
Acoustic yes yes yes yes yes small-medium
Video image processing yes yes yes yes yes yes small-high
Triple technology detectors yes yes yes yes yes yes small-medium
1
Speed can be measured by using two detectors placed at an established distance.
2
It can be obtained by using some specialised electronic devices.
The magnetic detector is the most limited, because it can work only in pulse mode (the
detector produce a short exit pulse, in the moment of detection). This type of detector can be
used only for movement detection at junction entries or as numbering equipment.
Magnetometers and inductive loops can be used for presence and passing detection.
Magnetometers can produce satisfying results for vehicle counting if the detection zone is
correctly defined. The magnetometers detection area is, normally, less than 1 meter, and the
detector may „miss” some vehicles.
The inductive loop detector is recommended especially for presence detection. The size
of the loop permits adaptation for many applications.
For small detection area, magnetometers and conventional loops are interchangeable. All
three types of detectors are interchangeable only for movement detection (pulse mode). Some
designs for high speed junctions are using conventional loops or normal exit magnetometers.
Other designs are using control unit, which permit the call extension, even after the vehicle
has left the detection area (extended call detectors).
4.3. Advantages and Disadvantages
The vehicle detector systems present many alternatives which are used according to the
specific location and to the data which must be obtained, due to the fact that the detectors
have different characteristics (measurement distance, influence of external factors etc.). The
detectors can be classified in terms of installation mode and detection mode as: in/on
pavement detectors and roadside detectors. In table 1 we present a comparison between the
most used detectors characteristics. Table 2 present a comparison of detectors, according to
the data provided [6].
4.4. Application Examples
4.4.1. Presence and Speed Detection, Using Inductive Loops

Inductive magnetic loops are the most used type of detector. The loops generate an
electromagnetic field, which is modified by passing vehicle, hence detecting the respective
vehicle. Size and form of loops are different, according to specific situation. The main
components of inductive loop detectors are one or more coils from an insulated conductor,
placed in a road seating, a power cable connected to a junction box placed roadside, an
equipment cabinet ant the interior electronic unit, connected to the power supply. Inductive
loops are used to detect passing vehicles, stationary vehicles, for counting vehicles or for size
determination. Also, vehicle speed can be calculated by using a pair of loops, placed at a
determined distance between each other, on the same lane or by using a single loop, using an
algorithm that take into account the loop length, medium vehicle length, loop occupation time
and the number of counted vehicle. New version of detectors can provide data for vehicle
classification, by using higher frequencies.
In order to control the entire circulation direction (all the lanes), it can be used the
scheme presented in Figure 4.
Figure 4. Road monitoring system, by using inductive loops.
The inductive loops can be classified in permanent loops (placed in the pavement and
used for a long period of time) or temporary (“glued” to the pavement and used for short
periods of time).
The loop is functioning like an inductive element connected to an electronic unit. When a
vehicle pass or is stopped over the detector, the inductance decrease. Inductance decreasing
conducts to an increase of oscillator frequency. The electronic unit observed this increased
frequency and send a pulse to the controller, indicating the vehicle presence [6].
The magnetic flow is evenly generated along the loop, except the end portions. The
intensity of the generated magnetic flow is:
NI
H (1)
l
where, H = magnetic flow intensity [A/m], N = coils number and l = resistor length [m].
Because the generated magnetic flow is even, the magnetic flow is:
  BA (2)
where,  = [Wb], B = magnetic flow density [T] and A = loop section area [m2].
The magnetic flow is dependent on the magnetic permeability:
B  0 r H (3)
where, 0 = 4 π×10-7 [H/m], r = relative permeability of the material (1 for air) [H/m].
Loop inductance is:
N  NBA N 0 r HA 0 r N 2 A
L    (4)
I I I l
where, I = electrical current intensity [A].

The main parameters that can be obtained based on inductive loops data are:
Volume:
Q  N /T (5)
where, Q = detected vehicles/hour, N = detected vehicle in time interval T, and T = time

interval, in hours.
Occupancy:
100 N
  (ti  D)
T i 1
(6)
where, θ = occupancy[%], T = time interval [hours], ti = total detector pulse time and D =
descendent slope time - ascendant slope time.
Speed:
3.6  106 d
V (7)
5.280(t1  t0 )
where, V = vehicle speed [km/h].

In case of using one detector: d = medium vehicle length + detector length [meters], t0 =
detection start time [ms] and t1 = detection end time [ms].
In case of using two detectors: d = the distance between the two detectors [meters], t0 =
detection start time for first detector [ms] and t1 = detection start time for second detector
[ms].
Based on the previous formula, it can be deduced the exact length of the vehicle:
1  5.280V 
Lv     t11  t10    t21  t20    (8)
 
2  3.6  106 
where, V = previously established speed, ti0 = detection start time for detector i[ms] and ti1 =
detection end time for detector i[ms].
Another formula for speed calculation, function of the volume and the occupancy is:
Q
V C (9)

where, C = calibration coefficient, experimental determined.

Vehicle density:
1 N  1 
K     
 T  i 1  Vi  (10)
where, K = vehicle density, N = detected vehicles number per time unit T, T = time unit
[hours] and Vi = speed of detected vehicle i.
(Vehicle length  Loop length)

Speed (Figure 5)  K * (11)
(Pulse end time)  (Pulse start time)
D
Speed (Figure 6)  K *
(Pulse start time 2nd Detector)  (Pulse start time 1st Detector)
(12)
4.4.2. Automatic Plate Recognition Systems with Video Cameras

Video image processing sensors use visible light range and a closed to infrared light
range, in order to determine more traffic parameters, according to the implemented
configuration. Video image processing sensor permits obtaining traffic data for complex
areas, with multiple lanes. However, the necessity of video image digitization and of model
recognition routines imposes spending a lot of energy in the field. Such a sensor can provide
data regarding speed and traffic volume and it can also be used for incident detection and
traffic signal control. Video image sensors represent a very fruitful solution for traffic flow
detection, due to the increasing of performances, high flexibility of characteristics and price
decreasing.
Figure 5. Speed detection with one inductive loop.
Figure 6. Speed detection with two inductive loops.

A video detection system is composed by a number of video cameras, digitisation and

image processing system and software for data processing and data generation. These systems
are looking at the images from an area, in order to observe any modification. Complex
algorithms are implemented to identify vehicles (cars, trucks, motorcycles, bicycles),
irrespective of external condition, shadows, day-night transitions etc. Dependent on
algorithms complexity, it can be obtained data that are not available from other detectors,
such as cars registration numbers, it can be provided data about complex situations, such as
road incidents or it can be monitored the drivers behaviour.
The following categories of video detection systems are frequently used:
 Tripline – The user can define only a limited number of detection zones. Vehicles are
detected through pixels changes of the image, against the situation when the road is
clear. There are two detection systems: one realises a surface analysis, identifying the
objects profile; the other realises a grid analysis and identifies in each square vehicle
in movement, stationary vehicle or the absence of vehicles. Speed measurement is
made by computation, the system providing data about the time needed by a vehicle
to transit a known distance.
 Close loop tracking – is an extension of tripline system and offers the advantage of
extended zone detection. The system track permanently the vehicles situated in the
(visual) detection area. In order to validate a vehicle, it is necessary to obtain more
detections of the respective vehicle along its route. After validation, vehicle speed
can be obtained, besides other data, such as lane changing. The exit data can be sent
to variable message signs or to on-board equipment.
 Data association tracking – these systems are tracking certain vehicles or vehicle
groups, by searching of connected pixels areas. These areas are tracked then on all
the frames, to obtain the identified vehicle or vehicles group tracking. The objects are
identified by using some markers, based on gradient and morphology. Gradient based
markers identify contours, and morphology based markers identify characteristics
and dimensions combinations, which are associated with a known vehicle or a known
vehicle group. By successively identifying a vehicle though video system, it can be
determined its route, travel time, medium speed, driver behaviour and other useful
statistics data.
Image formatting and data extraction are performed through powerful operating systems,
which allow real time running of processing algorithms. After data digitisation it is performed
a first vehicle identification (identification errors may appear in this stage). A second data
processing eliminates the erroneous vehicles and maintains only the real vehicles.
Figure 7 presents the event succession that lead to vehicles identification and tracking.
The following paragraphs present vehicle detection modalities and a registration number
recognition algorithm (useful for access control, automated taxes collection, ramp control
etc.).
Digitizing and storing Obtaining the

Detection Monitored traffic
images images
Segmenting the Extraction of Classification and

Vehicle following
image features identification
Figure 7. Video image processing.
Figure 8. Cases of wrong vehicle detection.
Identification of
Identification of the
Drawing the framing the lower side of Identification of
higher side of the Captured image
rectangle(s) the framing perimeters
framing rectangle
rectangle
Identification of
Identification of Composing vectors
objects based on Binary image
neighbor vehicles based on neighbors
compoment blocks
Figure 9. Contour detection.
4.4.2.1. Vehicle Recognition

The first detection method described regards using symmetry and framing scale to
identify vehicles. The algorithms are based on the idea that most vehicles are symmetrical and
can be framed in rectangles, on the video camera images. In order to determine symmetric
elements it is not enough an image grey tones analysis, due to the possibility of anomalies. In
Figure 8 it can be observed three such examples: in the first image, the vehicle is not
symmetric due to the solar rays reflection on the left side; in the second image it is detected
an uniform zone of the lane; in the third image, a background area has symmetrical
characteristics [7].
In order to eliminate reflections and uniform areas it can be performed an object contour
detection. Next, symmetrical elements search is performed. The first and second problems are
eliminated by this method. In the first case, the object contour can be detected, despite the
stronger brightness from the left side of the vehicle. In the second case, there are no contours
to be detected, for the uniform area.
The next step is to determine the framing rectangles. First, it is searched the lower side of
the rectangle.
As we mentioned before, the framing rectangle determination it is not made only for the
original picture, but also for the image with object contours identified. In Figure 9, it is
depicted the process to determine the lower side of the framing rectangle, and the last step is
vehicle identification.
After the lower side identification, the same criteria are applied to determine the higher
side of the rectangle. This area is determined by taking into account the camera location and
its perspective over the road.
The last step is a backtracking algorithm which eliminates the determined objects which
cannot be vehicles, such as background areas. These areas cannot contain vehicles, due to
their position in relation to the road.
Another solution uses the Block Matching Algorithm (BMA) – the movement detection
algorithm for MPEG (Moving Picture Experts Group) standard. This algorithm has the
advantage that it can be used to determine not only the movement, but also the movement
direction.
The image is divided in specified dimensions blocks (6x6 pixels, 8x8 pixels etc.) and for
each part of it there are analysed the differences between the current image and the previous
image. It has to be done estimation for blocks position in the previous image, in order to
calculate the movements and speed. Each block defines a search area, with the respective
block in the middle. The initial block is translated in the searching area, to determine the
correspondence level for each translation. A movement vector for the block it is generated, in
the moment of location determination.
Based on the speed or direction differences, a distinction can be made even for close
vehicles. The algorithm can produce erroneous vectors, which can see, for example, the same
vehicle as two different objects. In order to eliminate this problem, the next step performs a
medium filtering of values, using the eight adjacent vectors and taking into account only their
non-zero values, to the effect that it will be avoided the situation of transforming in a zero
vector a non-zero vector, surrounded by zero vectors.
The objective is to minimise the sum of distances to all neighbours and the distance
between two vectors is:
u  v  ( xu  xv )2  ( yu  yv )2
(13)
This regularisation significantly reduces the noises and produces block sets with more
evenly disposed vectors.
After data filtering and obtaining compact blocks which represent vehicles, it will be
performed blocks numbering. The numbering is also performed based on the adjacent blocks.
The numbering process is an iterative process. Two blocks are considered adjacent if they
have adjacent borders with identical vectors and the distance between the two blocks is lower
than a specified admissible detection error.
Vehicle tracking is based on current image blocks identification and numbering. The
numbering is temporary, being updated according with the determined movement vectors.
Figure 10. Object identification through constituent blocks.
Each identified block from the current image is relocated back to the previous image,
with a vector that is equal with the opposite of the current movement vector ( v  v ). Next,
it is estimated the superimposing between the current label and the previous label. In Figure
10, the left side shows the previous numbering, and the right side shows the current
temporary numbering. It is possible that the superimposing is not perfect, being reliant to the
object detection accuracy for each image.
After verifying each “3” pixel superimposing over the previously image pixels, it will be
decided that the previous label is better covered than the current image. The respective label
will be transferred to the current object (in our example, the current object will have the final
label: “5”). The procedure is repeated in order to update the numbers for all last image
identified objects. In case that the number for a certain object cannot be identified from the
previous sequence, this will be classified as a new object.
4.4.2.2. Vehicle Registration Number Recognition

This procedure can be used for automatic vehicle recognition, by identifying the unique
registration number.
Some of the methods are: recognition by OCR (Optical Character Recognition)
techniques, model based recognition or neuronal network based methods. Here it is explained
a method which identify the registration number zone, decompose plate characteristics and
identify the existing characters on this zone.
The first step is to extract registration number zone. The video camera image is
transformed in a binary image (0-black and 1-white). This decision is taken by comparing
each pixel‟s luminosity with an established value.
The resulted image is processed in order to decide the registration number area. A lines
and columns scanning is performed, and in case that the white pixels number is not included
in a specified range, the white pixels are turned black.
Extracting the Identification of

Extracting the
Location of vehicle vehicle
vehicle registration Binary image
characters registration plate registration plate
number (as image)
image area
Cross-check with
Identification of Normalizing the Vehicle registration
database for sets of
individual characters characters number
characters
Figure 11. Obtaining of binary image.

Next step is registration number extraction from the digitised image, based on the decided
area.
Finally, an image containing the registration number plate is obtained.
The following step aims to identify the characters. First an image filtering is carried out,
to eliminate noise or unwanted points. Scanning procedure is again applied, to extract the
areas that contain characters.
Individual characters are extracted, by applying this mask on the previously identified
number.
Before starting the recognition procedure, the characters are normalizes. This operation
consists in modifying characters in order to eliminate any white space between the character
and the circumscribed cassette.
Finally, o model comparison is made, by determining the best conformity of each
character with a model character. It is possible to use more models, which different character
sets, in order to correctly identify all the vehicles, irrelevant of the region or country of
provenience.
In conclusion, video detection systems have the larger application range, but imply a high
calculation power and complex algorithms for image processing.
5. COMMUNICATION TECHNOLOGIES
5.1. Introduction
Data communications are required for data collection and information sharing. In terms
of road infrastructure, communications used are stationary. Data from fixed sources such as
loop detectors and closed circuit television may be sent to management centres and then
distributed through fixed and mobile communications. Vehicles need mobile
communications. Data from mobile sources, such as helicopter patrols and outputs of data
processors located inside the vehicles, such as GPS coordinates, must be submitted to the
management centre through wireless or mobile communications. All existing and future
telecommunications infrastructure, stationary and / or mobile can and should be used for ITS
functions in order to minimize costs and benefit from the continuing technological progress in
the telecommunications industry.
5.2. Wireless Technologies
Wireless communications are used more and more for ITS services. In many countries,
drivers of vehicles report often incidents on highways through mobile phones. New personal
communication systems are digital. Data communications via digital cellular systems have
become feasible due to cellular digital packet and particular techniques of wireless data
communication packet switching. The new digital cellular telephone services based on
standards like TDMA (Time Division Multiple Access), CDMA (Code Division Multiple
Access) as well as European GSM standard have already been launched on the market.
Mobile communications may be used in large areas, in which the transmitter and receiver
are located far away from each other (the order of kilometres). Usually in communications
over large areas, ITS data is grouped with other messages. In contrast to these are dedicated
short range communications (DSRC) for ITS, where the range is so short that the
communications are useful for certain dedicated purposes. In general, those who invest in
DSRC transmitters (beacons) infrastructure are not part of the telecommunications industry,
but are public or private organizations interested in some type of ITS systems using dedicated
short-range communications.
ITS services based on dedicated short-range communications are:
• Electronic toll collection;

• Commercial vehicles operation;
• Parking management;
• Signal programming;
• Signalling inside the vehicles;
• Information of the passengers inside the vehicles;
• Route guidance based on the beacons.
5.3. Radio Data Systems (RDS)
Radio Data System represents a form of channel data coding design to operate along with
FM commercial radio emissions. The RDS principle is the following: commercial FM radio
has some "spaces" between frequencies. This space, which serves as a buffer between radio
stations can be used to transmit special audio programs or to transmit digital information.
This sub-sender has been used in the past to broadcast background music in restaurants and
shops. The efficiency of the system is based on the fact that there are no connections between
the sub-sender and the main audio program.
RDS main motivation is to allow commercial radio stations to be identified on the radio
screen and to admit self-regulating radio transmissions, that would automatically select the
strongest signals corresponding to a particular type of program (such as news, sports,
different types of music and traffic information). From the ITS perspective, an important
feature of RDS is TMC (Traffic Message Channel), that can provide passengers information
about current traffic. Traffic information transmitted to vehicle passengers by RDS digital
coded silent messages will get to a display and / or voice synthesizer receiver inside the
vehicle. Messages are language independent; they can be decoded in the language chosen by
the user. Incoming messages are stored in the memory of the receiver and may be selected by
the user as required.
RDS-TMC requires a special decoder. The reason is that RDS-TMC is a sub-sender
consisting of ultrasonic signals beyond human hearing ability. Moreover, conventional radios
cannot decode these signals without adding the special decoder. Basic implementation allows
a radio to indicate whether the receiver outputs traffic information as part of the current
program and if the receiver is now getting a traffic message. This information could be used
to stop a cassette or CD or turn on the radio during the announcement. Usually RDS radios
allow the auditor to use this information as search criteria for identification of radio stations
and choosing only the ones that transmit traffic information. While the receiver/decoder
provides the user interface to the travel information service, its functionality may vary
substantially according to the technical development and market needs.
5.4. VHF-PMR (Professional Mobile Radio) Communications
PMR is a concept which provides complete support for radio networks users. Generally,
PMR sets technical parameters for mobile, fixed or portable radio stations. The following two
requirements are fundamental in determining the usability of a PMR system:
• Full rural coverage in normal operating conditions;

• Mechanisms that ensure the coverage in exceptional circumstances.
Additionally, there is often the need to provide dedicated frequencies for air-ground or
ground-air communications in an organizational network. Dedicated frequencies are required
to eliminate the interferences that the air emitter could induce in ground stations, even for
larger distances. PMR technology creates special configurations to provide radio coverage
even in case of network downtime such as faults of base stations. PMR key user needs can be
classified as follows:
• Specialized functionalities for group and dispatched communications, including

groups‟ management in a safe and dynamic mode, emergency calls management,
prioritizing of communications and so on;
• Response times, normally required to establish a voice communication, are in the
range 0.3 to 1 second, with an average of 0.5 seconds for operation in large areas;
• Continuous radio coverage throughout the service area, including guaranteed
availability of coverage even in exceptional circumstances;
• Voice quality without compromise, allowing voice recognition, even under excessive
external noise.
Critical users of the radio system (which currently use private radio networks and PMR
technologies) have specialized requirements for functionality, response times, radio coverage
and quality of service especially during major incidents - with few possibilities for
compromise (if any) when the safety of society or human life must be protected.
5.5. Area Radio Network (ARN)
Area radio signals radiate uniformly in all directions, like point-to-point links between
microwaves. The dispersion and reflection allow the signal to expand on built surfaces,
although these reduce the signal strength. Land barriers may limit the feasibility of this
technology.
There are various frequency bands available for ARN applications; typically, two
frequencies are assigned to an application. For voice transmissions all receivers are set to the
same broadcaster and the users respond only to messages directed to them. ARN networks
have been primarily used for fleets of vehicles.
A more sophisticated version of this technology is ESMR (Enhanced Specialized Mobile

Radio). This radio uses digital technology to allow more users per channel and provide more
services to the users. Services based on this technology may represent a major competition for
cellular phone services in the near future.
5.6. TETRA (TErrestrial Trunked Radio) – Professional Mobile

Communications for Closed User Groups
TETRA is a global wireless communications standard that presents a lot of similarities

with GSM. In both cases are used cellular systems for digital voice and data communications.
Unlike public GSM network, TETRA focuses especially on support offered for special
services and restricted professional groups such as police, fire brigade, ambulance, transport
companies etc. TETRA networks offer professional support, redundancy and priority levels
for the users in the group. In addition, modern TETRA networks are deployed on IP
platforms, in order to ensure 100% compatibility with the terrestrial networks. Being
standardized since 1994, TETRA is now available in all European countries for non-
governmental group of users. In the TETRA network can be defined closed user groups. It
can perform communication between two users or between two groups.
One of the specific features of TETRA is the dispatcher function. A dispatcher fulfils a
central role for the communications in company or group. The dispatcher can, optionally, take
part in all the connections made (listening and communicating). It is noticeable that TETRA
can make a connection without network support (base stations) – in direct mode. Direct mode
provides point-to-multipoint communications without any instructions from a dispatcher. A
TETRA unit can also function as a repeater between a base station and another terminal,
situated outside the base station coverage. Relay option allows a station within the network
coverage area to be used as a relay station to another station outside the coverage area, using
direct mode.
An important requirement is redundancy of trunk connection circuits‟ arrangements,
allowing traffic routing in case of line faults. In an extreme case, of full isolation, each
network element is capable of standalone operation. It is necessary to ensure that the
transmitting nodes in the network will not affect the overall performance of the network, even
from the design phase. Network architecture must be tolerant to failures in order to provide a
good redundancy. To achieve this tolerance, all the modules are redundant, even doubled, or
provided with backup.
The system software provides automatic supervision of the operation, even "watch-dog"
routines, automatic generation of alarms, fault diagnosis and recovery functions to bypass a
faulty module or for routing.
TETRA system supports emergency calling on different priority levels. The emergency
call shall get access to a higher priority level, in terms of both radio and network resources. In
case of too many emergency calls received at a time, for example in case of major incidents,
the system will automatically divide its resources according to priority classes. All radio
subscribers have the right to initiate emergency calls, but the calls will be segregated into
priority classes to allow end users to distinguish real emergencies from a message sent to the
dispatcher. All users are able to initiate emergency calls routed directly to the national
emergency number 112. The network is designed to have sufficient resources for emergency
calls in any situation that would arise and no user or group of users will have a higher priority
level than emergency calls. But, if necessary, network resources can be de-allocated to
perform emergency transmissions even if it means to disturb other on-going conversations.
5.7. Special Radio Systems and Short Distance Communication Systems
Some radio systems or radio networks support a specific type of application or are even
dedicated to a specific target group. Below we shall present special radio systems used in
transport and traffic management applications. DSRC, WLAN or Bluetooth systems are
developed for proprietary indoor enterprise applications. These systems may also be used in
outdoor applications, for transport and traffic management.
5.7.1. DSRC (Dedicated Short Range Communication)

DSRC is a communications standard developed in Europe, USA and Japan to support ITS
type systems. DSRC is created by short-range communication devices capable of transferring
data at high speed, using air interface, between the moving or stationary vehicle and devices
normally stationary, which is often mounted on structures along the road, or are hand-held
type.
DSRC is suitable for communications on limited range of approximately 30 m.
Applications are in the field of road traffic, for communications between the RSE (Road Side
Equipment) and OBE (On-Board Equipment). RSEs are radio beacons placed above or along
the road. OBEs are small and represent units installed in vehicles, trailers or containers.
DSRC communications supports bidirectional half-duplex mode, for which RSE controls the
connection. For simple applications, such as automatic fare collection, OBE can be pseudo-
passive, if installed behind the windshield of the vehicle. This type of OBE does not generate
radio signals independently, the energy used to emit the response being obtained from the
RSE signal. The energy received is sufficient to achieve temporary high-speed links. The
DSRC standard describes only communications between road side equipment and equipment
inside vehicles. It contains no description of the network structure and data protocols used for
RSE interconnections.
DSRC radio channel bandwidth is 3 MHz and modulation type is binary AM (Amplitude
Modulation). Due to very low range of DSRC and that the antennas used are unidirectional
(small bundle), the same frequencies can be reused after short distance (several hundred
meters).
ITS architecture requires DSRC in those specific applications that necessitate a close
physical interaction between vehicles and roadside infrastructure such as fare collection,
electronic authorization of commercial vehicles, safety inspections etc. The following
applications have been identified, that may use DSRC as the main mechanism and
communication technique:
• Electronic payment of taxes and parking

• Commercial Vehicle Operations
• Authorizations at international borders
• Electronic authorization
• Safety inspection
• Automatic identification of equipment and freight management

• Off-line check
• Operations with emergency and transit vehicles
• Fleet management
• Vehicle use as probes to obtain connection times
• Information on-board the vehicle
• Avoiding collisions at intersections and automated highway systems
• Commercial applications
• Safety warnings at intersections.
5.7.2. WLAN (Wireless Local Area Network)

Wireless networks, such as wireless LAN (WLAN) are used to achieve mobility and
flexibility, such as the extension of a wired network. There are currently a number of
standards for these systems, including the US IEEE 802.11 and the European HIPERLAN.
It is customary to link the different equipment and in LANs - Local Area Networks.
Every aspect (physical connection, the signals used, transmission protocols) for such
networks are standardized so that minimum effort is needed to connect different entities.
The structure of a WLAN can be flexibly changed to meet all requirements. The simplest
form of a network is the connection of two computers, in which case the communication will
be an “ad-hoc” one. For more complex situations the equipment may be connected using a
cellular network concept. This implies that the transmission of data between them is managed
and done through a base station (Access Point). To ensure coverage of larger areas there is the
need of more cells. This type of connection between different base stations can be done via an
existing network (usual LAN) or via a wireless network. WLAN standard supports user
mobility: participants may have access to different access points by roaming. However, the
mobility is limited, the connection being maintained as long as the user moves relatively
slowly (10 km/h).
In WLAN systems are applied some techniques to ensure noise immunity: FHSS
(Frequency Hopping Spread Spectrum) and DSSS (Direct Sequence Spread Spectrum). In
FHSS frequencies are changed regularly and may jump between different channels in the
spectrum. The DSSS system uses a single coding system to spread the information using an
extremely wide range of frequencies. Both methods give good results in case of
electromagnetic disturbances and contribute to anonymity (sender and recipient addresses)
and immunity from unauthorized interception.
WLAN systems based on the IEEE 802.11 standard provide a wireless connection for
data transfer speed of 54 Mbps. The effective data transmission rate per terminal is relatively
modest. This is because it is necessary to resend the messages and the available capacity of
radio channel per unit is dependent on the current requests from other cells operating in the
WLAN. Typically, in the absence of other units, the effective rate of transfer between the
units is 10 Mbps.
In addition to the frequency band of 2.4 GHz there are WLAN systems available for
frequencies around 5.8 GHz. In these WLAN systems, based on HIPERLAN standard, are
attainable higher transfer speeds.
5.7.3. Wi-Fi (Wireless Fidelity)

“Wi-Fi” is a trademark of the Wi-Fi Alliance, which uses IEEE protocol 802.11b, a
special WLAN version, which was described above. Wi-Fi uses access-points that connect to
high-speed data lines from the data network and the Internet.
Wi-Fi is designed to transmit data over wireless network portions at very high speeds,
easily reaching Mbps. These high-speed Wi-Fi connections are governed by several factors,
including: the number of active users that utilize a single access point, the operating distance
from the access point, any obstructions blocking the signal. Generally, Wi-Fi signals are
transmitted relatively well through the glass and several types of walls, but do not propagate
well through metal, concrete or buildings.
The term “hotspot” became popular, referring to a publicly available access point in a
location other than home or work. In general, all hotspots and Wi-Fi access points have a
limited radius of action, around 100 m or less. New technologies and the use of multiple
access points will make possible the extension of the signal‟s range.
5.7.4. Bluetooth
Bluetooth is a relatively newly developed technology for data transmission over short
distances. It aims at first to find an alternative for the need to interconnect different
peripherals of PCs with cables: cable interconnecting keyboard, mouse, printer, etc. Bluetooth
can also be used for various applications such as wireless telephony, shopping, electronic
parking payment, public transport or access control.
Bluetooth standard (IEEE 803.15) contains the description of all issues related to data
sharing, in this case radio interface, transmission protocols and signal characteristics. Each
Bluetooth network has one “master” equipment and one to seven “slaves”. Each unit has a
unique 12-bit address. The “master” controls the access and internal transmissions.
The Bluetooth transmission power is divided into classes. Class 3 has a maximum
transmission power of 100 mW (covering about 100 m) and Class 2 transmit power is only
2.5 mW (covering about 1 m). Bluetooth also support limited mobility. In the coverage area
of the “master” station, the link is maintained even if the cell is slowly moving. Bluetooth
radio interface uses frequency hopping over the whole ISM band (79 channels). A radio
channel has a bandwidth of 1 MHz and is used for uplink and downlink.
Bluetooth connection capability is not necessarily symmetrically divided between uplink
and downlink. In the case of a symmetrical partition, the data transmission rate for each link
is 432.6 kbps. In the case of an asymmetric division, the maximum speed for a link is 721
kbps, and for the other is 57.6 kbps. The exchange of information is done the data packets.
Similarly to WLAN, information exchange takes place without errors based on resending the
messages received with errors. Information can be protected against interception using
encryption keys.
5.7.5. NTCIP (National Transportation Communications for ITS Protocol)

NTCIP was originally designed as an extension of the NEMATS-2 standard, used in
traffic control communications. For the functional implementation of ITS, all the components
that make up the traffic and transportation control must be able to communicate through a
common language, or at least an understandable one. Words that are "spoken" must have a
clear and unambiguous meaning for everyone. NTCIP developers started by defining a
language for traffic control and expanded it to include Traffic Management Centres (TCM).
Later it was diversified, resulting an open set of protocols to meet the various requirements of
ITS.
6. INFORMING PUBLIC SYSTEMS

The continuous increase of traffic in major cities led to the mandatory usage of traffic
management systems in order to have an organized traffic. There are many levels such
systems may be implemented, depending on the traffic volume, starting from the static
signalling systems up to management systems based on complex algorithms that consider a
series of key factors.
Information obtained by the management traffic systems must be disseminated to the
users of the system: the travellers. Beside the static signalisation methods, it is essential to
obtain dynamic information of the traffic participants, meaning an information regarding the
real traffic condition.
There are many technologies that can be used to realise this goal, the most usual being:
variable message signs, highway advisory radio, personal information systems, vehicle
information systems, commercial radio broadcasting etc.
Below we present a project focused on Multimodal Real Time Traffic and Travel
Information (RTTI) services with the goal to reduce drastically energy consumption in urban
areas across the different modes of transport by changing the mobility behaviour (modal shift)
of the single traveller. The same project can be used to transmit information also for the
travellers that are using the interurban road network.
6.1. The In-Time Project - Introduction
The In-Time (Intelligent and Efficient Travel management for European Cities) project is
developing a traffic management system that provides mainly 3 services:
 Business-to-business (B2B) services that will enable European-wide Traffic

Information Service Providers (TISPs) to get access to regional traffic and travel data
and services of the single pilot cities via a harmonized standardized open interface.
This will enable the TISP to provide interoperable and multimodal RTTI services (e-
services) to their end-users.
 E-services will influence the on-trip travel behaviour by optimizing journeys taking
the energy consumption into account. The community will be the users of mobile
devices or navigational devices.
 Web based interoperable and intermodal pre-trip information will be provided by the
pilot operators and has the potential to influence the travel behaviour in the trip
planning stage by taking environmental aspects into account. The typical users are
persons that are planning an urban trip on short term. [8]
The fundamental concept and goal of the In-Time project is to pilot and validate – in
terms of transport benefits, user acceptance and market potentials – an innovative pan-
European approach to RTTI service provision based on an open, standardized service oriented
infrastructure and B2B services that will facilitate access to urban traffic related data, RTTI
service provision and interoperability by TISPs.
The central part of the In-Time concept is the interoperable and multimodal Regional
Data/Service Server (RDSS), a service-oriented middleware infrastructure providing a
standardized B2B interface and a number of data/services covering individual traffic, public
transport, weather, location based services, inter-modal transport planning. Investigated and
tested at a prototype level in the EU project eMOTION the In-Time RDSS enables the
operation of end-user applications (web based B2C applications and services) as well as the
B2B access via the standardized interface and e-service provision by European TISPs. The In-
Time RDSS will be set up in all pilot sites to ensure the easy access of real-time multimodal
traffic data for external TISPs. This model ensures the easy access to all urban traffic related
data within one region resulting in the distribution to the end-users via several information
channels and in parallel enhancing user acceptance.
The RDSS can be seen as a service-oriented middleware infrastructure providing a
number of data/services, covering:
 individual traffic,
 public transport,
 weather,
 location based services,
 intermodal transport planning,
and enables the operation of the In-Time services.
Figure 12. The In-Time concept.

Figure 13. RDSS concept.
6.2. Description of In-Time Services
Web based interoperable and intermodal pre-trip and on-trip information are very
common interfaces from the single operators to their users. "Pre-trip information is already
an important contribution to sustainable mobility" (Final Report of the eSafety RTTIWG),
but has in many cases the disadvantage that a door-to-door trip planning taking different
modes into account, is currently impossible.
The new approach within In-Time will ensure that on the one hand each pilot site will
have the complete travel information (public and individual traffic) combined and interpreted
at one location - the Interoperable and Multimodal In-Time Regional Data/Service Server
(RDSS). This ensures that the user gets accurate and precise door-to-door information on the
current travel times within the city, including alternatives. But as shown in above usage
example, the setup of the RDSS will give also the possibility for a seamless transition of
data/services between the single regions and to result in interoperable cross-border travel
information. It is expected that this service will facilitate inter-modal changes, thus attracting
travellers to less energy consuming transport modes.
For the implementation of web based interoperable and intermodal pre-trip and on-trip
information the costs will be mainly with the operators, that might be public authorities and
ppp-models.
Accurate reliable intermodal real-time on-trip Traffic and Travel information as planned
in the In-Time e-services has the potential to avoid congestions and bottlenecks, to select
alternative routes and/or alternative transport modes. Hereby one key factor is, that the user
needs to be used to the device he uses, therefore within In-Time standard products of
Wayfinder will be enhanced by the e-service.
Figure 14. In-Time pilot sites.

With the assessment of the operating Traffic Management the environmental impact of
new traffic management technologies, that have reduced energy efforts, it is possible not only
to improve visibility but also to cut the energy consumed to at least one fifth (in terms of CO2
emissions, comparable to the local impact of vehicle control measures!). The pilot of
Bucharest is an opportunity to make direct measures of the savings in terms of overall energy
efficiency across the city network, as here historical data from before the implementation of
modern traffic management technologies are available.
Beside the energy consumption also the efficiency of the Infrastructure Device Product
Life Cycle needs to be taken into consideration. A global approach must consider also energy
efficiency of the whole life-cycle of products used for traffic control. Again the example
comes from the signal heads, specific LED-based products have gone through the assessment
of their Global Warming Potential and demonstrated dramatic improvement if compared with
traditional productions. This technology is used in the field in the Bucharest demonstrator and
analysis for future “green-certification” rules will be performed.
6.3. IN-TIME Pilot Sites
The In-Time services is piloted and operated in following European cities (Figure 14):
 Vienna: 3.5 M inhabitants

 Brno: 0.4 M inhabitants
 Munich: 1.3 M inhabitants
 Bucharest: 2.3 M inhabitants
 Oslo: 0.5 M inhabitants
 Florence: 0.6 M inhabitants
6.4. IN-TIME Bucharest Pilot Site
Bucharest is one of the major growing cities within Europe with a potential of approx. 7
million inhabitants until 2015. In Bucharest it was installed and it is functional a complete
traffic management system, which consists of:
 140 intersections controlled with the UTOPIA UTC system;

 300 vehicles monitored and managed by the MIZAR Fleet Management system
FLASH
 about 300 probe vehicles equipped with monitoring equipment
 40 CCTV cameras used for security monitoring purposes.
The dynamic traffic control is implemented in order to measure the impact of the
adaptive traffic control system configured to reduce fuel consumption (i.e. implement the eco-
flow concept) with selective bus priority function the data collected used as a basis will
include:
 travel time both private cars and public transport vehicles (300) on selected fixed
routes within the area where the UTC system is active (these are referred to in the
following as fixed route trials);
 measurement of number of stops made at traffic lights and overall time vehicle is
stationary (based on probe vehicle data);
 average speed of private vehicles when in motion.
To assess the impact of pre-trip and on-trip traveller information and guidance on
traveller route and modal choices, it is planned to collect the following data:
 travel time broken down into component areas (e.g. on board, waiting time at bus
stop etc.) for selected OD relations within the area where the Bucharest city systems
are available. This will be done where users have free path selection and will be
carried out with the Bucharest systems ON and OFF. These trials will be referred to
as OD trials. The survey will be designed mainly to assess the differences in total
distances travelled and, where possible, vehicle speed. The “Politehnica” University
participated in 2007 as subcontractor in the project “Bucharest Transport
Masterplan”, with the Municipality of Bucharest as beneficiary. Part of the project
involved collecting of OD data and also Household Surveys (questionnaires about
public travel by private cars and public transport: time of departure, time of arrival,
mode of transport, waiting time and so on). This data can be updated and completed
with other necessary information.
 Energy consumption estimates: it is intended to apply energy consumption models
and algorithms (sing a methodology to be developed during the first phase of the
project) to the above data in order to produce estimates of energy consumption in the
different ITS areas. In addition, overall fuel consumption and motorisation statistics
for Bucharest will be used to calculate:
 the change in energy consumption (volume of fuel consumed per million
inhabitants/year for vehicles within the controlled system).
The efficiency of technology choices with regard to energy consumption of the

infrastructure and especially the adoption of LED technology for signal heads will be
assessed by means of:
 an estimate of the amount of energy needed to operate the traffic control system in
the area concerned; Here, in addition, it is also planned to make
 a product life cycle analysis, with information from the manufacturer to permit an
estimate of total costs over full lifespan of the LED technology;
 an estimate of CO2 emissions is also possible.
The pilot project is an opportunity to make direct measures of the savings in terms of
overall energy efficiency across the city network.
As a part of the pilot, it is planned to use data collected from probe vehicles to constantly
monitor the operational effectiveness in terms of energy efficiency of the measures described
above. The continuous measurement of actual average fuel consumption of vehicles driving
through the network will provide evidence of the overall trend and quantify the impact of any
new strategies. This approach will guarantee constant improvement in the policies adopted, as
for other network parameters are already commonly measured (e.g. travel time).
7. COMMAND AND CONTROL CENTRE

7.1. Introduction
In the actually world when more and more actions are centralized managed by
specialized experts, and the information are transferred and stored using electronically
devices, it is absolutely necessary to ensure reliable data infrastructure, including central
managing points and communication data networks. As a beginning, a special attention in this
domain has been ensured by the emergency organizations around the world, those being some
of the first entities which have to involve in such cases. For effective activities, they need the
most realistic information from the field and well coordination in action, all of this needing a
reliable communication infrastructure.
7.2. Concept
The concept of Command and Control Centre for traffic management starts from the idea
that the action has to be fully controlled from one central point.
The Command and Control Centre are permanently working systems, typically 24/7,
representing the “heart” and the “brain” of the complex integrated system. The operators are
responsible of all systems and actions, their job being the coordination and supervising of the
automatic process in the field.
Each place in the Command and Control Centre has to ensure well working conditions,
both for operators, support teams and, not at last, the equipment.
The main spaces in any Centre are:
 Operators room (Dispatch) – probably the most important area and de operative
core, this area will be dimensioned so as to ensure enough space for all regular
operators, supervisors and visitors. From the design phase, any control room has to
ensure enough space for all future extensions. Technically, the space provides a high-
reliability powerful data network, redundant, capable to cover all voice and data
needs. The control room ensures conditions for at least the follow categories of
operators:
 Permanent operators;
 Supervisors;
 Experts;
 Technical referents and specialists;
As an option, in the Dispatch there will be possible to add more operators, permanent or
visitors. For this, the infrastructure can be developed so as to ensure enough technical
resources.
 Equipment room – dedicated to all technical facilities in the Centre, it provides

enough space for all servers, wall-display equipment, telecommunication equipment
and network. The space must be well ventilated, secured and dotted with inert gas
fire protection. Preferable, the space should be designed in according with TIER 4
standards suite.
 Other spaces – dedicated spaces, equipped with modern facilities, will be delivered.
The role of those spaces is to ensure all facilities for normal working conditions for
operators.
7.3. Architecture
The Dispatcher should contain equipment with high processing capacity, redundant to
ensure availability, built-in security solutions, easy to manage and high performance, capable
of processing data traffic received from video cameras and other sensors place in the field and
which need to be monitored.
In intersections where FO is available industrial switches should be used, fan-less and
optimised for ensuring connectivity between the surveillance / miniaturization and dispatcher.
Industrial switches offer the possibility of using ring topology that can suffer single point of
failure (broken cable or removed connector) while continuing to work as if no incident
occurred in the network. This solution ensures continuous functioning to the surveillance
cameras, traffic lights and other sensors with the command centre even in case of an incident
within the communication network.
Figure 15. Command and Control Centre typical design.

The architecture of the system is based on a high reliable core data-network, partitioned
in sub-network which covers all the IT&C needs. Those sub-networks are grouped in
categories, as follow:
 Operators network;
 Server farm;
 Communications link;
 Wall-display and console.
The network uses a unique time reference, such a GPS time-extractor. This solution is
considered the most reliable case, especially because most of the linked operators (public,
private or governmental) use the same time reference, making the interface simple and more
efficient. More than this, the time generator equipment is standard, easy configurable and well
known by most of the technical support personnel.
The server farm is implemented by using a complex infrastructure, made by independent
machines or virtualised servers – the final solution can be chosen in according with the
budget, software licence policy of specific needs. The servers should provide enough
computing power and storage volume in order to ensure the general capacity to manage and
integrate all the processes driven from the central point.
Display walls for Control Centres usually consist of a matrix of rear projection modules.
Main task of a controller is combining these projection modules to one single display wall, in
other words to one giant monitor where application windows can be moved anywhere without
any impact on the application itself. Hence applications can benefit from the entire resolution
of the giant monitor, e.g. displaying the entire network of an energy utility, and several video
windows with information.
Moreover, the controller integrates the display wall seamlessly in a modern digital IT
infrastructure via standardized digital networks (LAN).
Basically a controller consists of a high performance PC based digital graphical cards for
controlling the individual projection modules. As an option, additional video or RGB input
cards (“frame grabbers”) allow to hook up analogue signals which are displayed in moveable
and sizeable windows on the display wall.
The controller and the projection modules are connected via DVI-D (Digital Video
Interface), providing a pure digital configuration from LAN up to the projectors for maximum
display quality and reliability.
The connection between the controller and the projection modules has to be ensured on
the standardized DVI-D (Digital Video Interface).
Transportation medium is either Copper or optical cable.
All wall management functions have to be accessible via a convenient graphical user
interface, running either on any operator‟s workstation or on the display wall controller itself
(remote mode).
Using wall-display management software, operators have to be able to switch predefined
layouts on the display-wall, containing prearranged windows as previously defined or other
functions.
Figure 16. Command and control centre IT architecture.
Figure 17. Wall-display architecture.

For remotely operating all functions on the display wall, remote pointer functionality is
required, enabling to re-route a local keyboard and mouse of an operator workstation to the
display wall.
A pointer finder application assists finding the pointer on large display walls, which
otherwise could be troublesome.
Every working station has to be equipped with complete facilities for all specific
activities. Thus, all working station have to be technically equivalent and be able to operate
full specific jobs, no matter what operator is working on it (traffic operator, CCTV operator,
supervisor, expert, maintenance technician etc.).
The main wiring structure should be mounted in the suspended floor, so as to enter
simply in each cabinet. All the wires should have fire-safe insulations and halogen-free. The
voice / data cables should be separated by the power supply cables in according with general
accepted standards and equipment supplier specifications. Also, a special attention should be
accorded to radio communication cables, which have to be separated by any other cables.
7.4. Redundancy and Back-up Solution
Through the benefits of implementing such solutions, we can present:
 Reliable telecommunication infrastructure both for voice and data, capable to work in
any conditions, including damaged-field (for example in case of an earthquake);
 Command and control infrastructure, implemented in dedicated spaces which ensure
well conditions for 24/7 work, with specialized operators and experts. The IT
infrastructure will be based on a power-full data centre, equipped at least according
with TIER-4 standards. This facility is connected to the reality in the field by using
the previews communication infrastructure;
 For major incidents and disasters, there will be a second Command and Control
Centre, mobile, capable to be placed in the field and ensure all the coordination
activities. More than this, by using a extended communication infrastructure, the
Mobile Command and Control Centre (MCCC) will be able to ensure all the
operational facilities, including a virtual local radio network. More than this, the
MCCC will cover the operational back-up in case of Command and Control Centre
malfunction;
 Back-up and recovery solution, placed far away from the main command and control
centre, which will ensure all the data-space for real time storage of all information
from the main data-centre. Using specific applications, form the Back-up Centre
there will be the possibility to remote all the applications so as to manage a recovery
procedure in the main data-centre, in case of a damage. The communications
between the mail data-centre, the mobile centre and the back-up solution will be
ensured using a high-speed data link, implemented on a dedicated infrastructure, with
at least one loop.
 The IT&C infrastructure has to be designed so as to be reliable enough for operating
with all terminal full active, all communications channel running and all data
recorded. For this the newest technological solutions has been adopted:
 Modern data-centre, designed in according with Tier-4 standard, capable to work

permanently, in any external conditions. This facility will be organized in dedicated
space, with air conditioning system and equipped with monitoring applications,
power back-up system (with UPS and generator) and a multi-layer electronic alarm
system installed for security reasons;
 Virtualized servers, installed on powerful machines, capable to ensure all processing
and delivery data needs: applications servers, unified communications, storage, video
streaming and wall display management applications. For compatibility reasons,
keeping account that a extensive application like this imply lot of suppliers which
have to collaborate, the virtualization solution will ensure 100% compatibility with
all modern operating system;
 Redundant data network, capable to ensure flow and jitter parameters for all
terminals (data and voice), in fully redundant conditions. The network will be real
separated in at least 4 networks, organized in according with the category of
operators: external access network, operator‟s network, server farm network, and
wall display. All those networks will be managed using dedicated tools and it will be
interconnected by OSI Layer 3 equipment, security full granted;
CONCLUSION
Modern highway depend on ITS to increase the safety level. Many sensors that are used
in cities may also be employed on highways with the purpose of traffic management or even
specific applications like speed detection, license plate recognition, vignette check or
automatic fare collection. The use of these detectors requires communications with RSU
(Road Side Units) and then data transmission to a regional Command Centre. After data
processing the communication loop is closed by sending the information back to the field
equipment. Depending on the application, this may imply a large bandwidth and hence
powerful communication systems. Other systems require short distance communications with
a higher level of security. In this case a dedicated short range communications may be used,
and some of them was being developed especially for road traffic applications.
All these aspects were covered in this chapter, where the most relevant devices and
techniques were detailed. The data collected may also be used to inform the drivers about
different aspects related to traffic. This may be done via variable message signs or by
transmitting the information inside the vehicle. In this chapter an information system was
presented, which wad the purpose of providing updated traffic data with the purpose of
generating routes in a navigation system depending on the current road traffic status (free
flow, pre-congested, congested, etc.).
All the equipment and techniques presented in this chapter are part of intelligent transport
systems for highways, providing a safer journey and more information to the drivers, but also
contributing to law enforcement on these high speed routes.
REFERENCES
[1] Timnea, R.; Minea, M.; Nemţanu, F. C.; The ITS Architecture - One of the Most
Important Component for Planning and Developing of the Intelligent Transportation
Systems and a New Approach of the Information and Communication Systems in
Transports Field. International Congress CONAT 2004, Brasov, Romania, p.178, ISBN
973-635-394-X, 20-22 October 2004.
[2] Federal Highway Administration, Traffic Management Centers - The State-of-the-
Practice. US Department of Transportation, Washington, DC, US.
[3] Federal Highway Administration, Innovative Traffic Control - Technology and Practice
in Europe. US Department of Transportation, Washington, DC, US.
[4] Banciu, D.; Hrin, R.; Alexandrescu, M.; Anghel, L.; Intelligent Transport Systems, Ed.
Tehnica. 2003. Timnea, R.; Minea, M.; Nemţanu, F.C.; Public Transport Management
Development Based on ITS Architecture. ITS 05 Prague Conference Proceedings,
Prague, Czech, pp. 42-45, 2005.
[5] Di Stefano, L.; Viarani, E.; Vehicle Detection and Tracking Using the Block Matching
Algorithm Federal Highway Administration. US DoT Traffic Detector Handbook,
Washington, DC, US, 2003.
[6] Pline, J. J.; Traffic Control Devices Handbook. Institute of Transportation Engineers,
2001.
[7] Bates, T.; Accuracy Is Everything: Perfecting the License Plate Reader. Traffic
Technology International, The international Review of Advances Traffic Management,
1999.
[8] UPB CEPETET Group; In-Time Project. Technical Materials, 2011.
Chapter 6
CYBER SECURITY EVALUATION OF CRITICAL

INFRASTRUCTURES SYSTEMS
Alecsandru Patrascu* and Emil Simion†

Advanced Technologies Institute, Bucharest, Romania
ABSTRACT
Over the last 25 years we have seen a massive change in how we manage and think
about information. The Internet is a rather new technology for us, but nevertheless it has
become a force that is working its way into all our lives and there is no doubt in this.
However it is important to realize the security issues involving it. In the 1980s the
information was mostly stored and used in analogue form, as it was stored on papers,
binders, shelves and safes. Today almost all information is digital. It is created and stored
on individual computers and transmitted over computer networks. From the security point
of view, this means that the information sent, often secret or private, can now be
intercepted from anywhere over the planet. This has strong consequences because now
we no longer have to be physically where the data is.
Cyber security has emerged within the latest years as an issue of vital importance for
everyone as governments, businesses and individuals are under constant attack from other
governments, cyber criminals and hackers. These attacks steal large amount of money, in
form of personal-business data, and compromise sensitive government operations.
Industrial software and equipment represents a new attack vector for our modern industry
that involves both the risk of unintentional employee errors as well as aimed attacks.
Attack types such as social engineering, port scanning, packet sniffing, password
cracking, denial of service attacks, machine bugs, server bugs and exploits are no longer
targeted for the desktop and server environments only as they are becoming a real threat
for all non-IT industries. This makes work harder even for security specialists and
researchers, as these directions are new and have not been yet properly evaluated. Since
almost every critical infrastructure like transportations systems, automotive factories,
telecommunication, water supply and electricity generation has a computer network for
decision-making and control in the background, this is vulnerable to a series of threats.
This chapter is focused on these threats with accent on new forms of it, more exactly the
*
E-mail: alecsandru.patrascu@gmail.com, ati@dcti.ro.
†
E-mail: esimion@fmi.unibuc.ro.
186 Alecsandru Patrascu and Emil Simion
issues about industrial control systems like SCADA, which are configured to control and
monitor specific industrial processes, with a case study on Stuxnet. State of the art
vulnerabilities like car viruses are also covered. For in-depth understanding of the
concept involved in cyber security we have included some attack scenarios explained in
detail.
Keywords: Cyber Security, Cryptography, Steganography, Critical Infrastructure, Industrial

Control Systems, Car Viruses, Information Warfare, Smart Grid, Malware, SCADA,
Transportation Network, Intelligent Transport Systems
1. INTRODUCTION
In our days, Internet Technologies optimize the time needed for taking human and
organizational decisions. For example, we are using Internet for electronic communications,
electronic commerce, transactions and banking, for accessing different data bases in order to
process the information. All these actions, which interact with virtual activities, must be
“protected” from electronic fraud. Thus, we need to implement, in the virtual space, security
measures similar to real security measure. But there is a difference: behind real life thefts
there are humans which interact with the goods and in the virtual space the thefts are made by
viruses, worms, malware applications which interact and monitor the actives of the system. If
we think that all these weapons (viruses, worms, malware applications etc.) are produced by
humans we can conclude that in the protection of the system actives the human intelligence
plays a definitive role and the critical decisions must be taken by humans. The problem is
how we can order these decisions, which are in large number, in such a way that to be made,
corrected and modified by a limited human intervention [1, 2].
In this chapter we are going to talk about a new and controversy topic, which is closely
related to the problems that affected until now only traditional computers and networks –
cyber security of critical infrastructures and intelligent transport systems.
Traditionally, critical infrastructures represent the field of utilities and facilities that we
are using daily. These include among many other: electricity grids, gas, transport and
transportation systems. More recent actors are represented by the more and more involvement
of computer systems within all other classic actors that were presented earlier. And since we
rely more and more on computers for proper system function we open a new window of
opportunity to the hackers and all cyber warfare manifestations.
These manifestations had until recently only a single goal: to silently gather data and
multiply across all vulnerable devices reachable from an infection point. This was ended in
2010 when a computer worm called Stuxnet was discovered across different critical
infrastructures across the globe. We can say for sure that everything that is linked to a
traditional computer network can be applied to modern intelligent transport systems,
including the use of zero-days attacks, botnets and all other malware.
Along with other specific tools and techniques that are currently used in traditional
computer networks we are going to present how can we use the same instruments for this
specific field. In this case, the use of scenarios is the main way in which people involved in
incident response can be trained. Often these scenarios are presented to the regular people in a
masked form, and a good example in this field is represented by an international challenge
Cyber Security Evaluation of Critical Infrastructures Systems 187
called “Cyber MITRE”, organized by the Federal Bureau of Investigation and the Fordham
University. We will present in this case seven basic scenarios that can be used in real life
transportation network security breach:
1. Identification of encrypted data in a file

2. Decrypted a piece of encrypted data;
3. Identification of steganography and revealing the hidden data;
4. Identification of a suspect communication between two computers and reveal the
stolen data;
5. Identification of the incorrect usage of cryptographic algorithms usage and finding
the private key used for signing.
6. Identification of host to host wireless communication
7. Interception of host to host communication
The chapter is organized as follows. Section 2 contains a brief introduction in critical

infrastructures; we talk about what is a critical infrastructure from the industry point of view,
what are the security problems that surround these infrastructures and we provide examples
for problems and solutions in this field. In section 3 we talk about the concept of intelligent
transport systems and in section 4 we talk about the concept of smart grids, a rather new topic
which combines a classic approach of electric grids with the new approach of information and
communication technology. Section 5 is dedicated entirely for the newer botnets and malware
threats that are intended for desktops and industrial computer networks. During section 6 we
talk about cyberattacks on SCADA systems; we present what are these systems, how they
work and we provide real examples of attacks, like the Stuxnet malware. In section 7 we
present a new and controversy domain in the security of automotive industry - the protection
of the transportation networks from malware; we present the main types of these new threats,
how they affect our daily lives and as a case study we talk about a subject that is gaining more
and more attention to security specialists, the problem of car viruses. In section 8 we briefly
talk about some of the most used tools in case of cyber security investigations and we present
some practical scenarios used in cyber trainings. Section 9 is dedicated for presenting
different views about critical infrastructures in different countries, like United Kingdom,
Germany, and USA. Finally section 10 of the chapter contains conclusions and an outline of
the directions for future research.
2. CRITICAL INFRASTRUCTURES
People and businesses today are connected using some kind of device or through the
cellular network, Wi-Fi or a local area network. Nevertheless, organizations have not yet fully
understood that endpoint security is vital. Saying this, we can define a critical infrastructure
as a network or system which is vital for an organization, society or economy day-to-day
operations. Historically [3], critical infrastructures have been gathered under terms like
“utilities” and “facilities” that are commonly associated with:
 electricity, gas, oil products generation, transmission and distribution;

 telecommunication;
 water supply;
 agriculture
 food production and distribution;
 public hospitals and ambulances;
 transportation systems (like railways, airports, etc.);
 banking and other financial services;
 people security services.
The criteria that we can use to determine what asset is a critical infrastructure or not, have
been changing over time. According to a report created for the USA Congress in 2002 [4],
“originally were considered as critical the infrastructures whose prolonged unavailability
could cause significant military and/or economic disruption. Today, the term expanded and
covers even national monuments, where an attack can lead to loss of life and affect the
attacked nation morale, and chemical industry, where an attack can alter the proper
functioning of a factory and threaten the safety of surrounding communities”.
It‟s easy to see that in order to have means to deal with threats affecting these systems,
the governments across the globe are working together for a multi-purpose action [5].
Nevertheless, if we find Interpol as the international agency for all criminal activity, for the
cyber space there is no such agency yet, with responsibility closer to its relative. This can be
found in international statements, such as the one made by Subhendu Sahu, director of
Symantec Asia: “the doctrine of national security has to change. There is a need to be aware
and to have the right intelligence. Security is not enough, resilience is required.” This means
that entire economies must know about, understand and prioritize what is important for their
computer networks, especially for those parts of the critical infrastructures.
In order to protect critical infrastructures from different cyberattacks, system
administrators can use many technologies and tools existing today. And since the lately
emergence of central computation nodes or cloud computing technologies are used more and
more, new technologies and tools must be developed if we want to migrate all or part of the
computation made inside such infrastructure to them. Current technologies, like secure access
control, remote system management and checking for system integrity help administrators
and security expert to avoid unauthorized access to information and data that is stored,
processed and transported inside such infrastructures. We can say that all these technologies
should be gathered under a single cyber security framework for “Critical Infrastructure
Protection”, as presented in [6]. According to the authors of this document, “such a
framework must include:
1. determining the business requirements for security;

2. performing risk assessments;
3. establishing a security policy;
4. implementing a cybersecurity solution that includes people, processes, and
technologies to mitigate identified security risks;
5. continuously monitoring and managing security.”
Nevertheless, we must pay attention when implementing and using such frameworks
because they have great costs, which must be justified to the owners. In this entire process we
must not neglect two important factors.
The first one represents the fact that each used technology has a certain limitation and
due the fact that these are designed to work in pair with other technologies they must be used
by professional personnel.
The second one is the lack of international or standardized solutions for these problems.
When working in such environments we often state the fact that, in order to make security a
first-class element, we must be able to develop and impose a long-term plan that contains
such standards and all other practical solutions for the problems involved. In this part, a lot of
help must be given by the government and the designated authorities for it in deploying a
national plan for critical infrastructures protection and allowing agencies in charge with
national security to intervene and collaborate for all the matters involving cyber-security.
3. INTELLIGENT TRANSPORT SYSTEMS

Intelligent transport systems, or ITS for short, are systems designed to provide solutions
to modern transportation modes. We can see them as an automatic traffic management system
that uses modern technology such as sensors and computers to make safer, coordinated and
smarter decisions that affect transport systems.
Why we have need for such ITS? The problem appeared together with the exponential
increase in number of cars and humans on the globe: traffic congestion. To be more efficient,
computing technology was used, both on the field and in dedicated laboratories, for
simulations and models. Of course, along with traffic congestion comes a lot more problems
like pollution, fuel consumption and large travelling times.
Over time, the concept of ITS has become larger, covering more transportation fields
than cars. For example railroads, water and air transport, and all other means of transportation
for both critical and non-critical infrastructures, benefits from this technologies.
Since ITS are very complex systems, they are designed to use multiple technologies for
different modules management. We can find some basic management for cars like traffic
lights, number plate recognition, speed cameras, etc. up to advanced guidance and monitoring
systems, like parking guidance, weather monitor, etc.
In order for these systems to communicate with each other, ITS use different
communication channels such as wireless communication. Wireless communication is a good
alternative to wired communication because it can cover a lot more areas with much less
impact over the environments. In this type we can find technologies used for short-range or
long-range communication.
In case of short-range communications we find the standard wireless IEEE 802.11
protocols. These come along with all their advantages and disadvantages. As advantages we
can mention the fact that they are flexible, as with radio coverage, the nodes can
communicate without many restrictions and furthermore, radio waves can pass through walls.
Of course, a deployment for such technology does not require so much planning as in case of
a wired network. As a downside, a major problem is given by the fact that a lot of devices and
interfere with this kind of communication.
In case of long-range communications we find standards like IEEE 802.16 (WiMAX),

GSM, 3G or satellite. The problem when using such technologies is the cost for the
infrastructure that is going to be used and the response time for the sub-systems.
The development of ITS was marked substantially by the evolving technologies. If ten
years ago a medium priced car contained several tens of chips and microcontrollers for
controlling all important car functions, like braking, today‟s cars contains fewer, but more
powerful processing units, with dedicated memory, sensors and operating system.
We can find applications for ITS all over the world, starting from the notification systems
existing on highways, up to dynamic semaphore lights and ending with those used for
collision avoidance.
4. SMART GRIDS
In the context of critical infrastructure systems we consider a smart grid an evolution
from the classic electrical grid. It is “smart” because it uses the latest evolvements in
information technology and computer network communications in order to automate its
internal work queues. This leads directly to an improved efficiency, lower maintenance costs
and lower effort used for electricity distribution to end clients.
The term of “smart grids” is considered new, since it is used broadly since 2005. It was
first mentioned officially in an article written by Amin and Wollenberg [7] called “Toward a
Smart Grid”. Nevertheless we can trace back this concept, in a primitive form, as far back as
the early 2000. We can give many definitions to this term, from a functional and/or
technological perspective, but all of them have the same central idea: the involvement of
computation technologies to create a centralized managed electric grid.
Currently there are a lot of technologies that we can use in the context of smart grids.
Technically, we can see the entire picture split in two distinct layers. The first one is
composed from the technologies and physical computing resources that are used, and the
second one, relative to the classical electric grid, is composed from the business management
units, which are responsible with the interaction with the end clients.
The main threats surrounding these infrastructures are represented by attacks to the first
layer presented in the previous paragraph. As presented in [8] by the US Council on Foreign
Relations, “concerns chiefly center around the communications technology at the heart of the
smart grid. Designed to allow real-time contact between utilities and meters in customers‟
homes and businesses, there is a very real risk that these capabilities could be exploited for
criminal or even terrorist actions. One of the key capabilities of this connectivity is the ability
to remotely switch off power supplies, enabling utilities to quickly and easily cease or modify
supplies to customers who default on payment. This undoubtedly a massive boon for energy
providers, but also raises some significant security issues”.
But what are those “security issues” mentioned before? To answer this question we must
first try and put ourselves in the place of a cyberattacker. The first thing he would try is to
paralyze the entire activity of a chosen target. And since all their equipment is running on
electric power, we find that attacking a smart grid can be very attractive. We can make a
parallel with a classic war, meaning that this kind of attack can be compared with dropping an
atomic bomb over a targeted region – nothing will stand in its way.
Looking back into our history we can find clear examples, like the one from the Kosovo
war in the late 90s in which electric grids were key targets for fighting the enemies. This is
the main reason for which these infrastructures are critical and a great deal of attention and
money are spent to secure it. This also leads to a psychological war because people panic
once they see they do not have electricity to power their utilities, like cooling or heating
devices.
The concepts behind smart grids were always on the first page on the standardization
institutes. On the “US infrastructure” website we can find some interesting paragraphs
regarding our topic. In their vision [9], “over the next 10 years there will be an opportunity
for ill-intentioned foreign governments to have their own agents inserted in key facilities. This
is generally outside the system of national vetting and clearances in most countries. Second,
it is straightforward to install malware on peoples‟ PCs. Social engineering, spear phishing
can be used easily”.
Furthermore, in the same declaration we can see that “there are other ways in which the
smart grid revolution could bring unwelcome side effects. In the rush to implement security
standards on a national scale a number of organizations are getting involved, including
NIST, the Department of Homeland Security and even the Securities Exchange Committee.
With new standards come new compliance rules, along with all their associated costs. The
desire to circumvent that compliance burden is already having negative impacts.”.
5. CYBER THREATS TARGETING CRITICAL INFRASTRUCTURES

Today‟s critical infrastructures and all its parts, including intelligent transport systems,
suffer the same weaknesses that can be found, for example, in a computer operating system.
And since more and more infrastructures migrate to using operating systems and other
software programs the same or derived from personal computers software, we consider this a
great threat to all infrastructures and its components.
In this following subchapter we will talk about the most important and most often
encountered cyber threats. We will present the concept of zero-day attacks, that stands the
ground for every great cyberattack, and continue with botnets and malware.
5.1. Zero Day Attacks
A “zero-day” attack or threat is an attack that aims to exploit previously unknown

vulnerabilities in software applications. The term comes for the fact that the attack occurs on
“day zero” of awareness of the vulnerability and the developers of the affected application
had no days available for patching it.
These attacks come in a large number. Malware, viruses and Trojans all represent attacks
vectors that target modern software and delivery networks. In this equation the web browsers
and the operating systems on top, which they are running, represent the most widely targets
because they are widespread on all devices, starting with mobile phones and ending with
desktops. Mail delivery networks are also targeted because they can carry to a potential
victim an infected e-mail attachment. To cope with these threats organizations like US-CERT
[2] and Zero Day Initiative [10] dedicate their work in providing users cyber security.
5.2. Botnets
A botnet is a number of Internet computers that, although their owners are unaware of it,
have been mangled and have been set up to send and forward different types of transmissions,
including spam or viruses, to the other computers on the Internet. The term comes from the
fact that any infected computer becomes a “robot”, or “bot” for short, that serves an attacker.
Reports from well knows security companies like Symantec and Kaspersky Labs reach a
common conclusion: botnets currently are the biggest threat to the Internet.
Computers that are used inside a botnet are those whose owners fail to provide effective
firewalls or other safeguards from the Internet. Furthermore, we see that an increasing
number of home computers benefit from high speed Internet connections, thus aiding the
efforts of the attackers. A bot is a program attached to one of the computer ports that is left
open and through this port a remote program can connect to it.
One example is the usage of a botnet to redirect HTTP traffic to another specific
computer or website, in a Distributed Denial-Of-Service (DDoS) attack. The remote website
will be closed down because it cannot handle all the traffic.
5.3. Malware
Malware represents the software used or created by hackers to alter computer and system
operations. The goal is to gather sensitive information or to gain access to private computer
systems. Its form varies from a full software program to a script. It is a general term that is
used to refer to all forms of hostile and intrusive software, like computer viruses, Internet
worms, Trojan horses, spyware, adware, and rootkits.
In the evolution of malware we can establish two big periods: before 2010 and after 2010.
Malware before 2010 was mainly targeted to single computers or medium-sized computer
networks. Since 2010 we can find the so-called “modern malware”, which now has migrated
from personal computers to large systems, even critical industrial systems.
We will present on the following paragraphs one of the main treats in this field,
responsible with industrial systems malfunction called Flame.
Flame, also known as Flamer or sKyWiper [11] is a computer malware that attacks
computer and industrial systems running Microsoft Windows operating system. It was used
for regular and critical infrastructure penetration in several countries around the world [12].
This malware is capable of spreading to other systems over a local area network (LAN)
or through USB sticks. It can record audio, screenshots, keyboard activity and network traffic.
Furthermore it can record instant messaging conversations and can turn mobile infected
computers and auxiliary wireless modules, into Bluetooth beacons which attempt to connect
and download information from other Bluetooth devices nearby. This data was then sent to
one of the several master servers existing in the Internet.
What is interesting on this malware is its built-in capability to “suicide”. The creators of
Flame implemented a “kill” command that automatically removed itself from the infected
computers. This command was caught by the Symantec Security Company using computers
set up to watch the malware‟s actions. More exactly, when it received that command, Flame
located every file existing on the victim PC, deleted it and then overwrote its memory
location with random data to prevent a forensic examination.
According to cryptographic experts [13] [14], Flame was the first malware to use a rather
obscure cryptographic technique called “prefix collision attack”. This allowed it to fake
digital credentials that had helped it to spread. The exact method for this kind of attack was
demonstrated in 2008, but the creators of Flame implemented their own variant. This
determined Marc Stevens to state that “the design of this new variant required world-class
cryptanalysis”. All these findings give support to claims that Flame must have been built by a
nation state rather than cybercriminals due to the large amount of time, effort and resources
that have been put into its creation.
6. CYBERATTACKS ON SCADA SYSTEMS

Supervisory Control and Data Acquisition (SCADA) systems, created by Siemens,
represent an industrial large scale control system used to automate different industrial
processes existing in facilities such as power generation, water supply, etc. One of its main
advantages is that this technology can be used in more delicate fields, where precision is vital,
such as nuclear fusion facilities. Their main purpose is to monitor and control all the
operation inside such facility using a large network of sensors and computers for central
decision making. Since SCADA systems are designed to be used as a control system, they
must be complex and safe. This is one of the main strength points for this system, as it can
manage up to million subsystems input and output channels.
As a brief description, we will present the main components of a SCADA system. The
main parts are represented by the signaling hardware for input and output process lines, the
network between subsystems, the control devices, the user interface to the subsystems
(Human-Machine Interface, or HMI), communication equipment and modules, and, of course,
the operating system and all the software for the modules and subsystems. The measurement
and control modules of a SCADA system has one so-called Master Terminal Unit (MTU),
which is considered the central management node. The chosen operating system for it is
Microsoft Windows. This MTU handles a series of Remote Terminal Units (RTU), which are
responsible with local data gathering.
In this current form, SCADA systems use open-loop control systems with a series of
some closed-loop characteristics built in. Because of this, SCADA systems cannot use
feedback from remote modules to check the final results. In other words, in SCADA we
cannot use machine-learning software and algorithms.
Because SCADA is a complex system that can be used with success even in critical
infrastructures, it represented an attraction to cybercriminals. In the following subchapter we
will talk about Stuxnet, a computer worm that targeted Siemens industrial software and
equipment running on Microsoft Windows. The analysis that follows represents a compilation
of information that has been gathered from different international sources [15, 16] regarding
Stuxnet.
Stuxnet is a computer worm discovered in June 2010 and it is the first public menace that
targeted Siemens industrial control systems – SCADA [15]. It used advanced techniques for
spreading by taking advantage of several Microsoft Windows vulnerabilities in the USB
stack. Also, it is the first malware that used malicious STL code (Statement List) and the first
to use a PLC (programmable logic controller) rootkit.
Researchers at Symantec have cracked Stuxnet cryptographic system, and they say it is
the first worm built not only to spy on industrial systems, but also to reprogram them. Once
installed on a PC, Stuxnet uses Siemens default passwords to seek out and try to gain access
to systems that run the WinCC and PCS 7 programs, the PLCs that are used to manage large-
scale industrial systems on factory floors and in military installations and chemical and power
plants. The software operates in two stages following infection. First it uploads configuration
information about the Siemens system to a command-and-control server. Then the attackers
are able to pick a target and actually reprogram the way it works.
Based on information in the code, Stuxnet was targeting a facility that had 33 or more of
the frequency converter drives installed, all operating at between 807Hz and 1210Hz. The
malware would sit quietly on the system doing reconnaissance for about two weeks, then
launch its attack swiftly and quietly, increasing the frequency of the converters to 1,410Hz for
15 minutes, before restoring them to a normal frequency of 1,064Hz. The frequency would
remain at this level for 27 days, before Stuxnet would kick in again and drop the frequency
down to 2Hz for 50 minutes.
The drives would remain untouched for another 27 days, before Stuxnet would attack
again with the same sequence. The extreme range of frequencies suggested Stuxnet was
trying to destroy whatever was on the other end of the converters.
7. PROTECTION OF TRANSPORTATION NETWORKS

The area of intelligent transport systems and intelligent cars is becoming more and more
a reality surrounding us. And since we rely more on more on technology to make our daily
tasks easier, we tend to forget about security involving these kind of systems.
The topic of protection transportation network and attacks that can be made to a car is a
new and emerging field of research and more and more scientists are joining every day. This
is important because, as we have seen until now, securing critical infrastructures require
mixing many of the technologies existing for regular computer or networks.
We can use a brief introduction in this topic from Zhang and Levinson‟s document,
“Investing for reliability and security in transportation networks” [30]. They consider that
“transportation networks support various vital human activities, and their performance
affects the efficiency of virtually all economic transactions. Decision makers need to consider
the potential trade-off between network efficiency and reliability when they forge policies that
could shape future transportation networks.”
Checkoway, Mc Coy and Kantor present in their paper entitled “Comprehensive
experimental analyses of automotive attack surface”[17], how a modern car is built, what
electronics does it have, and most important, how can someone hack into them. According to
them, “a modern automobile is controlled by tens of distinct computers physically
interconnected with each other via internal (wired) buses and thus exposed to one another. A
non-trivial number of these components are also externally accessible via a variety of I/O
interfaces.”
Current automotive systems can be considered as a network of embedded systems, like
Electronic Control Units (ECUs) and sensors, which are connected to each other via different
bus systems and thus, exhibit a considerable complexity. Moreover, in the recent years,
automotive systems became more and more software intensive systems. This is caused by the
fact that more and more functionality of a car is implemented by software functions.
Of course we must not neglect the software modules existing in these embedded systems.
In this direction we must agree with Schulze et al. in their work “IT security in automotive
software development”[18] saying that “in order to develop software in an efficient way and
to overcome the still increasing complexity anyway, techniques and concepts from the field of
software engineering (SE) and requirements engineering (RE) have been proposed and
applied to the development process. Although the software should contribute, amongst others,
to the reliability of the system, one aspect has been neglected so far in the development
process: holistic concepts for IT security. Successful attacks on an automotive IT system can
have negative implications on the safety of its human users or on the reliability of the system
itself. However, today it is common that the mostly generated code is retrofitted at the end of
the development process in order to satisfy security concerns. This, in turn, counteracts the
effort of reducing complexity by using modern concepts of SE and RE. Furthermore, this
process is prone to simply ignore serious vulnerabilities and thus, allows for security leaks
remaining in the system”.
This is the main reason why we must not concentrate only on hardware, and we must pay
more attention to the particular software engineering problems that appear when developing
for cars. As Shulze continues, “along with the mentioned fundamental changes within
automotive systems, the IT security becomes an important issue. Per definition, IT security
means reliability in terms of preserving security aspects of information, namely integrity,
availability, authenticity, non-repudiability, confidentiality and privacy. Because of its
networked character, automotive systems exhibit vulnerabilities to malicious attacks, which,
in turn, can violate one or more aspects of IT security. The access for the execution of an
attack on the system can take place in multiple forms from outside or inside the car.
Regardless of how access is achieved by the user, the basic attack principle is always the
same: it aims at influencing a certain behavior or state of the automotive system. Since this is
done by manipulating the respective functionality (e.g., by introducing malicious code,
communication or manipulating data), it directly addresses the software responsible for this
functionality. Subsequently, it is reasonable to ensure the IT security of the respective
software in order to increase the security of the overall system”.
7.1. Car Viruses
In the last years there was a lot of debate regarding the concept of car viruses. The reason
is because, as we can see in [19, 20], in the past, car viruses were rare because one of the only
ways to infect a vehicle was by a mechanic and via the computer or software he used to
diagnose problems with the car.
A good definition of this concept can be extracted from [21]. In this article we are
introduced to the idea of car viruses and what they require. As the authors stated, “modern
cars are full of computers and they seem more the realm of an IT expert than a regular car
mechanic. And it certainly is true that modern vehicles have plenty of computers, although
they‟re not exactly like PCs. “Cars have much simpler processors than a home computer and
are designed to do simple, dedicated tasks,“ says Cameron Camp, a researcher at ESET, a
technology security company.
Indeed, most cars today have numerous embedded systems, which are small computers
controlling very specific aspects of the car„s functioning, such as air bag deployment, cruise
control, anti-lock braking systems and power seating. While these embedded systems share
the same architecture as a PC - they utilize hardware, software, memory and a processor -
they are more akin to a smartphone in sophistication than a laptop. Automotive computers
have been more or less immune to hackers and viruses because, unlike PCs, there have been
few ways for outside computers or people to connect with vehicle computers.
Not surprisingly, automakers are said to be working on ways to prevent hackers from
introducing viruses into cars and otherwise making mischief, although details of their efforts
are not readily available. Still, opinions are mixed about how much of a concern this really is
for future drivers”.
7.2. Protecting the Transportation Network from other Types of Software

Attacks. Solutions
Schulze et al. also present in [18] the connection between automotive systems and
software. “As already stated, automotive systems more and more rely on software to fulfill
certain functionalities. Furthermore, the complexity of such systems steadily increases, while
the reliability has to be ensured. Altogether, this is a challenging task to be managed during
the software development process of such systems. Hence, different approaches of software
development found their way into the systems engineering domain in order to overcome these
challenges. For instance, software product lines (SPL) as a special concept of software
engineering are used to manage commonalities and variability of automotive software.
Another common practice is model-based development (MBD) of software for automotive
systems, where functionality is described by models. Afterwards, the code is generated
automatically based on these models. Finally, requirements engineering gains more and more
importance within automotive systems since a good requirements analysis is inevitable for all
other stages of the development process”.
Also as can we see in [17],”past work has illuminated specific classes of threats to
automotive systems, such as the technical security properties of their internal networks; we
believe that it is critical for future work to place specific threats and defenses in the context of
the entire automotive platform. Technical capabilities describe the assumptions concerning
what the adversary knows about its target vehicles as well as her ability to analyze these
systems to develop malicious inputs for various I/O channels. For example, we assume that
the adversary has access to an instance of technical capabilities that describe our
assumptions concerning about what the adversary knows about its target vehicles as well as
his ability to analyze these systems to develop malicious inputs for various I/O channels.”.
In the following paragraphs we have presented also some interesting parts from
Checkoway‟s et al. [17] research regarding the vulnerabilities surrounding a car along with
means to physical or wireless access it.
“Modern automobiles provide several physical interfaces that either directly or indirectly
access the car‟s internal networks. We consider the full physical attack surface here, under
the constraint that the adversary may not directly access these physical interfaces herself but
must instead work through some intermediary.
A. OBD-II. The most significant automotive interface is the OBD-II port, federally
mandated in the U.S., which typically provides direct access to the automobile‟s key
CAN buses and can provide sufficient access to compromise the full range of
automotive systems. While our threat model forbids the adversary from direct access
herself, we note that the OBD-II port is commonly accessed by service personnel
during routine maintenance for both diagnostics and ECU programming.
In both situations Windows-based computers directly or indirectly control the
data to be sent to the automobile. Thus, if an adversary were able to compromise
such systems at the dealership she could amplify this access to attack any cars under
service.
B. Entertainment: Disc, USB and iPod. The other important classes of physical
interfaces are focused on entertainment systems. Virtually all automobiles shipped
today provide a CD player able to interpret a wide variety of audio formats like raw
“Red Book” audio, MP3, WMA, and so on. Similarly, vehicle manufacturers also
provide some kind of external digital multimedia port, typically either a USB port or
an iPod/iPhone docking port, for allowing users to control their car‟s media system
using their personal audio player or phone. Some manufacturers have widened this
interface further; BMW and Mini recently announced their support for “iPod Out,”
a scheme whereby Apple media devices will be able to control the display on the
car‟s console.”
8. TOOLS AND SCENARIOS THAT HELP INVESTIGATING INTELLIGENT

TRANSPORT SYSTEMS INCIDENTS
Attacks on intelligent transport systems can be handled using tools and methods common
to the field of classic computer networks and regular computing systems. As stated in the first
section, it is common in this field to use of scenarios. These scenarios represent the main way
in which people involved in incident response can be trained. Often these scenarios are
presented to the regular people in a masked form. This means that an expert handling a
modern transport system can use in case of incident handling and response the same tools as
an expert in computer networks.
In this part we will present some of these tools, along with their capabilities. Also, in
order to be more relevant to the reader, we present these tools in action, applied to the “Cyber
MITRE” international challenge mentioned before.
8.1. Tools
In general, the tools used in incident response tools are designed as a general application,
which can have multiple other uses. We are going to briefly talk about three main
representative categories: stand-alone tools, statistical tools and security oriented
distributions.
8.1.1. Stand-Alone Tools

Stand-alone tools represent tools that can be used independently over an operating
system. Representative to this field is CrypTool. This is a software package dedicated to
cryptographic simulation, analysis and cracking which has a user graphic interface.
CrypTool has been developed in cooperation with prestigious universities and thus has
become excellent educational software and also a tool for learning cryptology. CrypTool
covers both branches of cryptology: cryptography and cryptanalysis. Thus, the product has
implement facilities of each field, such as classic cryptography (Caesar and Vigenère ciphers,
mono-alphabetic substitution, etc.), symmetric cryptography (IDEA, RC2, AES, etc.),
asymmetric cryptography (RSA and elliptic curves, etc.), hash functions (MD2, MD5, SHA-
1, etc.), cypher text attacks, plaintext attacks, adaptive attacks, side channel attacks.
8.1.2. Statistical Tools

In order to test the degree of randomness for input or output for such transportation
systems, we need a set of different tools, tests and theoretical models – we need statistical
tools. Using this kind of tools we can test for true randomness of functions that are part of the
software implementation of these systems. Good examples in this direction are:
 “A Statistical Test Suite for Random and Pseudorandom Number Generators for
Cryptographic Applications” [22], is a publication of sixteen statistical tests. The
authors provide also an implementation for it.
 “The Art of Computer Programming, Seminumerical algorithms, Volume 2”, by
Donald Knuth [23] contains the theoretical description for some of these tools, that
are based on permutations, birthday spacing, serial correlations, etc.
 The Crypt-XS suite developed by the researchers from Queensland University of
Technology, Australia. This suite contains the implementation for some tools
described by Knuth, along with other custom ones for binary derivative, sequence
and linear complexity measurement, etc.
 The DIEHARD suite [24] developed by George Marsaglia adds to the tools
mentioned before, tests such as random spheres, overlapping sums, etc.
8.1.3. Security Oriented Distributions

Penetration testing [25] and security auditing are now part of every system
administrator‟s “other duties as assigned”. In this direction, BackTrack Linux (BTL) or its
newer version, Kali [26], comes to help. BTL is a custom distribution build on top of Debian
Linux, designed for security testing for all skill levels from novice to expert. It is the largest
collection of wireless hacking, server exploiting, web application assessing and social-
engineering tools available in a single Linux distribution. It contains more than 300 security
tools and utilities that are all open source, grouped in major categories like: information
gathering, vulnerability assessment, exploitation tools, privilege escalation, maintaining

access, reverse engineering, RFID tools, stress testing, forensics, reporting tools, services.
8.2. Practical Scenarios
Of course, to use such tool, the incident response experts must know how to use them,
and especially what tool fits better in a context. In order to cross this border different sets of
tests have been created. Next we will present five of the most important and most used one,
that have also a great impact on today‟s intelligent transport systems: finding what kind of
encryption is used over a system, identifying what data is leaked from or into our system
computer network, identifying and intercepting eventually live transmissions between an
infected host from our network and its malicious command and control servers over the
Internet, identifying and recognizing fake or mangled signatures keys for public certificate
access and finally, in case of using large transport system deployments with many nodes and
a lot of intermediary wireless communication, identifying the weakest points that can be used
by attacker to gain unauthorized access to the system.
8.2.1. Identification of Encryption Systems

In this scenario the investigator finds a suspect file on the system or computer. This file
represents an encrypted data with a classical encryption system. We need to recognize
encryption system, decrypt the data and find the password hidden in the encrypted file. The
investigator will need this password in solving the second task.
To solve this scenario we will use CrypTool. If we take a look at the encrypted file we
see that this file contains only 26 characters A to Z. If we perform a statistics of these letters
we see that the characters A-Z appears to be random. Thus, we can think that there is a classic
encryption such as substitution (Playfair, Caeser, Vigenère etc.). Using cipher text only attack
on a Vigenère cipher we find the encryption password <<SQUARE>>. If we take a look at
the end of the decrypted file we find the text THEPASSWORDFORTOMORROW
ISSTRONGPASSWORDSAREGOOD. Thus, we conclude that the password we are looking
for is: STRONGPASSWORDSAREGOOD.
8.2.2. Identification of Hidden Data inside other Files

Over the investigated system network have been intercepted some images. One of these
images has a huge size reported to its format. We need to find the data founded in the image
and decrypt it if necessary.
We start by analyzing the intercepted image. It is in gif format but has a large size,
approximately 13 MB. If we investigate this image with a hex editor like UltraEdit we can see
at the end of the file a zip file header (PK). Thus, if we exchange the extension of gif file to
zip file and open it with a supported archiver like WinZip we find an encrypted archive. After
cracking the password, we find three gif files. Opening each file we can see that one of these
one has the content: hollenger.dll. Thus, the malicious file name that the attacker is trying to
access is hollenger.dll.
8.2.3. Identifying a Suspect Communication between Two Computers and Reveal

Stolen Data
By analyzing our system network we find regular traffic between a node and another host
over Internet. The local administrator gives to the investigator a traffic capture between the
node and an outside unknown source. The task is to find what information is stolen.
Using the Wireshark application we are opening the target file. This tool is used for
network traffic analysis or any other general network troubleshooting. Using Follow TCP
stream option we locate within the capture a file found on the source node and dump it into a
file. Opening the dumped file with a hex editor like UltraEdit we see the zip file header (PK);
thus we change the extension of the dumped file into zip and open it. This archive contains a
file and after a visual inspection with UltraEdit we find that it contains the magic header GIF.
Thus, if we exchange the extension of the file into gif and open it with an image viewer we
can see an image which has the content The Root Password is Pengu1nsRock.
8.2.4. Identifying and Recovering Mangled Signature Keys

In this scenario we are given two ECDSA (Elliptic Curve Digital Signature Algorithm)
signatures. After a close inspection we find out that something about them looks strange.
Using the known public key and its public parameters we must find a way to recover the
private key used to generate the signatures.
Before presenting how an investigator can resolve the scenario, we will present briefly
the concept behind ECDSA. Basically, ECDSA is a variant of DSA algorithms that use
elliptic curve cryptography in order to be more reliable. The public parameters are the prime
number p, an elliptic curve E[Fp], a point G E[Fp] with ord(G)=q, q prime number. The
public key VE[Fp] is derived from the signing key 1≤d≤q-1: V=dG. The signature of the
hash h is computed using the ephemeral key k mod q is the pair (r,s)=(xkG mod q,(h+dr)k-1
mod q), where xeG is the first component of the point eGE[Fp]. To verify the signature (r,s)
of the hash h we need to check if xv1G+v2V mod q=r, where v1=hs-1 mod q and v2=rs-1 mod q. It
is essential to have for different signatures (r1,s1) and (r2,s2) different ephemeral keys k1≠k2. If
this two keys are equal then the signatures of the two hashes looks like (r,s1) and (r,s2). Thus,
we can derive s1-s2=k-1(h1-h2) mod q and find the ephemeral key k=(h1-h2)(s1-s2)-1mod q.
Since s1=k-1(h+dr) mod q we derive the private key d=(s1k-h1)r-1 mod q.
The investigator receives three files. The first file contains the hash, in hex codification,
of two messages h1, h2 and their ECDSA signatures (r1,s1) respectively (r2,s2) :
h1=DE37B3145DB7359A0ACC13F0A4AFBD67EB496903
r1=ACB2C1F5898E7578A8A861BDF1CA39E7EF41EAC0B6AAA49468DD70E2
s1=BE4FA99C9D261C5F387A3ACE025702F6FB7884DD07CE18CAD48654B8
h2=28469B02BF0D2CFC86FF43CB612EE8FC05A5DBAA
r2=ACB2C1F5898E7578A8A861BDF1CA39E7EF41EAC0B6AAA49468DD70E2
s2=D3540E2B13E51605F5FEB8C87EE8E176E59213F31EA8B8FFDAD077E2
The second file, parameters.der, contains, in der codification, the public parameters of the
EC. This file can be interpreted using OpenSSL:
opensslecparam -inform DER -in /cygdrive/e/parameters.der -outformPEM -out

/cygdrive/e/parameters.pem
opensslecparam -text -in /cygdrive/e/parameters.pem -noout

Field Type: prime-field
Prime:
00:ff:ff:ff:ff:ff:ff:ff:ff:ff:ff:ff:ff:ff:ff:
ff:ff:ff:ff:ff:ff:ff:ff:ff:fe:ff:ff:e5:6d
A: 0
B: 5 (0x5)
Generator (uncompressed):
04:a1:45:5b:33:4d:f0:99:df:30:fc:28:a1:69:a4:
67:e9:e4:70:75:a9:0f:7e:65:0e:b6:b7:a4:5c:7e:
08:9f:ed:7f:ba:34:42:82:ca:fb:d6:f7:e3:19:f7:
c0:b0:bd:59:e2:ca:4b:db:55:6d:61:a5
Order:
01:00:00:00:00:00:00:00:00:00:00:00:00:00:01:
dc:e8:d2:ec:61:84:ca:f0:a9:71:76:9f:b1:f7
Cofactor: 1 (0x1)
The third file, public.oct, contains the public key:
XV =85CEEE9C98EFDFDFCF64CB522A773F1435D568173677D1D28FC00643
YV =58A105CC1AB1A53D77B278850776E144197F3FA4E27AA676408DFE22
At this point, because the two signatures collide on the first half, we have all the elements
to finalize the investigation. The only thing we need to do is to compute the private key using
the formula:
d=(s1k-h1)r-1 mod q,
where k=(h1-h2)(s1-s2)-1mod q. We can perform these computations using, for example,

MAPLE:
h1:=convert(“DE37B3145DB7359A0ACC13F0A4AFBD67EB496903”,decimal,
hex);
h2:=convert(“28469B02BF0D2CFC86FF43CB612EE8FC05A5DBAA”,decimal,
hex);
r:=convert(“ACB2C1F5898E7578A8A861BDF1CA39E7EF41EAC0B6AAA49468DD7
0E2”,decimal,hex);
s1:=convert(“BE4FA99C9D261C5F387A3ACE025702F6FB7884DD07CE18CAD48654
B8”,decimal,hex);
s2:=convert(“D3540E2B13E51605F5FEB8C87EE8E176E59213F31EA8B8FFDAD077E
2”,decimal,hex);
q:=convert(“010000000000000000000000000001DCE8D2EC6184CAF0A971769FB1F
7”,decimal,hex);
d:=(-h2*s1+h1*s2)*(r*(s1-s2))^(-1) mod q;
convert(d,hex,decimal);
After the compilation of the program we find the private key:
d=8E88B0433C87D1269173487795C81553AD819A1123AE54854B3C0DA7
8.2.5. Identification of Insecure Wireless Connection Points

In this scenario the investigators are trying to find what wireless access points are unsafe
in public use and they are trying to gain access to a private Wi-Fi network that is secured
using the WPA (Wi-Fi Protected Access) protocol. WPA is a security protocol and security
certification program developed by the Wi-Fi Alliance. It is known as the IEEE 802.11i
standard. More exactly a Temporal Key Integrity Protocol (TKIP) is used, that involves using
a dynamic 128 bit key for every packet transmitted. The newest version, WPA2 also includes
Counter Cipher Mode with Block Chaining Message Authentication Code Protocol (CCMP),
a new AES based encryption mode with strong security.
In order to solve this scenario, the investigators will use a series of intercept nodes that
will have a single goal: to continuously scan the entire wireless networks available and try to
crack their password. For this, all the intercept nodes will run two tools called Reaver and
Airmon-ng.
Reaver implements a brute force attack against Wi-Fi Protected Setup (WPS) registrar
PINs in order to recover WPA/WPA2 passphrases. The algorithm behind this tool is
described in detail in http://sviehb.files.wordpress.com/2011/12/viehboeck_wps.pdf. Currently
there is no wide-spread solution for this and it affects even big network vendors like Linksys
and Cisco.WPS is a feature that exists on many modern routers and its intention is to provide
to the users an easy setup process. The problem is that it is tied to a PIN that is hard-coded
into the device.
The first step is to set our node into a monitor mode using the command airmon-ng start
wlan0. After this, we need to find the BSSID of the target node that we want to test. The
BSSID is a unique series of letters and numbers that identifies a target router. We find it by
using the command airodump-ngwlan0. In the list shown inside the terminal we copy the one
for our network. We will assume the following made up BSSID: 8D:AE:9D:65:1F:B2.Now,
with the BSSID and the monitor interface name in hand we have everything we need to
startup Reaver. Inside a terminal we issue the command reaver –imon0 –b
8D:AE:9D:65:1F:B2 –vv After this, Reaver will try a series of PINs on the router in a brute
force attack, one after another. After cracking is complete we will have the output of the total
time needed to crack the password, the router PIN and its password.
9. PRACTICAL ISSUES AND EXPERIENCES IN DIFFERENT COUNTRIES

As we have seen until now, the field of critical infrastructure protection, including the
most important of them - water, gas, and electricity providers, and especially the area of
intelligent transport systems, is very important for a country, mainly for those that have
strong industries and big population. As we can see in national reports such as those from
United Kingdom [27] or Germany [28 - 29] these infrastructure represent important assets.
On one side we have the technological requirements and on the other side we have the
public opinion in case of cybercrime, such as banking fraud, identity and content theft. A
warning signal was triggered with the discovery of Stuxnet malware as it offered the level of
progress made by the cybercriminals. We can say for sure that every country in this world
must pay attention to all its infrastructures and “do something” about cyber security.
For a better image we have briefly presented the situation existing in three different
countries around the world. They have been chosen because they are representative for their
power and importance.
9.1. United Kingdom
In the United Kingdom we can find the National Security Strategy (NSS) and the
Strategic Defense and Security Review (SDSR) documents that specify the need for critical
infrastructures protection. These documents were released in 2010, and we can see how in the
late years this problem has become more and more important. This is also supported by the
fact that their annual budget for cyber security is around 650 million pounds.
In charge with applying these documents is the Center for Protection of Critical National
Infrastructure (CPCNI) and since their publishing, it has implemented many incident response
services, such as energy, economies, health, etc.
9.2. Germany
In Germany we can see a good example of a secure critical infrastructure deployment.

They have all their important infrastructures connected through a full mesh network. This
means that any failure is covered thoroughly. As examples we can give health, industries,
water supplies, IT and food supplies.
This implementation was possible thanks to the mature legal environment existing in this
country that threats security as a vital element. For example, we can analyze their power
distribution infrastructures; both public and private companies are obligated by the law to
have an internal failure and recovery system.
Of course all these implementation are monitored constantly by different enforcers, such
as “Energiewirtschaftsgesetz” – the Act on the Supply of Electricity and Gas and the
“Bundesnetzagentur” – the Federal Network Agency in the field of smart grids and power
distribution. Telecom services also have their own procedures and laws regarding protection
from network penetration, unknown call interception and any other form of unauthorized
access.
9.3. USA
In the USA, a lot of effort has been put in order to secure their critical infrastructures
since the terrorist attacks from 11th September 2001. Since then, an unauthorized access to
this kind of infrastructure was never spotted.
In charge with this field is the Department of Homeland Security (DHS). It has the means
to coordinate the cyber activity to all these systems and it also can use the help from the other
governmental agencies in case something happens.
The entire plan can be found in the USA “Critical Infrastructure Protection Plan” which
enforces each agency with a specific task. For them, these infrastructures represent a top
priority.
CONCLUSION
As we can see from this chapter the problem of critical infrastructures represents a
current and menacing threat, which involves strong knowledge of computer communications
techniques, secure programming techniques, algorithm and software implementation analysis,
cryptography, steganography, probability and finally applied mathematics. The process of
high level of assurance of cyber security must take into account all the above-specified
domains.
Fast advances in cybercrime technology and techniques have resulted since the beginning
of 2012 in an unprecedented rise in data breaches. We think that planning to ensure that our
critical infrastructures and intelligent transport systems are trustworthy and secure we need to
consider the fundamental changes that are occurring in the cyberspace and try to adapt to
them. In our opinion, looking forward into the future of more than 3 billion Internet users
existing today we can see four big directions for resolving the cyber security issues: online
users security education, cryptography, online data obfuscation and cloud services
transparency and security.
REFERENCES
[1] Hough, P.; Understanding Global Security. Routledge, 2009.
[2] www.us-cert.gov/.
[3] www.expresscomputeronline.com/index.php/sections/events/796-critical-infrastructure-
protecting-the-heartbeat-of-the-nation.
[4] Critical Infrastructures: What Makes an Infrastructure Critical?. Report for Congress,
2003, http://fpc.state.gov/documents/organization/13839.pdf.
[5] Kramer, F.D.; Starr, S.H.; Wentz, L.; Cyberpower and National Security. Potomac
Books Inc., 2009.
[6] Cybersecurity for Critical Infrastructure Protection. United States General Accounting
Office, www.gao.gov/new.items/d04321.pdf.
[7] Massoud Amin, S.; Wollenberg, B.F.; Toward a Smart Grid. IEEE P&E Magazine,
2005.
[8] U.S. Infrastructure: Smart Grid, Renewing America. Council on Foreign Relations, 16.
[9] www.americainfra.com/article/Smart-grid-security/.
[10] www.zerodayinitiative.com/.
[11] www.crysys.hu/skywiper/skywiper.pdf.
[12] www.certcc.ir/index.php?name=news&file=article&sid=1894&newlang=eng.
[13] www.securelist.com/en/blog/208193522/The_Flame_Questions_and_Answers.
[14] www.bbc.co.uk/news/technology-18365844.
[15] www.codeproject.com/Articles/246545/Stuxnet-Malware-Analysis-Paper.
[16] Zhang, L.; Levinson, D.; Investing for Reliability and Security in Transportation
Networks. 2008.
[17] Checkoway, S.; McCoy, D.; Kantor, B.; Comprehensive Experimental Analyses of
Automotive Attack Surface, Autosec, 2011.
[18] Schulze, S.; Pukall, M.; Hoppe, T.; ITSecurity in Automotive Software Development.GI
Softwaretechnik-Trends, 2009.
[19] Clayton, M; Scientists Hack into Car Computers - Control Brakes Engine, The
Christian Science Monitor, www.csmonitor.com/USA/2010/0813/Scientists-hack-into-
cars-computers-control-brakes-engine.
[20] Goretsky, A.; Researcher at ESET:A Technology Security Company. Personal
correspondence, 2011.
[21] http://auto.howstuffworks.com/car-computer-virus.htm.
[22] A Statistical Test Suite for Random and Pseudorandom Number Generators for
Cryptographic Applications. 800-22, NIST Special Publication, 2010.
[23] Knuth, D; The Art of Computer Programming, Seminumerical Algorithms.3rd Ed., vol.
2, Addison Wesley, Reading, Massachusetts, 1998.
[24] Marsaglia, G.; DIEHARD Statistical Tests. http://stst.fsu.edu/geo/diehard.html.
[25] Engebretson, P.; The Basics of Hacking and Penetration Testing. Syngress, 2012.
[26] www.backtrack-linux.org/.
[27] www.chathamhouse.org/sites/default/files/public/Research/International%
20Security/r0911cyber.pdf.
[28] www.bmi.bund.de/SharedDocs/Downloads/EN/Broschueren/cip_stategy.pdf?__
blob=publicationFile.
[29] ww.bsi.bund.de/SharedDocs/Downloads/EN/BSI/Kritis/paper_studie_en_pdf.
pdf?__blob=publicationFile.
[30] www.pcworld.idg.com.au/article/360645/siemens_stuxnet_worm_hit_industrial_
systems/.
[31] Hills, R.; Senior Education Program Manager at Universal Technical Institute.
Personal Correspondence,2011.
[32] Mills, E.; Hacking a Car (Q&A).CNET, http://news.cnet.com/8301-27080_3-
20005047-245.html.
[33] Mollien, C.; Information and Communication Technology Strategist at Basic Blue.
Personal Correspondence, September 14, 2011.
[34] Petraglia, D.; Director of Forensic and Information Security Services at Chartstone
Consulting. Personal Correspondence, 2011.
[35] Shaer, M.; Disgruntled Hacker Remotely Disables 100 Cars. The Christian Science
Monitor, www.csmonitor.com/Innovation/Horizons/2010/0318/Disgruntled-hacker-
remotely-disables-100-cars.
[36] www.cfr.org/united-states/us-infrastructure-smart-grid-security/p26842.
[37] http://wwwiti.cs.uni-magdeburg.de/~sanschul/papers/GI-WS_regensburg2009.pdf.
[38] www.autosec.org/pubs/cars-usenixsec2011.pdf.
[39] http://nastfenews.org/index.php?option=com_content&view=article&id=107:making-
automotive-software-safe&catid=27:features&Itemid=49.
[40] www.isticom.it/documenti/news/pub_003_eng.pdf.
Chapter 7
SUSTAINABLE INFORMATION SYSTEMS

MANAGEMENT FOR SPATIAL GOVERNANCE
Angela Ionita*1, Marcel Foca†2 and Maria Visan‡2

1
Research Institute for Artificial Intelligence,
“MihaiDrăgănescu” of Romanian Academy, Romania
2
Intergraph Computer Services S.R.L.
ABSTRACT
This chapter has the intention to discuss the complexity of sustainable information
systems management in the framework of spatial governance. There is a diversity of
motivations to introduce sustainability into Information Systems based on Information
Technology and Communication. The discussion about sustainability and Information
Systems based on Information Technology and Communication innovation appears in
different contexts. Based on the current literature, this article is a synthesis and proposes
a reorientation of the way the concept of sustainability is dealt with, positioning spatial
governance based on human behavior and the processing of spatial knowledge at the
centre of the concept. The sustainability approach presented in this chapter is then
translated into requirements needed for a Sustainable Information Systems Management
for Spatial Governance focusing on (Cooperative) Intelligent Transportation Systems,
Spatial Data Infrastructure and emphasizing Spatial Governance.
Keywords: Sustainability, Spatial Governance, Intelligent Transportation Systems,

Cooperative Intelligent Transportation Systems, Spatial Data Infrastructure
*
E-mail: aionita@racai.ro.
†
E-mail: office@ingr.ro.
‡
E-mail: office@ingr.ro.
208 Angela Ionita, Marcel Foca and Maria Visan
1. INTRODUCTION
“Moving forward with sustainability is no more a subject matter of intellectual
discussion. Instead it has become a harsh reality in evolution of human civilization. However,
scope and definition of sustainability has constantly expanded manifold to include essential
utilities and requirements of day to day life. Inherent consumerism of growth/development
has constantly evolved luxuries into commodities causing incremental consumption per
capita” [1]. Sustainability is in many cases, but not correctly, associated with the
environmental aspects only, while corporate social responsibility is more often perceived as
primarily social-oriented. Additional confusion is created by the various social and
environmental-focused initiatives related to climate change or fair trade. Current literature [2,
3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24) acknowledges that
sustainability is a broad, complex concept [25] involving environmental as well as social
issues, and which requires continuous learning in order to be understood and tackled.
However all of them are about vision for the future.
Rio+20, a joint project of the entire United Nations System, is engaging all constituencies
and programs of the global governance body to encourage governments to adopt measures for
implementing sustainable practices developed over the past 20 years [26]. The global
economy and economic progress paradigm as in the late 20th century was completely
revolutionized with the development of the personal computer and rise of the Internet.
Analytics and metric tools enabled by Information Technology and Communications (IT&C)
now allows for better and more efficient reporting for all sorts of fields - including sustainable
development. As Daniel Kent mentioned [26]: “by investing further in information and
environmental technologies, it is possible for some countries to become exporters of clean
technologies and thus turn the environment into a new engine for competitiveness and
growth, rather than a cost as currently perceived. Embedding environmental perspectives
into social and economic policy requires a more holistic and cross-sectoral way of working,
strengthening local to global interactions and improving scientific understanding. In this
respect, greater transparency about assumptions, shared access to data and better
communication of scientific findings will become critical to gathering and retaining the
support of society and policy-makers”.
The discussion about sustainability and Information Systems (IS) based on Information
Technology and Communications (IT&C) innovation appears in different contexts. For
instance, models and tools have been developed to assess corporate sustainability [27] and
sustainability of Management Information Systems [28]). The notion of sustainability of IS
based on IT&C innovation (IS IT&C) stems from the broader notion of Sustainable
Development (SD) and is applied in a specific domain. Also, contributions concerning
sustainability and IS IT&C originate from joining domains, sustainability and systems [29,
30].
2. SPATIAL GOVERNANCE AS AN ASPECT OF GOVERNANCE

'Governance' relates to human organization of various kinds: government and non-
government, commercial and non-commercial, private and public, formal and informal. It is
Sustainable Information Systems Management for Spatial Governance 209
thus a broader term than 'government' and deals with a variety of more general human social
interactions and their social, economic and ecological outcomes. In generally sense, spatial
governance (SG) is an aspect of governance. Spatial governance involves modern ideas about
governance; but it focuses more particularly on a wide variety of situations where knowledge
and mutual understandings about both time and place are important to successful social,
economic and ecological outcomes (http://spatialgovernance.com/).The Spatial Data
Infrastructures (SDI) could be considered as kernel of Spatial Governance. An analysis of
SDI in Governance shows that the idea of SG has a wide variety of practical applications
starting with the built environment, ways of creating enabling scenarios for making timely
and information-driven decisions to solve existing or identified spatial problems. These
aspects are important to the ownership, occupation, government and administration of
territory in all its forms; as well as the activities that occur in a particular territory.
2.1. Multiple Facets of Spatial Data Infrastructures
There are many definitions of spatial data infrastructure (SDI) [31, 32, 33, 34) and many
critical discussions about existing definitions and the complexity of the SDI concept,
perspectives [35, 36, 37, 38, 39].
In this chapter, we accepted a more comprehensive and easy to understand definition of
the International Society for Photogrammetry and Remote Sensing (ISPRS) Commission IV -
Geodatabases and Digital Mapping, 2008-2012: “SDI is generally defined as the
technologies, policies, and people necessary to promote sharing of geospatial data
throughout all levels of government, the private and non-profit sectors, and the academic
community. Goal of SDI is to reduce duplication of effort among agencies, improve quality
and reduce costs related to geographic information, to make geographic data more
accessible to the public, to increase the benefits of using available data, and to establish key
partnerships with states, counties, cities, tribal nations, academia and the private sector to
increase data availability”.[40]
2.2. Spatial Information and Sustainable Development
Hadley [41] notes the often-quoted statistic that 80% of data used by government is
spatial and that the proportion used for planning could well be even higher. Information on
resources, (e.g. ecosystems and environment, human resources, linguistic resources, cultural
and economic aspects etc.) becomes imperative for decision-making in SG. The
implementation of Sustainable Development strategies should be based on scientific policy
making, which demands great deal of real-time information. It is one essential step to obtain
real-time information and to use IT&C in order to build IS IT&C for Sustainable
Development. The nature of the built environment requires an IS capable of handling both the
descriptive characteristics (attributes) and the spatial component of this environment. This
important feature is what an IS for management spatial information offers. Such system
implemented by IT&C specialists for the acquisition and verification, storage, updating,
management and exchange, manipulation, retrieval and representation, analysis of geospatial
data [42, 43] and using the expertise of specialized people, stressed the relevance of spatial
information in decision making process when defined it [44, 45, 46]. Considered as superset
of Geographic Information Systems (GIS) formed by the intersection of GIS and other
techniques, the Spatial Decision Support Systems (SDSS) which contain specific decision
models, and using the general spatial capabilities of the GIS [47], becomes a more relevant
domain to the generation of more conducive decision making environment [48, 49, 50] for the
problems of SG.
3. TRANSPORTATION SYSTEMS AND SUSTAINABLE DEVELOPMENT

“Every problem has an „environment,‟ to which it is inextricably united. If you stop x
from growing (or declining), you will make other things grow (or decline), and these changes
you have created may very well be as serious, and as disastrous, as the growth of x.”
(Churchman‟s classic „environmental fallacy [51]).
Each country recognizes that transportation is an important tool to help meet overall
sustainability objectives. The development of new technologies and its introduction in
everyday life has been a regular scenario in several areas including also transportation. The
introduction and development of new technology for the transportation field has been
occurring with different rhythms, being expected that the future lead unconditionally to the
presence of such systems in all transportation areas. Intelligent Transportation Systems (ITS),
whose introduction has been bringing several benefits to the transportation field and, more
specifically to transports users, are the most visible consequence of this evolution. Attributes
of sustainable transport follow from the expanded definition of sustainable development -
sustainable transport is safe, high quality, and accessible to all; ecologically sound;
economical; and a positive contributor to regional development.
According to European Roadmap – Infrastructure for Green Vehicles (2012) “a
sustainable transportation system, or to say sustainable mobility, is one that:
 Allows the basic access and development needs of individuals, companies and
society to be met safely and in a manner consistent with human and ecosystem
health, and promotes equity within and between successive generations.
 Is affordable, operates fairly and efficiently, offers a choice of transport mode, and
supports a competitive economy, as well as balanced regional development.
 Limits emissions and waste within the planet‟s ability to absorb them, uses
renewable resources at or below their rates of generation, and uses non-renewable
resources at or below the rates of development of renewable substitutes, while
minimizing the impact of the use of land and the generation of noise. In the context
of sustainable mobility, with its individual mobility, public and goods transport, the
pertinent infrastructure technology carry an important role to overcome the problems
of environmental impact, air quality and sustainable mobility.” [52].
Specific goals for sustainable transport may include improved service quality and quality
of access to goods and services, safety, improved air quality, noise reduction, improved water
quality, protection of habitat and open space, historic preservation, reduced carbon emissions,
increased social equity, economic development, and a satisfying quality of life, as well as
local goals consistent with the overall objective.
One of the main ideas of this chapter is that “to be successful, sustainable transportation
policy must avoid the common transportation pitfall of ignoring the larger systems in which
transport is embedded” [53]. As mentioned also in [53] a fundamental characteristic of the
transportation sector that continually frustrates policy development is “the porous nature of
the transport system itself. Transportation is not a closed, self-contained system; rather, it is
tightly intertwined with other systems”.
3.1. Intelligent Transportation Systems
A first report, made by ITS America and published in 1994, tried to make a comparison
between the Intelligent Transportation Systems (ITS) developed in Japan, USA and Europe.
In [54] considered to be a very important base to build a common idea of ITS, is considered
as opened way to the development of a second work, published in 1997 [55]. In this more
recent report an international comparison between ITS and the related activities was made as
an effort to collect information and gather consensus and in [54] suggested that categories for
ITS technology need to be identified. Besides the classification in each country may not
exactly be the same, they presented a table (Table 1) that indicates ITS technologies utilized
for the three mentioned regions: Europe, Japan and America. Beside this observed division in
the ITS subsystems, in order to accomplish for their work the authors choose to focus on the
seven major categories of ITS commonly present in the three studied areas: ATMS, ATIS,
AVCSS, CVO, APTS, EPS &ETC, and EMS.
Table 1. Major ITS technology categories in the US, Europe and Japan. Based in data
collected from ([56], ITS Japan and ITS America [54])
Region
Designation
USA Europe Japan
ATMS Advanced Traffic Management Systems   
ATIS Advanced Traveller Information Systems   
AVCSS Advanced Vehicle Control and Safety Systems   
CVO Commercial Vehicle Operations   
APTS Advanced Public Transportation Systems   
EPS&ETC Electronic Payment System and Electronic Toll Collection   
EMS Emergency Management Systems   
Electronic Navigation System 
Road and Safety Management System 
Information Management  
Maintenance and Establishment Management 
Now talking about at least two dominant concepts in this research area: Intelligent
Transportation Systems (ITS) and Cooperative Intelligent Transport Systems and Services
(C-ITS)[56].
According to European Telecommunications Standards Institute (ETSI): “Intelligent
Transport Systems (ITS) add information and communications technology to transport
infrastructures and vehicles in an effort to improve their safety, reliability, efficiency and
quality. ITS services are also designed to optimize transportation times and fuel consumption
thus providing greener and safer transportation. However, the deployment of Intelligent
Transport Systems and provision of corresponding services are not limited to road transport
sector only, but includes other domains such as railways, aviation and maritime as well.”
[57]
Intelligent Transportation Systems (ITS) are a vision of a future that integrates existing
transportation infrastructure with communication networks and spatial data infrastructures in
an effort to reduce congestion and travel time. In doing so on a mass scale, the larger effect of
ITS is to limit the release of carbon emissions into the atmosphere, cut back on fuel
consumption and improve road safety.
In [58] has been identified five trends as most relevant for transportation needs:
urbanization; demographic developments; individualization; informatization; globalization.
And, as mentioned in [59], applying IT&C, ITS delivers five classes of benefits by:
increasing safety; improving operational performance, particularly by reducing congestion;
enhancing mobility and convenience; delivering environmental benefits and boosting
productivity and expanding economic and employment growth.
3.2. Cooperative Intelligent Transport Systems and Services
As mentioned in [56], modern vehicles equipped with driver assistance systems can
“feel” (by sensors), “see” (by cameras) and – in future – “speak” (by communication
systems). The new technology of C-ITS enables communication between vehicles and traffic
infrastructure (Figure 1). It is based on the principle that cooperative parties (ITS stations, i.e.
in vehicles, road side units) exchange information among each other in terms of standardized
message sets. Deployment of C-ITS in Europe based on ITS G5 technology beginning in
2015. C-ITS applications are manifold and may not all be implemented at the same time.
Instead, a phase model will be applied.
The ITS world is vast and includes a wide variety of sub-taxonomies and definitions that
are currently utilized to describe intelligent systems. A considerable variety of equipment‟s
are exclusively dedicated to the road sector and much of them are directly related with single
vehicles. A specific cluster of systems have received a great deal of attention from
psychologists and human factor professionals that study road safety [60]. This cluster is
characterized by on-board systems that can, in a direct or indirect manner, influence the
driving task and the driving activity. Two main categories are usually distinguished according
to their impact on driving: In-Vehicle Information Systems (IVIS) and Advanced Driver
Assistance Systems (ADAS). This classification has been used in some European scientific
projects that aimed to study the implications of these new technological systems, like for
example the CONSENSUS [61], AIDE [62], HASTE [63], COST Action 352 [64] and
HUMANIST [65].).
The In-Vehicle Information Systems “provide drivers with information or communication
while the vehicle is in motion. The delivered information may not be related to driving (e.g.
conversing by phone, consulting emails, or listening to the radio) or related to the trip
management (e.g. traffic or weather information, navigation map, or route guidance).” The
Advanced Driver Assistance Systems “are designed to support the driving task on vehicle
manoeuvring by informing, warning and actively assisting drivers on the basis of vehicle
surroundings analysis and with the requirement of immediate driver actions (e.g. intelligent
speed adaptation, lateral and longitudinal control, blind spot warning, or night vision)” [66].
Figure 1. Mind map of the most important facts on cooperative Intelligent Transport Systems and
Services (C-ITS) (Source: ITS Niedersachsen cited in [56])
3.2.1. Interaction with Navigation/Guidance Systems

In the last decades, with the introduction of new technologies, several systems have been
developed with the main aim of present to drivers‟ useful information for their navigation.
Once navigation can be described as: “to achieve movement through a space” [67] these
systems can help drivers to travel throughout unfamiliar places and avoid making deviations
from the wanted course [68, 69].
To accomplish this task, systems can transmit guidance instructions in strategic decision
points [70] so drivers can make a more efficient use of the road network (Jeffrey as cited in
[69]). In a broader way, the advantages obtained by their use include preventing drivers to get
lost, and if traffic information is available, prevent them to travel through congested roads
[71]. As a consequence to the presentation of other route options, these systems give the
possibility of more equal distributions of the traffic through the transportation network,
helping to solve quite a lot of congestion problems in several cities around the world [72].
Additionally, they can also contribute to a decrease in the consumption of fuel and to a
decline in the release of pollutant gases ([73, 72, Jeffrey as cited in [69]). The use of these
equipments can also augment the confidence and security of road users, once they can receive
useful instructions to avoid or outline dangerous situations [73, 74, 71, 69].
In order to make the navigation possible, based on local SDI, these systems provide
updated information about the place where the vehicle is at, offering also guidance
instructions to select the most adequate roads. The information supplied can have different
features, and be transmitted in different modes in order to guide the driver to the destination.
Some of these systems are also capable of transmitting on-line information about traffic jam,
to calculate the time needed to complete the course, or notify about incidents or other kinds of
malfunctions on the road [75].
Some years ago, navigation systems could be characterized as being an electronic help
that offered the driver several kinds of information upon request [76]. With the development
of equipments that displayed a map which could give current location of the vehicle as well as
the destination position; or a display that used symbols or voice instructions and also the
increased variety of systems available in the market, these two groups of systems received
different denominations: Navigation Systems and Route Guidance Systems.
Navigation Systems are considered as passive systems due to the fact that they do not
provide route guidance information, offering only “geomatic information” [77]. They usually
display a digital map on a screen that might highlight a route, the current position of a vehicle
and the destination point [78]. With this information, drivers are expected to actively make
their own navigation decisions at each intersection in order to get to the destination [79].
On the other hand, Route Guidance Systems are able to calculate the best route to follow,
based on specific criteria. The suggestion of a course is given through route guidance
information, being these messages considered as important and useful to guide the driver [78,
77, 79].
On the other hand, Route Guidance Systems are able to calculate the best route to follow,
based on specific criteria. The suggestion of a course is given through specific instructions,
also named as route guidance information, being these messages considered as important and
useful to guide the driver [78, 77, 79].
Besides this systems‟ classification based on the type of the information transmitted, they
can also be classified in a different way [60]: those fitted as standard to the vehicle and the
nomadic devices. Devices fitted as standard to the vehicle are originally installed in the car,
meaning that when the driver buys the car the navigation systems is already initialized. There
are also some options that allow drivers to include a system afterwards; however theses
aftermarket navigation devices are professionally installed and can only be used inside the
vehicle.
Contrarily, nomadic systems are generally introduced inside vehicles by their own drivers
and one of the characteristics that distinguish this type of system is the fact that they can be
used for besides the driving context [80, 81]. They can be activated in different situations and
with different aims because these equipments can include also mobile phone functions, access
to the internet, and electronic agenda, among several other utilities. Concerning the man-
machine interaction in a road environment, nomadic systems are similar to systems fitted as
standard to the vehicle as they allow the transmission of audio and written messages that can
help drivers guide themselves through the road network [80]. However, the way they are
positioned and attached inside the vehicle can be considerably different. Unlike the fitted as
standard systems, nomadic devices are placed according to drivers‟ wish. This aspect can
foresee two problems: the first is related with the proper positioning of the system and
consequently with the efficiency on the acquisition of messages; the second is a safety
concern as in case of accident the system can be loose and hurt the driver [80]. Problems
related to the single, isolated automotive vehicle and its subsystems are challenging enough
(see the grand challenge on Advanced Driver Assistance Systems), but the research
community is also exploring the “big picture” of intelligent road transportation - the system,
or system of systems, consisting of many vehicles and their drivers interacting on roads. Two
related topics are included in this vision: Vehicle-to-infrastructure (V2I) interaction and
Vehicle-to-vehicle (V2V) interaction.
3.2.2. Vehicle to Infrastructure Communication

In Vehicle-to-infrastructure (V2I) concept, the infrastructure plays a coordination role by
gathering global or local information on traffic and road conditions and then suggesting or
imposing certain behaviors on a group of vehicles. One example is ramp metering, already
widely used, which requires limited sensors and actuators (measurements of traffic density on
a highway and traffic lights on ramps).
In a more sophisticated scenario, the velocities and accelerations of vehicles and
intervehicle distances would be suggested by the infrastructure on the basis of traffic
conditions, with the goal of optimizing overall emissions, fuel consumption, and traffic
velocities. Suggestions to vehicles could be broadcast to drivers via road displays or directly
to vehicles via wireless connections. Looking further ahead, in some cases suggestions could
be integrated into the vehicle controls and implemented semi automatically (always taking
onto account the restrictions on automatic vehicle driving imposed by the Vienna Convention
on Road Traffic [82]).
V2I promise revolutionary improvements in transportation - greater energy efficiency,
less road construction, reduced collisions, and safety of vehicle occupants as well as
pedestrians and bicyclists. Control is a key contributing discipline for both topics. Some
experts [83] predict that the first V2I systems may be developed and deployed in the 2015-
2020 time frame.
3.2.3. Vehicle-to-Vehicle
In the Vehicle-to-Vehicle (V2V) concept, when two or more vehicles or roadside stations
are in radio communication range, they connect automatically and establish an ad hoc
network enabling the sharing of position, speed, and direction data. Every vehicle is also a
router and allows sending messages over multihub to more distant vehicles and roadside
stations. The routing algorithm is based on the position of the vehicles and is able to handle
fast changes of the network topology. Control technology comes into play at local and higher
layers of the architecture. Uncertainties, delays, partial measurements, safety and performance
objectives, and other aspects must be considered, and the system must be capable of making
automatic or semiautomatic decisions, providing warnings/information and potentially
effecting actions.
A lot of infrastructure technologies and processes are already known, some already
developed, interesting and promising ideas appear and research initiatives have been
undertaken. The smart electricity grid, charging devices in car parks, filling stations for
hydrogen, gas and renewable fuels, logistics and services, information systems for users,
traffic management are only some infrastructure needs for new research, development and
implementation.
As mentioned in [84] there are currently many initiatives across the globe in V2V and V2I
communications. Though performed by different groups with different objectives, the overall
goal is the same: to provide drivers with real-time information beyond what is available
through the senses alone. This allows for better decision making, which enhances safety and
driving efficiency. Between these initiatives has been enumerated: FRICTI@N [85],GST[86],
PReVENT [87], VSC[88], Centrico [89], COMeSafety [90], CVIS[91], eCall [92],
COOPERS [93], NoW [94], Staufreies Hessen 2015[88], INFONEBBIA “Safety in the
Fog”[95] etc.
Within SmartCities and Communities [96] has been mentioned some initiatives such as
listed:
 in [97] - projects addressing the problems of ITS, such as: in Germany [98], on 6
August 2012 has been proposed a Pilot Project regarding the improvement of road
safety and efficiency and a Pilot Project concerning the enhancement of application
of C-ITS; in The Netherlands, on 20 September 2012 has been proposed a Pilot
Project towards a Smart Mobility Roadmap and on 24 September 2012, Schneider
Electric proposed a Pilot Project addressing V2I on-board information and a Pilot
Project for City Congestion Management;
 in [99] - Schneider Electric proposed on 24 September 2012 a Pilot Project
concerning the data exchange between existing systems (multiagency interaction);
EIGSI proposed on 1 October 2012 a Pilot Project addressing visibility and
accessibility of medium sized cities and in 6 October 2012 launch an idea regarding
the integration city/logistics in global transport system; GMV proposed on 21
December 2012 a Pilot Project addressing efficient and sustainable mobility.
 In [100] it is presented an initiative undertaken by the U.S. Department of
Transportation (DOT) to provide an infrastructure where vehicles can identify threats
and hazards on the roadway and communicate this information over wireless network
to alert and warn drivers, named Vehicle Infrastructure Integration (VII). This
initiative, recently renamed IntelliDrive, has significantly impact DOT operations
and should be monitored on a continual basis.
In summary, for ensuring of a sustainable development of the society, what intelligent

transportation systems do is empowers actors in the transportation system – from commuters,
to highway and transit network operators – with actionable information (or, intelligence) to
make better-informed decisions, whether it‟s choosing which route to take; when to travel;
whether to mode-shift (take mass transit instead of driving); how to optimize traffic signals;
where to build new roadways; what the true cost of roadways are and how best to price their
use; or how to hold providers of transportation services accountable for results. “The big
opportunity at hand is to bring information to bear on transportation networks, transforming
them into truly intelligent transportation systems.” [59]
4. SUSTAINABLE INFORMATION SYSTEMS MANAGEMENT

The increasing dissemination of IS IT&C into all areas of business and personal life has
drawn attention to its economic, environmental and social effects. The purpose of sustainable
IS IT&C management is to contribute to the vision, the overall strategy and the strategic
objectives of the society as whole, which are not only economic, but also ecological and
social [101]. These objectives do not only apply to information as a resource but also all other
resources identified by the resource-based view. A sustainable IS IT&C management covers
not only the economical, but also the ecological years [101] and social perspective, extending
traditional concepts of IS management. The social dimension of sustainable IS management
includes topics of IT&C utilization by individuals and organizations, but also comprises the
challenges of generating, preserving and safeguarding knowledge, information and data.
Sustainability is sometimes compared with the internet revolution, given its wide-ranging
impact on businesses.
According to [102], a Sustainable Information System (SIS) is an Information System
which: adapts to its environment, involves relevant stakeholders, and supports the knowledge
lifecycle, i.e. knowledge creation, knowledge evaluation and knowledge integration/
application.
The new generation of SIS will not be a single system but “multiples connected
infrastructures based on multiple technological platforms and on open access and will
require new governance models” [103].
Starting from the above considerations, in our opinion, the development of Sustainable
Information Systems Management for Spatial Governance (SISMSG) requires the integration
of various IT&C tools, models, formats of data and methodologies. The selection of IT&C
tools and the merging of methodologies should result in specifications of domain-specific
software architecture, suitable for development a particular application domain. The SISMSG
development requires taking into consideration of cultural and political factors in addition to
technological components, comprehensive studies by multidisciplinary terms of researchers
with diverse cultural background and domain expertise.
4.1. Requirements for SISMSG
“Today, we live in a world that is dramatically transformed with technological advances.

The immediate opportunity is to integrate the data residing in disparate places and create
value out of it. The convergence of various stakeholders including users, technology
providers and the industry will enable us to meet the needs of the future” (KK Singh in [1]).
In the generally sense, the IS IT&C consists of stakeholders, process and equipment
(hardware and software) to collect, analyze and proliferate information across relevant
decision makers (http://www.managementstudyguide.com/measuring-market-demand.htm).
Starting from SIS‟s accepted definition [103] and from the article elaborated by
Jakobsson and Tsoulos [104], we try to exemplify the evaluation of SISMSG by using an
architecture that seems suitable to support a multi-layered approach, underling the benefit to
users, benefit to society and the gain (Figure 2). Based on our experience, we would like to
discuss some relevant characteristics of a SISMSG to constitute a base of evaluation.
All of our considerations start from the fact that must to be a demand for any SISMSG
which comes from the society, from the users and/or stakeholders in order to provide a
harmonized content and of good quality as support for decision making in spatial
governance.
Demand. As an economic principle that describes a consumer‟s desire and willingness to
pay a price for a specific good or service, the demand for spatial information systems has
been growing sharply in recent years, especially with the proliferation of SDI at different
levels and with entering SG in political and economical areas.
Figure 2. Evaluation of SISMSG.
The resources, which are playing an important role in this context, are primarily the infra-
structure(s), the internal and external relationships, and the market/demand responsiveness.
For this reason the model of sustainable IS IT&C management has the ability to illustrate
complex relationships and effects between stakeholders, objectives and measures. High
demand is also a prerequisite for building a stable and significant user base, which definitely
would help to sustain a system over time.
4.2. Resources for SISMSG
In order to satisfy the existing and future demand for SISMSG and having Spatial
Governance as goal, the resources are represented by spatial data, spatial knowledge,
geospatial information and human resources (HR) represented by specialized people and/or
simple citizens.
Spatial Data also known as geospatial data (GD) or geographic information (GI) it is
the data that identifies the geographic location of features and boundaries on Earth, such as
natural or constructed features, oceans, and more. Geospatial Data is usually stored as
coordinates and topology, and is data that can be mapped. Geospatial Data is often accessed,
manipulated or analyzed through GIS.
Spatial Knowledge. Creating geospatial knowledge (GK), as mentioned in [105], is an
effortful process the analyst undertakes; it is a cognitive intellectual endeavour that arrives at
a conclusion through geospatial reasoning. The notion of “geospatial reasoning” is the core of
geospatial knowledge creation. Geospatial reasoning is the cognitive process of integrating
spatial evidence and geographic problem representations in the process of making judgment
about space and spatial relationships. Evidence is geospatial information helpful in forming a
conclusion or judgment. Problem representation is what we know about the structure in the
system and its properties. “There is a need for greater collaboration and association between
industry and government to develop a global platform for sharing geospatial knowledge” [1]
and GIS is used as a knowledge management platform to create, disseminate, renew and
manage explicit (codified) and tacit (experiential) knowledge toward organisational
sustenance and survival. Also is used to collaboratively integrate multiple forms of GK to
develop meaningful solutions to complex environmental and natural resources management
issues, challenges and opportunities.
Geospatial Information. “Geospatial information is becoming more and more central to
many capabilities … The nature of geospatial is a combination of three different technologies
– sensors that capture data, software that turns data into information and the wireless
communication that enables that information to be used everywhere in relatively seamless
manner” (Steven Berklund in [1]).
The terms Geospatial Data (GD), Geospatial Knowledge (GK) and Geospatial
Information are difficult to define, but cannot be used interchangeably. Geospatial
information (GI) provides the basic framework for space visualization. It is information
produced by multiple sources to common interoperable data standards. It may be presented in
the form of printed maps, charts, and publications; in digital simulation and modeling
databases; in photographic form; or in the form of digitized maps and charts or attributed
centerline data. Geospatial services include tools that enable users to access and manipulate
data, and also include instruction, training, laboratory support, and guidance for the use. It
should be noted that the definitions and usage of the terms GD, GK and GI are not used
consistently or unambiguously further exacerbating the situation. Geospatial Data (GD) can
(usually) be applied to the output of a collector or collection system before it is processed,
i.e., data that was sensed. GI is geospatial data that has been processed or had value added to
it by some human or machine process. Geospatial Knowledge (GK) is a structuring of
Geospatial Information (GI), usually with some interpretation or analysis.
4.3. The Characteristics of a SISMSG
In order to evaluate the really gain, in our opinion, for the characteristics of a SISMSG it
is necessary to pay attention at least to two points of view: of the users and of the society.
a) From the point of view of users, we have to pay attention to: longevity, accessibility,
responsiveness, and robustness.
Longevity. Given the amount and the complexity of investments in the collection,
storage, handling and visualization of the resources of SISMSGS mentioned above, one of the
important characteristic of a SISMSG is that it is to be able to survive over time. Digital
longevity is an area of growing concern [106]. It is widely accepted that IT&C is
revolutionizing the concepts of documents and records – in general sense - in an upheaval at
least as great as the introduction of printing, if not of writing and/or designing itself. The
current generation of digital records therefore has unique historical significance; yet the
digital documents and records are far more fragile than paper. In fact, the record of the entire
present period of history is in danger. The content and historical value of many governmental,
organizational, legal, financial, and technical records, scientific databases, geospatial
databases and personal documents may be irretrievably lost to future generations if we do not
take steps to preserve them. The longevity of the GI is one of the most complex problem if it
is considered only the discouraging characteristic of being software-dependent. Preserving

digital documents and records containing GI, GK, and GS is almost analogous to preserving
ancient written texts [107].
Accessibility. One of the most prominent features of the Web is the easy and ubiquitous
access. Services channeled through standard browsers reach the most distant corners of our
globe. The illusion that standards provide an answer is only an illusion and the promise of
“migration” seemly to be false and time and costs consuming. Although defining ultimate
standards for digital documents and for the resources of SISMSG may be an admirable goal, it
is premature [107]. The field seemly to be too new to have developed an accepted, formal
understanding of the ways that humans manipulate GI and GK. It would be presumptuous to
imagine that we are ready to enumerate the most important kinds of digital applications
containing GI and GK, let alone to propose that we are ready to circumscribe their capabilities
by standardizing them. Any attempt to force users to settle for artificial limitations imposed
by such standards would be futile, since the momentum of the information technology
revolution derives directly from the attraction of new capabilities. It may become feasible to
define long-term standards for digital documents containing spatial information sometime in
the future, when information science rests on a more secure, formal foundation, but standards
do not yet offer a perfect solution to this problem.
Responsiveness. Specialists and/or simple citizens need feedback and the feeling of
being heard and participating in the processes of spatial governance. Most systems will be
strengthened by processes allowing users to contribute to the development but as mentioned
above, we cannot imagine till now the complexity of the applications in the area of SG
starting from the GIS, SDSS including Artificial Intelligence techniques and sensors.
Robustness. Robust systems are fault tolerant in the sense that sub systems may fail
without bringing the system down. This might be implemented by exploiting redundancy or
by providing backup systems/methods. Robustness also implies that it should be easy to
replace modules of the system without redesigning the whole system. This aspect is very
important in Spatial Governance having in mind the diversity of processes and relation
between them.
b) Considering the society‟s point of view, we have to discuss about adaptivity,
scalability, simplicity, stability and environmentally friendly.
Adaptivity is the challenge that faces all researchers in this field is the development of
adaptive systems for unpredictable and unforgiving applications domain [108]. A system
should have an underlying structure that allows for easy adapt of technological innovations
and improvements.
Scalability. In this context the scalability does not refer to the scalability from geospatial
domain but it refers to the meaning from IT&C domain. Accelerated interest in and rapidly
increasing number of request against a web server may bring a system to halt, and in several
cases start a domino effect which eventually will crash the system fundamentally. A system
has to gracefully handle an increasing work load, and also be prepared for a rapid growth of
the content.
Simplicity. “To facilitate easy adoption of these technologies by the user community, the
industry needs to make the applications cheaper and simpler for them” [1]. This is a common
denominator for long-time survivors in the world of IT&C systems and in geospatial world. A
sustainable system should be well planned with easy-to-use-and-understand design and clear.
Stability. A sustainable system should be run by an organization, or one or more key

responsible if needed, with regards to maintenance and development. Commercial companies,
public and/or private, may administer sustainable systems, but would most likely drop a
project if becomes financially uninteresting.
Environmentally Friendly. Environmentally friendly is a term used to refer to goods
and services, laws, guidelines and policies claimed to inflict minimal or no harm on the
environment.
c) But the really gain is acquired by applying process management and it is expected to
reach interoperability and affordability in order to use the harmonized content and of good
quality, preserving the previously investments.
Process Management in Spatial Governance. The generic process management
mechanisms needed to provide SG are those pertaining to: authority; planning;
monitoring[109]. Interaction is necessary between the generic mechanisms. Authority can
involve rights and responsibilities in respect of planning - and planning can have implications
for the kind of authority necessary to implement plans. Planning has implications for
monitoring to indicate whether planning is successful - and monitoring provides the
experience on which further planning can take place. Monitoring has implications for whether
an authority is performing satisfactorily - and an authority can instigate changes in
monitoring regimes as a test of its own performance.
Spatial Planning Processes. Practices of spatial planning (SP) in Europe have borne
witness to considerable challenges in recent decades as a consequence of changing policy
priorities, heightened expectations and increased socio-economic and demographic disparities
[110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122]. However, recent decades
have, also witnessed the emergence of new strategic approaches to SP at local, regional,
national and transnational scales, which have sought to broaden the scope and enhance the
governance capacity of SP in practice [113, 114]. As presented in [115] the purpose of
strategic planning is to inform decision-making in practice, rather than to directly influence
material outcomes (i.e. the object of project plans). From this perspective, strategic SP is
viewed as a continuous process, which provides a frame of reference for subsequent decisions
and actors [115]. The concept of strategic SP places emphasis on the development of
coordinated or integrated perspectives that transcend traditional sectoral policy divisions
through a specific focus on the spatial impacts of sectoral policies [114, 116, 117].
Challenging an instrumentalist view of planning, a number of authors have stressed the role
of values in strategic SP. For Healey [113] the concept of strategic SP places particular
emphasis on the development of approaches to policy-making where objectives and values
relating to future socio spatial development within a territory or functional space are shaped,
framed and negotiated as part of the strategy-making process. In [114] similarly defined
strategic Spatial Planning as a: “public-sector-led sociospatial process through which a
vision, actions and means for implementation are produced that shape and frame what a
place is and may become”. The concept of a re-emergence or revival of strategic SP in
Europe since the 1990s is more specifically associated with the development of SP policy and
perspectives at the European scale. In particular, the publication of a European Spatial
Development Perspective [123] and the adoption of the principle of territorial cohesion policy
have served to highlight the significance of territorially differentiated perspectives on
processes of social and economic development in the context of the European Union ([118,
119, 120, 121, cited in [122]). Planning involves a range of information-intensive activities
aimed to improve performance by design or 'on purpose' for improving performance taking
advantage of future possibilities - or avoiding - future possibilities. Design involves working
out how to achieve some purpose within constraints of what is deemed possible in: a technical
sense (something that can be done); an economic sense (something that is affordable) and a
political sense (something that is socially permissible and acceptable) [109]. This issue could
be expanded in SDI in Spatial Governance: “Today, every local action has a global reaction.
So, every global challenge influences local living. Geospatial technology helps understand
and manage our resources. I believe geospatial industry has the true potential to be a game
changer in the current world order" (Sanjay Kumar in [1]).
Interoperability. “We were working in silos and this is particularly prevalent in
intelligence communities. Increasingly now, that is not the way things work. We need to fuse
across traditional boundaries, inculcate the culture of sharing and enable the databases to
become interoperable. It is also about the interoperability of the mind” (Marchal Stuart Peach
in [1]).
As mentioned in the literature [124, 125, 126, 127, 128, 129, 130, 131] information
system interoperability occurs at multiple levels within and across platforms, providing the
capacity of interchanging data, services, and applications among different organisations as
well as users [130, 131]. Several interoperability levels have been proposed to achieve system
interoperability, including technological, syntactic and semantic levels (ISO 2002, 126, 127].
In a GIS context, interoperability has emerged as a research issue on computation
implementations of GD exchange since a decade ago when eight interoperability levels were
identified: these being the information community and institution, enterprise, application
tools, middleware, data store, distributed computing environment and network [132]. In a SDI
context, the Open Geospatial Consortium (OGC) and International Organisation for
Standardisation (ISO) TC211 have played a major role in improving GD and System
Interoperability through the specification of object models and XML schemas to store and
transfer data, Open Service Interfaces, and data and services metadata standards. The
formalisation of interoperability models became essential for developing abstractions of
critical measures for the connectivity of different interoperability levels.
“In order to work more efficiently and in a unified manner, there is an increased need to
have more integrated and active SDIs and standardization of data to facilitate
interoperability” [1].
In our opinion, in connection with SISMSG, it is necessary to pay attention also to the
aspect closely related to interoperability named in the literature persistence, the
“interoperability with the future”, i.e. the independent systems able to exchange meaningful
information and initiate actions from each other are separated by time.
Affordability. This aspect is in connection with identification of bottlenecks, missing
links and quality of service in infrastructure networks. This poses two types of challenges:
 How to speed up a process in order to achieve a sustainable development with legal

instruments which are its key;
 How can effectively improve the implementation management mechanisms and how
much could rely on cooperation with the partners.
Harmonized Content. “Opportunities are tremendous and it is time for geospatial

revolution, but we all face challenges, we need to work together on standardisation and
harmonisation on sharing data. That is how we can create a geospatial revolution” (Melanie
Schultz van Haegen-Maas Geesteranus in [1]).
A resource of SISMSG (such as GD, GK and GI) can be part of more than one domain
(document, records, scientific domain etc.), and can be identified by different systems, so it is
necessary to guarantee interoperability between different identification systems as well as
implementations based on the same namespace [125]. OGC in parallel with World Wide Web
Consortium (W3C) and ISO develop standards and specifications to support the
interoperability between repositories with geospatial content. The scope of these standards
and specifications is to enable an application developer to use any geospatial content
(resources and services) available on “the net” within a single environment and a single
workflow [133]. However, standardization may not solve the problem of interoperability by
itself, because [134] the construction and maintenance of a single and integrated model is a
hard task and the requirement to communicate with geospatial sources that do not complying
the adopted standard will be always present; existing geospatial sources have their own
models which may not always be mapped to the common model without information loss; the
standards/specifications are subject to continuous change, but systems will not all
simultaneously change to conform.
Quality can be described from different viewpoints [135, 136]. Based on [104] quality
can be defined as fitness for use, including both quality of design, conformance to the design
(production oriented quality), customer satisfaction and the fulfilment of the needs of society
or environment. Most of the quality descriptions of the resources of SISMSG have been
developed to serve the production-oriented approach. ISO 19113 and ISO 19114 standards
follow this approach based on the data quality concepts developed already in the 1980‟s [137,
138, 139, 140, 141], on earlier standards like the Spatial Data Transfer Standard (SDTS)
(FGDC, 1991) and CEN pre-standard on quality in 1998. In the context of SISMSG, the
concept of spatial quality (SQ) could be developed, in our opinion, by means of three
components: use value, perception value and future value. These three terms have become the
point of reference in discussions on SQ [142]. A high use value occurs when space can be
used in a safe manner for several purposes (to live and to work, but also to recreate and to
move through). The idea is that these different functions do not hinder and possibly reinforce
each other. Perception value plays an important role in people‟s living environment. Cultural
awareness and diversity, the presence of characteristic properties (identity) and of history and
beauty, are all elements that can be attributed to perception value. Perception value also
involves spatial diversity and variation. Future value includes characteristics such as
sustainability, bio-diversity, robustness and flexibility, both concerning suitability for new use
forms and admissibility for new cultural and economic meanings [142]. The future value are
in direct relation with persistence.
Preserving the Previously Investments. Despite the technological advances, the
problem of archiving remains: there is, as yet, no satisfactory solution for to the long term
preservation of the vast amounts of the resources for SISMSG (spatial data, spatial
knowledge, geospatial information) already created. It is hard at this point to see how century
long preservation of digital media can be achieved, together with the means to search and
retrieve useful information.
Regarding Human Resources, significant research work that considers how C-ITS
impacts on the citizens as drivers [60] and passengers has been carried out over a number of
years in research projects and by in-house work from vehicle manufacturers and road
transport testing laboratories. Much of this work has been undertaken in European research
projects. A detailed analysis has been made of existing ITS projects (National and European)
and also of the work of Consortiums/Organizations/Institutes working in the ITS related area,
and the high-level observations on what work are contained in ([57] annex B). In the same
time, in order to improve the infrastructure “… there is a need to build capacity for enhanced
integration of geospatial technologies into the mainstream education by means of
partnerships between the industry, government and academia” [1].
ITS has been attracting attention from many professionals in several countries all over the
world. Groups of discussion have been created in order to contribute to the effective
development of ITS and also to take an active role in sharing its future standards since the
eighties [54].
CONCLUSION
Starting from the fact that it is recognized that “ICT reinforces the convergence of all
three pillars of development -- social, economic and environmental, and that ICT by
definition embraces the fundamental principles of sustainability” [26] and from the fact that
United Nations Istanbul Plan of Action for Least Developed Countries (2011-2020) [143] and
based on the current literature, this chapter proposes a reorientation of the way the concept of
sustainability is dealt with, Spatial Governance based on human behavior and the processing
of Geospatial Knowledge at the centre of the concept. The sustainability approach presented
in this chapter is then translated into requirements needed for a Sustainable Information
Systems Management for Spatial Governance focusing on (Cooperative) Intelligent
Transportation Systems, Spatial Data Infrastructure and emphasizing Spatial Governance.
To conclude this chapter, the discussion of the characteristics of Sustainable Information
Systems Management for Spatial Governance has been used to briefly revisit the complexity
of involved in the concept of sustainability, sustainability of IS IT&C innovation, Spatial
Data Infrastructures and his multiple facets in connection with Spatial Governance and the
resources. Positioning Spatial Governance based on human behavior and the processing of
Geospatial Knowledge at the centre of the concept, on the one site, and inspired from the
paper elaborated on the way quality should be taken into account in the development of
innovation, Spatial Data Infrastructures – the Spatial Governance‟s kernel - of [60] based on
four quality management viewpoints, i.e. a production-centred, planning-centred, customer-
centred and a system-centred perspective, to exemplify has been used an architecture that
seems suitable to support a multi-layered approach, underling the benefit to users, benefit to
society and the gain.
Based on the experience of the authors and on the literature, has been highlighted that
such Sustainable Information Systems Management for Spatial Governance:
 will come from the holistic and multilevel vision and possibility of for real time
actions of the various systems constituting the mobility space because such approach
is the starting point for integration and interoperability between systems;
 will be not just about space and spatial relations but also about place, culture,
identity, spanning physical and virtual space;
 will emphasise the analysis of networks and flows and relationships between places
(including cultural influences) in addition to the traditional spatial analysis;
 will be easy to use with different levels of functionality for different types of users;
 the competition for limited resources implies reduced redundancies in the system and
a large vulnerability to random failures.
In order to highlight the complexity, the authors suggested some characteristics of a

Sustainable Information Systems Management for Spatial Governance that should be
applicable, in general, for any infrastructure based on IT&C innovation but also to any
domain‟s application including (cooperative) intelligent transportation systems and defining
spatial governance.
It is necessary to recognize that the implementation of the Sustainable Information
Systems Management for Spatial Governance require changes in behaviour and adaptation to
new modes of governance, from the design and implementation of new governance to
individual awareness and acceptance of new mobility modes. The consequences of the
implementation of such Sustainable Information Systems Management for Spatial
Governance will be the empowerment of individuals, organizations, governments and society.
REFERENCES
[1] Geospatial World Forum Report. RAJ Convention Center, Amsterdam, The
Netherlands, 23-27 April 2012, http://anax1em.pressmart.com/GeospatialMedia
Proceedings/In dex.aspx.
[2] Glavic, P.; Lukman, R.; Review of Sustainability Terms and their Deﬁnitions.
Journal of Cleaner Production, vol. 15, no. 18, pp. 1875-1885, 2007.
[3] Cato, S.M.; Green Economics. ISBN 9781844075713, Earthscan, London, UK, pp. 36-
37, 2009.
[4] Adams, W.M.; The Future of Sustainability: Re-Thinking Environment and
Development in the Twenty-First Century. Report of the IUCN Renowned Thinkers
Meeting, 29-31 January 2006, http://cmsdata.iucn.org/downloads/iucn_future_of_
sustanability.pdf.
[5] Dictionary.com. http://dictionary.reference.com/browse/sustain.
[6] Onions, C.T.; The Shorter Oxford English Dictionary. Clarendon Press, Oxford, p.
2095, 1964.
[7] United Nations General Assembly; Report of the World Commission on Environment
and Development: Our Common Future. Transmitted to the General Assembly as an
Annex to Document A/42/427 - Development and International Co-operation:
Environment; Our Common Future, Chapter 2: Towards Sustainable Development;
Paragraph 1, March 20, 1987, www.un-documents.net/ocf-02.htm.
[8] United Nations General Assembly; World Summit Outcome. Resolution A/60/1,
Adopted by the General Assembly, 15 September 2005.
[9] Forestry Commission of Great Britain; Sustainability, www.forestry. gov.uk/forestry
/edik-59fmzf.
[10] Manning, S.; Boons, F.; Von Hagen, O.; Reinecke, J.; National Contexts Matter: The
Co-Evolution of Sustainability Standards in Global Value Chains, Ecological
Economics, Forthcoming, 2011, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=
1752655.
[11] Reinecke, J.; Manning, S.; Von Hagen, O.; The Emergence of a Standards Market:
Multiplicity of Sustainability Standards in the Global Coffee Industry. Organization
Studies, Forthcoming, 2012, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=
1970343.
[12] SAI Platform; Sustainability Indicators. Sustainable Agricultural Initiative. 2010,
www.saiplatform.org/activities/alias/sustainability-indicators.
[13] Alvarez, G.; Sustainable Agriculture and Value networks, Lausanne, Switzerland:
Latitude. 2012, www.intracen.org/WorkArea/DownloadAsset.aspx?id=51770.
[14] What is Sustainable Development?. International Institute for Sustainable
Development, 2009, www.iisd.org/sd/.
[15] Sustainable Development: Introduction. EurActiv, 2004, www.euractiv.com/
sustainability/sustainable-development-introduction /article-117539.
[16] Kates, R.; Parris, T.; Leiserowitz, A.; What Is Sustainable Development?. Environment,
vol. 473, pp. 8-21, 2005, www.euractiv.com/sustainability/sustainable-development-
introduction/article-117539.
[17] Holling, C.S.; Theories for Sustainable Futures. Conservation Ecology, vol.42, p. 7,
2000, www.ecologyandsociety.org/vol4/iss2/art7.
[18] Redclift, M.; Sustainable Development 1987-2005: An Oxymoron Comes of Age.
Sustainable Development, vol. 134, pp. 212-227, 2005.
[19] Porritt, J.; Capitalism as If the World Mattered. ISBN 9781844071937, Earthscan,
London, UK, p. 46, 2006.
[20] IUCN/UNEP/WWF, Caring for the Earth: A Strategy for Sustainable Living. Gland,
Switzerland, 1991, http://coombs.anu.edu.au/ ~vern/caring/caring.html.
[21] Markus, J.; Milne, M.K.; Kearins, K.; Walton, S.; Creating Adventures in Wonderland:
The Journey Metaphor and Environmental Sustainability. Organization, vol.136, pp.
801-839, 2006, http://org.sagepub.com/content/13/6/801.abstract, retrieved on 2012-04-
08.
[22] The Earth Charter. The Earth Charter Initiative, 2000, www.earthcharterinaction.org/
content/pages/Read-the-Charter.html.
[23] Costanza, R.; Patten, B.C.; Defining and Predicting Sustainability. Ecological
Economics, vol. 15, no. 3, pp. 193-196, 1995.
[24] Pezzoli, K.; Sustainable Development: A Transdisciplinary Overview of the Literature.
International Congress of Collegiate Schools of Planning and the European Schools of
Planning, Toronto, July 1996.
[25] Faber, N.R.; Jorna, R.J.; Van Engelen, J.M.; The Sustainability of Sustainability: A
Study into the Conceptual Foundations of the Notion of Sustainability. Journal of
Environmental Assessment Policy and Management, vol. 71, 2005.
[26] Kent, D.; Should Information and Communications Technology Experts Have a Seat at
the Table During the Rio+20 Summit?. 2012, www.huffingtonpost.com/daniel-kent/rio
20_b_1516056.html.
[27] Zwetsloot, G.I.; Van Marrewijk, M.N.; From Quality to Sustainability. Journal of
Business Ethics, vol. 55, pp. 79-82, 2004.
[28] Caldelli, A.; Parmigiani, M.L.; Management Information System - A Tool for
Corporate Sustainability. Journal of Business Ethics, ISSN: 0167-4544, vol. 55, issue 2,
pp. 159-171, Dec. (I) 2004.
[29] Clayton, A.M.H.; Radcliffe, N.J.; Sustainability: A System Approach. Earthscan
Publications Ltd, ISBN: 1 85383319 3, ISBN: 1 85383314 2, 1997.
[30] Edwards, A.R.; The Sustainability Revolution: Portrait of a Paradigm Shift. New
Society Publishers, ISBN: 0-86571-531-9, 2006.
[31] TWHOMB. The White House, Office of Management and Budget, Circular no. A-16
Revised, August 19, 2002, http://www.whitehouse.gov/omb/circulars_a016_rev/.
[32] Masser, I.; GIS Worlds: Creating Spatial Data Infrastructures. Redlands, ESRI, CA,
USA, 2005.
[33] Nebert, D.; Developing Spatial Data Infrastructures. The SDI Cookbook, Version 2.0,
2004, www.gsdi.org/docs 2004/Cookbook/cookbookV2.0.pdf.
[34] Kuhn, W.; Introduction to Spatial Data Infrastructures. Presentation, March 14, 2005,
www.docstoc.com/docs/2697206/Introduction-to-Spatial--Data-Infra structures.
[35] Hendriks, H.J.; Dessers, E.; Van Hootegem, G.; Reconsidering the Definition of a
Spatial Data Infrastructure. International Journal of Geographical Information Science,
doi:10.1080/136588 16.2011.639301, 2012, www.tandfonline.com/doi/pdf/10.1080/
136588 16.2011.639301.
[36] Masser, I.; All Shapes and Sizes: The First Generation of National Spatial Data
Infrastructures. International Journal of Geographical Information Science, vol. 13, pp.
67-84, 1999.
[37] European Commission, Establishing an Infrastructure for Spatial Information in the
European Community (INSPIRE). Directive 2007/2/EC of the European Parliament and
of the Council, 14 March 2007.
[38] Giff, G.A.; Crompvoets, J.; Performance Indicators: A Tool to Support Spatial Data
Infrastructure Assessment. Computers Environment and Urban Systems, vol. 32, no. 5,
pp. 365-376, 2008.
[39] Salvemini, M.; Van Den Broucke, D.; Results of the INSPIRE State of Play 2010: First
Lessons Learnt from the Application of INPIRE Directive in EUROPE. Third
Preparatory Meeting of the Proposed United Nations Committee of Experts on Global
Geospatial Information Management, New York, 4-6 April 2011.
[40] www.commission4.isprs.org/wg1/.
[41] Hadley, C.; Technological Innovations in Land Survey and Large-Scale Mapping: New
Techniques for Revision of Large-Scale Map Data: The Great Britain Experience and
Lessons Learned. UNCHS Habitat, Land Survey and Large-Scale Mapping in Sub-
Saharan Africa, p. 118, 2001.
[42] Benhardsen, T.; Geographic Information Systems. VIAK IT, Norway, pp. 9-17, 1992.
[43] Grimshaw, D.J.; Bringing Geographical Information into Business. Longman Group
Ltd, pp. 22-40, 1994.
[44] Densham, P.J.; Spatial Decision Support Systems. Maguire, D.J.; Goodchild, M.S.;
Rhind, D.W.; (Eds.), Geographical Information Systems: Principles and Applications.
Longman, London, pp. 403-412, 1991.
[45] Jankowski, P.; Nyerges, T.; Smith, A.; Moore, T.J.; Horvath, E.; Spatial Group Choice:
A Spatial Decision Support Tool for Collaborative Decision Making International
Journal of Geographical Information Systems, vol. 11,no. 6, pp. 577-602, 1977.
[46] Keenan, P.B.; Geographic Information Systems. Encyclopedia of Information Systems,
Ed. by Bidgoli, H.; pp. 421-432, Elsevier Science USA, San Diego, CA, USA, 2003.
[47] Keenan, P.B.; Spatial Decision Support Systems: A Coming of Age. Control and
Cybernetics, vol. 351, pp. 9-27, 2006.
[48] Pick, J.B.; Geographic Information Systems. Business, Idea Group Publishing, Hershey,
PA, USA, 2005.
[49] Huerta, E.; Navarrete, C.; Ryan, T.; GIS and Decision-Making in Business: A
Literature Review. Pick, J.B.; Ed., Geographic Information Systems in Business. Idea
Group Publishing, Hersey, PA, USA, pp. 151-174, 2005.
[50] Jarupathirun, S.; Zahedi, F.; GIS as Spatial Decision Support Systems. Pick, J.B.; Ed.,
Geographic Information Systems in Business. Idea Group Publishing, Hersey, PA,
USA, pp. 151-174, 2005.
[51] Churchman, C.W.; The Systems Approach and its Enemies. ISBN: 9780465083428,
Basic Books Inc., New York, USA, 1979.
[52] European Roadmap - Infrastructure for Green Vehicles. 2012, www.ertrac.org/pictures/
downloadmanager/6/54/infrastructure-for-green-vehicles_final-october-2012_65.pdf.
[53] Goldman, T.; Gorham, R.; Sustainable Urban Transport: Four Innovative Directions.
Technology in Society, vol. 28, pp. 261-273, 2006.
[54] Wu, Y.C.J.; Lee, P.L.; The Use of Patent Analysis in Assessing ITS Innovations: US,
Europe and Japan. Transportation Research, Part A, vol. 41, pp. 568-586, 2007.
[55] HIDO; A Bible for Those Involved in ITS Around the World: ITS Review. Ministry of
Land, Infrastructure and Transport, Highway Industry Development Organization,
Japan, 1997, Retrieved February 2008, www.hido.or.jp/ITSHP_e/Rev/review2-
4/review4/1-3e.html.
[56] C-ITS;Key to Innovation Integrated Solution: Cooperative Intelligent Transport
Systems and Services (C-ITS). April 2013, www.eu-smartcities.eu.
[57] ETSI; Human Factors (HF). Intelligent Transport Systems (ITS):ICT in Cars. ETSITR
102 762 V1.1.1,DTR/HF-00117, 2010, www.etsi.org.
[58] Van Voorst Tot Voorst, M.P.; Future of Super Intelligent Transport Systems. EFP
Brief, no. 202, 2011, www.foresight-platform.eu/wp-content/uploads/2011/11/EFP-
Brief-No.-202_Future-of-Superintelligent- Transport-Systems.pdf.
[59] Ezell, S.; Explaining International IT Application Leadership: Intelligent Transportation
Systems. ITIF, 2010, www.itif.org/files/2010-1-27-ITS_Leadership.pdf.
[60] De Oliveira Pereira, M.S.; In-Vehicle Information Systems - Related Multiple Task
Performance and Driver Behavior: Comparison between Different Age Groups.
Dissertation Submitted Public Advocacy for the Degree of Doctor in Business Human
Kinetics, Specialty Ergonomics, 2009, http://www.repository. utl.pt/bitstream/
10400.5/1232/1/In-Vehicle%20Information%20System-Related%20Multiple%20Task
%20Performance%20and%20Driving%20Behaviour-%20Comparison% 20of%20Two.
pdf.
[61] http://ec.europa.eu/research/social-sciences/projects/360_en.html.
[62] http://www.aide-eu.org/.
[63] http://ec.europa.eu/transport/road_safety/projects/doc/haste.pdf.
[64] http://www.ifsttar.fr/en/europe-international/colonne-1/projets-europeens/actions-cost/.
[65] http://www.noehumanist.org/news.php.
[66] Brusque, C.; The Influence of In-Vehicle Information Systems on Driver Behaviour and
Road Safety: Synthesis of Existence Knowledge Cost 352. Collections of INRETS.
Synthesis, no. 54, 2007.
[67] Lee, J.; Forlizzi, J.; Hudson, S.; Studying the Effectiveness of MOVE: E-Contextually
Optimized In-Vehicle Navigation System. In-Vehicle Interfaces, Portland, Oregon,
USA, 2005.
[68] Brooks, A.; Lenneman, J.; George-Maletta, K.; Hunter, D.R.; Green, P.; Preliminary
Examinations of the Time to Read Electronic Maps: the Effects of Text and Graphic
Characteristics (UMTRI-98-36). University of Michigan Transportation Research
Institute, 1999, http://www.umich.edu/~driving/publications/UMTRI-98-36.pdf
[69] Burnett, G.; Usable Vehicle Navigation Systems: Are We There Yet?. 2000,
www.cs.nott.ac.uk/~geb/pubs.html.
[70] Ross, T.; Burnett, G.; Evaluating the Human-Machine Interface to Vehicle Navigation
Systems as an Example of Ubiquitous Computing. International Journal of Human-
Computer Studies, vol. 55, pp. 661-674, 2001.
[71] Brooks, A.; Nowakowski, C.; Green, P.; Turn-by-Turn Displays versus Electronic
Maps: An on-the-Road Comparison of Driver Glance Behaviour (UMTRI-98-37). The
University of Michigan, 1998.
[72] Jackson, P.J.; In Search for Better Route Guidance Instructions. Ergonomics, vol. 41,
no. 7, pp. 1000-1013, 1998.
[73] Adler, J.L.; Blue, V.J.; Toward the Design of Intelligent Traveler Information Systems.
Transportation Research, vol. 6, issue 3, pp. 157-172, June 1998.
[74] Pauzie, A.; Aging Population and Ergonomics Technological Innovation
Communication in Driving. Research in Transport Security, vol. 81, pp. 203-212, 2003.
[75] Pauzie, A.; Ergonomics and Safety of Mobile Systems in the Vehicle: A Reality Pauzie,
A. (Ed.); Ergonomics of Communicating Systems in the Vehicle Use and Safety. pp. 97-
102, INRETS. LESCOT, 2001.
[76] Schraagen, J.M.C.; Information Presentation in In-Car Navigation Systems. Parkes,
A.M.; Franzen, S. (Eds.); Driving Future Vehicles. pp. 171-185, Taylor & Francis,
1993.
[77] Zaidel, D.M.; Noy, Y.I.; Automatic versus Interactive Vehicle Navigation Aids. Noy,
Y.I. (Ed.); Ergonomics and Safety of Intelligent Driver Interfaces. pp. 287-307,
Lawrence Erlbaum Associates, 1997.
[78] Daimon, T.; Kawashima, H.; Akamatsu, M.; Drivers' Cognitive Process and Route
Guidance. Noy, Y.I. (Ed.); Ergonomics and Safety of Intelligent Driver Interfaces. pp.
273-286, Lawrence Erlbaum Associates, 1997.
[79] Pauzie, A.; Ergonomic Guidelines for the Driver System Interface: Help to the Road
Safety Design Process.4th Asia Pacific Conference on Human Computer Interaction,
Singapore, Amsterdam, Elsevier Science Ltd, New York, 2000.
[80] Pauzie, A.; Ergonomics and Safety of Mobile Systems in the Vehicle. Ergonomic
Inspection Guidance Function.vol.0501,INRETS. LESCOT, 2005.
[81] Thompson, S.; Nomadic Devices: Distraction Issues and Specific Results. HUMANIST
(Ed.), Impact of IVIS on Driver Workload and Distraction: Review of Assessment
Methods and Recent Findings. pp. Annex 5, HUMANIST, 2004.
[82] Vienna Convention on Road Traffic. www.unece.org/fileadmin/DAM/trans/conven
tn/crt1968e.pdf.
[83] Samad, T.; Annaswamy, A.M.; The Impact of Control Technology. Glielmo, L.;
University of Sannio, Italy,2011, http://ieeecss.org/sites/ieeecss.org/files/documents/
IoCT-Part4-13VehicleToVehicle-HR.pdf.
[84] Motsinger, C.; Hubing, T.; A Review of Vehicle-to-Vehicle and Vehicle-to-
Infrastructure Initiatives. Technical Report,CVEL-07-003,2007, www.cvel.clemson.
edu/pdf/CVEL-07-003.pdf.
[85] http://friction.vtt.fi/.
[86] http://www.ertico.com/gst-website.
[87] http://ec.europa.eu/information_society/activities/esafety/doc/rtd_projects/fact_sheets/p
r event.pdf.
[88] Kellum, C.C.; Deegener, M.; An Analysis of the Differences Between the Evolving
United States and European Dedicated Short Range Communications (DSRC) Systems.
Detroit, Society of Automotive Engineers, MI, 2007.
[89] www.centrico.ten‐t.com/documents/briefing%20notes/03_InnovativeProjects_I.pdf.
[90] http://ec.europa.eu/information_society/activities/esafety/doc/rtd_projects/fact_sheets/c
all_4/ comesafety.pdf.
all_4/cvis.pdf.
[92] http://www.heero-pilot.eu/view/en/ecall.html
all_4/coopers.pdf
[94] Gerlach, M.; Trusted Network on Wheels. October 2005, www.ercim.org/publication/
Ercim_News/enw63/gerlach.html.
[95] Alampi, C.; Euromobility. 2007, www.euromobilityforum.compresentazioni_11_04_
07/ EMF_Seminar_11‐04‐07_Alampi‐ANAS.pdf
[96] www.eu-smartcities.eu.
[97] Proskawetz, K.O.; Klug, S.; Beckert, B.; Key to Innovation Integrated Solution,
Cooperative Intelligent Transport Systems and Services (C-ITS). 2013,www.eu-
smartcities.eu.
[98] www.its-nds.de.
[99] Klug, S.;Breuil, D.;De Wilde, H.; Key Innovation Integrated Solution. Smart
Organization of Traffic Flow and Logistics,2013, www.eu-smartci ties.eu.
[100] Ammana, A.; Overview of IntelliDrive/Vehicle Infrastructure Integration (VII). Report
Edited by Virginia Tech Transportation INSTITUTE,2009, http://filebox.vt.edu/users/
aaman na/web%20page/VII-IntelliDrive%20Report-edited.pdf.
[101] Schmidt, N.H.; Kolbe, L.M.; Erek, K.; Zarnekow, R.; Sustainable Information Systems
Management. Business & Information Systems Engineering, vol. 5, pp. 400-402, 2009.
[102] Maruster, L.; Faber, N.R.; Peters, K.; Sustainable Information Systems: A Knowledge
Perspective.CAiSE'08 Forum, pp. 57-60, 2008.
[103] Craglia, M.; De Bie, K.; Jackson, D.; Pesaresi, M.; Remetey-Fulopp, G.; Wang, C.;
Annoni, A.;Bian, L.; Campbell, F.; Ehlers, M.; Van Genderen, J.; Goodchild, M.; Guo,
H.; Lewis, A.; Simpson, R.; Skidmore, A.; Woodgate, P.; Digital Earth 2020: Towards
the Vision for the Next Decade. International Journal of Digital Earth, vol. 5, issue 1,
pp. 4-21, 2012.
[104] Jakobsson, A.; Tsoulos, L.; The Role of Quality in Spatial Data Infrastructures. 23rd
International Cartographic Conference, Moscow, Russia, 2007.
[105] Bacastow, T.S.; Bellafiore, D.J.; Redefining Geospatial Intelligence. American
Intelligence Journal, The Magazine for Intelligence Professionals, pp. 38-40, Fall 2009.
[106] Besser, H.; Digital Longevity: Handbook for Digital Projects, A Management Tool for
Preservation and Access. Sitts, M. (Ed.); Northeast Document
ConservationCenter,Andover MA,pp. 155-166,2000, www.gseis.ucla.edu/~howard/
Papers/sfs-longevity.html.
[107] Rothenberg, J.; Ensuring the Longevity of Digital Information. An Expanded Version of
the Article, Ensuring the Longevity of Digital Documents. Scientific American, vol.
272, no. 1, pp. 42-47,1999,www.clir.org/programs/otheractiv/ensuring.pdf.
[108] Patel, N.V.; Adaptive Evolutionary Information Systems. ISBN: 1-59140-034-1, 9-
781591-400349, IDEA GROUP Publishing, 2003.
[109] Bradbury, R.H.; Sustainable Development as a Subversive Issue.UNESCO Flagship,
Nature & Resources, October 1998.
[110] Albrechts, L.; Strategic Spatial Planning Re-Examined. Environment and Planning B,
Planning and Design, vol. 31, no. 5, pp. 743-758,2004.
[111] Nadin, V.; The Emergence of the Spatial Planning Approach. Planning Practice and
Research, England, vol.22,no. 1, pp. 43-62, 2007.
[112] Schmidt, S.; Land Use Planning Tools and Institutional Change in Germany: Recent
Developments in Local and Regional Planning. European Planning Studies, vol. 17, no.
12, pp. 1907-1921, 2009.
[113] Healey, P.; Collaborative Planning: Shaping Places in Fragmented Societies. Palgrave
Macmillan, Hampshire, 1997.
[114] Albrechts, L.; Healey, P.; Kunzmann, K.R.; Strategic Spatial Planning and Regional
Governance in Europe. Journal of the American Planning Association, vol. 69,no. 2, p.
113, 2003.
[115] Faludi, A.; The Performance of Spatial Planning. Planning Practice and Research, vol.
15, no.4, pp. 299-318, 2000.
[116] Salet, W.; Thornley, A.; Kreukels, A.; Institutional and Spatial Coordination in
European Metropolitan Regions. Salet, W.; Thornley, A.; Kreukels, A (Eds.),
Metropolitan Governance and Spatial Planning: Comparative Case Studies of
European City-Regions. Spon Press, London and New York, pp. 3-19, 2003.
[117] [Adams, N.; Alden, J.; Harris, N.; Introduction: Regional Development and Spatial
Planning in an Enlarged European Union. Adams, N.; Alden, J.; Harris, N. (Eds.),
Regional Development and Spatial Planning in an Enlarged European Union. Ashgate,
Aldershot, pp. 3-16, 2006.
[118] Commission of the European Communities CEC; Green Paper on Territorial Cohesion:
Turning Territorial Diversity into Strength. 2008, http://ec.europa.eu/regional_
policy/consultation/terco/index_en.htm.
[119] Faludi, A.; Making Sense of the Territorial Agenda of the European Union. European
Journal of Spatial Development, Refereed Articles, vol. 25,2007, www.nordregio.
se/EJSD/refereed25.pdf.
[120] Faludi, A.; Territorial Cohesion and the European Model of Society. Lincoln Institute
of Land Policy, Cambridge, MA, 2007.
[121] Faludi, A.; European Spatial Planning: Past, Present and Future. Town Planning
Review, vol. 81, no. 1, pp. 1-22, 2010.
[122] Allin, S.; Walsh, C.; Strategic spatial Planning in European City-Regions: Parallel
Processes or Divergent Trajectories. NIRSA Working Papers Series, September 2010,
www. nuim.ie/nirsa/research/documents/WP60_Allin_Walsh.pdf.
[123] European Spatial Development Perspective.1999, http://ec.europa.eu/regional_policy/
sources/docoffic/officeal/reports/pdf/sum_en.pdf.
[124] Janowicz, K.; Schade, S.; Broring, A.; Kebler, C.; Maue, P.; Stasch, C.; Semantic
Enablement for Spatial Data Infrastructures. Transactions in GIS, 2010, vol. 142, pp.
111-129, 2010.
[125] Paskin, N.; Identifier Interoperability- A Report on Two Recent ISO Activities. D-Lib
Magazine, ISSN 1082-9873, vol. 12, no. 4, April 2006, www.dlib.org/dlib/april06/pas
kin/04paskin.html.
[126] Tolk, A.; Beyond Technical Interoperability - Introducing a Reference Model for
Measures of Merit for Coalition Interoperability. Command and Control Research and
Technology Symposium (CCRTS), Washington DC, 2003.
[127] Turnitsa, C.; Tolk, A.; Battle Management Language: A Triangle with Five Sides. Paper
06SSIW-016, Spring Simulation Interoperability Workshop, Simulation Interoperability
Standards Organization, IEEE CS Press, April 2006.
[128] Groot, R.; McLaughlin, J.; Introduction: Geospatial Data Infrastructure - Concepts,
Cases and Good Practice. Groot, R.; McLaughlin, J. (Eds.), Oxford University Press,
Oxford, 2000.
[129] Bernard, L.; Kanellopoulos, I.; Annoni, A.; Smits, P.; The European Geoportal - One
Step Towards the Establishment of a European Spatial Data Infrastructure. Computers,
Environment and Urban Systems, vol. 29, pp. 15-31, 2005.
[130] Arms, W.Y.; Hillmann, D.; Lagoze, C.; Kraft, D.; Marisa, R.; Saylor, J.; et al.; A
Spectrum of Interoperability. D-Lib Magazine, vol. 81,2002, www.dlib.org/dlib/
january02/ arms/01arms.html.
[131] Gordon, D.; What Is Interoperability and Why Does It Matter?.2003,
http://resource.mapinfo.com/static/files/document/1074288321775/mapinfomag_summ
er2003.pdf.
[132] Goodchild, M.; Egenhofer, M.; Fegeas, R.; Interoperating GISs: Report of a Specialist
Meeting held under the Auspices of the VARENIUS Project.1997, www.ncgia.
ucsb.edu/Publications/Varenius_Reports/Interop.pdf.
[133] McKee, L.; Buehler, R.; The Open GIS Guide. Open GIS Consortium Inc.,1998,
www.OpenGIS.org/techno/specs.htm.
[134] Stoimenov, L.;Derdevic-Kajan, S.; Framework for Semantic GIS Interoperability.
University of Facta, Ser. Math. Inform., vol. 17, pp. 107-125, 2002.
[135] Garvin, D.; Managing Quality: The Strategic Competitive Edge. The Free Press, New
York, p. 319, 1988.
[136] Lillrank, P.; Laatuajattelu Quality Thinking. 1998, Otava, Cited in Jackobson and
Tsoulos, Keuruu, 2007.
[137] Chrisman, N.R.;A Theory of Cartographic Error and its Measurement in Digital Data
Bases. Proceedings of Auto Carto, vol. 5, pp. 159-168, 1982.
[138] Goodchild, M.; Gopal, S.; The Accuracy of Spatial Databases. Taylor & Francis Ltd, p.
290, 1989.
[139] Buttenfield, B.P.; Research Initiative 3: Multiple Representations. Closing Report,
National Centre for Geographic Information and Analysis, Buffalo, 1993, http://www.
ncgia.ucsb.edu/Publications/Closing_Reports/CR-3.pdf
[140] Guptill, S.C.; Morrison, J.L.; Elements of Spatial Data Quality. International
Cartographic Association, Elsevier, Oxford, 1995.
[141] Goodchild, M.; Jeansoulin, R.; Data Quality in Geographic Information, From Error to
Uncertainty. Hermes, Paris, 1998.
[142] Assink, M.; Groenendijk, N.; Spatial Quality, Location Theory and Spatial Planning.
Regional Studies Association Annual Conference, Understanding and Shaping Regions,
Spatial, Social and Economic Futures Leuven, Belgium, April 6-8, 2009,
http://doc.utwente.nl/71952/1/Assink09spatial.pdf.
[143] www.unohrlls.org/UserFiles/File/IPoA.pdf.
Chapter 8
STANDARDS OF COMMUNICATIONS IN THE

INTELLIGENT TRANSPORT SYSTEMS (ITS)
Maaruf Ali*
Department of Computer Science & Software Engineering,
College of Computer Science and Engineering, University of Hail, Hail, Saudi Arabia
ABSTRACT
This chapter covers the latest standards in communication in intelligent
transportation systems, especially related to inter-vehicular communication. An overview
of the IEEE 802.11 protocol standard relevant to ITS communication is discussed. Along
with the latest standardizations in the Dedicated Short Range Communication, DSRC,
specification and the IEEE 1609 Wireless Access in Vehicular Environments (WAVE)
family of standards. Fully distributed technology as well as infrastructure based and
centralised cellular networks such as 3G and 4G are also briefly covered.
Keywords: IEEE 802.11p, Dedicated Short Range Communication, DSRC, IEEE 1609,
Wireless Access in Vehicular Environments, WAVE, Inter-Vehicular Communications,
IVC, Telematics, Continuous Air Interface Long and Medium range, CALM, Traffic
Message Channel, TMC, Transport Protocol Experts Group, TPEG, Multimedia
Broadcast Multicast Service, MBMS, VeINS, Vehicles in Network Simulations
1. INTRODUCTION
This chapter is a high level tutorial overview based partly on the presentation by Dressler
and Casetti [1] on inter-vehicular communications (IVC) presented at IEEE ICC 2012. This
chapter will concentrate purely on those aspects of the various standards and protocols related
to ITS communication, covering briefly some applications then the DSRC/WAVE standard.
*
E-mail: m.maaruf@uoh.edu.sa.
236 Maaruf Ali
The protocols and standards are explained with relevance to vehicular communication
including the use of centralised and distributed systems.
The need for inter vehicular communication grew out of mainly safety considerations
along with the need to maintain and increase road safety between the drivers with the ever
increasing traffic density. Initial communication was mainly of the broadcast type from a
central station disseminating traffic information. The need for a feedback channel between
vehicles was realized from early on and then consequently the need for inter-vehicular
communication for safety applications such as maintaining safe braking distances. The main
constraints of a vehicular network are the need to develop a system within the constraints of a
real-time system, keeping network latency to safe reaction time limits.
Communication between vehicles may take place using the existing standards whether
via cellular means such as 3G or 4G or by means of satellite communications. This chapter
will address specifically those radio communication technologies pertinent to inter-vehicular
communication.
2. THE TRANSPORTATION SYSTEMS

For a better understanding in devising suitable communication systems suited to the
application, the differing types of transportation systems need to be defined.
Figure 1. The Intelligent Transportation System (ITS) network (original diagram inspired by [2]).
Standards of Communications in the Intelligent Transport Systems (ITS) 237
Intelligent Transportation Systems deals with the long distance transmission of digitised
multimedia or hypermedia realtime and non-realtime information, that is, telematics between
vehicles and vehicles to other network entities, stationary and non-stationary, terrestrial and
non-terrestrial. Thus the flow of information also includes to those encompassing the railway,
airway and maritime domains. Their fundamental basis of access and disseminating this
information is currently based on radio transmission using specialised protocols and currently
established international standards.
Figure 1, shows the rather complex inter-relationships between the various and rather
diverse vehicular networks ranging from satellites, cars, ships, planes, trains to cellular
networks and the LAN and MAN computer networks. The arrows between the various
network entities have been omitted so as to make the diagram less cluttered. It can clearly be
seen that communication between the network entities will require the use of different
protocols, frequencies and power of transmission. The ITS network is clearly not static and is
constantly evolving in terms of both technology and diversity of uses, beyond its adoption by
merely automobiles.
3. CURRENT AUTOMOTIVE RELATED ITS PROJECTS

These are the three major projects for ITS currently under investigation. These are the
Dedicated Short-Range Communications (DSRC) system, the Telematics using the 5 GHz
and 63 GHz radio bands and the Continuous Air interface Long and Medium range (CALM)
initiative. These three projects also come under the umbrella of the European Commission‟s
eSafety drive specially related to road safety.
The DSRC is designed for short range communication between the vehicle and the road
side curb, specifically at designated places such as near parking meters or toll gates. Using
such a system will also mean having to accept electronic forms of payment too for such things
as parking charges or toll fees.
Specific wireless systems for ITS are being tested and trialled to offer telematics using
the 5 GHz and 63 GHz frequencies. Using such frequencies by their nature means the systems
are designed to be line-of-sight operation only using relatively low power. However, they do
offer higher bandwidth and hence higher data rates to carry complex and detailed information.
To ensure international operability, these frequencies and channels do off course have to be
standardised and agreed up, specially ratified by the International Telecommunications
Union.
The CALM system is designed to use both existing networks and future networks to
provide long and medium range continuous services between the vehicles and to their non-
vehicles. They can use the existing cellular networks such as 3G, 4GLTE, 4G, WiFi and or
WiMax. They are also designed to operate on the newer standardised ITS frequencies of 5
GHz and 63 GHz. Because of the higher frequencies and greater bandwidth, both safety and
entertainment applications for the passenger and driver can be easily accommodated. The
type of data may include 3D video and immersive multi-surround sound.
It should be noted that the railways have adopted the use of GSM for signalling, known
as GSM-R, whose frequencies are mainly in the 400 MHz band. Thus ITS systems for
238 Maaruf Ali
railways should also be able to access the GSM-R network. The main difference between
standard GSM and GSM-R is the use of different frequency bands.
The ITS system not only encompasses land based systems but also the maritime and
aeronautical systems extending to communication with satellites. Typical services include
providing cellular and WiFi onboard connectivity onboard aeroplanes to those serving the air
traffic controllers for critical real-time navigation purposes. Maritime usage also concerns
primarily navigation and safety use both for sea and for the inland waterways of many
countries. For both the maritime and aeronautical specifications, these are governed by the
European Telecommunications Standards Institute, ETSI.
3.1. Infrastructure-Assisted Data Exchange
This is using the existing technologies to exchange data between vehicles and between
vehicles and other network entities, whether terrestrial, maritime or space based. These
technologies are currently based on 2G, 3G and 4G networks. These technologies are used
mainly by navigation units in vehicles. But there are off course limitations with such
technologies, such as radio coverage outages when transiting through a tunnel or loss of
satellite signal when driving under heavy foliage. Thus other technologies are being
experimented with and trialled such as Ad Hoc Multihop Broadcasting to counteract the
problems of loss of cellular coverage including the problems of deep fading.
Willke et al., [3] partitions into two the major types of data that need to be exchanged in
inter-vehicle communications, these being non-safety and safety messages. Safety messages
are identified as those that alert the driver to danger and require immediate action. These are
best served by short-range radio broadcast with its low network latency. The non-safety data
encompasses such things as entertainment data as one example. Such non-critical and non-
time sensitive information can make use of the existing cellular infra-structure as they are
network delay tolerant.
Historically, it all started with TMC, traffic message channel and TMC Pro. This was a
centralized management of traffic information using FM radio broadcasting. The major
drawback was the limitation of providing an uplink channel from the vehicle to the central
station, there was not any. However, emergency service providers like the Police force could
update information to the central station. The other major drawback was the limited data
channel rate. Some of these problems were addressed by the implementation of TPEG
(Transport Protocol Experts Group), founded in 1997.
Experiments have already been carried out using infra-structure based network utilising
the 3G/3.5G network. However, MBMS (multimedia broadcast multicast service) of the 4G
network system is ideal to be utilised for inter-vehicular communication. MBMS is currently
not being deployed as no application has been found that requires it, however, this may
change with more pressure from the inter-vehicular communications pressure group. Using
MBMS can be good for non-safety applications as the end-to-end delay being 120 ms is
considered too slow for safety critical applications.
Many problems need to be addressed in considering the type of technology to be adopted
for ITS. If wireless communication is used, then interference is a big factor, as it leads to
errors in the data rate due to packet collisions. These have to be minimized. Utilising Ad Hoc
networks has its own particular topological concerns such as the length of the multi-hop
forwarding link, the number of unidirectional links and co-existence with multi-radio and
multi-network systems. The very nature of the ITS is both mobility and fluidity of the
network topology with often disappearing links as the cars go into tunnels, or cars following
each other take different directions and routes. This calls for re-routing by the routing
protocols. The speed of the vehicles is also very variable and this leads to a constantly
changing and thus a dynamic topology. Privacy and security is also an increasing matter of
concern and the permanent presence of an internet connection is also not guaranteed.
3.2. DSRC/WAVE (Dedicated Short Range Communication/Wireless Access

in Vehicular Environments)
The IEEE 1609 specification is the latest current standard in ITS communication that
covers vehicular networks. As shown in Figure 2, all the layers are designed for vehicular
communication including accessing the network and security features. Short distance
communication is dealt with by the new IEEE 802.11p[3] standard, this being similar to the
WiFi IEEE 802.11a standard.
IEEE 1609.1: Core Systems IEEE 1609.3: Network Services

IEEE 1609.2: Security IEEE 1609.4: Channel Management
Figure 2. The IEEE 1609/WAVE protocol stack (adapted from [4]).
The physical layer is almost similar to the IEEE 802.11a standard and uses orthogonal
frequency division multiplexing (OFDM) using 16 level quadrature amplitude modulation
(16QAM) for its modulation. This leads to reduced inter-symbol interference due to multipath
and Doppler shift effects, which is commonly faced by fast moving vehicular traffic. The
timing parameters have also been doubled and the channel bandwidth reduced to 10 MHz
instead of 20 MHz. This consequently leads to a reduced throughput ranging from 3 to 27
Mbit/s instead of 6 to 54 Mbit/s. The system is designed to operate for close inter-vehicular
traffic and thus the effective radio wave operating distance is set in the standard to be
effective up to a range of 1000 metres whilst the vehicle is travelling at a maximum speed of
200 km/h (125 mph). So the timings had to be changed and the protocol made more robust
because of the speed of travel of the vehicles. Thus to make the protocol more robust the
240 Maaruf Ali
timings had been doubled. Interference considerations were also an important factor, thus 5.9
GHz has been reserved for the USA and Europe.
The IEEE 802.11p MAC layer has features with extensions to the IEEE 802.11a standard.
The MAC layer has randomized MAC addresses, quality of service (QOS) built-in with
priorities, features that are also present in IEEE 802.11e. The MAC layer also supports multi-
channel and multi-radio access with a new ad hoc mode. The protocol allows for both non-
BSS and BSS mode of operation.
QoS is an important factor so that safety critical messages can be transported and relayed
as a priority whilst entertainment type of data can be relegated to have a lower class of
service. These differing QoS and access control mechanisms have been borrowed from the
IEEE 802.11e protocol.
The WAVE channel spacing for both the US and Europe are shown in Figure 3. The
frequency range and channels are strictly regulated but there will be no licensing cost to the
user. Both the US and the European channel allocation allow for four service channels to be
used by applications and one control channel. However, the US having an extended frequency
range has two further channels, channel 172 for “Critical Safety of Life Use” and channel 184
for “Hi-Power Public Safety” applications. With both the US and European systems the
channel are of 10 MHz bandwidth. Even though they may have the same channel number,
they do not all carry the same service of operation.
ETSIES 202 663 V1.10 (2010-01): Intelligent Transport Systems (ITS) is an European
Profile standard from the physical and medium access control layer of Intelligent Transport
System operating in the 5 GHz frequency band.
1 control and 4 service channels to be used by application
In Europe, ETSI has reserved 5 channels of 10 MHz bandwidth each
Figure 3. The WAVE channel spacing (adapted from [5]).
The control channel relays management and safety information. Two-way

communication takes place on the service channel. Slot management is critical and the
synchronization of timing is achieved using the timing signal from the atomic clock from the
GPS satellite. The standard specifies a 100 ms sync interval including spending 50 ms on the
control channel (CCH). Single radios are used, so they have to switch to the CCH at specific
known times. To allow for jitter, the timeslots all start with a guard interval.
Car-2-Car communication is not defined in DSRC/WAVE but a frame format has been
provided to carry such messages. First applications include an application that make the drive
aware of his/her current surroundings which include the time, speed, position and heading of
other vehicles. This is known as “Cooperative Awareness Messaging” or CAM.
To summarize, there are two methods of relaying traffic information systems. Those
which belong in the domain of a centralized system, such as using FM Radio, DAB (digital
audio broadcasting), 3G/3.5G and those which belong in the decentralized domain, such as
using broadcasting, Ad Hoc routing and peer-to-peer networking.
Historically an early traffic information system was known as “SOTIS” or “Self
Organizing Traffic Information System”. This was a fully distributed system with no
communication infrastructure. It was a basic and simple system with no congestion control. It
worked on exchanging a local knowledge base using beacons. The beacons would aggregate
all the sensed and received data with the beacon transmitting every five seconds. This system,
however, suffered from many shortcomings such as limitation of channel capacity and beacon
interval. To gain market penetration of any system, it has to be realized that real networks are
heterogeneous and cross platform interoperability is a must for future adoptability and market
penetration for the system to be adopted by the public. Thus any system that is to be designed
must be able to operate with the existing legacy systems.
Further details regarding beaconing approaches may be found in [6,7,8] and peer-to-peer
in [9,10].
Peer-to-peer traffic information system may also be implemented using a hybrid 3G/4G
network [5]. The idea behind this was to use all the cars as a massively distributed system. To
ease the data processing, the data for the geographical map can be tiled with each vehicle
carrying a portion of the map.
Communication between vehicles will also be dependent on traffic density, speed of
movement, terrain, obstructions in the propagation and permitted frequencies of operation.
There will also be the sensitive issue of different cultural patterns in driving behaviour
between nations, age groups, socio-economic background of the driver and the gender. Such
factors may affect how communications between these vehicles are in fact modelled. It may
be appropriate to use different models for different road types and nations, for example for
fast moving autobahn traffic versus those suited to slow moving urban traffic with dense
populated traffic volume interacting with the data services of the conurbation. Behaviour and
swarm studies are also applicable in modelling group driver behaviour.
Some vehicle networking simulations that can be explored by the independent researcher
include such tools as: VeINS – Vehicles in Network Simulation. This simulator is based on
tools such as OMNeT++ for network simulation and SUMO for road traffic micro-simulation.
Further information on VeINS can be found in [12, 13].
Security and Privacy is also a major research topic in its own right and is intimately
linked with the various inter-vehicular communication protocols. Tied in with this also
probably the most important issue and these are human computer interaction and usability
issues. The protocols must be able to incorporate all these features for its adoption by humans
from diverse backgrounds.
242 Maaruf Ali
4. SECURITY AND PRIVACY IN INTER-VEHICULAR

COMMUNICATION (IVC)
Security provisioning is important not only for security but to actually protect the data in
the hazardous and noisy electromagnetic environment that exist in the fast moving and
contentious vehicular and road traffic environment. The key factors that need to be considered
in a practical working IVC network are such things as the identity of the user or IVC network
entity and its management and privacy protection. As mentioned above, it is not always a
question of protection from ominous users or situations but simply the protection of data in a
very noisy environment. Messages exchanged between IVC network components also require
authentication to prevent spoofing or “man-in-the-middle” attacks from occurring. The data
being transported between and through the IVC network need to be protected and also be
consistent. This off course requires the use of standardized protocol with international
recognition. Thus the IEEE 1609/WAVE protocol is one very important and likely contender
for global adoption in the IVC network. The European equivalent IVC protocol is the ETSI
Working Group 5. Any IVC network will off course require both the determination and
secure exchange of location services.
The IVC network must be a tolerant network such that it can prevent attacks from
happening in the first place, as prevention is always better than the cure. This also requires the
detection of attacks and recovering from them ideally with the user data or signalling
information intact. The IEEE 1609/WAVE Protocol does have a security layer in the protocol
stack, this being detailed in the IEEE 1609.2 Security specifications. Protection from
undetectable attacks, that is, those that are unknown to the user are tackled by focussing on
prevention; whereas attacks that are aggressive in nature are best dealt with by focussing on
detection. The issues and experiences from wireless mobile communications and WiFi usage
can be applied to the wireless IVC network. Extensive knowledge can be applied from the use
of MANETs (Mobile Ad Hoc Networks) and VANETs (Vehicular Ad Hoc Networks) as
well. The concepts of encryption using keys, digital signatures, data integrity mechanisms
such as message authentication have all been adopted to ensure security in the IVC network.
Figure 4 shows how a smart vehicle integrates various modules necessary for the proper
functioning of an IVC network. Some of the features shown are actually present in most
modern manufactured cars these days, such as radar and GSM mobile communication. The
presence of a hardware security module (HSM) clearly shows the importance attached to
network security and integrity. Cryptographic data will be processed by the HSM. Each
module requires a high end processor in its own right for the realtime processing of the vast
amounts of data that need to be processed. Radar both in the front and rear are necessary for
collision avoidance for example. It should also be noted that there is an event data recorder,
much like the black box of an aircraft. This can keep a log of the journey taken and provide
vital data during an breakdown or accident. The information from the vehicular black box can
also be useful for the insurance companies where claims are disputed. The vast amounts of
realtime data generated need to be communicated between vehicles for realtime traffic
management and accident avoidance.
Figure 4. The smart vehicle (adapted from [14]).
5. KEY APPLICATIONS OF IVC

A vitally important application of IVC is that of emergency vehicles switching traffic
lights to force a green so that it can get to an emergency site as quickly as possible or to a
hospital. This can actually be taken further in that traffic lights further across the route of the
ambulance for example can be controlled to regulate traffic and relieve congestion in certain
hotspots along the trajectory of the emergency vehicle.
Dynamic pricing [14] is also one possible revenue generating stream likely to be adopted
by the various international highway agencies. “The driver (possibly assisted by a navigation
system) decides on a route; the payment of any tolls automatically occurs when entering the
toll road or bridge” [14]. This would require the seamless communication between the vehicle
and the toll gate charging communication system.
A stationary vehicle also needs to be aware of its surrounding traffic. This can be
important if the car is hit by another car. Another example may be that the car can track and
monitor another car that has been involved in a criminal incident. The speed, data and time
may be used in a court case or help in the identification of the fleeing vehicle [14].
Verifiable multilateration is also possible with IVC networked vehicles [14]. If there are
four such vehicles then the target vehicle can be positioned in 3d space. If three IVC
networked vehicles are present, then the target vehicle can be positioned in 2d space. The
positioning accuracy can be further augmented with GPS and other satellite navigational data.
What most drivers will feel happy with is the concept of “cooperative driving”. This is
when a vehicle has a special “electronic token” that allows it to safely navigate through a
busy blind crossing or approach a busy congestion safely. This is made possible because of
244 Maaruf Ali
the sharing and exchanging of vehicular positioning information and access rights and
privileges with each other using a standardized IVC protocol.
6. DISCUSSION
The need for a robust, secure and internationally agreed standard for inter-vehicular
communication is now reaching the stage of absolute necessity. This is specially driven by the
increasing traffic density in urban cities and the need to manage traffic flows safely and
efficiently.
Routing of traffic needs to be managed at peak times to avoid congestions and also to
minimize potential accidents at junctions. Safety is of paramount importance as well as the
smooth and fast flow of traffic through cities.
If vehicles are considered as network entities then the reliable inter-vehicular
communication becomes a necessity. This dictates the adoption of an internationally agreed
standard or standards. The current scenario is of two emerging standards, one based on the
IEEE 1609/WAVE protocol driven by the North American IEEE institution and the other
standard is the European ETSI Working Group 6. Both standards share similarities and have
security as one of their utmost concern. It is envisioned that future ITS communication will
have to adopt a multi-standard approach, just like cellular communication.
All the applications rely on the secure and robust transport of data between vehicles both
moving and importantly also sedentary ones. Communications between other types of
vehicles such as aircrafts ships and trains are also incorporated into the standards.
Communication is not just limited to terrestrial network entities but also with space bound
vehicles such as satellites orbiting at various altitudes.
Applications which cater for the emergency services have a special place in the protocol
standards, an example being to allow vehicles such as ambulances safe and fast passage
through congested traffic including the control of traffic ahead of the emergency services
vehicle. Monitoring of vehicles by each vehicle can also be used for intelligence gathering
and crime prevention.
Accurate estimation of a vehicle‟s position is essential for the safe monitoring of the
trajectories of vehicles to avoid collisions and thus avoid multiple fatalities. This is a special
application area that is being researched intensively. The position of a vehicle can be
determined by multilateration in three dimensions using four other vehicles equipped with
ITS and IVC systems. Again the position of a vehicle can be used for security and crime
prevention and detection. GPS and other satellite navigational systems can be augmented with
terrestrial and inertial systems for added safety derived from redundancy.
Most applications are safety and traffic flow management based. Revenue generating
applications are also starting to appear with the concept of toll collecting based on routing
information provided or gathered from the vehicle.
CONCLUSION
The ITS system is a complex, ever evolving network that not only consists of vehicles but
other types of mobile and non-mobile entities. This chapter has presented the latest current
projects and specifications being applied and adopted for ITS. An overview of the DSRC and
CALM systems were given. Technologies covered included the use of the current cellular
networks such as the GSM-R network adopted for the railway network and those of the 3G
and 4G networks for ITS use. An overview of the IEEE 1609/WAVE specification was also
discussed including the coverage of the lower layers of the protocol stack. For an efficient and
functioning ITS to operate, interoperability with existing communications protocols and
emerging protocols must be ensured. The IEEE 1609/WAVE specification appears to be the
successor and the most likely communication protocol for inter-vehicular communication to
be internationally adopted as a ratified standard.
REFERENCES
[1] Dressler, F.; Casetti, C.; Standards, Protocol Design, and Integrated Security Metrics.
IEEE Tutorial Lecture on Inter-Vehicular Communication (ICC-2012), http://host.
comsoc.org/freetutorial/wiley5/wiley5.html [Accessed 23rd Nov., 2013].
[2] http://www.etsi.org/technologies-clusters [Accessed 23rd Nov., 2013].
[3] Jiang, D.; Delgrossi, L.;IEEE 802.11p: Towards an International Standard for Wireless
Access in Vehicular Environments. IEEE Vehicular Technology Conference (VTC
Spring 2008), doi:10.1109/VETECS.2008.458, pp. 2036-2040, 11-14 May 2008.
[4] Willke, T.L.; Tientrakool, P.; Maxemchuk, N.F.; A Survey of Inter-Vehicle
Communication Protocols and their Applications. IEEE Communications Surveys and
Tutorials, doi: 10.1109/SURV.2009.090202, vol. 11, no. 2, pp. 3-20, Second Quarter
2009.
[5] Uzcategui, R.; Acosta-Marum, G.; Wave: A Tutorial. IEEE Communications Magazine,
doi: 10.1109/MCOM.2009.4939288, vol. 47, no. 5, pp. 126-133, May 2009.
[6] Korkmaz, G.;E kici, E.; Ozguner, F.; An Efficient Fully Ad-Hoc Multi-Hop Broadcast
Protocol for Inter-Vehicular Communication Systems. IEEE International Conference
on Communications (ICC'06), doi: 10.1109/ICC.2006.254764, vol. 1, pp.423-428, June
2006.
[7] Sommer, C.; Tonguz, O.K.; Dressler, F.; Traffic Information Systems: Efficient
Message Dissemination via Adaptive Beaconing. IEEE Communications Magazine,
doi: 10.1109/MCOM.2011.5762815, vol.49, no.5, pp.173-179, May 2011.
[8] Kargl, F.; Schoch, E.; Wiedersheim, B.; Leinmuller, T.; Secure and Efficient Beaconing
for Vehicular Networks.5thACM International Workshop on Vehicular Ad-Hoc
Networks (VANET 2008), ACM, San Francisco, USA, pp.82-83, 2008.
[9] Lee, K.C.; Uichin L; Gerla, M.; TO-GO: Topology-Assist Geo-Opportunistic Routing
in Urban Vehicular Grids. 6th International Conference on Wireless On-Demand
Network Systems and Services (WONS 2009), doi: 10.1109/WONS.2009.4801842,
pp.11-18, 2-4 February 2009.
246 Maaruf Ali
[10] Rybicki, J.; Scheuermann, B.; Koegel, M.; Mauve, M.; PeerTIS: A Peer-to-Peer Traffic
Information System. In Shorey, R.; Weimerskirch, A.; Jiang, D.; Mauve, M.;Ed.,
Vehicular Ad-Hoc Networks. ACM, pp. 23-32, 2009.
[11] http://veins.car2x.org/ [Accessed 23rd Nov., 2013].
[12] Sommer, C.; German, R.; Dressler, F.; Bidirectionally Coupled Network and Road
Traffic Simulation for Improved IVC Analysis. IEEE Trans. on Mobile Computing, vol.
10, no. 1, pp. 3-15, Jan. 2011.
[13] Sommer, C.; Dietrich, I.; Dressler, F.; Realistic Simulation of Network Protocols in
VANET Scenarios. 26th IEEE Conf. on Computer Communications (INFOCOM
2007), IEEE Workshop on Mobile Networking for Vehicular Environments (MOVE
2007), Poster Session, Anchorage, AK, pp. 139-143, May 2007.
[14] Hubaux, J.P.; Capkun, S.; Luo, J.; TheSecurity and Privacy of Smart Vehicles. IEEE
Security and Privacy, doi: 10.1109/MSP.2004.26,vol. 2, no.3, pp. 49-55, May-June
2004. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.127.1164&rep=rep1&
type=pdf.
PART II. SMART TECHNOLOGIES APPLIED IN
THE ELECTRICAL AND HYBRID VEHICLES
Chapter 9
POWER ELECTRONICS AND DRIVES USED IN

AUTOMOTIVE APPLICATIONS
Ersan Kabalci*
Electrical and Electronics Engineering Department, Faculty of Engineering and
Architecture, Nevsehir University, Turkey
ABSTRACT
This chapter deals with power electronics and drives used in electric vehicles (EV).
The power electronic devices are classified in four categories as rectifier, inverter, DC-
DC converter, and cycloconverter according to AC and DC conversions together and vice
versa. The converters are mostly utilized after the rectifier block to ensure supplying the
battery charger with a stable DC voltage, at the output of battery or fuel cell, and before
DC motor drives. The propulsion systems of electric, hybrid electric and plug-in hybrid
electric vehicles are based on electrical machines. An electric machine used in an EV is
chosen according to its construction structure, efficiency, maintenance requirements,
electromagnetic interference (EMI), reliability, and speed range parameters. The rapid
improvements seen in electrical machines and power electronics devices also contribute
the development of EVs. There are a few electric motors among others are known as
workhorse in power-train applications. The induction machines (IM), permanent magnet
synchronous machines (PMSM), and brushless DC machines (BLDC) are the several
electric machine types used in EV power-train applications with specific controllers. The
machine controllers are power electronics systems that are combined with sensors and
observers in order to ensure the wide speed range in constant torque situations and
extended speed range for constant power requirements. Although the power electronics
devices and motor drives cover a wide research area, this chapter is dedicated to specific
applications used in EV technology. The related topics are arranged from plug to
propulsion system for an EV.
Keywords: Power Electronics, Converter, Inverter, Cycloconverter, Electric Vehicle, Electric

Machines, PMSM, BLDC, ACIM
*
E-mail: kabalci@nevsehir.edu.tr.
250 Ersan Kabalci
NOMENCLATURES
ARCP Auxiliary Resonant Commutated Pole
BLDC Brushless DC Machine
CCM Continuous Conduction Mode
CHB-MLI Cascaded H-Bridge Multilevel Inverter
CSI Current Source Inverter
DCM Discontinuous Conduction Mode
DC-MLI Diode Clamped Multilevel Inverter
EMF Electromagnetic Force
EMI Electromagnetic Interference
EV Electric Vehicle
FACTS Flexible AC Transmission Systems
FC-MLI Flying Capacitor Multilevel Inverter
FCV Fuel Cell Electric Vehicle
FOC Field Oriented Control
HEV Hybrid Electric Vehicle
ICE Internal Combustion Engine
IM Induction Machine
MLI Multilevel Inverter
PEEM Power Electronics and Electrical Machines
PEV Plug-In Electric Vehicle
PHEV Plug-In Hybrid Electric Vehicle
PI Proportional-Integral
PID Proportional-Integral-Derivative
PMBL Permanent Magnet Brushless
PMSM Permanent Magnet Synchronous Machine
UPS Uninterruptible Power Supplies
VSI Voltage Source Inverter
1. INTRODUCTION
The plug-in electric vehicles (PEV) are known as a subcategory of electric vehicles that
also include pure electric vehicles (EVs) and plug-in hybrid vehicles (PHEVs). Furthermore,
the PEVs are defined as electric vehicle conversions of hybrid electric vehicles (HEVs) and
conventional internal combustion engine (ICE) vehicles. Although a HEV has battery
configuration that is consistently recharged by using the power obtained from the ICE and
regenerative braking. Since HEVs cannot be recharged from any off-vehicle electric energy
source, they are not categorized in plug-in electric vehicles. A PEV can be defined as an
electric motor vehicle that obtains the required energy for propulsion over the electric grid
and/or the charged batteries located on board [1-3].
The PEVs require a detailed power electronics and electrical machines (PEEM) control to
operate efficiently as other advanced electric drive vehicles such as HEVs, PHEVs, fuel cell
electric vehicles (FCVs), and EVs. The power electronics devices convert the grid electricity
Power Electronics and Drives Used in Automotive Applications 251
to the required ratings and various current waveforms as direct or alternating. The electric
drive system in EVs is based on direct current (DC) and alternative current (ac) energy
conversion stages. The complete drive system is defined with functions of converter, battery,
inverter, electric motor, and wheels. The inverter converts the DC line voltage to AC and
supplies the electric motor to generate propulsion in wheels [3,4]. On the other hand, the
power electronics and electrical machine stages of PHEVs are more complex according to
HEVs. This vehicle configuration uses two electrical machines; one as a generator connected
to the engine and another as a motor to drive the wheels. Each of these machines is connected
to an inverter that runs at a higher voltage levels comparing to the battery.
The boost converter increases the DC voltage that will be applied to inverter. Main
components of a hybrid and an electric vehicle include motors, motor controllers, DC-DC
converters, and inverters that condition the electrical energy between the power supply (a fuel
cell or battery) and electric motor to provide power to various components, sensors, control
systems, and other interface electronics. The electric motor drive converts the stiff DC battery
voltage to DC (for DC motor) or AC (for AC motor) voltage levels with controllable
amplitude and frequency. Besides performing the propulsion of PEVs with induction motors,
the DC motors are used in order to manage the auxiliary systems such as air conditioning
(A/C) compressor and pumps [5-7].
The input command of the motor drive is converted to torque command that sets the
operating point parameters of the electric motor. The torque command, in conjunction with
the feedback signals from sensors, performs by adjusting the turn-on and turn-off transitions
of the power switches through the drive system. The motor drive then shifts the power at the
desired voltage and frequency to the motor, which in turn delivers the desired torque and
speed for propulsion [7,8].
This section covers power electronic converters and inverters, DC and AC electric motors
that are most widely utilized in vehicle technology and motor drives specifically for PEVs.
The PEEM subsections focus on actual and practical electric vehicle technologies instead of
pure power electronics and introduces the application areas of widely known power
electronics in PEVs.
2. POWER ELECTRONICS USED IN PEVS

Power electronics is a wide research area owing to its circuit topologies converting the
power from AC to DC or vice versa. In addition to current waveform conversions, AC-AC
cycloconverters or DC-DC converters are used to generate the required various frequency,
current, and voltage levels. The power electronics circuits and components used in a PEV are
depicted in Figure 1 where all the components are introduced in the following subsections.
Even though the electric motor in the figure is shown as actuating the rear wheels, the front
wheel propulsion is widely used referring to actual automotive technology.
The main topics of power electronics required for a PEV design cover:
 Rectifiers: The rectifiers are used in ac-DC conversion stages of PEVs where the grid
electricity is firstly met with this device. The grid voltage obtained from plug is
110V/60Hz or 220V/50Hz single-phase AC input in household application of PEV.
Rectifiers are analysed in ideal circuit topology and uncontrolled or controlled power
252 Ersan Kabalci
switch topologies. Since the automotive applications are mostly based on

uncontrolled rectifiers, controlled rectifiers will not be analysed in this section.
 Power Converters: Power converters are classified according to their input and
output types namely DC-DC converters and AC-AC cycloconverters. The DC-DC
converters are designed to increase or decrease the input voltage stage to desired
value at the output. Therefore, a device known as buck or step-up converter
decreases the output value according to a high input voltage while boost or step-up
converter generates an increased output voltage. The cycloconverters are used to
meet high voltage and high power requirements of large motors where frequency and
voltage levels are adjusted by converter. The AC-AC converters are not used in
electric vehicles and the frequency or AC voltage requirements are provided by
inverters that may be named as DC-AC converters. The cascaded inverter topologies
are introduced in brief besides widely used full bridge inverters in PEVs. Control
techniques of inverters are also focused in a following subsection.
 Motors and Motor Controllers: Motors or electric drives used in electric vehicles are
classified in two main categories as brushed or brushless. Induction motors, brushless
DC motors (BLDC), and permanent magnet brushless (PMBL) motors that are
widely used PEV electric drives are analysed in terms of physical structures and
controllers in a separate subsection.
 Controller and sensors: Almost all the power electronics in PEV require control
signals as met in industrial applications. The required control signals are generated
by microcontrollers and commutate power electronic devices with switches to
provide fixed speed or torque. To obtain the instant control action, interface
electronic devices are used besides current and voltage sensors located in the related
parts.
Figure 1. Power electronics and electrical components used in a PEV.

The power electronic as being a discipline of electronics itself, the devices mentioned
above are mostly known by anyone who is familiar to this research area. Therefore, the
following topics are dedicated to specific power electronics used in PEVs instead of analysing
entire variations of devices.
3. RECTIFIERS
The rectifier is a circuit that converts AC to DC owing to diodes in uncontrolled structure
or thyristors in controlled structure. A general overview is given in Figure2 where the input
utility voltage is shown with Vin, while f and q stand for line frequency (50Hz/60Hz) and
phase number, respectively.
Figure 2. General rectifier overview.
The half-wave rectifiers are used in low power or industrial applications to isolate load
from power line and are widely preferred because of its low cost installation. The power
output and therefore, the rectification efficiency is quite low besides its low transformer
utilization ratio and high ripple factor requiring a robust filtering[9-11]. The full-wave or
bridge rectifiers generate DC output voltage during each half cycle of a period on the contrary
of half-wave rectifiers [12, 13].
Figure 3. Single phase full-wave rectifier.
The full-wave rectifiers can be accomplished with two different circuit topologies that are
known as center-tapped and bridge rectifier. The bridge rectifier depicted in Figure3 requires
254 Ersan Kabalci
four diodes while it is enough to use two diodes in the center-tapped rectifier due to
transformer structure [9]. In the full-wave rectifier with center-tapped transformer, each diode
acts as a half-wave rectifier using the tapped half of line transformer and output generates a
full-wave waveform to supply the load [11]. The average value of the load voltage (VL) is
defined as Vdα in Figure 3 and calculated as given below;
T
1
Vd   vL (t )dt
T0
(1)
If this generic equation is rearranged for full-wave bridge rectifier;

1
Vd 
 Vm sin t d t
0 (2)
is obtained. Therefore,
2Vm
Vd   0.636  Vm (3)

The root mean square (rms) value of load voltage,
Vm
VL _ rms   0.707  Vm (4)
2
Figure 4. Single-phase full-wave rectifier simulations with RL loads.

The rms value of a full-wave rectified voltage should be equal to that of the original AC
voltage as proved in Equation (5) [11-13]. The supply and load waveforms of rectifier with
RL load are seen in Figure4. Each diode pairs (D1D3-D2D4) are switched on during positive
and negative half cycles of source voltage (Vd) provided by transformer. Therefore, the output
current flow is continuously obtained in sinusoidal waveform and there is not any phase shift
occurred with output voltage. The load current waveform seen in Figure4 is obtained with an
R-L load where a 2mH inductance is selected to observe effect of the inductance on current
waveform.
The load voltage is seen in sinusoidal waveform while Id is distorted by odd harmonics
and a phase shift is occurred between Vd and Id. There a distortion is also seen in source
current because of source inductance. The bridge diode currents are seen in square waveform
in high inductive loads that is a usual case seen in industrial applications [9,11,13].
4. DC-DC CONVERTERS
The DC-DC converters in PEVs are required to keep the output voltages stable against
varying source voltages supplied by fuel cells or batteries. Furthermore, the variable and/or
stable torque and speed requirements under several conditions also involve DC-DC
converters. All the electronic equipments in PEVs require a fixed DC input voltage. This
requirement is accomplished by decreasing or increasing the input voltage of the converter
and performing circuit is called with its operating principle where a buck converter (step-
down) adjusts the output voltage by decreasing the source voltage or a boost converter (step-
up) that increases the output voltage to a fixed value. There is a third type of converter
namely buck-boost is widely used in PEVs that performs specific features of both buck and
boost converters together. These converters are constituted with a high frequency switching
devices and reactive components to ensure operating in the continuous conduction mode
[14,15].
The converters are mostly utilized after the rectifier block to ensure supplying the battery
charger with a stable DC voltage, at the output of battery or fuel cell, and before DC motor
drives that are fed by auxiliary batteries as illustrated in Figure1 [16-17]. The following
subsections deals with the most widely used converter topologies such as buck converter,
boost converter and buck-boost converter with design criteria for PEVs.
4.1. Buck (Step-down) Converter
The buck converter, which is also known as step-down converter, is shown in Figure5. It
involves input source (Vs), controlled switching device (S), a diode (D), low-pass filter with
L and C, and load resistance. Inductor and capacitor ensure to generate output voltage at a
limited ripple ratio and filters the output according to requirements of the load. The filtering
operation affects the circuit analysis mode depending to inductor value. The inductor current
never drops to zero if the inductor value is high enough and this analysis mode is called
continuous conduction mode (CCM). In case of inadequate inductor values, inductor current
drops to zero for a while in each half-cycle where this mode is known as discontinuous
256 Ersan Kabalci
conduction mode (DCM). There are separate analysis required for CCM and DCM and the
relation between these modes is described with boundary analysis method.
Figure 5. Circuit diagram of buck converter.
The relationship among the input voltage, output voltage, and the switch duty ratio D can
be derived from the inductor voltage VL waveform. The DT in Figure6 expresses duty ratio or
the interval where the switch is on while the left interval shows off state. Inductor and output
currents are seen in the first curve of Figure6 [9-11].
Figure 6. Waveforms of buck converter.
The voltage equation of the circuit is calculated by using maximum and minimum
inductor currents,
 V  VC 
I max  I min   S  DT (5)
 L 
 V 
I min  I max    C  (1  D)T (6)
 L 
VC  DVS (7)
The minimum inductor value providing the converter to operate in CCM is determined
as,
 TR 
L    1  D  (8)
 2 
4.2. Boost (Step-up) Converter
A boost converter, which the circuit diagram is shown in Figure7, is a power converter
with an output DC voltage greater than its input DC voltage. It is a class of switching-mode
power supply containing at least two semiconductor switches (a diode and a controlled
switch) and at least one energy storage element (capacitor and/or inductor). Filters made of
capacitors are normally added to the output of the converter to reduce output voltage ripple
and the inductor connected in series with the input DC source in order to reduce the current
ripple. The diode is reverse biased when the switch is on and therefore conduction does not
occur in this state. The load is supplied in the off state of power switch. The value of output
capacitor should be high enough to obtain a stable output voltage in steady-state analysis
[18,19].
The maximum and minimum values of inductor currents are seen in Figure 8 where
currents of switch, capacitor, and load voltage are also given respectively.
The boost converter is also analysed for CCM and DCM, and the relation between these
modes is described with boundary analysis method. The inductor current (iL) is differentially
increased during on state,
diL VS
 (9)
dt L
Figure 7. Circuit diagram of boost converter.

258 Ersan Kabalci
Figure 8. Waveforms of boost converter.
The capacitor voltage (VC) should be greater than VS in order to decrease the inductor
current. The inductor current in the off state of switch is calculated as follows,
diL VS  VC
 (10)
dt L
Since the increment and decrement of iL should be equal in on and off states, the voltage
relation of converter is defined as,
diL VS  VC
 (11)
dt L
4.3. Buck-Boost Converter
The buck-boost converter namely step-up/down is a popular non-isolated, inverting

power stage topology as shown in Figure 9. The buck-boost converter is selected because of
the output voltage is inverted from the input voltage, and the output voltage can be either
higher or lower than the input voltage. The diode voltage is reversed when the switch is
turned to on state and the diode current (iD) is obtained at zero value in this case.
Furthermore, the source voltage is transferred to inductor and iL increases as a function of
time in conduction mode. The connection between the source and inductor is leaved by
switching off the S and iD gets equal to iL in this state. The mentioned on and off states are
used in steady state analysis as given in the following equation for each operation mode
[18,19]. The output voltage equation is obtained by equalling the variations of inductor
currents as given in Equation(12).
D
VC  VS
1 D (12)
The operation mode is determined according to value of D where higher values than 0.5
provides to operate in boost mode and lower values to operate in buck mode. The output
voltage is negative as seen in Figure 10.
Figure 9. Circuit diagram of buck-boost converter.
Figure 10. Waveforms of buck-boost converter.
5. INVERTERS
The main objective of the power inverters is to generate an AC output waveform by using
the DC input. Several industrial applications such as uninterruptible power supplies (UPS),
static VAR compensators, active filters, flexible AC transmission systems (FACTS), and
voltage compensator require AC waveforms. The magnitude, frequency, and phase should be
controllable for sinusoidal AC outputs. In electric vehicles, inverters are used to provide the
260 Ersan Kabalci
required voltage and/or current for AC motors, and consequently the inverter control
algorithms perform motor control operations in terms of torque and speed. Inverters are
considered as voltage source inverters (VSI) and current source inverters (CSI) according to
the type of output generation where VSIs are widely used in many industrial applications and
novel topologies are developed to meet the power requirements.
The efficiency parameters of an inverter such as switching losses and harmonic reduction
are principally depended on the modulation strategies used to control the inverter [19-21].
There is various control techniques that can be classified into two groups such as feed-back
and feed-forward control have been proposed to reduce the switching losses and harmonics in
occurred in DC link inverters [22-24]. A complete AC motor drive and control system is
shown in Figure 11, where a microcontroller dynamically controls the three-phase full bridge
inverter according to feedback data inherited from sensors. The switches of any phase leg in
the inverter (S1 and S4, S3 and S6, or S5 and S2) cannot be switched on simultaneously to
prevent a short circuit across the DC link voltage supply as considered in the single-phase
VSIs. Similarly, the switches of any leg in the inverter cannot be switched off simultaneously
in order to avoid undefined switching states and undefined AC output line voltages in the
VSI. The phase outputs are mutually phase shifted by 120º angles. The line voltage rates of
inverter are determined according to modulation indexes (mi) which define the operating area
as in linear modulation (mi  1) or as over-modulation (mi  1) range. The line voltages are
4 3Vd
limited to ( 3Vd 2 ) of DC line in linear modulation range and are limited to ( )( ) in
 2
over modulation range [25-27].
Figure 11. A complete AC motor drive system with full-bridge inverter for EVs.
The Sinusoidal PWM (SPWM) modulator acts as an amplifier in the linear modulation
range with the gain parameter (G) seen as in Equation (13).
0.5miVd 0.5Vd
G  (13)
Vp Vt
The closed loop control operation is performed to ensure torque or speed stabilisation of
AC motor. In recent years, multilevel inverters have gained much attention in the application
areas of medium voltage and high power owing to their various advantages such as lower
common mode voltage, lower voltage stress on power switches, lower dv/dt ratio to eliminate
higher harmonic contents in output voltage and current. The multilevel inverters (MLIs) also
have the advantages of lower harmonic ratios of line-to-line voltages fed to load are reduced
owing to its switching frequencies when compared to two-level inverter topologies at the
same power ratings. The most common MLI topologies classified into three types are diode
clamped MLI (DC-MLI), flying capacitor MLI (FC-MLI), and cascaded H-bridge MLI
(CHB-MLI). Since MLIs are mostly used in large hybrid electric vehicles instead of EV or
PEVs, these topologies are not analyzed in this section. Further information about MLIs can
be found in [25].
6. ELECTRICAL MACHINES AND CONTROLLERS

The propulsion systems of electric, hybrid electric and plug-in hybrid electric vehicles are
based on electrical machines. The rapid improvements seen in electrical machines and power
electronics devices also contribute the development of EVs. There are a few electric motors
among others are known as workhorse in power-train applications. Induction machines (IM),
permanent magnet synchronous machines (PMSM), and brushless DC machines (BLDC) are
the several electric machine types used in EV power-train applications with specific
controllers. Due to the developments met in power electronics, various machine control
techniques are implemented to increase the efficiency [28-30].
An electric machine used in an EV is chosen according to its construction structure,
efficiency, maintenance requirements, electromagnetic interference (EMI), reliability, and
speed range parameters. The machine controllers are power electronics systems that are
combined with sensors and observers in order to ensure the wide speed range in constant
torque situations and extended speed range for constant power requirements [3]. Fixing the
gear ratio allows to easily adjust the speed where a high speed electric machine can be used
due to its smaller size and weights less in contrast to low speed machines. This section deals
with electrical equivalents and controllers of IMs, PMSMs, and BLDCs for EVs.
Furthermore, there are various textbooks and application notes introducing electrical
machines in detail can be found in literature.
262 Ersan Kabalci
6.1. AC Induction Motor (ACIM) and Controllers
The induction machine, which has two separate operation mode as generator or motor, is
used in motor mode for power-train applications. An equivalent circuit can be used to model
the induction motor for electric vehicles. The efficiency of the motor can be easily calculated,
if the parameters of the equivalent circuit model seen in Figure 12 are known. Besides the
starting and maximum torque characteristics of an induction motor, some other steady-state
performance characteristics include the variation of current, speed, and losses as the load-
torque requirements.
The rotor of an induction motor can be wounded or squirrel-cage type. A wound rotor is
built with a poly-phase winding similar to the stator where the rotor terminals are externally
available on the motor. On the other hand, a squirrel-cage rotor has a winding that consists of
conductor bars embedded on the rotor iron, which are short-circuited at each end by
conducting end rings. The wound rotor is the most common type of ACIM in sizes ranging
from fractional horsepower on up. The difference between synchronous speed and the rotor
speed is commonly referred to the slip of the rotor [29,31]. The rational slip factor (s) is
defined as
ns  nr
s % (14)
ns
where the synchronous speed is nsand the rotor speed is depicted with nr both in terms of
revolutions per minute (rpm). The synchronous speed is calculated with Equation (15).
120  f
ns  (15)
P
where f is the line frequency in hertz and P is the number of poles in the machine. The
equivalent circuit shown in Figure 12 contains the electrical and mechanical power
conversion properties of an induction motor in a schematic diagram.
The stator equivalent circuit that is seen on the left hand side up to Rm//Xm magnetizing
circuit of Figure 12 is combined with stator resistance (Rs) and stator inductance (Xls). The
core losses of stator and rotor are depicted in the magnetizing circuit. The power flow
diagram of a polyphase induction motor is shown in Figure 13 by considering input power
(Pi), air gap power (Pg), mechanical converted power (Pm), and mechanical output power
(P0). The m parameter is the number of phase applied to motor while T is torque, and ωm is
the angular velocity.
An induction motor can be operated by vector controls with appropriate feedback signals
that are obtained from encoders. This method performs the speed control by a PWM
synthesized sinusoidal waveform.
The vector controlled motors that are typically designed include a high efficiency
winding, an efficient lamination design, and high-temperature insulation materials. The
tighter speed regulation approaching to 0.01% of fixed speed can be attained by vector
control according to scalar control methods. A vector controller allows obtaining controllable
speed ranges from zero to five times of base speed. The constant horsepower range is also
about 3.5 times of the base speed. The smooth stopping and highly efficient operation is
obtained by using a vector control. Standard vector controls with dynamic brake resistor are
used in many applications.
Figure 12. Modified equivalent circuit model of an induction motor.
Figure 13. Power flow diagram of an induction motor.
The advantages of this approach include low initial cost, reliability, the capability of
providing full rated torque from rated speed down to zero, precise speed and torque control,
constant output power above the rated speed, and such programmable features as controlling
acceleration/deceleration time and tuning. The vector control of an ACIM can be applied with
high-performance adjustable-speed applications of EVs. The block diagram of a vector
control system is shown in Figure 14. The speed controller is based on a PI regulator that
controls the motor slip ratio where the value computed by the regulator is added to the motor
speed in order to generate the demand frequency. The latter frequency is also used to generate
the demanded voltage in order to maintain the constant V/f ratio of the motor.
The space vector PWM modulator contains three-phase generator, low-pass bus filter, α-β
transformation, α-β vector sector, ramp generator, and switching time calculator [32,33].
Three-phase generator provides the modulating signals in 120° degree phase shift and the α-β
transformer generates three-phase to two-phase conversion in order to be used in sector
preparation. The low-pass bus filter is used to remove fast transients from the DC bus voltage
measurement. The feed-forward method is used to compute the voltage vector applied to the
motor [34,35].
Another widely used control technique of the induction motor is known as field-oriented
control (FOC). The FOC of an induction motor achieves torque (T) and rotor flux (ψr)
dynamics in a decoupled way by orthogonal projection of the stator current into torque (T)
and flux (ψr) generating components. The orientation angle that allows a complete decoupling
264 Ersan Kabalci
between those two variables is the rotor flux angle, θ. This technique is performed by two
basic methods that are namely direct and indirect vector control. The required instant value is
obtained by direct measurements using flux sensors or flux estimators in direct field
orientation, whereas indirect field orientation is based on the inverse flux model dynamics
and there are three possible implementation based on the stator, rotor or air gap flux
orientation. The indirect control technique of rotor flux is the most widely used due to its
simplicity. Although the FOC methods are attractive, they suffer from one major
disadvantage that is related to sensitivity.
Figure 14. Vector control block diagram of an AC induction motor.
The FOC methods are sensitive to parameter variations such as rotor time constant and
incorrect flux measurement or estimation at low speeds. The field oriented control block
diagram of an ACIM is shown in Figure 15 [36-38]. Since all variables of Figure 15 are
expressed in terms of rotating coordinate frame, the electromagnetic torque can be controlled
via the imaginary component of the stator current isq [39, 40].
The rotor flux can be commanded by its real part isd according to a first order dynamic
with a time constant τr, as shown in the following equations [37];
Figure 15. Field oriented control block diagram of an AC induction motor.

Lm
r  i (16)
 r  s  1 sd
3 Lm
T p  r  isq (17)
2 Lr
Equation(16) expresses the flux estimation while torque estimation is given in

Equation(17). The θ calculation is used to find the phase angle of the rotor flux-rotating field
[41-43].
6.2. Brushless DC Motor and Controllers
Two separate brushless motor types are known as brushless DC motor (BLDC) and
permanent magnet synchronous motor (PMSM). The brushless motors are widely used as in-
wheel motors due to their high efficiency in electric vehicles. The BLDC term means an AC
motor with combination of semiconductor inverter control and rotor position sensor. This
combination provides a stable and linear moment-speed characteristic such as classical DC
motors. The major advantages of BLDC are compactness, easy control and cooling, low
maintenance and low noise, and high reliability besides its efficiency. The BLDC
commutation is performed by using simple hall-effect sensors that are compactly located on
the motor and provides appropriate current waveforms. The permanent-magnet DC machines
are widely found in a wide variety of low-power applications. The field winding is replaced
by a permanent magnet, resulting in simpler construction. The permanent magnets provide a
number of benefits in these applications. However, a high-resolution encoder is required for
the PMSM control mode to generate sinusoidal currents comparing to BLDC [11, 12].
The permanent magnets provide the DC excitation of the field winding in a synchronous.
By replacing, the electrical excitation with permanent magnet ensures to eliminate the copper
losses. The BLDC machines can be categorized by the position of rotor permanent magnet in
which the magnets are mounted on the rotor. The magnets can either be surface mounted or
interior mounted. There are several textbooks and publications define the physical structure of
BLDC in terms of rotor and stator [28, 31, 40].
Therefore, this part will focus on back electromagnetic force (EMF) properties of BLDC
and PMSM and control techniques of brushless motors. The back EMF of BLDC motors are
classified in two ways such as trapezoidal-shaped and sinusoidal-shaped. The basic torque
and voltage equations of BLDC closely resemble those of brushed DC motors. The PMSM is
sinusoidal excited while the BLDC is trapezoidal excited machine motor.
The construction differences between are that while the stator windings of the trapezoidal
PM machines are concentrated into a narrow-phase pole, the windings of a sinusoidal
machine are typically distributed over multiple slots to approximate a sinusoidal distribution.
These differences in construction are reflected in their corresponding motion characteristics.
This implies the consequence that the first type of PMSM provides sinusoidal back-
electromotive force (back-EMF) generation, and the second type provides trapezoidal back-
EMF. The Back EMF is directly proportional to the motor speed while the torque production
is almost directly proportional to the phase current. The idealized back EMF and current
266 Ersan Kabalci
waveforms are seen in Figure16 where the BLDC is operating during half cycles of each
three-phase. The back EMF waveforms indicate the switching orders of three-phase inverter
in torque or speed controller that are shown in Figure 17 and Figure 18, respectively.
Figure 16. Idealized back EMFs (EA, EB, EC) and current signals (IA, IB, IC) of a three-phase BLDC
motor.
Figure 17. Block diagram of the torque control of a three-phase BLDC motor.
Figure 17 depicts a torque control scheme for a BLDC motor drive. The desired current
I* is derived from the commanded torque T* through a torque controller. The current
controller and commutation sequencer acquires the desired current I* from limiter, the
position information from the position sensors, and the required current feedback through
current transducers, and then generates gating signals. The gating signals are applied to the
three-phase inverter to generate the desired phase current to the BLDC machine.
The three-phase inverter is widely used in full-bridge topology. Most of the high-
performance traction control systems require torque control to sustain required performance.
The speed control of a BLDC shown in Figure 18 is based on torque controller with
additional equipments such as speed sensor and speed controller block. The speed control is
used in cruise control of traction application where stabilizing the speed is required. The
speed sensors acquire the actual rpm (ωr) value of the BLDC to compare the desired reference
speed (ωr*). The speed controller block is constituted with classical proportional-integral (PI),
proportional-integral-derivative (PID) controllers or advanced software controllers such as
artificial intelligence or neural network controllers.
The remainder blocks in the controller perform as in torque controller. The switching
signals are generated according to inherited values and are applied to inverter in order to
generate three-phase line voltage of BLDC [40, 44, 45].
Figure 18. Block diagram of the speed control of a three-phase BLDC motor.
7. THE LATEST POWER ELECTRONICS IN EVS

The circuit diagram seen in Figure 19 is known as resonant inverter application used for
AC motor driving operations. The additional switches located between phase legs and neutral
points constitutes the auxiliary resonant commutated pole (ARCP) that provides to operate the
inverter in soft switching mode to decrease the switching losses and to increase the efficiency.
The auxiliary circuit helps enable the load to be swung to the opposite rail to assure zero turn-
on voltage. If the auxiliary circuit is not included, then there is a load current constraint to
insure zero turn-on voltage.
268 Ersan Kabalci
Figure 19. A zero-voltage-transition inverter for AC motors.
(a) (b)
(c)
Figure 20. Bidirectional DC-DC converter configurations; (a) cascaded buck-boost, (b) half-bridge,(c)
split-π.
Emerging power electronics in electric vehicles also cover the DC-DC converter
topologies. There are several topologies developed to increase the conversion efficiency and
decrease the switching losses. Besides the unidirectional DC-DC converters introduced in 4th
section, the novel applications seen in electric vehicles are based on bidirectional converter
topologies. The bi-directional DC-DC converter plays an important role in the four-quadrant
EV propulsion system where it covers motoring and generator modes. While the EV drive
operates in motoring mode, it consumes supply energy from the batteries. On the other hand,
when the EV driver switched into braking mode the batteries are charged by electric motor
owing to bidirectional structure of converter. Therefore, the bi-directional DC-DC converter
will be the key device in EVs. Bidirectional converters can operate in both motoring and
regenerative braking modes, and it is therefore dedicated to four-quadrant operation of EV
propulsion systems. The most widely used bidirectional converters are shown in Figure 20
where Figure 20(a) illustrates the cascaded buck-boost while half-bridge and split-π
configurations are seen in Figure 20(b) and Figure 20(c), respectively [46, 47].
Although the cascaded buck-boost converter topology decreases the electrical and
thermal stresses on switching devices, it requires twice of the active devices according to
regular buck-boost topology. The required switch number can be decreased by using half-
bridge topology seen in Figure20.b where the passive components are at the same number.
The split-π configuration is relatively a new topology among other DC-DC converters,
and has numerous application areas as regenerative braking in electric vehicles. The passive
components that are used in a split-π converter are quite smaller in terms of sizes and yields
efficiency over 97%. Furthermore, the switching losses, distortions, and ripples on the output
current waveform are reduced owing to the topological configuration [46, 47].
8. DISCUSSION
The power electronics and drives have the key role on traction control of electric vehicle
technologies. The latest trends involve the most efficient power flow that is realized from
battery through the entire vehicle. In order to achieve this, there are numerous studies
performed on innovative issues such as highly efficient converters, bidirectional power
control, dynamic feedback, and regenerative braking. On the other hand, motor drives are also
extensively studied owing to emerging magnetic and electrical materials. This chapter covers
the most widely used power electronics and motor drives referring to classical circuit
topologies and presenting the novel configurations. It is seen that the energy efficient
converters such as resonant or hybrid ones sustain their popularity due to reliable switching
specifications at higher frequencies. Another important issue discussed for converter circuits
is bidirectional configuration that allows to operate the traction motor as a generator during
the braking mode. In this mode, the generated energy is used to charge the batteries if there is
a convenient converter configuration used. Consequently, more efficient and high power
electric vehicles depend on high performance power electronics.
270 Ersan Kabalci
CONCLUSION
This chapter is dedicated to specific applications used in EV technology although the
power electronics devices and motor drives cover a wide research area. The related topics are
arranged from plug to propulsion system for a PEV as shown in Figure1. The rectifier circuits
are used to convert AC grid voltage to DC voltage, and supplies the converted voltage to
DC/DC converters as introduced in fourth subsection. There are several novel converter
topologies are also announced in the literature to decrease costs and increase the efficiency
besides the introduced main topologies. A converter ensures the charge control of battery
system and allows getting the batteries charged in several modes by controlling rectifier
output and regenerative braking feedbacks. Furthermore, DC-DC converters also manage the
DC motor controls. Inverters cover a wide research area in medium and high power
application due to proposed topologies such as full bridge, multilevel and resonant structures.
Multilevel inverters generate output voltage in staircase levels that increases the number of
voltage steps and decreases the electromagnetic interference (EMI) rates. Low cost topologies
such as full-bridge or resonant types are widely used in EVs since the inverters are used in
motor drive operations. Three-phase full-bridge inverter that is analyzed in 5th subsection is
the main energy conversion part of motor control applications shown in Figures 14, 15, 17,
and 18. Resonant inverters, namely soft switching inverters, tackle switching power losses by
managing commutating intervals during zero voltage or zero current transitions.
All the power electronics devices introduced up to 6th section require a microcontroller
or microprocessor to generate switching signals considering the reference values obtained
from voltage or current sensors. The most widely used electric motors in EVs are surveyed in
6th section with drive methods and systems. Induction machines (IM), permanent magnet
synchronous machines (PMSM), and brushless DC machines (BLDC) are considered as
several electric machine types used in EV powertrain applications with specific controllers.
The machine control techniques and controllers are implemented to increase the efficiency of
the machine due to developments met in power electronics. Speed or torque control systems
can be summarized as a synthesis of electric motor, observers (position and/or speed sensors),
and inverter. The microcontroller unit (MCU) manages the closed loop control by acquiring
the reference values form sensors, determining the required arrangements, and generating the
required switching signals for inverter. The efficiency of closed loop is mostly based on the
algorithm that runs on MCU besides self-performance of each device. The estimating
algorithms, which allows to sensorless control are developed to remove the sensor
requirement in speed and torque control. Other widely known algorithms such as Park, Clarke
and Inverse Park, Inverse Clark provide 3-phase/2-phase conversion and generate d-q
reference values in current control. BLDC and PMSM motors are frequently used in
propulsion systems due to their higher rpm rates comparing to IMs. The BLDCs with 230V
supply can generate up to 20000 rpm and 250 HP power. The internal sensor and tachometer
options are evaluated as an important advantage. Furthermore, hub or in-wheel BLDCs
provide independent control for each wheels as a 4x4like application.
REFERENCES
[1] Sandalow, D.B.; Plug-In Electric Vehicles: What Role for Washington?. 1st Ed., The
Brookings Institution, ISBN 978-0-8157-0305-1, 2009.
[2] Pistoia,G.; Electric and Hybrid Vehicles Power Sources, Models, Sustainability,
Infrastructure and The Market. Elsevier B.V., ISBN 978-0-444-53565-8, 2010.
[3] Mi, C.; Masrur, M.A.; Gao, D.W.; Hybrid Electric Vehicles: Principles and
Applications with Practical Perspectives. John Wiley & Sons Ltd, ISBN 978-0-470-
74773-5, 2011.
[4] Mapelli, F.L.; Tarsitano, D.; Mauri, M.; Plug-In Hybrid Electric Vehicle: Modeling,
Prototype Realization, and Inverter Losses Reduction Analysis. IEEE Transactions on
Industrial Electronics, vol. 57, no. 2, pp. 598-607, 2010.
[5] Emadi, A.; Lee, Y.J.; Rajashekara, K.; Power Electronics and Motor Drives in Electric,
Hybrid Electric, and Plug-In Hybrid Electric Vehicles. IEEE Transactions on
Industrial Electronics, vol. 55, no. 6, pp. 2237-2245, 2008.
[6] Lee, Y.J.; Khaligh, A.; Emadi, A.; Advanced Integrated Bidirectional AC/DC and
DC/DC Converter for Plug-In Hybrid Electric Vehicles. IEEE Transactions on
Vehicular Technology, vol. 58, no. 8, pp. 3970-3980, 2009.
[7] Gao, Y.; Ehsani, M.; Design and Control Methodology of Plug-in Hybrid Electric
Vehicles. IEEE Transactions on Industrial Electronics, vol. 57, no. 2, pp. 633-640,
2010.
[8] Chau, K.T.; Chan, C.C.; Liu, C.; Overview of Permanent-Magnet Brushless Drives for
Electric and Hybrid Electric Vehicles. IEEE Transactions on Industrial Electronics,
vol. 55, no. 6, pp. 2246-2257, 2008.
[9] Bose, B.K.; Modern Power Electronics and AC Drives. Prentice Hall Inc., ISBN 978-
0130167439, 2002.
[10] Luo, F.L.; Yee, H.; Rashid, M.; Digital Power Electronics and Applications. Elsevier
Publishing, ISBN 978-0120887576, 2005.
[11] Rashid, M.; Power Electronics Handbook. Academic Press, ISBN978-0125816502,
2001.
[12] Emadi, A.; Handbook of Automotive Power Electronics and Motor Drives.
Taylor&Francis CRC Press, ISBN 978-0824723613, 2005.
[13] Mohan, N.; Undeland, T.M.; Robbins, W.P.; Power Electronics-Converters.
Application and Design, John Wiley & Sons Inc., ISBN 978-0471142089, 1995.
[14] Soylu, S.; Electric Vehicles - Modelling and Simulations. Intech, ISBN 978-
9533074771, 2011.
[15] Larminie, J.; Lowry, J.; Electric Vehicle Technology Explained. John Wiley & Sons
Inc., ISBN 0-470-85163-5, 2003.
[16] Kabalci, E.; Irmak, E.;Colak, I.; Design of an AC-DC-AC Converter for Wind
Turbines. International Journal of Energy Research, Wiley Interscience, vol. 36, no. 2,
pp. 169-175, 2011.
[17] Kabalci, E.; Kabalci, Y.; Develi, I.; Modelling and Analysis of a Power Line
Communication System with QPSK Modem for Renewable Smart Grids. International
Journal of Electrical Power and Energy Systems, vol. 34, no. 1, pp. 19-28, 2012.
272 Ersan Kabalci
[18] Colak, I.; Irmak, E.; Yesilbudak, M.; Kabalci, E.; Design of a Parameter Determination
System for Non-Isolated Converter Topologies. IEEE International Aegean Conference
on Electric Machines and Power Electronics & Electromotion Joint Conference
(ACEMP 2010),Istanbul, Turkey,pp.241-244, Sept. 8-10, 2011.
[19] Babaei, E.; Hosseini, S.H.; Gharehpetian, G.; Reduction of THD and Low Order
Harmonics with Symmetrical Output Current for Single-Phase AC/AC Matrix
Converters. International Journal of Electrical Power and Energy Systems, vol. 32, pp.
225-235, 2010.
[20] Ghaemi, A.H.; Abyaneh, H.A.; Mazlumi, K.; Harmonic Indices Assessment by
Wavelet Transform. International Journal of Electrical Power and Energy Systems,
vol. 33, pp. 1399-1409, 2011.
[21] Ramasamy, M.; Thangavel, S.; Photovoltaic Based Dynamic Voltage Restorer with
Power Saver Capability Using PI Controller. International Journal of Electrical Power
and Energy Systems, vol. 36, pp. 51-59, 2012.
[22] Colak, I.; Kabalci, E.; Increasing the Efficiency of an Inverter Used in Renewable
Energy Sources. 2nd International Conference on Nuclear and Renewable Energy
Resources (NURER 2010), Ankara, Turkey,pp. 853-859, July 4-7, 2010.
[23] Colak, I.; Kabalci, E.; Bayindir, R.; Sagiroglu, S.; The Design Analysis of a 5-Level
Cascaded Voltage Source Inverter with Low THD. IEEE Int. Conf. (POWERENG),
Lisbon, Portugal,pp. 575-580, 18-20 March 2009.
[24] Colak, I.; Kabalci, E.; Bayindir, R.; Bal, G.; Modelling of a Three Phase SPWM
Multilevel VSI with Low THD Using Matlab/Simulink. EPE 2009, Barcelona, Spain, pp.
1-10, 8-10 September 2009.
[25] Colak, I.; Kabalci, E.; Bayindir, R.; Review on Multilevel Voltage Source Inverter
Topologies and Control Schemes, Energy Conversion and Management. vol. 52, pp.
1114-1128, 2011.
[26] Colak, I.; Kabalci, E.; Developing a Novel Sinusoidal Pulse Width Modulation
(SPWM) Technique to Eliminate Side Band Harmonics. International Journal of
Electrical Power and Energy Systems, doi:10.1016/j.ijepes.2012.08.024.
[27] Colak, I.; Kabalci, E.; Practical Implementation of a DSP Controlled Multilevel
Inverter with Low THD for Motor Drive Applications. Journal of Power Sources, vol.
196, pp. 7585-7593, 2011.
[28] Veltman, A.; Pulle, D.W.J.; Doncker, R.W.; Fundamentals of Electrical Drives.
Springer, ISBN 978-1-4020-5503-4, 2007.
[29] Vukosavic, S.N.; Digital Control of Electrical Drives. Springer, ISBN 978-0-387-
48598-0, 2007.
[30] Leitman, S.; Brant, B.; Build Your Own Electric Vehicle. McGraw-Hill, ISBN 0-07-
154373-2, 2009.
[31] Fitzgerald, A.E.; Kingsley, C.; Umans, S.; Electric Machinery. 6th Ed., McGraw-Hill
Higher Education, ISBN 978-0073660097, 2002.
[32] Zheng, C.; Zhang, B.; Qui, D.; Digital Natural Sampling SPWM Based on Inverse
Operator Method. IEEE Power Electronics Spec. Conf., Orlando, USA, 2007.
[33] Kanchan, R.S.; Gopakumar, K.; Kennel, R.; Synchronised Carrier-Based SVPWM
Signal Generation Scheme for the Entire Modulation Range Extending up to Six-Step
Mode Using the Sampled Amplitudes of Reference Phase Voltages. IEEE Elec. Power
App., vol. 1, pp. 407-415, 2007.
[34] Beig, A.R.; Narayanan, G.; Ranganathan, V.T.; Modified SVPWM Algorithm for
Three Level VSI with Synchronized and Symmetrical Waveforms. IEEE Transactions
on Industrial Electronics, vol. 54, pp.486-493, 2007.
[35] Jacobina, C.B.; Lima, A.M.N.; Silva, E.R.C.; Alves, R.N.C.; Seixas, P.F.; Digital
Scalar PWM: A Simple Approach to Introduce Non-Sinusoidal Modulating
Waveforms. IEEE Transactions on Power Electronics, vol. 16, pp. 351-359, 2001.
[36] Cirrincione, M.; Accetta, A.; Pucci, M.; Vitale, G.; MRAS Speed Observer for High-
Performance Linear Induction Motor Drives Based on Linear Neural Networks. IEEE
Transactions on Power Electronics, vol.28, no.1, pp.123-134, Jan. 2013.
[37] Rodriguez, J.; Kennel, R.M.; Espinoza, J.R.; Trincado, M.; Silva, C.A.; Rojas, C.A.;
High-Performance Control Strategies for Electrical Drives: An Experimental
Assessment. IEEE Transactions on Industrial Electronics, vol.59, no.2, pp.812-820,
Feb. 2012.
[38] Patel, H.M.; Shah, P.T.; Patel, H.V.; Field Oriented Control of Induction Motor Using
Matlab Simulink. Journal of Information, Knowledge and Research in Electrical
Engineering, vol. 1, no.2, pp. 56-60, 2011.
[39] Biswas, D.; Mukherjee, K.; Kar, N.C.; A Novel Approach towards Electrical Loss
Minimization in Vector Controlled Induction Machine Drive for EV/HEV. IEEE
Transportation on Electrification Conference and Expo (ITEC), pp.1-5, 18-20 June
2012.
[40] Ehsani, M.; Gao, Y.; Emadi, A.; Modern Electric, Hybrid Electric, and Fuel Cell
Vehicles: Fundamentals, Theory, and Design. 2nd Edition, CRC Press, ISBN 978-
1420053982, 2009.
[41] Yi, W.; Kaiqi, Z.; Field-Oriented Vector Control of Induction Motor for Electric
Vehicles. 31st Annual Conference of IEEE Industrial Electronics Society, pp. 1610-
1614, 2005.
[42] Rehman, H.U.; Xu, L.; Alternative Energy Vehicles Drive System: Control, Flux and
Torque Estimation, and Efficiency Optimization. IEEE Transactions on Vehicular
Technology, vol.60, no.8, pp.3625-3634, Oct. 2011.
[43] Mauri, M.; Mapelli, F.L.; Tarsitano, D.; A Reduced Losses Field Oriented Control for
Plug-in Hybrid Electrical Vehicle. XIX International Conference on Electrical
Machines (ICEM), pp.1-6, 6-8 Sept. 2010.
[44] Texas Instruments Europe; DSP Solutions for BLDC Motors. 1997.
[45] Balazovic, P.; Designer Reference Manual: Sensorless PMSM Control for an H-axis
Washing Machine Drive. Freescale Czech Systems Laboratories, Roznov pod
Radhostem, Czech.
[46] Tie, S.F.; Tan, C.W.; A Review of Energy Sources and Energy Management System in
Electric Vehicles. Renewable and Sustainable Energy Reviews, vol. 20, pp. 82-102,
2013.
[47] Zhang, X.; Wang, Z.; Cheng, M.; Shichuan, D.; Wang, W.; The Random PWM Based
Bi-Directional Buck-Boost Cascade Converter for Electric Vehicles. International
Conference on Electrical Machines and Systems (ICEMS), pp.1-5, 20-23 Aug. 2011.
Chapter 10
POWER CONTROL OF PLUG-IN

ELECTRIC VEHICLES IN SMART GRIDS
Amin Hajizadeh1,, Amir Hossein Shahirinia2,†

and David C. Yu2,‡
1
Department of Electrical Engineering, Shahrood
University of Technology, Shahrood, Iran
2
Department of Electrical Engineering and
Computer Applied Science,
University of Wisconsin-Milwaukee, Milwaukee, US
ABSTRACT
In this chapter, Plug-in Electric Vehicle (PEV) is considered with dual energy
storage sources. In order to respond to a transient power demand, usually battery energy
storage device is combined with super capacitor. Battery-super capacitor system is used
to create an energy storage system with the high energy density attributes of batteries and
the high power density of super capacitor. In order to simulate the PEV model, dynamic
state space models of bidirectional DC-DC converter and grid connected voltage source
converter with energy storage models are considered to connect the PEV to the main grid.
In order to stabilize the DC-link power and distribute the power between dual energy
storage sources in PEV during normal and disturbance conditions on the grid voltage, a
Lyapunov based Neuro-Fuzzy control strategy has been developed. This controller
determines the super capacitor power that should be generated according to the amount of
available energy in DC-link. Moreover, a robust sliding mode control strategy for voltage
source converter based on positive and negative symmetrical components is presented to
investigate the voltage disturbance ride-through and voltage control capability.
Keywords: Plug-in Electric Vehicle, Intelligent Control, Smart Grids, Neuro-Fuzzy,

Voltage Disturbances

E-mail: aminhajizadeh@shahroodut.ac.ir.
†
E-mail: shahiri2@uwm.edu.
‡
E-mail: yu@uwm.edu.
276 Amin Hajizadeh, Amir Hossein Shahirinia and David C. Yu
1. INTRODUCTION
Gridable Electric Vehicles (GEV), such as plug-in hybrid electric vehicles and battery-
electric vehicles, have recently become more attractive options over combustion engine-based
vehicles, due to the high fuel price and environmental concerns. They have the ability to
discharge their batteries and send electricity back to the power grid when they are parked and
connected to an electrical outlet. The electricity flowing from electric vehicles to the power
grid is then called “vehicle-to-grid” power or V2G [1-3].
The success of widespread electric vehicle implantation depends on the reliability of both
the electric grid and the charging equipment. The cost of the batteries make the energy price
of battery supplied power expensive, which implies that not only a robust and reliable charger
is needed, but also with high power efficiency and minimum impact on the power quality of
smart grid distributions systems. The electric vehicle storages can act as controllable loads or
generation units to buffer the variable power produced by the wind turbines. V2G systems
can act as a provider of the important power system ancillary services like regulation (up and
down) and spinning reserves [3]. The EV owners can benefit by the revenues earned from the
electricity market by participating in the power balancing act.
The EV storages act as distributed sources with smart metering, intelligent control and
communication interfaces with which it is connected to an aggregator [2].
The power flow in V2G can be both unidirectional and bidirectional, though in
unidirectional V2G only services such as frequency regulation and delivering power, are
provided to the grid. Through V2G, EV owners can potentially generate revenue while
charging their cars and at the same time mitigate the negative impacts on the grid from
charging. With V2G, an EV can participate in most energy markets, from bulk energy, to
spinning reserves and frequency regulation [4-6].
Plug-in Electric Vehicles (PEV) are connected to smart grid via power electronic
converters to support the grid.
So it is important that the control strategies are designed to keep the system stable under
any disturbance and parameter variations in the electrical distribution system.
The grid-connected power electronic converters are highly sensitive to grid disturbances
and it is important to emphasize the necessity to reduce the effects of voltage disturbances on
their operation. In this chapter, PEV is considered with dual energy storage sources.
In order to respond to a transient power demand, usually battery energy storage device is
combined with supercapacitor. Battery-super capacitor system is used to create an energy
storage system with the high energy density attributes of batteries and the high power density
of super capacitor.
For simulation studies, dynamic models of bidirectional DC-DC converter and grid
connected voltage source converter are considered to connect the hybrid energy storage
sources to the main grid. In order to stabilize the DC-link power and distribute the power
between dual energy storage sources in PEV during normal and disturbance conditions on the
grid voltage, a Lyapunov based neuro-fuzzy control strategy has been developed. This
controller determines the super capacitor power that should be generated according to the
amount of available energy in DC-link. Moreover, a robust sliding mode control strategy for
voltage source converter based on positive and negative symmetrical components is presented
to investigate the voltage disturbance ride-through and voltage control capability.
Power Control of Plug-In Electric Vehicles in Smart Grids 277
2. ARCHITECTURE OF THE PLUG-IN ELECTRIC VEHICLE

The modeling of Plug-in Electric Vehicle (PEV) is an important issue that needs to be
carefully addressed. In Figure 1, overall structure of PEV is shown. Many articles deal with
static models that are built up from maps and static relationships between parameters in the
model.
These models allow for fast simulation, but they cannot show the oscillations and other
dynamic phenomena when switching occurs between different modes of operation. However,
a good model should consider both accuracy and simulation time. The electric components of
electric vehicle used in this chapter comprise a battery bank, DC/DC and DC/AC converters.
The details about modeling of battery energy storage and power electronic converters have
been described in [7, 8].
3. POWER MANAGEMENT ON V2G - STATE OF THE ART

EV battery‟s energy can be utilized as distributed energy storage and used for reactive
power support as well as voltage sag reduction at a particular node. A schematic diagram of a
hybrid energy storage including battery and supercapacitor and a bidirectional converter
coupled with the distribution node via line reactance X is shown in Figure 2. In this figure, the
hybrid energy storage system represents a lumped parameter for the energy available from the
charging station in that particular area. With the bidirectional converter, this dual energy
storage can behave both as a source or a sink, i.e., the vehicles at the charging station can
collectively discharge or charge respectively. The bidirectional converter, interfaced with the
EV energy storage, is synchronized with the grid system at point a. Power injection by the
battery is the power at connection point b, which is at the utility side.
The grid-connected power electronic converters are highly sensitive to grid disturbances
and it makes to emphasize the necessity to reduce the effects of voltage disturbances on their
operations [9]. In spite of the growing number of V2G units, their contribution of power
delivered to the utility grid remains small, as compared to the power injected by the large
centralized power plants.
Figure 1. Overall structure of PEV.

Figure 2. Diagram of charging and discharging modes of V2G.
However, to support the grid in case of disturbances, it will become necessary to keep the
V2G units connected to the grid.
In the wide range of power quality disturbances, the interest focuses on voltage sags,
which can severely affect the performance of the voltage source converter (VSC). A voltage
sag is a drop in voltage with duration between one half-cycle and one minute [9], which is in
most cases caused by a short-circuit fault.
Therefore, the operation of V2G under voltage sag has not received much attention in the
past. However, as the power generated by V2G increases, this behavior stresses the utility
grid and could cause power unbalance, which may turn into instability.
So, the interaction between V2G and the grid during the voltage sag is very important and
it must be considered for designing the proper control strategy.
Hence in this chapter a control strategy is proposed for V2G including energy storages
under voltage sag conditions. During the voltage sag in distribution systems, a decrease in
voltage magnitude ate the grid-connected converter is occurs.
In this case, the current controllers to avoid overloading the converter limit the power that
V2G can supply to the grid during voltage sag. For V2G, the power limitation can be a
problem resulting the slow dynamics of battery energy storages. According to this limitation,
in order to respond the transient of power demand, usually a supercapacitor is combined with
battery energy storage. The comparison between battery and supercapacitor energy storages
shows that the use of supercapacitor is better than battery for power quality problem studies
and there are some limitations to use the battery energy storage.
Due to the low power density of the battery than supercapacitor, it cannot release its
charge or discharge fast enough during the voltage sag [7].
Additionally, the main drawback of the batteries is a slow charging time, limited by a
charging current, in contrast, the supercapacitor can be charged in a short time, depending on
the availability of a high charging power from the main source.
Although the batteries are considered to be the main energy storage device for EV
application, their cycle and calendar life should still be improved.
For these reasons, the use of the supercapacitor as an auxiliary source is expected that the
very fast power response and high specific power can complement the slower power output of
the main source.
So it is important to study the operation and the behavior of whole components of V2G
system under voltage sag not only the response of power electronic converters.
Hence, in this chapter a robust power control strategy has been presented for V2Gsystem
during voltage sag in smart grid.
4. POWER MANAGEMENT OF GEV BASED ON

DC-LINK POWER AND VOLTAGE CONTROL
In this section, the control strategy of V2G system has been presented. The overall
control structure is illustrated in Figure 3, which includes the power flow controller and local
controllers for power conditioning units and the hybrid energy storages system.
The term, “power flow control”, refers to the design of the higher-level control algorithm
that determines the proper power level to be generated, and its split between the two power
sources [10, 11].
In fact, during voltage sag conditions, the power flow control strategy must be design to
stabilize the DC-link power and regulate the DC-link voltage consequently. During the
voltage sag, a decrease in voltage amplitude occurs at the converter terminal. To keep the
power supplied to the grid constant, the current should increase. It will be limited by the
current controller however, to avoid overloading of the converter. This will thus limit the
power that the V2G can supply to the grid during voltage sag, resulting the DC-link voltage
will increase.
Figure 3. Power Control structure of V2G with hybrid energy storages and power electronic converters.
To avoid a too high DC-link voltage, the power balance between inverter power and V2G
power must be satisfied. According to three operation modes for V2G, the DC-link voltage is
given by three different state equations:
Discharging mode:
dvdc
Cdc vdc  PV 2G  Pinv
dt
PV 2G (k )  Pbatt (k )  Psup (k )
(1)
Charging mode:
dvdc
Cdc vdc  Pinv  PV 2G
dt
(2)
Disconnected Mode:
dvdc
Cdc vdc  PV 2G  Pload
dt
(3)
In this mode, the V2G is not connected to grid and it could be used for supplying
electrical motor to propel the vehicle. According to the above mentioned equations, PV2G, Pinv,
Pload, Pbatt, Psup, Cdc and Vdc are V2G power supplied to the DC-link, inverter power (grid
power), electrical load power, battery power, supercapacitor power, DC-link capacitor and
voltage respectively. During the voltage sag, the reference power of inverter is determined by
the maximum power that can be supplied to the grid. So, the V2G power must be decreased to
avoid a too high DC-link voltage. Hence, the power management in hybrid battery/ super-
capacitor energy storage system is very important task and its effect on the operation of the
system during voltage sag. So it should be robust under any variations in voltage and power.
Moreover, the distribution of power between power sources must be done optimally.
4.1. Lyapunov Based Neuro-Fuzzy Control Strategy
In this section, a power flow control structure has been developed for hybrid energy
sources during voltage sag. It is based on the Neuro-fuzzy algorithm and distributes the power
between the supercapacitor and battery energy storage according to the input parameters [10].
The configuration of the proposed control strategy is shown in Figure 4.
In this control structure, the battery power is determined according to the error of DC-
link power (power difference between inverter power and V2G power) P(k), the error of
DC-link voltage (voltage difference between desired voltage and bus voltage) Vdc(k) and the
error of battery power Pbatt(k) (battery power difference between Pbatt(k-1) and Pbatt(k)).
Figure 4. Proposed neuro-fuzzy power flow control strategy.
These inputs are calculated as follow:
P(k )  PV 2G (k )  Pinv (k )
Vdc (k )  Vdc ref VDC (k )
Pbatt (k )  Pbatt (k )  Pbatt (k  1) (4)
In fact, the P(k) and Vdc(k) and its derivative are very important during the decision
process to stabilize the dc bus voltage during the occurring of voltage sag. Also, Pbatt(k) is
very important to battery operation and it makes to minimize the changes of the battery
power. The output of neuro-fuzzy controller is the battery system duty ratio (Kbatt). It is a
representation of mount power of battery that must be delivered and varies between 0 and 1.
Then according to this parameter, the amount of battery power must be delivered is
determined. The following equations show the battery.
Pbatt (k )  Kbatt (k )  Pbatt  nom

Psup (k )  PV 2G (k )  Pbatt (k )
(5)
where, Pbatt-nom is the nominal power of battery.

During the voltage sags, the power flow control strategy must be designed to manage the
power between the battery and supercapacitor energy sources and the utility grid. Moreover,
it is important to consider the dynamic limitation of the battery, its power must change
smoothly so that the battery operating point deviation is minimized and its life time is
increased. In proposed neuro-fuzzy control strategy, for each input, four fuzzy subsets have
been used. These are ZE (zero), L (low), M (medium) and H (high). For each of these fuzzy
sets, Gaussian membership function has been used.
As each of the three inputs has seven four sets, there are altogether 64 control rules in the
neuro-fuzzy logic controller.
The neuro-fuzzy algorithm developed by Wang [12] uses the hybrid model developed by
Takagi-Sugeno in [13]. In this type of model, the condition part uses linguistic variables and
the conclusion part is represented by a numerical value which is considered as a function of
system condition expressed with the variablesx1, x2, …., xm. These models are suitable for
neural-based-learning techniques as gradient methods to extract the rules and generate models
with a reduced number of rules.
 (l )  g ( x1, x2 ,...., xn ) (6)

The neuro-fuzzy algorithm uses membership functions of Gaussian type. With Gaussian
fuzzy sets, the algorithm is capable of utilizing all information contained in the training set to
calculate each rule conclusion, which is different when using triangular partitions. Figure 5
illustrates the neuro-fuzzy scheme for an example with two input variables (x1, x2) and one
output variable (y). In the first stage of the neuro-fuzzy scheme, the two inputs are codified
into linguistic values by the set of Gaussian membership functions attributed to each variable.
The second stage calculates each rule R(l) its respective activation degree. Last, the inference
mechanism weights each rule conclusion ω(l), initialized by the cluster-based algorithm, using
the activation degree computed in the second stage.
The error signal between the model inferred value Y and the respective measured value
(or teaching value) y', is used by the gradient descent method to adjust each rule conclusion.
The algorithm modifies the values of ω(l) to minimize an objective function E usually
expressed by the mean quadratic error (16). In this equation, the value y'(k) is the desired
output value related with the condition vector x'(k) = (x'1, x'2,…,x'm). The element Y(x'(k)) is
the inferred response to the same condition vector x'(k) and computed by Equation (7).
1
E  [Y ( x(k ))  y(k )]2
2 (7)
c  m 
    A ( xj (k ))  . (l ) (k )
(l )
j 1 
Y ( x(k ))   
l 1 j 1
c  m 
    A(jl)1 ( xj (k )) 
l 1  j 1 
 ( xj (k )  m j ) 
2
 A ( xj (k ))  exp  
 ( j ) 2  (8)
where, the subscript j indicates the jth input xj.
Figure 5. The neuro-fuzzy structure.

By using the back propagation (BP) learning algorithm, the weighting vector of the FNN
is adjusted such that the error defined (7) is less than a desired threshold value after a given
number of training cycles. The well-known BP algorithm may be written briefly as:
E
 (l ) (t  1)   (l ) (t )  
 (l ) (9)
The symbol α is the learning rate parameter, and t indicates the number of learning
iterations executed by the algorithm.
The inference function (8) depends on ω(l) only through its numerator. The expression
composing the numerator is now denoted by a and is shown in (10).
c  m 
a      A( l ) ( xj (k ))  . (l ) (k )

l 1  j 1
j 1 
 (10)
The denominator of function (8) is dependent on a term d(l), defined in (11), and denoted
by b in (12).
m
d (l )    A( l ) ( xj (k ))
j 1
j 1
(11)
c
b   ( d (l ) )
l 1 (12)
To calculate the adjustment of each conclusion value ω(l), it is necessary to compute the
variation of the objective function E, ∂E, in relation to the variation that occurred in ω(l) in the
anterior instant, ∂ω(l). Therefore, using the chain rule to calculate ∂E/∂ω(l) results in Equation
(13).
E E Y a

 (l )
Y a  (l ) (13)
The use of chain rule looks for the term contained in E that is directly dependent on the
value to be adjusted, i.e., the conclusion value ω(l). Therefore, we can verify by chain
Equation (13) that it starts with E dependent of Y value. The Y value depends on term  and,
at last, the expression a is a function of ω(l). After some computation, the adjustment to be
made in ω(l) can be interpreted as being proportional to the error between the neuro-fuzzy
model response and the supervising value, but weighted by the contribution of rule (l),
denoted by d(l), to the final neuro-fuzzy inference.
(Y ( x(k ))  y(k ))d (l )

 (l ) (t  1)   (l ) (t )   c
 ( d (l ) )
l 1 (14)
Next, a convergence theorem has been developed to guarantee the stability of learning
algorithm used for the above-mentioned FNN. A Lyapunov energy function is defined as
follows:
1
Vk  J k  Ek2
2 (15)
From Equation (15), we can get
1
V  Vk 1  Vk  ( Ek21  Ek2 )
2 (16)
The error difference, Ek, can be defined as
Ek
Ek  Ek 1  Ek  
 (17)
where
Ek
   k 1  k   Ek
 (18)
Using Equation (16), we can get
1 1
V  ( Ek 1  Ek )( Ek 1  Ek )  (Ek )(2Ek  Ek )
2 2 (19)
Substituting Equation (18) into Equation (19), we have
1 Ek E E E 1 E  E 
V   Ek k (2Ek  k  Ek k )  ( Ek k ) 2 ( k ) 2 2  2 
2     2     (20)
If V<0, the convergence of the algorithm described in Equation (20) can be guaranteed.
Therefore, we have
2
 Ek  2
     2  0
  (21)
From Equation (21), we can obtain
2
0   2
 Ek 
  
  (22)
4.2. Robust Sliding Mode Control Strategy for Voltage Source Converter
Based on Positive and Negative Symmetrical Components
Control of grid connected voltage source converter (VSC) is an important problem during
voltage disturbances like voltage sag. A voltage sag is a reduction in the RMS voltage in the
range from 0.1 to 0.9 pu of the nominal voltage for duration greater than half cycle and less
than one minute. The drawback of using VSC is its sensitivity to voltage disturbances. For a
VSC, a sudden decrease in grid voltage normally causes an increase in current, as the control
attempts at maintaining the power to the DC-link constant.
Moreover, most faults are unbalanced and result in unbalanced voltage sag, which
produce undesirable power oscillations of low order frequencies in current harmonics and
poor DC-link voltage regulation. Ultimately, this can also lead to tripping of the converter due
to DC overvoltage.
The VSC is connected is connected to the grid via a filter inductor. The dq-components
of currents and voltages are then used along with the reference current signals. The VSC
controller is required to have main functions: Current control and DC-link voltage regulation.
A comparison between different types of current controllers for shunt connected VSC,
has proved that Dual Vector Current Controller (DVCC), is capable of providing sinusoidal
grid currents and regulated DC-link voltage during unbalanced faults. The current controller
used in this chapter consists of two sliding model controllers that control the positive and
negative sequence current separately and are implemented in two different rotating coordinate
systems [8].
A simplified scheme for the DVCC is shown in Figure 6. The VSC is connected to the
grid via filter inductors. The three phase grid currents and voltages are sampled and
transformed into its positive and negative sequence components. The positive and negative
sequence of dq-components are then used along the with the reference current signals in the
DVCC to produce the reference voltage signals for the PWM regulator.
According to the proposed control strategy in [8], the purpose of the current controller is
to synthesize a voltage correction vector so that the current error vector can be kept to a mini-
mum value. The error current-vector is defined as:
i(k )  iref (k )  i(k )

(23)
where, iref(k) is the specified current vector command for the positive and negative sequence
of dq-components and i(k) is the measured current vector feedback. In order to generate
proper current references, consider the complex apparent power from the grid:
S g  (edqp e jt  edqn e jt ).(idqp e jt  idqn e  jt )* 

 ( P  P2c cos(2t )  P2 s sin(2t ))  j (Q  Q2c cos(2t )  Q2 s sin(2t ))
(24)
In order to work with an invertible matrix (4×4), oscillating reactive power (Q2c, Q2s) can
not be included in the current reference calculation. Therefore, oscillating reactive power is
not controlled and will flow through the system.
By expanding Equation (24), the following expression in matrix form can be written:
 Pg  e dp e qp e dn e qn  idp 
    
Qg  e qp  e dp e qn  e dn  iqp 
 P   e  e dn  e qp e dn  idn 
 s 2 g   qn  
 Pc 2 g  e dn e qn e dp e qp  iqp 
   (25)
where, Pg and Qg are the constant active and reactive power, respectively, while the subscripts
Ps2g and Pc2g represent the second harmonic sine and cosine component of the active power.
These are the oscillating powers due to the unbalance in the grid voltages.
The reference currents can be calculated as follow:
idp
* 
e e qp e dn e qn 
1
 
   dp P*
  
iqp  e qp
*
 e dp e qn  e dn   Q 
*
 
idn
*  e  e dn  e qp e dn  
 Ps 2 

 qn  
*  
iqn  e dn e qn e dp e qp    Pc 2 
(26)
Figure 6. Block diagram of dual vector current controller.

P  R(idp
2
 iqp
2
 idn
2
 iqn
2
)
(27)
Pc 2  2R(idpidn  iqpiqn )  2 L (idpiqn  iqpidn )

(28)
Ps 2  2R(idpiqn  iqpidn )  2 L (idpidn  iqpiqn )

(29)
This algorithm calculates current references (26) by setting active and reactive power
references (P*, Q*), and by forcing the oscillating active power demanded by the filter to be
delivered from the grid ( P2*c  P2c ; P2*s  P2 s ) . Then, no oscillating active power flows
between the DC-link and the filter.
5. SIMULATION RESULTS
In order to show the effectiveness of proposed control strategy, the simulation model of
the proposed hybrid V2G system has been built in Matlab/Simulink environment. The
parameters of the hybrid battery/ supercapacitor energy storage system in this study are given
in [8, 10]. The supercapacitor operates as a buffer of energy to meet load demand that cannot
be met by the battery energy source, particularly during transient or disturbances periods. In
this case study, the output power of the battery is limited to 50 kW and the supercapacitor
bank has is capable of sustaining the extra load of 20 kW during the voltage disturbances.
Also it is supposed that the initial state of charge for supercapacitor is 0.7. For the
investigated system, simulations have been done with unbalanced voltage sag (type C) with
the magnitude of between 0.3 and 0.9 pu Maximum grid current and peak- to-peak DC
voltage ripple during the sag are shown for sag type C in Figures 7 and 8. In order to ride
through sag with minimum retained voltage of 30%, the converter switches should be able to
carry a maximum current of 3.65 pu.
The maximum DC voltage ripple is ±0.25% around the nominal value for sag magnitudes
below 30%. Moreover, for the evaluation of proposed control strategy, simulations have been
run with unbalanced voltage sag (type C) with magnitude 40%. The sag starts at 0.3sec and
ends at 0.4sec. The DC-link voltage is shown in Figure 9. The grid currents increase to above
2 pu and DC voltage shows a variation during the transients at the beginning and end of the
sag. However, the ripple during the transient is not bigger than 10% peak-to-peak and is quite
quickly damped to almost zero. As shown in Figure 10, the generated power of the hybrid
energy storage systems decreases during voltage disturbance to increase the voltage sag ride-
through capability. In fact, during the voltage sag, the current controllers limit the output
current to avoid the overloading of the converter. As a result the average and instantaneous
active power that converter supplies to the grid deceases (Figure 10), resulting in an increase
in DC-link voltage (Figure 9). The grid converter now controls the power (with its set point
determined by the maximum current). In these conditions, to stabilize the DC-link voltage,
the neuro-fuzzy controller manages the power flow between battery and supercapacitor,
resulting the reference power of the energy storage bank changes for decreasing the input
power to dc bus during voltage disturbances and neuro-fuzzy control strategy switches the
energy storage bank to the charging mode and energy storage‟s state of charge (Figure 11)
goes high and remain constant on a new value. Because after occurring voltage sag, the
energy storage dose not discharge and the output power of supercapacitor is equal to zero.
From Figures 12 and 13, it is evident that the battery and supercapacitor together share
this load requirement. Because of the slow variations of battery, the output power of battery
energy source remains constant during the voltage sag (Figure 13).
The output voltage of battery is shown in Figure 14. The battery voltage has not been
affected during the voltage sag. According to the simulation results, it is clear that the
parameters of battery system has not changed during the occurring the voltage disturbances in
distribution systems.
On the other hand, during occurring voltage sag, the supercapacitor makes to improve the
performance of V2G. In this condition, the proposed power flow control strategy finds
suitable operating point for supercapacitor and battery energy sources.
Figure 7. Maximum grid current in pu for voltage sag type C with magnitude between 0.3 and 0.9.
Figure 8. DC-link voltage ripples in pu for voltage sag type C with magnitude between 0.3 and 0.9.
Figure 9. DC-link voltage for 40% sag type C (pu).
Figure 10. Generated average and instantaneous active power by hybrid distributed generation system
during voltage sag.
CONCLUSION
In this chapter, PEV is considered with dual energy storage sources. In order to respond
to a transient power demand, usually battery energy storage device is combined with super
capacitor. Battery-super capacitor system is used to create an energy storage system with the
high energy density attributes of batteries and the high power density of super capacitor.
For simulation studies, dynamic models of bidirectional DC-DC converter and grid
connected voltage source converter are considered to connect the hybrid energy storage
sources to the main grid.
Figure 11. Energy storage‟s state of charge during voltage sag.
Figure 12. The supercapacitor power during voltage sag (pu).
Figure 13. The battery power during voltage sag (pu).
In order to stabilize the DC-link power and distribute the power between dual energy
storage sources in PEV during normal and disturbance conditions on the grid voltage, a
Lyapunov based neuro-fuzzy control strategy has been developed. This controller determines
the super capacitor power that should be generated according to the amount of available
energy in DC-link.
Figure 14. Output voltage of battery during voltage sag.
Moreover, a robust sliding mode control strategy for voltage source converter based on
positive and negative symmetrical components is presented to investigate the voltage distur-
bance ride-through and voltage control capability.
REFERENCES
[1] Han, S., Sezaki, K., Development of an Optimal Vehicle-to Grid Aggregator for
Frequency Regulation. IEEE Transaction on Smart Grid, vol. 1, no. 1, pp. 65-72, 2010.
[2] Guille, C., Gross, G., A Conceptual Framework for the Vehicle-to-Grid (V2G) Imple-
mentation. Energy Policy, vol. 37, no. 11, pp. 4379-4390, 2009.
[3] Han, S., Sezaki, K., Design of an Optimal Aggregator for Vehicle to Grid Regulation
Service. IEEE Innovative Smart Grid Technologies (ISGT 2010), Gaithersburg, Mary-
land, US, pp. 1-8, 2010.
[4] Nyns, K. C., Haesen, E., Driesen, J., The Impact of Charging Plug-In Hybrid Electric
Vehicles on a Residential Distribution Grid. IEEE Transaction on Power System, vol.
25, no. 1, pp. 371-380, 2010.
[5] Kisacikoglu, M. C., Ozpineci, B., Tolbert, L. M., Examination of a PHEV Bidirectional
Charger System for V2G Reactive Power Compensation. Twenty-Fifth Annual IEEE
Applied Power Electronics Conference and Exposition (APEC 2010), Palm Springs,
California, US, pp. 458-465, 2010.
[6] Lee, Y. J., Khaligh, A., Emadi, A., Advanced Integrated Bidirectional AC/DC and DC/
DC Converter for Plug-In Hybrid Electric Vehicles. IEEE Transaction on Vehicular
Technology, vol. 58, no. 8, pp. 3970-3980, 2009.
[7] Barton, J., Infield, D., Energy Storage and its Use with Intermittent Renewable Energy.
IEEE Transaction on Energy Conversion, vol. 19, no. 2, pp. 441-448, 2004.
[8] Hajizadeh, A., Golkar, M. A., Feliachi, A., Voltage Control and Active Power Manage-
ment of Hybrid Fuel-Cell/Energy-Storage Power Conversion System under Unbalanced
Voltage Sag Conditions. IEEE Transactions on Energy Conversion, vol. 25, no. 4, pp.
1195-1208, 2010.
[9] Bollen, M. H., Understanding Power Quality Problems: Voltage Sags and Interrup-
tions. IEEE Press, 1999.
[10] Hajizadeh, A., Golkar, M. A., Intelligent Power Management Strategy of hybrid
Distributed Generation System. International Journal of Electrical Power and Energy
Systems, vol. 29, no. 10, pp. 783-795, 2007.
[11] Hajizadeh, A., Golkar, M. A., Fuzzy Neural Control of a Hybrid Fuel Cell/Battery
Distributed Power Generation System. IET Renewable Power Generation, vol. 3, no. 4,
pp. 402-414, 2010.
[12] Wang, L. X., Back-Propagation of Fuzzy Systems as Nonlinear Dynamic System
Identifier. IEEE International Conference on Fuzzy Systems, San Diego, California, US,
pp. 1409-1418, 1992.
[13] ChingHung, L., ChingCheng, T., Identification and Control of Dynamic Systems Using
Recurrent Fuzzy Neural Networks. IEEE Transactions on Fuzzy Systems, vol. 8, no. 4,
pp. 349-366, 2000.
Chapter 11
OPTIMAL SHORT-TERM SCHEDULING OF

AGGREGATORS FOR PLUG-IN ELECTRIC VEHICLES
Mohammad Moradi-Dalvand*1 and Behnam Mohammadi-Ivatloo†2

1
Department of Electrical Engineering,
Power and Water University of Technology, Tehran, Iran
2
Faculty of Electrical and Computer Engineering,
University of Tabriz, Tabriz, Iran
ABSTRACT
Plug-in hybrid electric vehicles (PHEVs) are fast growing options for the next
generation of transportation systems. With the increasing number of PHEVs in power
systems, the system load will increase accordingly. If PHEVs are treated as conventional
loads (uncontrolled loads), the large amount of charging power will take place at peak
hours when drivers head back home in the evening. Since most drivers are expected to
charge their vehicles in the morning and have no preferences about the exact time of
charging, PHEVs are more flexible loads than the traditional ones. It is predicted that
there will be one million PHEVs on U.S. roads within 5 to 10 years. A new organization,
which is called a PHEV aggregator, is required for utilizing the flexibilities and charging
control of PHEVs. PHEV aggregator is a new participant of electricity markets which is
responsible for providing power and controlling the charging pattern of PHEVs under its
territory. The aggregator should buy power from the wholesale electricity market. Thus,
it needs to forecast power prices for the next day. Moreover, the aggregator should
predict PHEVs' consumption, charging duration, and arrival and departure time. This
chapter deals with the short-term scheduling of aggregators in day-ahead electricity
markets considering price uncertainty. Market prices, as the most important uncertainty,
will be modeled using a stochastic programming technique. Proper scenario generation
and reduction methods are utilized for generating efficient scenarios. The objective of
PHEV aggregator is to maximize its profit by charging PHEVs at low-price hours.
Constraints of electricity market and characteristics of PHEVs are incorporated into the
proposed method. An illustrative example is provided to describe the proposed method.
*
E-mail: moradi.dalvand@gmail.com.
†
E-mail: mohammadi@ieee.org.
294 Mohammad Moradi-Dalvand and Behnam Mohammadi-Ivatloo
Keywords: Plug-in electric vehicle (PHEV), Energy market, PHEV aggregator, Stochastic
programming
1. INTRODUCTION
1.1. Background and Review of the Related Works
Plug-in hybrid electric vehicles (PHEVs) are a new paradigm in transportation. Some
reasons of introducing this paradigm can be expressed as follows:
1 Increasing emissions due to burning fossil fuels and related problems such as global
warming,
2 Restriction on fossil fuel resources and need for using other sources of energy for
vehicles, and
3 Advances in storage and battery technologies.
Introducing PHEVs to power system will bring some advantages and disadvantages.
Using the storage capability of PHEVs may have some advantages for power systems such as
frequency regulation and load-shifting. However, since vehicles usually connect to the grid in
the evening when the system load is high, charging PHEVs may introduce new challenges to
the power system such as a need for more generation capacity and enforcing transmission and
distribution systems. A good idea which has been proposed for handling the disadvantages of
the connection of PHEVs to the power system is smart charging that can be helpful because
the main characteristic of PHEVs, i.e. having storage capability and no need for immediate
charging generally. Based on this idea, PHEVs can be charged at midnight hours when the
system demand is very low and also grid integration of the renewable generation is facilitated.
A new entity which should be introduced for implementing smart charging and relating
PHEVs to the electricity market is the PHEV aggregator. PHEV aggregation can result in
lower charging cost for PHEV owners through smart charging at low-price hours. Moreover,
PHEV aggregator may bring some benefits to the power system via shifting demand to off-
peak hours and therefore decreasing the power demand at peak hours. Two main patterns of
PHEV charging include unidirectional and bidirectional ones. In the unidirectional charging,
PHEV aggregator only purchases energy from grid to charge batteries of PHEVs and can also
participate in the ancillary service market using smart charging which controls charging rate
of PHEVs. In the bidirectional charging, PHEV aggregator can sell power to the grid in
addition to buying electricity from it. Unidirectional charging is the first step in the
integration of PHEVs into the electricity grid and markets [1, 2], while bidirectional charging
requires additional hardware that is not currently installed on the produced PHEVs [3]. It
should be mentioned that using bidirectional charging increases battery cycles and
degradation [4]. Hence, in this chapter, the unidirectional charging is considered. Although
PHEV aggregator may provide ancillary services for the system [5, 6] and stability [7], this
chapter is focused on risk constrained optimal participation in the energy market. The driving
cost of PHEVs is less than that of conventional vehicles (CVs) and hybrid electric vehicles
(HEVs). Annual cost of a CV is $1,196, while it is $549 for PHEV when no control strategy
Optimal Short-Term Scheduling of Aggregators for Plug-In Electric Vehicles 295
is imposed on PHEV charging [8]. If the charging profile of a PHEV is controlled, this cost is
reduced to $393 [8].
In this chapter, the optimal bidding of a PHEV aggregator to the electricity market which
is inspired by the Iran electricity market will be discussed and formulated.
1.2. PHEV Configurations
Series Configuration
The required traction force of the vehicle is solely provided by the electric motor in the
series PHEV. Fig, 1 shows this type of configuration. As shown in this figure, the required
power for electric motor is provided by either battery pack or a generator driven by an
internal combustion engines (ICE). If the battery state of charge (SOC) drops below a certain
level, the ICE would drive the electric generator and the battery would be charged by the
produced power of the generator. The series configuration requires two electric machines and
two converters. The electric motor should be rated for maximum power. There is no
mechanical linkage between ICE and mechanical transmission which lets engine to operate in
the most efficient region [9].
Figure 1. Series configuration of PHEVs (adapted from [9]).
Parallel Configuration
Fig. 2 shows the parallel configuration of PHEVs, in which the traction force is provided
by either electric motor or ICE. The eclectic motor and ICE are connected to mechanical
transmission system using a mechanical coupling. The motor can also be operated as a
generator. Unlike the series configuration, parallel configuration uses only one electric
machine and one converter. The engine and electric motor of PHEV can be rated at reduced
levels for short distance trips [9].
Figure 2. Parallel configuration of PHEVs (adapted from [9]).
Series-Parallel Configuration
Series-parallel configuration takes the advantages of both series and parallel
configurations. Fig 3. represents the series-parallel configuration of the PHEV. As can be
observed in this figure, this PHEV can operate using the electric motor alone or along with
ICE. Operating point of the ICE affects the efficiency of the ICE and also power transfer [10].
Similar to the series configuration, it is possible to charge batteries by either grid power or
electric generator.
It is worth mentioning that there are other configurations of PHEV that cannot be
classified into the aforementioned series/parallel configurations, but are categorized as
complex configuration [11].
Figure 3. Series-parallel configuration of PHEVs (adapted from [9]).

1.3. Chapter Organization
The remainder of this chapter is organized as follows: an introduction to stochastic

programming in addition to the description of scenario generation/reduction methods is
presented in Section 2. Risk controlling approaches in stochastic programming are also
discussed in this section. The proposed problem formulation is presented in Section 3. A
comprehensive illustrative example with different case studies is provided in Section 4.
Finally, the concluding remarks are provided in Section 5.
2. STOCHASTIC PROGRAMMING
2.1. Stochastic Versus Deterministic Programming
Different solutions exist in the real-world decision making problems. Finding the best
solution is called programming or optimization. For each problem, there exists some input
data, which are called parameters, and some decisions which should be made in the
optimization procedure and are called variables. In some cases, value of the parameters is
definite. These problems are called deterministic programming problems. However, in most
of the real-world problems, all or at least some parameters are uncertain; this type of
problems is called stochastic programming. For this type of programming, decision should be
made in an uncertain framework. In most problems, despite the exact values of uncertain
parameters, their probability distribution functions are known.
Since the solving procedure of deterministic programming problems is usually more
straightforward than that of stochastic problems, it is desired to cast stochastic programming
problems into the deterministic ones. The main key for this conversion is the probability
functions of the uncertain parameters. Generally, such a conversion is done using two
approaches. The first way is by replacing uncertain parameters with their expected values.
The second one is generating some probable instances of uncertain parameters using the
information of their probability functions which are called scenarios and replacing them with
uncertain parameters.
2.2. Risk Control in Stochastic Programming
Objective function of the stochastic programming problem can be a random value.

Characteristics of this random value should be utilized to obtain a single and deterministic
value. The expected value (mean) is the most widely used index for forming the objective
function.
The expected value of a random variable does not provide any information about its
deviation and therefore can lead to a great risk. Some decision makers are concerned about
the associated risk of their decisions. For these decision makers (called risk-averse), risk
measurement should be defined and incorporated into the optimization problem as a term of
the objective function or constraints in addressing their concerns. Different risk management
approaches have been proposed for scheduling conventional market participants (generation
companies, large consumers, and retailers) in the literature [12, 13]. A review of the
application of stochastic programming methods in deregulated energy markets was provided
in [14]. Value at risk (VaR) is a measure that estimates the financial loss of a decision-maker
resulting from price fluctuations with a certain confidence level [15]. Conditional value at risk
(CVaR) is an extension of VaR which is defined as the difference between the expected profit
and the average of potential profit values which are less than VaR [16]. Compared with VaR,
CVaR is a more conservative index, because it measures the average loss exceeding the VaR
value for a given confidence level [13].
2.3. Scenario Generation and Reduction
Figure 4. Flowchart of the scenario generation approach using MCS.
Each uncertain parameter is a random variable and its probability function is usually
known. Some random variables which are varying over time are called stochastic process. A
single realization of a stochastic process is called a scenario [17]. In scenario generation
approach, a probability distribution function based on the previous experiences is considered
for each stochastic variable. Generally, normal distribution is used for modeling the market
price [18]. For normal distribution, mean and variance of the stochastic variable should be
calculated. It should be mentioned that every distribution can be utilized to model the market
price and the proposed solution procedure is not restricted to the normal distribution. In this
study, Monte Carlo simulation (MCS) was used to generate scenarios [19, 20]. Fig. 4 shows
the flowchart of MCS for scenario generation. Market price sampling will be described in
Section 4.
Figure 5. Flowchart of the implemented scenario reduction approach.

Generally, so many samples are generated for extracting the characteristics of the
stochastic process. Despite the necessity of generating many scenarios, a large number of
scenarios can lead to computational intractability. For solving this problem, some scenario
reduction methods have been developed, the main goal of which is decreasing cardinality,
while keeping the maximum possible information of the deleted scenarios [17]. In this
chapter, backward method was used for scenario reduction [21, 22], the main idea of which is
eliminating the scenario with the lowest information and repeating this action until the
stopping criterion is satisfied. Two common stopping criteria are reduction percentage and
number of preserved scenarios. Reduction percentage is defined in terms of relative distance
between reduced and original scenarios. Algorithm of backward method is presented in Fig.
5. Original scenarios and their probabilities are input data. The eliminated set at the beginning
of the algorithm is empty. Distance of two scenarios is calculated and denoted by D (.,.) . For
each scenario, the nearest scenario and its associated distance are determined. The nearest
scenario is denoted by index r . The probability distance for each scenario, i.e., PD r (s ) , is
calculated as its probability, i.e.,  s , times distance with its nearest scenario, i.e., Ds (r ) . The
scenario which has the minimum probability distance is chosen for elimination, i.e., PDd .
Probability of the deleted scenario is added to the probability of the nearest scenario.
2.4. Two-Stage Stochastic Programming
Stochastic programming problems happen in a horizon. Stage in the problem horizon is

defined as a point of time, at which some decisions should have been made when
uncertainties partially or totally vanish [17]. Based on this definition, stochastic programming
problems can be divided into two-stage and multi-stage stochastic programming problems.
Since the studied problem in this chapter was a two-stage stochastic programming and the
multi-stage stochastic programming problems can be considered as the extension of the two-
stage one, only two-stage stochastic programming was introduced.
In two-stage stochastic programming, uncertainties vanish in one step. Therefore, these
problems consist of two stages: before and after scenario realization. Thus, two types of
decisions can be distinguished: first-stage or here-and-now decisions and second stages or
wait-and-see decisions. The first stage decisions should be made before the realization of
uncertain parameters. Since the scenarios are not realized at this stage, these decisions do not
depend on the value of scenarios. After making the here-and-now decisions, decision-makers
should wait until the uncertainties are vanished and make some decisions afterward. The wait-
and-see decisions depend on the scenarios.
Scenarios can be visualized using a scenario tree. A typical scenario tree with three
scenarios is shown in Fig. 6. Each scenario tree is composed of nodes and branches. Nodes
represent the state of the problem and branches show transition between stages where
scenarios are realized. The first node of the problem is called “root” and the nodes of the last
stage are called “leaves”. Each way between the root node and leaves is a scenario.
A typical two-stage stochastic programming problem can be formulated as follows [17]:
max c x  E(Q (s ))
x
subject to (1)
Ax  b
x X
where Q (x , s ) is a stochastic part of the objective function obtained using the following
formulation.

Q (x , s )  max d (s ) y (s )
y (s )
subject to (2)
T (s )x W (s ) y (s )  h (s )
y (s ) Y  , s  S
where x and y (s ) are the first and second stage decision variables and A , b , d (s ) , T (s ) ,
W (s ) , and h (s ) are parameters, and E (.) represents the expected value of its argument. This
problem can be expressed as deterministic problem considering some assumptions using the
following formulation [17]:
max c x    (s )d (s ) y (s )
x , y (s )
s S
subject to
Ax  b (3)
T (s )x W (s ) y (s )  h (s ), s  S
x  X , y (s ) Y , s  S
CVaR can be used as a risk controlling measure and incorporated into two-stage
stochastic programming. Mathematical formulation of the CVaR is as follows [17]:
 
E s max   f (x , s ), 0 ,   (0,1)
1
CVaR( , x )  max   (4)
 1  
CVaR can also be used as a term in the objective function and/or a constraint for risk
limitation. Finally, the formulation of the extended model can be stated as follows:
 1 
max c x    (s )d (s ) y (s )       (s )sc 
 
s
x , y ( s ), , sc s
s S  1 s S 
subject to
Ax  b
  c x  q (s ) y (s )   sc s , s  S (5)
sc s  0, s  S
T (s )x W (s ) y (s )  h (s ), s  S
x  X , y (s ) Y , s  S
where  is an auxiliary variable, sc s equals max{  c x  q (s ) y (s )  ,0} , and  is a

positive weighting parameter of risk measure.
Figure 6. Simple illustration of two-stage scenario tree.
3. PROBLEM FORMULATION
Main objective of PHEV aggregator is maximizing its total profit. Revenue of the PHEV
aggregator comes from selling power to PHEVs and cost of aggregator is related to
purchasing electricity from the grid. It is assumed that PHEV charging tariffs are
predetermined based on the regulatory constraints and market gaming with other aggregators.
If an aggregator sells electricity at a more expensive cost than others, PHEV owners will
migrate to other aggregators. Determining energy tariff is a long- or mid-term decision which
is considered as the input data for the short-term scheduling problem. Therefore, PHEV
aggregator should charge PHEVs at the low-electricity-price hours such as off-peak hours to
maximize its profit.
For increasing total profit and avoiding potential risks, uncertainties should be handled by
the aggregator in the decision making procedure. The main uncertainty in this current
framework is energy price. Economic effect of the price forecasting errors of four forecasting
methods on the short-term scheduling of thermal and hydro producers was presented in [23].
Considering this uncertainty, the optimal short-term operation scheduling of the aggregator
would become a stochastic programming problem.
3.1. Market Structure
One side of electricity market was considered in this chapter, in which only electricity
producers were competing. The generators participate in the day-ahead market through
offering value‒price pairs and consumption side of the hypothesized market, such as retailers
and PHEV aggregators, only submit their demands to the system operator. If demand side
players deviate from their stated demand, they incur imbalance costs. This market model is
inspired by Iran electricity market.
3.2. Decision Framework
In this model, the PHEV aggregator should decide about its consumption for all day-
ahead hours. Thus, the aggregator should estimate the required power for charging PHEVs.
This decision is made before the realization of energy prices. Afterward, the cost associated
with the consumed energy is determined. The scenario set in this problem consists of
electricity prices for each hour of the decision horizon and their associated probabilities. After
scenario generation and reduction, the proposal of the aggregator to the market should be
optimized. Since the power required for charging PHEVs ( E ,t ) should be determined before
the realization of energy price, it is a first stage or here-and-now decision and does not
depend on the scenario realization. Other state variables which depend on E ,t such as state
of charge ( SOC ,t ) and scheduled demands ( Ft ) are first stage decisions. Electricity cost is
determined after the realization of scenarios as the second stage variables.
3.3. Mathematical Formulation
The problem objective is to maximize the aggregator benefit which is formulated in (6).
T Ns
max   s ( t  s ,t )Ft (6)
Ft , E ,t
t 1 s 1
where,  s is the probability of the occurrence of the s th scenario, t is the price of selling
energy to PHEV owners at the t th hour ($/KWh), and s ,t is the s th realized electricity
price at the t th hour ($/KWh). In this formulation, it was assumed that the aggregator sells
energy to PHEV owners with constant and predetermined tariffs. Based on this assumption,
the benefit term is fixed and therefore the objective function can be rewritten as (7):
T Ns
max   s (s ,t )Ft (7)
Ft , E ,t
t 1 s 1
Formulation of the risk-averse two-stage stochastic programming problem using CVaR

risk measure results in the following object function:
T Ns
 1 
max
Ft , E ,t ,sc s , 

t 1 s 1
s (s ,t )Ft    
 1  

s S
 s sc s 

(8)
Also, the two following constraints should be added to the model to limit the risk.
T
  ( s ,t Ft )  sc s , s  S (9)
t 1
scs  0, s  S (10)
Arrival, departure, and battery usage of PHEVs are considered known according to [1,
26]. The sold electricity to PHEV owner should be provided by the purchased energies from
the grid. This relation is shown in the following equation:


E ,t  Ft t (11)
SOC of the battery of each PHEV is updated using power which is drawn from grid, the
power used for driving ( G ,t ), and its SOC at the previous hour using the following equation:
SOC ,t  SOC ,t 1  ( E ,t G ,t ) / C t ,  (12)
where, C  is the battery capacity of the  th vehicle and  is its charging efficiency.
SOC and rate of charge bounds are forced using (13) and (14), respectively.
SOC   SOC ,t  SOC  t ,  (13)
E  E ,t  E t ,  (14)
where SOC  is minimum SOC and SOC  is the capacity limit of battery of the v th vehicle.
Moreover, minimum and maximum charging rates are denoted by E and E  , respectively.
Constraints (15) imposed on PHEV can be charged only at plugged times.
E ,t  0 {t ,  | G ,t  0} (15)
PHEVs should be charged to a sufficient level before departure time, representing by the
following constraints:
SOC ,t  SOCdep
,t {t ,  | I dep
,t  1} (16)
,t is a flag that indicates the departure of vehicle  at hour t and the required level
where I dep
of charging is denoted by SOCdep
,t .
4. ILLUSTRATIVE EXAMPLE
In this section, a simple case study is used to demonstrate different bidding strategies of a
typical PHEV aggregator. Characteristic data PHEVs were adopted from [27]. For the sake of
simplicity, a few numbers of PHEVs were considered, i.e. five PHEVs in the six hour
scheduling period. Each PHEV battery had maximum and minimum allowable capacities of
25 KWh and 3 KWh. Charging efficiency of PHEVs was assumed to be 100% for the sake of
simplicity. The maximum charging rate of PHEVs was limited to 50% of their capacities and
therefore approximately 2 h was required for full charging. Minimum charging rate was
considered zero. The optimization problem was solved using CPLEX solver [29] under
general algebraic modeling system (GAMS) framework [28].
Means and variances of the electricity price are shown in Table 1.. Fig. 7 shows the
typical discretization of the normal probability distribution function. The continuous normal
probability distribution function is discretized at seven levels. Each error level has the width
of one forecast error standard deviation (  ) [30]. Fig. 8 shows the flowchart of the
generating electricity price scenarios.
Table 1. Expected and standard deviations of electricity price ($/KWh)
Hour te ($/KWh)  te ($/KWh)

1 50 10
2 47.5 9.74
3 45 9.48
4 42.5 9.22
5 40 8.94
6 42.5 9.22
Departure and arrival times of PHEVs along with the their initial SOC at the beginning of
the scheduling period, arrival SOC, and the required SOC at the end of the scheduling period
are shown in Table 2..
The Monte Carlo method which is described in Fig. 4 and Fig. 8 was used to generate
100 scenarios for energy prices. These scenarios were reduced using backward method. The
SCENRED tool under GAMS environment was used for scenario reduction. Ten scenarios
were retained after scenario reduction, as shown in Table 3..
0.4
0.3
0.2
0.1
0
-3 -2 -   + +2 +3
-5/2 -3/2 -/2 +/2 +3/2 +5/2
Figure 7. Discretization of normal probability density function.
Figure 8. Flowchart of the generating electricity price scenarios.

Table 2. Characteristics of PHEVs
Arrival SOC Required final

PHEV number Initial SOC Departure Time Arrival Time
(%) SOC (%)
1 50.00% 13:00 14:00 60% 100%
2 50.00% 13:00 15:00 20% 100%
3 15.00% 14:00 16:00 20% 100%
4 15.00% 15:00 17:00 20% 60%
5 15.00% 16:00 17:00 60% 100%
Table 3. Reduced scenarios for power prices over the planning horizon
Scenario Hour 1 Hour 2 Hour 3 Hour 4 Hour 5

1 40 37.76 63.96 42.5 48.94
2 50 47.5 45 51.72 40
3 50 47.5 35.52 42.5 31.06
4 60 47.5 54.48 42.5 48.94
5 40 47.5 54.48 42.5 40
6 60 47.5 54.48 42.5 40
7 50 57.24 45 42.5 48.94
8 50 37.76 45 42.5 48.94
9 50 57.24 35.52 33.28 48.94
10 40 47.5 35.52 42.5 31.06
Different cases were investigated. At first, deterministic scheduling was done. Then,
stochastic scheduling was performed without considering risk and finally CVaR as a risk
measure was added to the constraints. In this step, risk-neutral and risk-averse scheduling
were compared. Moreover, the effect of the degree of risk aversion was studied.
4.1. Deterministic Scheduling
In deterministic scheduling, uncertain parameters, i.e. energy price, are replaced with
their expected values. SOC of PHEV in the planning horizon is shown in Fig. 9. Expected
cost of supplying PHEVs in the scheduling horizon was $6.489.
Table 4. Cost of each scenario in the case of stochastic programming
Scenario Cost ($) Scenario Cost ($)

1 7.147 6 7.648
2 7.686 7 7.242
3 6.983 8 7.081
4 8.334 9 6.971
5 7.48 10 6.138
4.2. Stochastic Scheduling
The reduced ten scenarios were used for the realization of energy prices. SOC of PHEV
is shown in Fig. 10. In comparison with the results of deterministic scheduling, the schedule
was different for the 4th hour of the first and second PHEVs and the 5th hour of the second
PHEV. Schedules were the same for other PHEVs at other hours. The expected cost of
supplying PHEVs was $6.863, which seemed to be larger than the expected cost of
deterministic scheduling of PHEVs. Costs of all the scenarios are shown in Table 4..
PEV1 PEV2 PEV3

100 100 100
80 80 80
SOC (%)
SOC (%)
SOC (%)
60 60 60
40 40 40
20 20 20
0 0 0
1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6
Time Time Time
PEV4 PEV5
100 100
80 80
SOC (%)
SOC (%)
60 60
40 40
20 20
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Time Time
Figure 9. SOC of PHEVs during planning hours for the case of deterministic programming.
PEV1 PEV2 PEV3

100 100 100
80 80 80
SOC (%)
SOC (%)
SOC (%)
60 60 60
40 40 40
20 20 20
0 0 0
1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6
Time Time Time
PEV4 PEV5
100 100
80 80
SOC (%)
SOC (%)
60 60
40 40
20 20
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Time Time
Figure 10. SOC of PHEVs during planning hours for the case of stochastic programming.
In order to evaluate the effect of stochastic programming on the final cost of aggregator,
it was assumed that scenarios #1 and #8 were realized at real operation hours, which could
also be called after the fact energy prices. The real operation cost for the stochastic and
deterministic cases are compared in Table 5.. It can be observed that the real cost of supply
was less for both of the after the fact values when stochastic scheduling was employed.
4.3. Risk-Constrained Scheduling
The scheduling was done for  -CVaR with different values for  .  -CVaR with  =
0.75 was used and results are shown in Table 6.. SC equaled zero for all scenarios, expect
scenario #8. It can be seen that the expected cost decreased and the objective value increased
with increasing  .
Table 5. Comparing real cost of aggregator for the realizations of scenarios 1 and 8 for
deterministic and stochastic programming
Scenario #1 as after the fact values

Deterministic 7.2432
Stochastic 7.1466
Scenario #8 as after the fact values
Deterministic 7.1778
Stochastic 7.0812
Table 6. Results of risk-constrained stochastic scheduling
 Objective Value Cost SC(8) 

0 -6.864 6.864 0.368 -1.487
1 -8.288 6.877 0.368 -1.102
2 -9.699 6.877 0.368 -1.102
3 -11.111 6.877 0.305 -1.102
5 -13.932 6.921 0.305 -1.146
10 -20.944 6.921 0.368 -1.146
CONCLUSION
In this chapter, optimal short-term operation scheduling of a typical PHEV aggregator
was studied. A typical aggregator faced different uncertainties for day-ahead scheduling and
an important one was energy price. Stochastic programming can be used as a powerful
method for robust scheduling of PHEV charging. The objective of the aggregator in the
proposed method was to maximize its total profit, while supplying the required energy of
PHEVs. For a typical aggregator, the uncertainty of electricity price forecasting was
considered and the short-term operation of aggregator was formulated as a stochastic
optimization problem. Different stochastic and deterministic results were presented and
compared for the aggregator, demonstrating that stochastic programming can result in more
benefit (through lower cost) than deterministic programming for two after the fact energy
prices. Moreover, for risk-averse aggregator, different schedules with naturally more costs
were obtained.
REFERENCES
[1] Sortomme, E.; El-Sharkawi, M.; Optimal charging strategies for unidirectional vehicle-
to-grid. IEEE Transactions on Smart Grid, 2 (1) (2011) 131-138.
[2] Sortomme, E.; El-Sharkawi, M.; Optimal combined bidding of vehicle-to-grid ancillary
services. IEEE Transactions on Smart Grid, 3 (1) (2012) 70-79.
[3] Peterson, S. B.; Whitacre, J.; Apt, J.; The economics of using plug-in hybrid electric
vehicle battery packs for grid storage. Journal of Power Sources, 195 (8) (2010) 2377-
2384.
[4] Peterson, S. B.; Apt, J.; Whitacre, J.; Lithium-ion battery cell degradation resulting
from realistic vehicle and vehicle-to-grid utilization. Journal of Power Sources, 195 (8)
(2010) 2385- 2392.
[5] Guille, C.; Gross, G.; A conceptual framework for the vehicle-to-grid (v2g)
implementation. Energy Policy, 37 (11) (2009) 4379-4390.
[6] Tomic, J.; Kempton, W.; Using fleets of electric-drive vehicles for grid support.
Journal of Power Sources, 168 (2) (2007) 459-468.
[7] Kempton, W.; Tomic, J.; Vehicle-to-grid power implementation: From stabilizing the
grid to supporting large-scale renewable energy. Journal of Power Sources, 144 (1)
(2005) 280-294.
[8] Sioshansi, R.; Miller, J.; Plug-in hybrid electric vehicles can be clean and economical in
dirty power systems. Energy Policy, 39 (10) (2011) 6151-6161.
[9] Zhang, X.; Mi, C.; Vehicle Power Management Modeling, Control and Optimization.
Springer-Verlag, 2011.
[10] Park, J.; Oh, J.; Park, Y.; Lee, K.; Optimal power distribution strategy for series-parallel
hybrid electric vehicles. The 1st International Forum on Strategic Technology,
Gyeonggi, South Korea, (2006) 37-42.
[11] Zhang, P.; Yin, C.; Zhang, Y.; Wu, Z.; Optimal energy management for a complex
hybrid electric vehicle: Tolerating power-loss of motor. Journal of Shanghai Jiaotong
University (Science), 14 (4) (2009) 476–481.
[12] Kazempour, S.; Moghaddam, M. P.; Haghifam, M. R.; Yousefi, G. R.; Risk-constrained
dynamic self-scheduling of a pumped-storage plant in the energy and ancillary service
markets. Energy Conversion and Management, 50 (5) (2009) 1368–1375.
[13] Mohammadi-Ivatloo, B.; Zareipour, H.; Amjady, N.; Ehsan, M.; Application of
information-gap decision theory to risk-constrained self-scheduling of gencos. IEEE
Transactions on Power Systems, 28 (2) (2013) 1093-1102.
[14] Most, D.; Keles, D.; A survey of stochastic modelling approaches for liberalised
electricity markets. European Journal of Operational Research, 207 (2) (2010) 543–556.
[15] Garcia-Gonzalez, J.; Moraga, R.; Mateo, A.; Risk-constrained strategic bidding of a
hydro producer under price uncertainty. 2007 IEEE Power Engineering Society General
Meeting, Madrid, Spain, (2007) 1–4.
[16] Garcia-Gonzalez, J.; Parrilla, E.; Mateo, A.; Risk-averse profit-based optimal
scheduling of a hydro-chain in the day-ahead electricity market. European Journal of
Operational Research, 181 (3) (2007) 1354–1369.
[17] Conejo, A. J.; Carrion, M.; Morales, J. M.; Decision Making Under Uncertainty in
Electricity Markets. Springer, 2010.
[18] Li, T.; Shahidehpour, M.; Li, Z.; Risk-constrained bidding strategy with stochastic unit
commitment. IEEE Transactions on Power Systems, 22 (1)(2007) 449–458.
[19] Kall, P.; Wallace, S. W.; Stochastic Programming. John Wiley & Sons, 1994.
[20] Birge, J. R.; Louveaux, F.; Introduction to Stochastic Programming. Springer, 2011.
[21] Growe-Kuska, N.; Heitsch, H.; Romisch, W.; Scenario reduction and scenario tree
construction for power management problems. 2003 IEEE Bologna Power Tech
Conference, Berlin, Germany (2003).
[22] http://www.gams.com/dd/docs/solvers/scenred.pdf, [accessed on September, 2012].
[23] Mohammadi-Ivatloo, B.; Zareipour, H.; Ehsan, M.; Amjady, N.; Economic impact of
price forecasting inaccuracies on self-scheduling of generation companies. Electric
Power System Research, 81 (2) (2011) 617-624.
[24] U. D. of Energy, Benefits of demand response in electricity markets and
recommendations for achieving them, A report to the United States congress, February,
2006.
[25] Chang-Chien, L.R.; An, L. N.; Lin, T.W. ; Lee, W.J.; Incorporating demand response
with spinning reserve to realize an adaptive frequency restoration plan for system
contingencies. IEEE Transactions on Smart Grid, 3 (3) (2012) 1145-1153.
[26] Sortomme, E.; El-Sharkawi, M.; Optimal scheduling of vehicle-to-grid energy and
ancillary services. IEEE Transactions on Smart Grid, 3 (1) (2012) 351–59.
[27] Pantos, M.; Stochastic optimal charging of electric-drive vehicles with renewable
energy, Energy. 36 (11) (2011) 6567-6576.
[28] Brooke, A.; Kendrick, D.; Meeraus, A.; Raman, R.; GAMS: A user‟s guide. GAMS
Development Corporation, 1998.
[29] Brooke, A.; Kendrick, D.; Meeraus, A.; aman, R.; GAMS/Cplex 7.0 User Notes. GAMS
Development Corporation, 2000.
[30] Amjady, N.; Aghaei, J.; Shayanfar, H.; Stochastic multiobjective market clearing of
joint energy and reserves auctions ensuring power system security. IEEE Transactions
on Power Systems, 24 (4) (2009) 1841-1854.
Chapter 12
ENERGY EFFICIENCY OF THE HYBRID

POWER SOURCE USED IN THE PLUG-IN
FUEL CELL VEHICLES
Nicu Bizon1,2,, Marian Raducu1, Mihai Oproescu1

and Luminita Mirela Constantinescu1
1
University of Pitesti, Pitesti, Romania
2
University Politehnica of Bucharest, Bucharest, Romania
ABSTRACT
In this chapter a mathematical analysis of the energy efficiency for the series and
parallel Multiport Power Converter (MPC) topologies used in the Plug-in Fuel Cell
Vehicles (PFCVs) is presented. Besides this study, that is performed and validated in the
second part, the first part presents a brief overview over the electric vehicles with a range
extender, which also include the electrical motors and inverter systems used, the main
parameters of the DC input sources for electrical vehicle, and the usual strategies for the
energy management unit. The aim of the analytical study shown in the second part is to
provide some general strategies for the energy management unit operating under a wide
range of PFCV regimes. The goal is to choose the optimal parameters of the MPC control
for an efficient use of the PFCV during each particular drive cycle. In relation with the
FC system of PFCV, the Energy Storage System (ESS) can be operated in the following
regimes: (1) charge-sustaining, (2) charge-depleting, and (3) charge-increasing.
Considering the imposed window for the ESS state-of-charge, the MPC can be
connected to the renewable plug-in charging stations to exchange power with a smart
grid, when it is necessary for both. The energy management unit that communicates with
the smart grids will establish the moments to match the PFCV power demand with the
grid supply availability, stabilizing it. The MPC energy efficiency of the PFCVs is
studied when the ESS is charged (discharged) from (to) the smart grid. The comparative
results were shown for both PFCV architectures through the analytical calculation
performed and the appropriate Matlab/Simulink® simulations presented.

Corresponding author: Nicu Bizon. University of Pitesti, Pitesti, Romania. E-mail: nicubizon@yahoo.com, nicu.
bizon@upit.ro.
314 Nicu Bizon, Marian Raducu, Mihai Oproescu et al.
Keywords: Plug-in Vehicles, Fuel Cell Hybrid Power Source, Energy Storage System, State-
of-Charge, Multiport Power Converter, Energy Efficiency
NOMENCLATURES
RES Renewable Energy Sources
PV Photovoltaic panels
WT Wind Turbines
HT Hydraulic Turbines
PEMFC Proton Exchange Membrane Fuel Cell
ESD Energy Storage Devices
HPS Hybrid Power Source
HEV Hybrid Electric Vehicles
FCV Fuel Cell Vehicles
EGS Energy Generation Systems
DG Distributed Generation
UPS Uninterruptible Power Supplies
EMU Energy Management Unit
ESS Energy Storage System
MPPT Maximum Power Point Tracking
MEP Maximum Efficiency Point
SOC State-of-Charge
EV Electric Vehicle
ICE Internal Combustion Engine
PCS Plug-in Charging Stations
PHEV Plug-in Hybrid Electric Vehicles
PFCV Plug-in Fuel Cell Vehicles
FEV Full Electric Vehicle
EREV Electric Vehicle with a Range Extender
PI Power Interface
IM Induction Motor
SRM Switched Reluctance Motor
PMSM Permanent Magnet Synchronous Motor
BLDC Brushless DC motor
PMBLDC PM Brushless DC
ZSI Z-Source Inverter
DSP Digital Signal Processing
CVS Controlled Voltage Source
CCS Controlled Current Source
UC Ultracapacitor
AGM Absorbed Glass Mat
VRSLAB Valve Regulated Sealed Lead-Acid Battery
CS Charge-Sustaining
CD Charge-Depleting
Energy Efficiency of the Hybrid Power Source … 315
CI Charge-Increasing
V2G Vehicle-to-Grid
MPCs Series MPC
MPCp Parallel MPC
1. INTRODUCTION
In the last decade an increased interest in flexible and effective systems of power
management, which are able to harvest energy from various sustainable sources of energy,
such as the Renewable Energy Sources (RES) (photovoltaic panels - PV, wind turbines - WT,
hydraulic turbines - HT etc.) or hydrogen-based energy sources (Proton Exchange Membrane
Fuel Cell - PEMFC) appeared.
Using at least one energy source from those mentioned above and Energy Storage
Devices (ESD: batteries, ultracapacitors, etc.) combined in a hybrid stack (passive, semi-
active or active controlled mode). These merged technologies are used as Hybrid Power
Source (HPS) in Hybrid Electric Vehicles (HEVs) and Fuel Cell Vehicles (FCVs) [1], in
renewable Energy Generation Systems (EGS) [2], in distributed generation (DG) in smart
grids [3], in uninterruptible power supplies (UPS) [4], and so on. Usually, the voltage-current
characteristics of the energy sources and storage devices are nonlinear and their voltage levels
are different from those of the loads. Consequently, a Multi-Port Converter (MPC) must be
used to manage the power flow between the inputs and the outputs (energy sources, storage
devices and loads) based on an Energy Management Unit (EMU) [5].
The concept of MPC (or Multi-port Power Electronic Interface) is commonly adopted to
process the renewable power from multiple sources and loads [6-8], having the following
main features: (1) maximum energy harvesting from renewable sources, (2) optimal
management of energy from multiple sources, (3) optimal Energy Storage System (ESS)
management, and (4) adaptive energy management system for the best performance. The
MPC represents a particular case of energy hub concept that is considered as a unit where
multiple energy carriers can be converted, conditioned, and stored, representing an interface
between different energy sources and loads [9-11]. For example, the MPC interfacing the
HEV DC buses must meet the following requirements:
› Electrical isolation between the low voltage (LV) and the high voltage (HV) buses;
› Bidirectional power flow for all input and output ports;
› Continuous supplying of the loads;
› Short response time to load changes;
› High energy efficiency and power density
› Safe start-up without complex circuits.
The MPC creates an interface between the loads, the renewable sources, and the storage
devices to efficiently provide and recover the power. Consequently, the Maximum Power
Point Tracking (MPPT) control guarantees the optimal energy harvested from the energy
sources that have a power characteristic with a maximum at their MPP. Generally, it is better
to operate the FC system at the Maximum Efficiency Point (MEP), but MPP can also be used
because this is easier to track in comparison with the MEP that is more difficult to be
estimated [12]. The generic FCV architecture shown in Figure 1 is based on the flexible MPC
topology, generalizing the most used MPC topologies in automotive applications such as the
series and parallel MPC topologies. These topologies request the use of the ESS on the LV
and HV bus, respectively [13, 14].
In [14, 15] it is shown that a hybrid MPP topology is more efficient than both series and
parallel MPC topologies when the ESS is operated to have final State-Of-Charge (SOC)
different from the initial SOC. This case may be of interest in operating the PFCV. In this
chapter will be shown for basic MPC topologies that each could work more efficiently under
certain conditions that will be specified by analytical calculation performed. Consequently,
the basic MPC topology helps to boost the energy efficiency performance as well as the
flexibility in operating the PFCV to increase the driving range and fuel economy. Also, the
mass of the MPC could be reduced by using the modularization and integration techniques for
MPCs [6], but this is off the topic of this chapter. Finally, it can be noted that the MPC
concept represents both integrated PIs and their appropriate control, and also the fact that they
operate efficiently under the energy management strategies implemented in the EMU.
A brief comparison of the energy management approaches applied for the PEMFC
systems in the literature is made in [16]. The energy management of multiple sources and
loads is a complex and difficult task that needs to be solved by the specialists using different
strategies based on (1) intelligent concepts [17, 18], (2) local and global optimization
approaches [19, 20], (3) frequency decoupling techniques [21, 22], (4) linear and nonlinear
controllers used in multiple control loops [23] etc.
The ECU design took recently a more important and detailed place in literature,
especially based on the fuzzy logic and on the nonlinear optimization [16, 24].
Figure 1. Plug-in Fuel Cell Vehicle - generic MPC topology [14].

This design issue is mainly related to the ECU features of easy adaptation to more
complex MPC topologies, computational efficiency, driving rules integration, capacity to
modeling uncertainties, etc. [25].
Real-time optimization based on equivalent consumption minimization strategy may be a
good choice for the online implementation of the ECU [20]. Finally, all the proposed ECU
strategies compare the MPC energy efficiency in term of equivalent hydrogen cost.
In the first part of this chapter (Introduction and the next four sections) the state of art for
EVs and HEVs was briefly presented, especially for the FCVs with or without the plug-in
feature integrated.
So, these sections shortly present the common electrical motors and inverter systems
usually used in automotive applications, the main features of the FC stack and ESS, and the
appropriate EMU strategies.
Consequently, a MPC must be used to manage the power flow between the inputs and the
outputs (energy sources, storage devices and loads) based on an optimum EMU strategy.
Section 6 presents the main MPC topologies proposed in the literature in order to understand
the energy efficiency analysis of the MPC performed in the last sections.
The features of the MPCs, which are scalable to meet different power requirements, are
presented in relation with the DC input sources in Section 7.
The analysis of the two MPC topologies is presented in the last five sections in term of
energy efficiency related to the ESS power flow and the MPC parameters‟ values.
This part is organized as follows. Sections 8 and 9 present the series and parallel MPC
topologies operating in the three operating modes of the ESS. The next two sections (Sections
10 and 11) will compare the energy efficiency for the series and parallel MPC topologies in
each ESS operating mode over another.
This analysis is performed to increase the energy efficiency of the whole MPC for series
and parallel HPS architectures, which means the use of a LV or HV ESS type, respectively.
The analysis is made in term of energy efficiency related to the ESS power flow and MPC
parameters. Finally, some conclusions are given related to the PFCV capacity to stabilize the
grid by charging/discharging the ESS under EMU control.
2. PLUG-IN ELECTRIC VEHICLES WITH A RANGE EXTENDER

The Electric Vehicle (EV) is a technology that promises to drastically reduce emissions
associated with the road transport. In the last decade the technology has been supported by
different manufacturers and specialists as a key element in reducing CO2 emissions (as well
as emissions of pollutants and noise) of cars and light commercial vehicles.
But at the same time, EV technology is still far from being projected as necessary,
emphasizing too many uncertainties regarding the issues to be addressed, such as [26]:
› Well-to-wheel impacts on emissions;

› Interaction with the DG system;
› Cost of large scale introduction;
› The batteries technology (their capacity related to vehicle and road range, fast
charging, durability, availability and environmental impacts of used materials).
In addition, advantages such as high energy-efficiency and zero environmental pollution

will impose the battery-powered EV technology to the internal combustion engine (ICE)
vehicles conventional, even if the EV performance is far less competitive than ICE vehicles,
due to the much lower energy density of the batteries than that of gasoline.
Consequently, to overcome their disadvantages, the HEV technology was promoted by
the main companies that design and produce cars in order to exploit the advantages of the
both ICE and EV technologies. HEV combines an ICE with an on-board rechargeable ESS to
achieve better performances regarding the fuel economy and road range. The HEVs models
are currently produced by Toyota (Lexus), Honda, Chevrolet, Ford, Mercedes Benz etc.
The plug-in concept means the interaction of the vehicle with the DG system through an
internal or external charger. Plug-in vehicles can be classified into different categories such as
EV, HEV, Plug-in Hybrid Electric Vehicles (PHEV), and plug-in fuel cell vehicles (PFCV)
[27]. EVs classification is given in [28], as below:
- Full Electric Vehicles (FEVs) that have an electric motor and no ICE or Fuel Cell
(FC) system.
PHEVs have in addition a charger;
- Electric Vehicles with a Range Extender (EREVs) that have one or more electric
motors and an ICE or a FCS, which is used to extend the road range.
A FCV is an EV with a FC system operating as range extender. The car producers

announced their FCVs based on ongoing research into the development of the hydrogen
technology, even if hydrogen production, storage and distribution are in early stage of
implementation. Thus, the FCVs will have a long way before entering the market.
In this chapter will be shown that many design options for the PFCV-MPC topologies are
available. The FC system powers the DC bus via a DC-DC power converter (Power Interface
1 - PI1) integrated in the PFCV-MPC topology (see Figure 1), where the inverter system and
PI2 are of bidirectional type based on different power converter‟s topologies [29, 30]: bi-
buck, bi-boost, or hybrid integrated topologies.
Note that the power flow during the regenerative braking process is from the load to the
ESS via the inverter diodes and PI2 that operate in buck mode [30]. Considering the higher
cost of the bidirectional type against the unidirectional type, the bidirectional power flow is
split in two unidirectional power flows.
The load is powered via the unidirectional inverter and boost converter - PI2 and the ESS
is charged during the brake stage via the unidirectional PI3 and PI4 (of DC-DC and AC-DC
type) [29]. PI 4 operates equivalent to the series connection of PI5 and PI2 during the
regenerative braking stage.
Thus, the MPC shown in Figure 1 is a generic topology to study MPC energy efficiency,
considering the FC system and the ESS as energy sources, and the AC electrical machine(s)
as output(s).
3. ELECTRICAL MOTORS AND INVERTER SYSTEMS

USED IN THE HYBRID ELECTRICAL VEHICLE
Considering the efficiency, cost, size and weight, lifespan and maintainability as
performance indicators, the electric traction motors used for HEVs are of following type: the
Induction Motor (IM), the Switched Reluctance Motor (SRM), the Permanent Magnet (PM)
Synchronous Motor (SM) motor (PMSM), the Brushless DC motor (BLDC), and the new
motors such as PM Brushless DC (PMBLDC) [1, 13].
The electric motor operates as a generator to recharge the ESS during the braking stage
until to the emergency stopping phase or if the ESS is near to full charge. In a single foot
pedal both regenerative braking and mechanical braking modes appear. In the first part of the
foot pedal the regenerative braking power flow charge the ESS.
However, the braking requirements for all cars will have to be enhanced in the next
decade to meet the new safety stop requirements.
The PM motors have higher efficiency in comparison with the IM, due to the elimination
of magnetizing current and copper loss in the rotor. Note that high performance torque control
could be obtained using PMBLDC motors, even if these require complex control to operate in
both motoring and generating modes [31]. Also note that the IMs and SRMs have the highest
reliability, besides other new motors designed for the HEV applications [32].
The electric motor power varies from 10 kW to 100 kW. So inverter system must be
designed accordingly. The basic inverter topologies are shown in Figure 2.
The first (a) is a traditional PWM inverter that operates in rectified mode during the
regenerative braking. The second (b) is a DC-AC PI (also named inverter) plus a bidirectional
DC-DC PI. The third (c) is a Z-Source Inverter (ZSI), with the unidirectional variant shown in
plot (d).
Figure 2. (Continued)
Figure 2. Inverter topologies [14], (a): Traditional PWM inverter, (b): Inverter plus a bidirectional DC-
DC PI, (c): Bidirectional Z-source inverter, (d): Unidirectional Z-source inverter.
The bidirectional PIs (Figure 1) will operate in boost and buck mode during the motoring
and regenerative braking regime. The voltage on ESS bus usually has a large variation and
must be stabilized to the input level required by the inverter [1, 13]. The inverter system will
be designed to handle this ESS voltage range at full power load. Note that double of the rated
current must be considered at 50% of rated ESS voltage, so the cost of the inverter system
will increase. Consequently, the bidirectional PI2 must to be used in order to minimize the
inverter stress [33]. The ZSI topology shown in Figure 2c is a PI topology that has interesting
features such as buck-boost characteristics and single stage conversion, which can overcome
the above mentioned problems [34]. Note that any desired AC output voltage can be obtained,
even one greater than the input DC voltage, making the ZSI fed adjustable speed drive
systems competitive for the FC automotive applications [35, 36].
A bidirectional ZSI with nine-switches can replace the conventional HEV inverter [37],
using the resonant-phase leg inverters instead of hard-switching inverter and new control
techniques to increase the energy efficiency and reduce the ripple on the DC bus [38, 39].
Note that the inverter energy efficiency is usually from 0.8 to 0.95.
To better use the PIs must be modular and scalable [6], obtaining a higher production
volume and thus a lower cost. Besides this design, the system package, standardization and
interoperability among PIs and system are necessary to be considered in MPC design. Such of
MPCs will allow designers to use a hub-energy-block approach based on specific functions
defined. Thus, the following MPC features should be provided for the design [13]:
› Each MPC should have at least one DC-port and one AC-port;
› All MPC ports must be bidirectional and should be able to operate in a buck or boost
mode;
› MPC device‟s rating is limited by manufacturer datasheets, so paralleling of MPCs is
necessary for high power applications;
› The wiring will be based on the required load power;
› Filters, transformers, and other external circuit designs will be dependent on the
implemented operational power.
The overall cost of power system used in vehicle or Plug-in Charging Stations (PCS) will
be decreased based on this design concept.
4. THE DC INPUT SOURCES OF THE FC VEHICLE

The DC input sources for the FC vehicle shown in Figure 1 are the FC system and ESS.
4.1. Fuel Cell System
In comparison with the ICE, the FC system has the advantage of high energy efficiency
of the FC system, but its slow response requests the mandatory use of ESS. The ESS employs
high-power and high-energy density devices, such as hybrid stack of UCs and batteries.
Hybridization of the FC system with an ESS is an effective technology to overcome the
disadvantages of the FC-alone-powered vehicles.
The PEMFC stack represents one of the most used solutions as main energy source in
automotive applications because of its small size, the ease of construction, a fast start-up and
low operating temperature. Unfortunately, its relatively short life cycle is still an impediment
to FCV commercialization [13, 28]. The performances as energy efficiency [40, 41] and
PEMFC life cycle [42] are dependent to MPC topology and control. For example, the
magnitudes of low frequency harmonics of the PEMFC current ripple affect both performance
indicators mentioned above. Consequently, the PI2 must have a control loop for ripple
mitigation, which will force the ESS to supply the sharp load power profile [23]. The PEMFC
stack (PFC) and ESS stack (PESS) assure the power flows on the DC bus via the PI1 (Pout1) and
PI2 (Pout2), respectively. The power balance on the DC bus is PDC=Pout1+Pout2, where the load
power flow, PDC, is the input DC power required by the inverter system. The EMU control
the power flows of Controlled Voltage Source (CVS) and Controlled Current Source (CCS)
in order to operate the FC system near to MPP and ESS inside of the SOC window [1, 43].
The MPP current (IMPP) can be tracked based on current-mode control using the FC current as
control variable.
4.2. Energy Storage System
The on-board batteries of the ESS allow the energy to be stored during regenerative
braking and provide peak power to the motor controller during vehicle acceleration. To
achieve a higher energy balance on the ESS bus of the FCV, the ESS batteries stack is
charged by a fraction (kFC) of the FC power (PFC), too. Also, taking into account the plug-in
facility of the PFCV, all of these requirements mean the necessity of a high charge acceptance
capability for batteries used.
There is a wide variety of batteries‟ technologies from standard flooded and enhanced
flooded to next Absorbed Glass Mat (AGM) and lithium-ion polymer batteries, which are all
claimed to be “best choices” for automotive and DG applications.
Because there are other chapters dedicated to the ESS, in this section only a brief review
of main batteries‟ technologies is shown (Table 1) [44]:
● Nickel Cadmium (NiCd) - mature technology, but relatively low in energy density.
The NiCd is used where long life, high discharge rate and lower costs are important.
It is better to note that NiCd contains toxic metals, so is environmentally unfriendly.
● Nickel-Metal Hydride (NiMH) - has a higher energy density compared to the NiCd at
the expense of reduced cycle life. NiMH contains no toxic metals.
● Lead Acid - mature technology and most economical for high power applications
where weight is of little concern.
● Lithium Ion (Li-ion) - fastest growing battery technology. Li-ion is used where high-
energy density and lightweight is of main importance. The technology is still fragile
and a protection circuit is required to assure safety.
● Lithium Ion Polymer (Li-ion polymer) - takes the features of the Li-ion technology in
ultra-slim geometry and advantaged packaging. These batteries may also power the
next generation vehicles, but the cost of an electric car using this type of batteries is
prohibitive now.
Typically, the current Li-ion batteries use a carbon-based anode and a metal oxide
cathode. The research on the next Li-ion batteries will continue the development of electrode
and electrolyte materials and chemistries in order to increase the life cycle and energy density
of the battery while reducing size and weight. The most promising technologies appear to
involve silicon, sulfur and air (oxygen) (LiS and Li air technologies in development shown in
Figure 3) and another important development is research into nanotechnologies.
The AGM batteries are constructed differently than the traditional flooded or gel battery.
In AGM batteries (also called starved electrolyte), there is a thin ultra-fine fiberglass mat
sandwiched between the plates that are saturated with battery acid to about 95% of what they
can hold. AGM sealed battery technology was developed and introduced in 1985 for military
aircraft where energy density, safety, and reliability were the main design parameters.
An AGM battery is named as VRSLAB (Valve Regulated Sealed Lead-Acid Battery),
and sometimes is referred to as VRLA (Valve Regulated Lead-Acid). AGM batteries are not
the best choice for all applications - they are rather expensive compared to flooded batteries.
However, their safety and design features make them the battery of choice for many
applications (for example, applications where freezing temperatures could occur in some
situations), including automotive applications.
As it is known, the most current HEVs use NiMH batteries [44]. However, the most
likely alternative battery technologies for use in PHEVs, EREVs and FCEVs are Li-ion and
ACM batteries that have the advantage of higher energy densities and charge-discharge
efficiency (Figure 3).
Table 1 compares the characteristics of the six most commonly used rechargeable battery
systems in terms of energy density, cycle life, static and dynamic requirements and cost.
The efficiency of a battery can be calculated as the amount of power discharged by the
battery divided by the amount of power delivered to the battery.
This takes into account the loss of energy to heat, which warms up the battery. The
charge-discharge efficiencies of various batteries are summarized in Table 1, too. The internal
resistance of a battery represents an overall charge/discharge efficiency indicator.
A low value of it means a high dynamic capacity for example to deliver high currents
without significant drops in voltage, but also to obtain a short time in fast charging regime.
The new semi-solid flow batteries described in [45] and nanotechnologies in
development will overcome the limitation of energy density, providing a possible 10-fold
improvement in this performance over the present liquid flow-batteries, and a lower-cost
manufacturing than the conventional lithium-ion batteries [44] (Figure 3).
The SOC of the ESS should be kept inversely proportional to the vehicle‟s speed [1, 13]:
● At low speeds it is probable that vehicle will accelerate in the near time, so the ESS
should have a high SOC to power that acceleration.
● At high speeds, it is probable that vehicle will be brake soon, so the ESS should have
a low SOC to have capacity to accept as much of the regenerative braking energy as
possible.
Consequently, the durability and lifetime of the battery under SOC deep cycling are key
measures of the battery performance. Automotive batteries require a cycle life of at least
5,000 cycles of charge – discharge and a calendar lifetime of over ten years [44].
To overcome this situation, the hybridization of ESS by using UCs stack seems to be now
the best way to obtain high performance for ESS.
Several ESS topologies (passive, semi active and active battery–UC hybrid) have been
proposed and tested for usage in PFCV [13, 26]. To boost the UC stack voltage to ESS
voltage a bidirectional PI is required based on the UC semi-active ESS topology. The
operating window centered on SOC middle (around 0.7 SOC) assures a high efficiency and
flexibility of the ESS operation in all regimes [13] in comparison with the passive ESS
topology, which is still the most used because it is cheap and easy to implement [26]. The
active or semi-active ESS topologies use a large SOC window, reducing the incremental cost
[16, 46] and increasing the use of the overall energy [47].
Figure 3. Battery technology roadmap.

Table 1. Characteristics of commonly used rechargeable batteries
Li-ion AGM
NiCd NiMH Lead Acid Li-ion
polymer VRSLAB
Gravimetric Energy Density
45-80 60-120 30-50 100-200 150-250 >100
(Wh/kg)
200 to
Internal Resistance [mΩ] 100 to 200 200 to 300 <100 150 to 250 <50
300
(includes peripheral circuits) 12V pack 6V pack 12V pack 7.2V pack 12V pack
7.2V pack
Cycle Life [cycles] (to 80%
1500 500 300 1000 500 3000
of initial capacity)
Fast Charge Time 1h typical 2-4h 8-16h 2-4h 2-4h <1h
Very
Overcharge Tolerance moderate low high very low low
high
Charge/Discharge Efficiency
70% - 90% 66% 50% - 92% 80% - 90% > 90% 98%
[%]
Self-discharge / Month
20% 30% 5% 10%5 ~10% 1-3%
(room temperature)
Cell Voltage(nominal) 1.25V 1.25V 2V 3.6V 3.6V 2V
Load Current - peak / rated
2000 500 500 >200 >200 >2000
value [%]
Operating Temperature -40 to -20 to -20 to -20 to 0 to -40 to
(discharge only) 60°C 60°C 60°C 60°C 60°C 70°C
60 to 90 3 to 6
Maintenance Requirement 60 days not req. not req. not req.
days months
Typical Battery Cost (US$, $100 $100 $100
$100 (7.2V) $60 (7.2V) $25 (12V)
estimative only) (7.2V) (7.2V) (12V)
$0.03-
Cost / Life Cycle ($/cycles) $0.08 $0.12 $0.1 $0.1 $0.29
0.04
Commercial use since 1950 1990 1970 1991 1999 1985
However, until the PFCV will be plugged to PCS, EMU must decide how best will use
the FC stack to force the ESS to operate in (1) Charge-Sustaining (CS), (2) Charge-Depleting
(CD), or (3) Charge-Increasing (CI) mode. Thus, the battery energy must be sufficient to
operate the vehicle on anticipated use-restricted zone or distance to the next PCS based on the
stored electrical energy during initial CD operation. The ECU will switch to CS operation if
the next PCS is far away or to CI mode if the next PCS has communicated its energy demand
availability. The best EMU energy strategy for different load profiles [48] based on the
communication features implemented [49] will be presented briefly in next section.
5. ENERGY MANAGEMENT UNIT

From the power flows balance in hybrid MPCs point of view, the energy is balanced by
the DC bus voltage regulation, usually using an ESS [15]. This chapter is focuses on this kind
of MPC topology powered by a FC system and equipped with an ESS that usually has two
energy storage devices: UC and battery.
The UCs stack, as a high dynamic and high power density device, operates for supplying
power to regulate the DC bus voltage on load transitions. The batteries stack operates as a
high energy density device to satisfy the power flows balance on the DC bus and to keep the
UC stack charged, which could be connected to the ESS bus via a bidirectional PI to obtain
high performance in operating the ESS in CS mode [50].
Because the FC system has the slow response, the EMU will control the FC system to
operate at the average load power to keep the ESS stack in CS mode [1, 13]. Besides this
power control loop, there are other two voltage controls‟ loops are shown in Figure 1: the DC
bus voltage regulated by the ESS and the ESS voltage regulated by the FC system. If the ESS
active topology will be used, then the UC voltage will be regulated by the battery stack [51].
The FC system and hybrid ESS can assure the best performance for FCVs propulsion [52,
53]. The FC system and ESS can be downsized to fulfill the average and transient power
demand based on the load following strategy [52]. The dynamic performances of a FCV
operating with and without the assistance of the ESS are investigated in [53, 54].
As it was mentioned above, the MPC topology and EMU features implemented set the
(1) cost, (2) efficiency, and (3) reliability performances of the PFCV. The trade-off between
the complexity of the EMU and the system performance may provide different results for
different types of PFCVs that operate in the specified regimes under a given driving cycle
[13, 54]:
● CS Mode is the operating mode in which the ESS SOC may fluctuate, but on average
this is maintained at a certain level while driving.
● CD Mode - operating mode in which the average of ESS SOC decreases while
driving.
● CI Mode - operating mode in which the average of ESS SOC increases while driving.
The CI mode is a specific operating mode only for FCV in which the ESS is charged
(PESS< 0) while driving. Note that PESS represents the average power exchanged by the ESS
during a drive cycle:
● If PESS> 0, then this power represents the net power discharged by the ESS, which
operates in CD mode. The PFCV must be connected to the PCS as soon as the SOC
level reaches the low value imposed.
● If PESS< 0, then this power represents the net power from the FC system and
regenerative load that charged the ESS, which operates in CI mode. The PFCV must
be connected to the PCS as soon as the SOC level reaches the high value imposed.
In this way the PFCV could operate as grid stabilizers if EMU will match the power
demand to supply availability of the DG system via PCSs. It is obvious that connection to the
DG system allows more opportunities for all plug-in vehicles, including their autonomy. The
plug-in vehicles autonomy means that these can not only charge their ESS, but they can also
discharge the available ESS energy into the DG grid. In these operating regimes the plug-in
vehicles can support the DG grid and this concept is named vehicle-to-grid (V2G). However,
plug-in vehicles should not be confused with the V2G concept, in which the electric drive
vehicles can provide value to electric utilities through peak saving and auxiliary services.
V2G defines system architecture in which plug-in vehicles, such as EVs, PHEVs and PFCVs,
communicate with the DG system to sell demand response services by either delivering
electricity into the grid or by controlling their charging rate. Such operating regimes for
PFCV will be considered in the energy efficiency analysis performed in this chapter.
The EMU controls the PIs to operate the ESS in CD, CI or CS mode in order to increase
the energy efficiency of the whole MPC. The main design factors for EMU are the SOC
regulation, ESS life, power dynamic of load, power capacity of the FC system near to its
MPP, and available or required power to or from the grid.
In next sections it will be performed the analysis of the energy efficiency of the whole
MPC based on the SOC window width, the energy efficiency of each PI, the kL regenerative
braking factor, the kFC fraction of FC power used to charge ESS, and the road profile. To
implement the PCSs into DG system it is necessary to upgrade the level of [55]:
● Wide Area Monitoring Systems.

● Two way distribution grids.
● Two-way high speed communications.
● Renewable Energy Sources integration.
Furthermore, plug-in vehicles can be used as backup power to a home via the
bidirectional charger, which usually is a bidirectional two-stage topology based on high
frequency transformer for isolation and/or obtaining of a HV conversion ratio.
The EMU receives the measured state variables from the driver (signals from the
accelerator pedal and brake pedal) and feedback from the drive train (for example, the speed
signal) and other components (Figure 1) to set the levels for power flows based on the power
balance.
6. MULTI-PORT CONVERTER TOPOLOGIES

In this section the MPC topologies usually used for the PFCV that exchange power with
the PCS grid connected or autonomous renewable PCS (Figure 4) will be analyzed. The key
issue related to the grid integration will be the matching of the PFCV power demand to the
grid supply availability. It is obvious that a grid connection opens more opportunities for all
plug-in vehicles. The PFCV with a bidirectional charger can not only charge the ESS from the
EP system, but it can also discharge the available ESS energy into the grid. The PFCV based
on the EMU strategies supports the grid using two-way communication between the EP
system and the vehicle, enables a better management of the energy sources, and allows the
vehicle owners to earn money by selling power back to the grid. In this way plug-in vehicles
could become more like “cash cows” that produce income from existing equipment and less
like vehicles that merely consume energy [15]. The EREVs having as Range Extenders a FC
or/and ICE (Figure 5) are in production phase, but the research in reducing the losses is under
work. The data from Figures 4 and 5 give the values of the losses under an urban (highway)
cycle [56].
In last the decade a lot of work was devoted to reducing the MPC efficiency [56], besides
the work to obtain low cost and compact structure for the MPC topologies proposed in the
literature [57, 58].
Figure 4. PFCV architecture.
Figure 5. Electric Vehicles with a Range Extenders (FC and ICE).
The MPC are designed for various applications, such as achieving three bus voltages of
14 V/42 V/HV bus (voltage on HV bus been of around 500 V) in EV and HEV [1, 13],
interfacing the PV panel and a battery to a regulated 28-V bus in a satellite platform power
systems [59], PV energy harvesting [60], FC energy harvesting [61], and HPS [62].
From the topology point of view, the MPC are based on buck-boost, half and full bridges
and interleaved topologies [63] (see Figure 6). The basic bidirectional switching cells used in
the MPC are shown in Figure 6.
The MPC must meet the isolation requirement and the bidirectional capabilities needed in
vehicle applications. So, from the isolation requirement point of view the MPC topologies can
be classified in three main classes:
› Non-isolated (Figure 7);

› Partially isolated (Figure 8);
› Fully isolated (Figure 9);
The main research directions for the MPC are:

› Reducing the number of active switches and passive components;

› Finding a simple transformer windings connection that assures bidirectional power
transfer between each pair of ports;
› Simple control of the bidirectional power flows;
› Designing of compact interleaved topologies for high-power applications without the
use of bulky inductors or transformers to mitigate the current ripple;
› Reducing switching loss and stresses based on zero-voltage and zero-current
switching command techniques;
› Reducing conduction loss due to circulation current;
› Obtaining a higher power density based on a symmetrical topology that is favorable
to achieving magnetic integration and modularizing;
› Reducing the DC offset of the transformers caused by unbalance on the voltages or
the switching times.
7. PFCV WITH SERIES AND PARALLEL MPC TOPOLOGIES

The purpose of this chapter is to demonstrate that the EMU strategies could improve the
energy efficiency of the whole MPC for the both series and parallel MPC topologies, named
as MPCs and MPCp, respectively.
(a) buck-boost (b) half-bridge (c) full-bridge
(d) boost half-bridge (e) boost full-bridge (f) three-phases

interleaved buck-boost
(g) three-phases full-bridge (h) three-phases boost bridge
Figure 6. Basic bidirectional switching cells used in the MPC.

The FC system powers the DC bus (HV type) and the ESS bus (LV type), which is used
in the parallel and series MPC topologies, respectively, via a DC-DC power converter that
will boost the FC voltage to the HV level needed on the DC bus [1, 89].
a b c
d e f
(a) Adapted from [64]. (b) Adapted from [65]. (c) Adapted from [66]. (d) Adapted from [67]. (e)
Adapted from [68]. (f) Adapted from [69].
Figure 7. Non-isolated MPC topologies.
a b c
d e f
Adapted from [74]. (f) Adapted from [75].
Figure 8. Partially isolated MPC topologies.
a b c
d e
j
l
Adapted from [80]. (f) Adapted from [81]. (g) Adapted from [82]. (h) Adapted from [83]. (i)
Adapted from [84]. (j) Adapted from [85]. (k) Adapted from [86]. (l) Adapted from [87, 88].
Figure 9. Fully isolated MPC topologies.
The FC system usually operates in a range between the MAE and MPP for both high
efficiency and reliability. So, a PI is necessary and this must be appropriately controlled [20,
90]. The ESS exchanges power with the HV DC bus via a DC-DC bidirectional power
converter, named PI2 in the both MPC topologies. Thus, the PI2 is necessary in the parallel
and series MPC topology to assure the power flow balance on the DC bus (if it is properly
controlled) and to boost the LV of the EES bus to the HV DC bus, respectively. The inverter
system for the traction motor is usually of bidirectional type and the used topologies are of
classical or of Z-source inverter type in the both MPC topologies [1, 34]. The competitive
types of electric traction motors include the induction motor (IM), the switched reluctance
motor (SRM), the permanent magnet (PM) Synchronous Motor (SM) motor (PMSM), and the
Brushless DC motor (BLDC) [1, 13].
These motors have many advantages (such as efficiency, cost, size and weight, life span
and maintainability) if compared with other motor types.
The EMU controls the power flow in different operating modes (CS, CD, and CI) of the
ESS, monitoring its SOC. The energy efficiency depends by the charge/discharge ratio of the
ESS power (SOC level), energy efficiency coefficient of each PIs, regenerative braking
factor, and the drive cycle. For example, the EMU must to select the CD and CI mode when
the PFCV runs up and down a hill, but all CD, CI or CS modes could be used in efficient
manner until the PFCV will be plugged into a PCS. The most strategies propose to use the
stored electrical energy during the initial CD operation up to exhausting it, and then will
switch to CS mode and later to CI mode if a PCS on route has announced their availability to
sell power back to the grid. The best energy strategy will be shown in next sections for main
MPC topologies.
The on-board ESS allows energy storage from FC system and regenerative load and
supply the motor during the vehicle acceleration, which means a high charge/discharge
capability for the used batteries.
A wide variety of battery technologies have claimed to be “the best choice” for
automotive applications, starting from standard flooded and enhanced flooded to next AGM
and Li-ion polymer batteries [91-93]. Hybridization of the ESS for pulsed load is made using
UCs and batteries in different ESS topologies (passive, semi active and active) [94].
This section summarized few EMU strategies to assure the grid stabilization and cover
the distance over which the vehicle will likely operate before a connection to the near PCS.
The next sections will evaluate the energy efficiency of the whole MPC for different ESS
operating modes. The obtained results are a generalization of those shown in [50], where only
the CS mode is considered.
8. PARALLEL MPC TOPOLOGY

The parallel MPC (MPCp) topology is shown in Figure 10 without a plug-in feature. The
whole MPC has as inputs the FC system and the ESS bus, and as output the AC electrical
machine. The average power of the FC system and ESS on a load cycle is PFC and PESS,
respectively. The PFC is always positive, but the sign of the PESS depends on the SOC
regulation performed by the EMU.
The MPCp topology shown in Figure 10 contains the following PIs:
PI1 - unidirectional DC-DC power converter, having the energy efficiency coefficient
denoted by 1.
PI2 - bidirectional DC-DC power converter, having in direct and reverse mode the
2and2R coefficients, respectively.
PI5 - bidirectional inverter system, having in direct and reverse mode the 5 and 5R
coefficients, respectively.
The inverter system powers the AC electrical machine with PL from the DC bus via PI5.
During regenerative braking process the reverse power flow, kLPL, charges the ESS via the
PI5 and PI2 connected in series. The net power delivery to the AC electrical machine is
(1-kL)PL, where kL is the regenerative braking factor.
8.1. Equivalent Parallel MPC Topology
In terms of the power flow, an equivalent MPC topology of the MPCp topology is shown
in Figure 11.
The power flow will circulate in the motor regime from both FC and ES system to the
AC motor via the series connection of the PI1 and PI5, and the PI2 and PI5, respectively.
Figure 10. Parallel MPC topology [14].
Figure 11. Flowchart of power flow of the parallel MPC topology [15].
The FC system powers the load with PFC(L), and the rest of the needed power on the DC
bus is obtained from the ESS that is discharged by the average power, PESS(D), via the PI2. The
power flow balance on DC bus in the CD mode is:
2PESS(D)+1PFC(L)=PL/5 (1)
where the used notations are mentioned above.

The reverse power flow during regenerative braking process, kLPL, will circulate from the
AC electrical machine to the ESS via the PI5 and PI2 connected in series (that is formally
represented by PI4 in Figure 11). The FC system also powers with PFC(ESS) the ESS via PI1
and PI2 connected in series (that is formally represented by PI3 in Figure 11).
The power flow balance on the ESS bus related to the CI mode of the ESS (which is
charged in average with PESS(C)) is:
PESS(C)=3PFC(ESS)+4kLPL=12RPFC(ESS)+2R5RkLPL (2)
where the used notations are mentioned above.

The equivalent MPCp topology proposed in Figure 11 contains additional PIs:
PI3 - unidirectional DC-DC power converter, having the energy efficiency coefficient
denoted by 3=12R.
PI4 - unidirectional AC-DC power converter, having 4=5R2R. The plug-in feature of
PI 4 is not shown in Figure 11.
This equivalent representation of the MPCp is more clear and convenient for an analysis
of energy efficiency performed in this section.
Also, it can be noted that this topology is obviously better than the MPCp topology
because the power is less processed and easier to be controlled. The efficiency analysis will
be performed in the next section considering as parameters the kESS and kCD ratios, where
kESS(p)=PESS/PL and kCD=PESS(C)/PESS(D). As it was noted above, PESS(C) and PESS(C) represent the
average power exchanged by the ESS on CI and CD mode, respectively.
8.2. Energy Efficiency of Parallel MPC Topology
Note that the energy efficiency coefficients are defined as average value because their
values vary depending on the operating conditions.
The energy efficiency for the MPCp topology is given by the relationship:
p=MPCp=(1–kL)PL/(PFC+PESS) (3)
It is obvious that:
PFC=PFC(L)+PFC(ESS) (4)
PESS=PESS(D)– PESS(C) (5)
Relationships (6) result by simple mathematical manipulations of (5) using the definitions
for the kESS(p) and kCD ratios:
PESS(C)=kESS(p) kCD PL/(1 – kCD) (6a)
PESS(D)=kESS(p) PL/ (1 – kCD) (6b)
If relationships (6) will be used on (1) and (2), then relationship (4) becomes:
(7)
where
(8)
It is obvious that:
(9)
where the symbol used for comparison can be <, =, or >.

The case kCD=1 (kESS(p)=0) will be briefly discussed in section 11 of this chapter because
it has been widely analyzed in [50]. Considering kCD1, the energy efficiency for the MPCp
topology is finally given by relation:
(10)
9. SERIES MPC TOPOLOGY

The series MPC (MPCs) topology is shown in Figure 12. The MPCs topology contains
the same PIs, marked as it was mentioned in the MPCp topology. If the ESS bus is of HV
type (not of LV type), then PI2 will not be used, and their energy efficiency coefficients will
be considered unitary in all relations written in this section.
Here, for the simplicity of the analytical comparison performed, PI2 is considered having
the same values of the energy efficiency coefficients (as PI2 from the MPCp topology), even
if the design parameters are different. The same notations as above, PFC and PESS, will be used
for the average power on a load cycle of the FC system and ESS, respectively.
Also, the net power delivery to the AC electrical machine is (1-kL)PL.
Figure 12. Series MPC topology [15].

9.1. Equivalent Series MPC Topology
In terms of the power flow, an equivalent MPC topology to the MPCs topology is shown
in Figure 13. The inverter system powers the AC electrical machine with the PL power from
the the ESS bus via the PI2 (2) and PI5 (5) connected in series. Also, during regenerative
braking process the reverse power flow, kLPL, charges the ESS via the PI5 (5R) and PI2 (2R)
connected in series.
So, PI3 will represent the series connection of the PI2 and PI5, and PI4 will represent the
series connection of PI2 and PI5 that operate in direct and reverse mode, and have the 3=2
5 and 4=2R5R coefficient, respectively. The ESS bus is powered by the ESS and the FC,
directly and via PI1, respectively. Using the ratio of the ESS power exchanged on the ESS
bus, kESS(s)=PESS/PL, the power flow balance on the ESS bus is:
1PFC+PESS+2R5RkLPL=PL/25 (11)
The same relationship is obtained if the power flow balance on the ESS bus is written in
the CI and CD mode of the ESS, respectively:
PESS(C)=1PFC(ESS)+2R5RkLPL (12a)
PESS(D)+1PFC(L)=PL/25 (12b)
where the notations for the split power flows are the same.
The split power flows were not shown in the diagram for the clarity of the equivalent
MPCs topology. Considering (4) and (5), the same relationship as in (11) will be obtained by
subtracting (12a) from (12b).
9.2. Energy Efficiency of Series MPC Topology
In this case, the energy efficiency for the MPCs topology is given by (13):
s=MPCs=(1– kL)PL/(PFC+PESS) (13)
Figure 13. Flowchart of power flow of the series MPC topology [15].
Using the kESS(s) ratio, by simple mathematical manipulations of (11), results:
(14)
The case kESS(s) = 0 has been widely analyzed in [50], therefore will be briefly discussed
in section 11 of this chapter.
10. ENERGY EFFICIENCY OF MPCP VS. MPCS

If both MPC topologies (MPCp and MPCs) exchange the same ESS power (positive or
negative), i.e. kESS(p)=kESS(s)=kESS, it is also important to know which ones operates more
efficiently:
(15)
10.1. The Results of Analytical Calculations
Relationship (15) is equivalent with (16) by simple mathematical manipulations of its,

using (10) and (14):
(16)
Using (8) will result (17):
(17)
So, (16) will become equivalent with (18):
(18)
For an easy interpretation of the kESS values on the CI and CD mode, the following
notation is made related to the CS mode:
(19)
Finally, relationship (15) will become equivalent with (20):

(20)
Analyzing (20) and (9), some final conclusions related to the MPC energy efficiency on
both CD and CI operating modes are shown in (21a) and (21b), respectively, considering
kESS(p)=kESS(s)=kESS:
(21a)
(21b)
Relationships (21) can be rewritten in a concise equivalent form as:
(22)
where the  symbol is used for the logical OR operator.

Of course, if kESS(p)=kESS(s)=CCD, then both MPC topologies considered are equivalent in
terms of energy efficiency, so CCD is a border value. If kESS(p) kESS(s), then p/s ratio of
energy efficiency coefficients is computed in (23):
(23)
Obviously, if considering kESS(p)=kESS(s)=kESS in (23), the final conclusion (22) can be

obtained, too. The analytical results are validated by the simulation performed. Some relevant
results will be shown in next section.
10.2. Simulation Results
If it is not specifically mentioned, the energy efficiency coefficients of PIs are set, in all
simulations performed, to 1=0.9, 2=0.9, 2R=0.9, 5=0.9, and 5R=0.9. The values used for
the charge-discharge parameter, kCD, are 1/1.25 and 1.25 for the CD and CI operating mode,
respectively.
Figure 14. p/s(kESS, kL) surface [15].
The energy efficiency ratio for the MPCs topologies is shown in Figure 14 based on (23),
considering kESS(p)=kESS(s)=kESS. Using kCD=1/1.25 and kCD=1.25 in regions where kESS>0 and
kESS<0, respectively.
It can be noted that: (1) the unitary plane delimits the border lines where the MPCs and
MPCp topologies have the same energy efficiency; (2) if the ECU manages a low level of the
ESS power exchanged on a load cycle, then the MPCp is more energy efficiency than the
MPCs topology; (3) if the ESS accepts a large SOC window, then the ECU can manage a
higher value of the ESS power exchanged on a load cycle and thus the MPCs is more energy
efficiency than the MPCp topology; (4) the border lines (values of the CCD coefficient) are
less dependent by the regenerative braking factor [15].
The dependence of the CCD coefficient to kL parameter can be obtained based on (19)
computed for different values of the kCD parameter. The regions where p>s can be extended
for values of the kCD parameter that are far from unity value. A kCD value greater or less than
unity means that the ESS will be charged (PESS(C)>PESS(D)) or discharged (PESS(C)<PESS(D)),
respectively, but at higher values for the split powers, PESS(C) and PESS(D). For a PFCV under a
load cycle it is obvious that these high values for average power can produce unacceptable
peak values for the ESS. Consequently, in this case the ESS hybridization must be more
carefully designed. Note that each of the MPC topologies (MPCp or MPCs) can operate
efficiently in both operating modes (CD or CI) (Figure 14).
Also, the CCD coefficient is dependent on energy efficiency coefficients. It was shown
that CCD coefficient is less sensitive to energy the efficiency coefficients of PI5 (5 and 5R)
[15]. This aspect is explained trough topological considerations: PI5 has the same position
(between the DC bus and load bus) into both MPC topologies. As it was explained above, the
bidirectional PI5 can exchange power between the load port and the DC bus.
This dependence related to an energy efficiency coefficient () can be obtained by
computing the CCD/ coefficient sensitivity using (19). Also, relationship (19) clearly
shown that the CCD coefficient is independent to the PI1 energy efficiency coefficient. This
is explained by the same position of PI1 on both MPC topologies and the fact that the FC
system powers with the same power (PFC) the load and ESS.
Figure 15. kESSsvs kESSpfor 1=2=2R=5=5R=0.9 and different kL values [15].
It is obvious that both p and s coefficients increase if each of energy efficiency

coefficients increases. The exact variation of the p/s ratio is given by (23). A mathematical
analysis might reveal all dependencies [15], but this is not the main goal of this chapter. Only
a brief validation of the analytical results based on topological analysis is given below.
The load port is powered in both topologies with the same power flow from FC and ESS
via PI1, PI2, and PI5, but only the position of the PI2 is different on the MPC topologies.
If the PI2 is an ideal converter (2=2R=1), then MPCp and MPCs topologies are
identical, so p/s=1 must be obtained. Therefore, if 2=2R=1 is considered in (8), then
2m=1, and thus from (23) p/s=1 is obtained (q.e.d.).
It is obvious from (22) that the energy efficiencies of the MPCp and MPCs topologies
have different values, p and s, respectively, for a value of common kESS parameter, which is
different by the CCD value. On the other hand, if different kESS values are used in the
operation of the MPCp and MPCs topologies (kESSp and kESSs, respectively), then both
topologies could have the same efficiency on the conditions given by the relationship (24):
(24)
Relationship (24) is shown in Figure 15 for different kL values and1=2=2R=5=

5R=0.9. The same efficiency could be obtained for both MPC topologies if the ESS operates
in the CD or CI mode at different kESS values. If the both MPC topologies operate in the CS
mode, then the energy efficiency analysis will be performed in the next section based on (10),
(14) and (22), considering kESS(p)=kESS(s)= 0.
11. ENERGY EFFICIENCY OF MPCP

AND MPCS TOPOLOGIES IN CS MODE
The average value of charge exchanged by the ESS under a drive (load) cycle is zero in a
CS mode, so:
(25)
The energy efficient coefficients given by (10) and (14) will become as below:
(10a)
(14a)
where kL is the regenerative braking factor.

The kL-ESS0 coefficient is defined as a measure of load dynamic that requests an
appropriate ESS dynamic:
(26)
If both MPC topologies (MPCp and MPCs) operate in a CS mode (i.e. kESS(p)=kESS(s)=0),
it is also important to know which ones operates more efficiently (e.g. Figure 16).
Relationship (26) is equivalent with (27) by simple mathematical manipulations of its,
considering (10a) and (14a):
(27)
where the C0 coefficient defined in (27) is a positive value, which depends on the energy
efficiency coefficients of the PI2 and PI5, and the regenerative braking factor, kL.
Figure 16. p/s(kL, kL0) surface for 1=2=2R=0.93 and 5=5R=1 [15].
The energy coefficients of PI5 can be set to 1 (5=5R=1) for an easy comparison of
these particular results obtained in the CS mode with same results that have been presented in
[50] (e.g. Figure 16).
An example of a load cycle profile applied on the DC bus is shown in Figure 17, too [94].
The average (AV) and effective (RMS) value of the load power are dependent on the duty
cycle, DL, the ratio of positive and negative amplitudes, CL=PL(max)/PL(min), and the
regenerative braking factor, kL, as in (28):
(28)
The results presented in Figure 17 show that the MPCp topology may or may not be more
efficient than the MPCs topology, and this depends on the values set for the DL and kL-ESS0
parameters. Further details about MPCp and MPCs topologies operating on CS mode can be
found in [50].
CONCLUSION
In this chapter, besides a brief overview over the EV and HEV issues, a systemic analysis
of the MPC energy efficiency was presented for the series and parallel MPC topologies used
in PFCVs. The analysis has been focused on the FC/ESS HPS that supply the electrical
motors via a MPC and via inverter systems. The power flow flowchart of the parallel MPC
topology shown in [14] is similar to the MPC topology proposed in [15], to overcome the
limitations of both series and parallel hybrid MPC topologies for uncertain load profiles. The
main properties of the MPC topologies were underlined in this chapter.
The efficiency of MPC's topologies is evaluated and compared in CS, CD and CI
operating modes using different control's factors.
Figure 17. s(DL) - line width set to 2, and p(DL) - line width set to 1, for 1=2=2R=0.93, 5=5R=1,
CL=9 and different kL0 [15].
Therefore, the ESS exchanges the power under the EMU control and the PFCV must be
connected to the PCS when it is the case. The EMU sets these moments and notifies the driver
in due time. It was shown that each MPC topology can operate efficiently under certain
conditions, which have been specified.
The impact of the PFCVs on the electricity sector depends on the (1) magnitude of the
market penetration, (2) timing of ESS charging/discharging, (3) energy management, (4)
structure of the power sector, and (5) availability of the FC systems for automotive vehicle.
PFCVs can contribute by an efficient use of the renewable intermittent potential and can
potentially solve imbalances between electricity supply and demand.
Besides these conclusions, a number of questions also arise for further exploration: (1)
What type of energy management strategy is more suitable for the PFCV EMU in relation
with upcoming smart grids? (2) What are the optimum ESS requirements for various PFCV
configurations? (3) Can be quantified the impact of plug-in feature on battery life? (4) Can be
quantified the PFCV impact on renewable intermittent potential? (5) Can be evaluated the
PFCV consumption characteristics under a given EMU strategy? The CI mode will be used in
relation to the upcoming smart grids to sell the power back to the grid in order to stabilize the
grid. Overall, the potential to increase the MPC efficiency is shown for all CS, CD and CI
operating modes of the ESS. The use of multi-operating modes makes the PFCV attractive,
but also makes the estimation of its in-use fuel economy a challenging design action. If it is
not necessary to operate the ESS in the CD and CI modes, then the CS mode will be used, but
designing of appropriate EMU strategy remains a challenging action in order to approximate
real-world operation.
ACKNOWLEDGMENTS
Some figures, tables and text are reproduced from [14, 15] here with kind permission
from Elsevier Limited, UK [November 18, 2013].
REFERENCES
[1] Mehrdad, E., Yimin, G., Ali, E., Modern Electric: Hybrid Electric, and Fuel Cell
Vehicles. 2nd Ed., CRC Press, New York, 2010.
[2] Bizon, N., Advances in Energy Research: Distributed Generation Systems Integrating
Renewable Energy Resources. Nova Science Publishers Inc., US, 2012.
[3] Bizon, N., Tabatabaei, N. M., Advances in Energy Research: Energy and Power
Engineering. Nova Science Publishers Inc., US, 2013.
[4] Zhan, Y., Guo, Y., Zhu, J., Wang, H., Intelligent Uninterruptible Power Supply System
with Back-up Fuel Cell/Battery Hybrid Power Source. Journal of Power Sources, vol.
179, no. 2, pp. 745-753, 2008.
[5] Metaxiotis, K., Intelligent Information Systems and Knowledge Management for
Energy: Applications for Decision Support, Usage and Environmental Protection. IGI
Global, US, Chapter 2 (Bizon, N., Intelligent Control of the Energy Generation
System), pp. 40-96, 2009.
[6] Jiang, W., Fahimi, B., Multiport Power Electronic Interface: Concept, Modeling, and
Design. IEEE Trans. on Power Electron, vol. 26, no. 7, pp. 1890-1900, 2011.
[7] McDonough, M., Shamsi, P., Fahimi, B., Application of Multi-Port Power Electronic
Interface for Contactless Transfer of Energy in Automotive Applications. Vehicle
Power Propulsion Conference, pp. 936-940, 2009.
[8] Geidl, M., Andersson, G., Optimal Power Flow of Multiple Energy Carriers. IEEE
Trans. on Power Electronics, vol. 22, no. 1, pp. 145-155, 2009.
[9] Geidl, M., Koeppel, G., FavrePerrod, P., Kloeckl, B., Andersson, G., Frohlich, K.,
Energy Hubs for the Future. IEEE Power Energy Mag., vol. 5, no. 1, pp. 24-30, 2007.
[10] Krause, T., Andersson, G., Frohlich, K., Vaccaro, A., Multiple-Energy Carriers:
Modeling of Production, Delivery, and Consumption. Proceedings of the IEEE, vol. 99,
no. 1, pp. 15-27, 2011.
[11] Del Real, A. J., Arce, A., Bordons, C., Optimization Strategy for Element Sizing in
Hybrid Power Systems. Journal of Power Sources, vol. 193, no. 1, pp. 315-321, 2009.
[12] Loo, K. H., Zhu, G. R., Lai, Y. M., Tse, C. K., Development of a Maximum-Power-
Point Tracking Algorithm for Direct Methanol Fuel Cell and its Realization in a Fuel
Cell/Supercapacitor Hybrid Energy System. 8th International Conferenceon Power
Electronics, pp. 1753-1760, 2011.
[13] Chan, C. C., Chau, K. T., Modern Electric Vehicle Technology. Oxford University
Press, 2001.
[14] Bizon, N., Energy Efficiency of Multiport Power Converters Used in Plug-In/V2G Fuel
Cell Vehicles. Applied Energy, vol. 96, pp. 431-443, 2012.
[15] Bizon, N., Energy Efficiency for the Multiport Power Converters Architectures of
Series and Parallel Hybrid Power Source Type Used in Plug-In/V2G fuel Cell Vehicles.
Applied Energy, vol. 102, pp. 726-734, 2013.
[16] Erdinc, O., Uzunoglu, M., Recent Trends in PEM Fuel Cell-Powered Hybrid Systems:
Investigation of Application Areas, Design Architectures and Energy Management
Approaches. Renewable and Sustainable Energy Reviews, vol. 14, pp. 2874-2884,
2010.
[17] Gao, D., Jin, Z., Lu, Q., Energy Management Strategy Based on Fuzzy Logic for a Fuel
Cell Hybrid Bus. Journal Power Sources, vol. 185, no. 1, pp. 311-317, 2008.
[18] Erdinc, O., Vural, B., Uzunoglu, M., A Wavelet-Fuzzy Logic Based Energy Manage-
ment Strategy for a Fuel Cell/Battery/Ultracapacitor Hybrid Vehicular Power System.
Journal of Power Sources, vol. 194, no. 1, pp. 369-380, 2009.
[19] Rodatz, P., Paganelli, G., Sciarretta, A., Guzzella, L., Optimal Power Management of
an Experimental Fuel Cell/Supercapacitor-Powered Hybrid Vehicle. Control Eng.
Pract., vol. 13, no. 1, pp. 41-53, 2005.
[20] Feroldi, D., Serra, M., Riera, J., Energy Management Strategies Based on Efficiency
Map for Fuel Cell Hybrid Vehicles. Journal of Power Sources, vol. 190, no. 2, pp. 387-
401, 2009.
[21] Erdinc, O., Vural, B., Uzunoglu, M., Ates, Y., Modeling and Analysis of an FC/UC
Hybrid Vehicular Power System Using a Wavelet-Fuzzy Logic Based Load Sharing
and Control Algorithm. International Journal of Hydrogen Energy, vol. 34, no. 12, pp.
5223-5233, 2009.
[22] Liu, G., Zhanga, J., Sun, Y., High Frequency Decoupling Strategy for the PEM Fuel
Cell Hybrid System. International Journal of Hydrogen Energy, vol. 33, no. 21, pp.
6253-6261, 2008.
[23] Thounthong, P., Rael, S., Davat, B., Energy Management of Fuel Cell/Battery/ Super-
capacitor Hybrid Power Source for Vehicle Applications. Journal of Power Sources,
vol. 193, no. 1, pp. 376-385, 2009.
[24] Wang, Y., Chen, K. S., Mishler, J., Cho, S. C., Adroher, X. C., A Review of Polymer
Electrolyte Membrane Fuel Cells: Technology, Applications, and Needs on
Fundamental Research. Appl. Energy, vol. 88, no. 4, pp. 981-1007, 2011.
[25] Doucette, R. T., McCulloch, M. D., Modeling the Prospects of Plug-In Hybrid Electric
Vehicles to Reduce CO2 Emissions. Applied Energy, vol. 88, no. 7, pp. 2315-2323,
2011.
[26] Hacker, F., Harthan, R., Matthes, F., Zimmer, W., Environmental Impacts and Impact
on the Electricity Market of a Large Scale Introduction of Electric Cars in Europe -
Critical Review of Literature. ETC/ACC Technical Paper, vol. 4, pp. 56-90, 2009.
[27] www.pluginamerica.org/vehicles.
[28] http://evworld.com.
[29] Bizon, N., A New Topology of Fuel Cell Hybrid Power Source for Efficient Operation
and High Reliability. Journal of Power Sources, vol. 196, no. 6, pp. 3260-3270, 2011.
[30] Qian, Z., Al-Atrash, H., Modeling and Control of Three-Port DC/DC Converter
Interface for Satellite Applications. IEEE Trans. on Power Electronics, vol. 25, no. 3,
pp. 637-649, 2010.
[31] Singh, B., Singh, S., State of the Art on Permanent Magnet Brushless DC Motor Drives.
Journal of Power Electronics, vol. 9, no. 1, pp. 1-17, 2009.
[32] Zeraoulia, M., Benbouzid, M. E. H., Diallo, D., Electric Motor Drive Selection Issues
for HEV Propulsion Systems: A Comparative Study. IEEE Trans. on Vehicular
[33] Shen, M., Peng, F. Z., Converter Systems for Hybrid Electric Vehicles. Electrical
Machines System Conference, pp. 2004-2010, 2007.
[34] Peng, F. Z., Yuvan, X., Fang, X., Qian, Z., Z-Source Inverter for Motor Drives. IEEE
Trans. on Power Electronics, vol. 20, no. 4, pp. 857-63, 2005.
[35] Thangaprakashy, S., Krishnan, A., Current Mode Integrated Control Technique for Z-
Source Inverter Fed Induction Motor Drives. Journal of Power Electronics, vol. 10, no.
3, pp. 285-292, 2010.
[36] Haiping, X., Peng, F. Z., Chen, L., Wen, X., Analysis and Design of Bidirectional Z-
Source Inverter for Electrical Vehicles. APEC Conference, pp. 1252-1257, 2008.
[37] Dehghan, S. M., Mohamadian, M., Yazdian, A., Hybrid Electric Vehicle Based on
Bidirectional Z-Source Nine-Switch Inverter. IEEE Trans. on Vehicular Technoloy, vol.
59, no. 6, pp. 2641-2653, 2010.
[38] Yamanaka, M., Koizum, H., A Bidirectional Z-Source Inverter for Electric Vehicles.
Power Electronic Drive System Conference, pp. 574-578, 2009.
[39] Ding, X., Qian, Z., Yang, S., Peng, F. Z., A New Feedforward Compensation to Reject
DC-Link Voltage Ripple in Bidirectional Z-Source Inverter ASD System. APEC Confe-
rence, pp. 1809-1813, 2008.
[40] Fontes, G., Turpin, C., Astier, S., Meynard, T. A., Interactions between Fuel Cells and
Power Converters: Influence of Current Harmonics on a Fuel Cell Stack. IEEE Trans.
on Power Electronics, vol. 22, no. 2, pp. 670-678, 2007.
[41] Thounthong, P., Davat, B., Rael, S., Sethakul, P., An Overview of Power Converters for
a Clean Energy Conversion Technology. IEEE Industrial Electronics Magaz., vol. 3,
no. 1, pp. 32-46, 2009.
[42] Wang, Y., Chen, K. S., Mishler, J., Cho, S. C., Adroher, X. C., A Review of Polymer
Electrolyte Membrane Fuel Cells: Technology, Applications, and Needs on Funda-
mental Research. Applied Energy, vol. 88, no. 4, pp. 981-1007.
[43] Bizon, N., Nonlinear Control of Fuel Cell Hybrid Power Sources: Part I – Voltage
Control. Applied Energy, vol. 88, no. 7, pp. 2559-2574, 2011.
[44] Duleep, G., Van Essen, H., Kampman, B., Grunig, M., Impacts of Electric Vehicles -
Assessment of Electric Vehicle and Battery Technology. Delft, CE Delft, vol. 2, pp. 9-
29, 2011.
[45] Duduta, M., Ho, B., Wood, V. C., Limthongkul, P., Brunini, V. E., Carter, W. C.,
Chiang, Y. M., Semi-Solid Lithium Rechargeable Flow Battery. Advanced Energy
Materials, vol. XX, pp. 1-6, 2011.
[46] Gruning, M., Witte, M., Boteler, B., Kantamaneni, R., Gabel, E., Bennink, D., et al.;
Impacts of Electric Vehicles - Assessment of the Future Electricity Sector. Delft, CE
Delft, vol. 3, pp. 35-70, 2011.
[47] Tang, Y., Yuan, W., Pan, M., Li, Z., Chen, G., Li, Y., Experimental Investigation of
Dynamic Performance and Transient Responses of a KW-Class PEM Fuel Cell Stack
under Various Load Changes. Applied Energy, vol. 87, no. 4, pp. 1410-1417, 2010.
[48] Bayindir, K. C., Gozukucuk, M. A., Teke, A., A Comprehensive Overview of Hybrid
Electric Vehicle: Powertrain Configurations, Powertrain Control Techniques and
Electronic Control Units. Energy Conversion and Management, vol. 52, no. 2, pp.
1305-1313, 2011.
[49] Wagner, G. O., Real-Time Diagnosis and Maintenance for Automotive System. Journal
of Electronics and Computer Science, vol. 2, pp. 51-56, 2011, http://fecc_bulletin.upit.
ro/PublicSite/Main.aspx?cid=14.
[50] Ramos Paja, C. A., Romero, A., Giral, R., Calvente, J., Martinez Salamero, L., Mathe-
matical Analysis of Hybrid Topologies Efficiency for PEM Fuel Cell Power Systems
Design. Electrical Power and Energy Systems, vol. 32, pp. 1049-1061, 2010.
[51] Kuperman, A., Aharon, I., Battery Ultracapacitor Hybrids for Pulsed Current Loads: A
Review. Renewable and Sustainable Energy Reviews, vol. 15, pp. 981-992, 2011.
[52] Tang, Y., Juan, W., Pan, M., Wan, Z., Experimental Investigation on the Dynamic
Performance of a Hybrid PEM Fuel Cell/Battery System for Lightweight Electric
Vehicle Application. Applied Energy, vol. 88, no. 1, pp. 68-76, 2011.
[53] Barelli, L., Bidini, G., Ottaviano, A., Dynamic Analysis of PEMFC-Based CHP
Systems for Domestic Application. Applied Energy, vol. 91, no. 1, pp. 13-28, 2012.
[54] Thounthong, P., Rael, S., Davat, B., Control Strategy of Fuel Cell/Supercapacitors
Hybrid Power Sources for Electric Vehicle. Journal of Power Sources, vol. 158, no. 1,
pp. 806-814, 2006.
[55] Markel, T., Kuss, M., Denholm, P., Communication and Control of Electric Drive
Vehicles Supporting Renewables. Vehicle Power Propulsion Conference, pp. 27-44,
2009, www.nrel.gov/vehiclesandfuels/vsa/pdfs/46224.pdf.
[56] Zhao, C., Isolated Three-Port Bidirectional DC-DC Converter. Thesis, 2010, www.pes.
ee.ethz.ch/uploads/tx_ethpublications/Zhao_no_CV.pdf.
[57] Liu, Y., Chen, Y. M., A Systematic Approach to Synthesizing Multi-Input DC-DC
Converters. IEEE Trans. on Power Electronics, vol. 24, no. 2, pp. 116-127, 2009.
[58] Kwasinski, A., Identification of Feasible Topologies for Multiple-Input DC-DC
Converters. IEEE Trans. on Power Electronics, vol. 24, no. 3, pp. 856-861, 2009.
[59] Qian, Z., Abdel-Rahman, O., Al-Trash, H., Batarseh, I., Modeling and Control of
Three-Port DC/DC Converter Interface for Satellite Applications. IEEE Trans. on
Power Electronics, vol. 25, no. 3, pp. 637-649, 2010.
[60] Bizon, N., Energy Harvesting from thePV Hybrid Power Source. Energy, vol. 52, pp.
297-307, 2013.
[61] Bizon, N., FC Energy Harvesting Using the MPP Tracking Based on Advanced Extre-
mum Seeking Control. International Journal of Hydrogen Energy, vol. 38, no. 14, pp.
1952-1966, 2013.
[62] Jiang, W., Fahimi, B., Multi-Port Power Electric Interface for Renewable Energy
Sources. IEEE Applied Power Electronics Conference, pp. 347-352, 2009.
[63] Tao, H., Kotsopoulos, A., Duarte, J. L., Hendrix, M. A. M., Family of Multiport
Bidirectional DC-DC Converters. IEEE Proceeding on Electric Power Applications,
vol. 153, no. 3, pp. 451-455, 2006.
[64] Caricchi, F., Crescimbini, F., Napoli, A. D., Honorati, O., Santini, E., Testing of a New
DC/DC Converter Topology for Integrated Wind-Photovoltaic Generating Systems.
Fifth European Conference on Power Electronics and Applications, vol. 8, pp. 83-88,
1993.
[65] Napoli, A. D., Crescimbini, F., Solero, L., Caricchi, F., Capponi, F. G., Multiple-Input
DC-DC Power Converter for Power-Flow Management in Hybrid Vehicles. Industry
Applications Conference (37th IAS Annual Meeting), vol. 3, pp. 1578-1585, 2002.
[66] Chen, Y. M. M., Liu, Y. C. C., Lin, S. H., Double-Input PWM DC/DC Converter for
High/Low Voltage Sources. IEEE Transactions on Industrial Electronics, vol. 53, issue
5, pp. 1538-1545.
[67] Marchesoni, M., Vacca, C., New DC-DC Converter for Energy Storage System Inter-
facing in Fuel Cell Hybrid Electric Vehicles. IEEE Transactions on Power Electronics,
vol. 22, issue 1, pp. 301-308, 2007.
[68] Rodriguez, F. D., Imes, W. G., Analysis and Modeling of a Two-Input DC/DC
Converter with Two Controlled Variables and Four Switched Networks. IECEC Energy
Conversion Engineering Conference, vol. 1, pp. 322-327, 1996.
[69] Dobbs, B., Chapman, G., Lyle, P., A Multiple-Input DC-DC Converter Topology. IEEE
Power Electronics Letters, vol. 1, issue 1, pp. 6-9, 2003.
[70] Tao, H., Kotsopoulos, A., Duarte, J. L., Hendrix, M. A. M., Multi-Input Bidirectional
DC-DC Converter Combining DC-Link and Magnetic-Coupling for Fuel Cell Systems.
Industry Applications Conference, vol. 3, pp. 2021-2028, 2005.
[71] Su, G. J., Peng, F. Z., A Low Cost, Triple-Voltage Bus DC-DC Converter for Automo-
tive Applications. IEEE Appl. Power Electron. Conference, vol. 2, pp. 1015-1021,
2005.
[72] Su, G. J., Cunningham, J. P., Tang, L., A Reduced-Part, Triple-Voltage DC-DC
Converter for Electric Vehicle Power Management. IEEE Power Electronics Specialists
Conference, pp. 1989-1994, 2007.
[73] Al-Atrash, H., Tian, F., Batarseh, I., Tri-Modal Half-Bridge Converter Topology for
Three-Port Interface. IEEE Trans. on Power Electronics, vol. 22, no. 1, pp. 341-345,
2007.
[74] Al-Atrash, H., Pepper, M., Batarseh, I., A Zero-Voltage Switching Three-Port Isolated
Full-Bridge Converter. IEEE International Telecommunication Energy Conference, pp.
1-8, 2006.
[75] Al-Atrash, H., Batarseh, I., Boost-Integrated Phase-Shift Full-Bridge Converter for
Three-Port Interface. IEEE Power Electronics Specialists Conference, pp. 2313-2321,
2007.
[76] Matsuo, H., Lin, W., Kurokawa, F., Shigemizu, T., Watanabe, N., Characteristics of the
Multiple-Input DC-DC Converter. IEEE Trans. on Industrial Electronics, vol. 51, no. 3,
pp. 625-631, 2004.
[77] Chen, Y. M., Liu, Y. C., Wu, F. Y., Multi-Input DC/DC Converter Based on the Multi-
winding Transformer for Renewable Energy Applications. IEEE Trans. of Industry
Applications, vol. 38, no. 4, pp. 1096-1104, 2002.
[78] Krishnaswami, H., Mohan, N., Constant Switching Frequency Series Resonant Three-
Port Bidirectional DC-DC Converter. IEEE Power Electronics Specialists Conference
(PESC‟08), pp. 1640-1645, 2008.
[79] Liu, D., Li, H., A ZVS Bidirectional DC-DC Converter for Multiple Energy Storage
Elements. IEEE Trans. on Power Electronics, vol. 21, no. 5, pp. 1513-1517, 2006.
[80] Tao, H., Duarte, J., Hendrix, M., Three-Port Triple-Half-Bridge Bidirectional Converter
with Zero-Voltage Switching. IEEE Trans. on Power Electronics, vol. 23, no. 2, pp.
782-792, 2008.
[81] Zhao, C., Round, S. D., Kolar, J. W., An Isolated Three-Port Bidirectional DC-DC
Converter with Decoupled Power Flow Management, IEEE Trans. on Power
Electronics, vol. 23, no. 5, pp. 2443-2453, 2008.
[82] Tao, H., Kotsopoulos, A. J., Duarte, J. L.; Hendrix, M. A. M., A Soft Switched Three-
Port Bidirectional Converter for Fuel Cell and Supercapacitor Applications, In: Proc.
IEEE Power Electronics Specialists Conference (PESC‟05), Recife, Brazil, pp. 2487-
2493, 2005.
[83] Inoue, S., Akagi, H., A Bidirectional DC-DC Converter for an Energy Storage System
with Galvanic Isolation, IEEE Trans. on Power Electronics, vol. 22, no. 6, pp. 2299-
2306, 2007.
[84] Duarte, J. L., Hendrix, M., Simoes, M. G., Three-Port Bidirectional Converter for
Hybrid Fuel Cell Systems. IEEE Trans. on Power Electronics, vol. 22, no. 2, pp. 480-
487, 2007.
[85] Zhang, Z., Thomsen, O. C., Andersen, M. A. E., Nielsen, H. R., Dual-Input Isolated
Full-Bridge Boost DC-DC Converter Based on the Distributed Transformers. IET
Power Electronics. vol. 5, issue 7, pp. 1074-1083, 2012.
[86] De Doncker, R. W., Divan, D. M., Kheraluwala, M. H., A Three Phase Soft-Switched
High-Power Density DC/DC Converter for High-Power Applications. IEEE Trans. on
Industry Applications, vol. 27, no. 1, pp. 63-73, 1991.
[87] Chen, Y. M., Liu, Y. C., Wu, F. Y., Multi-Input DC/DC Converter Based on the Multi-
winding Transformer for Renewable Energy Applications. IEEE Trans. on Industry
[88] Tao, H., Duarte, J. L., Hendrix, M. A. M., High-Power Three-Port Three-Phase
Bidirectional DC-DC Converter. Industry Applications Conference (42nd IAS Annual
Meeting), pp. 2022-2029, 2007.
[89] Hwang, J. J., Chen, Y. J., Kuo, J. K., The Study on the Power Management System in a
Fuel Cell Hybrid Vehicle. International Journal on Hydrogen Energy, vol. 37, no. 5,
pp. 4476-4489, 2012.
[90] Saha, S. S., Efficient Soft-Switched Boost Converter for Fuel Cell Applications. Inter-
national Journal on Hydrogen Energy, vol. 36, pp. 1710-1719, 2011.
[91] Markel, T., Simpson, A., Energy Storage Systems Considerations for Grid-Charged
Hybrid Electric Vehicles. CP-540-38538, www.ctts.nrel.gov/.
[92] Park, C. R., Next Generation Lithium Ion Batteries for Electrical Vehicles. Vukovar, In-
Tech, 2010.
[93] Caux, S., Hankache, W., Fadel, M., Hissel, D., On-Line Fuzzy Energy Management for
Hybrid Fuel Cell Systems. International Journal on Hydrogen Energy, vol. 35, no. 5,
pp. 2134-2143, 2010.
[94] Kuperman, A., Aharon, I., Battery Ultracapacitor Hybrids for Pulsed Current Loads: A
Review. Renew Sustainable Energy Review, vol. 15, pp. 981-992, 2011.
Chapter 13
DYNAMIC RESPONSE OF ELECTRIC MACHINES

FOR ELECTRIC VEHICLES/HYBRID ELECTRIC
VEHICLES (EV/HEV)
Robert Beloiu, Mariana Iorgulescu† and Corina Savulescu

ABSTRACT
This chapter presents some of the most common traction solutions used machine for
electric vehicle/hybrid electric vehicle (EV/HEV) adopted by different car manufacturers.
It also covers the dynamic response of the most used electric machines applied in
EV/HEV. We describe three types of electric machines: brushed direct current machine
(DC), brushless machine (BLDC) and induction machine (IM). The brushed DC machine
presents the important disadvantage related to its maintenance cost. The need for brushes
change after a certain period of time makes it very costly. Brushless machines do not
have this maintenance cost related problem. . For the induction machine the cost of
production and maintenance is the smallest among the three analyzed machines.
Consequently, this type of electric machine is more fitted to be used than brushed or
brushless DC machines. .
This chapter describes the behavior of three types of electric machines used
nowadays in EV/HEV commercial vehicles. To compare these types of machines we
developed a testing regime based on the experience of a driver facing all functional
situations met. Thus, we took into consideration the rapid and slow start-stop regime,
speed change regime and the backward movement of the vehicle. We have studied the
behavior of the three types of machines using these testing regimes. In spite of their
advantages and disadvantages of one versus the others, these types of electric machines
are all used nowadays in the prototypes and commercial EV/HEV.
Keywords: Electric Machine, Electric Drive, Electric Vehicle, Hybrid Electric Vehicle,
DC Machine, Induction Machine, Brushless Machine, Dynamic Regime

E-mail: robert.beloiu@upit.ro.
†
E-mail: mariana.iorgulescu@upit.ro.

E-mail: corina.savulescu@upit.ro.
354 Robert Beloiu, Mariana Iorgulescu and Corina Savulescu
1. INTRODUCTION
Nowadays, there is a great interest in the field of EV/HEV. Many car companies develop
solutions for EV/HEV both commercial as prototypes still under investigation. These
solutions include different types of electric machines (DC, brushless, induction, synchronous,
etc.). However, these machines present advantages and disadvantages and there has not been
adopted a standardized unique solution adopted yet.
These solutions are available today on EV/HEV market [1,2]:
 DC machine: PSA Peugeot-Citroen/Berlingo

 Induction machine: GM Silverado ISG, Continental ISAD, Delphi-Automotive ISG,
Valeo ISA, Renault Kangoo, Chevrolet Silverado, BMX X5
 Brushless permanent magnet machine: Honda Insight and Civic mild hybrids, FCX-
V3 FCEV; Mannesman-Sachs
 Synchronous machine: Toyota Prius, Estima, Ford Hybrid Escape, Nissan Tino,
Honda Insight
The electric drive is an essential part in the EV/HEV structure. In order to ensure the
desired capabilities of a vehicle, the traction sub-ensemble has to fulfill certain characteristics
given by the electric drive system which includes the electric machine.
It is known that electric machines are manufactured in a variety of configurations and that
they are able to drive all kind of industrial applications. Theoretically, any kind of electric
machine could be used to drive EV/HEV, yet the car companies focused on some specific
electric machines that proved to provide advantages over the others.
The electric machine for EV/HEV needs to have various specific characteristics related to
the different functional regimes: acceleration, braking, very fast change in the functional
regime, capability of developing very low and also very high speeds, frequent starts and
stops, good proportion between mass and volume, robustness, functionality and low
maintenance cost, etc. [3]. The mechanical characteristics for electric machines used for
EV/HEV are presented in Figure 1.
Figure 1. Typical Performance Characteristics of Electric Machine for Traction [4].

Dynamic Response of Electric Machines for Electric Vehicles … 355
In this chapter we present some of the most common electric machines for EV/HEV
adopted by different car manufacturers. We also describe the dynamic response of the best
know electric machine used in EV/HEV, and we end by comparing the studied electric
machines.
2. ELECTRIC MACHINES USED IN EV/HEV

2.1. The DC Machine
There are four constructive types of DC machines, depending on the excitation field:
series, independent, shunt and compound. The DC permanent magnet machine falls in the
category of independent field excitation machines. The constructive type depends on the way
the winding field is connected to the winding armature shown in Figure 2.
(a) Series connection [6]
(b) Compound connection
(c) Shunt connection
Figure 2. DC machines configurations.

The use of DC machines in electric traction is not novel. The most used DC electrical
machines in traction applications are the series excitation winding, due to their mechanical
characteristics that allows a smooth start with a large resistive torque. Even though the
behavior of the DC machine is well known and well covered in the literature, there is still
interest in developing new functional regimes. The ideal traction machine should have both
the characteristics of series field connection (for starting regime) and of the shunt connection
(for regenerative braking and speed control). Such behavior should improve the EV/HEV
functionality and the battery use [5].
2.1.1. Mechanical Characteristics

The steady state equivalent circuit of the DC machine is indicated in Figure 3, and it is
widely covered in the technical literature.
Figure 3. Equivalent circuit for DC machines.
The basic equations, using the standard notations that describe the steady state functional
regime of a DC machine, are as follows:
U A  RA  I A  E; E  K    
 (1)
T  K    I A
where RA - armature resistance, I A - armature current, T - developed torque,  - field

excitation, K - machine constant,  - rotor angular speed.
The relation between the rotor angular speed and the developed torque T is derived from
the Equation(1), represents the mathematical form of the mechanical characteristic:
UA RA
  T (2)
K    K   2
The Equation (2) refers to all types of DC machines, and the graphical representation is
indicated in Figure 4.
Figure 4. DC machines‟ mechanical characteristics.
Before development of power electronic devices, the DC series machine was the only
fitted for electric traction applications, due to its mechanical characteristic, such as:
development of high rated torque values for low values of speed. Even though this special
characteristic is very important for electric traction vehicles, the use of DC series machines
has its limitations, especially because the speed rises as the torques values decrease.
Nowadays, when power electronics and numerical control developed very much,
manufacturers overcame the past limitation of using DC shunt machine for EV/HEV
applications. Due to the modern drive systems, they combined the ease in controlling the DC
independent/shunt connection machines and the advantages of DC series connection
machines.
Analyzing the Equation (2) for DC shunt machines, there are several functional regions
that could be defined as indicated in Figure 5:
 Nominal functionality
 Below nominal functionality
 Above nominal functionality
The “nominal mechanical characteristic” is obtained when the machine functions at the
nominal parameters. The point in Figure 5 represents the Static Operating Point (SOP) when
the voltage and torque are at nominal values.
The “below mechanical characteristic” is obtained by varying the armature voltage. The
point in Figure 5 represents the SOP when applied voltage is below the nominal value. The
“below mechanical characteristic” is characterized by lower speed comparing to the nominal
value. The mechanical characteristics are parallel to the nominal. When the machine works at
low and very low speed, the ventilation is lower and the temperature of the machine rises.
Thus when working at this parameters, cooling of the machine should be taken into
consideration.
The “above mechanical characteristic” is obtained by varying the field voltage. The point
in Figure 5 represents the SOP when applied voltage is above the nominal value. The “above
mechanical characteristic” is characterized by higher speed comparing to the nominal value.
The mechanical characteristics are not parallel to the nominal. The load capacity of the
machine is lower than the nominal and below nominal mechanical characteristics. Thus, when
working at these parameters, the load capacity has to be taken into consideration.
Figure 5. Mechanical characteristics for shunt DC machines.
2.1.2. Chopper Control of DC Machines

The chopper control is not the only controlling method by which the speed of a DC
machine can be varied. A different method of speed control is by adding supplementary
resistors. Because of energetic consideration, this method is not appropriate for EV/HEV, and
therefore it is not covered in this chapter.
The variation of speed for the machine can be controlled by variation of the field voltage
variation. This method is not covered in this chapter.
2.1.2.1. Chopper Control for Single-Quadrant DC Functionality

The control of the armature voltage applied to the DC machine can be obtained by the use
of chopper control techniques.
The generic chopper schematic consists in a power electronic switch connected in series
with the DC machine, like in Figure 6.
The power source has constant value V, which in case of an EV/HEV is provided by the
main battery. The voltage is applied to the DC machine through the power electronic switch
PES. The free regime diode (FRD) ensures power dissipation from the machine while PES is
open.
In the duty period of time 0  t   T , the PES is closed and the source voltage is applied
to the machine. In this period of time, the current rises from the minimum value I A _ min to the
maximum value I A _ max . In the freewheeling period of time  T  t  T , the PES is open and
no voltage is applied to the machine. In this time interval the current falls from the maximum
value to the minimum value I A _ min . The FRD ensures the electric current while the PES is
open, thus improving the current waveform.
2.1.2.2. Chopper Control for Multiple-Quadrant DC Functionality

The problem of the indicated schematic in Figure 6 is that the DC machine functions only
in one functionality quadrant. For EV/HEV the DC machine has to be able to operate in all
four functionality quadrants.
In order to obtain this, a generic schematic for the chopper is used in Figure 7. All
PES are closed and opened alternatively PES1 + PES3 and PES2 +PES4 in order to ensure
full functionality of the DC machine for all four quadrants. FRDs also work in groups of two
depending of the working quadrant of the DC machine.
Figure 6. Single-quadrant DC machine chopper drive.
Figure 7. Four quadrant DC machine chopper drive.
2.1.2.3. Driving System of Chopper Controllers for DC Machines

Either for single-quadrant or multiple-quadrant functionality of DC machines, the driver
is a closed loop system that controls both the current and speed.
The schematic of closed loop control of shunt DC machine is indicated in Figure 8.
In Figure 8, the is the current transducer and the is the speed transducer. The Rw and RIa
are the speed and current regulators placed in the speed and respectively current closed
control loops. The measured current and speed are compared with the prescribed values and
the result is applied to the respective regulator. The output signal of the current regulator, is
the G command signal for the PES. Thus the voltage level applied to the machine armature is
obtained.
Figure 8. Schematic for a drive system for chopper controllers for DC machine.
2.2. Induction Machine
There are two types of induction machines: squirrel cage (short circuit rotor) and
wounded rotor.
Because of their high maintenance cost due to brushes, the wound-rotor induction
machines are not as convenient as the squirrel cage induction machines for EV/HEV
applications.
The squirrel cage induction machines have a simple construction and thus their
fabrication and maintenance cost is lower comparing with other types of electric machines.
One disadvantage of the induction machines is the difficulty of controlling in closed loop
systems. Sometimes the controller is more expensive than the machine, making these
machines difficult to integrate in variable speed applications. However, these difficulties were
overcome by the evolution of power devices and industrial computing.

In order to determine the functional characteristics of the induction machine, its
equivalent schematic is used in the specialized literature. This is indicated in Figure 9.
In Figure 9, the U s , I s , Rs , and Ls are the stator quantities of voltage, current, resistance
Rr
and inductance. The I r , , and Lr are the rotor quantities of current, resistance and
s
inductance reported to the stator. The rotor slip is s, which indicates the delay between
the angular speed of the stator field and the rotor angular speed defined as:
1  2
s (3)
1
where 1 is the synchronous speed (the speed of the inductor field), 2 is the rotational rotor
speed and p the number of pole pairs.
Analyzing the equivalent schematic, the Kloss relation indicates the relation between the
torque and slip speed, is approximated by:
T 2
 (4)
Tcr s scr

scr s
Figure 9. The equivalent schematic of induction machine.
In the Equation (4), Tcr and represents the critical torque and respectively the critical slip
speed, while and s represents the torque and slip speed at certain functional conditions.
From Equation (4) we obtain the mechanical characteristic for the induction machine,
displayed in Figure 10.
(a) (b)
Figure 10. Mechanical characteristic of induction machines for U, f constant control, (a) T  f  s  ,
(b)   f T 
.
As mentioned above, Figure 10 presents the mechanical characteristic of an induction

machine when the supply voltage and frequency are constant. It can be observed that the
starting torque of the induction machine has low values and the speed variation is very small
with the load variation. When the induction machine is operated, the stable functional domain
s  scr
is when . This domain is very narrow and therefore the speed can vary in a very
limited domain. For the EV/HEV applications, this behavior is not acceptable.
2.2.2. Control of Induction Machines for Constant U/f

In the domain of constant U/f the induction machine behaves similar with DC machines.
The mechanical characteristics of its control are indicated in Figure 11.
The “nominal mechanical characteristic” is obtained when the machine functions at
nominal parameters. The point in Figure 11 represents the Static Operating Point (SOP) when
all physical quantities applied to the machine are at nominal values.
Figure 11. Mechanical characteristics for U/f constant control.
The “below mechanical characteristic” is obtained by varying the supply voltage and
frequency so that U/f is maintained constant. The point in Figure 11 represents the static
operating point (SOP) when the applied voltage and frequency are below the nominal values.
The “below mechanical characteristic” is characterized by lower speed compared to nominal
value. The below mechanical characteristics are parallel to the nominal. When the machine is
working at low and very low speed, the ventilation decreases and consequently its
temperature rises. Therefore when working at these parameters, machine cooling has to be
taken into consideration.
The “above mechanical characteristic” is obtained by varying the frequency of the supply
voltage. The value of the voltage is maintained constant at the nominal value. The point in
Figure 11 represents the static operating point (SOP) when the applied frequency is above the
nominal value. The “above mechanical characteristic” is characterized by higher speed than
the nominal. The mechanical characteristics are not parallel to the nominal. The load capacity
of the machine is lower than the nominal and below nominal mechanical characteristics.
Therefore, when working at these parameters, capacity load has to be taken into
consideration.
2.2.3. Driving System of U/f Controllers for Induction Machines

The U/f controller consists of 6 PESs, displayed in Figure 12. These PESs ensure that the
source voltage is applied to the stator windings of the induction machine IM. The PESs are
commanded with signals provided by three comparators. The command signal is obtained by
comparing the 3ph sinus voltage provided by U sin with the triangular signal provided by U tri .
Figure 12. Schematic for a U/f constant controller drive system of s for induction machines.
The PES1…6 are power electronic switches that could be of different types, depending
on the power of the machine, and consequently on the absorbed current. The groups of
PES1,3,5 and PES2,4,6 are closed alternatively depending on the output signal of COMP1-3.
When the sinusoidal voltage the output signal COMP1  0 . When the output signal
COMP1<0 (Figure 13).

Figure 13. The schematic of command circuit of U/f constant controllers for induction machines
Depending on the polarity of signal either PES1 or PES2 conducts providing maximum
or minimum values of the voltage supply applied at machine‟s phases, obtaining the
alternating current values in its windings.
2.3. Brushless Machine
Regarding construction principles, brushless (BLDC) machines are similar to the

permanent magnet synchronous machines. The similarity between brushless machines and
synchronous machine lies in the fact that both magnetic fields produced by the stator and the
rotor rotate at the same frequency.
In a brushless machine, the stator windings‟ configuration is similar to that of poly-phase
AC machines, especially to the induction machines. However, the winding configuration is
distributed differently.
The structure of the rotor of a brushless machine contains one or more permanent
magnets, and can vary from 2 to 8 pole pairs.
The brushless machines are manufactured in 2-phase or 3-phase configuration, although
the 3-phase is wider spread in industrial applications.
Figure 14. Mechanical characteristic of BLDC machines.
Compared to the brushed DC machines, the BLDC present some important advantages,
such as:
 Easier cooling due to windings placement in the stator

 Lower maintenance cost due to the lack of brushes
 Higher efficiency, as the Joule loses are lower than the corresponding DC brushed
permanent magnet machine due to the windings placement in the stator
In the technical [7] and scientific [8] literature, it is considered that the mechanical
characteristic of BLDC is very similar with the brushed DC machines.
Analyzing the mechanical characteristic of BLDC machine, it can be stated that its
behavior is similar with the brushed DC machine. At starting the machine develops a high
peak torque, which leads to a high peak current.
2.3.2. Control of Brushless Machines

The brushless driving system contains six PES controlled by a Gate Command System
(GCS). The GCS block receives signals from the machine and the imposed speed *. Based
on the input signals, the GCS forms the command signals for the PES1-6 that drives the
machine. Both the electronic system GCS and PES1-6 function as an electronic commuter
which drives the machine. In traction applications, the speed control for brushless machine
raises a special interest [7].
Figure 16 displays some waveforms that appear in a driving schematic of brushless
machines. Figure 16(a) indicates the currents waveforms. Depending on the inertia of the
machine and the resistant torque at the shaft, the current of the machine at starting has a high
peak value.
The CGS block forms the command signals for the PES1÷6 indicated in Figure 16(c).
The inputs of CGS are the machine currents, feedback signals and the output of a PI speed
regulator.
Figure 15. Driving system schematic for a brushless machine.
(a) Currents variation
(b) Feed-back signals

(c) PES command
Figure 16. Waveforms in the driving circuit of brushless machine.

3. DYNAMIC REGIME OF ELECTRICAL MACHINES USED

IN EV/HEV
In EV/HEV the functional regime of the electric traction machine could be of two types:
 transient or dynamic
 stable
The dynamic regime for EV/HEV has some particularities compared to the industrial
applications. In the normal use of a vehicle, various situations can occur consisting in:
acceleration, braking, very fast change in the functional regime, capability of developing very
low and also very high speeds, frequent starts and stops etc [9].
For the purpose of comparison between different types of electric machines used
nowadays for EV/HEV, a speed profile was created to be applied to each type of machine
(Figure 17).
Figure 17. Testing speed profile.
The following notations are used in Figure 17:
 0 periods of time when the vehicle is stopped

 1, 2 rapid variation of speed
 3, 4, 5, 8 stable speed regime when the vehicle is moving forward
 6 stable speed regime when the vehicle is moving backward
 7, 9 slow speed variation
The speed variation situations presented above are intended to simulate several functional
regimes most likely to appear during the use of a real vehicle. The 1 and 2 could be speed
regimes likely to appear at the traffic light crossroads. Situations 3, 4, 5, 8 could be various
levels of speed cruise, while 6 is moving backward. Situations 7, 9 are likely occur while
starting/stopping the vehicle. The transition 0-3 simulates very rapid starting, while 3-4-5
simulates rapid braking and accelerating situations. The 5-0 simulates a dangerous situation
when a very rapid stopping of the vehicle is needed. The situation simulated by 3-4-5 could
very easy be a cruise situation, when a rapid increase of speed is needed for overtaking other
vehicle. All these transitory regimes are very common in daily driving, especially when using
vehicles in crowded urban areas. For the experimental data, we applied the same resistant
torque at the shaft to the analyzed machines: DC, induction and brushless.
Nominal data of the machines are indicated in Table 1.
Table 1. Nominal data of tested machines
DC machine Induction machine BLDC machine

UAn = 220 V c.c. Un = 230/400 V M0 = 2.6 Nm
IA = 2 A, In = 3,5/2,1 A Mmax = 13 Nm
Ifield = 0,2 A fn = 50 Hz I0 = 4.2 A
nn = 1.500 rpm Pn = 1 CV = 0.735 kW Imax = 21 A
Pn = 0.373 kW nn = 1500 rpm nn = 4000 rpm
3.1. DC Machine
The DC machine functions in a closed loop control speed schematic. The schematic
contains also a subordinate current control loop. In order to accurately function, the driver
should be set before use and the most appropriate configuration for current and speed
regulators should be chosen. The procedure of setting up the DC machine‟s driver is
presented in previous works of the authors [10,11]. Briefly, this procedure consists in
adjusting the values for the current and speed regulators‟ parameters to obtain the most
appropriate behavior in the transient regime.

Figure 18. Set-up of DC machine driver control loops.
Figure 18 represents the excitation speed signal and the acquired output speed signal for
various configurations of the driving schematic. During the parameter adjustment, different
improved output signals are obtained. The best configuration of the driver is the one that
provides the quickest answer from the machine. The best response is a compilation of the
time delay, the override output signal and the delay of the steady state.
Figure 19. Dynamic behavior of DC machine.
Once the best considered configuration is chosen, the system is tested using the user-
defined speed testing profile as indicated in Figure 19. It is noteworthy that the output speed
of the machine follows the testing speed profile.
3.2. Induction Machine
The induction machines function in an open loop U/f converter. This means that there is
no speed feedback to the controller. For this reason, there is no need for adjustment of internal
regulators dynamic parameters.
Figure 20. Dynamic behavior of induction machine.

The dynamic behavior of the system containing the induction machine is indicated in
Figure 20. The output speed signal follows the testing speed input signal. The shape of the
input and output speed signals is similar, but there are substantial errors between the two
signals. The main reason for these errors is that the system functions in open loop
configuration. Thus the speed error is not corrected in any way through a feedback.
3.3. Brushless Machine
The brushless machine, likewise the DC machine, functions in a closed speed control
loop. The brushless machine‟s driver parameters should be set-up before any experiment is
conducted. The procedure of setting up the brushless machine‟s driver is covered in previous
works of the authors [12]. Just as for the DC machine, the procedure consists in establishing
the parameters for the current and speed regulators‟ parameters.
Figure 21 presents the set-up procedure for the brushless system. Equally as the DC
brushed machine, the BLDC machine drive includes a regulator parameter configuration in
order to achieve the best behavior. As in any closed loop systems, monitored parameters are:
the delay of the dynamic response, the override signal and the delay to reach the steady state.
Figure 21. Set-up of brushless machine driver control loops.
Figure 22. Dynamic behavior of BLDC machine.

The effect of the feedback loop of the driver for BLDC machine is displayed in Figure
22. The shape of the output speed signal is similar to the input speed excitation signal. There
are slight errors between the input and the output speed signals.
4. COMPARATIVE BEHAVIOR OF ELECTRIC MACHINES FOR EV/HEV

For the comparison of the dynamic regime for the three electric machines, an excitation
speed signal was applied to its‟ drivers (Figure 23). The dynamic response of these machines
was recorded (Figure 24)
Figure 23. Testing speed profile.
Figure 24. Comparative dynamic behavior of DC, induction and BLDC machines.
Firstly the speed responses of DC and BLDC machines are very similar, which is
consistent to other results from technical [8] and scientific literature [13] . Secondly, for the
forward move of the EV/HEV equipped with these kinds of machines, the error between the
input and the output speed signals is small. Therefore in all imposed functional regimes, the
output signal follows the input signal with an acceptable error.
Third, in the slow deceleration regime (9 in Figure 24) a smoother behavior for the
BLDC machine compared to the brushed DC machine can be noticed. In addition, at a sudden
change of functional regimes, the DC machine has bigger override value of measured speed
than the BLDC machine. When the slow acceleration is applied (7 Figure 24) the override of
the DC machine is again bigger than the one for the BLDC machine. Nevertheless, the
response is attenuated rapidly and the brushed DC machine has a very similar behavior with
the BLDC machine.
Forth, a different situation can be observed for the induction machine. The experiments
revealed a very large error of the induction machine‟s output speed compared to the input
speed-testing signal. At every regime change there is a notable delay, in both senses of
rotation. The profile of output speed for the induction machine follows the profile of the
testing signal, but there is an important difference in amplitudes.
One of the reasons for such behavior is the fact that in this circumstance, the induction
machine‟s controller is an open loop U/f driver. Therefore the output speed is never compared
with the input signal and speed regulators never process the error.
CONCLUSION
From the above mentioned analysis the following conclusions could be drawn:
 The most accurate dynamic behavior for regime changes is offered by the BLDC
machine as indicated by our tests
 The dynamic behavior of brushed DC machine is similar with the BLDC machine
 The dynamic behavior of the induction machine controlled by a U/f driver in an open
loop configuration presents very large errors
Each analyzed solution for traction for EV/HEV in this chapter has its advantages and
disadvantages.
The brushed DC machine fabrication technology and control systems arrived at their
maturity, it is envisaged its costs will decrease [14]. Nevertheless the brushed DC machine
presents the important disadvantage of high maintenance cost.
The brushless machines do not present the same problem as the brushed DC machines
related to maintenance cost, as the brushes are eliminated. Since the fabrication process of
magnets improves and the price for control and power electronics decreases, the BLDC
machines become a very suitable solution for replacing the brushed DC machines.
Comparing the brushless machines with induction machines, it is outlined in the literature
that the stators for a three-phase induction machine and for a BLDC machine are practically
identical [14]. However, these machines differ in the rotor construction and the electronic
controller.
The cost of production and maintenance for induction machines is the smallest among the
three types analyzed. For this reason they are more fitted to be used than the brushed or the
brushless DC machines in EV/HEV, but they are more difficult to control.
In spite of presented advantages and disadvantages, all three categories of machines are
used in prototypes and already commercial EV/HEV. There is still more room for further
studies to develop the best electric machine configuration to be implemented in EV/HEV.
NOMENCLATURES
BLDC Brushless DC
COMP Comparator
DC Direct Current
FRD Free Regime Diode
GCS Gate Command System
PES Power Electronic Switch
SOP Static Operating Point
REFERENCES
[1] Miller, J. M.; Propulsion Systems for Hybrid Vehicles. The Institution of Engineering
and Technology, London, UK, 2008.
[2] Zeraoulia, M.; Benbouzid, M. H.; Diallo, D.; Electric Motor Drive Selection Issues for
HEV Propulsion Systems: A Comparative Study. IEEE Transactions on Vehicular
[3] Xue, X. D.; Cheng, K. W. E.; Cheung, N. C.; Selection of Electric Motor Drives for
Electric Vehicles. Australasian Universities Power Engineering Conference
(AUPEC'08), Sydney, Australia, paper 170, pp. 1-6, 2008.
[4] Dong, G. G.; et al.; Studies of Electric Motor for Light-weight Electric Vehicle.
Malaysian Universities Transportation Research Forum and Conferences
(MUTRFC2010), Putrajaya, Malaysia, pp. 135-148, 2010.
[5] Khan, R. U.; Khan, M. A.; Arjumand, H.; Analysis of Innovative Applications of Single
DC Motor in Series and Separately Excited Mode for Hybrid Electric Solar Car.
International Journal of Engineering Science and Technology, vol. 2, no. 3, pp. 312-
316, 2010.
[6] www.powersimtech.com.
[7] Mehrdad, E.; Yimin, G.; Ali, E.; Modern Electric, Hybrid Electric, and Fuel Cell
Vehicles. CRC Press, 2010.
[8] Jung, G. H.; Yeom, J. H.; Kim, M. G.; Determination of Torque-Speed-Current
Characteristics of a Brushless DC Motor by Utilizing Back-EMF of Non-Energized
Phase. Journal of Magnetism and Magnetic Materials, vol. 310, pp. 2790-2792, 2007.
[9] Xue, X. D.; Cheng, K. W. E.; Cheung, N. C.; Selection of Electric Machine Drives for
Electric Vehicles. Australasian Universities Power Engineering Conference
(AUPEC'08), Sydney, Australia, pp. 1-6, 2008.
[10] Beloiu, R.; Iorgulescu, M.; Dumitru, O.; Combei, P.; Stancescu, A.; Electric Drive
System with DC Motor with Closed Control Speed Loop. Electronics, Computers and
Artificial Intelligence (ECAI 2005), Pitesti, Romania, pp. 47-50, 2005.
[11] Beloiu, R.; Iorgulescu, M.; Dumitru, O.; Combei, P.; Stancescu, A.; Electric Drive
System with DC Motor with Closed Current Loop and Open Speed Loop. Conference
on Electronics, Computers and Artificial Intelligence (ECAI 2005), Pitesti, Romania,
pp. 50-53, 2005.
[12] Beloiu, R; Iorgulescu, M.; Set up of Brushless Control Drivers in Hybrid Electric
Vehicle. Conference on Electronics, Computers and Artificial Intelligence (ECAI
2011), Pitesti, Romania, pp. 109-113, 2011.
[13] G. Ellis; Advances in Brushless DC Motor Technology, Control, and Manufacture,
PCIM-Europe, 1996, www.kollmorgen.com/uploadedfiles/Files/Website/Common/
Images/pcim1996brush_vs_brushless.pdf, (accessed January 2014).
[14] W. Rippel; Induction versus DC Brushless Motors, http://www.teslamotors.com/blog/
induction-versus-dc-brushless-motors, (accessed January 2014).
Chapter 14
USE OF THE EXTERNAL MAGNETIC FIELD FOR

SUPERVISION AND DIAGNOSIS OF ELECTRICAL
TRACTION MOTORS
Remus Pusca 1,2,* Raphael Romary1,2,† and Andrian Ceban1,2,‡

University of Lille Nord de France, Lille, France
University of Artois, LSEE, Bethune, France
ABSTRACT
The work presents a supervision method for detection of AC electrical machine
faults, generally used in electrical traction, exploiting the information given by the
external magnetic field in the vicinity of the machine. This magnetic field is an image of
the air-gap flux density and consequently, it contains information concerning the failure
presence. In this work, two methods based on the analysis of the magnetic field outside
the machine are presented. They are reliable, original, inexpensive and simple to
implement in electrical automotive vehicles. They present the advantage of being non-
invasive and just get rid from the main drawback presented by other diagnosis methods
based on the comparison with a healthy state assumed to be known. Both methods need
simple coil sensors placed in specific positions near the motors for flux density
measurements. These sensors induce an electromotive force whose derivative effect
amplifies the low and medium frequency harmonic components. These methods are
developed in the case of AC electrical rotating machines (induction and synchronous
machines) often used in electrical traction. The analysis is focused on the low frequency
components of the magnetic field which is not influenced by the presence of electronic
converters usually used in traction applications. The theoretical analysis and the
experimental validation are also presented.
*
E-mail: remus.pusca@univ-artois.fr.
†
E-mail: raphael.romary@univ-artois.fr.
‡
E-mail: andrian.ceban@univ-artois.fr.
380 Remus Pusca, Raphael Romary and Andrian Ceban
Keywords: AC Motors, Electrical Traction Application, Fault Diagnosis, Low Frequency,

Magnetic Field Spectral Analysis
1. INTRODUCTION
The work presented in this chapter is a part of the current concerns in the development of
electric traction systems where EV (electric vehicle) and HEV (hybrid electric vehicle) are
used. Although electric traction is not a new idea, its application to electrical automotive
vehicle has be previously abandoned, consequence of an insufficient level of technological
developments. Today, the interest in electrical traction grows in order to respond to current
problems like the development of a non-polluting vehicle with sufficient reliability and
autonomy. In this context, the automotive vehicle should not be a modification of the
conventional one but it must be a "new vehicle".
Taking that point into consideration the work presented in this chapter shows a non-
invasive method for supervision and diagnosis of electrical motors used in automotive
vehicles. The presence of new traction control systems in automotive vehicles allows faster
control of drive torque and improved dynamic performance [1]. However, these systems
require new methods of monitoring and diagnosis to detect the faults because strong
vibrations in embedded systems reduce the reliability of all components including electrical
machines (motors and generators). To improve the reliability of the "new vehicle", the
proposed method, presented in this chapter, can insure a preventive diagnosis based on the
analysis of the magnetic flux leakage measured outside of the electrical machines by non-
invasive sensors.
Considering the electrical motor as part of EV and HEV, this chapter presents the sensors
used, their location in the vicinity of motor and the analysis of measured signals allowing an
easy use of the method for supervision and diagnosis of AC motors. Generally, electrical
motor diagnosis methods require a good knowledge of the machine and its behaviour in the
presence of internal faults. To diagnose electrical machines used in classical applications,
different methods can be distinguished [2]:
 diagnosis techniques that rely on knowledge methods and which usually use artificial
intelligence methods. These methods are enriched by feedback and do not require the
mathematical model of a faulty machine.
 methods of identification and estimation of the physical parameters of the machine.
In this case, monitoring is established by comparing the measurement results with
those obtained by analytical modelling or by calculation of a state vector.
 methods based on signal modelling. These methods analyze the time evolution or the
spectral content of measured signals in order to detect a fault of the electrical
machine.
Today, many works are dedicated to motor fault analysis using different methods such as
the current signature analysis [3, 4, 5], exploitation of the slotting harmonics [6, 7], artificial
intelligence techniques [8], vibrations and noise measurement [9, 10, 11] or flux analysis [12,
13, 14]. All these techniques have been validated beyond the research laboratories in order to
Use of the External Magnetic Field for Supervision ... 381
find their place in industry [15], and it could be interesting to check if these methods can be
implemented in electric traction motors in order to diagnose incipient faults. In electric
traction applications, the speed variation is often insured by static converters which feed the
motors with adapted waveforms of the currents. These signals are generally not perfectly
sinusoidal and are source for electromagnetic disturbances. Moreover, the interpretation of
the results obtained with these methods requires an advanced level of expertise which hinders
a real democratization of the techniques among users of electrical machines. Indeed, only
systems where a breakdown of the machine can have disastrous consequences (e.g. the
electricity generating plant) are equipped with a supervision system.
Generally, all diagnosis methods require the knowledge of the healthy state of the
machine regardless of the exploited variable. The fault detection is then based on the
comparison of a given signature with that of the healthy state assumed to be known,
considering an indicator sensitive to a particular fault. The difficulty with this approach is
that, often, the sensitive indicators exist even for the healthy machine (case of a spectrum
lines). Then, it is their change, often their increase, that is likely to provide information on the
presence of a fault. On the other hand, the load of the machine can be considered for
diagnosis as a disturbing factor, because it induces several healthy states. Additional difficulty
lies in the fact that the supposed healthy state is almost never known before the failure
appears, because the operator of the machine has not taken the precaution of recording the
parameters that characterize the healthy machine.
Recently, methods based on the analysis of the external magnetic field have appeared,
they have the advantage of being non-invasive and easy to implement [13, 16, 17]. These
techniques exploit information in the stray field around the machine in order to respond to
industrial problems concerning the diagnosis [18], but also on monitoring the aging of
insulation or the monitoring of eco efficiency [19, 20]. The various applications differ in the
analyzed frequency range.
The diagnosis techniques exploiting the low and medium frequency ranges, up to the
slotting effect frequencies of the electric machine, give the possibility to detect stator faults
(inter-turn short circuit) and rotor faults (broken bars, inter-turn short circuit) by analyzing the
external magnetic field of induction and synchronous machines. These techniques are based
on an analytical model which considers the decomposition of the external magnetic field into
a radial and axial field. The both fields do not have the same properties, because each has a
different origin. This model is able to predict the spectral components of the stray field
affected by different faults.
The work presented in this chapter mainly considers the effects of a fault on the magnetic
stray field. It is oriented toward the diagnosis of rotating electrical machines including electric
traction motors since. The proposed diagnosis methods are not influenced by the static
converters and use the load operating condition for detection of the faults. Nowadays, rotating
electrical machines are presented in all areas of modern life including the electrical vehicle
and one of the most important objectives of the maintenance systems is the detection of
incipient faults in order to anticipate a shutdown of the system. In this work, two methods
based on the analysis of the magnetic field outside the machine are presented.. These methods
are developed in the case of AC electrical rotating machines (induction and synchronous
machines). The analysis focuses on the magnetic stray field in the vicinity of the machine,
especially its radial an axial distribution which presents different sensitivity according to
various faults.
To apply these methods in electrical traction, two coil sensors are used to measure the
stray field in the motor vicinity (Figure 1). These sensors induce an electromotive force (emf)
which derivative effect amplifies the low and medium frequency components. They are
simple to carry out, non-invasive and low cost. Moreover, they can be set up outside the
motor without needing much space in electrical system of hybrid vehicles. The diagnosis
procedure is able to indicate the failure presence and can be implemented within a simple
microcontroller. For diagnosis, each fault gives specific signature that are identified and
analyzed.
Power
Electronics Battery
Sensor 1
Control
Electric
Motor Generator ICE
Sensor 2
Figure 1. Positions of sensors for on-line supervision of HEV traction motor.
Two diagnosis methods are developed in this work. The first one deals with the inter-turn
short-circuit faults in the stator winding of an AC machine. It uses two external flux coil
sensors placed in specific positions that measure the external magnetic field. The signal
delivered by the both sensors is exploited for diagnosis. The analysis concerns the magnitude
of specific harmonics of the electromotive force delivered by the coil and the method does not
require any knowledge of a presumed healthy former state of the machine. Its originality is
the use of the load to improve the diagnosis. Here, the load is not longer a disturbing factor
but it is used for the fault discrimination. Actually a comparison between machine operating
conditions (no-load and load) enables to detect a stator fault.
The second method is based on the variations of the axial flux density in the presence of
rotor faults. Any changes of magnetic circuit conditions, such as a broken bar or the
appearance of an eccentric air-gap due to bearing wear, will be reflected with a corresponding
change in the axial flux. The analysis is focused on the low frequency components of the
magnetic field, which are not influenced by the presence of electronic converters usually used
in industrial and traction applications. Moreover, due to their low frequency, these
components are less attenuated during their penetration through the stator magnetic core or
the machine external frame. Both methods need simple coil sensors for flux density
measurement.
In the end of the chapter, the experimental implementation of the sensors and its results
are presented. The faults have been artificially created in the rotor and in the stator of
different AC machines, especially designed for the tests.
2. MAGNETIC FIELD
From a diagnosis point of view, the advantage of methods based on the measurement of
the external magnetic field is that they are non-invasive. These methods are simple to
implement and they need a sensor placed near the electrical machine, able to measure the
external magnetic field. Considering the position of the sensor, the measured stray magnetic
field does not originate from the same source, and does not result from the same physical
phenomenon [21].
2.1. Measurement of External Magnetic Field
In a rotating electrical machine, the leakage flux is created by the different elements of
the machine. A distribution of these field lines is shown in Figure 2 [22].
From this distribution, one can decompose the external field in two parts called axial field
and radial field. These two fields are combined to generate the stray magnetic field: the axial
field is in a plane that includes the axis of the machine and the radial field is in a plane
perpendicular to the axis of the machine. The stray field may give information about the
presence of a fault, and sometime its location. Figure 3 shows the different positions for the
measurement of the stray field using a coil sensor.
Leakage flux generated

by the magnetic circuit
Leakage flux generated Leakage flux in the

by the coil heads outside of the motor
frame
Leakage flux generated
by rotor heads Axial field
Ball bearing
Motor frame
Rotor
Short-circuiting
ring Stator
Coil head
Figure 2. Distribution of leakage flux lines in electrical machine.
Assuming the field lines shown in Figure 3, the sensor will be sensitive to different parts
of the field, according to its position:
 in position 1 (Position 1), the coil measures the axial field. In this case the amplitude
of the measured flux is maximum at the motor end-shield (near the end-windings),
and minimum at the middle of the machine length; this minimum is practically null.
 in position 2 (Position 2), the coil measures the normal component of the radial field.
Taking the coil sensor away from the motor end-shields increases the sensitivity to
the air gap flux density.
 in position 3 (Position 3), the coil is used to measure the external field called "pure
radial" sensitive to the tangential component. In theory, field lines generated by the
end-windings cannot cross the section of the coil sensor in this position.
RADIAL FIELD AXIAL FIELD
Pos.1 Pos.1
Pos.3
Pos.2
Pos.2
Figure 3. Measure of leakage field in near vicinity of the machine.
2.2. Measurement Sensors
Several types of sensors can be used to measure the external field (magnetoresistive
sensor, Hall effect sensor, coil sensor ...).
In the diagnosis method presented in this chapter, a coil sensor is used. The principle of
the coil sensor is to measure the flux linkage in a coil located in the vicinity of the electrical
machine [23]. The sensor provides an emf equal to the derivative of the magnetic field with
respect to the time. Therefore, the measurement is only possible for variable fields. With this
type of sensor, the measurement is simple and direct because it does not require associated
electronics, unless it is necessary to amplify the induced emf signal [21]. The performance of
the measurement depends on the size of the sensor. The main advantage of these sensors is
their simple design which allows users to build it in accordance to their needs [24]. The
sensor used for experimental tests and its frequency response are represented in Figure 4.
This small sensor will be used for all experimental measurements presented in this work.
It has been built for small and low power machines. The coil diameter is about 3.2 cm and it
has 1200 turns. Its frequency response shows a resonance at 93 kHz, so, the sensor is suitable
to be used for measurement of the field in a frequency range lower than this frequency, what
actually corresponds to the frequency range exploited by the proposed diagnosis methods.
This type of sensor is used for flux measurements because it is simple, non-invasive, low
cost, and it induces an emf with a natural derivative effect that amplifies the low and medium
frequencies [25].
1200
coils
32 mm
(a) (b)
Figure 4. Coil sensor: (a) photo; (b) frequency answer.
3. STATOR FAULTS IN AC ELECTRICAL ROTATING MACHINES

The most of stator faults are caused by the degradation of winding insulation. They
manifest themselves as a short-circuit between turns, by a short circuit between two phases or
by a short-circuit between a phase and the ground. In this section the analytical model is
presented for the short-circuit between turns of the same phase. This is a common defect that
can occur either at the coil ends or in the notches, causing a decrease of effective winding
turns. On the other hand, this fault leads to increase the stator current in the faulty phase and
to a slight variation of the amplitude in the others. In the case of induction machines, the
inter-turn short-circuit fault also amplifies the current in the rotor circuit [26, 27]. Thermal
stress induced by short-circuit can cause the spread of the fault to other coils and can lead to
the breakdown of the machine.
The analytical approach developed makes it possible to identify the properties of the air
gap flux density waves generated by the fault and proposes a diagnosis method based on
those flux density waves.
3.1. Analytical Model
For the analytical study of stator faults in AC motors, a simplified geometry of the
machine has been considered (Figure 5) to study the external magnetic field.
Flux coil sensor

Stator
x M
B x
B M
bn
bt  s   0s
ts
R
int 0 ds
Rotor
ts
Rext
Stator lamination Air gap
packet
Figure 5. Transverse section and considered parameters of simplified machine.
The point M‟ outside the machine has polar coordinates x, αs. The stator is composed of a
magnetic yoke and an external frame; the gap which can possibly exist between the stator
yoke and the external frame is neglected. To simplify, the considered machine presents
smooth air gap where the flux density is imposed. Figure 5 also gives the position of the
ts ts
sensor, placed at the level of the point M. Here the Rint and Rint are internal and external
radii of the stator.
For short-circuit model let us consider p  2 pole pair, three-phase stator winding
machine given in Figure 6 [25].
ds coil  n s turns  elementary coil  n s / ms turns 
s
Figure 6. Three phase stator windings  p  2 .

Each coil is made up of m s elementary coils of n s / ms turns, where n s is the machine
per phase, per pole pair turn number. The coils of the same phase are series connected. As a
s
result the per phase turn number is pn .
Considering the phase q of the machine, the proposed model supposes that an elementary
coil with y turns among n s / ms is short-circuited. In this model, it is considered that the
s
currents are unchanged when y is small compared to pn . So the model which implements
the inter-turn short circuit can preserve the healthy structure and the faulty stator winding are
composed by the healthy winding and y short-circuited turns. A iqfs fictitious current
circulates in this coil (denoted “additional coil”) which is composed of inter-turn short circuit
s
induced current isc and the current in opposite direction iqs :
iqfs  isc
s
 iqs (1)
A damaged elementary coil is shown in Figure 7. Here, the model preserves the initial
structure and simplifies the fault analysis by separating the fault effects and those created by
the healthy coils. For this case, the resulting air-gap flux density b can be expressed by:
b  b  bqsc
s
(2)
where b   bK ,H , bK ,H  bˆK ,H cos  Kt  H s  K ,H  (K is the frequency rank and H

K ,H
s
the polarity of bK , H ) and bqsc is obtained as consequence of the damaged coil.
n s / m sn‟s
turns yyturns
coils coils
ii'qfsq
s s
i qsq
Figure 7. Elementary coil with y short-circuited turns
s
Taking in consideration the y short-circuited turns, bqsc is obtained by connecting in
series the  qel
s
mmf generated by two healthy coils of the same phase and the  qsc
s
mmf
generated by the additional coil and shifted of  as presented in Figure 8(a). The mmf
generated by the faulty turns is shown in Figure 8 (b).
In this approach, the additional coil generates a unidirectional wave  qsc
s
compared with
 qel
s
which is a rotational one. Its polarity H is equal to 1 and creates a dissymmetry in the
stator mmf. The rotational mmf waves associated to the unidirectional one evolve per pair in
opposite directions. Therefore,  qsc
s
can be written in the stator reference frame d s tied to the
stator phase 1 axis:
s
n s iq  qel
s
ms 2
2
0
s s
n iq s
 s
m 2
(a)
 qsc
s
yiqfs 4 2
0
s
3 yi s
qf 4
(b)
Figure 8. The mmf components generated by elementary damaged coil created by: (a) two healthy
elementary sections; (b) y short-circuited turns
 qsc
s
h

 I sf  Ahs cos  t  h s  h  (3)
where I sf is the rms value of iqfs ,  is the angular frequency, h is a not null relative integer,
which takes among others, all the values of h s expressed by: h s  6k  1 ( k is an integer
which varies between  to  ). The winding coefficient tied to the rank h s is taken in
consideration by function Ahs . The  h is the phase lag which takes into account the place
where the short-circuit occurs and the position of the elementary coil concerned by the short-
circuit.
s
The flux density bqsc generated by the fault is obtained by multiplying  qsc
s
with air-gap
permeance  expressed by:
 
     
 k s k r cos  k s N s  k r N r p s  pk r N r 

(4)
k s  k r 
where  k s k r is a coefficient which depends on the ranks k s and k r and the air gap geometry,
and N s is the stator per pole pair tooth numbers.
3.2. Model for Salient-Pole Synchronous Machine
For a healthy salient-pole synchronous machine, K and H are defined as:
K  1  2k r 

r 
 
(5)
H  p h  k N  2k 
sr s s

where h sr includes all the values taken by hs and hr : h sr  hs hr and

h  6k  1; h  1,3,5,
s r
,   .
s
If a fault exists, the flux density bqsc can be expressed as:
s
bqsc  
K sc , H sc
bˆscK
sc , H sc

cos K sct  H sc s  scK
sc , H sc
 (6)
with
K sc  1  2k r 

 (7)
H sc  h  pk  N  2 pk  
s s r

k s and k r are equivalent to k s and k r . Consequently they vary from  to  .

After attenuation, it is possible to measure, near the machine, the resulting flux density
which traverses the frame and which generates in the signal given by the sensor, the
harmonics at K and K sc angular frequencies. The analysis of K and K sc values
expressed by (5) and (7), allows the following remarks:
 in failure presence, no new frequencies are brought by K sc . Only the variation of the
existing lines amplitudes will be visible.
 The spectral lines with higher sensitivity in presence of the fault correspond to
harmonics obtained for k r  1 , (  and 3 ). One of these low frequencies is
confused with the fundamental [12].
 the spectral components at angular frequencies 3 , 5 , 7 , which can exist for a
healthy machine(supply harmonic, saturation) are superposed to other sensitive lines
obtained for k  2, k  3,
r r
Considering all those remarks, it is difficult to achieve the analysis of the fault by
measuring the harmonics amplitude variation. So the following model takes into
consideration the dissymmetry generated by the fault and exploits the properties generated by
this dissymmetry.
3.3. Model for Induction Machine
Under healthy conditions the quantities K and H in the case of induction machine can
be written in the following form:
K  1  k r N r 1  s 


r r 
 
(8)
H  p h k N k N 
s s s

The sensitive lines to the fault are dependent by the parameters K sc and H sc which can
be expressed by:

K sc  1  k r N r 1  s  , H sc  h  p k s N s  k r N r  (9)
A numerical application for induction machine is presented in Appendix. The results

obtained by compilation show that each harmonic of the external field, at given k r is
essentially created by components with polarities the lowest as possible. The analysis will
focus on the diagnosis harmonic at 850 Hz, because it is generated by a component with the
lowest polarity both for the healthy machine and for the faulty machine  H  2, H cc  1 .
As the numerical application (see Appendix) shows this harmonic has a high amplitude value
and can be more easily measured outside of the motor frame.
3.4. Principle of the Tests
For the new diagnosis method, which exploits the change of external magnetic field in
presence of the faults, the measurements are carried out in two diametrically opposed
positions: 1  0 (Position 1) and  2  0   (Position 2) in Figure 9.
s s
The new principle consists in measuring at the same time and in both positions of the
machine, the variation of specific harmonic components that appear in presence of the faults.
Practically, taking into consideration the attenuation of the air-gap flux density in the stator
core, the method will be concentrated on the harmonics defined by the lowest values of the
parameters h , h , k , k , k  , k  . The objective is to obtain a value of H smallest as
s r s r s r
possible.
1 s  2s 1s
Figure 9. Sensors position for measurement of external magnetic field.

In this context, the spectral line with magnitude b̂1,1 at angular frequency   K  1 has
to be considered. However, for this component, bˆ1,1  bˆ1, H (with H different from 1) and
generates an important effect because it corresponds to the fundamental of the air-gap flux
density. So for the components obtained for H  1, the sensitivity of b̂1 is lower.
Therefore, in the analysis of salient-pole synchronous machine, the study will focus on
the spectral lines corresponding to K  3 , ( harmonic at 150 Hz). In the case of a four pole
machine  p  2 , the harmonics are obtained considering the flux density components given
by:
for healthy case: h  1, k  0 , k  1, h  1 in (5), which leads to K  3 and

r s r sr

H  3, p  6 ;
s r
 for faulty case: h  3, k  0, k  1 in (7), which gives K  K sc  3 , H sc  1 .
x
The flux density component b3 measured in the machine vicinity is obtained taking into
consideration a term relative to the faulty case (of amplitude bˆsc3 ) and a component with 3
x
angular frequency relative to the healthy case. In Position 1, the flux density component is
given by:
  
b3*x(1)  bˆ 3x cos 3 t  60  3x  bˆscx 3 cos 3 t  0  scx 3  (10)
In Position 2, it becomes:
  
b3*x(2)  bˆ 3x cos 3 t  60  3x  bˆscx 3 cos 3 t  0  scx 3  (11)
The difference between the flux density measured by the sensors in the position 1 and
position 2 is the sign of the component ( bˆ sc
x
3 ) corresponding to the short –circuit case. In the
healthy case this component is null ( bˆ scx 3  0 ) and theoretically the both components
expressed by (10) and (11) have the same amplitude. This one changes in faulty case because
the component bˆ scx 3 is added to the healthy component in Pos. 1 and subtracted in Position 2.
Practically the amplitudes of the components measured in Position 1 and Position 2 are
different even in the healthy machine and that can be the consequence of:
 a non symmetrical alignment of the sensors in Position 1 and Position 2

 a different attenuation of the signal for the both sensors as consequence of the
building techniques which are not perfect leading to a geometrical dissymmetry.
In order to solve this difficulty a test in load operating condition has been achieved. The
application of this test can be refined as follows:
 for healthy machine, practically the air-gap flux density does not change in the load
case ( when the stator current increases) if the supply voltage is maintained constant.
So, in the load case, the elements which influence the amplitudes bˆ x (1) and bˆ x (2) act
3 3
in order to influence the evolution of the signals measured in Position 1 and Position
2 in the same direction.
 for faulty machine, the test in load condition gives a variation of the component
(corresponding to angular frequency) measured in Position 1 in opposite direction
that those measured in Position 2, according to (10) and (11). Considering this
different evolution of the sensitive harmonic in the both positions, the originality of
the method is to use it as an indicator of the fault. It can be noticed here the interest
in using the load tests to determine the lines variation and so, to eliminate any initial
knowledge of a healthy state.
In the case of induction machine with Nr=32, the sensitive harmonic which can be easily
highlighted is to one at 850 Hz. This line corresponds to a flux density component having a
low pole pair number
for healthy case: k  1, k  1, h  7 lead to K  17, H  2 in (8)

r s s

for faulty case: k   1, k   1, h  15 lead to Ksc  17, H sc  1 in (9).
r s s

Expression similar to (10) and (11) can be obtained, considering K  K sc  17 (instead

of 3) and H  2 (instead of 6).
4. ROTOR FAULTS IN AC ELECTRICAL ROTATING MACHINES

This section briefly presents the main phenomena characterizing the rotor faults. It will
focus on the influence of a broken bar on the change of the external magnetic field of an
induction machine.
An analytical approach requires some simplifying assumptions but it will allow (one) to
give the specific harmonics of the fault, including their frequency and polarity. The analytical
study will highlight the need to consider the axial field as part of stray flux for which a
simplified model will be proposed. A specific signature will be presented, visible in the
spectrum of the external magnetic field.
4.1. Types of the Rotor Faults
4.1.1. Bearings Faults

The majority of faults in electrical machines are relative to bearing defects that have
mechanical (lubrication, fatigue, etc.) or electrical ones (circulation of leakage currents
induced by the converters [28]) causes. Bearing faults lead to more general mechanical effects
such as noise and vibration increase or by the appearance of movements of the rotor about the
longitudinal axis of the machine [29]. This type of faults induces also variations (oscillations)
of the load torque. The ultimate point of the bearing defect is the rotor blocking.
4.1.2. Eccentricity
The eccentricity of an electrical machine is a phenomenon that evolves over time and
exists from its manufacture. It passes through various stages of building and assembly
tolerances, which induces an off-centering of the rotor relatively to the stator. During
operation of the machine, two main causes increase rotor eccentricity. The first is inherent to
the mechanical environment in which the machine operates and can impose a radial force on
the shaft of the machine. This can lead to bearing wear and amplification of the eccentricity.
The second phenomenon that may amplify the eccentricity is itself inherent to the operation
of the machine. Actually, the rotor off-centering creates an unbalance in the distribution of
radial forces between the stator and the rotor [30, 31] that tends to increase the eccentricity.
The ultimate point of eccentricity is the friction between stator and rotor, that leads to a fast
destruction of the machine.
4.1.3. Broken Bar

Inter-turn short circuit fault may also occur in a wound rotor. For squirrel cage induction
machines, broken rotor bars and the breaking of short-circuit rings are the most common rotor
failures (Figure 10).
The broken bar or the breaking of the short-circuit ring may be due, for example, to
mechanical overload (frequent starts, etc.), to excessive local heating or to a manufacturing
defect (air bubble or bad welds). This generates oscillations of currents and electromagnetic
torque even more apparent when the inertia of the system has a high value (constant velocity)
[32, 33]. If the system inertia is lower, oscillations appear on the angular speed and the
amplitude of the stator currents. The occurrence of a broken bar does not cause an immediate
breakage of the machine, because the current crossing the broken bar flows over the adjacent
bars. Therefore, these bars are overloaded and the thermal and electromechanical constraints
can lead to their breaking and so on until the breakage of a sufficient large number of bars
that causes the shut-down of the machine.
Short - Short -
circuiting circuiting
ring ring
Rotor bars Rotor bars

(a) (b)
Figure 10. Fault of squirrel rotor cage: (a) broken bar; (b) breaking of the short-circuiting ring.
4.2. Rotor Broken Bar
4.2.1. Influence on the Speed Variation

The presentation of the phenomenon will consider only the flux density component of p
pole pairs. Under healthy operating conditions, the balanced three-phase current system
creates a fundamental clockwise rotating magnetic field in the air-gap. This field induces a
current in the rotor bars with a frequency proportional to the rotor slip s . The rotor bars then
generate a clockwise field rotating at s s in a rotor referential (Figure 11(a)). For a
symmetrical motor (healthy machine), there is not any anticlockwise field in the air-gap [8,
34].
Under rotor faults, in addition to the clockwise field an anticlockwise field appears,
created by the unbalance of rotor bar currents and which rotates at an angular velocity  ss
shown in Figure 11(b). It is possible to associate with this anticlockwise field a negative-
sequence current system with amplitude I rn , which is added to the positive-sequence current
system with amplitude I rp .
(a) (b)
Figure 11. Operation schema for: (a) healthy motor; (b) faulty motor.
Figure 11(b) shows the phenomenon due to the appearance of rotor failure. A flux density
component rotating at  fault appears in the stator referential:
 fault  r  ss  s 1  2s  (12)

f fault
which leads to the well-known frequency in literature , appearing in the radial external
magnetic field and in the line current:
f fault  1  2s  f (13)
The interaction of the rotor currents with the clockwise stator field produces the
electromagnetic torque. In case of broken rotor bar fault, the total torque can be expressed as
follows:
Ce _ tot  Ce _ srp  Ce _ srn  

RLem
0
   
ks I s kr I rp sin Bs  Brp  kr I rn sin 2st  Bs  Brn 
 
(14)
where Ce _ srp represents the mean torque due to the interaction of the stator and the rotor
fields in the clockwise direction, Ce _ srn is the pulsating torque of pulsation 2s due to the
interaction of the stator clockwise field with the anticlockwise rotor field;  Bs ,  Bs and  Brn
are respectively the phases relative to the stator phase one current of the stator flux density,
the rotor flux densities in the clockwise direction and in the anticlockwise direction; k s is a
coefficient that depends on the stator winding arrangement and kr is a coefficient that
depends on the rotor cage constitution.
The torque oscillation at frequency 2sf , also presented in [8], generates a speed
oscillation at this frequency whose amplitude depends on the rotor asymmetry, the motor load
level [34, 35], and the total moment of inertia J . The speed is calculated with the relation:
d r
J  Ce _ tot  Cr  Ce _ srn (15)
dt
where Cr is the load torque:
1
J
r  Ce _ srn dt  0 (16)
By calculating the integral (11) one obtains:
r  0  r  cos(2st   ) (17)
1 RLem
where r   ks Ikrn I rn  is the magnitude of the speed oscillation, 0 is the
2sJ 0
mean value of  r which changes, depending on the number of broken bars at given load
torque, and   Bs  Brn .
It is well known that the speed oscillation also generates an additional air-gap flux
density component at 1  2s  f .
4.2.2. Influence on the Axial Flux

For the modelling of the axial field, it is necessary to consider the field lines existing at
the machine extremities. The distribution of these ones strongly depends on the geometry and
the mechanical components that exist at the machine extremities. As mentioned in [32], the
total axial flux  AT outside the induction machine with an eccentricity fault can be written
in the following form:
ˆ cos t  p 
 AT  
(18)
with   0  r t ,  indicates the position of the minimum air-gap thickness. Under

broken rotor bar fault, a speed variation appears and will influence the external axial flux. In
this case (18) can be written as follows:
ˆ cos t  p     t 
 AT   (19)
 0 r 
Developing the argument of the cosine, using (17) one obtains:
ˆ cos  st   pt cos(2st   )  p 

 AT   r 0 (20)
A frequency modulation appears in  AT . The time variation of (20) for  ˆ  1 Wb ,

r  1 rad/s and s  2.5% is presented in Figure 12(a). A direct analysis of the signal is
difficult; hence the change in the frequency domain is performed by Fast Fourier Transform
(FFT). The spectrum obtained is given in Figure 12(b).
Figure 12. Simulation of axial flux variation: (a) axial flux signal; (b) FFT for the axial flux.
In the spectrum it can be observed the appearance of the harmonic 3sf caused by the
speed variation and sf due to dynamic eccentricity on the axial flux. One can notice that a
small speed variation can produce an important component at frequency 3sf . With the
considered speed variation that corresponds approximately to 2 broken bars; the component
3sf is higher than the initial component sf .
5. EXPERIMENTAL TESTS
5.1. Achievements of the Tests
The application of the proposed method for stator fault detection is presented considering
a salient-pole synchronous machine and an induction machine. For rotor faults an
experimental bench using two identically squirrel-cage induction motors (one healthy and
another with a broken bar) is used. The principle of measurement is shown in Figure 13.
Figure 13. Measurement and treatment of emf signal.
5.2. Stator Tests
5.2.1. Inter-Turn Short Circuit in Synchronous Machines

The application of the proposed method is presented considering a test carried out
on a salient-pole synchronous machine characterized by: 7.5 kW, 50 Hz, p  2 , 230/400V,
N5=18, grid connected. This machine has been rewound so that the terminals of the different
stator elementary sections are extracted from the winding and are brought back to a connector
block which is fixed above the machine as indicated in Figure 14. So it is possible to produce
short-circuits between elementary coils.
The tests consist in measuring and analyse the emf delivered by the sensors in order to
validate the suggested procedure when an elementary coil is short-circuited (what
corresponds to 16.6% of the whole winding of a phase). In order to avoid total damage of
the winding, the short-circuit current is limited by an external resistance R. Verification has
also been done in order to make sure that the currents in the wires going up from the machine
towards the connector block do not disturb the measurements. The Position 1 and Position 2
correspond respectively to  0 and 0   ,  0 is chosen in the axis of the tested phase. The
interest is focused on the harmonic at 150 Hz. Figures 15 a and b show the variations of the
harmonic at 150 Hz relative to the healthy machine (motor) when the state (marked with
arrow) varies from no-load to load.
Figure 14. Modified salient-pole synchronous machine and the sensors‟ positions.
Figure 15. Harmonic components at 150 Hz for a healthy salient-pole synchronous machine (motor) at
no-load and load operations.
It can be observed that the magnitudes vary in the same direction (decrease) for the two
positions. Under faulty conditions at full load (Figure 16) it can be observed that the
magnitudes in both positions vary in the opposite direction. This indicates the failure presence
and confirms the proposed method.
Other tests are carried out on the machine which operates as generator. The
corresponding results are shown in Figures 17 and 18. One can notice the same phenomenon:
under healthy case the magnitudes varies in the same direction, but in the faulty case they
vary in opposite direction.
Figure 16. Harmonic components at 150 Hz for faulty salient-pole synchronous machine (motor) at no-
load and load operations.
Figure 17. Harmonic components at 150 Hz for a healthy salient-pole synchronous machine (generator)
at no-load and load operations.
Figure 18. Harmonic components at 150 Hz for faulty salient-pole synchronous machine (generator) at
no-load and load operations.
Figure 19. Harmonic components at 150 Hz for a healthy permanent magnet synchronous motor at no-
Figure 20. Harmonic components at 150 Hz for a healthy permanent magnet synchronous motor at no-
The next test has been performed on a four pole, 4.4 kW, 50 Hz, 3×380 V permanent
magnet synchronous motor. The results obtained under healthy and faulty conditions are
given in Figures 19 and 20. It can be observed that under healthy case the harmonic
magnitudes at 150 Hz vary in the same direction, but when the inter-turn short circuit exists in
the stator winding these harmonic magnitudes vary in different direction.
5.2.2. Inter-Turn Short Circuit in Induction Machines

In order to generalize the proposed method, the tests have been also performed on a
p  2, 50 Hz, 11 kW , 380 / 660V, N s  24 and N r  16 squirrel cage induction motor
(Figure 21).
This machine, connected to the grid, has been rewound so that the terminals of the
different stator elementary sections are extracted from the winding and are brought back to a
connector block which is fixed above the machine. So it is possible to produce short-circuits
between elementary coils. Position 1 and Position 2 correspond respectively to  0 and
0   ,  0 is chosen “arbitrarily”. One will be interested to the spectral line at 850Hz
k r

 1 . The results obtained by considering the harmonic of rank K  K sc  17 ( 850Hz
at no-load and load), are presented in Figures 22 and 23.
Figure 21. Modified squirrel cage induction motor and sensors‟ positions
The phenomena are similar than those obtained in the case of synchronous machines:
 under healthy condition - the lines relative to K  K sc  17 evolve in the same way
with the load (Figure 22),
 under faulty condition - these lines evolve in opposite way with the load (Figure 23).
Figure 22. Harmonic components relative to K  K sc  17 for healthy induction motor at no-load and
load operations.
Figures 23. Harmonic components relative to K  K sc  17 for faulty induction motor at no-load and
load operations.
This method is thus applicable to salient-pole synchronous machines, permanent magnet

synchronous motors and induction motors often used in electrical traction applications and
that confirm the universal quality of the proposed diagnosis method.
Figure 24. Frequency spectrum of the external magnetic field (Position 1) for healthy squirrel cage
induction motor (s = 5%).
Figure 25. Frequency spectrum of the external magnetic field (Position 1) squirrel cage induction motor
with one broken rotor bar (s = 5%).
5.3. Broken Rotor Bar in Induction Machine
The tests are performed on a 4 poles, 4 kW / 50 Hz, 3×380 V squirrel cage induction
motor under healthy and faulty (broken rotor bar) conditions. The results obtained at full load
are shown in Figures 24 and 25.
If the rotor has a broken bar, the frequency component 3sf is more sensitive than other
components and it denotes the speed variation at frequency 2sf , caused by the fault. Figure
25 shows that the amplitude of this harmonic is higher in comparison with the component sf ,
which is an indicator of eccentricity.
Also, one can observe in the case of healthy machine, the presence of frequency
components sf and 3sf with low amplitude. It is due the dissymmetry that exists even for a
healthy machine, inherent to the building tolerances.
CONCLUSION
The chapter presents a simple method to insure the supervision of AC machines used in
electrical traction. This supervision can detect stator and rotor faults using original signatures
presented in the chapter. Under stator winding faults the signature concerns the harmonic at
150 Hz for synchronous machines and a slotting harmonic for induction machines. This
diagnosis method does not require the knowledge of the healthy signature. It is based on
comparison between the states at no-load and load operations.
The indicators of rotor faults are the low frequency components of the flux spectrum at
sf and 3sf frequencies. The diagnosis method consists in checking the magnitude of the
component 3sf generated by the speed variation that can be compared to this of the
component sf .
The advantages of the proposed non-invasive methods compared to other existing
methods are their simplicity to implement, to interpret and a low cost set-up. The processed
emf signal, delivered by a simple coil sensor, is able to indicate the presence of the fault.
Moreover, the low frequency analysis is not disturbed by other sources of magnetic field
presented in vicinity. Finally, the experimental tests realized on the different AC machines
confirm the validation of used models and proposed methods for stator and rotor fault
detection. The simulation and experimental analyses confirm the models and approaches
used.
APPENDIX
For a numerical application it is considered an induction machine with p  2 , N s  24 ,

N r  16 , Rint  0.0825  , Rext  0.121  and a short circuit in the phase 1. Here 12.5%
ts ts
of the phase coils are short-circuited and the short-circuit current is isc  3i . The values
s s
that can take the coefficients that contribute to the description of predominant
components |are given by k r   ;    ; k s   ;    ; k r   ;    ; h s  6k  1;
hr  1,3,5, ,   .
If one consider the first rotor slotting harmonic obtained for k r  1 , for a healthy
induction machine with a slip null and considering k r  1 , it comes k r N r  16 and
K  15 , f  750 Hz .
In order to have k s N s  k r N r the lowest possible (here equal to -6), we must choose
k s  1 . Also, we find that the dominant mode is obtained for h s  5 , H  6 (in absolute
terms and because p  2 ). For k r  1, we obtain k r N r  16 and K  17 , f  850 Hz .
As k s N s  k r N r must be the lowest possible (here equal to -2), we must have k s  1 .
Also, we find that the dominant mode is obtained for h s  7 , H  2 (in absolute terms).
Table 1. Components generated by the healthy IM
ks hs K H K (Hz)  Hts bˆKx , H / bˆ11x

kr
-1 1 -5 -15 6 -750 5.09e-05 2.99e-05
-1 1 -11 -15 -6 -750 5.09e-05 8.16e-06
1 -1 7 17 -2 850 8.16e-04 2.62e-04
1 -1 13 17 10 850 3.97e-06 5.32e-07
-2 1 7 -31 -2 -1550 8.16e-04 -2.49e-04
-2 2 -17 -31 -2 -1550 8.16e-04 5.66e-05
2 -1 -5 33 6 1650 5.09e-05 -2.85e-05
2 -1 -11 33 -6 1650 5.09e-05 -7.76e-06
2 -2 13 33 -6 1650 5.09e-05 3.81e-06
2 -2 19 33 6 1650 5.09e-05 4.04e-06
Table 2. Components generated by the faulty coils
k r k s h K cc H cc Kcc (Hz)  Hts cc bˆccx / bˆ11

x
Kcc , H cc
-1 1 -10 -15 6 -750 5.09e-05 1.82e-05

-1 1 -13 -15 3 -750 3.76e-04 -7.31e-05
-1 1 -14 -15 2 -750 8.16e-04 -2.07e-04
-1 1 -15 -15 1 -750 0.0023 -3.78e-04
-1 1 -18 -15 -2 -750 8.16e-04 1.60e-04
-1 1 -19 -15 -3 -750 3.76e-04 4.94e-05
-1 1 -21 -15 -5 -750 9.72e-05 -1.15e-05
-1 1 -22 -15 -6 -750 5.09e-05 -8.09e-06
1 -1 10 17 -6 850 5.09e-05 1.82e-05
1 -1 11 17 -5 850 9.72e-05 2.23e-05
1 -1 14 17 -2 850 8.16e-04 -2.07e-04
1 -1 15 17 -1 850 0.0023 -3.78e-04
1 -1 17 17 1 850 0.0023 3.33e-04
1 -1 18 17 2 850 8.16e-04 1.60e-04
1 -1 21 17 5 850 9.72e-05 -1.15e-05
1 -1 22 17 6 850 5.09e-05 -8.09e-06
-2 1 17 -31 1 -1550 0.0023 -3.16e-04
-2 1 21 -31 5 -1550 9.72e-05 1.09e-05
-2 1 22 -31 6 -1550 5.09e-05 7.70e-06
-2 2 -26 -31 6 -1550 5.09e-05 3.78e-06
2 -1 -10 33 6 1650 5.09e-05 -1.73e-05
2 -1 -13 33 3 1650 3.76e-04 6.96e-05
2 -1 -14 33 2 1650 8.16e-04 1.97e-04
2 -1 -15 33 1 1650 0.0023 3.60e-04
2 -1 -18 33 -2 1650 8.16e-04 -1.52e-04
2 -2 26 33 -6 1650 5.09e-05 3.78e-06
The other components defined for values k , k , k  , k  , h , h of the lowest

s r s r s r
possible for the healthy machine and for the faulty coils are given respectively in Table 1 and
Table 2 with the amplitude in x  Rext given relatively to the fundamental.
s
REFERENCES
[1] Pusca, R.; Contribution to the Study of the Driving of a Mobile Platform to Test
Electrical Power Train Components - Design of a Simulator and Real Time System.
Doctoral Thesis, Science for Engineering, University of Franche-Compte, Belfort,
France, 2002.
[2] Bellini, A.; Filippetti, F; Tassoni, C; Capolino, G.A.; Advances in Diagnostic
Techniques for Induction Machines. IEEE Trans. on Industrial Electronics, vol. 55, no.
12, pp. 4109-4125, 2008.
[3] Thomson, W.T.; Fenger, M.; Current Signature Analysis to Detect Induction Motor
Faults. IEEE Industry Applications Magazine, vol. 7, no. 4, pp. 26-34, 2001.
[4] Kliman, G.B.; Koegl, R.A.; Stein, J.; Endicott, R.D.; Madden, M.W.; Noninvasive
Detection of Broken Rotor Bars in Operating Induction Motors. IEEE Trans. on Energy
Conversion, vol. 3, no. 4, pp. 873-879, 1988.
[5] Kliman, G.B.; Stein, J.; Methods of Motor Current Signature Analysis. Electrical
Power Component System, vol. 20, no. 5, pp. 463-474, 1992.
[6] Khezzar, A.; Kaikaa, M.Y.; El-Kamel Oumaamar, M.; Boucherma, M.; Razik, H.; On
the Use of Slot Harmonics as a Potential Indicator of Rotor Bar Breakage in the
Induction Machine. IEEE Trans. on Industrial Electronics, vol. 56, no. 11, pp. 4592-
4605, 2009.
[7] Romary, R.; Corton, R.; Thailly, D.; Brudny, J.F.; Induction Machine Fault Diagnosis
Using an External Radial Flux Sensor. EPJ. Applied Physics, vol. 32, no. 2, pp. 125-
132, 2005.
[8] Filippetti, F.; Franceschini, G.; Tassoni, C.; Vas, P.; AI Techniques in Induction
Machines Diagnosis Including the Speed Ripple Effect. IEEE Trans. on Industry
[9] Arthur N.; Penman, J.; Induction Machine Condition Monitoring with Higher Order
Spectra. IEEE Trans. on Industrial Electronics, vol. 47, no. 5, pp. 1031-1041, 2000.
[10] Menacer, A.; Nat-Said, M.S.; Benakcha, A.H.; Drid S.; Stator Current Analysis of
Incipient Fault Into Induction Machine Rotor Bars. Journal of Electrical Engineering,
vol. 55, pp. 122-130, 2004.
[11] Li, W.D.; Mechefske, C.K.; Detection of Induction Motor Faults: A Comparison of
Stator Current, Vibration and Acoustic Methods. Journal of Vibration and Control, vol.
12, pp. 165-188, 2006.
[12] Thailly, D.; Romary, R.; Lecointe, J.Ph.; Brudny, J.F.; Suau, P.; Synchronous Machine
Diagnosis Using an External Magnetic Flux Sensor.17th International Conference on
Electrical Machines (ICEM 2006),Chania, Crete, Greece, pp. 462-468, 2006.
[13] Yazidi, A.; Henao, H.; Capolino, G.A; Artioli, M.; Filippetti, F.; Casadei, D.; Flux
Signature Analysis: an Alternative Method for the Fault Diagnosis of Induction
Machines. IEEE PowerTech Conference, St. Petersburg, Russia, pp. 1-6, 2005.
[14] Ceban, A.; Pusca, R.; Romary, R.; Eccentricity and Broken Rotor Bars Faults - Effects
on the External Axial Field. International Conference on Electrical Machines (ICEM
2010), Rome, Italy, pp. 1-6, 2010.
[15] Tavner, P.; Ran, L.; Penman, J.; Sedding, H.; Condition Monitoring of Rotating
Electrical Machines. 2nd Ed., Stevenage, IET, UK, 2008.
[16] Henao, H.; Demian, C.; Capolino, G.A.; A Frequency-Domain Detection of Stator
Winding Faults in Induction Machines Using an External Flux Sensor. IEEE
Transactions on Industry Applications, vol. 39, no. 10, pp. 1272-1279, 2003.
[17] Kokko, V.; Condition Monitoring of Squirrel-Cage Motors by Axial Magnetic Flux
Measurements. Doctoral Thesis, Department of Electrical Engineering, University of
Oulu, Oulu, Finland, 2003.
[18] Negrea, M.D.; Electromagnetic Flux Monitoring for Detecting Faults in Electrical
Machines. Doctoral Thesis, Electrical and Communications Engineering, Helsinki
University of Technology, Helsinki, Finland, 2006.
[19] Werynski, P.; Dielectric Aging and in Situ Monitoring of Electrical Machines. Doctoral
Thesis, Laboratory of Electrotechnical Systems and Environment (RSAA), Artois
University, Bethune, France, 2006.
[20] Zidat, F.; Designing a Monitoring Unit Performance Eco-Efficiency of Rotating
Electrical Machines with Alternating Current. Doctoral Thesis, Laboratory of
Electrotechnical Systems and Environment (RSAA), Artois University, Bethune,
France, 2012.
[21] Thailly, D.; Field Study of Dispersion of Alternating Current Machines: Application to
Diagnosis. Doctoral Thesis, Laboratory of Electrotechnical Systems and Environment
(RSAA), Artois University, Bethune, France, 2007.
[22] Thomson, W.T.; A Review of Online Condition Monitoring Techniques for Three-
Phase Squirrel Cage Induction Motors - Past, Present and Future. IEEE International
Symposium on Diagnostics for Electrical Machines, Power Electronics and Drives
(SDEMPED'99), Gijon, Spain, pp. 3-17, 1999.
[23] Tumanski, S.; Induction Coil Sensors - A Review. Measurement Science and
Technology, vol. 18, pp. 31-46, 2007.
[24] Ceban, A.; Pusca, R.; Romary, R.; Supervision of Electric Traction Motors by External
Magnetic Field Analysis. International Conference on Electronics, Computers and
Artificial Intelligence (ECAI 2011), Pitesti, Romania, pp. 1-10, 2011.
[25] Pusca, R.; Romary, R.; Ceban, A.; Brudny, J.F.; An Online Universal Diagnosis
Procedure Using Two ExternalFluxSensorsAppliedtoACElectrical Rotating Machines.
Sensors, vol. 10, pp. 10448-10466, 2011.
[26] Bonnett A.H.; SoukupG.C.; Cause and Analysis of Stator and Rotor Failures in Three-
Phase Squirrel Cage Induction Motors. IEEE Transactions on Industry Applications,
vol. 28, pp. 921-937, 1992.
[27] Nandi, S.; Toliyat, H.A.; Xiaodong L.; Condition Monitoring and Fault Diagnosis of
Electrical Motors: A Review. IEEE Transactions on Energy Conversion, vol. 20, pp.
719-729, 2005.
[28] Busse, D.F.; Erdman J.M.; Kerkman, R.J.; Schlegel, D.W.; Skibinski, G.L.; The Effects
of PWM Voltage Source Inverters on the Mechanical Performance of Rolling Bearings.
IEEE Transactions on Industry Applications, vol. 33, pp. 567-576, 1997.
[29] Vaseghi, B.; Contribution to the Study of Electrical Machines in the Presence of
between Failure Spires. Doctoral Thesis, INPL-GREEN University of Henri Poincare -
Nancy I, Nancy, France, 2009.
[30] DorrellD.G.; Smith A.C.; Calculation of UMP in Induction Motors with Series or
Parallel Winding Connections. IEEE Transactions on Energy Conversion, vol. 9, pp.
304-310, 1994.
[31] Dorrell, D.G.; Thomson, W.T.; Roach, S.; Analysis of Air Gap Flux, Current, and
Vibration Signals as a Function of the Combination of Static and Dynamic Air Gap
Eccentricity in 3-Phase Induction Motors. IEEE Transactions on Industry Applications,
vol. 33, pp. 24-34, 1997.
[32] Ceban, A.; Pusca, R.; Romary, R.; Study of Rotor Faults in Induction Motors Using
External Magnetic Field Analysis. IEEE Transactions on Industrial Electronics, vol.
593, pp. 2082-2092, 2001.
[33] Bossio, G.R.; De Angelo, C.H.; BossioJ.M.; Pezzani, C.M.; Garcia, G.O.; Separating
Broken Rotor Bars and Load Oscillations on IM Fault Diagnosis Through the
Instantaneous Active and Reactive Currents. IEEE Transactions on Industrial
Electronics, vol. 56, pp. 4571-4580, 2009.
[34] Didier, G.; Razik, H.; Rezzoug A.; An induction Motor Model Including the First Space
Harmonics for Broken Rotor Bar Diagnosis. European Transaction of Electrical
Power, vol. 15, pp. 229-243, 2005.
[35] Bellini, A.; Filippetti, F.; Franceschini, G.; Tassoni, C.; Quantitative Evaluation of
Induction Motor Broken Bars by Means of Electrical Signature Analysis. IEEE Trans.
on Industry Applications, vol. 37, no. 5, pp. 1248-1255, 2011.
Chapter 15
BATTERIES FOR ELECTRICAL VEHICLES:

A REVIEW
Bogdan Enache*1, Emilian Lefter †1 and Costin Cepisca‡2

1
Faculty of Electronics, Communications and Computers,
2
Faculty of Electrical Engineering, University Politehnica Bucharest,
Bucharest, Romania
ABSTRACT
Batteries represent one of the key elements of electric and hybrid vehicles and they
constitute an important factor in the development of the automotive industry in an
environment-friendly direction.
Nowadays, there is a wide range of batteries available on the market: lead-acid,
nickel, lithium, sodium, metal-air, etc., but none of these technologies managed to
overcome the gasoline in the terms of specific energy, specific power, density of energy
and power density.
The new developments in the field of nanotechnology and fuel cells aim at offering
solutions for the creation of batteries able to provide autonomy comparable to that
offered by gasoline, but a lot of the present results are only applicable to experimental
research.
Keywords: Batteries, Electric Vehicles, Hybrid Vehicles
†*
E-mail: b0gdan.enache@yahoo.com.
†
E-mail: emilianlefter@yahoo.com.
‡
E-mail: costin@wing.ro.
410 Bogdan Enache, Emilian Lefter and Costin Cepisca
1. INTRODUCTION
By the end of the 19th century, the electric vehicles (EV) seemed to have a bright future
in front of them. They represented the preferred choice of the society‟s elite due to the fact
that they were silent and they didn‟t require the starter‟s manual actuation at start-up. This can
be also noticed from the analysis of the USA sales statistics [1]: in 1899, 1575 electric
vehicles were sold as compared to only 936 internal combustion engine vehicles (ICE). The
performances of the two types of automobiles were comparable. Thus, in 1899 the Belgian
Jenatzy reached a record speed of 100 km/h with his electric automobile named «Jamais
Contente».
Within a few years, with all the improvements brought to it, the electric vehicle begins to
lose its place on the market. In 1910 the number of the sold electric automobiles increased
significantly, to almost 4000, but the sales of the internal combustion engine automobiles
exceeded to 120000 pieces, namely 30 times higher number [1]. This was due, especially, to
certain problems related to the vehicle‟s battery: a long recharging period, a reduced
reliability, a limited autonomy, a high cost price, etc.
One hundred years later, even though batteries technology made a considerable leap
forward, a great deal of the electric vehicles issues remained unsolved. But new elements
started to bring our attention back to them: the repeated petrol crises and the environment
pollution. The noxious emissions contributed to the deterioration of the environment quality,
to the development of the greenhouse effect, to climatic modifications, all leading to the
negative change of people‟s life conditions. Suzuki Foundation shows in its studies that diesel
engines are responsible for numerous diseases such as: asthma, pneumonia, other respiratory
diseases, cancer, heart diseases [2].
In this context of environmental care completed with the diminution of petrol reserves, a
series of organizations appeared, which have their sole purpose to reduce the automobiles
impact on the environment and on people‟s health. This is carried out by accelerating the
development process of batteries, for the electric and hybrid vehicles. Among these
organizations, the best known are: USABC and PNGV.
USABC is a consortium made up of the Department of Energy (DOE) of the United
States, Daimler-Chrysler, Ford, General Motors and the Electric Energy Research Institute of
the USA (EPRI). USABC determined the required performances, so that the batteries used
within EV will be compatible with ICE engines from the performance point of view. Among
these we can mention:
 High density energy, obtained by an single charge in order to provide increased

autonomy and high currents during the acceleration period;
 An increased life span with maintenance free;
 Nonpolluting materials in batteries structure.
The other mentioned organism, PNGV, was a research cooperative program in the USA,
initiated in 1993 and completed in 2003 and was composed by federal agencies, national
laboratories, universities and motor vehicles manufacturers. The program was completed with
the elaboration of three diesel-electric car prototypes: General Motors - Precept, Ford -
Batteries for Electrical Vehicles: A Review 411
Prodigy, Chrysler -ESX3, which complied with the imposed requirements of providing a
consumption of 3.78l/130 km/ (80 miles/gallon).
Both USABC and PNGV made up a series of tests, grouped in 2 manuals, which are used
for the batteries performances evaluation.
2. EXISTING ELECTRIC VEHICLES

Under the influence of the new 21st century technologies, the electric vehicles started to
appear again on the automobile market. At present, two categories of electric vehicles can be
found on the streets in the entire world: pure electric and hybrid electric.
2.1. Electric Vehicles
The Electric Vehicles (EV) use an electric engine and a battery as its main power
supply for the engine. These cars are also called zero-emission vehicles (ZEV).
Before use, these vehicles require the charging of the battery from an external source of
energy. This charging must be performed as fast as possible and it must provide high
autonomy. The corroboration of the two conditions leads to a series of difficulties because the
ensuring high autonomy implies a great energy density which cannot be obtained in a short
charging time. This is why the batteries meant for the ZEVs are permanently searching for the
right balance between the energy density and the charging time.
Transposing the problem numerically, the quantity of energy required for a medium-type
ZEV is of 32 kWh and for a 5 minutes charge the battery requires a power source which can
supply 384 kW, a power too big to be practically. Due to this reason, the medium charging
time, for this type of vehicle, is between 6-8 hours. A comparison between the different
existing technologies, in relation to a 250kg battery, is presented in Table 1.
Table 1. Performances for different batteries technologies [1]
Technology Pb-acid NiCd NiMH Li-Ion

Vehicle mass [kg] 1200 1200 1200 1200
Battery mass [kg] 250 250 250 250
Energy density [Wh/kg] 33 45 70 120
Supplied energy [kWh] 6.4 8.8 13.0 23.4
Power density [W/kg] 75 120 170 370
Battery supplied power [kW] 15 24 33 72
Autonomy for 120 Wh/ton/km [Km] 53 73 114 195
2.2. Hybrid Electric Vehicles
A hybrid electric vehicle uses an internal combustion engine and an electric engine, and
needs two different power sources, one for each engine. The goal of the hybrid vehicle is to
combine the autonomy of the internal combustion engine with the pollution-free and economy
of the electric one in a single vehicle [3].
The electric engine and implicitly the battery are used especially at start-up, but they
assist the internal combustion engine whenever it requires additional power. This operating
regime is characterized by high currents, and so the battery used must have a high power
density.
There are several hybrid cars architectures: Micro-Hybrids - “Stop & Start”, Soft-Hybrids
-“Stop & Go”, Mild Hybrids, Full Hybrids - Power Assist, Plug-in Hybrids -PHEV.
2.2.1. Micro-Hybrids - “Stop and Start”

The fundamental characteristic of these vehicles is that they maintain the structure of the
internal combustion engines unchanged. They usually append to the existing engine a starter-
alternator system that is powered from a battery-inverter device. The starter-alternator system
is only used at start-up.
At these vehicles the recovery of energy at braking doesn‟t take place. During idling the
internal combustion engine is stopped and only the starter system in powered up. This
operating mode produces a fuel economy of 15% [1].
The micro-hybrid vehicles are equipped with lead-acid batteries, usually called start-stop
batteries, which are maintained in 100% SOC.
2.2.2. Soft-Hybrids - “Stop and Go”

These vehicles borrow a lot from the micro-hybrids vehicles characteristics but, as
compared to these, they use regenerative braking. Within the braking process the electric car,
goes into a generator regime and the energy produced by it is transferred to the battery or to
certain ultra-cap.
The start-up of the soft-hybrids vehicles is performed in an electric mode and afterwards
the thermal engine is started, thus producing a maximum fuel economy of 25% [4].
The battery mounted on these vehicles is lead-acid or Ni-Cd because they have to endure
high current peaks. The operating mode of the battery is between 100% and 80% of SOC.
2.2.3. Mild Hybrids

The mild hybrids vehicles depart from the classic vehicles with internal combustion
engines and they propose a new topology characterized by the incorporation of the electric
engine in the drive.
The start-up for this type of vehicle is performed in the electric mode and this mode is
maintained until the speed of 20 km/h is reached, after which the thermal engine interferes.
While driving, the electric engine brings its contribution only during the acceleration periods,
thus reducing the fuel consumption.
The battery mounted on these vehicles is Ni-MH or Li-Ion because they have to endure
high operating currents and profound discharges.
2.2.4. Full Hybrids - Power Assist

These vehicles use simultaneously the electric engine and the internal combustion engine,
although they have the possibility, for a certain number of kilometers (approximately 30), to
operate only in electric mode. Due to this reason, the main characteristic of the batteries is
high energy density and the preferred choice of technology is Ni-MH.
The battery that equips these vehicles is either in a discharge process, or in a charging
one and it usually operates at 50% SOC in order to be able to endure the high currents at
discharge and the energy recovery at braking.
2.2.5. Plug-in Hybrids -PHEV

These vehicles combine the characteristics of the electric vehicles - their battery can be
charged from an external source and hybrid ones– the propulsion system has a thermal engine
and an electric one.
The vehicle can function in the electric mode until the battery reaches a certain discharge
level (around 30% SOC), which usually is translated into 50 km autonomy, afterwards the
thermal engine interferes. During the electric engine functioning period, in case a higher
speed is required, the thermal engine supplies the additional power.
The PHEV vehicles battery is charges from an external source when the vehicle is not
used, and from this reason, both the power density and the energy density are important
battery parameters.
The parameters that a battery should provide, in order to be used on one of the existing
electric vehicles, are synthesized in Table 2.
Table 2. Electric vehicles requirements [1]
Nr. Crt. Type Power [kW] Energy [kWh] Voltage [V]

1 Micro Hybrids 2.5-5 0.5 12-36
2 Soft Hybrids 3-6 0.75 12-48
3 Mild Hybrids 15-20 1 120-160
4 Full Hybrids 30-50 2-3 200-350
5 Plug-in Hybrids 30-100 5-15 200-350
6 Electric Vehicle 35-75 25-40 200-350
3. BATTERY TECHNOLOGY USED FOR ELECTRIC VEHICLES

The batteries used in electric vehicles are devices that transform chemical energy into
electric energy, at discharge, respectively the electric energy into chemical energy at charge.
A battery is made up of several interconnected cells, where a cell is an independent unit
that possesses all the chemical properties of the battery.
Usually, a cell is made up of: electrodes (anode, cathode), electrolyte, separators and
recipient. The functioning of a battery relies on the oxidation and reduction reactions that take
place at the anode and cathode, and on the flow of the electrons released by oxidation into the
battery exterior circuit [5].
If at the anode the active substance is A, the oxidation reaction will be:
A  Az   ze (1)
Namely, each molecule of substance A sends in the electrolyte the positive ions Az 
(cation radical), and z electrons remain on the electrode e , z being the cation radical valence.
At cathode, the active substance C accepts the electrons directed through the exterior
circuit (the reaction is called reduction), becoming a negative ion (anion):
C  ze  C z  (2)
The anions C z  sent in the electrolyte are combined with the cautions Az  brought in this
area due to the diffusion forces, producing the element AC:
Az   C z   AC (3)
As it can be seen, the process of electric energy generating can be resumed as follows:
pairs of charges with contrary names are generated at the anode: electrons and ions. The ions
are sent in the electrolyte, which is a sort of ionic diode, reaching the cathode. The electrons
can circulate only through the exterior circuit, if it is connected. Through the exterior circuit
the electrons reach the cathode where the described phenomena take place.
3.1. Lead-Acid Batteries
The lead-acid batteries are still the most used batteries in the automotive industry. They
are used especially in the vehicles with internal combustion engines having the role of electric
source during idle and cranking.
The power generation is based on the use of sponge lead as active material at the anode
and on the use of lead dioxide (PbO2) as active material at the cathode. The electrodes are
immersed in an electrolyte made up of a diluted solution of sulphuric acid.
At discharge, the sulphuric acid combines with the sponge lead and with the lead dioxide
thus forming the lead sulphate (PbSO4) and water, at charge the process is being reversed. As
the discharge process takes place, the electrolyte becomes more and more diluted. In
exchange, during charge the concentration of sulphuric acid in the electrolyte increases.
The total reaction is:
Pb + PbO2 + 2H2SO4  2PbSO4 + 2H 2O (4)
One of the most important characteristics of the lead-acid batteries is the very low
internal resistance, fact that makes the voltage loss (internal voltage drop) very small.
There are two important categories of lead-acid batteries: with maintenance and without
maintenance.
The lead acid batteries with maintenance have a liquid electrolyte and allow gas escape at
the end of charge, when water electrolysis takes place; because of this it requires a refill with
water in order to maintain the electrolyte concentration.
The lead-acid batteries with maintenance free (VRLA- Valve Regulated Lead-Acid),
allow water recombination at the end of the discharging process. The oxygen is released at the
cathode, passes through the electrolyte and it‟s recombined at the anode, so that the refill with
water is no longer required.
In order to be used in electric vehicles, the lead-acid batteries require certain

modifications:
 The lead plates (forming the electrodes) should be thicker;

 The quantity of active material deposited on these plates should be higher ;
 The cathode is covered with one or several layers of fiberglass in order to prevent the
active material from getting loose;
 The distance between the electrodes should be higher in order to facilitate the
interaction between the active material and the electrolyte.
At present, the lead-acid batteries are used in micro-hybrids vehicles. Citroen launched in
2005 the model C3 equipped with a 55Ah battery which managed to obtain a fuel economy of
approximately 10% in urban mode [3].
Table 3. Lead-acid battery characteristics [4]
Parameter Value
Specific energy [Wh/kg] 20-35
Energy density [Wh/l] 54-95
Specific power [W/kg] 250
Nominal voltage [V] 2
Internal impedance [Ω] 0.022
Lifespan [no. of cycles] 800
Self-discharging [%/day] 2
Temperature [°C] 10-55
Charging time [h] 8
3.2. Nickel Batteries
The nickel batteries were the first serious rival for the lead batteries offering, first of all, a
higher life span and also a better power and energy density.
There are four major technologies based on nickel: Ni-Fe, Ni-Zn, Ni-Cd and Ni-MH, out
of which, at present, only the last two are still in use.
3.2.1. Ni-Cd Batteries

Ni-Cd batteries started to be produced after thirty years from the entry in the production
of lead batteries, occupying, initially, areas non-covered by the lead batteries: railroads, safety
lighting, aviation etc. [6].
The Ni-Cd batteries use cadmium at anode, nickel oxy-hydroxide (NiOOH) at cathode
and an electrolyte made up of potassium hydroxide (KOH).
The total reaction is:
2NiOOH + 2H2O + Cd  2Ni  OH 2 + Cd  OH 2 (5)
There are two great categories of Ni-Cd batteries: with and without maintenance.
For the batteries with maintenance free, the gases resulting after the charge process are
recombined in order to form water, element leading to the electrolyte concentration
diminution.
For the batteries with maintenance, this process does not take place and the resulting
gases are eliminated through a valve, fact that leads, with time, to the necessity of adding
water in order to dilute the electrolyte.
The Ni-Cd batteries are equipped with a mechanism of protection against overcharge,
thus having from construction an excess of cadmium hydroxide (Cd(OH)2) at anode. From
here, it results that the cathode will be always the first to be charged. If the charge continues,
the following reaction takes place:
4OH  2H2O + O2 + 4e (6)
The resulting oxygen diffuses to the anode where the following reaction takes place:
O2 + Cd + 2H2O  2Cd  OH 2 (7)
As the charge process continues, at this electrode, the following reaction occurs:
2Cd  OH 2 + 4e  2Cd + 4OH (8)
From the analysis of the mechanisms described by these reactions we can observe that the
rate of cadmium hydroxide production is the same with the cadmium conversion rate, from
where it results that we have a stable system that does not consume materials from the
battery.
Table 4. Ni-Cd battery characteristics [4]
Parameter Value
Specific energy [Wh/kg] 40-55
Energy density [Wh/l] 70-90
Nominal voltage [V] 1.2
Self-discharging [%/day] 0.5
Temperature [°C] -40–80
Charging time [h] 1
At present, the Ni-Cd batteries are used in the electric vehicles, especially in Europe:
Peugeot 106, Citroen AX, Renault Clio, Renault Kangoo, but also in the USA - Ford Th!nk.
Considering the required voltage for powering these vehicles, between 8and 206V
modules are used. Each module is made up of 5 cells connected in series and provided with a
liquid cooling system, necessary in order to dissipate the heat resulting in the
charge/discharge processes.
3.2.2. Ni-MH Batteries

The study of these batteries started in the 80s, but the technology became commercial
available only in 1991, mainly due to the R&D programs. They are the result of the crisis
generated by the increase of Cd price in the 80s, the Ni-Cd battery being the only power
source for the portable electronic devices at that time. The R&D programs aimed at the same
performance of the batteries based on Ni, but without cadmium. The environment issues
related to Cd were revealed in the marketing period for this new type of battery – Ni-MH
battery.
These batteries are similar in construction to the Ni-Cd ones, but the anode uses hydrogen
instead of cadmium, which is absorbed in a metal alloy named metal hydride, MH. When the
battery is discharged, the metal hydride oxidizes and forms an alloy, and at the cathode the
nickel oxy-hydroxide is reduced and forms the nickel hydroxide.
The total reaction that occurs is the following:
MH + NiOOH  M + Ni  OH 2 (9)
The interesting fact in that Ni-MH batteries is that the anode behaves exactly as a fuel
cell. Thus, the hydrogen is attached to a metal hydride (an alloy specific to each
manufacturer) and released when required. The functioning of the Ni-MH batteries is
conditioned by keeping the hydrogen to a certain constant pressure (approximately 0.5 bars).
Also, air penetration within the cell should be prevented because it may react with the metal
hydride and replace the hydrogen.
Even though they have a quite high energy density, these batteries equipped specially
prototypes, but in 1996 they were the power source for the EV1 an electric vehicle
manufactured by General Motors, that provided a 250 km autonomy.
Table 5. Ni-MH battery characteristics [4]
Parameter Value
Specific energy [Wh/kg] 65
Energy density [Wh/l] 150
Temperature [°C] -40–80
Charging time [h] 1
The development of hybrid vehicles made the Ni-MH batteries the preferred choice for
several Full Hybrids vehicles, such as: Toyota Prius, Honda Civic, Honda Accord or Ford
Escape. The batteries for the hybrid Toyota Prius, were manufactured in 1997 by Matsushita
and contained 240D-type registered units. For the Prius II variant for Europe and USA they
use also Ni-MH batteries with a reported density of 1100W/Kg.
The first generation of Ni-MH batteries used cylindrical cells and an air-cooling system.
The new generations, in exchange, use rectangular cells loaded in metal cases. This implied a
diminution of the volume occupied by batteries with 26%, at the level of the entire module, as
well as a power increase with 55% (in relation to the volume). For a good thermal
management, the batteries are equipped with channels allowing the air coming from the fans
to reach all the cells.
Table 6. Evolution of the Ni-MH batteries [1]
First Generation First Generation Second Generation New Model

cylindrical rectangular rectangular
Nominal voltage [V] 7.2 7.2 7.2 7.2
Rated capacity [Ah] 6 6.5 6.5 6.5
Power [W] 872 1050 1352 1352
Mass [g] 1090 1050 1040 1197
Dimensions
l[mm] 19.6 19.6 19.6 18.4
H[mm] 106 106 106 96
L[mm] 285 285 271.5 279
3.2.3. Batteries Based on Sodium

In the 80s, following a new development direction, batteries using liquid sodium as active
material for the anode, started to be produced. Their main characteristic is that they operate at
high temperature. These batteries did not meet a rapid development because, in the first place,
they addressed only to one segment of the market, namely to the automotive industry.
There are two types of batteries based on sodium: with sulphur and with nickel chloride
(Zebra), out of which only the Zebra batteries imposed on the market.
Table 7. Zebra battery characteristics [4]
Parameter Value
Lifespan [no. of cycles] >1000
Charging time [h] 8
Zebra use the nickel chloride (NaCl2) as active material for the cathode and liquid sodium
for the anode. The electrolyte is made up of two parts: a ceramic electrolyte surrounding the
liquid sodium and a secondary electrolyte made up of sodium chloride and aluminum for the
cathode. The Cl ions are the mobile ions from the electrolyte (they provide the ionic current).
The electric energy obtained at discharge results from the combining of the sodium with the
nickel chloride, nickel and sodium chloride resulting as product.
The total reaction is the following:
2Na + NiCl2  Ni + 2NaCl (10)

In order to provide the high functioning temperature, Zebra batteries use two stainless
steel shells which include a 2-3 cm insulating layer. Nevertheless, except for short periods of
time of 2-3 hours, the batteries must be connected to a power source; otherwise they consume
approximately 100 W/h. It can be calculated that within 24 hours they consume nearly 13% of
the stored energy.
At present, Zebra batteries are mounted on the electric vehicle Renault Twingo within the
promotion program for electric vehicles in Mendrisio -Switzerland.
3.2.4. Batteries Based on Lithium

Lithium is the lighter of all metals used within batteries and presents interesting
electrochemical characteristics. It allows a high thermodynamic voltage, fact that leads to
very important energy and power densities. Li is a highly reactive substance. In 1970 the
researchers from the Oxford University discovered the solution for intercalating Li in a cobalt
or nickel crystalline network, thus creating the first batteries based on Li.
At present there are two major technologies based on lithium: lithium polymer (Li-P) and
lithium-ion (Li-ion).
3.2.5. Li-P Batteries

Li-P batteries use lithium at the anode and an oxide of a transition metal at the cathode.
This oxide presents a structure in which the lithium ions can be introduced or from which
they can be removed according to the process taking place: charge or discharge.
The electrolyte, for these batteries, is a solid polymer (SPE) obtained by the dissolution
of a lithium salt (Li(CFSO2)2N, LiTFSI), in a polyethylene-polypropyleneoxide (PEO-PPO)
[7].
xLi + M yOz  Li x M yOz (11)
At discharge, the lithium ions formed at anode migrate through SPE and they are
introduced in the cathode structure, at charge the process is being reversed.
Table 8. Li-P battery characteristics [4]
Parameter Value
Self-discharging [%/day] 0.5
Charging time [h] 8
In the 2000s, these batteries were mounted on the Ford Th!nk vehicles within the
programs sponsored by USABC. The battery structure was made up of 8 cells connected in
series (one module) and 8 modules connected in parallel in order to create a 24V and 80Ah
battery.
Starting with 2005 these batteries were withdrawn from the program, mainly because of
the security issues related to oxygen generation within the discharge process, the situation
presenting a risk of explosion. Now they are used especially in the field of
telecommunications.
3.2.6. Li-Ion Batteries

The term Li-ion hides, in fact, an entire family of batteries. There is no universal
structure, each manufacturer having a proper “recipe”. Nevertheless, there are a few
similarities related to the anode composition - usually made up of carbon or graphite and to
the electrolyte containing, in most of the cases, a lithium salt (LiPF6) in an organic solvent.
For the cathode, the most known structures are the oxides of Co, Mn, Ni: LiCoO 2,
LiMn2O4, LiNiO2, or more complex structures: LiNiMnCoO2, LiNiCoAlO2, LiFePO4, etc.
The LiCoO2 cells were among the first cells modules that used a lamellar structure for the
cathode (Figure 1) in order to allow the Li ions to pass from the anode to the cathode at
discharge, respectively from the cathode to the anode at charge.
Li Ions
Cobalt Oxide
Figure 1. LiCoO2 cathode structure [8].
These cells are characterized by a high energy density, but also by a reduced power
density, fact that makes them adequate for the automobiles requiring high autonomy and
reduced currents. These were used by Nissan for the automobiles in Japan, but due to the
overload issues they were withdrawn and now are used only for portable devices.
The LiMn2O4 cells present a crystalline (spinel) structure of the cathode (Figure 2) which
allows a high electron transfer rate fact that is translated in a high power density.
These batteries were tested on electric and hybrid vehicles prototypes in Japan, but, due
to their short lifespan they were withdrawn and are now used only for electric tools.
The LiNiMnCoO2 cells can have either a high power density or high energy density, but
never both at the same time. Their “secret” resides in finding the right balance between Ni-
providing a high energy density and Mn- which, due to its crystalline structure, provides an
increased flow of electrons and, implicitly, a high power density.
The LiNiCoAlO2 cells have the advantage of both high power density and high energy
density, fact that renders them ideal for the automobile industry. At present, Toyota tests this
type of cells on the Vitz model, but problems related to safety may lead to their withdrawal
from the market.
Li Ions Manganese Oxide
Figure 2. LiMn2O4 spinel structure [8].
The LiFePO4 cells came more recently on the market, but they drew attention of the
manufacturers in the automobile industry by the fact that they are the safest batteries based on
Li. This quality, corroborated with a high power density and with a short recharging time,
makes them ideal for the hybrid vehicles, being already in Toyota‟s attention. The only
impediments come from the relatively low voltage, of only 3.3V supplied, and from the
capacity loss at low temperatures.
The basic characteristics of Li-ion batteries are presented in Figure 3.
Figure 3. Li-Ion cells characteristics [8].
3.3. Metal-Air Batteries
The metal-air batteries followed a completely different development process as compared

to rest of the batteries, combining the technology of the fuel cells with the technology of
secondary batteries [9].
The defining characteristic for these batteries is their open structure, due to the necessity
of absorbing oxygen from the atmosphere, within the discharge process [10, 11].
There are several types of metal – air batteries using: Ca, Al, Fe, Cd, Zn, Li etc., and the
most known being those with Al, Zn and Li.
In the context where charging a battery consists in the reversal of the electric current
direction, the metal – air batteries proposed a new method, by the complete change of the
electrodes and electrolyte, because they are consumed during the discharge process. The main
advantage of this type of charge is the short time in which it can be performed, usually
between 8 and 10 minutes. The materials resulted after the charge can be afterwards recycled
in order to recover the main components.
3.3.1. Al-Air Batteries

The chemical reaction occurring within these batteries involves the combining of Al with
the oxygen in the air and with water in order to form the aluminum hydroxide (Al(OH)3),
energy being released during the process. The aluminum represents the negative electrode and
it is, usually, made up of a 1 cm thick plate. The electrolyte is an alkaline solution of
potassium hydroxide.
The total reaction is irreversible and is:
4Al + 3O2 + 6H2O  4 Al  OH 3 (12)
Table 9. Al-air battery characteristics [1,4]
Parameter Value
Charging time [h] 10
The battery is charged by the replacement of the anode and of the electrolyte, now
containing the aluminum hydroxide. The electrolyte and the anode are processed in a
recycling factory that recovers Al.
Due to the much reduced power density, this type of battery is not practical for being
mounted on electric or hybrid vehicles.
3.3.2. Zn-Air Batteries

The Zn – air batteries are similar to the Al – air batteries, but they have an approximately
10 times higher power density. A complete drawing describing the composition and the
operating process for these batteries is presented in Figure 4.
As we can see a Zn-air battery is made up of three parts: an anode composed of Zn, an air
cathode divided into two layers, a gas diffusion layer and a catalytic active layer and a
separator. This design was chosen to prevent the accumulation of reaction products in the air
electrode by improving the oxygen transport into the inner regions [12, 27].
Figure 4. Zn-air operating process [12].
The battery operation implies the diffusion of the oxygen from the atmospheric air,
through the carbon electrode, to the active catalytic layer from the cathode, where it‟s
transformed into hydroxyl with the help of the electrons generated from the anode after the
oxidation reaction of the metallic Zn [12].
2Zn + O2  2ZnO (13)
Table 10. Zn-air battery characteristics [1,4]
Parameter Value
Charging time [h] 10
The Zn-air batteries represent the most mature technology in the field of metal-air
batteries, combining in the most efficient way the low cost price, the operating safety and the
specific energy density [13].
3.3.3. Li-Air Batteries

Encouraged by the results obtained with the Zn-air batteries, the researchers focused on
lithium, which is the lightest material and which has the highest theoretical energy density. In
this context, appeared the Li-air batteries, batteries with an energy density of approximately
12 kWh/kg (Figure 5), being the closest to the gasoline energy density of 13 kWh/kg.
Figure 5. Energy density for several battery technologies [12].
There are 4 types of Li-air batteries (Figure 6), all using Li (metal) and air oxygen as
active materials at the anode and at the cathode. Differencing them is the electrolyte, which
can be:
 With Li salts dissolved in nonaqueous solvents;

 With Li salts dissolved in aqueous solvents;
 With Li salts dissolved in hybrid (nonaqueous/aqueous) solvents;
 With solid electrolyte.
(a) Nonaqueous (b) Aqueous
(c) Hybrid (d) Solid State Electrolyte

Figure 6. Li-air cell configuration [12]
Figure 6. Li-air cell configuration [12].
The use of metallic lithium could make it possible to have high theoretical specific
energy density value with very simple mechanism. During the discharge reactions, lithium
ions dissolve into the electrolyte, from the Li metal anode, and oxygen reduction was
occurred on air electrode. When oxygen reduction reaction is occurred, the processes are so
complex depending on the type of electrolyte. Li2O2 is the principal product of the reaction,
but in some case, Li2O2 is converted to Li2CO3 resulting a low cycle stability [14].
A key element for a long life for a Li-air battery is protecting the lithium metal anode
from air moisture [15]. This is done by using an O2 selective membrane, in the cathode,
which allows O2 into the cell while stopping or slowing the ingress of water vapor [16, 17].
At present, the Li - air batteries are not available on the market, but they present the
highest potential for solving the automotive industry problem related to batteries energy
density.
4. NEW TRENDS IN THE DEVELOPMENT OF THE BATTERIES FOR

THE ELECTRIC DRIVE
Of all the existing batteries, the batteries based on lithium raised the highest interest and
they have priority in the development as energy sources, due to the high energy density,
increased lifespan, increased tolerance to mechanical shocks etc. [18].
In order to improve the lithium batteries performances new materials were tested. The
conclusion was that in order to obtain high efficiency materials with a high lithium
interconnection in their structure must be used [19, 20, 21]. This analysis conducted to the
introduction of materials obtained by nanotechnology as electrodes for the lithium batteries.
A first step in this sense was made by the development of batteries using carbon nano-
tubes (CNT) instead of the graphite anode– Figure 7. Carbon nanotubes have extraordinary
thermal and mechanical stability and high electronic conductivity, which makes them an ideal
candidate for nanoscale matrix material for the anode [22]. The direct implication is that, now
the entire mass of the electrode becomes active material. Due to this, the surface area,
between the electrodes and the electrolyte, increases which leads to high electrical
conductivity and high capacity [23]. This technology applied to the LiCoO2, LiNiCoAlO2 and
LiFePO4 batteries lead to an increase with 80% of the specific energy and with 60% of the
energy density [24].
Figure 7. Carbon nano-tubes structure [13].

Encouraged by these results, other materials obtained by nanotechnology were also tested
as materials for the anode. Among these, the silicon was the most promising, due to its
specific capacity. Thus, the batteries using silicon nano-fibers at the anode presented a
significant capacity increase but also a performance increase to almost 90% [25].
Nano-tin powders supported by multiwall carbon nanotubes (MWCNT) were also tested
as materials for the anode. One of these is a Sn-MWCNT anode with 11-20% higher
volumetric capacity than a graphite anode and excellent reversible capacity stability. This
anode retains 71% of its initial reversible capacity up to the 280th cycle. In addition, the
anode has a significant advantage over the graphite anode in terms of battery safety [22].
Attempts were made also for the increasing of the lifespan, and in this direction, the
nano-materials using sulphur demonstrated a reduction of only 20% of the battery capacity
after 3000 cycles.
The research concerning the improvement of the Li batteries did not stop only to the
anode, it also involved the cathode. The main characteristics for the cathode materials should
be superior electrochemical properties without any rare materials. A way of obtaining these
materials is by „blending‟ two or more distinct lithium intercalation compounds. The
motivation for blending these compounds is to achieve a more balanced performance
compared to what is possible with any individual compound [25]. Among these, iron nitro-
disulphate combined with lithium bromide lead to the obtaining of batteries with125 mAh/g
capacity which that can be easily extended up to 200mAh/g, in the conditions where at
present, the highest performance present by a Li batteryis150 mAh/g capacity [26].
CONCLUSION
After more than 100years of continuous development, the electric vehicle reached a
decisive threshold, especially due to the development of hybrid vehicles. The preoccupation
for the environment and the diminution of the fossil fuel reserves will contribute to the
acceleration of their further development [1].
Due to the fact that there are several types of electric vehicles, each one with specific
requirements, several types of batteries will have to be developed. Figure 8 shows a chart
representing the evolution of batteries technologies for electric vehicles and it is the starting
point for predicting their new trends.
The lead-acid batteries are the best represented on the batteries market, and for a few
years they will continue to be, because of their superior cost to stored energy ratio.
Nevertheless, due to the specific reduced energy, it is hard to believe that they will one day be
mounted on the highly autonomous electric vehicles.
The Ni-Cd batteries are the only ones presenting a downward trend due to the fact that
their active element - Cd is a polluting material. Even though they have certain remarkable
performances, such as increased lifespan and excellent resistance to mechanic shock, they
will be replaced on the automotive market, but will find their way in the electric tools
industry.
The Ni-MH batteries dominate for now hybrid vehicles market segment. The clear
advantages related to the high energy and power density, as well as the fast recharging
capacity will help them to maintain themselves on this position on the medium term, but on a
long term they might be surpassed by the lithium batteries.
Figure 8. Evolution of the batteries technologies 1996-2012 [8].
The lithium batteries, propelled by new technologies and also by the existing financing
programs, will become a serious competitor for the Ni-MH batteries, and, on the long term,
they could even become the most used power source in the automotive industry [3]. Among
these, the LiFePO4 batteries, promoted by a much reduced price and by the highest operating
safety, will impose themselves in front of the other types of Li batteries.
For electric vehicles, the only considerable progress is made by the metal-air batteries
and, especially, by the Li ones. Due to the high energy densities and the short charging time,
these batteries can prove themselves to be the most important power source for this type of
vehicles.
REFERENCES
[1] Braussely, M; Pistoia, G; Industrial Applications of Batteries: From Cars to Aerospace
and Energy Storage. Elsevier, 2007.
[2] www.davidsuzuki.org/publications/downloads/2007/DSF-Prescription-for-Canada-Full-
Research.pdf.
[3] Dhameja, S; Electric Vehicle Battery Systems. Newnes, 2002.
[4] Larminie, J; Lawry, J; Electric Vehicle Technology Explained. John Wiley & Sons,
2003.
[5] Tomescu, A; Tomescu, I.B.L; Tomescu, F.M.G; Direct Energy Conversion. Matrix
Rom, Bucharest, Romania, 2008.
[6] Lefter, E; Power Supply Vehicles. Ed. Mediamira, Cluj-Napoca, 2006.
[7] Ehsani, M; Gao, J; Emandi, A; Modern Electric, Hybrid Electric, and Fuel Cell
Vehicles, Fundamentals, Theory and Design. CRC Press, 2005.
[8] http://batteryuniversity.com/learn/article/types_of_lithium_ion.
[9] Martin, J.J; Neburchilov, V; Wang, H; Qu, W; Air Cathodes for Metal-Air Batteries
and Fuel Cells. IEEE Electrical Power & Energy Conference (EPEC 2009), Montreal,
Canada, 22-23 Oct. 2009.
[10] Wang, J; Li, Y; Sun, X; Challenges and Opportunities of Nanostructured Materials for
Aprotic Rechargeable Lithium-Air Batteries. NanoEnergy Journal, vol. 2, issue 4, pp.
443-467, July 2013.
[11] Zhang, J.G; Wang, D; Xu, W; Xiao, J; Williford, R.E; Ambient Operation of Li/Air
Batteries. Journal of Power Sources, vol. 195, pp. 4332-4337, 2010.
[12] SooLee, J; Kin, S.T; Cao, R; Choi, N.S; Liu, M; TaeLee, K; Cho, J; Metal-Air Batteries
with High Energy Density: Li-Air vs Zn-Air. Advanced Energy Materials, no. 1, pp.
34-50, 2011.
[13] Cao, R; Lee, J.S; Liu, M; Cho, J; Recent Progress in Non-Precious Catalysts for Metal-
Air Batteries. Advanced Energy Materials, no. 10, pp. 1-14, 2012.
[14] Jang, I.C; Hidaka, Y; Ishihara, T; Li Metal Utilization in Lithium Air Rechargeable
[15] Shimonishi, Y; Zhang, T; Johnson, P; Imanishi, N; Hirano, A; Takeda, Y; Yamamoto,
O; Sammes, N; A Study on Lithium/Air Secondary Batteries: Stability of NASICON-
Type Glass Ceramics in Acid Solutions. Journal of Power Sources, vol. 195, pp. 6187-
6191, 2010.
[16] Crowther, O; Keeny, D; Moureau, D.M; Meyer, B; Salomon, M; Hendrickson, M;
Electrolyte Optimization for the Primary Lithium Metal Air Battery Using an Oxygen
Selective Membrane. Journal of Power Sources, vol. 202, pp. 347-351, 2012.
[17] Chervin, C.N; Long, J.W; Brandell, N.L; Wallace, J.M; Kucko, N.W; Rolison, D.R;
Redesigning Air Cathodes for Metal-Air Batteries Using MnOx-Functionalized Carbon
Nanofoam Architectures. Journal of Power Sources, vol. 207, pp. 191-198, 2012.
[18] Tao, H; Feng, Z; Liu, H; Kan, X; Chen, P; Reality and Future of Rechargeable Lithium
Batteries. The Open Material Science Journal, vol. 5, Suppl. 1: M2, pp. 204-214, 2011.
[19] Phan, A; High Performance Lithium-ion Battery Anodes Using silicon Nanowire.
COSMOS Cluster 8, 2012.
[20] Xiong, Z; Yun, Y.S; Jin, H.J; Applications of Carbon Nanotubes for Lithium Ion
Battery Anodes. Materials, vol. 6, no. 3, pp. 1138-1158, 2013.
[21] Bresser, D; Oschmann, B; Tahir, M.N; Tremel, W; Zentel, R; Passerini, S; Stabilizing
Nanostructured Lithium Insertion Materials via Organic Hybridization: A Step Forward
Towards High-Power Batteries. Journal of Power Sources, vol. 248, pp. 852-860, 2014.
[22] Menkin, S; Barkay, Z; Golodnitsky, D; Peled, E; Nanotin Alloys Supported by
Multiwall Carbon Nanotubes as High Capacity and Safer Anode Materials for EV
Lithium Batteries. Journal of Power Sources, vol. 245, pp. 345-351, 2014.
[23] Kang, C; Lahiri, I; Baskaran, R; Kim, W.G; Sun, Y.K; Choi, W; 3-Dimensional Carbon
Nanotube for Li-Ion Battery Anode. Journal of Power Sources, vol. 219, pp. 364-370,
2012.
[24] Landi, B.J; Cress, C.D; Raffaelle, R.P; High Energy Density Lithium Ion Batteries with
Carbon Nanotube Anodes. Journal of Material Research, vol. 25, no. 8, pp. 1636-1644,
2010.
[25] Chikkannanavar, S.B; Bernardi, D.M; Liu, L; A Review of Blended Cathode Materials
for Use in Li-Ion Batteries. Journal of Power Sources, vol. 248, pp. 91-100, 2014.
[26] Liu J; Yu, X; Hu, E; Nam, K.W; Yang, X.Q; Khalifah, P.G; Divalent Iron
Nitridophosphates: A New Class of Cathode Materials for Li-Ion Batteries. Chemistry
of Materials, vol. 25, no. 20, pp. 3929-3931, 2013.
[27] Williford, R.E; Zhang, J.G; Air Electrode Design for Sustained High Power Operation
of Li/Air Batteries. Journal of Power Sources, vol. 194, pp. 1164-1170, 2009.
Chapter 16
COMPUTATIONAL STUDY OF Li-ION BATTERIES FOR

AUTOMOTIVE APPLICATIONS
Inigo Aramendia*1, Unai Fernandez Gamiz1, Jose Antonio Ramos†2,

Javier Sancho1 and Ekaitz Zulueta‡3
1
Nuclear Engineering and Fluid Mechanics Department,
University of the Basque Country, Vitoria Gasteiz, Araba, Spain
2
Electrical Engineering Department, University of the Basque Country,
Vitoria Gasteiz, Araba, Spain
3
Systems Engineering and Automatics Department,
University of the Basque Country, Vitoria Gasteiz, Araba, Spain
ABSTRACT
The battery is one of the most important elements of an electric vehicle. The
implementation of these vehicles is determined by the performance to be derived from
their batteries, as described in [1]. Vehicle manufacturers must ensure that the life
expectancy for the battery of eight to ten years is viable before the beginning of series
production, because it is undesirable to inform to the buyer of a car the need to replace
the battery, associated with high costs, after only a few years.
Previous experimental work, e.g. [2], shows that the application of lithium-ion
batteries in the automotive world is growing significantly as a result of increased
commitment by electric and hybrid vehicles. Lithium batteries [3] are considered the
most attractive of all the storage solutions available today and they are expected to be
used in the next generation of hybrid and electric vehicles.
The main goal of this chapter is to analyze a battery submodule design consisting of
lithium ion cells that allows an optimum cooling battery, and assess the heat transfer and
temperature of the cells along the testing time. To that end, computational fluid dynamic
simulations (CFD) have been carried out by CFD. This is a very versatile and useful tool
which allows simulate many industrial cases as models and prototypes.
*
E-mail: email: inigo.aramendia@gmail.com.
†
E-mail: josean.ramos@ehu.es.
‡
E-mail: ekaitz.zulueta@ehu.es.
432 Inigo Aramendia, Unai Fernandez Gamiz, Jose Antonio Ramos et al.
Li-ion batteries performance and life are significantly affected by operating

temperatures of the battery. Thermal management is one of the main issues to be
considered in the design of the battery, [4]. The control of the maximum temperature and
of the maximum temperature difference in a cell is critical parameter in cell durability
and in element safety [8].
The battery model investigated in this work is based on cold plates, as showed in [9].
Two versions will be displayed, differentiated by varying the geometry of the cold plates;
in one, they will be solid and in the other one the cold plate is created by cold extrusion.
In both models will be shown the geometry and mesh of the elements. The physical
properties and initial conditions will be analyzed, as well as the convergence of waste
products, air speed, cells temperature and heat transfer.
Keywords: Li-Ion Batteries, CFD, Refrigeration, Automotive
NOMENCLATURES
VOC Open Circuit Voltage
V Nominal Voltage
Vmin, Vmax Maximum and Minimum Voltage per cell in volt
C Stored Electrical Charge in A-hr
n Recommended and Actual Number of Charge Cycles
SOC State of Charge
DOD Depth of Discharge
SOH State of Health
Specific Energy, Energy Density, Specific Potential in Wh/kg, Wh/L,
W/kg
Maximum Charge and Discharge Currents in Ampere
Internal Resistance in mOhm
Thermal Operating Range in ºC
Charge-discharge efficiency in %
1. INTRODUCTION
1.1. History
The first lithium battery prototypes were developed in the 1970s by M. Stanley
Whittingham, while he was working for Exxon, using titanium disulphide for the negative
electrode (cathode) and lithium metal for the positive electrode (anode). Nevertheless, the
first workable lithium batteries started to emerge from Bell Labs, where graphite anodes were
developed as a more efficient alternative to lithium metal.
By 1983, Michael Thackeray and his group had already identified manganese spinel as a
material for the cathode, as it has a high conductivity, relatively low cost and an internal
structure that offered good stability for use in next-generation storage batteries. Initially, the
drawback of this approach was that the spinel degraded after the first cycle of use, but this has
been addressed in subsequent designs.
Two years later, in 1985, Akira Yoshino proposed a prototype with a carbonaceous
material in which lithium ions could be inserted for one electrode and lithium cobalt oxide
Computational Study of Li-Ion Batteries for Automotive Applications 433
(LiCoO2), which is stable in air, for the other. This design markedly improved the safety of
the batteries, as it avoided the use of lithium metal. Moreover, the use of LiCoO2 meant that
production could readily be scaled-up to an industrial level.
In 1989, Arumugam Manthiram and John Goodenough at the University of Texas at
Austin took these early lithium batteries another step forward by incorporating so-called
polyanion cathodes, which were able to produce higher voltages than the oxides given the
inductive effect of the polyanion and their higher capacitance.
The first commercial battery, however, was launched by Sony and Asahi Kasei in 1991
with new features (lithium cobalt). The design made these storage batteries truly competitive.
Subsequently, in 1996, John Goodenough, AkshayaPadhi and coworkers discovered the
electrochemical advantages of using lithium iron phosphate (LiFePO4) and phospho-olivines
(lithium metal phosphates with the same structure as mineral olivine) as cathode materials.
These new materials further improved the safety of the batteries and were important in the
development of this technology, being used above all in mobile telephones and digital
watches.
In the last decade, the performance of Li-ion batteries has continued to grow: Yet-Ming
Chiang and his group at MIT demonstrating, in 2002, a considerable improvement by
increasing the conductivity of components by doping with aluminium, niobium and
zirconium; and the same group further improving the performance, in 2004, using iron
phosphate particles of less than 100 nm in diameter.
By 2011, Li-ion batteries accounted for 66% of the rechargeable battery market in Japan.
1.2. Background Knowledge
A battery is a set of electrochemical cells, generally connected in series to increase the

final voltage, that is able to store energy by means of a chemical reaction and later release the
chemical energy as electrical energy. All electrochemical cells are formed by anodes,
cathodes, electrolytes and an external circuit connecting the electrodes.
Nevertheless, not all batteries are alike. Some can only be used once: they cannot be
recharged as their internal structure does not enable the energy extracted from the battery to
be restored. This class of batteries would have to be disposed of after use and are known as
“primary cells”. There are different types:
 “Dry cells”, first those with Zn and C electrodes and mostly a ZnCl2 electrolyte and
later others, known as alkaline batteries, based on Zn/MnO2.
 Li-ion cells [1], the most widely used being those based on Li and MnO2. These are
relatively cheap to manufacture, but more expensive than dry cells. They are thought
to account for 80% of the batteries in general (rechargeable and non-rechargeable).
On the other hand, when the chemical process can be reversed by an electrical current
without producing an immediate degradation of the system, this is called a “secondary cell”
and the battery is rechargeable. Today, rechargeable batteries are a practical and effective
source of energy.
Until recently, batteries in cars have been an element that only provided the electrical
energy for them to work. For that, they have used small batteries kept charged by the running
of the car. Now, however, electric cars aim to go a step further; the goal is for their movement
to be powered by battery systems, abandoning the petrol- and diesel-powered motors we have
known to date. This type of batteries for electric cars can be divided into three types:
 Lead-acid batteries. The cheapest and most common type, these batteries have been
used to start combustion engines for decades. They have a good specific power
(W/kg), behave well at a wide range of temperatures, retain charge well over time
and are relatively easy to recycle. On the other hand, they have only a rather low
energy density (40 Wh/kg).
 Ni-metal batteries. Their specific power is satisfactory, they have a long lifecycle and
do not present environmental problems, but they have a relatively short shelf-life
(losing around 30% of their charge in a month when unused). Their production cost
is somewhat high given that rare earth metals are used in the anode. At around 60
Wh/kg, they provide a higher energy density than lead-acid batteries, but their
performance is not sufficiently good to allow a true expansion in electric car use.
 Lithium-ion batteries. These offer better performance than the other two types in
terms of voltage, energy density, specific power, usable charge, recharge efficiency
and lifecycle. They also have a slower self-discharge rate. They are, however, less
robust to voltage variations and are more expensive.
The following lithium-ion batteries are the most widely used subtypes.
 Lithium cobalt (LiCoO2): with an energy density of 170-185 Wh/kg, these provide
only 500 charge cycles. If damaged, there may be exothermic reactions and these can
even result in fire. They tend not to be used in electric vehicles.
 Lithium iron phosphate (LiFePO4): these offer an energy density of 90-125 Wh/kg.
They also offer up to 2000 charge cycles, and with this are the cheapest, together
with the cobalt batteries. Further, with a high thermal and chemical stability they are
safer. Indeed, they present a relatively low-risk option for powering vehicles.
 Lithium manganese (LiMn2O4): these provide an energy density of 90-115 Wh/kg.
This energy range is lower than that of the cobalt batteries, but they are more stable
and operate at higher voltages. What is more, manganese is less polluting.
 Lithium nickel cobalt manganese (LiNixCoyMnzO2): with an energy density of 155-
190 Wh/kg, these have a good cost-performance ratio. They are starting to be used on
a massive scale. They offer some 1500 charge cycles at among the highest voltages.
 Lithium titanium (Li4Ti15O12): these have the longest lifecycle providing some 12000
charge cycles, but offer the lowest range of energy density, 65-100 Wh/kg, and are
very expensive.
In a short time, lithium-ion batteries have become the main source of energy for
electronic equipment, due to the fact that they can store more energy per unit volume or
weight than lead-acid batteries and nickel-metal hydride or nickel-cadmium batteries (both
nickel-based designs being developments from alkaline batteries). Lithium-ion batteries are
usually stable for around 2200-2500 deep charge cycles, which corresponds to around 10
years of use. Moreover, a battery in fact still has 80% of its maximum charge capacity at this
stage. A wide range of sizes and designs of this type of battery is already manufactured for
various different applications, including information technology devices, and portable tools
and machines, as well as vehicles, see [2].
1.2.1. Charge and Discharge Processes

Charging: During this process, energy is supplied in the form of an electric current,
causing Li+ ions to flow from the anode into the electrolyte and thence to the cathode where
the host material is reduced and host ions (Li+) are inserted. At the same time, the positive
electrode is oxidised.
The electrolyte allows the flow of ions, but not of electrons. During charging, electrons
also flow from the anode to the cathode, but through the external circuit. In this way, the
cathode becomes more negatively charged and the anode more positively charged, increasing
the potential difference between them, and hence the voltage of the cell.
Discharging: The process is exactly the reverse of what occurs during charging; the
battery now supplies an electric current. This process forces lithium ions to leave the cathode
(extraction), oxidising this electrode, while the ions are inserted into the anode, leading to its
reduction. As the discharging process progresses, the electric potential (E) at each electrode
falls, decreasing the voltage across the battery.
The mobility of Li+ ions between the electrodes is possible due to the fact that both are in
contact with an electrolyte formed by a lithium salt dissolved in a non-aqueous solvent.
LiClO and LiPF6 are among the most commonly used salts, while ethylene carbonate (EC),
propylene carbonate (PC), dimethoxyethane (DME), diethyl carbonate (DEC) and dimethyl
carbonate (DMC) are the most common solvents.
1.2.2. Important Technical Considerations

From a technical point of view, there are five key factors to consider: safety, durability,
performance, cost, energy density and power.
 Good safety is undeniably essential for the success of a technology, and we tend to
associate Li-ion batteries with a risk of explosion in certain circumstances [3], this
being particularly severe in the case of large batteries, as are required for the
automotive application under study. A small number of accidents could result in the
technology being rejected by potential users. We should, however, keep this in
perspective given that, through it tends not to be perceived by society, a full 60-litre
tank full of petrol also presents certain hazards.
 Durability is not a trivial issue for current batteries. The manufacturers aspire to
batteries working for 10 years, this lifetime being considered to be around half that of
the vehicles themselves. On the other hand, cells progressively lose their capacity
with charge cycles, and this makes it necessary to design batteries with excess power
at the start, increasing their weight, volume and cost.
 Performance, another key characteristic, depends on many other factors such as, for
example, the temperature of the environment. With current batteries, we cannot
expect the same level of performance in a hot summer in Seville and on a cold day in
Ávila (both in Spain) or say Moscow. They could be optimised to improve
performance in hot or cold conditions, but many improvements need to be made to
obtain the same output in the full range of conditions that a given vehicle might be
expected to operate.
 The energy density and power must also be considerably improved, as the values of
these parameters offered by currently available batteries are far from that of petrol
(little over 1%), and this limits the autonomy of electric vehicles to a few hundred
kilometres at best.
 Lastly, all the aforementioned parameters should be optimised [4] without
excessively increasing the cost of the battery and, in turn, of the vehicle. For the year
2020, the goal set by the battery sector itself is some €175 per kWh, which is
approximately equivalent to the cost of the small batteries used in consumer
electronics. The batteries required are, however, much more complex to construct
and the current cost is still over four times this target. Clearly, large-scale production
of the batteries will allow the final price to be considerably lower than at present, but
it is nevertheless difficult to believe that this goal will be achieved.
1.3. Computational Fluid Dynamics (CFD) Tools
Currently, one of the most popular approaches to development in the industrial sector is
the simulation of processes. Design remains important in this context but it is now seen that
this has to go hand-in-hand with simulation. Thanks to simulation programs it is possible to
explore the details of multiple thermal processes that it was previously impossible to visualise
and analyse. By running simulations, time and money can be saved with respect to
investigating the real processes, which tend to be more expensive, lengthy and, in some cases,
more complex to interpret.
With the help of CFD tools it is possible to simulate and analyse any industrial process to
obtain a desired model. On the basis of such a CFD computer based model, a prototype can
then be validated and results obtained to later be applied to the real system. The usefulness of
CFD analysis lies in its ability to predict fluid behaviour in great detail in the domain studied,
support design and development of prototypes, and obtain solutions rapidly avoiding
expensive experiments, allowing visualisation and animation of processes in terms of
characteristics of the fluids involved. On the other hand, the CFD approach requires highly
trained users with extensive experience, and substantial hardware and software resources that
require significant investments, in some cases, the computational cost being high. In
particular, the equations of fluid mechanics do not have analytical solutions and hence need to
be solved numerical methods. For this, CFD requires discretisation of space. The working
approach will be:
1 Description of the problem

2 Base model/geometry
3 Discretisation/Mesh generation
4 Physical definition of the problem
5 Iterative calculations
6 Post-processing/Analysis
Given the increases in the price of petrol, its limited supply and the growing concern for
the global environment, the automation sector is committed to electrical and hybrid vehicles.
The device for the storage of electrical energy continues to be the component that is proving
the greatest headache for the manufacturers. This challenge requires the development of new
types of batteries which are more efficient and offer higher energy densities. Among the
solutions available today, lithium batteries are considered the most attractive candidates,
given their strengths over the conventional nickel-metal hydride batteries:
 Better performance and energy densities

 More compact size
 High charge/discharge efficiency
 More charge cycles
2. STATE OF THE ART: PAST AND PRESENT OF THE LITHIUM-ION

BATTERY COOLING TECHNOLOGY
Below it summarize the main candidate solutions, in terms of types of refrigeration and
arrangements of the cells.
2.1. Lithium-Ion Battery with Passive Cooling
This patent [5] is focused on high energy density batteries and, more particularly, on a
package of Li-ion batteries with a plate of metal, such as aluminum, to dissipate the heat of
the cells in the battery to the central chamber (Figure 1).
Figure 1. Passive cooling with an aluminum plate [5].

The battery consists of 28 cells, with rapid recharge times and long lifecycles and is
relatively environmentally friendly.
2.2. Modular Battery System with Cooling System
This is a modular system with at least two battery modules (14 15) (Figure 3) in which in
each module is a cooling element [6]. Figure 2 shows the distribution of batteries (in two
parallel rows and 3 in series in each row). A cooling bracket, a light alloy extruded section,
extends over the two rows of cells (Label 8 in Figure 3).
Figure 2. Distribution of the cells in a module [6].
Figure 3. System consisting of 6 modules [6].

If adjacent modules are connected in series a problem arises: the length of the ductwork
for the refrigerant near the main connection (the inlet) is shorter than for those at the back.
This translates into a greater flow of refrigerant to the front modules and, therefore, better
cooling.
To avoid this, the coolant return path could be on the opposite side of the connection and
then take it to the main common connection on the other side of the storage device through a
dedicated channel. However, this requires additional space and increases costs.
In principle, the coolant could be gaseous or liquid. In this patent, a liquid coolant was
proposed.
2.3. Battery Cooling Device
In this design, a first cooling medium is injected (air). It would enter by (112) and exit by
(114) (Figure 4). In addition, a fan can be installed to increase the volume of flow [7].
The second cooling medium can be water. It would be stored in the tank (136), supplied
by the cooling pipe, which can be a spiral to increase the refrigeration efficiency (332)
through the water pump (334).
Figure 4. Cooling device with air and liquid refrigerant (water) [7].
3. APPLICATION TO A REAL WORLD EXAMPLE

3.1. Technical Data
The base data/requirements considered are the key specifications equivalent to a battery
for a bus, tram or similar vehicle.
 Nominal Voltage: 380 V

 Nominal Capacity: 100 Ah
 Maximum Power: 100 kW
These base data, corresponding to the full battery pack, will be divided into modules and,
in turn, into submodules. The analysis will focus on the specific case of cooling a submodule,
for which is knows (from theoretical calculations) the number of cells to be placed in series
and parallel, as well as their specifications.
The following proposals, studied in more detail with simulations, are composed of Li-ion
cells LiFeBATT40138F1 (Figure 5). These cells have the characteristics in Figure 6.
Figure 5. LiFeBATT cell.

Figure 6. Physical characteristics and operation conditions of LiFeBATT cells.
 Each submodule is able to deliver 14 Ah at 48 V.
Total Capacity 100 Ah

Number of submodules    7.14  8 submodules (1)
Submodule Capacity 14 Ah
Actual total capacity  Submodule Capacity  No. of submodules  14.8  112 Ah (2)
 Hence, each module is able to deliver 112 Ah at 48 V.
Total Voltage 380

Number of submodules    7.916  8 modules (3)
Module Voltage 48
Actual total voltage  Module Voltage  No. of modules  48.8  384 V (4)
Then, is calculated the number of cells to be placed in series and in parallel in each
module and submodule to meet requirements.
Submodule:
Total Voltage Submodule 48 V

Cells in series    15 cells (5)
Nominal Voltage 3.2 V
Total Capacity Submodule 10 Ah

Cells in parallel    0.714  1 cell (6)
Nominal Capacity 14 Ah
Total Submodule Cells  15.1  15 cells (7)
Module:
Total Voltage Module 48 V

Cells in series    15 cells (8)
Nominal Voltage 3.2 V
Total Capacity Module 100 Ah

Cells in parallel    7.14  8 cells (9)
Nominal Capacity 14 Ah
Total Module Cells  15.8  120 cells (10)
Total cell Pack LiFeBATT  Total Module Cells  No. of modules  120.8  960 cell
(11)
To calculate the maximum current flowing through each cell is used the equation:
P  V .I (12)
First, the maximum power is divided by the total number of cells:
Maximum Power 100.000 W W

  104.16 (13)
Number total of cells 960 cells cell
Hence
P 104.16 W / cell
I   32.55 A (14)
V 3.2 V
Now the efficiency of each cell is estimated, with three possible cases, high, moderate
and low efficiency; from this, the heat generated in each cell (losses) is calculated to
determine how much heat needs to be dissipated from the system (heat losses).
Q  I 2 R  32.552  0.006  6.36 W (15)
 
Qtotal  I 2 R  No. of cells  6.36  960  6103.52 W (16)
If is focused on a submodule, then:
Qsub mod ule  VI low   3.2  32.55  1  0.8  20.83W (17)
First design: Forced convection cooling system

The first design studied is based on forced convection, as described in [8]. Is considered a
submodule of 15 cells, with 5 cells in series in 3 parallel rows. All this is placed inside an
aluminium housing, which has a series of inputs and outputs to ensure that the airflow is as
uniform as possible.
(a) (b)
Figure 7. (a) In-line and,(b) staggered configurations.
T = 298 K (25°C) P = 1.013 bar (atmospheric pressure) Result obtained

V = 1m/s b/a =1.5
c/a =2 Tw1=303K Avg. Flux = 0.4369 kW/m2
Tw2=323K
Cell surface area  πdh  π  0.0406m2  0.1388m  0.0177 m2 (18)
kW 1000 W
Total heat flux  0.4369 2
 0.0177 m 2  15 cells  115.99 W (19)
m 1 kW
Figure 8. Geometry of the linear system of cells with convection cooling.
The simulation was conducted with two different cell arrangements: in-line and
staggered. First, these were simulated with the "QTPro2" program. By entering the parameters
describing each configuration, you obtain the average heat flow.
The geometry of the model is first drawn with a CAD program and then imported to the
CFD solver, which generates the base mesh. For this analysis, is used a polyhedral mesh with
an average size of 0.002 m, obtaining 836,000 elements for the aligned cells and 975,000 for
the staggered configuration.
As for boundary conditions, is considered an air inlet velocity of 5.5 m/s. The physical
conditions (physical values) applied in the simulation are as Table 1.
Table 1. Physical Conditions of the model
Physics 1 (Cells) Physics 2 (Air)

Space Three dimensional Three dimensional
Motion Stationary Stationary
Time Steady Steady
Material Solid Gas
Equation of state Constant density Constant density
Flow (------) Segregated Flow
Viscous Regime (------) Turbulent
Reynolds Average Turbulence (------) K-Epsilon Turbulence
Optional physical model Segregated Solid Energy Segregated Fluid Temperature
Figure 9. Plot of temperatures.

The first result that is considered from the simulation is a map of the temperature across
the system. The air flows past all the cells to the exit. It can be observed that the cooling of
the cells is poorer the further they are from the inlet. This is due to the air being warmed as it
flows through the system by the heat generated by the cells.
In the in-line configuration, the temperature of the cells is around 333 K close to the inlet,
increasing to a maximum value of 356 K in the central cell of the last row. With the same air
velocity (5.5 m/s) we obtain the temperature patterns are illustrated in Figures 9 and 10.
Another output that from the simulation is the air velocity; this should be explored in to
determine whether there are problems such as the generation of eddies and vortices. These
effects cause the phenomenon of reverse flow.
The staggered configuration can provide significantly better results. On the other hand,
this system would require higher air flow velocities.
Figure 10. Detailed view for two cells of the velocity vectors field of the air to illustrate reverse flow.
3.2. Definition and Concept of Cold Plate Cooling
This consists of a plate of aluminum or another metal that contains an inner channel
through which a liquid coolant is circulated to absorb the heat transferred to the plate by
transistors and other elements mounted on it, in this case battery cells. The cooling method is
based on direct contact between the plate and the object to be cooled, direct contact allowing
the heat to be efficiently conducted away (Figure 11).
In fact, this system is a type of heat exchanger that uses metal plates (aluminum) to
transfer heat between two fluids. Its main advantage over a conventional heat exchanger is
that fluids are exposed to greater surface because they are distributed across the plate, as
showed in [9]. This facilitates the transfer of heat and increases the speed of the temperature
change. Generally, the pipe is formed into a serpentine shape in order that the liquid follows a
longer path and thereby the cooling is greater.
Figure 11. Cold plates [10].
3.3. System of Solid Aluminum Cold Plates
The cells are placed between cold plates (Figure 12), here made of solid aluminium, and
coolant pipes run through these plates, as shown in Figure 13. We illustrate the results for the
inline configuration. In this case, the geometry shown is only half of the system, the other half
being symmetrical. In this way, the simulation times are reduced significantly.
Figure 12. Geometry of the solid aluminum cold plate.
Figure 13. Geometry and meshing system with solid aluminum cold plates.
Table 2. Physical properties of the solid aluminum cold plates model
Physics 1 (Cells) Physics 2 (Cold plates) Physics 3 (Pipes)

Space Three dimensional Three dimensional Three dimensional
Motion Stationary Stationary Stationary
Time Implicit Unsteady Implicit Unsteady Implicit Unsteady
Material Solid (Aluminum) Solid (Lithium) Liquid (Water)
Equation of state Constant density Constant density Constant density
Flow (------) (------) Segregated Flow
Viscous Regime (------) (------) Turbulent
Reynolds Average
(------) (------) K-Epsilon Turbulence
Turbulence
Optional Physical Model Segregated Solid Energy Segregated Solid Energy Segregated Fluid Temperature
Table 3. Material characteristics
Density (kg/m3) Specific Heat (J/kg*K) Thermal conductivity (W/m*K)

Aluminum 2702 903 237
Lithium 2267 1016 1
Figure 14. Temperature fields.
Figure 15. The temperature of the cooling water: along the length of the pipe and a detailed.
Figure 14 shows the cell distribution in the computational domain. The mesh has 566,646
elements and the physical properties considered in the model are summarized in Table 2. The
characteristics of aluminum cold plates and lithium cells are in Table 3. The final solution of
the simulation is obtained with 6200 iterations.
It can be clearly seen how the temperature rises as the water progresses through the
system. Even so, most of the water increases only slightly in temperature (Figures 14 and 15).
The highest values correspond to the small fraction that is in contact with the aluminum plate.
Figure 16 shows the temperatures at the centre of each of the 8 cells.
Figure 16. Model chart with the internal temperatures of the 10 cells.
4. SYSTEM OF EXTRUDED ALUMINUM COLD PLATES

The second option that will be discussed is based on aluminum extrusions, the option of
producing solid sections being very costly in materials and hence expensive. In the extrusion
process, material is pushed or drawn through a die of the required cross-section. This method
has several advantages, such as being able to create complex sections and obtain an excellent
surface finish.
In Figures 17 and 18, can be seen both the new geometry and the meshing of the
complete model. This total system has 757,477 elements.
Figure 17. Geometry of the extruded aluminum cold plate.

Figure 18. Geometry and meshing system with extruded aluminum cold plates.
Figure 19. Temperature fields.
The boundary conditions and physical properties considered are equivalent to those
specified above for the system of solid cold plates (Figure 19). As for initial conditions, a
temperature of 298 K (25°C) corresponding to room temperature, is specified for the entire
system, while the inlet velocity of the cooling water is set at 0.33 m/s. It must also set the
"Energy Source" property for cells. This is the amount of heat generated by the cells divided
by the volume of the cell (W/m3).
The heat generated, obtained by theoretical calculations, is 20.83 W as Equation (20).
Heat Generated  W  20.83 W W

Energy Source    115919.86
Volume of the cell m  
3
π   0.0406 m 
2
 0.1388 m
m3
4
(20)
After calculating the full solution with 6200 iterations, the temperature field illustrated in
Figure 20 was obtained. It can be seen how refrigeration works at the surface of the cells, the
temperature decreasing as it moves away from the inside (Table 4).
Table 4. Cell Temperatures
Cell-surface temperatures Cell-internal temperatures

Entire cells Half cells Entire cells Half cells
Cell 2 305.873 K Cell 7 305.475 K Cell 2 313.83 K Cell 7 313.39 K
CONCLUSION
There is not a notable difference in temperatures between the systems with solid cold
plates and with cold plates formed by extrusion. In both cases the internal temperatures of the
cells do not exceed the operating range specified by the manufacturer, and hence both
systems could be considered valid approaches to cooling batteries for electric vehicles.
All things being equal, however, it could be argued that the system using cold plate
formed by extrusion is the best candidate, since it represents a considerable saving both in
materials, aluminium in this case, and in costs.
REFERENCES
[1] Vayrynen, A.; Salminen, J.; LithiumIon Battery Production. The Journal of Chemical
Thermodynamics, vol. 46, pp. 80-85, March 2012.
[2] Chacko, S.; Chung, Y.M.; Thermal Modeling of Li-Ion Polymer Battery for Electric
Vehicle Drive Cycles. Journal of Power Sources, vol. 213, pp. 296-303, 1 September
2012.
[3] Wang, Q.; Ping, P.; Zhao, X.; Chu, G.; Sun, J.; Chen, Ch.; Thermal Runaway Caused
Fire And Explosion of Lithium Ion Battery. Journal of Power Sources, vol. 208,pp.
210-214, 2012.
[4] Santee, S.; Xiao, A.; Yang, L.; Gnanaraj, J.; Lucht; B.L.; Effect of Combinations of
Additives on the Performance of Lithium Ion Batteries. Journal of Power Sources, vol.
194, issue 2, pp. 1053-1060, 1 December 2009.
[5] Polk, R.C.; Villarreal, R.; Sunderlin, T.A.; Lithium Ion Battery Pack Having Passive
Cooling. April 1,2011,www.epo.org, page bookmark WO2010141854 (A1).
[6] Maier, G.;Michelitsch, M.; Modular Battery System with Cooling System. April 1,
2011,www.epo.org, page bookmark US2011033742 (A1).
[7] Song, T.W.; Park, T.S.; Oh, D.J.; Jeong, J.Y.; Cooling Device of Battery Pack. April
1,2011,www.epo.org, page bookmark US2010136406 (A1).
[8] Mahamud, R.; Park, Ch.;Reciprocating Air Flowfor Li-Ion Battery Thermal
Management to Improve Temperature Uniformity. Journal of Power Sources, vol. 196,
issue 13, pp. 5685-5696, 1 July 2011.
[9] Zhang, Y.P.; Yu, X.L.; Feng, Q.K.; Zhang, R.T.; Thermal Performance Study of
Integrated Cold Plate with Power Module. Applied Thermal Engineering, vol. 29, issues
17-18, pp. 3568-3573, December 2009.
[10] www.aavid.com/product-group/liquidcoldplates.
Chapter 17
Ni-Cd BATTERIES FOR AUTOMOTIVE

APPLICATIONS: THE ac ELECTRICAL DATA OF
THE Ni-Cd BATTERY IN THE HAVRILIAK-NEGAMI
DIELECTRIC REPRESENTATION
Mohammad A. Alim1*, Sudip Bhattacharjee2, Sanjida Khanam3,

Nasiha Muna4 and Sabiha Runa5
1
Department of Electrical Engineering, Alabama A&M University,
Huntsville, Alabama, US
2
Department of Mechanical and Civil Engineering, Alabama A&M University,
3
ITT Technical Institute, Madison, Alabama, US
4
Department of Mathematics, University of Alabama in Huntsville,
5
Department of Chemical and Materials Engineering,
College of Engineering, University of Alabama in Huntsville, Huntsville,
Alabama, US
ABSTRACT
The non-Debye dielectric relaxation via Havriliak-Negami (H-N) empirical relation
has been observed for the ac small-signal electrical data of the Ni-Cd battery in the
impedance plane. Various analytical aspects of the same data are demonstrated using the
Davidson-Cole (D-C) representation. It is clear that the continuous relaxation process
corresponding to high frequencies yields skewness beside the breadth of the H-N
distribution function. Possible meaning of the depression parameters α and β is extracted
from the breadth and skewness or deviation of the semicircular relaxation loci which is
normally observed for either Debye or Cole-Cole dielectric relaxation. Both α and β are
overall responsible for the skewness of the loci in the high frequency segment. In reality
α and β cannot be separated and viewed as individual parameters causing breadth and
*
E-mail: mohammad.alim@aamu.edu.
454 Mohammad A. Alim, Sudip Bhattacharjee, Sanjida Khanam et al.
skewness in conjunction with the presence of the apparent depression angle for the
skewed relaxation loci.
Keywords: Complex Planes, Relaxation Time, Equivalent Circuit, Impedance Spectroscopy,

Immittance, Debye and Non-Debye Relaxations, Cole-Cole Relaxation, Davidson-Cole
Relaxation, Havriliak-Negami Relaxation, Ni-Cd Battery
1. INTRODUCTION
Batteries operate by converting chemical energy into electrical energy through
electrochemical discharge reactions. Interest exists in compact size solid-state thin-film
batteries that serve energy-handling issues for various applications associated with cell
phones, camcorders, lap-tops, toys, etc. popular electronic systems. Basically a battery is a
device that stores electrical energy which is consumed. It is, in general, used as a single unit
building block cell possessing termination and insulation. In reality a battery refers to a series
combination of individual cells assembled in a pack and provided with some means for
connecting to the system it serves. The attributes of the ideal battery includes: (a) high energy
density, (b) long life, (c) rechargeable capability, (d) safe handling, (e) flexibility in
applications, etc. There are two types of batteries available for use: (1) primary type
discharged after single usage and discarded, and (2) secondary type for repeated use which is
rechargeable by reversing the chemical reaction within it. The primary batteries are not of
concern in this context as the rechargeable batteries are drawing consistently increasing
interest for handling large energy density [1-4].
There are several types of secondary batteries drawn attention depending on the nature of
the applications. The Hg-Zn and Hg-Cd batteries got popularity for the miniaturized feature.
The Li-based batteries are used to power the microelectronic devices. There are two basic
rechargeable type batteries used in the commercial arena. The first type is very commonly
used as the automotive starting battery which is maintenance free during the service life. The
second type is sealed and does not allow venting of gas to the atmosphere. This battery
requires that the gasses generated when charging the cell be recombined as part of the
process. This recombinant technology is employed in all sealed Ni-Cd and in some sealed
lead cell types. Other types of rechargeable batteries require partial sealed structure having
electrolyte immobilization scheme to reduce the possibility of acid leakage. These batteries
are open to the atmosphere and release gases during charging and discharging [2-4].
The Ni-Cd batteries showing outstanding features of long durability and high power
performance beside very quick discharging and charging, and low cost have been in use in the
cordless electronic products and expanded the market for several decades as it reduces the
environmental effects significantly [3]. Three fundamental components of Ni-Cd battery are
the nickel species as the positive electrode active material, the cadmium species as the
negative electrode active material, and the aqueous alkali solution as electrolyte. In order to
ensure the sealing by means of the oxygen absorption by the negative electrode, the capacity
of the negative electrode is designed to be larger than that of the positive electrode so that the
negative electrode does not reach a state of full charge. Another advantage of Ni-Cd batteries
is the wide variety of the product line up which permits a diversity of applications [4].
Ni-Cd Batteries for Automotive Applications 455
The dielectric behavior of the Ni-Cd battery is of interest concerning characterization and
determining functional capabilities as it handles high energy density. It is used for the
acquisition of the ac small-signal electrical data in the form of immittance (impedance or
admittance). These data are analyzed in the complex plane formalisms, and then distinguished
between the Havriliak-Negami and the previously [5] analyzed Davidson-Cole
representations.
2. NI-CD BATTERY: STATUS AND PROSPECT

The continuous rechargeable batteries are used in the spacecraft [6] hybrid electric
vehicle (HEV) [7] for the capability of handling high power capacity and high energy density.
The major parameters adjudged for the rechargeable batteries are: (1) kilowatt hour, (2) mass-
to-volume ratio, and (3) operating (environmental) temperature. Based on functional
capabilities Ni-Cd batteries indicate better performance, operating temperature range, and
durability [8]. The Li-ion batteries though handle large energy density but charging cycle life
is less than the Ni-Cd batteries. Also Li-ion batteries have a chance to get accidental ignition
or violent reaction while evolving hydrogen gas. Otherwise Li-ion batteries, in general, are
safe to use in electronics. Other types of rechargeable batteries are not considered in
analyzing the dielectric data in the present investigation because those batteries possess less
advantage over Li-ion and Ni-Cd batteries when energy density, mass-to-volume ratio, and
safety aspects are considered.
The applications of the Ni-Cd batteries for high power requirements satisfy temperature
extremes, severe deep cycling, and mechanical (pressure or thrust or stress) abuse. Other
continuous rechargeable batteries exhibit high heat dissipation during charging, and very
limited tolerance to over-discharge. When the design capacity is low for the negative
electrode so that it reaches fully charged state, then a sharp pressure rise occurs in the Ni-Cd
battery. Since the reaction of oxygen with hydrogen evolve during over-charging of negative
electrode is slow, hydrogen evolution itself causes a serious pressure rise inside the battery
[4].
The Ni-Cd rechargeable battery uses nickel-oxide-hydroxide and metallic cadmium as
electrodes with alkaline electrolyte (such as potassium hydroxide) material [4]. The Ni-Cd
batteries suffer memory effect when discharged and recharged at the same state of charge
(SOC) several times. This means when a battery is repeatedly recharged before it has
discharged more than half of its power, it will “forget” its original power capacity. Thus, the
battery is capable of remembering charging cycle at the recharging point. While in use, the
Ni-Cd battery suffers abrupt drop in voltage as if the battery is discharged. Thus, memory can
be attributed to changes in the cadmium electrode reflecting charging involving conversion of
cadmium hydroxide to metallic cadmium which is microcrystalline (polycrystalline). This is
the nucleation process to get enlarged Cd polycrystals in the Cd electrode. Scientific literature
concerning this aspect is not documented but interchangeably words such as whiskers or
dendrites or crystals, etc. are used. This is undesirable for the Ni-Cd battery since it makes Cd
harder to dissociate during the high current discharge, and leads to high internal resistance
attributing to the Cd polycrystals having variation in grain-size.
The aforementioned difficulties identified require extensive research for the Ni-Cd
batteries in order to bring complete satisfaction form the users. Thus, improvement of the Ni-
Cd battery becomes obvious with the materials development [9]. For continuous rechargeable
applications the Ni-Cd batteries have potential due to longer time usage in the HEV and long-
range aircraft. This is primarily due to the safety and recycling charging issues despite some
minor problems with the memory effect.
3. THIN-FILM STACK OR LAYER BATTERIES

Battery power can be either increased or optimized by introducing parallel structure of
the stacks or layers as opposed to a single stack or layer. Multiple stacks or layers reflect
parallel capacitance-like structure as found in most junction devices. Stacks or layers cause
reduction in battery impedance by introducing more equivalent cell area which aids in voltage
regulation for the high current applications. Possible length or mass is likely to increase
slightly due to the thin-film stacks or layers. Thick-film stack or layer can be used as stack or
layer in the form of pellets due to the easiness of making pellets. Any kind of reduction in
materials for making batteries is a reflection of saving in the eventual manufacturing. This is
the overall fundamental concept of thin-film based stack or layer batteries.
The thin-film based Ni-Cd structure can be evaluated to compare the performance with
respect to the Li-ion technologies. This structure reflects Metal Foil Substrate as Cathode
Current Collector surrounded by a bi-metallic oxide material as Cathode having insulation
between them. The metallic electrode surrounds this structure as Anode. This basic structure
can be repeated sequentially to form multi-layers having Cathode Current Collector terminal
and Metal Anode Encapsulation with terminal for use [1, 10, 11]. Using the fundamental
concept of the thin-film technology, the Ni-Cd structure is expected to encounter severe
competition with the alternate technologies such as NiMH (nickel-metal-hydride) based
battery having very less memory effect. Electrode based on sintered or fiber material in the
Ni-Cd thin-film structure can be used in the NiMH technology.
Due to the memory effect, there is a great prospect for the NiMH technology [12] in lieu
of the Ni-Cd battery. This needs to be examined employing the acquisition of the dielectric
data and subsequent analysis [13] via the powerful novel tool/technique [13] known as the
Impedance or Dielectric Spectroscopy (IS/DS). The analysis of the dielectric data employing
complex plane formalisms may indicate direction for optimized performance of the NiMH
technology.
The comparison of the dielectric data for the Ni-Cd battery with the NiMH battery may
indicate overall performance. These data may play an important role in minimizing the
memory effect in the Ni-Cd battery or aid in complete removal of the memory effect in the
NiMH battery. This is not done as current concentration remained in understanding the
dielectric response via the data analysis and possible interpretation for the underlying
operative mechanisms. Various aspects of the relaxation mechanisms are explained utilizing
relevant complex functions and subsequent derivations from the analyses.
4. BACKGROUND AND ANALYTICAL APPROACH

OF THE DIELECTRIC DATA
The relaxation behavior of the ac small-signal electrical data of the Ni-Cd battery is well
understood in the complex plane formalism where the imaginary part of the immittance
function is plotted against the real part [13]. The ideal or Debye relaxation process from the
measured ac small-signal electrical data is very rarely observed [14-16] in the engineering
material systems and devices. The immittance response (I*) for the Debye behavior [16] given
by
I * ()  I   ( I dc  I  ) / (1  j ) (1)
is represented as the ideal response in the form of semicircular loci in the complex plane
formalism where the center lies on the real axis (x-axis). The non-ideal or non-Debye
behavior is usually observed via depression parameter and/or skewed nature of the
semicircular loci. In the Cole-Cole(C-C) or ideal-like non-Debye representation [16] the
semicircular response is depressed without skewing the loci or the relaxation curve. The other
non-Debye type responses occasionally observed via skewed behavior [17-20] of the
semicircular loci (curve) where the depression parameter is not accounted alone for that non-
Debye response. Overall, non-Debye dielectric response is often described as simplistic C-C
relaxation [16] as:
 ( ICCs  ICC )
ICC ( )  ICC  (2)
1  ( j CC )
while the complex non-Debye dielectric response is often described as the Davidson-Cole
[17, 18] (D-C)relaxation in the form
 ( I DCs  I DC )
I DC ( )  I DC  (3)
(1  j DC ) 
and Havriliak-Negami [20, 21] (H-N) relaxation in the complex plane formalism as
 ( I HNs  I HN  )
I HN ( )  I HN   
(4)
1  ( j HN ) 
 
where the measurement frequency increases in the anti-clock wise direction along the
relaxation curve. The subscript CC or DC or HN represents Cole-Cole or Davidson-Cole or
Havriliak-Negami relaxation, respectively. The subscript s denotes dc (or static condition)
measurement corresponding to f → 0 Hz, subscript ∞ indicates high frequency measurement
corresponding to f → ∞, and τ is the relaxation time for respective situation.
The depression parameter for the simplistic non-Debye (i.e., C-C) behavior is derived
from the semicircular relaxation [13, 16]. This is not the case for the relaxations in D-C and in
H-N loci. The curve fitting procedure [13] of the data points in the complex plane formalism
for the semicircular relaxation provides extraction of the associated parameters.
In an earlier work [5], the D-C depression parameter was found to be complex quantity
where real and imaginary parts of the D-C depression parameter were illustrated for the Ni-
Cd battery. In this work, extraction of the depression parameters is demonstrated for the H-N
relaxation using real quantity as well as complex quantity for the same ac small-signal
electrical data. The degree of precision depends on the user and the meaning of the extracted
parameters in the complex domain is not ascertained. However, the interpretation of the
extracted parameters is likely to remain the same as does exist for conventional analysis.
4.1. Dielectric Relaxation: Ideal (Debye)and Cole-Cole (Non-Ideal) Responses
The ideal or Debye relaxation is straightforward from the admittance expression [18, 19]
Y *  G p  jC p (5)
where the resistance Rp[(= 1/Gp) or the conductance Gp (= 1/Rp)] is in parallel with the
capacitance, Cp. The parallel Rp-Cp(or R1-C1) combination constitutes the Debye relaxation
time τ (= RpCp = R1C1). The semicircular relaxation function F*(ω) is given by
1
F  ( )  (6)
1  j
where
1 1
  (7)
 peak 2 f peak
satisfying
 peak    peak peak  1 (8)
when represented in the complex plane reveals zero depression angle. The ideal semicircular
relaxation containing an intercept on the left-side of the semicircle is presumed to be zero for
F*(ω) in Equation (6). The existence of a finite intercept on the left-side of the semicircle may
be referred to as the left-intercept. In representing a semicircular behavior the ordinate (y-
axis) and abscissa (x-axis) must have the same plotting grid scale so that the relaxation in the
complex plane can clearly be delineated. This means that each unit grid or graphical segment
in length on the ordinate must be equal to the abscissa.
0.6
1
0.4
2
F
3 4
0.2
5
0
 = 9
-0.2
-0.2 0 0.2 0.4 0.6 0.8 1 1.2
F
Figure 1. Generic representation of the relaxation curves complex plane for the relaxation function
F  ( ) given in Equation (11): (1) Debye formalism with depression angle θ = 0o or α = 1 and β = 1,
(2) Non-Debye Cole-Cole formalism with depression angle θ = 9o or α = 0.9 and β = 1, (3) Davidson-
Cole formalism with α = 1 and β = 0.5, (4) Havriliak-Negami formalism with α = 0.85 and β = 0.45, (5)
Havriliak-Negami formalism with α = 0.45 and β = 0.85.
A semicircle whose center lies on the x-axis is defined as Debye or Debye-like relaxation,
whereas for a non-Debye relaxation the center of the semicircle lies below the x-axis giving a
finite depression angle (θ) measured from the point at the left-intercept to the center of the
semicircle below the x-axis. This is displayed in Figure 1. The depression angle gives rise to
the amount of the depression demonstrated originally by Cole-Cole [17] as the ratio of θ and
 / 2 (= 90o) [13]. Thus, the Cole-Cole type depression parameter h is given by
 2
h  (9)
 /2 
The C-C depression parameter h ranges between 0 and 1 depending on the idealized
Debye and/or extreme simplistic non-Debye response of a semicircular relaxation. In
Equations (2) and (4) the parameter α is related to the Cole-Cole type depression parameter h
showing dependence to each other as:
1 h   (10)
The value of α ranges between 0 and 1. The Equation (2) takes Debye form as given by
Equation (1) when h→0. This implies that when α = 1 Equation (2) leads to Debye form as
given in Equation (1). The same Equation (2) leads to a point on the x-axis when h→1. This
implies that when α = 0 Equation (2) leads to extreme non-Debye response as a point on the
x-axis. From Equation (10) it is observed that as h0 then α1, and likewise as h1 then
α0. The ideal Debye or Debye-like behavior can be achieved as θ0o implying α1, and
an extreme simplistic non-Debye response can be visualized with θ90o implying α0. That
is why α can be termed as the “depression parameter” in the present context.
4.2. Analysis of the Havriliak-Negami Representation
The H-N relaxation function can be ascribed from Equation (4) representing the skewed
relaxation curve (arc) given by
F  ( ) 1/ [1  ( j ) ] (11)

         
F  ( )  1/ 1    cos    j   sin   (12)
  2   2 
Both depression and skewed nature of the H-N relaxation curve are represented by the
parameters α and β. In order to separate real and imaginary parts of Equation (11) F*(ω) is
expressed in the form:
1
F  ( )  (13)
 A  jB 
with
   
A  1    cos   (14)
 2 
   
B    sin   (15)
 2 
Then
1 1
F  ( )   (16)
 A  jB  
( X exp  j ) 
F  ()  X   exp   j    X   cos     j sin    (17)
where
2
     2 2   
X 2  A2  B 2  1    cos       sin   (18)
  2   2 
2    
X 2  1     2   cos   (19)
 2 
and
  
( ) sin  
B
tan       2  (20)
A   
1  ( ) cos  
 2 
Thus, real and imaginary parts become:
F   X   cos    (21)
and
F   X   sin    (22)
F 
 tan    (23)
F
The H-N response can be achieved either in the impedance plot or in the complex
capacitance plot where the measurement frequency moves anti-clockwise for each
measurement point on the curve representing the total data. The presence of the intercept on
the left side of the relaxation arc has a finite magnitude while the relaxation portion can be
expressed as in empirical Equation (11). Like the D-C representation the skewed nature of the
generic H-N curve is shown in Figure 1. The time constant is not necessarily corresponding to
the conventional ωτ = 1 at the peak of the skewed arc. Thus, ωτ = 1 is likely to be dubious
while deriving the simplified equivalent R-C(resistance-capacitance) circuit for the relaxation
function F*(ω) shown in Figure 2.
Differentiating imaginary part F˝(ω) in Equation (22) with respect to (ωτ)α:
dF  dX d

   X (  1) sin    
  X   cos    (24)
d ( ) d ( ) d ( )
where
1

dX 1     2     

 1  ( )2  2( ) cos   2 ( )  cos   (25)
d ( ) 2  2    2 
Figure 2. Simplified equivalent circuit for the semicircular relaxation in the impedance plane presuming
either only Debye or non-Debye (Cole-Cole) response.
1

dX      2     

 1  ( )2  2( ) cos  

( )  cos  2   (26)
d ( )   2    
with
          
  tan 1  ( ) sin  
  / 1  ( ) cos 
1
    tan  (27)
  2    2  
Differentiating γ in Equation (27) with respect to (ωτ)α:
d 1 d

 . (28)
d ( ) 1   d ( )
2
where
     
( ) 2 sin 2   1  ( ) 2  2( ) cos  
1 2 1  2    2  (29)
2 2
           
1  ( ) cos  2   1  ( ) cos  2  
     
and
       
 ( ) sin    sin  
d d  2   2 
   (30)
d ( ) d ( ) 1  ( ) cos        
2
   1  ( ) cos  2 
 2    
Rearranging Equation (28) it is obtained:
  
sin  
d  2 
 (31)
d ( ) 2    
1  ( )  2( ) cos  
 2 
In order to obtain the maximum of F˝(ω), setting Equation (24) as:
dF 
0 (32)
d ( )
Thus,
dF  dX d

   X (  1) sin    
 X    cos    0 (33)
d ( ) d ( ) d ( )
dF  dX d

  X 1 sin    
 cos    0 (34)
d ( ) d ( ) d ( )
1 1
 
dF           
sin    1  ( ) 2  2( ) cos 
2 2

  1  ( ) 2  2( ) cos    
d ( )   2    2 
  
cos    sin  
       2 
 ( )  cos    0
  2   1  ( ) 2  2( ) cos   
 
 2 
(35)
Switching the quantities of both sides of the equal sign Equation (35) becomes:
  
1 cos    sin  
 
    
    2 
1  ( )  2( ) cos  2   sin    ( )  cos  2   
2  
       1  ( )2  2( ) cos   
 
 2 
(36)
Rearranging Equation (36):
       
sin    ( )  cos     sin   cos(  ) (37)
  2   2 
Thus, at the maximum value of F˝, βγ = (βγ)peak, ωτ = (ωτ)peak, and then Equation (37)
becomes:
  
sin  
tan(  ) peak   2  (38)
  
( )peak  cos  
 2 
Equation (38) represents the condition when F˝ = F˝max. When α → 1, then Equation (38)
leads to:
  peak tan    peak  1 (39)
And Equation (20) yields

 
tan  peak =   peak (40)
Then Equation (39) becomes D-C response [5, 20, 21]:
 
tan  peak tan    peak  1 (41)
Expression of β appears as real quantity for α being real using Equations (38):
    
 1   sin   
 2 
   arctan   (42)
  peak   ( )  cos   
   
 peak
 2 
where, γ and απ are in radians; and (ωτ)peak corresponds to the maximum value of F˝(ω).
4.3. Geometrical Interpretation of H-N Relaxation at the Limiting Case
Figure 3(a) displays a generic representation of the H-N relaxation formalism. This
feature has a similarity with the D-C feature [17, 18] with α = 1. Equation (23) gives the ratio
between the imaginary and real part of the H-N relaxation function yielding tan (βγ). From
Equation (20) it is seen that tan (γ) increases as ω increases anti-clockwise, i.e., the angle βγ
increases. Further examination of the imaginary and real parts of the H-N relaxation function
and imposing limiting case, Equation (23) gives [19, 20]:
F 
lim  lim tan     tan   max  (43)
  F   
    
 sin   
 2    tan 
 
tan   lim    tan   max  (44)
          2 
    cos  2 
  

   max  (45)
  2
Combining Equations (43) and (45) it is obtained:
F   
lim
  F 
 tan     tan   max   tan   (46)
   2 
where

 max    max  (47)
2
The product α×β has a proven connection with the Kohlrausch-Williams-Watts (KWW)
function illustrated by Alvarez et al. [21]. Equation (44) indicates the ratio between imaginary
and real part of the H-N relaxation function attains its maximum value when ω→∞. This
situation is shown in Figure 3(a) where the bisector of the angle βγmax intersects the relaxation
curve at the point ωτ= 1. The angle created by this bisector with the x-axis is given by:
 max 
 (48)
2 4
Now, when ωτ = 1, Equation (20) yields:
  
sin  
 2    
tan   1   tan   (49)
    4 
1  cos  
 2 
Therefore,
  max
  1   (50)
4 2
and
      
tan  max
 2
  tan  4
  

  tan   1   FF  (51)
 1
Equation (51) proves that the bisector of the angle βγmax intersects the H-N relaxation
curve at a point where ωτ = 1. This important information can be used to determine the value
of τ.
Gerischer impedance [22] representation in the impedance plane lies for α = 1 and β = 0.5
which is a special case. This impedance gives ωτ = 1 in the low-frequency domain whereas
the peak of the loci gives ωτ = 3 ≈ 1.732 and the corresponding coordinate gives F'peak =
F'ωτ=1.732= 3 / 8 ≈ 0.6124 and F"peak = F"ωτ=1.732 = 1/ 8 ≈ 0.3536. Since the Ni-Cd battery
data indicated β = 0.6 with α < 1, the D-C relaxation formalism is not emphasized like
Reference 8. Figure 3(b) refers to the D-C formalism while Figures 3(c) and 3(d) represent
the H-N formalism. Thus, α = 0.999 and β = 0.6 as well as α = 0.995 and β = 0.6 indicate H-N
representation of the experimental data.
0.6
0.5
0.4
 = 1
0.3
F
0.2
max
0.1
max/2
0
-0.1
0 0.2 0.4 0.6 0.8 1
F
Figure 3(a). Havriliak-Negami representation of the relaxation function F  ( ) per Equation (6) where
measurement frequency runs from right to the left.
0.6
 = 1.497
0.4  = 1
F
0.2
-0.2
-0.2 0 0.2 0.4 0.6 0.8 1 1.2
F

Figure 3(b). Havriliak-Negami representation of the relaxation function F ( ) with α = 1, β = 0.6 that
corresponds to F'peak = F'ωτ=1.497 = 0.5843 and F"peak = F"ωτ=1.497 = 0.3905 along with F'ωτ=1 = 0.7237,
F"ωτ=1 = 0.3888.
0.6
 = 1.498
0.4  = 1
F
0.2
-0.2
-0.2 0 0.2 0.4 0.6 0.8 1 1.2
F
Figure 3(c). Havriliak-Negami representation of the relaxation function F ( ) with α = 0.999, β =


0.6 that corresponds to F'peak = F'ωτ=1.498 = 0.5842 and F"peak = F"ωτ=1.498 = 0.3899 along with F'ωτ=1 =
0.7236, F"ωτ=1 = 0.3682.
0.6
 = 1.501  = 1
0.4
F
0.2
-0.2
-0.2 0 0.2 0.4 0.6 0.8 1 1.2
F
Figure 3(d). Havriliak-Negami representation of the relaxation function F ( ) with α = 0.995, β =


0.6 that corresponds to F'peak = F'ωτ=1.501 = 0.5841 and F"peak = F"ωτ=1.501 = 0.3877 along with F'ωτ=1 =
0.7229, F"ωτ=1 = 0.3662.
4.4. Extraction of the Relaxation Time τ and the H-N Depression Parameters
α and β
The H-N impedance representation given by
Z LF  Z HF
Z  ( )  Z HF  (52)
[1  ( j  ) ]
can be ascribed as D-C representation for α = 1 with ZHF = high frequency (ω → ∞)

impedance, and ZLF = low frequency (ω → 0 or dc) impedance. The generic representation of
both H-N and D-C responses shown in Figure 1 depicts the real world situation for the Ni-Cd
battery impedance data shown in Figure 4 as the H-N response. The same data approximate
both D-C and H-N representations within the close range of the value of α. The iteration
process of the fitting of each relaxation illustrated in Figure 4 was previously employed for
the D-C response [5]. For β = 0.6 three separate values of α provide near identical fitting.
Thus, two H-N representations of Figure 4 cannot be easily distinguished from the D-C
representation for the same data. The error in each fitting was brought to as minimum as
possible. Table 1 shows the observed and calculated (predicted) values of the real and
imaginary part of the impedance for both D-C and H-N responses. The errors are indicated
therein.
The values of ZLF, ZHF, α, and β are obtained by trial and error for the measured data of
the Ni-Cd battery previously represented as the D-C dielectric behavior [5, 19]. Figures 5 and
6 represent the ratio of the imaginary to the real impedance of the relaxation domain versus
angular frequency for H-N response and D-C response, respectively. For the H-N case, when
α = 0.995 and β = 0.6, then βγmax = αβπ/2 = 0.9378 and tan(βγmax) = 1.3628. For the D-C case,
when α = 1 and β = 0.6, then βγmax = αβπ/2 = 0.9425 and tan(βγmax) = 1.3764. Realistically the
horizontal tangent on the curve in Figure 5 and/or 6 yields tan(βγmax) = 1.2449. This finite
difference between the theoretical and the experimental values is attributed to the difference
between the tangent of the line containing experimental data points and the tangent of the
fitted curve, respectively.
(a) α = 1 and β = 0.6
(b) α = 0.999 and β = 0.6
(c) α = 0.995 and β = 0.6
Figure 4. Havriliak-Negami representation of the Ni-Cd battery via impedance plot per reference 9 for
(a): α = 1 and β = 0.6; (b): α = 0.999 and β = 0.6; and (c): α = 0.995 and β = 0.6.
In order to determine the value of relaxation time τ for the H-N response as well as for
the D-C response, Equations (48) and (51) are used. According to Equation (51), a horizontal
line is drawn for tan(βγmax/2) which intersects the experimental curve at a point from which a
perpendicular line is drawn on the abscissa to extract the relaxation time corresponding to τ =
1/ω. This gives τ = 0.2025 sec for the H-N case (Figure 5) and 0.2009 sec for the D-C case
(Figure 6). The inset in Figures 5 and 6 shows enlarged intersecting region of the curve.
By setting various values of the product ω×τ in the relaxation function F*(ω) in Equation
(11) it is possible to obtain the impedance function given in Equation (52). Either of these two
functions can be used in plotting the product ω×τ on the abscissa versus imaginary part of the
relaxation or impedance function on the ordinate for specific values of α and β. Thus, the
imaginary part of Equation (52) can be plotted using Equation (11) for Debye, C-C, D-C, and
H-N responses. These responses are shown in Figure 7 wherein the values of α and β
prompted (ωτ)peak. It is clear that (ωτ)peak = 1 only for ideal Debye and ideal non-Debye (i.e.,
Cole-Cole) responses. Further (ωτ)peak ≠ 1 for non-ideal non-Debye D-C and H-N responses.
For the real world situation of Figure 4 represented in the F" format in Figure 7 that gives
(ωτ)peak ≠ 1 (i.e., (ωτ)peak> 1) is shown in Figure 8. The imaginary part F" as a function of the
product ω×τ is displayed in Figure 7 for Debye response with α = 1 and β = 1 yielding
(ωτ)peak = 1; C-C response with α = 0.5 and β = 1 yielding (ωτ)peak = 1.0; D-C response with α
= 1 and β = 0.6 yielding (ωτ)peak = 1.192; and H-N response with α = 0.5 and β = 0.6 yielding
(ωτ)peak = 1.273. Figure 8 displays the product ω×τ versus Z"(ω) for both D-C and H-N
responses. For D-C response with α = 1 and β = 0.6 yields (ωτ)peak = 1.4970; for H-N
response with α = 0.999 and β = 0.6 yields (ωτ)peak = 1.4980 and with α = 0.995 and β = 0.6
(ωτ)peak = 1.5010. The values of (ωτ)peak are within a close range presumably because of close
proximity of α or β for D-C or H-N response. It is interesting to demonstrate the same Ni-Cd
battery data for both D-C and H-N dielectric formalism.
To see the effect of change in the peak value of the product ω × τ on the parameter α and
β, different values of α and β are plotted against (ωτ)peak. Figure 9 shows the variation of β
with (ωτ)peak for various values of α indicating increase in β with decrease in α. Figure 10
shows the similar variation of α with (ωτ)peak for various values of β indicating increase in α
with decrease in β. In general, both α and β decrease as (ωτ)peak increases. Also when the
values of α and β approach zero (i.e., highly non-Debye situation), then (ωτ)peak becomes
extremely large β. Thus, the extreme non-Debye situation gives (ωτ)peak>> 1. The values of
(ωτ)peak for various values of α and β are shown in Tables 2 and 3.
4.5. Test of Depression Parameter β When α Is Real
In order to verify β being a real quantity, it is necessary to test its nature. Thus,
considering α as a constant real parameter for β being complex quantity given by
  1  j 2 (53)
and using it in Equation (13):

Table 1. The AC electrical data for the Ni-Cd battery
D-CData H-NData
Measured Data (× 10-4 Ω) (× 10-4 Ω)
f
Observation
Calculated Calculated
(× 10-4 Ω)
Number
(Hz) R2 Error R2 Error

0.998545 0.994314 0.997304 0.976946
Z′ Z″ Z′ Z″ Z′ Z″ Z′ Z″ Z′ Z″
1 50.110 14.320 0.240 14.38200 0.32400 0.062 0.084 14.40586 0.48162 0.086 0.242
2 39.800 14.420 0.380 14.44000 0.38580 0.020 0.006 14.46283 0.54686 0.043 0.167
3 31.610 14.500 0.490 14.50500 0.45830 0.005 0.032 14.52813 0.62064 0.028 0.131
4 25.110 14.590 0.610 14.57900 0.54290 0.011 0.067 14.60304 0.70386 0.013 0.094
5 19.950 14.710 0.720 14.66300 0.64130 0.047 0.079 14.68913 0.79750 0.021 0.078
6 15.850 14.820 0.840 14.76000 0.75530 0.060 0.085 14.78842 0.90265 0.032 0.063
7 12.590 14.940 0.950 14.87300 0.88670 0.067 0.063 14.90334 1.02037 0.037 0.070
8 10.000 15.070 1.070 15.00400 1.03700 0.066 0.033 15.03670 1.15135 0.033 0.081
9 7.9430 15.220 1.210 15.15800 1.20710 0.062 0.003 15.19195 1.29599 0.028 0.086
10 6.3090 15.390 1.370 15.34200 1.39680 0.048 0.027 15.37332 1.45410 0.017 0.084
11 5.0110 15.570 1.540 15.56100 1.60410 0.009 0.064 15.58587 1.62443 0.016 0.084
12 3.9800 15.790 1.740 15.82400 1.82350 0.034 0.083 15.83540 1.80402 0.045 0.064
13 3.1610 16.070 1.950 16.13900 2.04480 0.069 0.095 16.12824 1.98746 0.058 0.037
14 2.5110 16.390 2.160 16.51200 2.25180 0.122 0.092 16.47003 2.16557 0.080 0.006
15 1.9950 16.800 2.350 16.93900 2.42260 0.139 0.073 16.86462 2.32491 0.065 0.025
16 1.5850 17.270 2.500 17.40900 2.53390 0.139 0.034 17.31151 2.44755 0.042 0.052
17 1.2590 17.790 2.570 17.89700 2.56750 0.107 0.002 17.80198 2.51249 0.012 0.058
18 1.0000 18.350 2.530 18.37000 2.51860 0.020 0.011 18.31539 2.50046 0.035 0.030
19 0.7943 18.800 2.400 18.79600 2.39940 0.004 0.001 18.82081 2.40163 0.021 0.002
20 0.6309 19.220 2.200 19.15600 2.23310 0.064 0.033 19.28393 2.22184 0.064 0.022
21 0.5011 19.52 2.060 19.44400 2.04590 0.076 0.014 19.67722 1.98273 0.157 0.077
22 0.3980 19.730 1.890 19.66500 1.85860 0.065 0.031 19.98790 1.71435 0.258 0.176
23 0.3161 19.920 1.740 19.83200 1.68490 0.088 0.055 20.21853 1.44506 0.299 0.295
1
F (54)
 A  jB (   j )
1 2
cos   2 ln X  cos  1   sin   2 ln X  sin  1  

F  X  1 exp   2    (55)
  j sin  1  cos   2 ln X   cos  1  sin   2 ln X 
F  X  1 exp  2  cos  1  2 ln X   j sin  1  2 ln X  (56)
Therefore, Equation (23) becomes:

F 
 tan  1   2 ln X  (57)
F
1.4
tan(max) = 1.3628
1.2 tan( max) = 1.2449
0.6
tan() = Z /(Z - ZHF)
tan() = Z /(Z - Z )

HF
0.8
0.4
0.6
0.4 0.2
-5 0 5 10 15
 (rad/sec)
0.2
0 50 100 150 200
 (rad/sec)
Figure 5. H-N response for tan (βγ) versus angular frequency ω of the Ni-Cd battery data with α =
0.995 and β = 0.6. Maximum theoretical value of tan (βγmax) obtained as 1.3628, while experimental
value shows 1.2449. Inset shows ω = 4.9390 rad/sec corresponding to τ = 0.2025 sec.
1.4
tan(max) = 1.3764
1.2 tan( max) = 1.2449

0.6
tan() = Z /(Z - ZHF)
1
tan() = Z /(Z - ZHF)
0.8
0.4
0.6
0.4 0.2
-5 0 5 10 15
 (rad/sec)
0.2
0 50 100 150 200
 (rad/sec)
Figure 6. D-C response for tan (βγ) versus angular frequency ω of the Ni-Cd battery data with α = 1.0
and β = 0.6. Maximum theoretical value of tan (βγmax) obtained as 1.3764, while experimental value
shows 1.2449. Inset shows ω = 4.9766 rad/sec corresponding to τ = 0.2009 sec.
0.6
0.4
3
F
2
0.2
4
 = 1
0
0 2 4 6 8 10 12

Figure 7. The product ω×τ is displayed as a function of the imaginary part of the relaxation function
F*(ω) for (1) Debye response with α = 1 and β = 1 yielding (ωτ)peak = 1; (2) Non-Debye (i.e., Cole-
Cole) response with α = 0.5 and β = 1 yielding (ωτ)peak = 1; (3) Davidson-Cole response with α = 1 and
β = 0.6 yielding (ωτ)peak = 1.497; and (4) Havriliak-Negami response with α = 0.5 and β = 0.6 yielding
(ωτ)peak = 2.637, Vertical dotted line indicates ωτ = 1.
2.6
2.58 1
2.56 2
3
2.54
Z ( 10-4 )
2.52
2.5
2.48
2.46
2.44
2.42
2.4
1 1.25 1.5 1.75 2

Figure 8. The product ω×τ is displayed as a function of the imaginary part of Z*(ω) for curve 1:
Davidson-Cole response with α = 1.0 and β = 0.6 yielding (ωτ)peak = 1.4970; curve 2: Havriliak-Negami
response with α = 0.999 and β = 0.6 yielding (ωτ)peak = 1.4980; and curve 3: Havriliak-Negami response
with α = 0.995 and β = 0.6 yielding (ωτ)peak = 1.5010.
From Equations (45) and (46), it is seen that γmax (when ω→ ∞) is a finite quantity.
Similarly taking logarithm of Equation (19) and imposing the limiting case:
1  2    
lim ln X  lim ln 1     2   cos     (58)
  2     2 
0.9
0.8
0.7
0.6

0.5
0.4
0.3
0.2
 = 0.1
0.1
0.9 0.4 0.3 0.2
0 0 2 4 6 8
10 10 10 10 10
()peak
Figure 9. The parameter β versus (ωτ)peak for various values of α.
0.9
0.8
0.7
0.6

0.5
0.4
0.3
0.2  = 0.1
0.1
0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2
0 0 2 4 6 8
10 10 10 10 10
()peak
Figure 10. The parameter α versus (ωτ)peak for various values of β.
Taking limit on both sides of Equation (57) and using Equation (58), it is seen that no
definite finite limit exists. However, in reality the observed data indicate finite ratio F˝/F΄
when ω→ ∞ as depicted in Figure 1 and Figure 3. This gives
2  0 (59)
which in turn yields β as a real quantity. In the recent D-C analysis [5] it has been presented
that the parameter β is the complex quantity for α = 1. However, β is a real quantity for the H-
N analysis. If the same analytical approach applied to the D-C analysis [5], β will turn to real
quantity. The difference for the parameter β between this analysis and reference 4 is due to
the fact that when the linear least square fitting equations are used, the independent and
dependent variables are real against the complex quantities considered earlier.
Table 2. Values of ( )peak for various values of  for fixed 
  ( ) peak   ( ) peak   ( ) peak

0.1 0.1 9.660E+09 0.2 0.1 9.354E+04 0.3 0.1 1.98E+03
0.2 9.509E+06 0.2 2.958E+03 0.2 196.783
0.3 1.656E+05 0.3 393.550 0.3 51.754
0.4 9.375E+03 0.4 94.267 0.4 20.110
0.5 1009.5 0.5 31.133 0.5 9.672
0.6 163.750 0.6 12.598 0.6 5.322
0.7 35.145 0.7 5.864 0.7 3.213
0.8 9.271 0.8 3.02 0.8 2.075
0.9 2.862 0.9 1.686 0.9 1.412
0.4 0.1 275.991 0.5 0.1 84.345 0.6 0.1 37.733
0.2 50.042 0.2 21.727 0.2 12.375
0.3 18.552 0.3 9.932 0.3 6.496
0.4 9.204 0.4 5.717 0.4 4.135
0.5 5.353 0.5 3.732 0.5 2.92
0.6 3.441 0.6 2.637 0.6 2.201
0.7 2.369 0.7 1.967 0.7 1.734
0.8 1.715 0.8 1.527 0.8 1.411
0.9 1.290 0.9 1.221 0.9 1.176
0.7 0.1 21.019 0.8 0.1 13.412 0.9 0.1 9.363
0.2 8.177 0.2 5.951 0.2 4.611
0.3 4.762 0.3 3.749 0.3 3.093
0.4 3.263 0.4 2.718 0.4 2.346
0.5 2.440 0.5 2.124 0.5 1.900
0.6 1.927 0.6 1.740 0.6 1.603
0.7 1.580 0.7 1.471 0.7 1.389
0.8 1.331 0.8 1.273 0.8 1.228
0.9 1.145 0.9 1.121 0.9 1.102
0.999 1.001
4.6. Test of Depression Parameter α When β Is Real
In order to verify α being a real quantity, it is necessary to test its nature. Thus,
considering β as a constant real parameter for α being complex quantity given by
  1  j 2 (60)
Table 3. Values of ( ) peak for various values of  for fixed 
  ( ) peak   ( ) peak   ( ) peak

0.1 0.1 9.66E+09 0.2 0.1 9.51E+06 0.3 0.1 1.66E+05
0.2 9.35E+04 0.2 2.96E+03 0.2 393.55
0.3 1.98E+03 0.3 196.783 0.3 51.754
0.4 275.991 0.4 50.042 0.4 18.552
0.5 84.264 0.5 21.752 0.5 9.932
0.6 37.733 0.6 12.375 0.6 6.496
0.7 21.019 0.7 8.177 0.7 4.762
0.8 13.412 0.8 5.951 0.8 3.749
0.9 9.363 0.9 4.611 0.9 3.093
0.4 0.1 9.38E+03 0.5 0.1 1009.5 0.6 0.1 163.75
0.2 94.267 0.2 31.133 0.2 12.598
0.3 20.11 0.3 9.672 0.3 5.322
0.4 9.204 0.4 5.353 0.4 3.441
0.5 5.717 0.5 3.732 0.5 2.637
0.6 4.135 0.6 2.92 0.6 2.201
0.7 3.263 0.7 2.44 0.7 1.927
0.8 2.718 0.8 2.124 0.8 1.74
0.9 2.346 0.9 1.9 0.9 1.603
0.7 0.1 35.145 0.8 0.1 9.271 0.9 0.1 2.862
0.2 5.864 0.2 3.02 0.2 1.686
0.3 3.213 0.3 2.075 0.3 1.412
0.4 2.369 0.4 1.715 0.4 1.29
0.5 1.967 0.5 1.527 0.5 1.221
0.6 1.734 0.6 1.411 0.6 1.176
0.7 1.58 0.7 1.331 0.7 1.145
0.8 1.471 0.8 1.273 0.8 1.121
0.9 1.389 0.9 1.228 0.9 1.102
0.999 1.086
and using it in Equation (13):
1  j 2  
F 1 / 1   j  (61)
 

     1   2  
F  1 / 1    1 exp   2  cos   j   exp   2  sin   (62)
  2    

F 1 /  C  jD  (63)
where,
1
   2 ln  (64)
2
C  jD  X exp  j  (65)
  
 
exp   2  sin 
1
D
tan    2  (66)
C    
1    1 exp   2  cos 
 2 
sin 
tan  (67)

 1 exp  2 
  cos 
 2 
  
sin  1   2 ln  
tan   2  (68)
1   2   1 
  exp    cos    2 ln  
 2   2 
and
X 2  C 2  D2 (69)
Therefore,
1
F  X   e j   X    cos   j sin   (70)
 Xe 
j 
Now Equation (23) becomes:
F 
 tan  (71)
F
Taking the limiting case for Equation (67) as well as using Equations(64) and (68):
lim tan  lim tan  (72)

   
which does not exist, however, from the observed data and Equation (46):
F    
lim  tan   =finite value (73)
  F   2 
Therefore, using Equations (68), (71), and (73) it is seen as:
2  0 (74)
in order to have a finite limit as ω → ∞.
4.7. Effect of α and β on the H-N Distribution Function
The H-N distribution function is given by [19, 20, 23, 24]:
1 x  sin(  )
g HN ( x)  (75)
  x 2  2 x cos( )  1  /2
 
where
 sin( ) 
  tan 1    (76)
 x  cos( ) 
and
x   /  HN for 0     , 0  x   (77)
τ is the continuous relaxation time within a limiting range and τHN is the fitted relaxation
time obtained from the same data. So far the relaxation time τHN has been referred to as τ.
When α→1then gHN→gDC, where gDC is the D-C distribution function. When gHN is written as
a function of ln(τ), following properties is identified:

g HN  ln    0 , for –   ln   , and  g (ln )d ln  1 (78)

Figure 11 displays the effect of β on the H-N distribution function for a fixed value of α =
0.5. The inset shows the same response for the fixed value of α = 0.995 obtained from the H-
N fitting of the NiCd battery data (Table 1). The curves indicate shift of the peak of the H-N
distribution function as β increases with respect to x = ln(τ/τHN) = 0 line. As β increases,
curves become symmetrical about the x = 0 line for the H-N distribution function and
gradually tend to the C-C distribution function.
0.16
0.9
14
0.14
12
 = 0.9
0.12 10
gHN ()
 = 0.995
0.1  = 0.5
6
gHN ()
0.3
0.08 4
0.2 2
0.06
0
-0.2 -0.1 0 0.1 0.2
0.04  =0.1 ln( /HN)
0.02
0
-15 -10 -5 0 5 10 15 20 25
ln( /HN)
Figure 11. Effect of β on the H-N distribution function for α = 0.5, and τ = τHN.
The spread of the H-N distribution function (standard deviation) also reduces as it
approaches the C-C distribution function. When β = 1, the H-N distribution function
coincides exactly with the C-C distribution function with the peak at τ = τCC. Similar behavior
is observed at any other value of α.
Figure 12 displays the effect of the change in α on the H-N distribution function for a
fixed value of β = 0.6 which was obtained from the H-N fitting of the Ni-Cd battery data. The
figure indicates that the increase in α shifts the peak of the H-N distribution function toward
the x = ln(τ/τHN) = 0 line. Each curve becomes more skewed in the negative domain, i.e. larger
portion of the curve lies left to the x = ln(τ/τHN) = 0 line. This indicates the H-N distribution
function tends to the D-C distribution function. The spread of the H-N distribution function
(standard deviation) also reduces as it approaches the D-C distribution function. When α = 1,
the H-N distribution function coincides exactly with the D-C distribution function which lies
to the left of the x = ln(τ/τHN) = 0 line. Similar behavior is also observed at any other
value of β.
In Figure 11 the breadth of the distribution function becomes narrower as β→1 (i.e., C-C)
which indicates decrease in skewness reflecting near homogeneous distribution of the R-C
equivalent circuit values. In Figure 12 the breadth of the distribution function becomes
narrower as α→1 (i.e., D-C) indicating asymmetrical distribution of the R-C equivalent circuit
values. The value of the resistance is very much limited by the chord of the relaxation loci
indicating singular contribution to the relaxation time whereas capacitance is extracted from
the y-axis attributing to the skewness. The capacitance is presumed to be nonlinear causing
drift in the ω×τ = 1 from the peak of the relaxation arc.
Figure 13 shows the nature of the skewed behavior of the H-N distribution function with
limiting case of C-C for β = 1. Here the skewness is quantified as the difference between the
expected (mean) value of τ and the parameter τHN. The expected value of τ from the
distribution differs significantly from the fitted parameter τHN and the difference is dependent
on the parameters α and β. The expected value is always less than the fitted value, which
indicates the skewed nature of the HN relaxation function. With increase in β (i.e., H-N→C-
C) for fixed α = 0.5, the difference reduces and the expected value of ln(τ) coincides with the
fitted relaxation time constant at the C-C limit, i.e., E[ln(τ)]→ln(τCC). Similar behavior is also
observed in the case of H-N→D-C situation via the effect of α (for fixed β = 0.6). For the
limiting C-C case the difference between the expected value of ln(τ) and ln(τHN) is zero.
Increase in α reduces the difference between the expected value and the fitted parameter value
of τHN. However, unlike the C-C limit, the difference does not vanish at the D-C limit
condition since the distribution is highly unsymmetrical at the D-C condition.
0.5
0.4 0.9
()
0.3
HN
 = 0.6
g
0.2
0.1
0.3
 = 0.1
0
-10 -5 0 5 10
ln( /HN)
Figure 12. Effect of α on HN distribution function for β = 0.6.
Another way to determine the skewness of the H-N distribution function and the effect of
α and β on it is to calculate the skewness statistic of the distribution function. The skewness
of the distribution function is defined as [25]:
 1  3 / 23/2 (79)
whereμ2 and μ3 are the second and third moment about the mean of the distribution,
respectively. The skewness of gHN given by γ1 is plotted in Figure 14 as a function of α and β.
In reality α is not the skewness but amalgamated with β = 0.6 in Figure 14. The horizontal
line shows β = 1. This figure indicates that the skewness of the H-N distribution function
reduces when β increases for fixed α, and increases when α increases for fixed β. This is
expected since the increase in β indicates H-N→C-C condition, and the distribution becomes
more symmetrical. Thus, making it more gradual skew positively, i.e., the skewness vanishes
when the H-N distribution becomes C-C distribution. On the other hand, the increase in α
indicates H-N→D-C condition and the distribution becomes more non-symmetrical, thus
making it increasingly negatively skewed distribution.
 = 1 (C-C)
0
Effect of 
-5
E[ln ] - ln HN
-10
Effect of 
-15
-20
-25
-30
0 0.2 0.4 0.6 0.8 1
, 
Figure 13. Effect of α and β on difference between expected value of ln(τ), ln(τHN) for α = 0.5 and β =
0.6.
4.8. Meaning of the Depression Parameters α and β
The straightforward meaning of the depression parameter α implies the presence of the
depression angle of the relaxation arc in the absence of β as found in the C-C behavior. This
is not the case for the ac electrical data of the Ni-Cd battery displayed in the D-C
configuration [4]. The presence of the finite depression angle represents non-ideal conduction
processes wherein a considerable degree of non-uniformity is associated with the capacitance
term than the resistance term. It also indicates the breadth of the distribution of the relaxation
times. The depression parameter α indicates broadness of the frequency dependence of the
imaginary part of the impedance (Z" ≡ F"). This is related to the average relaxation time
attributing to the reactance. Often the average relaxation time referring to the singular lumped
R-C circuit (showing one relaxation time) is termed as the distribution of a set of relaxation
times [26, 27]. For all purposes α lies between 0 and 1 in order to satisfy Debye and C-C
responses. The straightforward meaning of the depression parameter β in the D-C behavior
implies the breadth and skewness [28] of the relaxation arc for the ac electrical data of the
Ni-Cd battery. The breadth and skewness of the relaxation arc is not necessarily overlapping
with the behavior of the depression angle for the relaxation process resulting skewness in the
relaxation arc. For the D-C behavior using the Ni-Cd data the relaxation arc is found to be
skewed when α = 1 and β = 0.6. The equivalent circuit model of the D–C behavior may be
viewed considering both ω×τ = 1 corresponding to the maximum of the y-axis parameter and
ωτ ≠ 1 corresponding to the maximum of the relaxation function F*(ω). Another way to view
equivalent circuit is to refer to Figure 3 where ω×τ = 1 does not correspond to the maximum
of the y-axis parameter but takes the value on the low-frequency side.
Considering the role of both α and β causing the H-N behavior it is understood that these
two depression parameters are not isolated from each other. The frequency domain product α
× β for the H-N parameters is related to the time domain KWW (Kohlrausch-Williams-Watts)
parameter [21, 24]. Further interpretation of combined α and β may be extended to both
diffusion and activation controlled dielectric behavior of the device under test.
The parameters α and β are not optimized as far as H-N representation is concerned. The
analytical approaches presented herein reflect that the Ni-Cd battery system is surely too
complex to comprehend the underlying conduction processes. Based on the present analysis it
is understood that the internal configuration of the Ni-Cd battery is responsible for the
evolution of the parameters α and β. The solution of the problems associated with the internal
construction and overall content of the Ni-Cd batteries may not further simplify the
parameters α and β. The previous analyses [5] indicate that there is a chance to view the
present H-N representation in the D-C format. This dual complexity requires clarity from the
internal construction viewpoint of the Ni-Cd battery. Another study [27] also indicates similar
situation for the Ni-Cd batteries though curve fitting procedure for the identical complex
plane analysis is not conducted. Nevertheless, it is clear that the Ni-Cd battery is too complex
for the impedance spectroscopy to simplify the conduction processes. It is possible to use the
information of the parameters α and β to achieve better performance characteristics. Perhaps
in this way impedance spectroscopic approaches can provide improved functions of the Ni-Cd
batteries.
0.5
 = 1 (C-C)
0
Effect of 
Skewness of gHN ()
-0.5
-1
-1.5 Effect of 
-2
-2.5
0 0.2 0.4 0.6 0.8 1
, 
Figure 14. Effect of α and β on the skewness of the H-N distribution function for α = 0.5 and β = 0.6.
CONCLUSION
The analysis of the ac small-signal electrical data in the H-N dielectric representation in
the impedance plane allowed understanding of the depression parameters α and β for the Ni-
Cd batteries. These two parameters are real quantities and cannot be separated to display
independent role for each of them for the H-N formalism. The breadth and skewness are
associated with these two parameters primarily at high frequencies when the data are
displayed in the impedance plane. The product (ω×τ)peak corresponding to the maximum value
of the imaginary part of the complex function F*(ω) is not equal to unity for the H-N
representation. This product depends on the parameters α and β and the extreme non-Debye
situation which is characterized by (ωτ)peak> 1. The geometrical relationships among various
parameters of the H-N relaxation function can be used to determine the relaxation parameter
τHN. Based on the behavior of the two depression parameters, it can be concluded that the
optimized values to be determined in future may yield useful battery performance concerning
the reduction of the memory effect.
ACKNOWLEDGMENT
The authors gratefully acknowledge Mr. Stephen R. Bissell for his initial contribution in
the D-C analysis using AC electrical data of the Ni-Cd battery presented in Reference 5. The
authors appreciate interest and intellectual support of Mrs. Jayati Bhattacharjee.
REFERENCES
[1] Greiner, J.; Microfabricated Thin-Film Batteries: Technology and Potential
Applications. Thesis, Massachusetts Institute of Technology, 2006.
[2] Sequeira, C.A.C.; Hooper, A; Solid State Batteries. NATO Science, Series E, Applied
Sciences, No. 101, Springer, 1985.
[3] Rydh, C.J.; Karlstrom M.; Life Cycle Inventory of Recycling Portable Nickel-
Cadmium Batteries. Resources, Conservation and Recycling, vol. 34, pp. 289-309,
2002.
[4] Morioka, Y.; Narukawa, S.; Itoh, T.; State-of-the-Art of Alkaline Rechargeable
[5] Alim, M.A.; Bissell, S.R.; Mobasher, A.A.; Analysis of the AC Electrical Data in the
Davidson-Cole Dielectric Representation. Physica B, vol. 403, pp. 3040-3053, 2008.
[6] NASA Practice No. PD-EC-1103, Ni-Cd Conventional Spacecraft Battery Handling
and Storage. October 1995.
[7] Dhar, S.K.; Ovshinsky, S.R.; Gifford, P.R.; Corrigan, D.A.; Fetcenko, M.A.;
Venkatesan, S; Nickel/Metal Hydride Technology for Consumer and Electric Vehicle
Batteries - A Review and Up-Date. Journal of Power Sources, vol. 65, pp. l-7, 1997.
[8] Kiehne, H.A.; Battery Technology Handbook. 2ndEdition, CRC Press, 2003.
[9] Dell, R.M.; Batteries: Fifty Years of Materials Development. Solid State Ionics, vol.
134, pp. 139-158, 2000.
[10] N.J. Dudney, N.J.; Solid-state Thin-Film Rechargeable Batteries. Materials Science
and Engineering B, vol. 116, pp. 245-249, 2005.
[11] Bates, J.B.; Dudney, N.J.; Neudecker, B.; Ueda, A; Evans, C.D.; Thin-Film Lithium
and Lithium-Ion Batteries. Solid State Ionics, vol. 135, pp. 33-45, 2000.
[12] Yang, C.C.; Polymer Ni-MH Battery Based on PEO-PVA-KOH Polymer Electrolyte.
Journal of Power Sources, vol. 109, pp. 22-31, 2002.
[13] Alim, M.A.; Electrical Characterization of Engineering Materials. Active and Passive
Electronic Components, vol. 19, no. 3, pp. 139-169, 1996.
[14] Von Hippel, A.R.; Dielectrics and Waves. John Wiley & Sons, 1954.
[15] Smyth, C.P.; Dielectric Behavior and Structure. McGraw-Hill Book Company, 1955.
[16] Cole, K.S.; Cole, R.H.; Dispersion and Absorption in Dielectrics: Alternating Current.
Journal of Chemical Physics, vol. 9, pp. 341-351, 1941.
[17] Davidson, D.W.; Cole, R.H.; Dielectric Relaxation in Glycerine. Journal of Chemical
Physics, vol. 18, no. 10, pp.1417-1417, 1950.
[18] Davidson, D.W.; Cole R.H.; Dielectric Relaxation in Glycerol, Propylene Glycol, and
n-Propanol. Journal of Chemical Physics, vol. 19, no. 12, pp.1484-1490, 1951.
[19] Havriliak, S.; Negami, S.; A Complex Plane Analysis of Dispersions in some Polymer
Systems. Journal of Polymer Science, Part C, vol. 14, pp.99-117, 1966.
[20] Havriliak, S.; Negami, S.; A Complex Plane Representation of Dielectric and
Mechanical Relaxation Processes in some Polymers. Polymer, vol. 8, pp.161-210,
1967.
[21] Alvarez, F.; Algeria, A.; Colmenero, J.; Relationship between the Time-Domain
Kohlrausch-Williams-Watts and Frequency-Domain Havriliak-Negami Relaxation
Functions. Physical Review B, vol. 44, no. 14, pp. 7306-7312, 1991.
[22] Sluyters Rehbach, M.; Impedances of Electrochemical Systems: Terminology,
Nomenclature and Representation - Part I: Cells with Metal Electrodes and Liquid
Solutions. Pure and Applied Chemistry, vol. 66, no. 9, pp.1831-1891, 1994.
[23] Zorn, R.; Applicability of Distribution Function for the Havriliak-Negami Spectral
Function. Journal of Polymer Science, Part B: Polymer Physics, vol. 27, pp.1043-
1044, 1999.
[24] Alvarez, F.; Alegria, A.;Colmenero, J.; A New Method for Obtaining Distributions of
Relaxation Times from Frequency Relaxation Spectra. Journal of Chemical Physics,
vol. 103. No. 2, pp.798-806, 1995.
[25] Zorn, R.; Logarithmic Moments of Relaxation Time Distributions. Journal of Chemical
Physics, vol. 116, no. 8, pp. 3204-3209, 2002.
[26] Alim, M.A.; Admittance-Frequency Response in Zinc Oxide Varistor Ceramics.
Journal of the American Ceramic Society, vol. 72, no. 1, pp.28-32, 1989.
[27] Alim, M.A.; Seitz, M.A.; Singular Nature of Preferential Conducting Paths at High
Electric Fields in ZnO Based Varistors. Journal of the American Ceramic Society, vol.
71, pp.C246-C249, 1988.
[28] Havriliak, S.; Havriliak, S.J.; A Rigorous Comparison of Theoretical Autocorrelation
Functions with Dielectric Relaxation Parameters for the Alpha Process of Polymers.
Journal of Polymer Science, Part B: Polymer Physics, vol. 35, pp.1887-1897, 1997.
[29] Hammouche,A.; Karden, E.; De Doncker, R.W.; Monitoring State-of-Charge of Ni-
MH and Ni-Cd Batteries Using Impedance Spectroscopy, Journal of Power Sources,
vol. 127, pp. 105-111, 2004.
Chapter 18
ADVANCED DRIVE ARCHITECTURES AND STATE

OF THE ART IN-WHEEL MOTORS
Blaz Stefe and Gorazd Lampic

Elaphe Propulsion Technologies Ltd., Slovenia
ABSTRACT
The following chapter describes different concepts of drive architectures for
conventional 4-wheel vehicles, most typical for passenger cars and delivery vans. Besides
the conventional drive architectures, a state of the art direct drive in-wheel propulsion
system and its key advantages are presented. Conventional drive architectures include an
internal combustion engine and a central motor electric drive, while the concept of the
direct drive in-wheel motor, boasting a high torque to mass ratio, zero maintenance
requirements, compactness and efficiency, is researched in more detail. This chapter
reviews the major mechanical design possibilities that are currently known in the field of
in-wheel electric motors, including direct drive motors and the concepts with gears or a
reduction gear. The drive cycle and extreme drive conditions, such as driving uphill and
driving at top speed, are simulated for several vehicle segments from B to E. In-wheel
motor requirements, such as peak torque and RPMs, are presented. As it is known, in-
wheel motors increase the total unsprung mass of the vehicle. For its consideration, a
mathematical model is described and the effect on the drivability is both tested and
theoretically analyzed and presented by charts. Being one of the crucial topics when
designing in-wheel motors for different kinds of vehicles, a table reviewing a notable
amount of different mechanical constructions is presented in this chapter.
Keywords: Direct Drive, In-Wheel Motor Propulsion System, Electric Vehicle, Central
Motor, Mechanical Design, Drive Architecture, Unsprung Mass

Email: blaz@elaphe-ev.com.
486 Blaz Stefe and Gorazd Lampic
1. INTRODUCTION
Energy supply for the internal combustion engines, low efficiency, high maintenance
costs, a requirement for its own compartment (usually under the hood of the car), and
environmental concerns all call for a radical change of the propulsion concept of the vehicle.
Centrally mounted electric motors have especially turned out to solve environmental issues
and to reduce maintenance costs. However, according to the research and market demand, the
most promising propulsion concept, providing ultimate solutions for the customer and urban
society, seems to be a highly adjustable direct drive in-wheel motor propulsion system.
Numerous electro-magnetic and mechanical designs and concepts of in-wheel motors have
already been produced by several companies and research institutions. Here we present the
findings and concrete results.
The drive architecture of an internal combustion engine (ICE) powered vehicle is limited
by the constraints related to the design and performance of their engines. Gasoline, diesel,
LPG and other internal combustion engines provide the required performance only at high
RPMs, while their performance is poor at low power and their mechanical design fit only as a
central motor. There, it takes a notable percentage of the vehicle‟s total space and keeps the
center of gravity high. This kind of drive architecture requires mechanical transmissions and
gears (gear-box, reduction gear, homokinetic joints, differentials, axles and kardan joint
couplings), both contributing to energy loss, weight and complexity [1] (Figure 1). Despite
the undisputable reliability and also respectful durability, there are needs and possibilities for
new solutions.
Figure 1. The drive architecture of an internal combustion engine (ICE) powered vehicle [2].
As a natural evolution, the central internal combustion engines are being gradually
replaced by electric motors. The improvements of the first generation of vehicles are done by
even more complex hybrid solutions, gaining benefits from both technologies. The drawbacks
are a complex production, increased total weight of the vehicle, required maintenance of both
propulsion systems and of course a higher price of the vehicle (which is in some jurisdictions
Advanced Drive Architectures and State of the Art In-Wheel Motors 487
kept competitive by the environmental law policy). The second generation of electric vehicles
runs purely on batteries and central electric motors, as shown in Figure 2. However, the
advantages of this system are still related only to the advantages of the electric motor
technology in relation to the internal combustion engine. No other benefits of this transition
are exploited as all the beforehand mentioned mechanical parts remain in place and contribute
to the higher weight of the car, energy loss and in general decrease the vehicle‟s efficiency.
When taking advantage of the electric motor, only the gearbox is optional and its necessity
depends on the type and characteristics of the electric motor. However, mechanical parts that
waste energy are not omitted. Therefore, the efficiency of this propulsion system (providing
around 70 kW of power) with a central motor can range between 60 to 65%. The mass of the
motor with an inverter normally turns out to exceed 100 kg and, including all the shafts,
gearing and the differential, exceeds the value of 150 kg. On the other hand, a 2WD
propulsion system with a low torque, high speed in-wheel electric motors with a reduction
gear (or planetary gear-box) and inverters weighs between 90 kg and 120 kg (saving on the
differential and axles) and can reach efficiencies in the range of 65 to 70%. But the state of
the art direct drive in-wheel in a permanent magnet synchronous motor (PMSM) [8],
providing the power around 70 kW for a 2WD propulsion system, can reach efficiencies up to
88% (including the loss of the inverter) and can, in the case of Elaphe construction, weigh
below 60 kg.
Figure 2. Vehicle propulsion system powered by a central electric motor [2].
2. HISTORY
Currently the most advanced drive architecture, using the real benefits of electric motors,
is the so-called direct drive propulsion architecture with implemented in-wheel motors.
Motors deliver the torque exactly where it is needed; at the wheel. This solution has already
been proposed and implemented for the testing purposes by innovators in the late 19th century
(Figure 3), but numerous technological, mechanical, economic and political constraints have,
for more than 100 years, kept such a motor from the market and every-day use. In the last
decades, technology has made a giant step forward in the fields of production, electro-
magnetic design and energy storage. The new era of mobility seems to be closer than ever
before. Using the technology of chemical energy storage on the one hand and a development
of fuel cells on the other, the development of advanced rechargeable batteries with high
energy/mass ratio has also pushed forward the development of in-wheel motors as a fusion of
advanced electro-magnetic topology and mechanical and production engineering, as shown in
Figure 3.
Figure 3. A car powered by in-wheel motors built by Dr. Ferdinand Porsche in 1900-1901.
3. DIRECT DRIVE IN-WHEEL PROPULSION

A direct drive in-wheel propulsion system brings several advantages in the following
fields:
 uncompromising performance, satisfying and exceeding a modern driver‟s demands;

 clean propulsion with zero on spot emissions (in most configurations, except the in-
series or parallel hybrid with the ICE), or total zero emissions when vehicles run on
electricity from renewable resources;
 unbeatable efficiency;
 zero maintenance;
 space saving technology.
3.1. Uncompromising Performance
Uncompromising performance comes from the very high specific torque state of the art
in-wheel electric motors and is usually provided from a standstill. In most configurations with
2 in-wheel motors, a torque applied on the road already exceeds the maximum torque applied
on the road by a standard ICE engine in the 1st gear. For the vast majority of vehicles, the
torque, applied on the road at the speeds where 2nd gear was engaged with a normal ICE
engine, is significantly in favor of the in-wheel motors. The in-wheel motor can supply very
high torque to the road at low RPM and full torque at a very broad RPM range, while the ICE
engine can only do that at higher RPM. The total torque reserve, which can be applied to the
road at any speed, is highly in favor of the in-wheel electric motor, as is evident from the
comparison of the charts in Figure 4.
As technology and economies are moving towards the oil-independent transport, it can be
expected that in-wheel motors will soon be implemented particularly into the vehicles with
rechargeable batteries or fuel cells instead of in-series hybrids with the ICE [7]. Clean energy
can in many cases be gained entirely from renewable resources, such as the wind, water or
solar energy. Even in cases where the infrastructure for the CO2 neutral energy resources does
not exist, emissions from the power plants (like gasses and particles) can be much better
controlled and cleaned than those exhausted from many independent vehicles. Moreover, they
can be emitted into the environment in uninhabited areas instead of in the city centers.
(a) Torque and power characteristics of the ICE
(b) Torque characteristics of the in-wheel direct drive electric motor
Figure 4.
The comparison of the torque and power characteristics of the ICE and in-wheel direct drive
electric motor: an ICE motor (Figure 4(a)) provides a very low torque at low RPM and
requires a transmission, while an in-wheel motor (Figure 4(b)) provides a full torque already
at the lowest RPM and does not require a transmission due to the high torque and low specific
speeds (the diagram is for an Elaphe in-wheel motor).
3.2. Unbeatable Efficiency
Unbeatable efficiency of the direct drive in-wheel motors comes from the already
mentioned absence of numerous mechanical parts, transmissions and gears, a low mass of the
active part of the motors and a state of the art electromagnetic topology of the motors.
3.3. Zero Maintenance
It is well known that a conventional ICE, besides the fuel, also requires regular
maintenance for a reliable operation during the years. Regular maintenance includes regular
checks and changing the motor oil, air, oil and fuel filters, timing belt, seals, and so on). Apart
from the bearing, there is no friction between the moving parts as the drag force of the electric
motor is literally generated in the air gap. No lubricating liquids and no filters are needed,
therefore such an electric motor is a maintenance free machine. In certain configurations,
seals and bearings might be a subject to change after a certain amount of miles are driven. But
even in conventional cars, the bearing must be changed after a longer period of continuous
use. What is also believed to become a revolutionary factor in the car industry in the next 5 to
10 years is a completely new concept of a car‟s chassis, body and passenger compartment.
That is particularly due to the absence of the ICE under the hood of the car and because of the
absence of the kardan joint couplings in case of the RWD vehicles, differentials, exhaust
collector and exhaust pipes with catalytic converter and other emission gases treatment
devices, such as diesel particle filter as shown in Figure 5. Moreover, the gearbox that usually
takes a large space in the engine compartment or is placed in the passenger compartment in
case of sporty or top-end vehicles is absent and the space can be used for the passenger‟s
comfort.
Figure 5. Major components of a conventional ICE vehicle that are not required in the in-wheel motor
vehicle. This way, saved space can be used for passengers, luggage, or simply saved, which can result
in more compact urban vehicles.
A new dimension of using a so far empty space inside the wheel of the vehicle, allows
constructors to design a completely flat floor of the car and use it for the passengers comfort
(Figure 6). A lot of space, previously occupied by the motor parts, can now be used as a
luggage compartment and the designers of the car are much less limited and can bring their
creativity to an entirely new level. City cars can therefore be smaller in all dimensions and
can with their agility and compactness contribute to a better traffic flow, less traffic jams and
can significantly help the utilization of the parking areas in the city.
Figure 6. Electric vehicle propulsion system powered by an in-wheel motor: the propulsion elements do
not take up a lot of space, which can therefore be used for the passenger‟s comfort [2, 3].
4. HIGH TORQUE AND LOW RPMS REQUIREMENTS

FOR THE DIRECT DRIVE IN-WHEEL MOTORS
The simplest way to design the drive system is without any gears. Such motors need to
produce up to ten times the torque required by geared motors, but the high speed can be
reached by relatively low RPMs. Beside the peak torque and RPM shift, the approach of
analyzing the efficiency of the motor is also different in geared and gearless propulsions. Let
us examine this on the example of the urban car [6].
The first step of the analysis includes the vehicle data, driving demands and any dynamic
driving cycle representing the expected typical use of the vehicle.
Table 1. Technical specifications of the compact city vehicle representing the

L7equadricycle/microcar segment
Parameter Value
Frontal area 2.2 m²
Total mass (with 2 people) 650 kg
Air drag coefficient 0.35
Friction coefficient 0.015
Number of in-wheel electric motors 4
Outer radius of a wheel 0.25 m
Figure 7. The selected complex urban driving cycle lasts for 1369 seconds and covers 12 kilometers
driven with an average speed of 31.5 km/h.
Figure 8. Probability density function of working conditions for an in-wheel direct drive motor: the
dark color represents the area where the electric motor runs most of the time. White areas are less
occupied. The working point of the stand still time is not shown, since it does not contribute to power
consumption in no hotel load case.
The example is made for the super compact city vehicle, also called the heavy
quadricycle, representing the L7e market segment with the following parameters (Table 1).
The main calculation has been performed using the Urban Dynamometer Driving Schedule
(Figure 7).
Using the vehicle and the driving cycle data, we can calculate a diagram showing torque
and wheel rotation speed demand (Figure 7). The torque and frequency demand diagram
shows what part of the driving time an electric motor spends at different torques T and
revolutions per second ν (Figure 8). By calculating such a diagram, we can determine the
torque and the frequency range that have to be achievable with an electric motor [6].
The Equation (1) shows Ptime, which is the probability density function showing the
probability to find the motor during the driving cycle (Figure 7) at the working point of given
rotations ν and torque T. In the equation, t is the duration of the driving cycle.
 max T max
t ∬Ptime  ,  d dT
 0 _T 0
(1)
In Figure 8, we can see for how long an electric motor works at different conditions
(torque and rev/s). From these data we can deduce that, for most of the time, the electric
motor is in regime at 15 Nm of torque and at mechanical frequency of 7 rev/s.
Figure 9. The function of energy distribution showing the relative parts of released energy at all torque
and mechanical frequency related working points during driving cycle.
The peak torque and the peak mechanical frequency needed can also be calculated. The
peak mechanical frequency depends only on the top speed in the driving cycle and on the
wheel diameter. It is found to be 16.3 rev/s or 978 RPM, which is far less than in electric
motors used in combinations with gears.
The peak torque has been determined from three different conditions:
 Firstly, the torque and frequency demand diagram in a standard driving cycle shows
a peak torque of 69.7 Nm. However, this torque may not be sufficient for some
special driving conditions where higher force or acceleration are required.
 Secondly, the peak torque needed can be calculated from acceleration requirements
(0 to 60 km/h in 6 seconds). It is found to be 121 Nm.
 The third way of the peak torque determination is the hill climbing ability. In order to
climb a 35% slope, the peak torque needed is 134 Nm, which is the highest number
and is therefore the required value of peak torque.
In order to develop an appropriate electric motor for a direct drive application, its
efficiency has to be optimized according to the working conditions in the predicted driving
cycle. By multiplying the probability of density function with torque and angular speed at
each working point of the diagram (2), a new diagram is obtained (Figure 9). It shows the
amount of energy released at every working point (torque and mechanical frequency) during
driving cycle.
The best overall efficiency is obtained when the electric motor and power electronic
efficiencies are optimized according to the working points where most of the energy is
released. Such diagrams are useful for the optimization of direct drive electric motors and
power electronics for all efficiency critical applications (Figure 9).
Total energy required in the driving cycle E is defined in Equation (2) and is only an
extension of the Equation (1).
 max T max
t ∬Ptime  ,  d dT
 0 _T 0
(2)
Absolute values of energy are used, so in case of energy recuperation at regenerative

breaking, the energy shown is positive.
Approximately 40% of released energy is in the middle area of the diagram, 21% at high
speeds, 9% at low speeds and 16% at regenerative braking area. This calculation shows that
the efficiency of the propulsion system has to be high in a wide area of torques and RPMs,
which differs from the systems with gears where narrower high energy efficiency area is
acceptable.
5. IN-WHEEL MOTOR CHARACTERISTICS FOR DIFFERENT

SEGMENTS OF PASSENGER VEHICLES
The characteristics of the motor, calculated in section 4, apply to a propulsion system of a
compact electric city car - A segment according to the EU classification. This kind of
passenger vehicles started to gain popularity around 2010 but the conventional cars are still
well demanded by the customers and a long transition period is expected before the market
will show a notable demand for the super compact electric vehicles. Until then, hybrids,
converted ICE and also some electric cars in the B, C, D and E market segments are expected
to appear on the market. We had to make a research of the characteristics of cars in the
mentioned segments in order to be able to develop appropriate propulsion systems for these
vehicles. In every car segment we have chosen one of the most common representatives and
calculated the required performance of the propulsion system for such a vehicle.
The representative vehicles are: Renault Clio – B segment, Kia Ceed – C segment,
Volkswagen Golf – C segment, Ford Mondeo – D segment and Audi A6 – E segment.
The performance (torque and power) of a direct drive propulsion system that meets the
hill climbing, top speed and acceleration requirements is presented in Tables 2-5.
Figure 10. Complex extra urban driving cycle on which the calculations for the segments B, C and D
have been performed.
Table 2. Torque, RPS and power requirements for different segments

of the vehicles to drive up a 30% slope at 30 km/h
Torque driving up a
RPS at 30km/h Power driving up a 30%
Vehicle 30% slope at 30km/h
[rev/s] slope at 30km/h [kW]
[Nm]
B segment* 898 4.40 24.8
C segment* 1266 4.18 33.3
C segment* 1063 4.18 28.0
D segment* 1441 4.20 38.0
E segment* 1709 3.92 42.1
Table 3. RPS, power and torque requirements for different segments of the vehicles to
reach top speed
RPS at top Torque at top Power at top

Vehicle Top speed [km/h]
speed [rev/s] speed [Nm] speed [kW]
B segment* 140 20.53 269 34.8
C segment* 150 20.92 408 53.6
C segment* 150 23.7 364 54.2
D segment* 160 22.38 350 49.3
E segment* 160 20.92 355 46.8
The results obtained from these calculations and presented in Tables 2-5 can help
developers of in-wheel or any other type of direct drive motor with defining the motor
requirements as far as peak torques and RPS are concerned. As the data are relevant for the
vehicle, the total torque values have to be divided by 2 or 4 to obtain required “per motor”
specifications for 2- or 4-wheel drive systems. From the data above, one can see that a direct
drive motor in class B segment car with a 4-wheel drive needs approximately 300 Nm of peak
torque and has to reach up to 1200 RPM. On the other hand, the direct drive motor for a class
E segment car with a 2-wheel drive needs 1200 Nm and also has to work at approximately
1200 RPM. In addition, they show the differences between car segments and the most
frequent working conditions in the complex urban and complex extra-urban driving cycle.
Table 4. Results of the calculations of the most frequent working point and peak torque
in the complex urban driving cycle (Figure 7) for the direct drive in-wheel motor in
different car segments
Complex urban driving cycle most frequent

Complex urban driving
Vehicle working point
cycle peak torque [Nm]
Torque [Nm] RPS [rev/s]
B segment* 533.3 91.4 13.2
C segment* 627.3 99.4 15.9
C segment* 729.6 106 12.4
D segment* 848.4 120 12.5
E segment* 1003 142 11.7
Table 5. Results of the calculations of the most frequent working point and peak torque
in the complex extra-urban driving cycle (Table 4) for the direct drive in-wheel motor in
different car segments
Complex extra-urban Complex extra-urban driving cycle most frequent

Vehicle driving cycle peak working point
torque [Nm] Torque [Nm] RPS [rev/s]
B segment* 1212 88.1 14.6
C segment* 1709 112.0 13.9
C segment* 1687 119 13.9
D segment* 1669 138 16.1
E segment* 2337 100.6 13.0
Physical dimensions and the weight of the vehicle are taken from the gasoline version of
the mentioned representative vehicles. Top speed and some other parameters are taken in such
a way that they best represent a possible electric version of the vehicle in the same segment.
6. MECHANICAL DESIGN POSSIBILITIES OF AN IN-WHEEL MOTOR

There are several design possibilities for in-wheel motor and other required elements near
the wheel. Among these elements, the brake, bearing, motor position sensor and seals are the
most important. In the case of front drive, the A-arms for the steering mechanism have to be
integrated as well and the attachment to the suspension has to be considered in all cases.
Let us examine all of them systematically and list their advantages and drawbacks.
In-Wheel motors
 In-wheel motors with gears (reduction gear or a gear-box)

– Michelin [5]
– Toyota Fine X
– Eliica (Keio University, Japan) (Figure 11)

– Direct drive in-wheel motors
– Elaphe Propulsion Technologies
– SIM-LEI [15]
– Protean Electric
– Siemens VDO [16]
– Schaeffler [17]
– TM4 [18]
– Evans Electric Limited [19]
– e-Traction [20]
– Fraunhofer hub motor
Generally, direct drive in-wheel motors are divided into motors with:
 Outer runner
 Inner runner
6.1. In-Wheel Motors with Gears
In-wheel motors are generally divided into motors with gears and direct drive motors.
Some producers have reported a use of a reduction gear [1] or even a whole gearing system,
such as Eliica research car from the Keio University in Japan [1]. Implementation of a gearing
system allows the usage of conventional high speed electric motors with a low specific torque
and in the applications with a gear-box a planetary gear-box system is usually implemented as
it provides a high gear ratio (3 to 35). Furthermore, from the point of view of the mechanical
structure, other types of gear boxes are even less convenient. But it must be taken into
account that a gearing system means an additional weight of the propulsion system,
approximately 0.1 kg/Nm [1].
It is typical for motors with any gearing system and a conventional low torque high speed
motor [1, 5] to be axially longer. This means they are not protected by the rim of the vehicle,
so damages can occur when parking on the pavements or simply by the stones and obstacles
on the road. Moreover, such motors present notable constraints when mounted into the
steering wheels as it is technically very inconvenient to steer axially long and overall complex
assembly and mechanical construction. A conventional low torque and high speed motors
connected to the gear-box or reduction gear usually have a lower diameter than a wheel itself
and therefore take space right in the center of the wheel, where a brake is placed in the
conventional cars. Therefore a standard brake usually has to be replaced by a newly
developed braking system and installed outside the center of the wheel. This not only
increases the cost of such braking topology and often reduces the braking ability of the
vehicle, but also causes steering torques on the steering system, which can result in a
dangerous situation when all the braking wheels do not have equal grip to the driving surface.
Regarding the energy efficiency, the coupling of the motor and gearbox itself implicate
additional energy dissipation and therefore reduce the power and range of such vehicles.
Figure 11. Axially long low-torque high speed motors of the Eliica research car from the Keio
University in Japan with a gear-box [1].
As already mentioned in section 4, the direct drive motors typically boast high torque
already at low RPM and therefore present a notable potential for a high performance
passenger car. Moreover, they are usually axially short so that they can be entirely
implemented into the rim of the car. This protects them from environment (stones, splashing
water) and makes it safer to park them on the pavement. As such motors are designed in a
way to have active parts of the motors distributed only on a large circumference and are
therefore “hollow” at the lower diameters, there is a space left in the center of the vehicle for
the implementation of a conventional mechanical brake. Due to a less complex construction
and absence of the gearing, these kinds of motors are also lighter, which means lower
unsprung mass and consequentially higher durability, driving comfort and car turning agility.
As the motor can be entirely integrated into the rim of the car, there are no more constraints
with the maximum steering angle of the car. Wheels can theoretically steer even at the angle
of 90° and can therefore benefit from this uncompromising maneuverability and can park
sideways.
Increased unsprung mass of the wheels is a concern engineers come across when
developing in-wheel electric motors. Higher mass of the wheel also means higher inertia of
the wheels, which results in higher acceleration forces from the road on the wheels. A lot of
skepticism has been expressed whether these factors can have a negative impact on the
vehicle‟s stability, driving comfort, safety, passenger‟s comfort in the cabin and the durability
of the bearing and mechanical components of the motor [11]. The distribution of the mass of
the vehicle is also notably changed as the mass of the engine is moved from the center of the
vehicle towards the outer edges. It increases the moment of inertia of the vehicle around the
vertical axis through the center of gravity of the vehicle so the handling performance of the
vehicle is decreased. However, tests showed that this can only influence the performance of
racing cars and sports cars and cannot be noticed in an everyday ride with a passenger car.
Schalkwyk and Kamper [6] have presented a simple mathematical model of the quarter of the
vehicle‟s suspension for the analysis of the mentioned parameters on the vehicle‟s stability
and comfort, as shown in Figure 12.
Figure 12. A mathematical model of the quarter of vehicle‟s suspension.
In the Equations (3) and (4), Mv and Ms are the sprung and unsprung masses, Kt and Ks
are the spring coefficients (of the tire and of the damper), while damping coefficients are
expressed by Bt and Bs. The dynamic equations of the system are as follows:
MV z   Ks  z  y   BS  z  y   MV g (3)
M S y  KS  z  y   BS  z  y   KT  y  x   BT  y  x   M S g (4)
Interestingly, Schalkwyk, Kamper [6] and Anderson and Harty [4] have shown a slight
impact of the in-wheel motors on the vehicle‟s stability, maneuverability and driving comfort.
The wheel hop (a phenomena where a wheel loses contact with the ground) almost never
occurs and the unsprung mass does not decrease the driving comfort of passengers. It does,
however, imply higher vibration forces to the motor hub, which can decrease the endurance of
certain components of the vehicle, such as the bearing. Furthermore, the components of the
motor must be constructed in a way that makes them resistant to such working conditions [9-
10].
Anderson and Harty [4] have presented a Cooper rating scale describing some particular
properties and performances of the vehicle with integrated 30 kg in-wheel motors in each
wheel.
As can be noticed from the figures below, only a few performances of the car can be
notably changed in a car powered by the in-wheel motors. Considering the ride, these two
parameters are pitch control and a great impact feeling. As already mentioned, these two
factors are only notable at higher speeds and sharp cornering so they do not really affect
passenger cars (Figures 13-14) [11-17].
Considering the handling properties, tracking stability and yaw stability can be
influenced, yet again it has been discovered that the most worrying cornering grip level
remains unaffected.
Figure 13. Ride behavior of the vehicle with 30 kg unsprung mass at each end [4].
Figure 14. Handling properties of the vehicle with 30 kg unsprung mass at each end [4].
6.2. Direct Drive In-Wheel Motors
As far as the mechanical design of direct drive in-wheel motors is concerned, they are
divided in two main groups: motors with an outer runner rotor and motors with an inner
runner rotor. During almost 30 years of research and prototyping, we have produced and
tested the following main types of the in-wheel motors.
6.2.1. Outer Runner Motors

The most typical and at the same time the most convenient topology of the direct drive
in-wheel motor is the one where the standard car brake can remain in the hub of the wheel. In
such cases an expensive development and homologation of a new braking system is avoided
and a braking performance remains unchanged. Drawbacks of such a topology can be heat
transfer from the brake to the motor, which demands a higher performance of the cooling
system and can, in extreme cases, also harm the magnets. Moreover, such motors demand
high diameter seals, which are usually expensive to make, complex to install and produce
high energy dissipation and high braking torque. As the seals have a large diameter, the
lowermost part of the seal is much closer to the road than a small diameter seal, which means
a higher probability of the ingress of water when sealing is not perfect. But on the other hand,
these types of seals allow such a design that the condensed water can exit either at standstill
or at high speed. Two seals are necessary for this solution as the motor has to be opened on
the outer side of the wheel to ensure the proper cooling conditions for the brake. The
described solution can be implemented around different types of the brakes [18-25].
A motor with a disc brake and a caliper between the motor and the attachment plate
(Figure 15) boasts the advantage of good cooling capabilities for the disc brake and brings no
stator lower radial dimension boundary constraints, as the caliper‟s upper boundary can
obviously be on a larger radius. Moreover, the arrangement of this kind of motor is very
convenient, but maintenance of brake discs and changes of brake pads can be complicated in
some implementations. A seal can be on a smaller diameter than in the previous solution,
which means lower friction force and low energy dissipation. The rotor plate can be closed on
the outer axial side of the wheel so only one seal can be used and even this one on a small
diameter, consequently causing little energy loss. Such a topology of the drive is axially wider
than the previously described one.
Figure 15. The disc brake mounted between the motor and the attachment plate/mounting point.
The already mentioned energy loss of the seal is also presented by the Equation (5),
where TB is braking torque, kfr is a coefficient of friction, Fpl is a preload force and R is the
radius on which the seal is mounted.
TB  k f Fpl R (5)
It is important to note that the preload force is proportional to the circumference of the
seal as presented in Equation (6):
Fpl  2 R (6)
Based on Equations (5) and (6) we can conclude that the braking torque TB is
proportional to the R2, clearly supporting engineers‟ aims to mount the seal on the lowest
radius possible and with a lowest preload possible, while still ensuring the appropriate sealing
properties.
The solution with the brake on the inner side of the attachment plate (Figure 16) shares
the majority of properties with the solution where the brake is between the attachment plate
and the motor, but is axially even wider, so that the steering constraints are much more
limiting.
Figure 16. The disc brake mounted on the inner side of the attachment plate.
An unconventional solution for refined products is a motor with a brake disc attached to
the rotor and a caliper attached to the strut of the stator. One large seal is required for such an
application but the disc and the caliper are unconventional, which requires a development and
homologation of a new concept of the brake. Unfortunately, the disc (and in some topologies
the caliper) is thermally coupled with the rotor. This can be a limiting factor for the
performance of the motor as too much heat from the disc could dissipate to the magnets.
The brake is mechanically outside and uncoupled with the motor and the rim. The main
advantages of such a construction of the motor are two seals on a very small diameter
producing the lowest amount of friction of all the solutions presented so far. Such a motor,
called Elaphe*3, requires a two-sided mounting. It is therefore not usable in modern cars but
has shown a great reliability as a scooter motor.
The solution of the brake between the motor and the rim is an absolute winner from the
point of view of the produced total friction, as it requires only one seal on a small diameter
but faces issues with a caliper arm design.
6.2.2. Inner Runner Motors

Generally speaking, motors with an inner runner can be slightly more difficult to produce
as the winding faces towards the center of the motor so that there is less space for the tooling
and machine manipulation. Moreover, the magnets are attached to the outer side of the rotor,
which means the centrifugal force tries to tear them off. This way, in case of glue (or other
attachment means) failure, the magnet is thrown into the airgap, where it is ground by the
friction between the stator and rotor and the motor loses its power and gets jammed. In case
water cooling is required for a particular type of motor, the motor with inner runner is also
less favorable. As we are speaking about in-wheel motors we have to be aware of the outer
diameter limit. It is defined by the rim size and, as the cooling system of the stator is on the
opposite side of the winding, it means the winding, and consequently the air gap, have to be
on a lower diameter. With this, some torque potential of the motor is lost. It is true that the
total cooling area can be larger in such cases, but it is usually not a critical factor.
There is a strong analogy of the properties, advantages and drawbacks between the
designs of motors with inner and motors with outer runners. As there is a clear logic behind it,
different types of inner runner motors are only graphically presented in Figure 17, while their
properties are presented in Table 6.
Figure 17. The brake inside the motor – the inner runner design.
7. SOME TYPES OF THE IN-WHEEL MOTORS

Different types of vehicles have their own specific requirements when talking about the
in-wheel motor integration, especially when the prototype motors are produced for the
conversion of the stock vehicle and when motors have to be developed for a newly designed
electric vehicle [14]. The boundary conditions for an in-wheel motor integration are usually
the following:
 Inner diameter of the rim;

 Outer diameter of the brake (usually the limiting factor is a drum brake or a disc
caliper);
 Attachment points of the motor
o Specific original mounting points on the car or specific knuckles
o Specific bearings and their integration;
 The motor has to be adjusted to the braking system of the vehicle, as newly designed
brakes are very unfavorable in the automotive industry, also because of the
homologation requirements and costs;
 Heat dissipation from the brake should not be limited after the integration of an in-
wheel motor. Heat should also not be transmitted to the in-wheel motor where it
could limit its performance or even cause damage. When there is no other design
option, the heat from the brake can only be dissipated to the motor‟s cooling system
instead on the active parts, were magnets are the component most at risk.
For research purposes we have designed, produced and tested several prototypes of in-
wheel electric motors with inner or outer runners in order to recognize specific challenges and
advantages of every single design and its compatibility with certain designs of a braking
system.
A closer look at Table 6 and Figures 15-17 shows us that, generally, a motor with an
inner runner provides much more conventional options for seal integration and its full
performance. It is also more compact, stiff and provides better options for the cooling system
integration but its cooling system cannot absorb the heat dissipated by the brake.
Table 6. Reviewing some types of the in-wheel Elaphe motors
EM
Example Key advantages
Solution
Elaphe*Chebela1
Elaphe*Smart1
Elaphe*Smart2
Compact design, access to the
Outer Brake inside the Protean Electric in-wheel
brake without removing the motor
runner motor Siemens VDO e-corner
from the car
Schaeffler E-wheel drive
e-Traction The Wheel
Fraunhofer hub motor
Easy rim mounting/removal, small
Brake between the
seal diameter and consequentially
attachment plate and Elaphe*Chebela2
low friction, only one seal, simple
the motor
brake pads change
Brake on the other Robustness, brake thermally
axial side of the decoupled from the motor, easy
attachment plate brake pads change, only one seal
Brake (disc brake)
attached to the rotor, Low axial space requirement, low
Elaphe*Sports Car
caliper attached to weight, only one seal
strut or stator
More simple realization of
Brake outside the
Elaphe*3 (scooter motor) environmental protection for the
motor and the rim
motor
Only one seal used, not much
Brake between the
additional weight, caliper from the
motor and the rim
inner side
Slim
Elaphe*1 and Elaphe*2
Bearings
Compact design, access to the
Inner Brake inside the
brake without removing the motor
runner motor
from the car
Brake disc attached to
Only one seal
the rotor
Elaphe*Hycat3 ATV
Brake on the other Robustness, thermally decoupled
motor
side of the attachment from the motor, easy removal of
Elaphe*Hycat1
plate the brake, only one seal
Elaphe*Fido motor
Brake outside the Only one seal, thermally decoupled
motor and the rim from the motor
Brake between the
motor and the rim
In reality, it turns out the brake is usually much more accessible in the motor with an
inner runner and the adjustability for different types of unconventional vehicles (e.g., with
tracks) is higher. The cooling abilities for the magnets turn out to be worse than in the case of
an outer runner for the following reasons: the total outer surface is smaller, circulation of the
air is at a lower rate than at the outer side of the motor and in cases when the brake is
integrated into the wheel itself, the probability for heat ingress to the magnets is higher than
in the case of an outer runner [16, 18, 19, 24, 25].
Crucial factors when talking about the mounting specifics of an electric motor are the
type and position of the position sensor (usually hall sensors, incremental or absolute
encoder). Its type depends also on the type and configuration of the inverter.
In both cases (inner or outer runner) a braking system which is thermally decoupled from
the motor is preferred. The easiest way is when the brake is mounted on the inner axial side of
the motor, though space saving advantages of the direct drive in-wheel motor are the most
obvious when the brake is integrated directly inside the wheel.
CONCLUSION
Mobility trends in western societies (with eastern following quickly) are obviously
moving towards technologies of cleaner personal transport and concepts of small, compact
and agile vehicles with a high level of space utilization. The concept of the direct drive in-
wheel motor is obviously the brightest star on the horizon for the mobility of the coming
decades, not only because of the technological advantages presented in this section, the high
efficiency, high power/weight and torque/weight ratios and a very competitive production
cost, but also because of its unbeatable convenience for the user and society. The research and
development trends are therefore expected to notably increase in the field of unsprung mass
of the vehicle‟s components as well as in the field of great space utilization from the point of
view of the active parts of an electric motor. In the majority of vehicles, all active components
in a motor have to be implemented on a relatively thin radial space between the brake and the
rim, which suggests a notable increase in usage of the high number of pole pairs in permanent
magnet motors. Moreover, from the electrical point of view, the present lack of standard
inverters for high currents in a relatively specific frequency range is expected to accelerate
and deepen the research in this field, in order to provide the demanded supply of controlling
units for electric motors in different segments and topologies of the electric propulsion
systems.
REFERENCES
[1] Gerling, D.; Dajaku, G.; Lange, B.; Electric Traction for Automobiles – Comparison of
Different Wheel-Hub Drives. University of Federal Defense Munich; FEAAM GmbH,
Munich & Neubiberg, Germany.
[2] Pavasovic, U.; Internal Combustion Engine. http://picture_Internal_combustion_engine.
[3] Lampic, G.; Slivnik T.; Detela, A.; Holistic Approach in Developing Propulsion System
for Urban Electric Vehicles. IEEE Vehicle Power and Propulsion Conference, Chicago,
USA, 2005.
[4] Anderson, M.; Harty, D.; Unspring Mass with In-Wheel Motors – Myths and Realities.
10th International Symposium on Advanced Vehicle Control, Lough borough, UK,
2010.
[5] Michelin Active Wheel. Paris Motor Show, Paris, France, 2008.
[6] Schalkwyk, D.V.; Kamper M.; Effect of Hub Motor Mass on Stability and Comfort of
Electric Vehicles. University of Stellenbosch, Stellenbosch, South Africa.
[7] Kim, D.; Shin, K.; Kim, Y.; Cheon, J.; Integrated Design of In-Wheel Motor System on
Rear Wheels for Small Electric Vehicle. World Electric Vehicle Journal, vol. 4, pp. 597-
602, 2010.
[8] Wang, P.; Guang Ning, W.; Gao, B.; Luo, Y.; Cao, K.; Study of Partial Discharge
Characteristics at Repetitive Square Voltages Based on UHF Method. China
Technological Sciences, vol. 56, no. 1, pp. 262-270, 2013.
[9] Luo, Y.; Tan, D.; Study on the Dynamics of the In-Wheel Motor Systems. IEEE
Transactions on Vehicular Technology, vol. 61, no. 8, pp. 3510-3518, 2012.
[10] Muller, S.; Uchanski, M.; Hedrick, K.; Estimation of the Maximum Tire-Road Friction
Coefficient. Journal of Dynamic Systems, Measurement, and Control, vol. 125, no. 4,
pp. 607-617, 2013.
[11] Besselink, I.; Nijmeijer, H.; Influence of In-Wheel Motors on the Ride Comfort of
Electric Vehicles. Eindhoven University of Technology, Department of Mechanical
Engineering, Dynamics &Control, Eindhoven, The Netherlands, 2010.
[12] Janulevicius, A.; Pupinis, G.; Damanauskas V.; Effect of Tires' Pressure on the
Kinematic Mismatch of a Four-Wheel-Drive Tractor. Mechanika, vol. 19, no. 1, pp. 73-
80, 2003.
[13] Hartani, K.; Bourahla, M.; Miloud, Y.; Sekour, M.; Electronic Differential with Direct
Torque Fuzzy Control for Vehicle Propulsion System. Turk Journal of Electrical
Engineering and Computer Science, vol. 17, no. 1, pp. 21-38, 2009.
[14] Gaing, Z.L.; Chiang, J.A.; Robust Design of a Direct-Drive PM Hub Motor by Fuzzy-
Inference Taguchi Method. International Journal of Applied Electromagnetic sand
Mechanics, vol. 36, pp. 295-307, 2011.
[15] Yirka, B.; SIM-Drive Corp Announces New 'In-Wheel' Electric Car.
http://phys.org/news/2011-04-sim-drive-corp-in-wheel-electric-car.html.
[16] Siemens VDO; Technology for the Environment – Environmentally Friendly
Transportation. www.siemens.com/innovation/en/publikationen/publications_pof/pof_
spring_2007/technology_for_the_environment/green_transportation.htm.
[17] Schaeffler; The Drive for the City Cars of Tomorrow – Schaeffler Displays Wheel Hub
Drive E-Wheel Drive. www.schaeffler.de/content.schaeffler.de/en/press/press-
releases/press-details.jsp?id=23019393.
[18] TM4; Electric Corner Module.www.tm4.com/en/electric_corner_module.aspx.
[19] Evans Electric; The Automotive Mega Trend of the 21st Century –The Electrification of
the Automobile. www.evans-electric.com.au/.
[20] E-Traction; TheMotion.www.evans-electric.com.au/.
[21] Whitehead A.; Plug-in Hybrids Made Easy - Selectable EV Drive Using In-Wheel
Motors. Protean Electric Ltd, Farnham, UK, 2011.
[22] Rahman, K.M.; Application of Direct-Drive Wheel Motor for Fuel Cell Electric and
Hybrid Electric Vehicle Propulsion System. IEEE Transactions on Industry
[23] Busse, M.; Development of Components for Electric Vehicles. Fraunhofer Institute for
Manufacturing and Advanced Materials (IFAM), 2010.
[24] Kock, A.; Mertens, A.; Fault Tolerant Wheel Hub Drive with Integrated Converter for
Electric Vehicle Applications. IEEE Vehicle Power and Propulsion Conference, Seoul,
Korea, pp. 19-23, 2012.
[25] Juris, P.; Brune, A.; Ponick, B.; A Coupled Thermal-Electromagnetic Energy
Consumption Calculation for an Electric Vehicle with Wheel Hub Drive Considering
Different Driving Cycles. IEEE Vehicle Power and Propulsion Conference, Seoul,
Korea, pp. 28-31, 2012.
Chapter 19
THE FUTURE OF ELECTRICAL VEHICLES
Wim J. C. Melis
University of Greenwich at Medway, Kent, UK
ABSTRACT
Electrical Vehicles (EV) entered the transportation market much earlier than their
combustion engine driven counterparts. After a long period of no development, currently
they have found their way back into use, but then mainly as small goods vehicles on large
estates where they then often carry a label in the sense of “Zero Carbon Vehicle”. Due to
the increasing social pressure for environmentally friendly transportation, their zero
emissions comes to their benefit also for mainstream goods and person transport.
However, there are still a few problems to be resolved to smoothen the path for a better
market penetration. Even though the electrical motor has a much higher efficiency, in
comparison to fossil fuel powered vehicles, overall, electrical vehicles suffer from a
limited range due to the lower energy density for electrical storage. Current hybrid
vehicles provide a solution to this problem by using an internal combustion engine used
as a range extender. The environmental impact of this range extending motor as well as
the actual electricity production itself both have an environmental impact unless green
electricity is being used as an energy source and used to charge the plug-in hybrid. The
environmental impact depends on the source-fuel being used, but in essence highlights
the importance of finding more environmentally friendly energy sources. Additionally, it
is essential to improve the overall efficiency of our transport vehicles and any energy
conversions that take place during transport, by for example recovering the kinetic energy
of the vehicle when it is being decelerated. As much as there has been a move towards
electrical vehicles, there has been a move towards more Information and Communication
Technology (ICT) within vehicles for a variety of reasons, moving towards electrical
vehicles the integration of ICT with the actual vehicle is likely to become easier and
therefore more efficient allowing for vehicles to become fully autonomous.
Keywords: Fuel efficiency, environmental impact, hybrid vehicles, bio-fuels, alternative

fuels, energy storage for transport

E-mail: wim.jc.melis@gre.ac.uk.
510 Wim J. C. Melis
1. INTRODUCTION
Even though electrical vehicles have been around since the early 20th Century, they did
not quite make it to the mass market for a variety of reasons. Looking at the current situation,
there are a number of reasons why they are more likely to break through now than they ever
could have in the past. One of them is the environmental concern in relation to fossil fuel
powered vehicles, which are generally considered to be none environmentally friendly. In the
past, vehicles with lower consumption have been presented, but they never quite made it to
the market. Within the search towards more environmentally friendly vehicles, hybrid and
electrical vehicles have been considered to form the solution. This is partly related to the
efficiency of an electrical vehicle being rated at 80-90% [1] in comparison to the efficiency of
fossil fuel cars, which is only 25-30% [2, 3]. One however needs to look at the full picture
before drawing any premature conclusions about the environmental friendliness of the
electrical vehicle, since a lot depends on how the actual electricity is generated.
One way to improve the integration of electrical vehicles has been the introduction of
hybrid vehicles, and these are gradually being developed towards plug-in hybrid vehicles.
This should allow for the vehicle to be charged by green power, while still managing a long
range due to the presence of another engine that provides for added acceleration and range.
Being able to charge the vehicle with green energy does not only make it more
environmentally friendly, but also helps with the need for grid storage when renewable
sources are being used for energy generation. Hybrid vehicles as such, benefit from the
advantages of two worlds when it comes to the acceleration, however, they do not manage to
fully recover energy during deceleration. Hence, fully regenerative breaking is being
presented, its efficiency investigated and advantages discussed [4].
Current hybrid vehicles have, besides an electrical motor, another energy source that
helps during acceleration. This other energy source is generally a combustion engine, or
alternatively a Fuel Cell can be used when the energy is stored as Hydrogen rather than using
batteries. The use of batteries as well as fossil fuels however still has an environmental
impact. Therefore, the chapter looks at alternative fuels as well as energy storage
mechanisms.
After a short look into the past and present developments of Electrical Vehicles, the first
section of this chapter will look into the potential future of the electrical vehicle, and why
now it is more likely than ever that electrical vehicles will become mainstream. The next
section then looks at current vehicles and their efficiency, comparing EVs with Combustion
Engine (CE) vehicles, how to improve their efficiency and also their respective impacts on
the environment. The final section then introduces some alternative fuels and storage
mechanisms that could be used to overcome some of the problems that EVs currently face.
2. PAST, PRESENT, FUTURE

Electrical vehicles have been around since the early 20th Century. Their history can be
divided into a few stages. The very first transportation vehicles were in essence electric and
they were well ahead of their combustion engine counterparts. In the early 1920‟s to 1930‟s
there was actually a 3:1 domination in relation to EV registration when compared to their
The Future of Electrical Vehicles 511
gasoline counterparts [5]. From 1930 onwards, gasoline vehicles overtook the electrical
vehicle domination, because of the following reasons: 1) mass production at a reasonable cost
which made personal transportation available to the general public; 2) a lack of charging
infrastructure combined with the limited range for electrical vehicles; and 3) mass discovery
of gasoline resulting in fuel being available in abundance and therefore cheap. From about
1935 till the 1970‟s onward, there was political sensitivity with the Organisation of Petroleum
Exporting Countries (OPEC), and when the oil-crisis hit in the 70‟s only about 300 Electric
Vehicles were purchased for fleet testing by the US postal service.
Most other countries did not consider EVs as an option, but purely focused on reducing
fuel consumption. Other from this small order, the political and economic environment at the
time did not stimulate any further development. Towards the end of the 90‟s, a similar, but
slightly larger design for EVs, to be tested in a fleet setting, was performed. Several
manufacturers either altered existing vehicles to sell them as EV, or designed new vehicles.
Even though these vehicles proved to be working well, all of them were rather quickly
discontinued, and most of them have meanwhile been destroyed and recycled.
According to a prediction for the next 30 years made in 1999 [6] the hybrid as well as the
hydrogen vehicle would have the longest market future, because they deliver the same
functionality as our current vehicles at a higher fuel economy for a higher cost.
The only issue to resolve is to find the correct balance between cost and economy.
Considering we are about half way through this 30 year prediction, it seems that hydrogen
vehicles are still in the prototype stage, however, Toyota has already announced to provide a
production model on the market in 2015 [7]. On the other hand, hybrid vehicles have started
to take off, as can be seen from the large set of vehicles already provided currently, as well as
the ones in the pipeline to market [8]. While Hybrid Electric Vehicles might still have a
slightly larger sale price, over their lifetime, they benefit from a better fuel economy when
compared to combustion engine based cars [9]. On top of that, they also benefit from the
increase in social responsibility in relation to environmental concerns and consequently
government has started to support this technology more substantially than it ever did in the
past.
The further success of the electrical vehicle now depends on a variety of different factors,
not all of them technical. First of all, there is the requirement for good education around EVs
to ensure that people understand their benefits over the existing systems. Secondly, there is
the financial support by the government to overcome the initial higher investment
requirements related to this “new” technology hitting the market. The duration of this
financial incentive might enable the technology to take off or, if stopped too early, stop
market penetration while still at a crucial stage. As in relation to the social responsibility, the
benefit of the electrical vehicle is that the “fuel”, i.e. electricity, can be generated in a variety
of different ways.
As mentioned further on in this chapter, unless the fuel production is green, the electrical
vehicle does not justify some of its currently green image, but considering that electricity
generation is changing that will certainly become the case in the not so distant future. On a
technical side there are challenges in relation to: energy storage, range expansion and a
charging infrastructure to be dealt with. However, the energy stored within the vehicle could
actually become a benefit in relation to the increased use of renewable energy sources. The
problem with this kind of sources is that they are less “reliable” in the sense of not providing
a continuous supply. For example, sunlight cannot be guaranteed, and the same applies to
512 Wim J. C. Melis
wind. In most cases, the time when electricity is generated will not be the time it is actually
needed; therefore temporary storage of energy is essential. EVs could provide the ideal
energy storage required as we move to more renewable energy sources [10]. This approach is
generally referred to as Vehicle-to-Grid (V2G) integration and has been studied in several
countries that plan to introduce more renewable energy generation [11]. The principle is
generally that the EV form a flexible load to the network so that it can charge when there is
an excess of energy and when required they can put energy back into the grid to compensate
for a shortage. To allow this to happen, certain intelligence would be required that could
either form part of the grid or of the actual electrical vehicle itself [12]. This intelligence
would not only need to take the grid into account, but also the specifics of the EV storage
system, e.g. battery type, and the required urgency of the actual charging [13].
The influence of ICT on vehicles has been significant, and this will be even more the case
for EVs, which will be even more ICT driven/controlled [14]. The move towards more ICT in
current vehicles has had several reasons. Firstly, there is the engine management system,
which due to the increasing performance and exhaust requirements has led to the need for a
more precise control of the actual combustion cycle. Hence, current engines have many
sensors to measure a variety of characteristics in order to improve the overall engine
performance.
The second reason has been the continuous drive to increase “luxury” within the vehicle,
which includes items like multimedia equipment, navigation systems among systems like
driving aids (i.e. cruise control) and so on. Although several of these systems have already
been available for a number of years, they are currently all being integrated and combined
with one another to improve the overall user experience and friendliness. This development is
combined with the continuous drive towards more intelligent vehicles, which started with the
so-called drive by wire, and is currently being driven by a move towards fully autonomous
vehicles. The latter make extensive use of different “measuring” mechanisms generally
combined with a large amount of processing power in order to replace the human driver.
Although a large skepticism exists when it comes to allowing computers to drive our vehicles,
it is said that if computers would be driving, then that would eliminate all our traffic jams due
to the possibility of more efficiently usage of the road [15].
One approach that could overcome some of the initial problems with these drive-by-wire
systems is the move towards so-called semi-automatic systems, where a car would platoon
onto a set of cars [16]. A professional driver drives the vehicle at the front of the platoon and
other vehicles would join a platoon based on their destination. This would allow for a driver
to be able to relax/work part of the journey whereas a certain part might still need to be driven
by him/her-self. This would not only provide benefits to economy, but could also lead to a
reduction in exhaust, as long as the energy required to deal with the processing for this
autonomous driving is not excessive. Reducing the amount of energy required for
autonomous driving could come from more intelligent processing, which could be achieved
by either different algorithms and/or processing platforms.
The problem with current processing platforms is that for most algorithms used for
autonomous driving, a large set of calculations need to be performed before an actual result is
obtained, and this result can then still have a significant number of false positives. To
overcome these problems more intelligent systems would be required. Even though neural
networks have promised such intelligence, they have not shown to be practical for a variety of
reasons. Considering current technological developments, there are however a number of
promising paths as in relation to achieving this intelligence [17, 18, 19] however a detailed
study of this subject would be beyond the scope of this chapter.
3. EFFICIENCY OF EXISTING VEHICLES

3.1. Fuel Efficiency
Energy or fuel efficiency for vehicles generally looks at the conversion from energy
source or fuel into kinetic energy or vehicle movement. Considering this aspect, electrical
vehicles are said to be running at a much higher efficiency than fossil fuel powered cars. This
does apply if one considers the definition of fuel to be the fuel on board of the vehicle.
For fossil fuel powered vehicles, the fuel or energy source would be considered to be
either petrol or diesel. Depending on which one is being used, the efficiency of turning that
into kinetic energy is about 25% or 30% respectively [2, 3]. In contrast with that, electrical
vehicles have an efficiency of somewhere between 80 and 90% [1], if one considers the
conversion of electrical energy into kinetic energy. This higher efficiency is due to the fact
that the electrical motor has been improved to reach these high efficiencies. So even if the
energy density of fossil fuels may not be bad, there is a substantial amount of energy that is
wasted, since it is converted into heat. Although the electrical vehicle seems to imply that
these losses are smaller, it is essential to look at the complete picture, which requires an
investigation of the efficiency of the actual generation of electricity.
Electricity generation can be performed in a variety of different ways and depending on
the method used, Electrical vehicles will be more or less efficient and environmentally
friendly. In order to make a fair comparison between fossil fuel vehicles and electrical
vehicles, we will assume that the electricity is generated from fossil fuels, which could be
through coal or gas powered power stations. Since coal was used for steam engines and most
of these have made it to the museum by now, we shall look at gases as an example fuel for
our calculation. Electricity generation from conventional fossil fuels has an efficiency of
between 33 and 60 %, depending on whether the source fuel is coal/oil or gas [20, 21]. The
reason for this range being rather large is that for pure steam driven power generation the
efficiency would be about 30 to 40%. However, if a gas powered station makes use of the so-
called combined cycle then efficiencies up to 60% can be obtained, which is due to the
extraction of energy from the exhaust gases as shown in Figure 1.
Before the electricity can be stored in the batteries of the electric vehicle, it still needs to
travel a long way.
First of all, it needs to be distributed across the grid, where it incurs a certain amount of
transport losses. These transport losses would add up to an average of about 8% [22]
depending on factors like distance from power station and so on. Hence, the efficiency of
generating and transporting the electrical energy to the vehicle, assuming average values for
any range provided, would be: 0.50*0.92=0.46. This leads to an overall efficiency from fuel-
source, i.e. gas, into kinetic energy of: 0.46*0.85=0.39, which still assumes a perfect
efficiency for the charging and discharging of the vehicle‟s batteries. The overall efficiency of
a battery again depends on a variety of factors, namely the type of battery, type of charger,
number of cycles it has been charged/discharged, and so on. However, efficiency numbers of
514 Wim J. C. Melis
85% are not uncommon for the types of batteries used in current electrical vehicles. This
results in the overall efficiency from gas through to electricity and all the way to kinetic
energy to become: 0.39*0.85=0.33 or 33%, which is only marginally better than our diesel
powered car, but quite a bit better than the petrol powered car. However, we have not quite
compared like with like.
In order to be completely correct, the efficiency number calculated above should be
compared with a fossil fuelled car that combusts natural gas, which generally has a slightly
higher efficiency than when it runs on normal petrol, although the difference is not massive.
Exact numbers of gas powered vehicles are not wide-spread due to the small number of
vehicles running on this fuel, but it is generally known that natural gas has a higher efficiency
and therefore an engine combusting gas will have a slightly higher efficiency. This higher
efficiency, which relates to the higher fuel burning efficiency of gas, can be derived from its
octane number. The octane number for gasoline is known to be 95, whereas natural gas would
have an octane number of about 130. These numbers can be taken as a measure of how the
efficiency of the engine will change depending on which fuel it burns. Therefore, the
efficiency improvement can be calculated to be: (130/95)*0.25=0.34. Therefore, as from an
efficiency perspective, is seems that burning natural gas in a combustion engine directly is as
efficient as using it to generate electricity, which is then used to drive an electric vehicle.
One might wonder why these efficiency values are so close to one another, especially if
one considers that when using electrical energy, there are quite a few energy conversions
being performed. Well, other from the fact that the efficiency numbers for our electricity
generation are rather ambitious when it comes to the power station efficiency, the engine in
our vehicle will continuously be accelerating and decelerating, which affects fuel efficiency
in a negative way. This clearly provides an opportunity for improving energy efficiency,
which will be looked at in more detail later on.
The above comparison has been fully based around one particular fuel, namely gas. There
are however other fuels, like diesel, bio/diesel and so on. Using diesel as a fuel provides a
fairly similar story, only that diesel engines are generally slightly more efficient due to the
higher calorific value of the fuel itself. Bio-fuels on the other hand will again slightly change
the efficiency aspects, and are discussed in more detail towards the end of this chapter.
Figure 1. Working principle of combined cycle gas plant: (1) Electrical Generators; (2) Steam Turbine;
(3) Condenser; (4) Pump; (5) Heat Exchanger; (6) Gas turbine [23].
Besides the burning of fossil fuels, there are also other methods of generating electricity,
which might provide an overall better efficiency. The first one that would need to be
considered due to it being rather wide spread is nuclear power, which is quite often classified
as a controversial method of energy generation, because, as much as it is being phased into
some countries [24], it is being phased out of others [25, 26]. Trying to determine overall
efficiency of a nuclear power station is not that straight-forward, since the general measure of
looking at input fuel value in comparison to electrical energy output cannot simply be
applied, since it is rather difficult to determine the energy value of the input fuel. Due to its
controversial nature and the produced waste, depending on the source being listened to it will
either be considered as a green way of generating energy or not.
However, considering that sustainability is often described as: “living on this planet
without leaving any trace for future generations”, one could question whether nuclear power
does comply with this definition, due to the waste products generated. Therefore, in order to
move to a more green and sustainable transport, nuclear power does not seem to be compliant
with this image.
This brings us to green energy generation, which would generally comprise: solar, wind
and hydro as energy sources. Either directly or indirectly most of these types of energy relate
back to the energy provided to this planet by the sun, and considering that the sun provides
more energy in one minute to this planet than we currently consume in one year, it seems that
this is the more prominent way towards sustainable energy collection. Considering that solar
energy is still provided to us free of charge and has no negative impact on the environment,
efficiencies should not play such a major role when considering this energy source other than
from the possible return on investment. It is however also interesting to compare these newer
ways of energy collection with existing energy generation methods. Since solar and wind are
the easier ones to install everywhere around our planet, these will be looked at in more detail.
The efficiency for solar panels largely depends on the type of solar panel being used, however
average efficiencies would range between 10 and 30% [2]. Considering that the theoretical
efficiency limits of a single cell are at 30%, multi-layered structures will be needed to reach
higher efficiencies.
Depending on the number of layers, these efficiencies can go up to 86% for concentrated
sunlight [27]. On the other hand, wind turbines would have a maximum efficiency close to
60 %, but on average most practical turbines would have an efficiency of 35-45% [28].
Hence, as in comparison with fossil fuel generated electricity, these efficiency percentages are
not that different, however these technologies might allow for the electricity to be generated
closer to the actual consumption and consequently reduce the distribution losses, not to
mention that the energy came completely free in the first place. However, these sources are
generally not continuous and cannot be switched on/off, therefore a storage system is of
significant importance, for which EVs could be useful. When one considers the electricity to
be produced from a renewable source, then the overall efficiency from solar energy to kinetic
energy produced in the electrical vehicle is not that different from using fossil fuels.
However, since the actual generation is that much more sustainable, this also means that the
vehicle as such will also become much greener, as will be worked out in detail later on.
Needless to say that this is an incomplete list of energy generation methods, but
considering that this book deals with electrical vehicles and not with energy generation, this
would bring us outside the scope of this book. It is however important to notice that an
electrical vehicle is not much more efficient than the current fossil-fuelled vehicles.
516 Wim J. C. Melis
Table 1. Efficiencies of energy conversions
Conversion Process Current Efficiency (%) Physical/Theoretical Limit (%)

Gas/Coal to electricity 30 - 40 -
Gas (combined cycle) to electricity 60 -
Solar - Photo Voltaic 10 - 30 86
Wind Turbine 35 - 45 60
Fuel Cell 60 - 85 -
Combustion Engine - Petrol 25 -
Combustion Engine - Diesel 30 -
Combustion Engine - Gas 34 -
Electrical Motor 80 - 90 -
However, if the electricity generation is greener, then using an electrical vehicle suddenly
becomes more beneficial, but more on this later. A summary of the different efficiencies can
be found in Table 1.
3.2. Increasing Efficiency
In an ideal world, all efficiencies would be 100%, but as seen in the previous section, this
is rarely the case for any type of energy conversion. The question is however whether these
efficiencies can be improved, as has been the case for the electrical motor where a significant
amount of research has led to efficiencies of up to 90% and sometimes even higher. When
considering electrical vehicles there are however a large set of factors involved, and the
largest improvements would clearly be reached by looking at the factors, which currently
have the lowest efficiencies.
In order to improve overall efficiency, the first factor that needs to be considered is the
actual energy generation. Vehicles running on fossil fuels have been around for quite a while
and although some improvements are still being discovered, very few of those actually reach
the mass-market for a variety of reasons. When looking at electricity generation based on
fossil fuels, then one can notice that this is largely based around steam turbine technology,
which has been running close to optimal efficiency for quite a while now. On the other hand,
there are the green technologies, which are still being improved on continuously, although
also these methods will run against their physical limitations one day sooner or later.
The next factor where improvement is possible is the actual electricity transportation or
distribution. With losses being on average only 8%, there is very little room for optimisation.
However, any optimisation that can be achieved would have to come from introducing more
localised generation, and therefore shortening the distance between generation and
consumption.
Another factor is the actual energy storage in batteries. Battery technology, especially
liteon, is continuously improving, which results in improved charge/discharge cycle
efficiencies. However, the efficiencies for this factor were already quite high and can
therefore only improve marginally.
When considering the actual use of vehicles, then a significant amount of energy is used
to accelerate the vehicle itself. However, in many traffic situations, like city traffic,
deceleration is as frequent as acceleration, which is generally reflected in higher fuel

consumption. As much as acceleration generally consumes energy, deceleration tends to
waste energy, because during braking, kinetic energy is converted into thermal energy.
Considering an average journey, the majority of fuel will be used for acceleration and
only a minor amount of energy will be required to keep the vehicle at speed. However, at
some point in time, the vehicle will need to decelerate again and rather than wasting the
kinetic energy it would be useful if this energy can be stored to reuse during acceleration.
Obviously any of these conversions from kinetic to storage and back to kinetic might not
have perfect efficiency, but overall, the effect of this temporary storage should still be better
than converting all kinetic energy into heat. Hybrid vehicles perform this conversion by using
the electrical motor as a generator during deceleration in order to recharge their batteries. This
system is however mounted besides a conventional hydraulic break system for several
reasons, the first being safety, the second one being purely technology related. A single
generator would namely not be able to bring the vehicle to a full stand still in a sufficiently
short timeframe for what could be described as emergency braking.
Hence, to be able to use regenerative braking to stop a vehicle completely a few problems
need to be overcome. Firstly, there is the safety aspect, which leads to the requirement that
the vehicle needs to be able to stop in case of emergency. This problem can be overcome by
over-designing the system so that it can deal with emergency cases, including when the
vehicle is fully loaded or even overloaded. Vehicle load plays an important part in braking
since the force that needs to be countered during the braking action depends on vehicle mass
and speed. Secondly, an electrical generator has certain efficiency at a certain rpm. This
means that below this threshold rpm, the generator will no longer convert kinetic energy into
electrical energy and therefore no longer extract kinetic energy from the system at all. This
means that for low rpms, it becomes more difficult to bring the system to a full stop.
Considering that from the safety aspect the system needs to be overdesigned that would
mean that unless the vehicle is overloaded and running at maximum speed, the vehicle users‟
would experience a shock when the generator is switched on for braking because the
generator will extract the maximum amount of kinetic energy and therefore immediately
perform a maximum braking force, whereas the actual mass of the system is lower and
possibly also its velocity, and therefore the vehicle would slow down very quickly. This step
response approach is clearly far off from the more gradual braking that currently applies
within our vehicles, therefore a more gradual mechanism for the regenerative braking needs
to be developed.
The work presented in [4] considers these two problems and presents a solution that
allows for fully regenerative breaking. The problem of generator efficiency is overcome by
using multiple generators, each one being connected to the main driving shaft through a
different gear ratio. This method ensures that there is always at least one generator within its
optimal rpm range and therefore able to extract energy from the system. Figure 2 shows the
difference between using a single and multiple generators for regenerative braking, in this
case: 4. One should keep in mind that this diagram is slightly simplified, since the braking
slopes for the multiple generator case on the right-hand side should have slightly different
slopes depending on how many generators are extracting energy. This is due to the fact that
the amount of energy extracted from these generators depends on the rpm they run at, and
also on whether they are switched on or off, since the current model would not switch them
all on at the same point in time, but introduce them depending on the braking demand as will
518 Wim J. C. Melis
be shown next. Figure 2 clearly indicates that multiple generators lead to a better efficiency,
since a full range of vehicle speeds can be dealt with.
The second problem can be overcome by connecting/disconnecting the load to these
generators at a certain frequency. When this pulsing is done quickly enough then the average
load presented to the generator depends on the length of the period that the generator is
connected to the load as shown in Figure 3. Figure 3 clearly indicates that if this pulse is
shorter, then the average load to the generator is lower, whereas if the pulse becomes longer
then the average load increases. This increased load also means that more kinetic energy will
be extracted from the moving vehicle and therefore the braking force will be higher.
This approach has certain additional benefits that extend beyond the provision of fully
regenerative braking. The main advantage of this system is that there are benefits during the
actual acceleration since the different gear ratios now ensure a more continuous acceleration
independent of vehicle speed. There is also likely to be a higher overall efficiency due to the
fact that multiple generators are being used and each one of them is supposed to run at
optimal efficiency at some point in time. Calculating the overall efficiency will depend on
many different factors, which include: speed of vehicle, required slow-down, efficiency of the
individual generators and their gear ratios. Intuitively considering that there are multiple
generators one has to keep in mind that not all of them will be running at optimal efficiency.
However, overall the efficiency of the system should be higher than that of a single generator.
This higher efficiency does however come at the price of requiring more materials, since one
requires not only multiple generators, which from a generator perspective would be similar in
price to a single larger generator, but one also requires a larger mechanical construction, since
each generator needs to be connected to the main shaft with a specific gear.
Figure 2. Efficiency difference between using a single and multiple generators for regenerative braking.
Figure 3. Pulsing the connection between load and generator allows for a different average load to be
presented to the generator.
In order to fully judge the impact of these extra materials, one would actually need to
perform a full life-cycle analysis.
3.3. The Environmental Impact of Electrical Vehicles
Ideally, one would perform a full life-cycle analysis of electrical vehicles and compare it
with other types of vehicles to draw a proper conclusion on environmental impact of each of
the different types of vehicles [29, 30]. For example, if one looks at electrical vehicles, then
there would be one or multiple electrical motors, and each of these motors would require a
certain amount of copper, which is a limited, finite resource. On top of that, an electrical
vehicle will generally have a substantial battery bank installed, and these batteries will have a
limited life span, while probably making use of some limited resource materials like liteon,
among others, depending on which battery technology is being used. All of this will play a
factor in a full life cycle analysis of such vehicles, however looking at such an analysis, it
seems that electrical vehicles are not too bad off [29, 30].
One aspect of difference that will play part during the actual transportation is the energy
density of the fuel, since for fossil fuel cars the fuel is put in a tank and as the tank empties
the weight reduces. For an electrical vehicle, the weight of the storage system, i.e. batteries,
will not change, independent of the amount of energy stored. On top of that, batteries are on
average rather heavy and this weight also needs to be transported along with the rest of the
weight of the vehicle and will therefore have a negative impact on the overall fuel
consumption. Additionally, one should know that batteries have a lower energy density than
fossil fuels [31]. Although all of these factors play a role when looking at the full carbon
footprint of a vehicle, this particular section will however only look at the carbon footprint
related to the fuel used for transportation.
The current measure for carbon footprint is generally related to the produced amount of
CO2. For vehicles this would be expressed as a certain amount of gram CO2 produced per
kilometre. Needless to say that such a measure is limited, since it is not only CO 2 that causes
harm to the environment. Hence, there are debates on whether CO2 is a correct measure of
“damage to the environment”, but these discussions are beyond this book and for the purpose
of this exercise we will assume this can be considered as an appropriate measure. Due to the
heavy weighting of CO2 in most studies, this section will first focus on the aspect of CO2 after
which it will look at the other harmful by-products of burning fossil fuel.
As mentioned before, the environmental impact of an electrical vehicle largely depends
on the manner in which electricity is generated, which then again changes across the different
continents. Figure 4 shows the amount of CO2 produced per kWh of heat and electricity on
the different continents. From this figure, one can clearly notice that energy production as a
whole is not very environmentally friendly, and the amount of CO2 in the world has been
increasing. Considering that the electricity production does play a vital part in how green an
electrical vehicle is; it will be essential to ensure greener production to ensure that electrical
vehicles are greener than fossil fuel vehicles.
As an example, let us compare the amount of CO2 generated for a similar size vehicle
powered by either fossil fuels or electricity. Considering that the average CO2 for all new cars
in UK is 158 g CO2/100km [33] and that not all vehicles are new, a value of 160 g
CO2/100km is probably quite an optimistic measure.
520 Wim J. C. Melis
Figure 4. CO2 emissions per kWh of Electricity and Heat Output for the EU-27 level compared to the
world, United States, the Russian Federation and China [32].
Over time this number should go down though due to the target set within EU to reach
120 g CO2/100km by 2012. Considering that 2012 is, at the time of writing, just turned past
and this target has not been achieved, it seems that quite hard work will be required to catch
up with what now seems to have been an ambitious target. On the other hand, this average
size car would probably have a consumption of about 7 l/100km and this would relate to
about 70 kWh/100km based on the fact that one litre of fuel relates to approximately 10kWh
of energy [2].
When one considers the numbers from Figure 4 as an indication of the amount of CO2
produced per kWh of energy, and taking that the electrical vehicle would consume the same
amount of energy to move forward, then our first conclusion would be that the electrical
vehicle is way more polluting than our fossil fuelled vehicle. There is, however, good news,
which lies in the aspect that for a similar size vehicle we will need the same amount of energy
for kinetic movement. After all, the required amount of energy depends on: the weight of
vehicle, the front surface of the vehicle which is a measure of how much air needs to be
moved and so on. Considering that in the fossil fuelled vehicle only about 25 to 30% of fuel is
converted into actual kinetic energy, we can reduce the amount of electrical energy required
to be the same percentage of this 70 kWh, which would be between 17 and 21 kWh for 100
km. Taking that from generation to actual energy usage an electrical vehicle is about 33%
efficient, that would mean that the amount of energy to be generated by the power station is:
19/0.33=58 kWh, if we make use of average values. Since our current electricity generation
has a CO2 production of at least 400 to 500 g/kWh (See Figure 5, for data about Europe), this
would mean a total of 58*450=26100 g CO2/100 km for an electric vehicle, which seems to
indicate that electrical vehicles are way less environmentally friendly than their fossil fuelled
counterparts. Considering that the CO2 production of the electrical vehicle is mainly due to
the actual electricity generation, if this generation becomes greener, then the vehicle as such
will also become greener.
The burning process of fossil fuel, does however not only lead to the production of CO2.
Depending on the type of fuel, and the conditions of the burning process the actual by-
products found in the exhaust would vary slightly. However, most commonly the other by-
products are: NOx, Carbon Monoxide, Small Particles, Hydrocarbon and Oxynitride, which
are all in varying degrees responsible for a reduction in air quality, including smog, acid rain
and consequently influencing health [7]. NOx in combination with volatile organic
compounds leads to the creation of ozone, which when close to the ground can lead to health
problems especially for people with asthma and can even hurt plants and crops. NOx is a by-
product from any burning of coal, gasoline or any other fossil fuel, and would therefore be
generated in any combustion engine or power station running on fossil fuel. Carbon
Monoxide is mainly produced due to inefficient burning of fuel within combustion engines, it
is therefore highly essential that combustion engines are well maintained and tuned properly.
Small Particles have made it to the press quite a few times in relation to diesel engines
producing a substantial amount of them, which has consequently led to exhaust filters
becoming a requirement and/or catalysts being installed, the latter of which not only
eliminating the small particles from the exhaust [34]. Small particles are also formed when
burning fuel in power stations. Hydrocarbons, Oxynitride among other pollutants are all side
effects of burning fossil fuel, quite often in none ideal conditions.
Considering that an Electrical Vehicle requires the production of electricity, so in terms
of air pollution, the pollution would depend on how the electricity is generated. If the
electricity is generated from fossil fuels, then the difference is that in case of a power station,
the pollution is going to be created local to the power station rather than where the actual
vehicle is used.
Figure 5. Estimated CO2 emission factors for public electricity production in EU-27, 2008 [32].
522 Wim J. C. Melis
According to [35] when the electricity is generated from fossil fuel, then electrical
vehicles do not benefit over internal combustion vehicles, which is fairly understandable
since it is in essence the same fuel being burned and taking into account the overall efficiency
from source to end, the electrical vehicle is worse off than the one with an internal
combustion engine. This leads to the same conclusion as with the CO2 calculation, which is
that electrical vehicles will only have a smaller environmental impact if the electricity is
generated from a renewable/green source.
4. ALTERNATIVES
The previous section indicated that electrical vehicles are not really better than our
current fossil fuelled vehicles, and one of the main reasons for that is the actual electricity
generation. The results presented clearly indicated the influence of fuels being used therefore
one of the questions is whether there are any other fuels with better efficiency and lower
environmental impact. The other option would clearly be to reduce actual fuel consumption of
the vehicle itself and/or reduce the fuel consumed per person for transport, which could be
achieved by moving towards smaller, lighter vehicles overall and/or more mass transport.
4.1. Hybrid Vehicles
One of the reasons that has prevented electrical vehicles from breaking through into the
mass market is their limited range on a full charge. Although the majority of distances
travelled are on average between 15 and 45 km [36, 37], the limited ranges offered by
electrical vehicles, combined with the limited charging locations has led to a limited market
penetration for electrical vehicles. On the other hand, adding in electrical motors and batteries
to a car with a fossil fuelled engine, generally referred to as hybrid vehicles, is starting to
know a fairly good uptake but that did take quite a number of years.
Hybrid vehicles can be classified into two types. The first type is the serial type, where
the electrical motor is driving the vehicle and a combustion engine combined with a generator
provides the electricity. The energy source within this serial type does not necessarily have to
be a combustion engine, but could also be powered by hydrogen, which will be discussed in
more detail further on. The second type of hybrid vehicle is the parallel type, and here the
combustion engine as well as the electrical motor(s) can be used to drive the vehicle. An
example of the latter type is the Toyota Prius.
Needless to say that both types have their respective advantages and disadvantages. The
benefits of the serial type lies in the fact that the combustion engine can be designed and run
at a certain optimal rpm, which stands in contrast to using the combustion engine to drive the
vehicle itself. In the latter case, the engine needs to have good torque over a wide range of
rpm, which is not always easily achieved and inherently implies a disadvantage when it
comes to engine efficiency. An additional benefit is that due to the continuous energy source
being available, the battery bank can be kept small, whereas the range of the vehicle itself is
not being limited by the storage capacity of the battery bank.
A parallel hybrid benefits from being able to combine the advantages of both a motor and
an engine. The advantage of an electrical motor is that it has high torque as soon as it is
switched on and therefore allows for fast acceleration. On the other hand, an engine will have
higher torque at higher rpms and will therefore be better for acceleration at higher speeds. The
range of a parallel hybrid will not be too different from a fossil fuelled vehicle. Depending on
how much electrical energy one wants to use, a smaller or larger battery bank will need to be
installed.
When comparing the consumption of normal vehicles with that of hybrid vehicles, then
the latter are said to run with higher fuel efficiency, generally about 20 % [38, 39]. This has
multiple reasons: firstly, they recover some of the brake energy and secondly, the combustion
engine in e.g. the Toyota Prius makes use of the Atkinson cycle, which is known to have an
improved thermal efficiency. Another motor that has an improved thermal efficiency is the
Reinhardt Turbine since it applies the principle of the combined cycle applied in some gas
powered stations to the combustion engine [40]. This implies that the heat produced during
the combustion cycle is used to drive another piston, which consequently increases the overall
efficiency of the engine to over 50%, which in the future can possibly be improved up to
75%. The construction of the engine itself is easier than the current engines since there is no
requirement for a crankshaft, while the principle can be applied to any type of combustion
engine and therefore increases the efficiency of a combustion engine running on any type of
fuel.
An engine used in a hybrid vehicle generally has slightly different requirements, as for
example it may be running only to drive a generator. Hence, the requirements for the engine
are slightly different, and the engine no longer needs to provide high output values in a range
of rpms. Therefore, it is likely that new engine designs, such as the shock wave engine [41,
42] which is about 3.5 times more efficient than the normal combustion engine, produces
95% less carbon dioxide and can run on any type of fuel: gas, petrol, hydrogen will find its
way into hybrid vehicles. On top of that, the shock wave engine has a smaller weight, is less
expensive to produce and has lower maintenance requirements.
4.2. Bio-Fuels
Together with the drive towards greener energy production and transport, there has been
a drive towards using more bio-fuels. The definition of bio-fuels has however also been
diversified. In the beginning, the term bio-fuel would be used if waste was turned into a fuel.
For example, in the early days when wood-chip waste or saw dust was used for running a
boiler system, it would be considered a bio-fuel since it made use of a waste product. Now the
demand for wood-chips and sawdust has overtaken the available waste and therefore there are
specialised factories producing sawdust and wood chips to fulfill the demand. This raises
questions as to whether fuel produced in this way can still be considered to be a bio-fuel,
since the fuel is no longer a by-product of another process that is being recycled, but a
product that is produced especially and actually requires quite some energy to be produced.
More generally, the current definition of bio-fuel would be based on the principle that a fuel
would be considered to be a bio-fuel if it compensates the CO2 produced when it is burned
during the actual growing and production of the fuel itself. This definition implies a zero
524 Wim J. C. Melis
carbon impact when performing a full life-cycle analysis, or in other words introduces the
natural recycling principle to fuels and energy.
Bio-fuels have caused another controversy, which lies in the fact that their production can
take up quite a number of fields normally used for food crops. Considering the rise in
population on one hand and therefore food demands, it seems rather contradictory that
legislation requires for an increasing percentage of bio-fuel to be produced. This brings us to
a slightly different and probably less harmful definition for a bio-fuel, namely: a bio-fuel is a
fuel that is either a by-product of, or a product that can easily be produced in nature around us
without affecting current crop production and does not provide any harmful output when
being burned.
Bio-fuels would currently be classified into first and second-generation fuels. The first
generation fuels would be based on sugar, starch or vegetable oil, and hence comprise fuels
like: bio-diesel, bio-alcohol, bio-gas among others. The second-generation bio-fuels, on the
other hand, would be made from sustainable feedstock, and comprise fuels like: wood diesel,
bio-methanol, bio-hydrogen, algae-fuel and the like.
Bio-diesel can be produced from a variety of different sources and is the lowest emission
diesel available. Depending on the source it is created from, the viscosity can however be
slightly different and can therefore only be used in the older type diesel engines without being
diluted. The latest generation diesel engines with direct injection require a fuel with a certain
viscosity and therefore bio-diesel has to be mixed with fossil fuel based diesel to ensure a
consistent viscosity so that the engine‟s injection system does not get damaged. One benefit
of bio-diesel is that it cleans the engine from any deposits that settled there from burning
normal diesel. Hence, especially when an engine is changed from normal diesel to bio-diesel
the filters will need to be changed more regularly since they need to deal with all this deposit.
Bio-ethanol/propanol/buthanol would generally be produced from the fermentation of
sugars or starches. The resulting fuel can be used in any normal gas-oil combustion engine
generally mixed with normal gas-oil except for buthanol, which can be burned on its own.
When compared to normal petrol, this fuel holds a lower energetic value and therefore more
fuel will be required for the same energy output.
Wood gas was particularly popular as a fuel during the Second World War [43]. When
wood is brought to a high temperature, it releases a gas that once purified from water vapour
and other impurities is highly flammable. This gas can be burned in a normal gas-oil type
engine with slight adjustments to the air-fuel ratio. The by-product of extracting the gas from
the wood is that the wood turns into charcoal, which can then again be used to heat other
wood. Considering that wood contains a burnable gas, it will be no surprise that one can also
extract oil from it to burn in a petrol engine.
Bio methanol is a rather controversial fuel. Although it can be produced cheaper than
ethanol, it has a much lower energy density. On top of that, it is toxic and therefore needs to
be dealt with, with care, not to harm the user. Algae fuels are generally considered to
overcome the food production problems associated with other bio-fuels. The added advantage
of algae is that they produce fuels very quickly, and can be grown in an industrial
environment to fulfill the current fuel requirements. A few fuels that can be produced by
algae are ethanol/methanol and glycol. The latter of which is a bio-fuel with extremely low
emissions and therefore a perfect solution for many situations other from road transport [44].
Although a variety of bio-fuels can be produced, quite a few of them have lower
energetic values and therefore more fuel will be required for the same output. The main
benefit however lies in the fact that most will have a reduced CO2 output in comparison to
fossil fuels.
4.3. Hydrogen
Hydrogen can be used as a fuel in two different ways: one can either combust it, which
results in a slightly higher efficiency than burning gas-oil [45], or it can be used to produce
electricity by providing it to a fuel cell, which acts as a catalyst. In this case, the catalyst
converts hydrogen and air into water and electricity. The latter method is currently more
commonly used in transportation. In essence this also means that hydrogen is actually only
used as an energy carrier within the vehicle and therefore replaces the batteries. Generally,
hydrogen will be stored under high pressure to ensure a proper amount of energy being
carried on board the vehicle, but this requires strong storage tanks which obviously also have
a certain weight associated with them. This weight will depend on the size of the tank,
materials being used for its construction and so on, but overall, the energy density of
hydrogen is smaller than that of fossil fuels, therefore to reach a similar range more storage
will be required. With hydrogen being used as a store for electrical energy, one also needs to
compare the storage efficiency in comparison to batteries. This efficiency would be
somewhere between 60 and 85 % depending on the fuel cell type used (Table 2). Considering
that batteries are at the latter level already, there seems to be little advantage to using
hydrogen, especially considering the fact that, because it is highly flammable, there are extra
problems for use in a transport context.
4.4. Compressed Air
Compressed air vehicles are not a new invention. They were used in the early 20th
century in different locations around the planet. One of the main problems with compressed
air vehicles is the actual compression/decompression cycles that are required. During
compression there is generally heat produced and during decompression cold is produced,
both reduce the efficiency of the overall process.
Table 2. Fuel cell technologies [46]
Operating Typical Stack

Fuel Cell Type Electrolyte Efficiency
Temperature Size
Alkaline Fuel Cell (AFC) Potassium hydroxide solution 90-100°C 60% 10-100 kW
Polymer Electrolyte
50-100°C,
Membrane Fuel Cell Perfluoro Sulfonic Acid 40-60% <1 kW-100 kW
typically 80°C
(PEMFC)
Phosphoric Acid Fuel
Phosphoric Acid 150-200°C 40% 400 kW
Cell (PAFC)
Molten Carbonate Fuel Molten mixture of alkali
600-700°C 45-50% 300 kW-3 MW
Cell (MCFC) metal carbonates
Solid Oxide Fuel Cell Oxide ion conducting
700-1000°C 60% 1 kW-2 MW
(SOFC) ceramic
526 Wim J. C. Melis
In order to increase this efficiency, several techniques are available. Firstly, there is the
option of regenerating the heat produced during compression for decompression [47],
alternatively one can compress air stored in a tank by adding/removing a fluid to the tank and
hence use a hydraulic system for compression and decompression.
Before moving to purely hydraulic systems, there are still a number of advantages about
compressed air storage that need to be mentioned. When compared to an electrical vehicle
with batteries, the life expectancy of a compressed air vehicle is longer, there are no nasty
chemicals being used and the storage capacity does not degrade over time. The actual
construction is also easier, and although most tank systems would be based on metal, these
come with a certain weight and therefore reduce the fuel efficiency in relation to weight.
However, new materials are likely to enable other tank constructions, and such lighter
materials could then improve on the fuel efficiency.
4.5. Hydraulic Vehicles
As mentioned previously, one of the problems with current fossil fuelled vehicles is that
their engines have to be designed to deal with a large range of requirements and therefore
need to have high torque across a large rpm-range. Having to over-design the system leads to
inefficiencies when it is used for the majority of the time. Therefore, it would be better when
the vehicle has an energy source that can deal with the average demand and delivers the peak
demands in a different way. The hydrid [48] is a system build around this model. It makes use
of a hydraulic transmission system, and even though this is less efficient than a mechanical
transmission, overall, the system still has a higher efficiency. The hydrid also overcomes the
complexity and extra requirements of hybrid vehicles, since these generally require a double
transmission system to enable the electrical and combustion engine to power the vehicle. On
the other hand, the hydrid uses the engine as an energy source to fill up the hydraulic
accumulator, similar to the serial type hybrid. However, once the storage system is full, the
engine would switch off and the car would drive completely based on the energy from the
accumulator. This accumulator is in essence a set of tanks where liquid is used to compress a
gas and therefore indirectly uses compressed air as storage. An added advantage of this
hydraulic system is that this system can also be used in reverse mode to perform regenerative
braking, which to some extend compensates the higher losses of the not mechanical
transmission system. Overall, the hydrid system results in a 50% fuel efficiency when
compared to a fossil fuelled vehicle, which consequently also halves the grammes of
CO2/100km. The hydrid system can also be combined with an electric motor to gain the
advantages of both systems, which are: the hydraulic system, which has a good power
density, but low energy density, whereas the electrical system has a poor power density, but
good energy density [49].
4.6. Public Transport
Considering that one of the influencing factors on fuel consumption for a vehicle is the
size of the front of the vehicle and therefore the required air displacement, it will be no
surprise that public transport can reach much higher efficiencies per person since there are
more people in the same vehicle. From this aspect it will also be no surprise that trains are
one of the most efficient ways of transportation. However public transport can only be
efficient by having a limited number of stops, which also means it cannot provide the same
comforts of a personal vehicle.
CONCLUSION
The current trend seems to propose that electrical vehicles form the solution towards the
pollution problem of current fossil fuelled vehicles. However, the environmental friendliness
of electrical vehicles largely depends on how the electricity itself is generated. As long as
electricity is mainly generated from fossil fuels, like in most Western countries, electrical
vehicles do not provide a benefit in this aspect. However, they do provide benefits when it
comes to the integration of ICT and automation of the vehicle, as well as allowing for an
increase in transport efficiency due to energy recovery during deceleration. It was shown that
fully regenerative breaking is essentially not hard to achieve and even brings benefits for the
actual acceleration of the vehicle.
Current hybrid vehicles already benefit from combining the two technologies. However
there are a number of improvements still possible to this technology. The first one has already
been made available, by creating the plug-in Hybrid, which allows for, preferably green
electricity to be used to charge the vehicle. A second improvement lies in the fuel being
burned in the engine in relation to extending the vehicle‟s range. The required improvement
here is related to the CO2 production of the fuel being burned, which can be solved by using a
different fuel. Replacing the engine with a Fuel cell would provide a better link to the third
problem, which is that of electrical energy storage. Currently, batteries are used for this
storage, whereas other principles, like compressed air, hydraulics and hydrogen would
provide for higher efficiencies and storage densities. In any case the energy storage provided
within EVs would come in handy when such vehicles are linked to the grid as a buffer for
storing locally generated renewable energy.
REFERENCES
[1] Binesti, D.; Core Losses and Efficiency of Electrical Motors Using New Magnetic
Materials. IEEE Transactions on Magnetics, vol. 32, no. 5, pp. 4887-4889, 1996.
[2] Mackay, D.J.C.; Sustainable Energy - Without the Hot Air. UIT Cambridge, 2009,
www.withouthotair.com/download.html.
[3] Bandivadekar, A.; Bodek, K.; Cheah, L.; Evans, C.; Groode, T.; Heywood, J.; Kasseris,
E.; Kromer, M.; Weiss, M.; On the Road in 2035: Reducing Transportation‟s Petroleum
Consumption and GHG Emissions. MIT Report, 2008. http://mitei.mit.edu/system/files/
On%20the%20Road%20in%202035_MIT_July%202008.pdf.
[4] Melis, W.J.C.; Chishty, O.; Fully Regenerative Braking and Improved Acceleration for
Electrical Vehicles. International Journal of Sustainable Energy Development (IJSED),
vol. 2, no. 1 and 2, pp. 75-80, 2013.
528 Wim J. C. Melis
[5] Situ, L.; Electric Vehicle Development: The Past, Present & Future. 3rd International
Conference on Power Electronics Systems and Applications, pp. 1-3, 2009.
[6] Chan, C.C.; The Past, Present and Future of Electric Vehicle Development. IEEE
International Conference on Power Electronics and Drive Systems (PEDS ‟99), Hong
Kong, pp. 11-13, 1999.
[7] Toyota; Green Innovation. 2012. www.toyota.com/about/environmentreport2012/
green_innovation.html.
[8] www.hybridcenter.org/hybrid-timeline.html.
[9] Shen, Z.J.; Zhang, Y.B.; Yao, Y.Q.; Liu, B.W.; Electric Vehicle: The Hope of the
Green Future. International Conference on Management and Service Science (MASS
2011), pp. 1-4, 2011.
[10] Zahedi, A.; Electric Vehicle as Distributed Energy Storage Resource for Future Smart
Grid. 22nd Australasian Universities Power Engineering Conference, pp. 1-4, 2012.
[11] Pillai, J.R.; Bak-Jensen, B.; Thogersen, P.; Electric Vehicles to Support Large Wind
Power Penetration in Future Danish Power Systems. IEEE Vehicle Power and
Propulsion Conference, Seoul, Korea, pp. 1475-1479, 2012.
[12] Mierau, M.;Kohrs R.;Wittwer, C.; A Distributed Approach to the Integration of Electric
Vehicles into Future Smart Grids. 3rd IEEE PES International Conference and
Exhibition on Innovative Smart Grid Technologies, pp. 1-7, 2012.
[13] Shafiei, A.; Williamson, S.S.; Plug-in Hybrid Electric Vehicle Charging: Current Issues
and Future Challenges. IEEE Vehicle Power and Propulsion Conference (VPPC),pp. 1-
8, 2010.
[14] Buckl, C.; Camek, A.; Kainz, G.; Simon, C.; Mercep, L.; Stahle, H.; Knoll, A.; The
Software Car: Building ICT Architectures for Future Electric Vehicles. IEEE
International Electrical Vehicle Conference (IEVC), pp. 1-8, 2012.
[15] Silberg, G.; Wallace, R.; Self-Driving Cars: The Next Revolution. KPMG LLP
&Center for Automotive. Research Report, 2012. www.kpmg.com/US/en/IssuesAnd
Insights/ArticlesPublications/Documents/self-driving-cars-next-revolution.pdf.
[16] Coelingh, E.; Solyom, S.; All Aboard the Robotic Road Train. IEEE Spectrum, 2012,
http://spectrum.ieee.org/green-tech/advanced-cars/all-aboard-the-robotic-road-train.
[17] [17] Wolfram, S., A New Kind of Science. Wolfram Media Inc., 2002.
[18] Melis, W.J.C.; Kameyama, M.; A Study of the Different Uses of Colour Channels for
Traffic Sign Recognition on Hierarchical Temporal Memory. Fourth International
Conference on Innovative Computing, Information and Control, pp. 111-114, 2009.
[19] Chua, L.O.; The Fourth Element. Proceedings of the IEEE, vol. 100, no. 6, pp. 1920-
1927, 2012.
[20] Sanpasertparnich, T.; Improved Thermal Efficiency of Coal-Fired Power Station:
Monte Carlo Simulation. IEEE EIC Climate Change Technology, Ottawa, USA, pp. 1-
9, 2006.
[21] http://en.wikipedia.org/wiki/Fossil-fuel_power_station.
[22] Ferreira S.; Eco-sheet: 1 kWh Generated Using Different Energy Sources. Leonardo-
Energy, 2008, www.leonardo-energy.org/eco-sheet-1-kwh-generated-using-different-
energy-sources.
[23] http://en.wikipedia.org/wiki/Combined_cycle.
[24] Siddique, A.H.; Lamont, L.A.; Al-Bloushi, H.K.; Nuclear Power for the Vision 2030 of
Abu Dhabi and Improvement of Air Quality. World Congress on Sustainable
Technologies, London, UK, pp. 14-18, 2012.
[25] Yamaguchi, M.; Nuclear Power to be Phased out Following Fukushima Disaster. The
Independent, September 15, 2012, www.independent.co.uk/news/world/asia/nuclear-
power-to-be-phased-out-following-fukushima-disaster-8139966.html.
[26] Evans, S.; Germany: Nuclear Power Plants to Close by 2022. BBC News, May 30,
2011, www.bbc.co.uk/news/world-europe-13592208.
[27] De Vos, A.; Detailed Balance Limit of the Efficiency of Tandem Solar Cells. Journal of
Physics D: Applied Physics, vol. 13, no. 5, pp. 839-846, 1980.
[28] Shepherd, W.; Zhang, L.; Electricity Generation using Wind Power. World Scientific
Publishing, 2011.
[29] Aquirre, K.; Eisenhardt, L.; Lim, C.; Nelson, B.; Norring, A.; Slowik, P.; Tu, N.; Life
Cycle Analysis Comparison of a Battery Electric Vehicle and a Conventional Gasoline
Vehicle. 2012, www.environment.ucla.edu/media_IOE/files/BatteryElectricVehicle
LCA2012-rh-ptd.pdf.
[30] Hawkins, T.R.; Singh, B.; Majeau Bettez, G.; Stromman, A.H.; Comparative
Environmental Life Cycle Assessment of Conventional and Electric Vehicles. Journal
of Industrial Ecology, vol. 17, no. 1, pp. 53-64, 2012.
[31] U.S. Energy Information Administration, Few Transportation Fuels surpass the Energy
Densities of Gasoline and Diesel.2013, www.eia.gov/todayinenergy/detail.cfm
?id=9991.
[32] www.eea.europa.eu/data-and-maps/indicators/ener002-co2-intensity-of-heat
/assessment.
[33] www.carpages.co.uk/co2/.
[34] Deeba, M.; Dettling, J.C.;Lui, Y.K.; Four-Way Diesel Exhaust Catalyst and Method of
Use. US6093378, 2000, http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=
HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=5
0&s1=6,093,378.PN.&OS=PN/6,093,378&RS=PN/6,093,378.
[35] Ji, S.; Cherry, C.R.; Bechle, M.J.; Wu, Y.; Marshall, J.D.; Electric Vehicles in China:
Emissions and Health Impacts. Environmental Science and Technology, vol. 46, no. 2,
pp. 2018-2024, 2012.
[36] FIA Foundation for the Automobile and Society; The Automobile and Society,
Executive Summary. 2003, www.fiafoundation.org/publications/Documents/ auto_
society_exec_summary.pdf.
[37] www.solarjourneyusa.com/EVdistanceAnalysis7.php.
[38] Imai, S.; Takeda, N.; Horii, Y.; Total Efficiency of a Hybrid Electric Vehicle. Power
Conversion Conference, Nagaoka, Japan, vol. 2, pp. 947-950, 1997.
[39] Kaneko, T.; Shimada, M.; Kujiraoka, S.; Kojima, T.; Easy Maintenance and
Environmentally-Friendly Train Traction System. Hitachi Hyoron, vol. 85, no. 8, pp.
549-552, 2003.
[40] Reinhardt, G.T.; Thermal Engine such as e.g. an Internal Combustion Engine and/or
Steam Engine. WO2011/042301, 2011, http://patentscope.wipo.int/search/en/
WO2011042301.
[41] Knight, H.; Shock wave puts hybrid engines in a spin. New Scientist, vol. 209, no. 2803,
p. 25, 2011.
530 Wim J. C. Melis
[42] Revkin, A.C.; The „Wave‟ of the (Car Engine) Future?. The New York Times, 2011,
http://dotearth.blogs.nytimes.com/2011/04/11/the-wave-of-the-car-engine-future/?
partner=rss&emc=rss.
[43] Food and Agriculture Organization of the United Nations; Wood Gas as Engine Fuel.
Food and Agriculture Organization of the United Nations, 1987, www.fao.org/docrep/
T0512E/T0512E00.htm.
[44] Day, P.; McNeil, J.; Sirovski, F.; (E)Mission Impossible ?.The Chemical Engineer, vol.
839, pp. 33-35, 2011.
[45] Berger, E.; BMW Hydrofen Near Zero Emission Vehicle Development. BMW Clean
Energy CARB ZEV Technology Symposium, pp. 1-28, 2006.
[46] US Department of Energy; Energy Efficiency & Renewable Energy: Fuel Cell
Technologies Program. 2011, www1.eere.energy.gov/hydrogenandfuelcells/fuelcells/
pdfs/fc_comparison_chart.pdf.
[47] http://lightsailenergy.com/tech.html.
[48] Achten, P.; Vael, G.; Murrenhoff, H.; Low-Emission Hydraulic Hybrid for Passenger
Cars.ATZ, 111, pp. 378-384, 2009.
[49] Vael, G.; Eggenkamp, S.; Achten, P.; The E-Hydrid. The Twelfth Scandinavian
International Conference on Fluid Power, Tampere, Finland, pp. 1-15, 2011.
INDEX
airports, 188
# algae, 524
Algeria, 483
20th century, 208, 525
algorithm, xiii, 15, 16, 29, 35, 81, 83, 84, 99, 100,
21st century, 411
102, 103, 110, 113, 121, 122, 124, 126, 127, 128,
3D images, 58
129, 130, 131, 132, 134, 135, 138, 157, 161, 163,
202, 204, 215, 270, 279, 280, 281, 282, 283, 284,
A 287, 300
alternative hypothesis, 35
abuse, 455 aluminium, 433, 443, 446, 450
AC motors, 260, 268, 380, 385 amplitude, 10, 11, 13, 14, 15, 32, 36, 38, 57, 58, 63,
accelerator, 326 70, 76, 239, 251, 279, 384, 385, 389, 390, 391,
access, xiv, 149, 151, 152, 161, 168, 170, 171, 172, 393, 394, 395, 403, 406
173, 183, 188, 192, 194, 195, 196, 197, 199, 202, anatomy, 60
203, 208, 210, 214, 217, 219, 220, 237, 238, 240, applied mathematics, 204
244, 504 aqueous alkali solution, 454
accessibility, 216, 219 arithmetic, 12, 22
acid, 322, 409, 411, 412, 414, 415, 426, 434, 454, artificial intelligence, 267, 380
521 Asia, 188, 229
actuation, 410 assessment, 82, 149, 176, 199, 529
actuators, 115, 215 assets, 202
ad hoc network, 215 asthma, 410, 521
adaptation, xiii, 67, 68, 148, 150, 157, 213, 225, 317 asymmetry, 68, 395
ADC, 61, 62 atmosphere, xii, 42, 49, 212, 422, 454
adjustment, 155, 283, 370, 371 attachment, 192, 496, 501, 502, 504
administrators, 188 attacker, 190, 192, 199
Admittance, 483 Attenuation Factors, 3
adverse weather, 118 audit, 198
adware, 192 authentication, 242
aerosols, 49 authenticity, 195
Africa, 227 authority(s), 175, 189, 221
age, 241 automate, 190, 193
agencies, 189, 203, 209, 243, 410 Automatic incident detection, 148, 150
aggregation, 294 automation, 59, 437, 527
Aggregator, 291 automobile(s), xi, xv, 143, 197, 237, 410, 420, 506,
agility, 491, 498 529
agriculture, 188 automotive application(s), ix, xiii, 117, 252, 316,
air quality, 210, 521 317, 320, 321, 322, 334, 435
air temperature, 44
532 Index
autonomy, xii, 82, 83, 325, 380, 409, 410, 411, 413, carbon, 210, 212, 322, 420, 423, 425, 426, 519, 523,
417, 420, 436 524
avoidance, 4, 5, 6, 51, 59, 74, 190, 242 carbon dioxide, 523
awareness, 191, 223, 225 carbon emissions, 210, 212
Azerbaijan, x, 111 carbon nanotubes, 426
case study(s), xvi, 84, 186, 187, 287, 297, 305
cash, 326
B catalyst, 525
cathode materials, 426, 433
background noise, 35, 36, 38, 39
cation, 103, 110, 413
bandwidth, 10, 19, 32, 34, 37, 40, 41, 52, 65, 66, 76,
C-C, xv, 457, 459, 469, 477, 478, 479, 480
156, 169, 171, 183, 237, 239, 240
CEC, 231
banking, 186, 188, 202
cell phones, 454
barriers, 167
ceramic(s), 418, 428, 483, 525
base, xii, 107, 148, 150, 167, 168, 170, 211, 217,
certificate, 199
218, 241, 262, 440, 444
certification, 176, 202
battery, viii, 275, 276, 289, 292, 314, 322, 323, 324,
challenges, xii, 83, 114, 115, 117, 196, 217, 219,
345, 346, 347, 348, 351, 411, 413, 427, 428, 437,
221, 222, 223, 294, 503, 511
438, 439, 450, 451, 453, 454, 455, 456, 482, 516,
chemical(s), xii, xv, 188, 194, 413, 422, 433, 434,
529
454, 488, 526
beams, 116, 155
chemical industry, 188
behaviors, 215
chemical properties, 413
Belgium, 233
chemical reactions, xii, xv
benefits, 129, 172, 182, 189, 209, 210, 212, 265,
chemical stability, 434
294, 486, 487, 511, 512, 518, 522, 523, 527
Chicago, 505
bias, 14, 65
China, 78, 506, 520, 529
Bible, 228
circulation, 157, 328, 392, 504
bio-fuel, 514, 524
cities, 120, 172, 176, 183, 209, 213, 216, 244
blind spot, 5, 59, 213
citizens, 220, 224
blogs, 530
civilization, 208
Bluetooth, 169, 171, 192
clarity, 338, 481
BMA, 163
classes, 113, 122, 125, 127, 132, 134, 135, 136, 138,
bounds, 304
139, 142, 168, 171, 196, 197, 212, 327
brain, 178
classification, 58, 82, 113, 119, 120, 121, 122, 123,
Brazil, 145, 350
124, 125, 126, 127, 129, 130, 131, 132, 134, 135,
breakdown, 242, 381, 385
136, 137, 138, 141, 142, 155, 157, 211, 212, 214,
burn, 524
318, 494
business management, 190
clients, 190
businesses, 185, 187, 190, 217
climate, 150, 208
buyer, 431
climate change, 208
by-products, 519, 521
clustering, 121, 130, 132, 134
clusters, 133, 245
C CO2, 176, 177, 317, 347, 489, 519, 520, 521, 522,
523, 525, 526, 527
cables, 171, 182 coal, 49, 50, 513, 521
CAD, 444 coal dust, 49, 50
cadmium, 415, 416, 417, 434, 454, 455 cobalt, 419, 432, 433, 434
calibration, 159 coding, 166, 170
CAM, 241 cognitive process, 218
cancer, 410 collaboration, 219
candidates, 132, 133, 437 collisions, 170, 215, 238, 244
car accidents, ix color, 117, 133, 492
Index 533
combustion, xi, xv, 250, 276, 295, 318, 410, 411, construction, xiv, 53, 74, 95, 215, 223, 249, 261,
412, 414, 434, 485, 486, 505, 509, 510, 511, 512, 265, 311, 321, 360, 364, 375, 416, 417, 481, 487,
514, 521, 522, 523, 524, 526 497, 498, 502, 518, 523, 525, 526
commerce, 186 consulting, 212
commercial, 73, 166, 169, 172, 208, 317, 353, 354, consumers, xi, 298
376, 417, 433, 454 consumption, ix, 177, 208, 213, 293, 303, 317, 345,
communication(s), xi, xiv, 6, 8, 11, 82, 149, 167, 410, 492, 510, 515, 516, 520, 523
168, 169, 170, 178, 179, 182, 183, 187, 188, 189, containers, 169
193, 195, 199, 208, 212, 215, 219, 222, 235, 236, Continental, 143, 354
237, 238, 239, 240, 241, 243, 244, 245, 276, 324, Continuous Air interface Long and Medium range
326 (CALM), 237
communication systems, 6, 82, 183, 212, 236 contour, 141, 162
communication technologies, 236 control measures, 176
communications channel, 182 controversial, 515, 524
community, xv, 172, 209, 215, 220, 222 convergence, xv, 31, 217, 224, 284, 432
compatibility, 103, 168, 183, 503 conversations, 169, 192
compensation, 5, 66, 74, 76 conversion rate, 416
competition, 168, 225, 456 cooling, xv, 47, 191, 265, 357, 362, 365, 416, 417,
competitiveness, 208 431, 437, 438, 439, 440, 442, 443, 445, 447, 450,
compilation, 193, 202, 371, 390 500, 501, 502, 503, 504
complement, 118, 119, 279 cooperation, 198, 222
complexity, 8, 38, 52, 113, 116, 119, 135, 161, 195, coordination, 60, 178, 182, 215
196, 198, 207, 209, 219, 220, 224, 225, 325, 481, copper, 265, 319, 519
486, 526 correlation(s), 5, 32, 35, 38, 39, 52, 74, 115, 148, 198
compliance, 191 covering, 117, 171, 173, 189, 235
composition, 420, 422 creativity, 491
compounds, 426 credentials, 193
compression, 525, 526 criminal activity, 188
computation, 8, 52, 74, 124, 129, 134, 135, 161, 188, criminals, xiii, 185
190, 222, 283 crises, 410
computer, 6, 58, 60, 62, 71, 74, 98, 132, 185, 186, critical infrastructure, ix, xiii, 185, 186, 187, 188,
187, 188, 190, 191, 192, 193, 194, 195, 196, 197, 189, 190, 191, 192, 193, 194, 202, 203, 204
199, 204, 205, 208, 237, 241, 436 crop(s), 521, 524
computer systems, 186, 192 crop production, 524
computing, 124, 128, 129, 130, 147, 150, 180, 188, cryptography, 198, 200, 204
189, 190, 197, 341, 360 Cryptography, 186
condensation, 47 crystalline, 419, 420
conditioning, xii, 183, 251, 279 crystals, 455
conductance, 458 cultural influence, 225
conduction, 255, 257, 258, 328, 480, 481 culture, 222, 225
conductivity, 425, 432, 433, 447 cure, 242
conductor, 157, 262 customers, 190, 494
confidentiality, 195 Cyber security, 185
configuration, xi, 65, 66, 94, 102, 154, 160, 180, cyberspace, 204
194, 250, 251, 269, 280, 295, 296, 364, 369, 370, cycles, 255, 266, 283, 294, 323, 324, 415, 416, 417,
371, 372, 375, 376, 424, 443, 444, 445, 446, 480, 418, 419, 422, 423, 426, 434, 435, 437, 513, 525
481, 505 cycling, 323, 455
conformity, 165
congress, 79, 184, 188, 204, 226, 311, 529
connectivity, 179, 190, 222, 238 D
consensus, 132, 211
damages, 497
conservation, 89
damping, 88, 499
danger, 219, 238
534 Index
data analysis, 456 distribution, 45, 122, 130, 173, 188, 190, 198, 203,
data availability, 209 265, 276, 277, 278, 280, 288, 294, 299, 305, 310,
data collection, 165 318, 326, 381, 383, 393, 395, 438, 448, 453, 477,
data communication, 165, 168 478, 479, 480, 481, 493, 498, 515, 516
data gathering, 193 distribution function, 305, 453, 477, 478, 479, 481
data generation, 161 diversity, 207, 220, 223, 237, 454
data processing, 62, 161, 183, 241 DME, 435
data rates, 237 doping, 433
data transfer, 155, 170 Doppler shift, 10, 11, 239
database, 132, 136, 138, 139, 142 DOT, 216
DDoS, 192 downlink, 171
decision makers, 217, 297 drawing, 422, 454, 510
decomposition, 381 durability, 317, 323, 432, 435, 454, 455, 486, 498
decoupling, 263, 316 Dynamic Modeling, 81
defects, 392 dynamic systems, 103
degradation, 28, 40, 41, 47, 48, 51, 52, 53, 294, 310,
385, 433
democratization, 381 E
dendrites, 455
economic development, 211, 221
denial, xiii, 185
economic policy, 208
denial of service attack, 185
economic progress, 208
Department of Energy, 410, 530
economics, 310
Department of Homeland Security, 191, 203
ecosystem, 210
Department of Transportation, 184, 216
editors, ix, xv, xvi
dependent variable, 474
education, 204, 224, 511
deployments, 199
educational software, 198
deposits, 524
elaboration, 410
depression, xv, 453, 457, 458, 459, 460, 480, 481
electric current, 358, 422, 435
depth, xiii, xv, 82, 83, 101, 105, 135, 186
electrical conductivity, 425
derivatives, 85, 92, 94, 109
electricity, xi, xiii, xiv, 185, 186, 188, 190, 191, 202,
designers, 320, 491
215, 250, 251, 276, 293, 294, 295, 302, 303, 304,
destruction, 393
305, 306, 309, 310, 311, 326, 345, 381, 488, 509,
detection system, 155, 156, 161, 165
510, 511, 513, 514, 515, 516, 519, 520, 521, 522,
detection techniques, 12, 51
525, 527
Detection threshold, 5
electrodes, 413, 414, 415, 422, 425, 433, 435, 455
deviation, xv, 151, 281, 297, 453
electrolysis, 414
DHS, 203
electrolyte, 322, 413, 414, 415, 416, 418, 419, 420,
diesel engines, 410, 514, 521, 524
422, 424, 425, 433, 435, 454, 455
differential equations, 98, 99
electromagnetic, xiv, 42, 63, 117, 157, 170, 242,
diffusion, 414, 423, 481
249, 261, 264, 265, 270, 381, 393, 394, 490
diodes, 253, 254, 318
electronic communications, 186
Direct drive, 497
electronic systems, 147, 454
direct measure, 176, 177, 264
electron(s), 413, 414, 420, 423, 435
disaster, 529
e-mail, 192
discharges, 412
emergency, xiii, 144, 147, 149, 150, 151, 152, 167,
discretization, 305
168, 170, 178, 238, 243, 244, 319, 517
discrimination, 74, 382
emission, 66, 74, 411, 490, 521, 524
discs, 501
employment, 11, 25, 38, 53, 212
diseases, 410
employment growth, 212
dispersion, 167
empowerment, 225
displacement, 526
EMS, 211
distortions, 68, 269
EMU, 314, 315, 316, 317, 321, 324, 325, 326, 328,
distributed computing, 222
333, 334, 345
Index 535
encryption, 171, 199, 202, 242 expertise, 209, 217, 381

endurance, 82, 499 exploitation, 199, 380
end-users, 172, 173 exporters, 208
enemies, 191 external relations, 218
energy consumption, 172, 176, 177 extraction, 69, 161, 165, 435, 458, 513
energy density, xii, 275, 276, 289, 318, 321, 322, extracts, 115
323, 325, 411, 412, 413, 415, 417, 420, 423, 425, extrusion, 432, 448, 450
434, 435, 436, 437, 454, 455, 509, 513, 519, 524,
525, 526
energy efficiency, ix, xiv, 176, 177, 215, 313, 315, F
316, 317, 318, 320, 321, 326, 328, 333, 334, 336,
fabrication, 69, 360, 375
337, 338, 340, 341, 342, 343, 344, 494, 497, 514
factories, xiii, 185, 523
energy prices, 303, 305, 308, 309, 310
false alarms, 26
energy recovery, 413, 527
false negative, 131
enforcement, 148
false positive, 131, 512
engineering, xii, xiii, xv, xvi, 6, 52, 95, 185, 191,
fault detection, xiii, 57, 58, 381, 397, 404
195, 196, 198, 457, 488
fault diagnosis, 168
England, 53, 231
feature detectors, 142
Environmental, 84, 226, 345, 347, 519, 529
Federal Bureau of Investigation, 187
environmental aspects, ix, 172, 208
Federal Highway Administration, 184
environmental care, xv, 410
feedstock, 524
environmental conditions, 59, 83
fermentation, 524
environmental effects, 454
FFT, 4, 8, 20, 21, 396
environmental impact, xv, 176, 317, 509, 510, 519,
FGDC, 223
522
fiber(s), 82, 426, 456
environmental issues, 486
fidelity, 81
environmental protection, 504
field tests, 95
environments, xiv, 4, 58, 59, 82, 83, 109, 115, 119,
filters, 32, 122, 134, 255, 259, 490, 521, 524
120, 185, 189
financial, 188, 219, 298, 511
EPS, 211
financial support, 511
equality, 34
Finland, 407, 530
equilibrium, 83, 84, 89
firewalls, 192
equipment, ix, xi, 61, 148, 149, 150, 157, 161, 169,
first generation, 417, 486, 524
170, 171, 176, 178, 179, 180, 183, 185, 190, 193,
fitness, 223
212, 217, 276, 326, 434, 512
flexibility, xii, 65, 160, 170, 223, 316, 323, 454
equity, 210, 211
flight, 83, 116
ESD, 314, 315
flow field, 95
ethanol, 524
fluctuations, 18, 19, 298
ethylene, 435
fluid, 63, 84, 86, 88, 93, 94, 95, 96, 431, 436, 526
EU, 173, 494, 520, 521
fog attenuation, 48, 49
Europe, 110, 147, 169, 176, 184, 211, 212, 221, 228,
food production, 188, 524
231, 240, 273, 347, 377, 416, 417, 520
force, 89, 90, 91, 92, 93, 94, 95, 96, 102, 105, 108,
European Commission, 227, 237
147, 148, 150, 185, 202, 220, 238, 243, 265, 295,
European Community, 227
321, 324, 379, 382, 393, 490, 493, 501, 502, 517,
European Parliament, 227
518
European Union, 221, 231
Ford, 318, 354, 410, 416, 417, 419, 494
everyday life, 210
forecasting, 303, 309, 311
evidence, 148, 150, 178, 218
formula, 89, 98, 99, 159, 201
evolution, 114, 115, 119, 190, 192, 208, 210, 220,
FOV, 114, 117
360, 380, 392, 426, 455, 481
framing, 162, 163
excitation, 265, 355, 356, 370, 374
France, xvi, 379, 406, 407, 408, 505
execution, 195
fraud, 186, 202
exercise, 519
freedom, 57, 60, 76, 85
536 Index
freezing, 322 hacking, 198

friction, 230, 393, 490, 501, 502, 504 halogen, 182
fuel cell, ix, xii, xiii, xiv, xv, 249, 250, 251, 255, handwriting, 124
318, 409, 417, 421, 488, 489, 525 harbors, 82
fuel consumption, 149, 176, 177, 189, 212, 215, 412, harvesting, 315, 327
511, 517, 519, 522, 526 hazards, 216, 435
fuel efficiency, 513, 514, 523, 526 headache, 437
fusion, 8, 74, 75, 76, 77, 193, 488 health, xv, 203, 210, 410, 521
fuzzy sets, 281, 282 health problems, 521
heart disease, 410
heat loss, 442
G heat transfer, xv, 431, 432, 445, 500
height, 133, 155
gas diffusion, 422
highway system, 170
Gaussian membership, 281
highways, 147, 149, 150, 151, 165, 183, 190
gel, 322
histogram, 129, 130
General Accounting Office, 204
historical data, 176
General Motors, 410, 417
history, 115, 191, 219, 223, 510
Geographic Information System, GIS, 210, 218, 219,
homes, 190
220, 222, 227, 228, 232
Hong Kong, 528
geometry, xv, 95, 96, 322, 385, 388, 395, 432, 436,
horses, 192
444, 446, 448
host, 187, 199, 200, 245, 435
Germany, 53, 54, 187, 202, 203, 216, 231, 311, 505,
hotel, 492
529
hotspots, 171, 243
global economy, 208
housing, 68, 443
global warming, 294
hub, 270, 315, 320, 497, 499, 500, 504
globalization, 212
human, xiv, 82, 166, 167, 186, 194, 195, 207, 208,
glue, 502
209, 210, 212, 218, 219, 224, 241, 512
glycol, 524
human behavior, xiv, 207, 224
goods and services, 210, 221
human resources, 209, 218
governance, ix, xiii, 207, 208, 209, 217, 220, 221,
humidity, 42, 116, 153
225
Hunter, 229
governments, 185, 188, 191, 208, 225
hunting, 67
GPS, 116, 165, 180, 240, 243, 244
hybridization, 323, 341
graduate students, xvi
hydrogen, 215, 315, 317, 318, 417, 455, 511, 522,
graph, 136
523, 524, 525, 527
graphite, 63, 65, 420, 425, 426, 432
hydrogen gas, 455
gravitational force, 91, 93
hydroxide, 415, 416, 417, 422, 455, 525
gravity, 89, 91, 92, 93, 486, 498
hydroxyl, 423
Great Britain, 226, 227
hypermedia, 237
Greece, 406
hypothesis, 16, 35, 36, 127, 133
greenhouse, 410
grids, ix, xiii, 72, 130, 186, 187, 190, 191, 203, 313,
315, 326, 345 I
grouping, 130
growth, 6, 208, 210, 220 ICC, 235, 245
guidance, 84, 110, 152, 166, 177, 189, 212, 213, 214, ICE, 250, 295, 296, 314, 318, 321, 326, 327, 410,
219 486, 488, 489, 490, 494
guidelines, 221 ICRA, 144, 145
Image generation, 73
Imaging, vii, 57, 58, 79
H imbalances, 345
immobilization, 454
habitat, 210
immunity, 170
Index 537
improvements, 215, 220, 249, 261, 410, 435, 486, isolation, 168, 315, 326, 327
516, 527 issues, x, xiii, 8, 131, 171, 185, 186, 190, 204, 208,
impulses, 71 219, 241, 242, 269, 317, 344, 410, 417, 420, 432,
impurities, 524 451, 454, 456, 502
income, 326 Italy, 230, 407
independence, 82 iteration, 31, 98, 99, 106, 467
individualization, 212
individuals, 185, 210, 217, 225
induction, xiv, 155, 156, 249, 251, 262, 263, 264, J
333, 353, 354, 360, 361, 362, 363, 364, 369, 371,
Japan, 142, 169, 211, 228, 420, 433, 497, 498, 529
372, 374, 375, 376, 377, 379, 381, 385, 389, 390,
joints, 486
392, 393, 395, 397, 400, 401, 402, 403, 404, 408
Induction machine, 261, 270, 354, 369
inductor, 34, 255, 256, 257, 258, 285, 360 K
industry(s), xii, xiii, xv, 6, 51, 57, 58, 82, 109, 165,
166, 185, 187, 202, 203, 217, 219, 220, 222, 224, kill, 192
381, 409, 414, 418, 420, 421, 425, 426, 427, 490, KOH, 415, 482
503 Korea, 507, 528
inertia, 88, 90, 91, 93, 96, 108, 365, 393, 395, 498 Kosovo, 191
infection, 186, 194
Information and Communication Technologies, 148
information exchange, 171 L
information sharing, 165
information technology, xiv, 190, 220, 435 lakes, 47
infrastructure, xi, xiv, 95, 148, 165, 166, 169, 173, lamination, 262
177, 178, 179, 180, 182, 187, 188, 190, 191, 202, languages, 71
203, 204, 205, 209, 210, 212, 215, 216, 218, 222, laptop, 196
224, 225, 228, 235, 241, 489, 511 lasers, 132
initial state, 287 latency, 236, 238
injuries, 6 law enforcement, xiii, 183
input signal, 13, 14, 375 laws, 89, 150, 203, 221
inspections, 169 lead-acid batteries, 434
insulation, 262, 381, 385, 454, 456 leakage, xii, 380, 383, 384, 392, 454
integration, ix, xi, 8, 98, 110, 118, 133, 216, 217, leaks, 195
224, 225, 294, 316, 317, 326, 328, 503, 509, 510, learning, 193, 198, 208, 281, 283, 284
512, 527 Least Developed Countries, 224
integrity, 195, 242 LED, 117, 176, 177
intelligence, 82, 186, 188, 216, 222, 244, 512 legislation, 524
intelligence gathering, 244 legs, 267
intelligent systems, xvi, 6, 212, 512 life cycle, 321, 322, 519
interface, xi, 8, 151, 167, 169, 171, 172, 173, 180, life expectancy, 431, 526
193, 197, 198, 202, 237, 251, 252, 315 lifetime, 323, 435, 511
interference, xiv, 16, 28, 37, 40, 53, 238, 239, 249, light, 6, 10, 45, 47, 48, 59, 117, 160, 317, 368, 438
261, 270 linear function, 30
interfering target, 20 linear model, 83
international standards, 237 linear systems, 32, 34
interoperability, 173, 221, 222, 223, 225, 241, 245, liquids, 490
320 lithium, xv, 322, 323, 409, 419, 420, 423, 425, 426,
intervention, 149, 152, 186 427, 428, 431, 432, 433, 434, 435, 437, 448
investment(s), 219, 221, 436, 511, 515 Lithium batteries, 431
ions, 413, 414, 418, 419, 420, 425, 432, 435 living environment, 223
Iran, xvi, 275, 293, 295, 303 localization, xiii, 57, 69, 129, 149
iron, 262, 426, 433, 434 loci, xv, 453, 457, 458, 465, 478
logistics, 215, 216
538 Index
longevity, 219, 231 microwaves, 167

low temperatures, 421 migration, 220
LPG, 486 military, 82, 109, 188, 194, 322
luggage, 490, 491 miniaturization, 179
luminosity, 164 mission(s), 81, 82, 84, 110
Luo, 246, 271, 506 mixing, 194
mobile communication, 165, 242
mobile device, 172
M mobile phone, 165, 191, 214
mobile robots, 57, 69
machine learning, 129
modelling, 241, 380, 395
magnet(s), 249, 252, 261, 265, 270, 333, 354, 355,
models, 18, 19, 87, 95, 99, 102, 103, 106, 135, 155,
364, 365, 375, 400, 402, 487, 502, 500, 502, 503,
165, 175, 177, 189, 196, 198, 208, 210, 217, 222,
504, 505
223, 241, 275, 276, 277, 281, 289, 318, 404, 431,
magnetic field, x, xiv, 364, 379, 381, 382, 383, 384,
432
385, 390, 392, 394, 402, 403, 404
modifications, 35, 410, 415
magnitude, 96, 259, 278, 287, 288, 345, 382, 391,
modules, 155, 168, 180, 189, 192, 193, 195, 220,
395, 404, 461
242, 416, 419, 420, 438, 439, 440
majority, 137, 138, 392, 488, 501, 505, 517, 522, 526
moisture, 425
Malaysia, 376
molecules, 42
malware, 186, 187, 191, 192, 193, 194, 202
momentum, 89, 90, 91, 93, 220
man, 214, 242
Monte Carlo method, 305
MANETs, 242
Monte Carlo Simulation, 528
manganese, 432, 434
morale, 188
manipulation, 88, 209, 502
morphology, 161
manufacturing, 393, 456
Moscow, 231, 435
mapping, 84, 123, 124, 125
motion control, 81, 84
market penetration, 241, 345, 509, 511, 522
motivation, 166, 426
market segment, 426, 492, 494
motor control, ix, 251, 260, 270, 321
marketing, xi, 417
multilateration, 3
masking, 22, 23, 24, 52
multimedia, 197, 237, 238, 512
mass, xv, 47, 63, 85, 88, 89, 91, 96, 108, 109, 116,
multiples, 217
118, 119, 212, 216, 316, 354, 411, 425, 455, 456,
multiplication, 12
485, 487, 488, 490, 491, 498, 499, 500, 505, 510,
music, 166
511, 516, 517, 522
materials, x, 62, 69, 262, 269, 317, 322, 410, 416,
422, 424, 425, 426, 433, 448, 450, 456, 518, 519, N
525, 526
matrix, 31, 32, 87, 88, 90, 96, 104, 105, 120, 121, nanotechnology(s), 322, 323, 409, 425, 426
122, 136, 138, 139, 180, 286, 425 National Academy of Sciences, x
matter, 151, 182, 208, 239 national emergency, 168
media, 77, 144, 197, 223, 529 national security, 188, 189
medicine, xiii, 57, 58, 69 NATO, 53, 54, 482
medium access control, 240 natural evolution, 486
membership, 282 natural gas, 514
memory, 69, 71, 166, 190, 193, 196, 455, 456, 482 natural resources, 219
messages, 166, 167, 170, 171, 200, 214, 215, 238, navigation, 53, 58, 78, 111, 143, 145, 146, 211, 213,
240, 241 214, 229
metals, 419, 434 navigation system, xii, 6, 183, 214, 243, 512
meter, 42, 155, 156, 157 neglect, 189, 195
methanol, 524 network operators, 216
methodology, 58, 69, 177 networking, 241
microcrystalline, 455 neural network(s), 267, 512
Microsoft, 192, 193, 194 Neuro-fuzzy, 280
Index 539
neutral, 93, 267, 307, 489 329, 333, 334, 335, 344, 357, 362, 419, 438, 440,
next generation, 293, 322, 431 441, 442, 456, 458, 488, 522, 523
Ni-Cd batteries, 415, 416, 426, 454, 455, 456, 481 parallel implementation, 135
nickel, 409, 415, 417, 418, 419, 434, 437, 454, 455, parameter estimation, 83, 102, 104
456 partial differential equations, 84
niobium, 433 participants, 115, 170, 172, 297
nodes, 168, 188, 189, 199, 202, 300 partition, 171
noise power estimation, 11, 37, 52 password, xiii, 185, 199, 202
nonlinear systems, 83 pattern recognition, 58, 69
non-renewable resources, 210 pedal, 319, 326
normal distribution, 298 performance indicator, 319, 321
North America, 244 permeability, 158
Norway, 227 permission, 345
NPS, 83, 110 permit, 149, 150, 155, 157, 177
nucleation, 455 permittivity, 49
null, 384, 388, 391, 404 personal communication, 165
numerical computations, 95 personal computers, 191, 192
personal life, 216
PES, 358, 359, 360, 365, 367, 376, 528
O Petroleum, 511, 527
phosphate(s), 433, 434
obstacles, 59, 116, 117, 119, 132, 133, 497
photovoltaic panels, 314, 315
oceans, 218
physical interaction, 169
octane, 514
physical properties, xv, 432, 448, 449
octane number, 514
physical structure, 252, 265
Office of Management and Budget, 227
pipeline, 511
oil, 188, 489, 490, 511, 513, 524, 525
pitch, 85, 86, 92, 499
one dimension, 124, 130
plants, 521
operating range, 450
platform, 196, 219, 228, 241, 327
operating system, 161, 183, 190, 191, 192, 193, 198
playing, 218
operations, xi, 82, 84, 185, 187, 192, 216, 260, 267,
PM, 265, 314, 319, 333, 506
270, 277, 398, 399, 400, 401, 402, 404
pneumonia, 410
opportunities, xii, 219, 325, 326
Poland, 53, 54
optimal performance, 13
polar, 386
optimization, 60, 122, 135, 297, 305, 309, 316, 317,
polarity, 364, 387, 390, 392
494
police, 147, 150, 168
organism, 410
policy, 180, 188, 208, 209, 211, 221, 231, 232, 487
originality, 382, 392
policy making, 209
oscillation, 395
politics, 149
overlap, 130, 141
pollutants, 317, 521
ownership, 209
pollution, xi, 149, 189, 318, 410, 411, 521, 527
oxidation, 413, 423
polymer, 322, 324, 334, 419
oxygen, 322, 414, 416, 420, 422, 423, 424, 425, 454,
poor performance, 59
455
population, 202, 524
oxygen absorption, 454
portability, 75, 76
ozone, 521
Portugal, 272
postal service, 511
P potassium, 415, 422, 455
power generation, 193, 414, 513
Pacific, 229 power plants, 194, 277, 489
parallel, xiv, 15, 60, 66, 86, 114, 120, 127, 134, 135, precipitation, 155
173, 190, 223, 295, 296, 310, 313, 316, 317, 328, preparation, 263
preservation, 210, 223
540 Index
prevention, 242, 244 radius, 10, 29, 63, 107, 130, 171, 491, 501
primary cells, 433 rain attenuation, 45, 46
principles, 35, 36, 103, 224, 364, 527 rainfall, 45, 47, 51
prior knowledge, 28 ramp, 149, 152, 161, 215, 263
private sector, 209 rate of change, 90
probability, 5, 11, 12, 14, 15, 16, 17, 18, 19, 20, 21, rating scale, 499
25, 36, 37, 38, 39, 46, 47, 52, 104, 123, 125, 127, RB1, 66
128, 131, 134, 135, 204, 297, 298, 300, 303, 305, reaction time, 236
306, 493, 494, 501, 505 reactions, xii, 413, 416, 425, 434, 454
probability density function, 18, 306, 493 reading, 13, 14
probability distribution, 297, 298, 305 real numbers, 105
Probability of Detection, 46, 54 real time, 5, 11, 37, 52, 59, 62, 76, 134, 135, 147,
probe, 176, 177 149, 150, 161, 182, 225
producers, 303, 318, 497 reality, 44, 116, 182, 194, 208, 453, 454, 473, 479,
product life cycle, 177 504
professionals, xii, xv, 212, 224 reasoning, 218
profit, xiv, 209, 293, 298, 302, 303, 309, 311 recall, 113, 131, 135, 136, 137, 138, 139
programming, xiv, 166, 197, 204, 293, 294, 297, reception, 63
298, 300, 301, 303, 304, 307, 308, 309 recognition, xii, 11, 57, 58, 74, 124, 147, 148, 150,
project, 53, 59, 68, 77, 172, 173, 177, 208, 221 154, 160, 161, 164, 165, 167, 183, 189, 242
proliferation, 218 recombination, 414
propagation, 8, 58, 241, 283 recommendations, 311
proportionality, 35 recovery, 82, 168, 182, 203, 412
propylene, 435 rectification, 253
protection, xiii, 179, 186, 187, 189, 194, 202, 203, recurrence, 155
210, 242, 322, 416 recycling, 422, 456, 524
prototype(s), 114, 115, 116, 119, 131, 132, 173, 353, redundancy, 168, 220, 244
354, 376, 410, 417, 420, 431, 432, 436, 503, 511 reference frame, 86, 98, 387
public opinion, 202 reference system, 89
publishing, 203 reflectivity, 116
pumps, 251 regression, 103, 125
PVA, 482 rejection, 129
relaxation, xv, 453, 454, 456, 457, 458, 459, 460,
461, 464, 465, 466, 467, 469, 472, 477, 478, 479,
Q 480, 482, 483
relaxation process, 453, 457, 480
quadratic programming, 122
relaxation times, 480
quality of life, 211
relevance, 209, 236
quality of service, 167, 222, 240
reliability, ix, xii, xiv, 70, 118, 155, 178, 180, 194,
quantification, 121
195, 196, 212, 249, 261, 263, 265, 276, 319, 322,
quantization, 71
325, 333, 380, 410, 486, 502
Queensland, 198
remote sensing, 11
renewable energy, 310, 311, 511, 527
R renewable fuel, 215
requirements, ix, xiv, 8, 82, 119, 167, 170, 172, 188,
racing, 498 195, 196, 202, 207, 208, 224, 249, 252, 255, 260,
radar, ix, xiii, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 18, 19, 261, 262, 315, 317, 319, 321, 322, 345, 410, 413,
20, 24, 28, 29, 30, 31, 32, 36, 38, 41, 42, 43, 44, 426, 440, 441, 455, 485, 493, 494, 495, 503, 511,
45, 47, 48, 49, 51, 52, 53, 59, 113, 116, 118, 120, 512, 523, 524, 526
132, 144, 147, 148, 150, 155, 156, 242 RES, 314, 315
radiation, 47, 63, 117, 155, 156 research institutions, 486
radio, 11, 155, 166, 167, 168, 169, 170, 171, 172, researchers, xii, xv, xvi, 185, 198, 217, 220, 419, 423
182, 189, 212, 215, 236, 237, 238, 239, 240 reserves, 276, 311, 410, 426
Index 541
residential, 291 sawdust, 523

resilience, 188 scaling, 12, 14, 15, 16, 17, 25, 37
resistance, 34, 63, 255, 262, 323, 356, 360, 397, 414, scattering, 45
426, 455, 458, 461, 478, 480 schema, 65, 394
resolution, 19, 61, 73, 76, 117, 118, 120, 147, 150, science, 220
180, 265 scientific understanding, 208
resources, xv, 149, 168, 179, 190, 193, 209, 210, scope, 208, 221, 223, 513, 515
217, 218, 219, 220, 222, 223, 224, 225, 294, 436, SDSR, 203
488, 489 sea level, 43, 44
response, xiv, 32, 64, 71, 75, 102, 149, 167, 169, second generation, 487
186, 190, 197, 198, 199, 203, 279, 282, 283, 311, Second World, 524
315, 321, 325, 326, 353, 355, 371, 372, 374, 375, sectoral policies, 221
384, 456, 457, 459, 461, 464, 467, 469, 471, 472, security, ix, xiii, 84, 147, 176, 179, 183, 185, 186,
477, 517 187, 188, 189, 190, 191, 192, 194, 195, 196, 198,
response time, 167, 190, 315 202, 203, 204, 205, 213, 239, 242, 244, 311, 420
responsiveness, 218, 219 security services, 188
restaurants, 166 semicircle, 458, 459
restoration, 311 semiconductor(s), 118, 144, 257, 265
restrictions, 147, 148, 150, 189, 215 senses, 216, 375
revenue, 243, 276 sensing, ix, xiii, 62, 69, 82, 113, 114, 115, 116, 119,
rights, 221, 244 142, 143
rings, 262, 393 sensitivity, 4, 52, 62, 68, 117, 264, 285, 341, 381,
risk(s), 82, 133, 151, 185, 188, 190, 294, 297, 301, 384, 389, 391, 511
302, 303, 304, 307, 309, 310, 311, 420, 434, 435, sensor network, 8, 28, 29, 30, 31, 32, 53
503 servers, 179, 180, 183, 192, 199
risk assessment, 188 service provider, 238
risk aversion, 307 service quality, 210
risk management, 297 services, xiii, 147, 150, 165, 166, 168, 172, 173, 174,
robotics, 57, 59, 96 175, 176, 188, 199, 203, 204, 212, 215, 216, 219,
Romania, xvi, 57, 78, 113, 143, 144, 147, 184, 185, 222, 223, 237, 238, 241, 242, 244, 276, 294, 310,
207, 313, 353, 377, 407, 409, 427 311, 325
room temperature, 324, 449 shape, 58, 62, 67, 68, 74, 82, 95, 109, 116, 130, 194,
root, 254, 300 221, 372, 374, 445
rotations, 87, 493 shock, 426, 517, 523
routes, 147, 149, 151, 175, 177, 183, 239 shortage, 512
routines, 160, 168 showing, 83, 131, 454, 459, 480, 492, 493
ROVs, 82 side effects, 191, 521
rules, 176, 191, 281, 317 signalling, 154, 172, 237, 242
Russia, 111, 231, 406 signals, 5, 8, 10, 14, 22, 23, 24, 25, 37, 38, 40, 41,
52, 60, 62, 65, 66, 67, 70, 71, 74, 76, 154, 166,
167, 169, 170, 171, 180, 216, 251, 252, 262, 263,
S 266, 267, 270, 285, 326, 363, 365, 366, 370, 372,
374, 375, 380, 381, 392
safety, ix, xiii, 3, 4, 5, 6, 7, 8, 20, 51, 53, 59, 143,
signal-to-noise ratio, 4
144, 147, 148, 149, 150, 167, 169, 183, 188, 195,
signs, 115, 116, 120, 146, 148, 149, 150, 161, 172,
210, 212, 214, 215, 216, 229, 236, 237, 238, 240,
183
244, 319, 322, 415, 421, 423, 426, 427, 432, 433,
silicon, 322, 426, 428
435, 455, 456, 498, 517
Singapore, 229
Safety Management System, 211
skewness, xv, 453, 478, 479, 480, 481
salinity, 96
slaves, 171
salts, 424, 435
Small particles, 521
saturation, 389
smog, 58, 521
Saudi Arabia, 235
social interactions, 209
savings, 176, 177
542 Index
social responsibility, 208, 511 439, 488, 509, 510, 511, 515, 516, 517, 519, 522,
society, 167, 187, 208, 210, 216, 217, 219, 220, 223, 525, 526, 527
224, 225, 230, 410, 435, 486, 505, 529 strategic planning, 221
sodium, 409, 418 strategy use, 100
software, 6, 59, 62, 74, 76, 95, 148, 149, 150, 154, stress, 199, 261, 320, 385, 455
161, 168, 180, 185, 191, 192, 193, 194, 195, 196, stress test, 199
198, 204, 205, 217, 219, 220, 267, 436 stress testing, 199
solution, xiii, 28, 30, 31, 51, 52, 58, 60, 66, 69, 76, structuring, 219
92, 122, 147, 148, 160, 163, 179, 180, 182, 183, subscribers, 168
188, 202, 220, 223, 297, 299, 354, 375, 414, 419, substitutes, 210
422, 448, 450, 481, 487, 501, 502, 509, 510, 517, substitution, 198, 199
524, 525, 527 subtraction, 121
solvents, 424, 435 succession, 134, 161
Sonar, 58, 59, 77, 78, 110 suicide, 192
South Africa, 506 sulfur, 322
South Korea, 3, 54, 310 sulphur, 418, 426
Spain, 78, 79, 143, 146, 272, 310, 407, 431, 435 Sun, 77, 111, 347, 428, 451
spam, 192 Super Bowl, 145
spatial information, 209, 217, 220 supervision, x, xiv, 168, 379, 380, 381, 382, 404
specialists, xii, 178, 185, 187, 209, 316, 317 supervisor(s), 178, 182
species, 454 supplier(s), 182, 183
specifications, 182, 217, 223, 238, 242, 245, 269, surface area, 108, 425
440, 491, 495 surveillance, ix, xiii, 82, 147, 148, 149, 150, 151,
spectral component, 381, 389 152, 179
spectroscopy, 481 Surveillance, vii, 79, 147, 150
speech, 11 survival, 219
Speed detection, 148, 150, 160 survivors, 220
spending, 160, 240 sustainability, xiii, 207, 208, 210, 217, 223, 224,
spin, 117, 134, 529 225, 226, 227, 271, 515
sponge, 414 sustainable development, 208, 209, 210, 216, 222,
Spring, 232, 245 225, 226, 231
spyware, 192 sustainable energy, 515
St. Petersburg, 111, 406 Switzerland, 226, 419
stability, 37, 66, 94, 102, 118, 220, 284, 294, 425, symmetry, 91, 162
426, 432, 498, 499 synchronization, 8, 240
stabilization, 334 synthesis, 207, 270
stabilizers, 325
stakeholders, 217, 218
standard deviation, 37, 305, 478 T
standardization, 191, 222, 223, 320
tanks, 525, 526
starch, 524
tariff, 302
statistics, xiii, 5, 14, 16, 18, 35, 52, 148, 150, 161,
taxes, 161, 169
177, 199, 410
teachers, v
steel, 419
teams, 148, 150, 178
Steganography, 186
technical support, 180
stimulation, 82
technician, 182
stochastic model, 310
techniques, xiii, 11, 38, 52, 57, 59, 81, 110, 117, 125,
Stochastic programming, 294, 300, 309
152, 164, 165, 170, 183, 186, 194, 195, 204, 210,
stock, 503
220, 252, 260, 261, 265, 270, 281, 316, 320, 328,
storage, xi, xii, xiv, xv, 71, 180, 182, 183, 209, 219,
358, 380, 381, 391, 526
257, 275, 276, 277, 278, 280, 287, 289, 290, 294,
technological advances, 217, 223
310, 315, 317, 318, 324, 334, 431, 432, 433, 437,
technological developments, ix, 380, 512
technological progress, 165
Index 543
telecommunications, 165, 166, 420 troubleshooting, 200

Telematics, 235, 237 turbulence, 155
telephone(s), 165, 433 Turkey, 81, 249, 272
temperature, xv, 42, 43, 44, 95, 155, 262, 321, 357,
362, 418, 419, 431, 432, 435, 445, 447, 448, 449,
450, 455, 524 U
temporal variation, 14
UK, 54, 225, 226, 345, 376, 407, 505, 506, 509, 519,
terminals, 62, 183, 262, 397, 400
529
territorial, 221
Ultracapacitor, 314, 346, 348, 351
territory, 209, 221, 293
ultrasonic frequency, 57
terrorist attack, 203
ultrasound, xii, xiii, 57, 58, 61, 69, 70
test data, 136
underwater vehicles, ix, xii, xiii, 95
testing, xiv, xv, 65, 95, 114, 125, 131, 136, 198, 224,
UNESCO, 231
353, 371, 372, 375, 431, 487, 511
unidirectional links, 239
textbooks, 261, 265
uniform, 45, 72, 162, 443
texture, 58
United Kingdom, 187, 202, 203
theft, 202
United Nations, 208, 224, 225, 226, 227, 530
thermal energy, 517
United States (USA), 53, 54, 77, 145, 146, 169, 187,
threats, xiii, 185, 187, 188, 190, 191, 192, 196, 203,
188, 203, 204, 205, 211, 227, 228, 229, 230, 240,
216
245, 272, 311, 410, 416, 417, 505, 520, 528
time frame, 98, 215
universities, 114, 198, 410
tin, 426
updating, 209
titanium, 432, 434
uplink, 171, 238
topology, 179, 215, 218, 239, 251, 258, 267, 269,
urban, 115, 119, 143, 151, 172, 173, 241, 244, 326,
316, 318, 320, 321, 323, 324, 325, 326, 327, 328,
369, 415, 486, 490, 491, 492, 495, 496
333, 334, 335, 336, 337, 338, 341, 344, 345, 412,
urban areas, 172, 369
488, 490, 497, 500, 501
urbanization, 212
total costs, 177
user data, 242
toxic metals, 322
Toyota, 318, 354, 417, 420, 421, 496, 511, 522, 523,
528 V
toys, 454
tracks, 504 vacuum, ix
trade-off, 132, 194, 325 valence, 413
traffic information, 166 validation, xv, 125, 161, 342, 379, 404
traffic violations, 148, 150 valve, 416
training, 122, 125, 126, 127, 128, 129, 136, 137, 219, variables, 83, 102, 122, 264, 281, 282, 297, 298, 301,
282, 283 303, 326
trajectory, 83, 95, 243 variations, 12, 49, 95, 155, 253, 258, 264, 276, 280,
transactions, xii, 186, 194 288, 382, 393, 397, 434
transducer, 62, 63, 64, 65, 66, 68, 70, 71, 75, 359 vector, 30, 32, 87, 88, 89, 91, 92, 95, 97, 98, 99, 103,
transformation, 86, 88, 263 104, 121, 122, 123, 124, 126, 127, 128, 129, 133,
transformation matrix, 86, 88 134, 135, 163, 164, 185, 262, 263, 264, 282, 283,
transistor, 67 285, 286, 380
transition metal, 419 vegetable oil, 524
transition period, 494 velocity, 4, 5, 6, 7, 8, 9, 10, 18, 19, 22, 51, 83, 84,
translation, 163 85, 86, 87, 88, 89, 90, 92, 95, 97, 101, 105, 108,
transmission, 8, 62, 71, 170, 171, 183, 188, 214, 237, 135, 262, 393, 394, 444, 445, 450, 517
259, 294, 295, 490, 526 ventilation, 357, 362
transparency, 204, 208 versatility, 82
transportation infrastructure, 212 vibration, 63, 393, 499
treatment, 397, 490 virtualization, 183
trial, 467 viruses, 186, 187, 191, 192, 195, 196
544 Index
viscosity, 524 wholesale, 293

Visibility, 48, 50 Wi-Fi, 171, 187, 202
vision, 57, 58, 117, 146, 191, 208, 212, 213, 215, wind turbines, 276, 315, 515
216, 221, 225 windows, 13, 15, 16, 26, 38, 130, 180
visualization, 219 wireless networks, 202
vocabulary, 130 wireless systems, 237
volatile organic compounds, 521 wires, 182, 397
Volkswagen, 494 Wisconsin, 146, 275
vote, 125, 126 withdrawal, 421
voting, 125, 126, 132, 137, 138, 140 wood, 523, 524
vulnerability, 191, 199, 225 workflow, 223
working conditions, 178, 179, 492, 494, 496, 499
workstation, 180, 182
W world order, 222
World Wide Web, 223
waste, xv, 210, 432, 487, 515, 517, 523
worldwide, 59
water, xiii, 42, 43, 44, 47, 49, 50, 84, 86, 95, 96, 115,
worms, 186, 192
185, 188, 189, 193, 202, 203, 210, 414, 416, 422,
425, 439, 447, 448, 450, 489, 498, 501, 502, 524,
525 X
water quality, 210
water supplies, 203 XML, 222
water vapor, 425
waterways, 238
weapons, 186 Z
wear, 382, 393
Zebra batteries, 418, 419
web, 173, 175, 191, 198, 220, 230
zirconium, 433
web browser, 191
ZnO, 483
weight ratio, 505
Western countries, 527

Autonomous Vehicles-Intelligent Transport Systems and Smart Technologies PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Autonomous Vehicles-Intelligent Transport Systems and Smart Technologies PDF

Uploaded by

Copyright:

Available Formats

ENGINEERING TOOLS, TECHNIQUES AND TABLES

Additional books in this series can be found on Nova‟s website

Additional e-books in this series can be found on Nova‟s website

NOTICE TO THE READER

Library of Congress Cataloging-in-Publication Data

Published by Nova Science Publishers, Inc. † New York

February 15, 2014

Academician Arif M. HASHIMOV

Environmentally friendly, high efficiency, green power charging, renewable sources

applications in intelligent electrical vehicles and automotive applications in smart grids

Pitesti, Romania Nicu Bizon

Poitiers, France Lucian Dascalescu

Tabriz, Iran Naser Mahdavi Tabatabaei

February 20, 2014 Editors

RADAR SENSOR DETECTORS FOR VEHICLE

Modar Safir Shbat and Vyacheslav Tuzlukov†

AWGN Additive White Gaussian Noise

2. INTELLIGENT VEHICLE AND TOTAL SAFETY DRIVING APPROACH

Figure 1. Safety driving applications.

Table 1. Typical radar sensor systems used in safety driving applications

Range/Velocity/View Sensor Radar Sensor Operation

Table 2. Comparison of the main features of radar sensor systems

Criteria Radar Sensor System

3. OVERVIEW OF MRR AND SRR SYSTEMS

Figure 2. Overview of MRR and SRR systems.

Figure 3. LFMCW radar sensor waveform.

where AT is the amplitude, and T (t ) is the instantaneous phase defined as

Figure 4. Normalized detection threshold versus the probability of false alarm.

Figure 5. CFAR detector functioning diagram.

 There are no any interfering target vehicles.

4.1. CA-CFAR Detector

Figure 6. The main structure of the CA-CFAR detector.

The CA-CFAR detection threshold is defined as follows

THRCA   CA n2CA

4.2. OS-CFAR Detector

Figure 7. The main structure of the OS-CFAR detector.

The scaling factor in the case of OS-CFAR procedure is given by

Finally, the OS-CFAR detection threshold is determined as:

THROS   OS  n2OS   OS X ( k ) . (21)

4.3. GCML Detector

order of the amplitude as {X (1)  X (2)    X ( M ) } and {X ( M 1)  X ( M 2)    X ( 2M ) } ,

Figure 8. The main structure of the GCML detector.

THRGCML   GCML n2GCML (23)

4.4. ACGO-CFAR Detector

The ACGO-CFAR detector can be considered as a modified version of the GCML

Figure 9. The main structure of the ACGO-CFAR detector.

The noise power estimation is set to be a maximum of U and W :

 n2ACGO  ˆ n2  max(U ,W ) . (25)

THRACGO   ACGO n2ACGO . (26)

5. CFAR DETECTION PERFORMANCE

 There are no interfering target vehicles;

where  av denotes the average RCS fluctuations of total number of targets.

Figure 11. Losses of the fluctuations described by the Swerling models.

Table 3. Main simulation parameters

Simulation Parameters Values

Based on simulation, the probability of detection PD is determined after K observations, as

Based on simulation, the probability of detection PD is determined after K observations, as

Figure 19 demonstrates a comparison of detection performances between the CA-CFAR

Figure 17. The target masking phenomenon of the CA-CFAR detector.

Figure 18. OS-CFAR detection of two closely spaced target vehicles.

Table 4. The scaling constants for censoring procedure and PFC  10 3

Figure 21 demonstrates a comparison between the OS-CFAR, GCML, and ACGO-CFAR

Figure 25. Comparison of detection performance six interfering target vehicles.