PETTERI MÄKELÄ LOCAL POSITIONING SYSTEMS AND INDOOR NAVIGATION

PETTERI MÄKELÄ LOCAL POSITIONING SYSTEMS AND INDOOR NAVIGATION
PETTERI MÄKELÄ
LOCAL POSITIONING SYSTEMS AND INDOOR NAVIGATION
Licentiate of Science Thesis
The subject has been approved at the
Department of Electrical Engineering
council meeting on May 7, 2008
Examiners: Professor Markku Kivikoski
Professor Jarmo Takala
iii
ABSTRACT
TAMPERE UNIVERSITY OF TECHNOLOGY
Degree Programme in Electrical Engineering
Mäkelä, Petteri: Local Positioning Systems and Indoor Navigation
Licentiate of Science Thesis, 116 pages
June 2008
Major: Electronics
Examiners: Professor Markku Kivikoski, Professor Jarmo Takala
Keywords: local positioning system, indoor navigation, WLAN, UWB, AGPS
The market of location-based services is anticipated to grow strongly in the near future.
Map-based guidance and navigation, intelligent transport services and tracking of
people and valuable assets, are examples of typical positioning applications. Many
location-based services depend on Global Positioning System (GPS) as a position data
provider. However, there is also a growing interest in positioning applications which are
independent from satellite-based navigation technologies. These positioning systems,
which are designed to operate inside a limited geographical area, are called local
positioning systems. Local positioning systems are needed in indoor navigation and in
wireless sensor networks, which measure various environmental parameters.
This thesis is a literature research, which considers the most promising current and
emerging local positioning technologies. The research aims to match the characteristics
of the identified local positioning techniques to the needs of various location-based
applications. Various types of signals, sensors, observables and computation algorithms
used with these positioning technologies are presented and compared.
There is no single technology that may be relied upon in all environments to provide
accurate location information. However, there clearly are a few promising technologies
which outperform certain others, depending on what the purpose of usage is. The most
widely used local positioning systems today are based on the use of wireless local area
network (WLAN) signals. Of these systems, the system based on location fingerprinting
method was found the most promising. On the other hand, technology based on signal
propagation time measurements of ultra-wideband (UWB) was found the most
interesting emerging local positioning technology due to the accuracy and performance
of UWB in multipath conditions. Also some other local positioning technologies, such
as computer vision based positioning and dead-reckoning method, were identified as
suitable technologies for certain applications.
Based on this research it can be anticipated that in the future fusing indoor positioning
techniques with the satellite-based navigation systems would be one of the most
important research topics. Integrating a local positioning technology, such as UWB
positioning, with GPS as a single location based service would open remarkable new
market opportunities.
iv
v
TIIVISTELMÄ
TAMPEREEN TEKNILLINEN YLIOPISTO
Elektroniikan koulutusohjelma
Mäkelä, Petteri: Paikalliset paikannusjärjestelmät ja sisätilanavigointi
Lisensiaattityö, 116 sivua
Kesäkuu 2008
Pääaine: Elektroniikka
Tarkastajat: professori Markku Kivikoski, professori Jarmo Takala
Avainsanat: paikannus, sisätilapaikannus, WLAN, UWB, AGPS
Paikannuspalveluiden markkinoiden on ennustettu kasvavan lähivuosina voimakkaasti.
Esimerkkejä paikannuspalveluista ovat erilaiset kartta- ja reittipalvelut sekä tavaroiden
ja ihmisten seuraamiseen tarkoitetut sovellukset. Nykyiset paikannuspalvelut perustuvat
enimmäkseen satelliittipaikannusjärjestelmien tuottamaan sijaintitietoon. Monissa
sovelluksissa halutaan kuitenkin käyttää satelliiteista riippumattomia paikannusteknologioita, jotka perustuvat paikallisten verkkojen käyttöön. Paikallisiin verkkoihin perustuvia navigointiteknologioita tarvitaan erityisesti sisätiloissa tapahtuvassa paikannuksessa
ja ympäristön mittaamiseen tarkoitetuissa langattomissa anturiverkoissa.
Tämä lisensiaattityö on kirjallisuustutkimus, jonka tarkoituksena on esitellä tärkeimmät
paikallisiin verkkoihin perustuvat paikannusteknologiat. Työn tavoitteena on tunnistaa
kunkin paikannusteknologian erityispiirteet ja etsiä tiettyyn käyttötarkoitukseen
parhaiten soveltuva paikannusteknologia. Työssä käsitellään yleisimmät paikannusjärjestelmissä käytetyt signaalit, anturit, havaintosuureet sekä laskentamenetelmät.
Mikään yksittäinen paikannusteknologia ei sovellu kaikkiin ympäristöihin ja sovelluksiin. Parasta tarkkuutta vaativiin sovelluksiin tarvitaan eri teknologia kuin sovelluksiin,
joissa paikannuspalvelun tulee toimia laajalla alueella. Monet nykyisistä sisätilapaikannusjärjestelmistä perustuvat langattoman lähiverkon (wireless local area
network, WLAN) käyttöön. Tässä tutkimuksessa korrelaatio- eli sormenjälkipaikannukseen (location fingerprinting) perustuva paikannus todettiin sisätiloihin sopivimmaksi
WLAN-pohjaiseksi paikannusmenetelmäksi. Uusimmista teknologioista ultralaajakaistaisen (ultra-wideband, UWB) signaalin kulkuajan mittaamiseen perustuva paikannusteknologia todettiin lupaavimmaksi. UWB-signaalin lyhyt pulssi mahdollistaa tarkan
saapumisajan mittauksen ja hyvän suorituskyvyn olosuhteissa, joissa esiintyy
monitieheijastuksia. Toisaalta rajatuille alueille tarkoitettuihin sovelluksiin voivat sopia
myös muut menetelmät, kuten sijainnin haku videokuvasta tai vektoripaikannus (deadreckoning).
Tämän tutkimuksen perusteella on nähtävissä, että satelliittipaikannuksen ja paikallisten
navigointiverkkojen yhteiskäyttö tulee olemaan eräs keskeisimmistä tutkimuksen
aiheista. GPS:n ja esimerkiksi UWB-paikannuksen yhdistäminen yhdeksi paikannuspalveluksi avaisi merkittäviä uusia markkinoita.
vi
vii
Preface
First, I would like to express my appreciation to the supervisor of this research,
Professor Markku Kivikoski of Tampere University of Technology. His guidance and
the excellent lectures given in Seinäjoki have made this research possible. I am also
thankful to the second supervisor of this thesis, Professor Jarmo Takala of Tampere
University of Technology.
Also my employer, Seinäjoki University of Applied Sciences, has supported this
research. I am especially grateful to the directors of the School of Engineering, Dr.
Tech. Jukka Aarnio and Lic. Tech. Jorma Nevaranta for enabling to write this thesis.
Finally, I would like to thank my wife M.S.S., M.A. Maria Mäkelä for helping me to
improve the language of this text, and my children Anna, Ossi and Elina for their
understanding and patience.
Seinäjoki, May 2008
Petteri Mäkelä
viii
ix
Table of Contents
ABSTRACT .................................................................................................................. III
TIIVISTELMÄ............................................................................................................... V
PREFACE .................................................................................................................... VII
LIST OF FIGURES ...................................................................................................XIV
LIST OF TABLES .....................................................................................................XVI
LIST OF ABBREVIATIONS ............................................................................... XVIII
LIST OF SYMBOLS ............................................................................................... XXII
1
INTRODUCTION ................................................................................................... 1
1.1
1.2
1.3
1.4
1.5
2
BACKGROUND .................................................................................................... 2
RESEARCH OBJECTIVES AND CONTRIBUTION ..................................................... 3
SCOPE OF THE RESEARCH ................................................................................... 3
MAIN RESULTS .................................................................................................. 4
OUTLINE OF THE THESIS..................................................................................... 5
PRINCIPLES OF LOCAL POSITIONING SYSTEMS ..................................... 7
2.1
INTRODUCTION................................................................................................... 7
2.2
DEFINITION OF LOCAL POSITIONING SYSTEM .................................................... 7
2.3
NEED FOR THE LOCAL POSITIONING SYSTEMS ................................................... 8
2.4
TAXONOMY OF POSITIONING SYSTEMS .............................................................. 9
2.4.1 Classification Based on Type of the Location Information ......................... 10
2.4.2 Classification Based on where Position Estimation Takes Place ............... 11
2.4.3 Classification Based on Scale of the Positioning System ........................... 12
2.4.4 Classification Based on Signal Technologies ............................................. 12
2.4.5 Classification Based on Observables .......................................................... 13
2.4.6 Classification Based on Location Estimation Techniques .......................... 13
2.5
ACCURACY METRICS FOR POSITIONING SYSTEMS ........................................... 14
2.5.1 Root of Mean Square Error ........................................................................ 15
2.5.2 Distance Root Mean Square Error ............................................................. 15
2.5.3 Circular Error Probability.......................................................................... 16
2.5.4 The Effect of Geometry and Range Error Factors ...................................... 16
2.6
OTHER PERFORMANCE METRICS ...................................................................... 18
2.7
PRIVACY OF LOCATION INFORMATION ............................................................. 19
x
3
POSITIONING SENSORS AND SIGNALS ...................................................... 23
3.1
RADIO FREQUENCY .......................................................................................... 23
3.1.1 Modeling Radio Signal Propagation .......................................................... 24
3.1.2 Multipath ..................................................................................................... 24
3.2
INFRARED-BASED SYSTEMS ............................................................................. 25
3.3
ULTRASOUND................................................................................................... 26
3.4
OPTICAL ........................................................................................................... 27
3.5
INERTIAL NAVIGATION SYSTEMS ..................................................................... 28
3.6
DC ELECTROMAGNETIC ................................................................................... 29
4
OBSERVABLES ................................................................................................... 31
4.1
PROXIMITY ....................................................................................................... 31
4.2
RECEIVED SIGNAL STRENGTH .......................................................................... 32
4.3
ANGLE OF ARRIVAL ......................................................................................... 34
4.3.1 Signal Strength Direction Finding .............................................................. 35
4.3.2 Phase Difference Direction Finding ........................................................... 36
4.3.3 Doppler Direction Finding ......................................................................... 36
4.4
TIME OF ARRIVAL ............................................................................................ 37
4.4.1 Measuring Time of Arrival at Multipath Conditions .................................. 37
4.4.2 TOA Measurement Methods........................................................................ 42
4.5
TDOA MEASUREMENT METHODS ................................................................... 48
5
LOCATION ESTIMATION ALGORITHMS ................................................... 51
5.1
TIME OF ARRIVAL TECHNIQUES ....................................................................... 51
5.2
TIME DIFFERENCE OF ARRIVAL POSITIONING TECHNIQUES ............................. 56
5.2.1 Taylor Series Method for Hyperbolic Equations ........................................ 57
5.2.2 Comparing Taylor Series Method to Other Methods.................................. 59
5.3
ANGLE OF ARRIVAL TECHNIQUES .................................................................... 61
5.4
LOCATION FINGERPRINTING ............................................................................ 63
5.4.1 Offline Phase ............................................................................................... 65
5.4.2 Online Phase ............................................................................................... 66
5.4.3 Improvements to the NNSS Method ............................................................ 67
5.4.4 Effect of the Environment and Infrastructure on Performance .................. 67
5.4.5 Radio Propagation Model ........................................................................... 70
5.4.6 Aliasing ....................................................................................................... 70
6
INDOOR GPS ....................................................................................................... 73
6.1
6.2
INTRODUCTION ................................................................................................ 73
ASSISTED GPS AND HIGH SENSITIVITY GPS ................................................... 74
xi
6.2.1 GPS Signal Acquisition ............................................................................... 74
6.2.2 Assisted GPS ............................................................................................... 76
6.2.3 Weak GPS Signal Processing ..................................................................... 76
6.3
PSEUDOLITES ................................................................................................... 77
6.4
GPS REPEATER ................................................................................................ 78
7
LOCAL POSITIONING SYSTEMS ................................................................... 79
7.1
ULTRASOUND-BASED POSITIONING SYSTEMS.................................................. 79
7.1.1 MIT Cricket System ..................................................................................... 80
7.1.2 Active Bat System ........................................................................................ 81
7.2
WLAN-BASED POSITIONING SYSTEMS ............................................................ 81
7.2.1 Aeroscout Positioning System ..................................................................... 82
7.2.2 Ekahau System ............................................................................................ 84
7.3
UWB-BASED POSITIONING SYSTEMS .............................................................. 86
7.3.1 Advantages of UWB .................................................................................... 87
7.3.2 Applications ................................................................................................ 87
7.3.3 UWB in Positioning .................................................................................... 88
7.3.4 Standards and Regulations ......................................................................... 89
7.4
FUSING POSITIONING TECHNOLOGIES .............................................................. 94
8
CONCLUSIONS AND DISCUSSION ................................................................ 97
8.1
8.2
8.3
8.4
8.5
9
COST AND MARKET ISSUES .............................................................................. 97
ACCURACY AND AVAILABILITY REQUIREMENTS OF LOCATION-BASED
SERVICES ......................................................................................................... 98
COMPARISON OF THE SIGNAL TECHNOLOGIES ............................................... 100
CONCLUSIONS ................................................................................................ 101
FUTURE RESEARCH ........................................................................................ 104
SUMMARY ......................................................................................................... 107
REFERENCES ............................................................................................................ 109
xii
xiii
xiv
List of Figures
Figure 1.
Relation between received signal strength and distance. ............................ 32
Figure 2.
Intersection of three circles ......................................................................... 33
Figure 3.
Location solution derived from angle of arrival measurements ................. 34
Figure 4.
Correlation of DSSS signal ......................................................................... 39
Figure 5.
Approximate range resolution versus signal bandwidth ............................. 40
Figure 6.
Estimated TOA of the DLOS path and normalized time domain
responses obtained using three different techniques. ................................ 41
Figure 7.
Signal propagation delay estimation in one-way ranging ........................... 43
Figure 8.
Two-way ranging ....................................................................................... 45
Figure 9.
Symmetric double sided two-way ranging ................................................ 48
Figure 10.
Intersection of two hyperbolas................................................................... 49
Figure 11.
Cross correlation method for TDOA estimation ....................................... 49
Figure 12.
Trilateration by using three measurements ................................................ 51
Figure 13.
Hyperbolic positioning .............................................................................. 57
Figure 14.
AOA Positioning technique ....................................................................... 61
Figure 15.
The geometry of AOA location method .................................................... 62
Figure 16.
Example of location fingerprints estimated ............................................... 65
Figure 17.
Two-dimensional acquisition search space ............................................... 75
Figure 18.
Aeroscout system components .................................................................. 83
Figure 19.
Ekahau system components ....................................................................... 85
Figure 20.
FCC definition of UWB............................................................................. 86
Figure 21.
Multipath effect of narrowband signal and ultra-wideband signal ............ 88
Figure 22.
FCC Emission masks for indoor and handheld devices ............................ 90
Figure 23.
ECC draft spectrum mask .......................................................................... 91
Figure 24.
Illustration of the IEEE802.15.4a ranging protocols ................................. 94
Figure 25.
Comparison of positioning technologies ................................................. 100
xv
xvi
List of Tables
Table 1.
Relationship between RMSE, 2 drms and CERP ....................................... 16
Table 2.
Summary of positioning technologies....................................................... 101
Table 3.
Merits and drawbacks of the positioning technologies ............................. 102
xvii
xviii
List of Abbreviations
AGPS
Assisted Global Positioning System
AOA
Angle of Arrival
API
Application Programming Interface
BER
Bit Error Rate
BS
Base Station
CDMA
Code Division Multiple Access
CEPT
European Conference
Administrations
CERP
Circular Error Probability
COO
Cell of Origin
CRLB
Cramer-Rao Lower Bound
CSMA-CA
Carrier Sense Multiple Access with Collision Avoidance
CSS
Chirp Spread Spectrum
C/A
Coarse/Acquisition
DAA
Detect and Avoid
DC
Direct Current
DDP
Dominant Direct Path
DoC
Department of Commerce
DoD
Department of Defense
DOP
Dilution of Precision
DR
Dead Reckoning
DRMS
Distance Root Mean Square
DSP
Digital Signal Processor
DSSS
Direct Sequence Spread Spectrum
EC
European Commission
of
Postal
and
Telecommunications
xix
ECEF
Earth Centered Earth Fixed
EGNOS
European Geostationary Navigation Overlay System
EIRP
Equivalent Isotropically Radiated Power
ETSI
European Telecommunications Standards Institute
EU
European Union
E112
Enhanced 112 emergency call requirements
E911
Enhanced 911 emergency call requirements
E-OTD
Enhanced Observed Time Difference
FCC
Federal Communications Commission
FHSS
Frequency Hopping Spread Spectrum
GDOP
Geometric Dilution of Precision
GNSS
Global Navigation Satellite System
GPS
Global Positioning System
HDOP
Horizontal Dilution of Precision
HSGPS
High Sensitivity GPS
HTTP
Hypertext Transfer Protocol
HW
Hardware
IEEE
The Institute of Electrical and Electronics Engineers
IMU
Inertial Measurement Unit
INS
Inertial Navigation System
ISM
Industrial, Scientific and Medical radio frequency bands
LAN
Local Area Network
LBS
Location-Based Services
LDC
Low Duty Cycle
LPS
Local Positioning System
LOS
Line of Sight
LR-WPAN
Low Rate Wireless Personal Area Network
LS
Least Squares
MEMS
Micro Electro Mechanical System
xx
MIT
Massachusetts Institute of Technology
MS
Mobile Station
MUSIC
Multiple Signal Classification
NAVSTAR
Navigation System by Timing and Ranging
NDDP
Non-Dominant Direct Path
NLOS
Non Line Of Sight
NNSS
Nearest Neighbor in Signal Space
OTD
Observed Time Difference
PAN
Personal Area Network
PDA
Personal Digital Assistant
PDOP
Position Dilution of Precision
PPS
Precise Positioning Service
ppm
parts per million
PRN
Pseudorandom Noise
P(Y)
Precision(Encrypted)
RAIM
Receiver Autonomous Integrity Monitoring
RAKE
Radio receiver designed to counter the effects of multipath fading
RDF
Radio Direction Finding
RF
Radio Frequency
RFID
Radio Frequency Identification
RMSE
Root Mean Square of Errors
RSS
Received Signal Strength
RSSI
Received Signal Strength Indicator
RTK
Real Time Kinematics
RTLS
Real-Time Location System
RTT
Round Trip Time
RX
Radio Receiver
SDK
Software Development Toolkit
SNR
Signal-to-Noise Ratio
xxi
SPS
Standard Positioning Service
SW
Software
TCP/IP
Transmission Control Protocol/Internet Protocol
TDOA
Time Difference of Arrival
TDOP
Time Dilution of Precision
TG
Task Group
TOA
Time of Arrival
TW-TOA
Two-way TOA
UDP
Undetected Direct Path
UERE
User Equivalent Range Error
US
United States
UWB
Ultra-wideband
VDOP
Vertical Dilution of Precision
VHF
Very High Frequency
VOR
VHF Omnidirectional Range
WAAS
Wide Area Augmentation System
WGS
World Geodetic System
WLAN
Wireless Local Area Network
WPAN
Wireless Personal Area Network
XML
Extensible Markup Language
2D
Two Dimensional
3D
Three Dimensional
xxii
List of Symbols
ε meas
Vector of pseudorange measurement errors
εx
Vector of position error components
T
Oscillator time offset

Phase of the signal

Wavelength
θ
Angle of arrival
ρ
Pseudorange
̂
Approximate (predicted) pseudorange
σ
Standard deviation
a
Standard deviation of the position
r
Standard deviation of the range

Time of arrival

Phase

Phase difference

Difference between the approximate (predicted) pseudorange and the
true pseudorange
Δρ
Vector of  i ’s
r
Vector of differences between the measured TDOA and predicted
TDOA
t
Time difference
Δx
Vector offset of the user’s true position and time bias from the values
at the linearization point
c
Speed of light
d
Distance
xxiii
fc
Carrier frequency
F
Design matrix for AOA method
G
Design matrix for TDOA method
H
Design matrix for TOA method
n
measurement error of angle of arrival
P0
Empirical coefficient
Pr
Received signal strength
r
Range
r̂
Approximate (predicted) range
r
Residual vector
R
Vector of RSS samples measured at online phase in LF method
~
R
Vector of RSS samples measured at offline phase in LF method
RSE
Square of residual vector length
t
Time
tu
Time offset between the mobile terminal clock and the system time
T
Temperature
Treply
Reply time of the signal
Tround
Round trip time of the signal
Tt
Time of flight of the signal
vus
Speed of sound
xmeasured
Measured location
xtrue
True location
xi , y i , z i
Location of the ith base station
xu , y u , z u
Location of the user or mobile station
xˆ u , yˆ u , zˆu
Approximate (predicted) location of the user
1
1 Introduction
Due to the increasing number of mobile computing devices and wireless local-area
networks, a growing interest in location-aware systems and services has emerged.
Commercial location-based services, such as map-based guidance and navigation,
intelligent transport services and tracking of people and valuable assets, have a huge
market potential. The recent company acquisitions in the telecommunications and
navigation industry especially emphasize this market potential (Nokia, 2007).
In order to serve their users well, emerging mobile computing applications will need to
know the physical location of people and assets. Today the most common technology
to provide the physical location information is Global Positioning System (GPS). In
the future there may also be other satellite navigation systems, for example European
Galileo. However, there is also a growing interest in positioning applications which are
independent from global satellite navigation systems. These positioning systems,
which are usually designed to operate inside a limited geographical area, are called
local positioning systems.
The most important reason to use local positioning systems is the relatively poor
performance of GPS inside the buildings and urban canyons. Despite the development
of high-sensitivity GPS receivers and assisted GPS (AGPS) technologies, the
performance of GPS is not sufficient for many indoor applications. To obtain location
information also inside buildings, several indoor positioning systems have been
developed.
The need for the local positioning systems is not only restricted to the indoor
environment. Local positioning systems are used in applications where people or
objects are tracked both indoor and outdoor environments. For example, in Legoland
Billund the parents can track the movements of their children in the amusement park
area by using the local positioning system based on wireless local area network
(WLAN) infrastructure (InformationWeek, 2004).
The cost of the positioning technology is also an important factor. Sometimes local
positioning systems are used because the sensors to be tracked in the local positioning
system are less expensive than the GPS receivers. For some organizations,
2
independence from global satellite navigation systems may be a good reason to use
local positioning systems.
In addition to the global satellite navigation systems, there are many different local
positioning and indoor navigation technologies. None of these systems has optimal
performance in all circumstances where location-aware services are needed. Different
positioning technologies have to be used with different applications. However, there is
a need for a location-based service, which would provide accurate position data both
indoors and outdoors. Which local positioning technologies will be integrated with
global positioning technologies, remains to be seen in the future.
1.1
Background
The first indoor location systems were developed some 20 years ago. The first systems
were intended for research purposes in laboratories and they used pattern matching,
ultrasound propagation or magnetic field as a navigation observable. Early indoor
location systems were either too expensive or too sensitive to electromagnetic
interferences, and thus not suitable for industrial or consumer markets.
Since the late 1990’s less expensive and more robust indoor navigation systems have
appeared in the market. Most of these systems utilize wireless LAN or other local
radio network. In WLAN-based systems, the position of a mobile terminal is most
often calculated from the received signal strength (Ekahau, [Cited January 24, 2008]).
However, received signal strength (RSS) indicator does not provide a good estimate
for the distance, but the RSS may give a reasonable accurate position fix if it is
correlated to a map of previously measured signal strengths. This type of positioning
method is called location fingerprinting. The problem of the location fingerprinting
technique is that it requires extensive calibration work before the location service can
be used.
The other type of WLAN-based system measures the time difference of arrival
(TDOA) of a mobile WLAN transmitter (Aeroscout, [Cited January 24, 2008]). In
these systems additional hardware and signal processing is needed, since the standard
WLAN access points are not able to perform the TDOA measurements.
The two major error sources in TDOA estimation at the indoor environment are
multipath fading and no line of sight (NLOS) condition due to shadow fading
(Pahlavan 2002). Because of the multipath problem present in indoor environment,
traditional cross correlation techniques of wide-band direct sequence spread spectrum
3
(DSSS) signal do not perform well (Pahlavan & al., 2002). Performance of TOA
estimation of wide-band signal can be improved by employing super-resolution TOA
estimation techniques or increasing the signal bandwidth.
In ultra-wideband (UWB) technology extreme short duration pulses (sub-nanosecond)
are used instead of continuous waves to transmit information. The short pulse
generates directly a very wide bandwidth signal. One important advantage in indoor
navigation and communication is that UWB is relative immune to multipath fading
(Sahinoglu & Gezisi, 2006). The problem in the ultra-wideband technology is the
limited range of the signal.
1.2
Research Objectives and Contribution
The objective of this research is to identify the most promising current and emerging
local positioning techniques. The research also aims to match the characteristics of the
identified local positioning techniques to the needs of various location-based
applications. These needs depend considerably on the physical environment where the
local positioning systems are used.
This licentiate thesis is a literature research. The aim of the research is to make a
comprehensive survey on the field of local positioning systems. A classification of the
local positioning systems is presented to find a match between the application needs
and properties of the local positioning systems. The principles of the localization are
discussed and various positioning technologies are presented. Also the most common
location estimation algorithms are presented in such level of details that the algorithms
can be easily converted to a computer program.
1.3
Scope of the Research
The overall purpose of the research is outlined in its title, but a closer look into the
scope is discussed in this section. This research discusses local positioning systems and
indoor navigation.
In this thesis the local positioning system refers to the positioning services, which are
available only within a limited geographical area. Global positioning systems and
services, such as NAVSTAR GPS, are out of scope in this research, with the exception
of the indoor use of GPS signals. Also the positioning services based on the cellular
network signals are considered as ―global‖ positioning services, and thus they are out
of the scope in this research.
4
The other part of the title is indoor navigation. In this thesis, indoor navigation is
considered as a special case of the local positioning. Several positioning techniques
developed for the indoor applications are discussed. In spite of that the NAVSTAR
GPS is not a local positioning system, the assisted GPS technology and highsensitivity GPS (HSGPS) are explained in this thesis. The indoor use of the
NAVSTAR GPS can be considered as a competing technology to the local positioning
systems intended for indoor environment.
A wide range of positioning methods such as dead-reckoning, ultrasound-based
systems and use of optical sensors, are discussed. However, the most commonly used
local position services are based on the use of radio frequency signals. The other
sensors will provide only supplementary information for the positioning service. Thus,
the main emphasis of this research is in the positioning observables and algorithms of
the radio frequency devices.
Only the most commonly used positioning algorithms are explained in detail in this
thesis. In the systems based on trilateration, the position estimation algorithm based on
Taylor series expansion is most widely used in real applications. In the literature, there
are also many other algorithms presented, whose performance over the Taylor series
method is often questionable or whose implementation is relatively complex. These
algorithms are discussed only shortly, and references to them are given to the reader.
1.4
Main Results
No single technology was found which would have provided optimal performance for
all possible applications and environments. The technology based on location
fingerprinting method was found the most promising technology in heavily multipath
environment, such as office buildings. On the other hand, the technology based on
measuring the WLAN signal propagation time performs better in large open areas,
such as sport halls. Neither of these WLAN-based technologies is feasible, if better
than 1–2 meters accuracy is required. When accuracy is important, utilizing UWB
signals for positioning will be an interesting alternative. However, opposite to WLAN,
UWB is not a mature technology yet and in Europe the legislation processes of UWB
signals are still going on. When only a building level accuracy is sufficient indoors,
use of the assisted GPS can be recommended.
Several positioning technologies based on other positioning signals than radio
frequency were also identified. For example, the location of an object can be
determined with few centimeters accuracy by using computer vision system and video
5
image analysis. Also dead-reckoning method can be considered in applications where
the calibration to the local frame of reference is not a problem. However, the RF-based
local positioning systems are more popular because of the excellent coverage of the
radio signal.
Fusing satellite-based navigation system with local positioning systems is suggested as
a future research topic. Another interesting future research topic is better utilization of
redundant measurements in location fingerprinting systems.
1.5
Outline of the Thesis
Chapter 1 reviews the background and purpose of this work. The main objectives and
research questions are formulated. Finally, the scope and limitations of the work are
defined.
Chapter 2 defines the concept of the local positioning system. It also defines the
taxonomy of the local positioning systems and shows examples of applications
utilizing the local positioning technologies. Finally, the performance criteria for
evaluating the performance of the position systems are introduced.
Chapter 3 presents the positioning sensors and metrics used in the positioning.
Positioning sensors and systems based on infrared, ultrasound, DC electromagnetic,
optical, gyroscope and radio frequency signals are explained.
Chapter 4 presents the observables used in positioning systems based on the use of the
radio frequency signals. The following observables are explained: received signal
strength (RSS), angle of arrival (AOA), time of arrival (TOA), and time difference of
arrival (TDOA). Also proximity-based methods and location fingerprinting methods
are explained.
Chapter 5 reviews the position estimation algorithms, which compute the location of a
mobile terminal from the observables presented in chapter 4.
Chapter 6 illustrates the use of the indoor GPS as an alternative to the local positioning
systems.
Chapter 7 presents the most promising current and emerging local positioning systems.
Examples of local positioning systems based on ultrasound, WLAN and ultrawideband signals are given.
6
Chapter 8 contains the discussion and conclusions.
Chapter 9 summarizes the results.
7
2 Principles of Local Positioning Systems
2.1
Introduction
This chapter provides an overview of the applications which utilize the local position
systems. A user’s location will become common information in the near future. GPS
chips and other navigation devices have become smaller and less expensive. They also
consume substantially less energy than 10 years ago. Furthermore, the WLAN-based
positioning systems requiring no extra components at the mobile device have been
developed during the recent years. This development in the navigation systems makes
it possible to equip even smaller and less expensive devices with the knowledge of
their location. This opens completely new application areas for the location-based
systems.
Traditionally, positioning systems have been used in sea and air navigation and
surveying. Consumer level navigation devices became common after the mid-1990’s,
when GPS became operational. Markets of the handheld and car installed GPS devices
have been growing since that. However, in many applications the positioning
information is not interesting if this information cannot be communicated to other
users or devices. Until the early 2000’s the GPS receivers have been too expensive and
too power consuming to be integrated in to a mobile phone. The first high volume
mobile phones containing an integrated GPS receiver came to European markets only
in 2007. Integrating positioning technology with the communications technology
provides many new applications for the consumer and industrial markets.
Even though modern GPS receivers are much more sensitive than GPS receivers in the
1990’s and even though they are able to locate indoors, the accuracy and reliability of
GPS is not sufficient for many indoor applications. Local positioning systems are
needed for applications that are used in environment where the GPS signal is weak.
2.2
Definition of Local Positioning System
There is no commonly accepted definition for the local positioning systems. In this
thesis the local positioning system refers to the localization services, which are
available only within a limited geographical area. In the literature, the term local
8
positioning system usually refers only to indoor navigation. However, there is a need
to deploy proprietary positioning services for small geographical areas, which may be
located either indoors or outdoors – or both.
The local positioning systems have the following properties:

Short-range signals are used and thus the positioning service is available only
within a relatively small geographical area

The positioning infrastructure is often maintained locally by an organization
that operates at the same area where the local positioning service is available.

The positioning service is often independent on the global positioning services
or positioning services provided by telecommunications operators

Locally defined coordinate frame of reference is often used
A local positioning system can be considered as contrary to a global positioning
system. Here, a global positioning system is considered to be a globally available
service, which is operated by governments or by large organizations such as
telecommunications operators. NAVSTAR Global Positioning System, which is
operated by the Department of Defense (DoD) of the United States, is the most well
known example of global positioning systems. Other examples of global or ―semiglobal‖ navigation systems are LORAN-C and VHF Omni-directional Radio Range
system (VOR). The positioning services dependent on the cellular network signals may
be considered semi-global positioning services as well.
2.3
Need for the Local Positioning Systems
Today, the most common reason to deploy a local positioning service is the poor
performance of NAVSTAR GPS at inside buildings and urban canyons. Most of the
local positioning systems used are either indoor navigation systems or systems based
on dead-reckoning. However, the need for the local positioning systems is not
restricted to the indoor environment alone. There is a need for positioning services,
which are able to provide accurate positioning information both indoor and outdoor
areas. The high-sensitive GPS receivers provide high accuracy position data outdoors,
but the positioning accuracy for certain indoor location-aware applications is relative
poor. Local positioning system based on WLAN signals may provide a better
performance in the environment that consists of indoor and outdoor areas.
In wireless sensor network applications the mobile nodes must be extremely low-cost
and energy-efficient. Today, the prices of GPS modules intended for mobile devices
9
are relative high, varying from 5 € to 30 €, depending on the volumes purchased. The
GPS receivers are also too power-hungry for the nodes of sensor network. The local
positioning techniques developed can provide a cost-efficient solution for indoor and
outdoor wireless sensor network applications. For some organizations, independence
from global satellite navigation systems may be a good reason to use local positioning
systems.
2.4
Taxonomy of Positioning Systems
There are many different positioning technologies developed to locate people and
valuable assets. Each positioning technology solves a slightly different problem or
supports a different kind of location-aware application. These positioning technologies
vary in many parameters, such as the physical phenomena used for location
determination, the size of the sensing device, energy consumption, and resolution in
time and space.
Different kinds of taxonomies have been developed for positioning systems
(Hightower & Borriello, 2001; Zeimpekis & al., 2003; Muthukrishnan & al., 2005;
Kolodziej & Hjelm, 2006). According to Hightower & Borriello (2001), the purpose of
these taxonomies is to help developers of location-aware applications evaluate their
options better when choosing a positioning system. The taxonomy may also help
researchers in identifying opportunities for new positioning techniques.
In this research, the following taxonomy is presented:

Classification based on type of the location information. A positioning system
may provide physical or symbolic location information. On the other hand, the
physical coordinates may be given in absolute (global) or relative frame of
reference.

Classification based on where location estimation takes place. The location of
the mobile terminal may be estimated either in the mobile terminal itself or in
the network server associated with positioning system infrastructure.

Classification based on scale of the positioning system. The coverage of a
positioning system may be worldwide or it may be able to locate objects within
a limited geographical area only.

Classification based on signal technologies. Most of the positioning systems
are based on radio signal. However, also magnetic compass, gyroscope, video
camera picture or ultrasound pulses can be employed.
10

Classification based on observables. The position of the object can be derived
from time of arrival (TOA), time difference of arrival (TDOA), or angle of
arrival (AOA) observables.

Classification based on location estimation techniques. Triangulation and
trilateration methods compute the position of the target object by using ranging
measurements. There are also range free methods, such as proximity-based
methods and location fingerprinting.
2.4.1 Classification Based on Type of the Location Information
Hightower & Borriello (2001) classify the positioning systems in two categories
according to the type of the location information. The positioning system may provide
either physical coordinates or symbolic location of the target device. The location
information in physical coordinates can be classified further in two categories:
absolute and relative.
The physical coordinates, having numerical values, may be given in absolute or
relative frame of reference. In location-aware applications Cartesian or polar
coordinate system is usually used. In contrast to the physical coordinates, the symbolic
location tells to the user, in which place the target object is. The symbolic location may
refer to a particular room or building or it may tell that the target object is approaching
to the point of interest. The physical location information can be usually transformed
to symbolic location information by using a database containing interesting places and
their physical coordinates (Hightower & Borriello, 2001).
The most well known positioning systems, like GPS and LORAN-C, are clearly
physical-coordinate positioning technologies. On the other hand, many technologies
provide inherently symbolic location information. Bar code scanners and access
control terminals provide symbolic location information based on proximity to known
objects (Hightower & Borriello, 2001).
The physical-positioning system must be accurate enough, so that it can place the
target object to the correct symbolic location. For example, positioning system having
10 meter accuracy is not very useful, if the symbolic locations are office rooms having
size of 3 x 3 meters. On the other hand, purely symbolic location systems typically
provide only very coarse-grained physical positions (Hightower & Borriello, 2001).
They may tell that the target object has passed a certain access control terminal, and
thus it must be in a certain room in the building. However, it is impossible to tell in
which part of the room exactly the target object is.
11
The physical-positioning systems can be further classified in two categories: the
systems providing absolute and relative coordinates (Muthukrihnan & al., 2005,
Hightower & Borriello, 2001). An absolute positioning system uses a shared frame of
reference for all located objects, while relative positioning system has its own frame of
reference. GPS and LORAN-C are examples of absolute positioning systems. GPS
provides location data in WGS-84 frame of reference. A relative location can be
converted into an absolute position if the translation, rotation and scaling matrices
between the reference frames are known.
2.4.2 Classification Based on where Position Estimation Takes Place
The position of the user may be calculated either at the user’s terminal or at a remote
computer. Systems where the position calculation of the device takes place in the
device itself are called self positioning systems, while systems where the position of
the object is calculated at a remote site are called remote positioning systems
(Zeimpekis & al., 2003). Often the remote positioning systems are referred as networkcentric or network-based systems.
In the self positioning systems, the mobile device being located calculates its own
position. The base stations can be either terrestrial or satellites, and the locations of the
base stations are known. The self positioning model ensures privacy by mandating that
no one else may know where the mobile device is, unless the object specifically takes
action to publish its location (Hightower & Borriello, 2001). Also the raw
measurements, from which the location is computed, are not known by other entities in
the system. The mobile terminal receives the signals transmitted from the base stations
and calculates its own position by using this data. GPS is an example of the self
positioning systems.
In the network-based system, the mobile terminal can be located by measuring signals
traveling to and from a set of base stations. The position estimation takes place at a
remote site, which is often a network server. If the mobile terminal transmits the
navigation signal and the signal is received at the base station, the system is inherently
network centric. Also the pure symbolic positioning systems, which employ devices
such as access control terminals or radio frequency identification tags, are obviously
network centric. Indoor positioning systems utilizing the WLAN signal are typically
network-based.
12
2.4.3 Classification Based on Scale of the Positioning System
One of the classification argument presented by Hightower & Borriello (2001) is the
scale of the location system. The scale of the positioning system can be associated to
the size of the geographical area the system serves. Moreover, the scale of the
positioning system is associated to the number of the objects the system is able to
locate with a certain amount of infrastructure or over a given time.
A global system like GPS is able to locate objects worldwide, while the service of a
local positioning system is limited within a certain area only. Local positioning
systems may be able to locate objects within a certain part of a city, in a particular
building, or within a single room. Usually the number of objects located is not limited
in the self positioning systems, like GPS and Cricket indoor positioning system
(Priyantha, 2005). On the other hand, in network centric systems, the number of
mobile terminals is limited due to the restricted server capacity and bandwidth
available in communication channel. Because of these limitations, the base station can
receive only a certain amount of positioning signals per time interval.
2.4.4 Classification Based on Signal Technologies
Muthukrishnan & al. (2005) identifies the following signal technologies used in
positioning systems:

Radio frequency. Radio navigation uses radio waves to determine the user’s
position. Using radio signals has been popular in navigation, because of radio
wave’s ability to travel long distances. Most of the indoor positioning systems
are based on radio signals as well.

Infrared. Infrared positioning systems are based on proximity, not ranging. The
location of the mobile device is assumed to be the same as the location of the
base station emitting the infrared signal.

Ultrasound. Ultrasound technology provides a great ranging accuracy with low
cost electronics. High probability of interference from other ultrasound sources
reduces the reliability of the ultrasound-based systems.

Optical. The location information can be derived from analysis of video
images.

Inertial. Inertial navigation systems collect information from gyroscopes and
accelerometers to determine the position and orientation of the device.
13

DC Electromagnetic. The direct current (DC) electromagnetic systems generate
axial DC magnetic-field pulses from a transmitting antenna in a fixed location.
The system measures the position and orientation of one or more receiving
antenna sensors with respect to transmitting antenna.
These signal technologies are explained in more detail in Chapter 3.
2.4.5 Classification Based on Observables
The positioning systems can be classified according to observables used in position
determination. The most common observables are:

Time of arrival (TOA). A range between the transmitter and receiver is
estimated by measuring the signal propagation delay between these two
devices.

Time difference of arrival (TDOA). Difference of arrival times of two signals is
measured. The TDOA measurement and the known coordinates of the two base
stations define a hyperbola. The position of the object is obtained from the
intersection of multiple hyperbolas.

Angle of arrival (AOA). Direction angle of the signal is measured. Each
estimated angle defines a line between the base station and a mobile device.
The location of the object is obtained from the intersection of these lines.

Received signal strength indicator (RSSI). The intensity of a transmitted signal
decreases as the distance from the transmitter increases. The range between the
transmitter and receiver can be estimated from the received signal strength.
Muthukrishnan & al. (2005) use the term ranging technology in context of the
observables mentioned above. However, the term ranging technology is somewhat
misleading, since the angle of arrival method provides an angle measurement instead
of a range measurement. On the other hand, positioning systems based on RSSI do not
usually perform geometric ranging. Instead, the position estimate is obtained by
location fingerprinting algorithm, which is explained in the next section.
2.4.6 Classification Based on Location Estimation Techniques
Hightower & Borriello (2001) divide the positioning systems into three main
categories: triangulation-based systems, proximity-based systems, and scene analysis
systems. In addition to these techniques, location fingerprinting and dead-reckoning
can be considered as examples of location estimation techniques too.
14
Muthukrishnan & al. (2005) subcategorize the triangulation into lateration and
angulation. Trilateration method computes the position of an object by measuring its
distance from multiple reference positions. In three dimensional positioning, three
distance measurements are needed to solve the position of the object. The position of
the object is obtained from the intersection of spheres whose centers are at the
reference positions and whose radius is the measured distance.
In triangulation method, angles of arrived signals are used for determining the position
of an object. Each measured angle defines a line between the base station and the
target object. The location of the object is obtained from the intersection of these lines.
Proximity measures the nearness to a known point or set of points. Knowing only
whether or not two devices are in communication range is enough to give a position
estimate. In this method the location of the target object is considered to be the
location of the base station. If a signal is received at several known locations, it is
possible to intersect the coverage areas of that signal to determine a ―containing‖
location area (Kolodziej & Hjelm, 2006, p. 145). Another method is to choose the
location of the base station which has the strongest signal as the target object location.
Scene analysis examines a view from a particular vantage point to draw conclusions
about the observer’s location. Video cameras can be used to describe spatial
relationship in scenes using image processing techniques, and thereby determine the
position (Hightower & Borriello, 2001).
In location fingerprinting method, the position estimate is obtained by comparing the
measured RSS values to the database containing previously measured RSS values at
each place in the positioning area. In the dead-reckoning (DR) method, the current
position estimate of the target object is based upon a previously determined position
estimate. This position estimate is then advanced by using the known speed, elapsed
time, and course.
Location Estimation techniques are discussed in more detail in Chapter 5.
2.5
Accuracy Metrics for Positioning Systems
Generally, the most important performance metric for the positioning systems is the
accuracy of the estimated position.
15
Accuracy is usually reported as the error distance between the estimated location and
the actual location, while a location precision is reported in percentages of position
information, which is within the distance of accuracy. A self explanatory definition for
accuracy is defined in (CDG, 2000), as: ―Accuracy of the geolocation technology is a
measure that defines how close the location measurements are to the actual location of
the mobile station to be located‖.
2.5.1 Root of Mean Square Error
The root of mean square error (RMSE) is a widely used measure of the difference
between the true location and multiple of measured locations. RMSE is calculated
from square root of the mean value of the squared error:
N
RMSE 
 (x
k 1
measured
(k )  xtrue ) 2
(2.1)
N
where N is the number of measurements in the set and k is the index of the
measurement. RMSE equals the standard deviation of the measurements.
The accuracy metric can be defined for various dimensions. Three-dimensional
accuracy is evaluated by taking into account all position dimensions. The result is
independent from coordinate system. Two-dimensional accuracy refers to horizontal
accuracy. One-dimensional accuracy refers either to vertical accuracy measuring
deviations in altitude or to radial accuracy measuring deviations in distance from the
true location to the measured location (Syrjärinne, 2001, p. 8).
2.5.2 Distance Root Mean Square Error
The distance root mean square error (drms) in two dimensions is defined by formula
(Kaplan, 1996, p. 280):
drms   x2   y2
(2.2)
where x and y are the standard deviation (RMSE) of the error along the x and y axes.
The probability that the computed location is within a circle of radius drms from the
true location is about 0.63 if the error distribution in both axis is about the same.
Probability that the error is within a circle of radius two times drms (2drms) varies
16
between 95% and 98% depending on the difference of error distributions on the
coordinate axis. The 2drms value is commonly taken as the 95% limit for the
magnitude of the horizontal error (Kaplan, 1996, p. 280).
2.5.3 Circular Error Probability
A second accuracy measure in common use is the circular error probable or circular
error probability (CERP). The CERP is defined as the radius of a circle that contains at
least 50% of the measurements. Thus, probability that the magnitude of the error is less
than the CERP is precisely ½. In addition to 50%, CERP is sometimes defined also to
other probabilities. Kaplan (1996, p. 281) gives a following approximation for
relationship between the CERP and drms:
CERP50  0.75 drms
CERP80  1.28 drms
CERP95  2.0 drms
Table 1 summarizes the relationship between the RMSE, 2 drms and CERP.
Table 1.
Relationship between RMSE, 2 drms and CERP
Accuracy Measure
Probability (%)
RMSE, drms (root mean square error)
63 to 68
CERP50 (circular error probability 50%)
50
CERP95 (circular error probability 95%)
95
2 drms
95 to 98
2.5.4 The Effect of Geometry and Range Error Factors
The accuracy of the location estimate depends on the accuracy of the individual
measurements and the mutual geometry of the mobile object and base stations. If the
measurement observable is range, the accuracy of the estimated location depends on
the accuracy of the individual range measurements and the mutual geometry of the
mobile object and base stations.
In the positioning systems based on range observable the accuracy of the position
solution is ultimately expressed as the product of a geometry factor and a range error
factor. Error in the position solution can be approximated by the formula (Kaplan,
1996, p. 238):
17
(error in position solution) = (geometry factor) * (range error factor)
User Equivalent Range Error
In GPS the effective accuracy of the measured range value is termed the user
equivalent range error (UERE) (Kaplan, 1996, p. 238). The UERE for a satellite is
considered to be a statistical sum of various error sources associated with the satellite.
Usually the error components of total UERE are considered independent with
Gaussian distribution. UERE is usually assumed to be independent and identically
distributed from satellite to satellite. However, it is also possible to model UERE of
certain satellites (e.g. EGNOS/WAAS satellites) with a different variance.
In GPS the total range measurement errors consist of the following factors (Kaplan,
1996, p. 239):

delays due to the atmosphere (very small in local positioning systems)

receiver noise and resolution offset

multipath offset

receiver hardware offsets

errors in base station location

errors in system time reference
The error components are root-sum-squared (rss) to form the total system UERE. For a
GPS C/A code the UERE is 8.0 m when Selective Availability is off (1) (Kaplan,
1996, p. 261). In reality some of the error components, like multipath offset, are not
independent nor Gaussian distributed.
In spite of the concept of UERE is used mostly in GPS literature, the UERE concept
can be applied to other positioning techniques using the range observable as well.
Dilution of Precision
Dilution of precision is a GPS term which describes the geometric strength of satellite
configuration on GPS accuracy. When visible satellites observed from the user
position are close together in the sky, the geometry is said to be weak and the DOP
value is high. When the satellites are far apart, the geometry is strong and the DOP
value is low. The term DOP can also be applied to other positioning systems based on
range measurements and to the systems using angle-of-arrival (AOA) measurements
(Dempster, 2006).
18
The geometric dilution of position (GDOP) is defined as follows (Kaplan 1996, p.
267):

GDOP  trace (H T H) 1

(2.3)
where H is the direction cosine matrix containing unit vectors pointing from the
linearization point to the location of the ith base station. Direction cosine matrix H is
the same matrix which is used in TOA position calculation explained in Chapter 5.1.
If H is constructed using a local coordinate system (East, North, Up), then similar
DOP factors can be defined for 3D position, horizontal position, vertical position and
time. Position dilution of precision (PDOP) is defined by three first diagonal elements,
horizontal dilution of precision (HDOP) by the first two, vertical dilution of precision
(VDOP) by the third, and time dilution of precision by the fourth.
The expected value of the positional error can be related to expected range error via
(Dempster, 2006):
 a  DOP r
(2.4)
where r is the standard deviation of the range and a is the standard deviation of the
position. This relationship can be expressed also by using terms RMSE and UERE:
RMSE  DOP  UERE
(2.5)
For example, if the expected value of the range error (UERE) is 8 meters and PDOP is
3, the expected position error is 24 meters (drms). Two-sigma value for the position
error would be 2 * PDOP * UERE = 48 meters.
2.6
Other Performance Metrics
In addition to the accuracy of the positioning system Syrjärinne (2001) defines four
more measurements of performance: reliability, availability, latency and applicability.
In addition to these, Muthukrishnan & al. (2005) have found other performance
metrics, such as scalability and privacy. However, the last two metrics are rather
properties of positioning systems than performance metrics.
19
Reliability is as important performance metric as the accuracy in local positioning
systems. Reliability measures the probability of exceptional phenomena, which would
totally destroy the location measuring process. Reliability should be automatically
taken into account when accuracy of the positioning technology is determined
(Syrjärinne, 2001, p. 10).
Syrjärinne (2001, p. 11) considers availability as a performance metric. Positioning
system availability describes whether or how often a system is available for
positioning by its intended users. Usually the concept availability also includes the
concepts of coverage and capacity (Syrjärinne, 2001, p. 12). Availability measures
different aspects of positioning than accuracy and reliability. Syrjärinne gives an
example: If a GPS receiver is taken deep underground, the GPS signals cannot be
tracked and thus position determination becomes impossible. The fact that signals are
blocked does not imply that GPS was not very good in terms of accuracy and
reliability. The lack of GPS tracking implies that GPS signals are not available for the
user in underground facilities. The concept of performance metric availability relates
closely to the taxonomy property scale discussed earlier in this chapter.
Responsiveness (Muthukrishnan & al., 2005) describes how quickly the location
system outputs the location information. Syrjärinne (2001) uses term latency for the
same purpose.
Applicability measures the physical limitations and requirements associated with the
implementation of a certain technology in terms of technical and financial issues
(Syrjärinne, 2001, p. 15). The most important metrics are the cost, power consumption,
hardware size and network dependency.
2.7
Privacy of Location Information
Location-based services have a wide field of promising applications. Unfortunately,
the emerging location-based services raise a lot of privacy issues due to their ability to
collect, store, use, and disclose personal information. These privacy issues represent
new challenges to the law makers, location service providers and the users of these
systems. To maintain trust and confidence among the different participants, these
privacy issues and the associated threats must be addressed (Gadzheva, 2007). If
people do not have confidence about their privacy, the widespread deployment of the
location-based systems will be hindered.
20
Location information of a person is very sensitive and private data. When used
together with other information of a person, location information identifies the person
and allows his or her whereabouts to be tracked anytime and anywhere. Continuously
tracking of a person’s location, storing it centrally and transmitting it to other users
potentially compromise an individual’s location privacy (Gadzheva, 2007). To make
the users feel confident in using location-based services, they should be informed
about the possibilities of the system and be assured that their privacy is well protected.
The user should not only be conscious of the data collection, but also have an
opportunity to explicitly give permission to the collection of location data (Baumann &
al., 2002).
The distinction between self positioning and network-based positioning is important
from the privacy point of view. The self positioning systems, such as GPS, are less
risky in terms of privacy, since the measuring of the ranging observables and the
position computation are performed by the user’s terminal equipment. This means that
the location data are originally in the control of the user and thus not disclosed to
location service providers or operators unless the user decides so (Simojoki, 2003).
Most of the local positioning systems are network-based systems. In network-based
systems the computation of the location information takes place at a network server
maintained by the service provider. For example, the WLAN-based positioning
systems on the market are network-based. In addition, many proximity-based systems
are also network-based as well as all cellular network-based positioning techniques. In
these systems, location data is generated by the service provider and the loss of privacy
is higher. Gadzheva (2007) sees the growing use of services based on radio frequency
identification (RFID) technology as a potential risk for the location privacy. When
combining the RFID techniques with the proximity-based positioning techniques,
people can be tracked via their objects or clothing, and a wide array of information can
be gathered. Shop owners can create profiles of their customers by tracking their
movements in the shops by using RFID.
The network-based approach provides many advantages from the technology point of
view. Usually the cost of components and the power consumption are lower, if the
mobile device does not contain the hardware needed to detect and measure the ranging
signal. Also the position computation requires some processing power and memory,
which also raises the cost and adds the power consumption.
Many local positioning systems used today are intended to track things rather than
people, and the concerns related to privacy in these systems are naturally obsolete.
21
However, when position data of the people is tracked, consent of the user for
positioning is always required. The service provider must also be able to guarantee that
the data is well protected and not used for any other purposes than the user has
accepted. The markets of the indoor and outdoor location-based systems will not
develop, if people do not trust the privacy of the location information.
22
23
3 Positioning Sensors and Signals
Different types of sensors and signal technologies are used in various positioning
systems. In this chapter radio frequency, infrared, ultrasound, optical and inertial
signal technologies are presented.
3.1
Radio Frequency
Radio navigation uses radio waves to determine the user’s position. Using the radio
signals has been popular in ship and aircraft navigation, because of radio wave’s
ability to travel long distances. Most of the indoor positioning systems are based on
radio signals as well.
Radio direction finding (RDF) is the oldest method of radio navigation. A radio
direction finder works by pointing a directional antenna in various directions and then
listening to the direction in which the signal comes through most strongly. Measuring
the angle of arrival (AOA) at the base station be can also employed in local positioning
systems.
Many radio navigation systems are based on determining the distance between the
radio transmitter and the radio receiver. This distance can be derived either from the
signal propagation time between the devices, or from the received signal strength
(RSS), which depends on the distance between the transmitter and the receiver.
A GPS receiver measures the time of arrival (TOA) of the radio signal. The signal
propagation time between the transmitter and the emitter is obtained by subtracting the
known signal transmit time from the measured arrival time. Measuring the signal
propagation time is used in some local positioning systems also. Some systems like
Decca, OMEGA and LORAN-C measure the time difference of arrival (TDOA) of two
signals instead of the signal propagation time. These systems are called hyperbolic
systems, since the coordinates of the two stations transmitting these signals and the
measured time difference of arrival defines a hyperbola. The position of the device is
obtained from the intersection of two or more hyperbolas.
24
Examples of radio navigation systems are

Satellite-based systems: GPS and Galileo

Global terrestrial systems: LORAN-C, VOR

Systems based on cellular phone network signals

Systems based on short range communications signals, such as WLAN, RFID
UWB and Bluetooth
From these technologies WLAN, RFID, UWB and Bluetooth are applicable in local
positioning systems. Also the satellite navigation systems like GPS and Galileo can be
utilized in indoor navigation. Global terrestrial systems like LORAN-C are not
accurate enough for indoor navigation.
3.1.1 Modeling Radio Signal Propagation
Pahlavan & al. (2002) characterize the indoor radio propagation channel as a sitespecific and a severe multipath propagation path between the transmitter and receiver.
Also, inside buildings the availability of a direct line of sight (LOS) signal is relative
low. The two major error sources in the indoor environment are multipath fading and
no line of sight (NLOS) condition due to shadow fading.
Radio propagation channel models are developed to provide a means to analyze the
performance of a wireless receiver. However, the performance criteria for
telecommunications and positioning systems are quite different. In telecommunication
systems the most important performance criteria is the bit error rate (BER) of the
received data stream, while for the positioning systems the performance measure is the
accuracy of the estimated position. The accuracy of the estimated position depends on
the location metrics used in the positioning technology. The metrics for radio
navigation applications are angle of arrival (AOA), received signal strength (RSS) and
time of arrival (TOA). According to Pahlavan & al. (2002) there are suitable models,
which can be used to analyze the RSS and AOA for indoor positioning systems.
However, the existing wideband indoor channel models designed for the
telecommunication applications are not suitable for the analysis of the behavior of
TOA for indoor positioning systems.
3.1.2 Multipath
Multipath is a phenomenon of a radio device receiving the same signal multiple times,
slightly offset in time. Multipath phenomenon occurs when a radio signal is received
25
directly from the transmitter, and also reflected off one or more nearby objects, such as
walls, roofs or furniture inside a building. Because a reflected signal takes a longer
path, it is slightly delayed compared to the direct signal. At the receiver, the multiple
copies of the received signal interfere with each other. Multipath not only distorts the
data modulated on the carrier, but also the phase of the carrier itself.
Pahlavan & al. (1998) present three signal reception conditions, which may occur at
multipath environment

Dominant direct path (DDP) case

Non-dominant direct path (NDDP) case

Undetected direct path (UDP) case
In dominant direct path condition (DDP) the direct line of sight (DLOS) path is
detected by the measurement system and it is the strongest path in the channel profile.
In this condition the receiver can lock onto the LOS path and detect its TOA
accurately.
In non-dominant direct path (NDDP) condition, the LOS path is detected by the
measurement system but it is not the dominant path in the channel profile. Many
systems, like traditional GPS receivers, lock onto the strongest path and make an
erroneous decision on the time of arrival in NDDP condition. The amount of error
made is the time difference between the TOA of the strongest path and the TOA of the
LOS path. To avoid the wrong decision in NDDP condition, more complex receiver
architectures have been researched. One example is a RAKE type receiver, which tries
to resolve the multipath and make an intelligent decision on the TOA of the DLOS
path.
In undetected direct path condition (UDP) the system cannot detect the LOS path, and
therefore neither traditional nor RAKE type receivers can detect the LOS path
(Pahlavan & al., 1998).
3.2
Infrared-based Systems
Infrared radiation is electromagnetic radiation of a wavelength longer than visible
light. Infrared data transmission has been used for short-range communication among
computer peripherals and cellular phones. Infrared is also very popular in remote
controls, because the infrared radiation does not penetrate walls and interfere with
other infrared controlled devices in adjoining rooms.
26
Infrared has also been used for indoor location finding (Kolodziej & Hjelm, 2006, p.
109). Most of the infrared positioning systems are based on proximity, not ranging.
The location of the mobile device is assumed to be the same as the location of the
infrared base station. Active Badge (Want & al., 1992) system uses short-range
transmissions of pulse-width modulated infrared light for positioning. A mobile device
transmits periodically its identification code, which is received by fixed receivers
located at known locations. Multiple infrared transmitters sending their identification
code can be positioned throughout the building. Typically, each room in the
positioning area has an infrared receiver. The granularity of proximity-based infrared
system is low; it is not possible to determine the location of the badge at a higher
resolution than the known locations of the base stations (Kolodziej & Hjelm, 2006, p.
109).
The infrared signal can also be used to calculate the spatial position of the target
device. These systems utilize computer vision. They consist of infrared emitting tags
and a stationary mounted stereo camera. The stereo camera measures the angle of
arrival of the emitted light at two different points in space. The spatial positions of the
tracked tags are then calculated by triangulation (Aitenbichler & Muhlhauser, 2003).
The infrared light transmission range is only a few meters and it is restricted to the
optical line of sight between the transmitter and the receiver. This may be an
advantage in the proximity-based systems. The infrared signals received by the base
station must be originated from the tags located at the same room. Often the limited
range and line of sight requirement is a disadvantage, since the infrared receivers must
be deployed to every site separated by walls or other obstructions. The infrared
systems also suffer from direct sunlight and high ambient heat. The infrared
components are inexpensive, but the cost of deploying the base stations is quite high
(Want & al., 1992).
3.3
Ultrasound
Ultrasound is cyclic sound pressure with a frequency greater 25 kHz, which is the
upper limit of human hearing. Ultrasound has many application areas like medical
sonography, chemistry and positioning. In positioning applications the ultrasound
technology provides great ranging accuracy with low cost electronics.
The range between the ultrasound emitter and listener can be computed from the
ultrasound propagation time. The distance is obtained from the following equation
27
d  vus t
(3.1)
where vus is the speed of sound and t is ultrasound propagation time between the
emitter and the transmitter. Because the speed of sound is relative low (344 m/s), the
accuracy requirement for measuring the propagation time is also low. For example,
only 3 ms timing accuracy is needed to measure the distance within one meter
accuracy. In contrast, using radio frequency signal propagating at the speed of light, 3
ns timing accuracy would have been needed.
The speed of sound in air depends on environmental factors like atmospheric pressure,
temperature and relative humidity. The variations on the temperature affect most to the
speed of sound. In completely dry air the speed of sound depends only on the absolute
temperature T (in Kelvin), and is given by Priyantha (2005, p. 62).
vus  20.05 T
(3.2)
Priyantha (2005, p. 62) calculates that at room temperature 295 K the speed of sound
changes by 0.18% per one K. Temperature change of 10 K would change the ranging
measurement by almost 2%. Since the speed of sound has a relative large sensitivity to
temperature variations, the ultrasound range measurements should be temperature
compensated. In contrast, the speed of sound is not very sensitive to relative humidity
and atmospheric pressure variations. For example, a humidity change from 0% to
100% at room temperature and 101.325 kPa pressure changes the speed of sound only
by 0.5%. A pressure change from 101.325 kPa to 30 kPa at 298 K and 50% relative
humidity would change the speed of sound only by 0.6%. (Priyantha, 2005, p. 63)
Ultrasound signals used in most positioning systems (40-50 kHz) do not penetrate
physical objects such as walls. High probability of interference from other ultrasound
sources reduces the reliability of the positioning systems based on ultrasound.
Ultrasound-based positioning systems also need an RF link for communications.
3.4
Optical
The location information can be derived from analysis of video images. Vision is a
natural way to track the location, because it does not require any special devices
attached to the objects to be positioned. However, vision systems typically need huge
amounts of processing power to analyze the frames captured. The computer vision
28
systems also have the same line of sight problems as infrared systems, and they may be
complex and expensive. They are designed to capture position well in a single room
immersive environment. Many of these systems are not designed to be scalable, not
even to a building level (Kolodziej & Hjelm, 2006, p. 125-126).
Several research groups have explored using computer vision technology for
positioning. Example of such systems is Microsoft Easy Living (Microsoft Research,
2008).
3.5
Inertial Navigation Systems
Inertial navigation systems were first developed for navigating rockets. These systems
typically collect information from gyroscopes and accelerometers to determine the
position and orientation of the system. Such a system is also called an inertial
measurement unit (IMU).
Gyroscopes measure the angular velocity of an object. If the initial orientation of the
system is known in the inertial reference frame, the system’s current orientation can be
obtained by integrating the angular velocities. There are various types of gyroscopes
such as laser gyros, vibrating gyros and mechanical gyros. Also MEMS gyros have
been researched during recent years (Saukoski, 2008).
Accelerometers measure the linear acceleration of the system in the inertial reference
frame. Because the accelerometers are not aware of their own orientation and they
rotate with the system, the accelerations can only be measured relative to the moving
system. Orientation relative to the Earth can be measured with 3D accelerometers only
(Wikipedia, [Cited January 24, 2008]).
The linear acceleration of the system in the inertial reference frame can be calculated
by measuring both the angular velocity of the system and the linear acceleration of the
system, measured relative to the moving system. The position of the system can be
obtained by integrating the linear accelerations twice. First, the velocity vector is
calculated by integrating the acceleration vector and using the original velocity as
initial condition. Second, the position is obtained by integrating the velocity vector
obtained from previous step and using the original position as initial condition
(Wikipedia, [Cited January 24, 2008]).
The inertial navigation systems suffer from integration drift. Small errors in measured
accelerations and angular velocities cumulate progressively into large errors in position
29
and velocity. Inertial navigation system is often used to complement GPS when the
satellite signals are lost or to aid GPS satellite tracking when the user’s dynamics is
high (Kaplan, 1996).
Gyroscopes are too expensive for many applications, and accelerators cannot be used
alone to determine user’s orientation. In pedestrian navigation a better method to
utilize low cost MEMS accelerometers is to detect the steps of the pedestrian. The
distance walked or run is obtained by counting the steps and multiplying the step count
by estimated step length. A dead reckoning navigation system can be constructed by
combining a pedometer with a magnetic compass or gyroscope, which provides the
azimuth angle.
Magnetic Compass
A magnetic compass determines the azimuth by sensing the components of the
terrestrial magnetic field. Magnetic compass provides absolute direction, while
gyroscope provides relative direction to initial direction. However, the magnetic
compass suffers from disturbances of magnetic field. Magnets, electric currents or
large iron bodies can cause local disturbances, resulting large azimuth errors.
Combining Gyroscope and Magnetic Compass
Gyroscope does not suffer from local magnetic fields, but it needs the initial direction
and the direction error grows over time, due to non-constant bias. Thus, the gyroscope
needs external input for continuous calibration. A better direction measurement system
can be constructed by combining the gyroscope and magnetic compass into a single
unit. The gyroscope has excellent short time accuracy, while the magnetic compass
provides the absolute direction and long term accuracy. Usually the gyroscope is used
as the main source for direction. Magnetic compass provides the initial direction and
absolute direction values to compensate the gyroscope drift. Magnetic disturbances can
be detected by comparing the gyroscope and magnetic compass output. If the
difference in the angular rate exceeds the predetermined threshold, a magnetic
disturbance is detected. (Brännström, 2002, p. 11).
3.6
DC Electromagnetic
The direct current (DC) electromagnetic systems generate axial DC magnetic-field
pulses from a transmitting antenna in a fixed location. The system measures the
position and orientation of one or more receiving antenna sensors with respect to
30
transmitting antenna. The transmitting antenna is driven by a pulsed DC signal. The
receiving antenna measures transmitted magnetic field pulse and earth’s magnetic
field. (Hightower & Borriello, 2001)
The DC electromagnetic systems are used in the following applications (Ascension
Technology, [Cited January 24, 2008]):

head tracking in flight simulators

head, hand and body tracking for virtual reality environment

body tracking for virtual prototyping and visualization

manipulation of telerobotic controls

measuring a pilot’s line of sight for aiming weapon systems and interacting
with helmet mounted displays
Tracking systems utilizing the DC electromagnetic sensing compute the position and
orientation of the receiving antennas by measuring the response in three orthogonal
axes to the transmitted field pulse. The constant effect of the earth’s magnetic field is
taken into account in the computations. These systems provide 1 mm positioning
accuracy and 0.1 degrees orientation accuracy. Disadvantages are the high cost and
short range of the signal. The sensors must be located within 1 to 3 meters of the
transmitter. Magnetic trackers are designed to capture position in small scale and
controlled environment. They are not designed to be scalable for use in large locationaware applications. (Hightower & Borriello, 2001)
31
4 Observables
The classical observables used in radio navigation are:

Received signal strength (RSS)

Angle of arrival (AOA)

Time of arrival (TOA)

Time difference of arrival (TDOA)

Doppler shift
In addition to these metrics the proximity of the target device and base station is used
as a positioning observable. These positioning observables are discussed in the
following sections.
4.1
Proximity
The simplest positioning method is to use proximity measurements. Knowing only
whether or not two devices are in communication range is enough to give a position
estimate. In this method the location of the target object is considered to be the
location of the base station. If there are several base stations, the strongest base station
or the base station serving the target object is selected.
Cell of origin (COO) method used in cellular network positioning is one example of
methods based on proximity (Fikouras & al., 2001). In this method the caller’s location
is assumed to be the cell’s location serving the mobile phone. Cell of origin is not a
very precise location technique. The base station serving the mobile phone is not
necessarily the closest base station, and in the urban area the accuracy may be within
one hundred meters of the target. On the other hand, the accuracy may be 30
kilometers away from the target where base stations are less densely concentrated.
Because of its inaccuracy, Cell of origin is often used in conjunction with some other
positioning technology.
Proximity-based methods can also be applied in local position systems and indoor
navigation. Various different signal technologies and sensors are used to detect the
proximity of two objects. When radio frequency signals are used, the proximity
32
estimate can be considered as a binary quantization of RSS measurement. Often
proximity measurement devices are used in conjunction with distance-based
positioning technology. For example, some indoor positioning systems utilize access
control ports to monitor the presence of an active RFID tags near the access control
ports (Aeroscout, [Cited January 24, 2008]). Also infrared technology is used often in
proximity-based positioning systems (Kolodziej & Hjelm, 2006).
4.2
Received Signal Strength
The intensity of a transmitted radio frequency signal decreases as the distance from the
transmitter increases. This phenomenon can be used to estimate the location of a
mobile device. Measuring the received signal strength (RSS) is quite straightforward
and it is done in most of the radio receivers.
In free space the electromagnetic waves obey the inverse square law, which states that
the attenuation of the electromagnetic wave is inversely proportional to the square of
distance between the transmitter and the receiver. Inverse square law can be also
utilized in positioning techniques by converting the measured signal strengths into
range estimates. The location of the target object can be computed from these range
measurements by using trilateration. Another method to estimate the position of the
object by using RSS measurements is the location fingerprinting method, which is
explained in Chapter 5.4.
Figure 1. Relation between received signal strength and distance on the free space
model.
33
In the free space model, relation between received signal strength (RSS) and distance
is described as (Kitasuka, 2005b)
 4d 
Pr (d )  P0  20 log 10 
 [dBm]
  
(4.1)
where P0 is empirical constant,  is the wave length of the signal, and d is the distance
between the receiver and transmitter. Figure 1 illustrates the relation between received
signal strength and distance on free space model.
When at least two RSS measurements are available, a two-dimensional position of the
object can be solved by using trilateration. The RSS measurements form circles around
base stations, where the radii of the circles correspond to the estimated signal
strengths. The position estimation using three received signal strength measurements is
illustrated in Figure 2.
Figure 2. Intersection of three circles
The radio signal propagation is not only affected by the inverse square model, but also
by a number of other factors determined by its path from the transmitter to the
receiver. For example the air temperature and pressure affect to the signal attenuation
as well as any obstructions on the signal path. Also the multipath propagation affects
to the received signal strength. In indoor positioning systems, the position of the target
object is seldom computed by using geometric range estimates and trilateration.
Instead, pattern matching algorithms, such as location fingerprinting, are used more
often.
34
4.3
Angle of Arrival
Angle of arrival method (AOA) is sometimes also referred as direction of arrival
(DOA) method. Direction finding systems utilize antenna arrays and try to estimate the
angle of arrival of the signal of interest. The estimated AOA restricts the transmitter
location along line in the estimated AOA. When several AOA estimates are available
from multiple base stations at different locations, the position of the transmitter can be
located at the intersection of these lines. Figure 3 illustrates two dimensional
positioning using AOA measurements of three base stations. While only two AOA
measurements are needed to estimate the location of the transmitter, multiple AOA
measurements can be used to improve the estimation accuracy.
Figure 3. Location solution derived from angle of arrival measurements
Usually the angle of arrival (AOA) is measured at a base station. The hardware of the
antenna measuring the AOA of the signal is relative complex. In addition, the antenna
has to be precisely calibrated to the correct orientation. Measuring the direction of the
arrived signal is seldom practical at a mobile terminal, because it is difficult to keep
the antenna in the proper attitude in a moving object. Thus, AOA positioning systems
are usually network-centric. This means that the AOA measurements made by the base
stations are collected to a network server, where the position estimation takes place.
Positioning systems based on AOA measurements have some advantages with respect
to the other techniques. For example, in TOA and TDOA techniques the clocks of the
base stations have to be accurately synchronized to a common time reference. In AOA
systems the complex time synchronization is not needed. In AOA method, twodimensional position estimate can be calculated by using AOA measurements of two
35
base stations only. In contrast, the TDOA method needs measurements of at least four
base stations. TOA method would need measurements from three base stations if twoway communication (two-way ranging) is used, and four measurements if one-way
ranging is used.
Among disadvantages, the accuracy of the position estimate calculated from the AOA
measurements diminishes with increasing distance between the transmitter and the
receiver. In addition, AOA performs worse than TOA/TDOA methods in indoor
environment due to multipath propagation (Pahlavan & al., 2002). AOA estimation
techniques estimate the direction of a transmitter based on the strongest received
signal, which is assumed to be the LOS signal (Mizusawa, 1996). However, the NLOS
condition, where the direct LOS path is obstructed, is quite common in indoor
environment. In NLOS condition, the AOA estimate will be the direction of the
strongest multipath component, which causes errors to the AOA estimate. Also in nondominant direct path (NDDP) case, where LOS signal is available but not the strongest
signal, multipath causes significant errors to the AOA estimation. Antenna arrays are
also relative complex and thus quite expensive.
Due to the multipath problems, AOA technique is seldom used in indoor positioning
systems. However, AOA is sometimes used in outdoor local positioning systems as a
complementary measurement to TOA or TDOA measurements.
The AOA can be measured by using various techniques. The following AOA
estimation techniques are explained in the following sections:

Measuring the direction having the strongest signal strength

Measuring the phase difference between two or more antenna elements

Measuring the Doppler shift from a rotating antenna
4.3.1 Signal Strength Direction Finding
In the direction finding based on the signal strength, the AOA is found by rotating the
antenna. The source of the transmission is determined from the direction where the
maximum signal level is obtained. Another technique is to arrange an array of antennas
in circular form and measure the signal strength from all of the antennas. The AOA
estimate is obtained from the direction which has the strongest signal.
36
The signal strength direction finding does not provide very accurate results. To
measure the angle of arrival with a better accuracy requires much more sophisticated
techniques, like phase difference direction finding described in the next section.
4.3.2 Phase Difference Direction Finding
The angle of arrival can be measured by an accuracy of less than one degree by
measuring the phase difference between two or more antenna elements. This method is
also called short baseline interferometry or phase interferometry.
In this method the phase of the signal is measured in two antennas separated by a
distance d. If a wavefront arrives to the antennas at an angle θ, the extra path between
the two antennas causes a phase difference,  , between the two antennas. The angle
of arrival can be calculated from the phase difference using the equation:
  

 2d 
 ( )  arcsin 
(4.2)
where  is the wavelength of the signal. For unambiguous results, the antennas must be
spaced half a wavelength apart, or less.
The severe multipath condition of the indoor environment causes substantial errors in
phase measurements and thus to the AOA estimated by using phase difference
direction finding.
4.3.3 Doppler Direction Finding
The Doppler technique is a direction finding method that produces an estimate on the
received signal by measuring the Doppler shift induced on the signal by sampling
around the elements of a circular array.
The original method used a single antenna that physically rotated. When a rotating
antenna moves towards the wavefront, a maximum frequency occurs. When the
antenna moves to the same direction as the wavefront, a minimum frequency occurs.
The direction of the wavefront is obtained from the antenna attitude where the Doppler
shift is getting smaller and approaching to zero. The modern approach uses a multiantenna circular array with each antenna sampled in succession.
37
4.4
Time of Arrival
Many navigation systems like GPS utilize the concept of time of arrival (TOA)
ranging to determine user position. This concept is based on measuring the time it
takes for a signal transmitted by an emitter to reach a receiver. This time interval,
referred to as the signal propagation delay, is then multiplied by the speed of the signal
to obtain the distance between the emitter and the receiver. The signal may be an
electromagnetic wave, audible sound or ultrasound signal. The position of the target
object can be determined by measuring the propagation times from multiple base
stations at known locations.
In network centric systems the user terminal emits the signal and the base stations
receive the signal. In user centric systems, like GPS, the roles of the emitter and
receiver are opposite.
This chapter explains the time of arrival estimation process using radio signal
(electromagnetic wave). However, the TOA positioning algorithms explained in
section 5.1 can be used with ultrasound measurements as well. There are two major
challenges when estimating the TOA of the radio signal: the multipath phenomenon
distorts the TOA measurements especially at indoor environment, and the base stations
of the positioning system must be synchronized accurately in time. These challenges
are illustrated in the subsequent sections.
4.4.1 Measuring Time of Arrival at Multipath Conditions
The arrival time of the signal is measured in the positioning systems which are based
on measuring the signal propagation delay or time difference of arrival. The arrival
time of the signal can be measured by measuring the phase of received narrowband
carrier signal or directly measuring the arrival time of a wideband narrow pulse. In
wideband systems the pulses can be generated directly or using spread spectrum
technology (Pahlavan & al., 2002).
Narrowband Signals
In the narrowband systems, the distance between two points is obtained by measuring
the phase difference between the received and the transmitted carrier signals. The TOA
 of the signal is calculated as (Pahlavan & al., 2002)
38


2 f c
(4.3)
where  is the measured phase of the carrier signal and f c is the carrier frequency. If
the distance between the transmitter and the receiver is longer than the wave length of
the carrier signal, there are also multiple integer carrier waves between the two points.
The number of integer carrier waves can be obtained from lower frequency signal such
as data modulated on the carrier signal.
The severe multipath condition of the indoor environment causes substantial errors in
phase measurements. In multipath environment, the composite received carrier signal
is the sum of a LOS and multiple reflected carriers, all arriving along different paths.
The components of the received signal have the same frequency but different phase
and amplitude. The frequency of the composite received signal remains unchanged,
but the phase will be different from the phase of the directly arriving signal. The
composite signal contains no information from which the correct TOA or distance
could be estimated. Pahlavan & al. (2002) conclude that phase-based distance
measurement using narrowband carrier signal cannot provide accurate distance
estimate in a heavy multipath environment.
Wideband signals
Many navigation systems transmitting wideband signals utilize direct sequence spread
spectrum (DSSS) modulation. The most well known example of these systems is GPS.
The DSSS wideband signals are used also in most of the current WLAN
communication systems. The same principles as used in GPS can be used in ranging in
local positioning systems also.
In GPS satellite navigation system the DSSS provides the structure for the
transmission of ranging signals and navigation data (Kaplan, 1996, p. 39). The ranging
signals are coded by a known pseudo random number (PRN) code and transmitted by a
transmitter. Then a receiver cross correlates received signal with a locally generated
replica PRN code sequence using a sliding correlator. The distance between the
transmitter and the receiver is obtained from the arrival time of the first correlation
peak. The correlation process is illustrated in Figure 4.
39
Figure 4. Correlation of DSSS signal
A specific code phase generated by the satellite at t0 arrives at the receiver at t1. The
signal propagation time is t  t1  t 0 . At the receiver an identical replica code is
generated at t0, with respect to the receiver’s own clock. This replica code is shifted in
time until it achieves correlation with the satellite generated ranging code. If the
receiver clock is synchronized with the transmitter perfectly, the correlation process
would yield the true propagation time, and the distance between the transmitter and the
receiver is obtained by multiplying this propagation time by the speed of light (Kaplan,
1996, p. 41).
However, the transmitter and receiver clocks are generally not synchronized. In this
case the propagation time measurement includes the receiver clock offset from the
system time and the measurement is denoted as pseudorange. Another way of utilizing
the correlation process described above is to measure the time difference of arrival of
two signals from different transmitters with respect to receiver’s inaccurate clock. The
time difference of arrival measurement is then used in hyperbolic positioning.
Wideband DSSS signal is commonly used for the TOA-based ranging systems because
of several advantages as compared with other alternatives. From Figure 5 we observe
that the timing accuracy improves as the bandwidth increases (Szewczyk, 2004). In
addition, the large bandwidth of the DSSS signal helps to resolve multipath signals.
The correlation process also provides processing gain to the receivers of the DSSS
systems. This processing gain makes DSSS ranging systems perform much better than
other systems in suppressing interference from other radio systems operating in the
same frequency band (Li, 2003, p. 60).
40
Figure 5. Approximate range resolution versus signal bandwidth (Szewczyk, 2004)
In single-path radio propagation channels the variance of the time of arrival estimate is
bounded by the Cramer-Rao Lower Bound (CRLB). In practice this means that the
accuracy of the time of arrival estimate can be only a small fraction of the wavelength
of the signal. For example, in GPS the C/A-code noise at the receiver is on the order of
1.5 m while the wave length of the C/A code is about 300 m (Kaplan, 1996, p. 256).
However, the multipath radio propagation channel is very complex, and the CRLB is
not directly applicable to multipath environment. Pahlavan & al. (2002) state that in
multipath environment the resolution of TOA estimation in DSSS is roughly
determined by the signal bandwidth. Also, according to Dumont & al. (1994) the
resolution of multipath delays using the correlation method is limited by the chip
interval of the PRN sequence, and the range errors normally range from zero to half of
the chip interval. For example, if the PRN sequence is transmitted at a rate of 10 MHz,
the resolution of multipath arrivals is limited to 30 m, and the errors range from 0 m to
15 m.
If the same correlator designed for the single-path channels is used in the multipath
channel, the TOA estimation accuracy degrades significantly. Li (2003, p. 63) presents
the following general principles, which can be used to improve the TOA estimation in
multipath channels:

Increase the receiver sensitivity and the receiver dynamic range

Improve the resolution of estimation techniques by increasing the signal
bandwidth
41

Improve the resolution of estimation techniques by employing advanced signal
processing techniques

Place the transmitter and receiver in a way to minimize the occurrence of the
NLOS propagation scenarios between the transmitter and the receiver.
Improving the TOA estimation by increasing the signal bandwidth is not always
possible due to the scarcity of the available bandwidth. On the other hand, it is always
desirable to improve ranging accuracy using the same bandwidth. The accuracy of the
TOA estimation can be improved by employing super-resolution techniques. The
super-resolution techniques have been studied in the field of the model-based
parameter spectral estimation for a variety of applications (Li, 2003, p. 82).
The super-resolution techniques are used to improve the ranging resolution beyond
that what is possible with a traditional PRN correlation technique. Li (2003) and
Dumont & al. (1994) have studied performance of Multiple Signal Classification
(MUSIC) super-resolution algorithm in multipath conditions. MUSIC algorithm uses
an eigenvector-based approach to model the data, and it is described in detail by Li
(2003, p. 83).
Pahlavan & al. (2002) have compared the performance of MUSIC algorithm to inverse
Fourier transform (IFT) method and to traditional cross correlation with DSSS signals
(DSSS/xcorr). The simulation results of the comparison are shown in Figure 6.
Figure 6. Estimated TOA of the DLOS path and normalized time domain responses
obtained using three different techniques. The vertical dash-dot line denotes the
expected TOA. The x-axis delay in ns. (Pahlavan & al., 2002)
42
From Figure 6 it is observed that the MUSIC algorithm has much higher time domain
resolution than the other two methods. The MUSIC algorithm detects the expected
TOA accurately while the other two methods fail. According to Pahlavan & al., the
MUSIC method is preferred, especially when the signal bandwidth is small. However,
it should be noted that even super-resolution techniques could not eliminate large
ranging errors at some locations because of NLOS conditions between transmitter and
receiver. Faulty measurements should be identified and removed in the position
calculation process. It is worth mentioning that while the super-resolution techniques
can improve the TOA estimation performance, they also increase the complexity of
system implementation (Pahlavan & al., 2002).
Ultra-wideband Signals
As mentioned in the previous chapter, the accuracy of the TOA estimation can be
improved by increasing the signal bandwidth. UWB systems, whose bandwidth is
typically more than 1 GHz, have been researched in order to develop indoor
positioning systems with high TOA estimation accuracy. It has been shown that the
UWB signal does not suffer from multipath fading, which is desirable for accurate
TOA estimation in indoor environment (Ramirez-Mireles, 2001).
The use of ultra-wideband signals in indoor navigation is explained in more detail in
section 7.3.
4.4.2 TOA Measurement Methods
The techniques to estimate the time of arrival were discussed in the previous sections.
To convert the time of arrival estimation to the signal propagation delay estimation for
the purpose of ranging, the spatially separated transmitter and the receiver have to be
synchronized in time. Because the electromagnetic wave propagates at the speed of
light, only a one microsecond error in timing would cause a 300 meter error in range.
If the requirement for the distance measurement accuracy is 3 meters, the time
synchronization has to be made with 10 nanoseconds accuracy. Synchronizing the base
station clocks and measuring the time of arrival with 10 nanoseconds accuracy is a
challenging task.
43
The following methods to form the propagation delay estimate are discussed in this
section:

Synchronized transceiver method (one-way ranging)

Pseudorange method (one-way ranging)

Round-trip TOA method (two-way ranging)

Symmetric double-sided two-way ranging
In addition to these methods, the arrival time measurements can be used to form a time
difference of arrival (TDOA) estimate. The TDOA method is explained in section 4.5.
Synchronized Transceiver Method
In the synchronized transceiver method the transmitter and the receiver are accurately
synchronized in time. Also the transmission time has to be known. The signal
propagation delay estimation process is shown in Figure 7.
Figure 7. Signal propagation delay estimation in one-way ranging
The terminal A sends a radio signal to terminal B at time t0. Then the terminal B
receives the signal and measures the arrival time t1. The signal propagation delay t is
obtained by subtracting the known transmission time t0 from the measured arrival time
t1. The range between the terminals A and B is obtained from
rAB  c  t
(4.4)
where c is the speed of light and t  t1  t 0 . If there are at least three terminals
located at known positions, three dimensional position of a mobile terminal (xu, yu, zu)
can be obtained from the following set of equations
44
ri  ( xi  xu ) 2  ( yi  yu ) 2  ( zi  zu ) 2
(4.5)
where ri is the range estimate of the ith base station and (xi, yi, zi) denote the ith base
station position in three dimensions. The position computation algorithms to solve
Equation (4.5) are explained in 5.1.
Pseudorange Method
In all timing-based navigation systems the clocks of the base stations are accurately
synchronized in time with each other. The base stations may have atomic clocks or
GPS time reference, or the known distances between the fixed base stations are utilized
in the time synchronization process. However, synchronizing the clock of the mobile
terminal to the common time reference is a challenging task. Typically the base station
implementation can be relative expensive, but the price of the mobile terminal must be
very low. Use of accurate time references in the mobile devices is usually impossible
due to their high price. Many systems like GPS allow the use of a low cost time
reference in the user terminal.
The receivers of these systems measure the arrival time of the radio signal with an
inaccurate clock, whose time is biased from the common system time. The positioning
observable can be formed from these biased arrival time measurements in two ways:

A pseudo-propagation delay estimate is formed by subtracting the known
transmit time from the biased arrival time

A time difference of arrival (TDOA) estimate is formed by subtracting two
arrival time measurements from each other. The transmit time of the signal
need not be known.
The first method is called pseudorange method in this thesis. The TDOA method is
explained in section 4.5.
The pseudo-propagation delay consists of the geometric range time equivalent and the
unknown time offset tu between the mobile terminal clock and the system time. The
pseudorange  AB between the terminals A and B is
 AB  rAB  ct u
(4.6)
45
where rAB is the geometric range between the terminals A and B. The clock offset tu is
the same for all measurements. The unknown time offset adds one unknown variable
to the positioning equations. Thus, one additional measurement is needed if the mobile
terminal clock is not synchronized to the base station clocks.
If there are at least four terminals located at known positions, three-dimensional
position of a mobile terminal (xu, yu, zu) and the time offset tu can be obtained from the
following set of equations
 i  ( xi  xu ) 2  ( yi  yu ) 2  ( zi  zu ) 2  ct u
(4.7)
where  i is the range estimate of the ith base station and (xi, yi, zi) denote the ith base
station position in three dimensions. The position computation algorithms to solve
Equation (4.7) are explained in section 5.1.
Two-way Ranging Method
In synchronized transceiver method the remotely located transmitter and receiver must
be synchronized to a common time reference. To avoid the time synchronization
requirement, the two-way ranging method can be employed to measure the signal
propagation delay.
Figure 8. Two-way ranging (Hach, 2005)
Two-way ranging technique does not require exact time synchronization between the
clocks of the originator and recipient devices. On the other hand, two way
communications is needed. A range between two devices A and B is determined via
46
two-way exchange of a message and measuring its arrival time. This method known
also as two-way time-of-arrival (TW-TOA) or Round Trip Time (RTT) is illustrated in
Figure 8 (Hach, 2005).
The range between the devices A and B is measured as follows. First the originator
device A sends a ranging message D1 and records the transmitting time ta1 with respect
of clock A. Then the recipient device B receives the message and records the receiving
time tb1 with respect of clock B. Next the recipient B sends an acknowledgement
message A1 and records the transmission time tb2. Finally, the originator A receives the
acknowledgement message and records the receiving time ta2. The elapsed time
between the departure of ranging message from A, and the reception of the
acknowledgement at A can be approximated as
TroundA  2Tt  TreplyB
(4.8)
where Tt is the one way time-of-flight of the ranging message and TreplyB = (tb2- tb1) is
the reply time at the recipient device B. Usually the recipient transmits the reply time
TreplyB or tb1 and tb2 to the originator in a separate message after the ranging message
exchange (802.15.4a). The time-of-flight of the ranging message is solved from the
following equation
Tt 
TroundA  TreplyB
2
(4.9)
The range between the devices A and B is
rAB  Tt  c
(4.10)
where c is the speed of light.
Usually the reply time TreplyB is much larger than the time-of-flight Tt. Thus the round
trip time TroundA and the reply time TreplyB will be almost equal. Because the reply time
is relative long, the oscillators in both of the devices should be very stable. In other
words, drift in the oscillator A or B would cause significant error to the time-of-flight
measurement.
Hach (2005) gives an example. Let the signal propagation time Tt be 30 ns and reply
time TreplyB be 1.000000 ms. Then the round trip time TroundA will be 1.000060 ms.
47
Assume that clock A drifts +10 ppm and clock B drifts -10 ppm. Then the time-offlight will be
Tt 
TroundA  TreplyB

2
1.000060  10 3 s  (1  10 6 )  1.000060  10 3 s  (1  10 6 )s
 40 9 s  40ns
2
(4.11)
This example shows that clock accuracy of order 10 ppm may cause 10 ns error to the
time-of-flight measurement.
Symmetric Double Sided Two-way Ranging
Although the time synchronization between the devices is not needed in the two-way
ranging, the clock drift of the devices is still a problem. This is because the reply time
TreplyB is a rather long interval compared with the time-of-flight of the signal. The error
due to the clock drift can be eliminated by making the two-way ranging measurement
transaction two times (Hach, 2005). The first ranging measurement is calculated based
on a round trip from device A to device B and back to device A (like in two-way
ranging). The second measurement is calculated based on a round trip from device B to
device A and back to device B. This method, called Symmetric Double Sided Two-way
Ranging is illustrated in Figure 9 (Hach, 2005).
From Figure 9 it is obtained that time of flight of the ranging message is (Hach, 2005)
Tt 
TroundA  TreplyB  TroundB  TreplyA
4
(4.12)
TroundA and TreplyA are measured with the oscillator of the device A. Both of these
measurements are biased by the oscillator offset TA of the device A. Similarly TroundB
and TreplyB are biased by the oscillator offset TB of the device B. Symmetric double
sided two-way ranging protocol cancels the oscillator offsets (Hach, 2005):
d  c
TRound, A  TReply, B  TRound, B  TReply, A
4
T  TB  TB  TA
c A
0
4
(4.13)
48
Figure 9. Symmetric double sided two-way ranging (Hach, 2005)
According to Hach (2005) the symmetric double sided two ranging can work with up
to 40 ppm crystal tolerances, while better than 10 ppm crystal tolerances are needed in
the straightforward version of the two-way ranging.
4.5
TDOA Measurement Methods
The most common method to convert the time of arrival estimation to the signal
propagation delay estimation is the time difference of arrival (TDOA) method. In this
method the difference of arrival times of two signals is measured. Like in TOA
methods, the base stations must be synchronized in time, but the mobile terminal does
not need to be time synchronized. In contrast to TOA method, the transmit time of the
signal is not needed in TDOA method. In TDOA method the position of the mobile
terminal is obtained from intersection of multiple hyperbolas (Figure 10), while in the
TOA method the position of the mobile terminal is determined by intersection of
multiple spheres.
49
Figure 10. Intersection of two hyperbolas
In principle the time difference of arrival of two signals can be measured on the mobile
terminal. However, most local positioning systems are designed so that the mobile
terminal transmits the signal, whose arrival time is then measured at multiple base
stations with respect to a common time reference. The arrival times are then sent to the
server computer, which computes the TDOA estimates by subtracting the arrival time
of a pivot base station t1 from the arrival times of the other base stations ti :
tTDOA,i  t i  t1
(4.14)
where i ranges from 2 to N. Thus, N-1 TDOA estimates are obtained from N arrival
time measurements.
In addition to obtain TDOA estimate from the difference of two arrival times, the
TDOA estimate can be obtained by using cross correlation technique. In this technique
the received signal at one base station is correlated with the received signal at another
base station. The TDOA estimation using cross correlation technique is illustrated in
Figure 11 (Aatique, 1997).
Figure 11. Cross correlation method for TDOA estimation (Aatique, 1997)
50
In Figure 11 signals x1(t) and x2(t) are first filtered by H1(f) and H2(f), then correlated,
integrated and squared. This is performed by incrementing the value of the time delay
by small steps, until a peak correlation is obtained. The time delay causing the cross
correlation peak is an estimate of the TDOA.
51
5 Location Estimation Algorithms
This chapter discusses the location estimation processes using AOA, TOA, TDOA and
RSS observables.
5.1
Time of Arrival Techniques
The time of arrival technique exploits trilateration to determine the position of the
mobile users. Position estimation by the trilateration is based on knowing the range
from the mobile unit to at least three (3D) base stations at known locations. The range
is determined from the propagation delay of the signal or received signal strength
indicator (RSSI). Two-way ranging method illustrated in section 4.4.2 provides a range
estimate without the need to synchronize the base station or mobile device clocks.
In the positioning systems measuring the range, the position is estimated by
intersecting circles (2D) or spheres (3D) with radius ri and centre (xi, yi, zi), as
illustrated in Figure 12. Radius of the circle ri is obtained from the propagation delay
of the signal or received signal strength indicator (RSS). Point (xi, yi, zi) is the known
location of the base station transmitting or receiving the signal.
Figure 12. Trilateration by using three measurements
52
When distance measurements ri are available from at least three base stations, the
three-dimensional location of the receiver (xu, yu, zu) can be solved from the following
set of non-linear equations (Kaplan, 1996, p. 44):
ri  ( xi  xu ) 2  ( yi  yu ) 2  ( zi  zu ) 2  f ( xu , yu , zu )
(5.1)
where i ranges from 1 to 3 and references the base stations at known locations, (xi, yi,
zi) denote the ith base station coordinates in three dimensions, and ri is the range
measurement from ith base station.
In one way ranging the base station clocks must be synchronized with each other. In
these systems the time-of-flight of the signal is obtained by subtracting the transmit
time from the measured time of arrival of the signal. The transmit time is usually
obtained from the data of the signal used for the time of arrival estimation. If the
transmit time is not known, the time difference of arrival (TDOA) technique must be
used.
Usually in one-way ranging systems the clock of the mobile device is not synchronized
with the base station clocks. Thus the range derived from the signal propagation time
is denoted as the pseudorange. The pseudorange consists of a geometric range between
the transmitter and the receiver, and an offset between mobile device clock and base
station clock. This clock offset adds one unknown variable to the positioning
equations. Thus, one additional measurement is needed if the mobile terminal clock is
not synchronized to the base station clocks.
In order to determine the user position in three dimensions (xu, yu, zu) and the user
clock offset tu, four pseudorange measurements are needed. The user position can be
solved from the set of equations (Kaplan, 1996, p. 44)
 i  ( xi  xu ) 2  ( yi  yu ) 2  ( z i  z u ) 2  ct u  f ( xu , yu , z u , t u )
(5.2)
where i ranges from 1 to 4 and references the base stations at known locations, and (xi,
yi, zi) denote the ith base station coordinates in three dimensions, and tu represents the
advance of the satellite clock with respect to the system time, and  i is the measured
pseudorange to the ith base station.
The nonlinear equations (5.1) or (5.2) can be solved for the unknowns by using either
closed form solutions, iterative methods based on linearization, or Kalman filtering.
53
The linearization of the range equations (5.2) using Taylor series expansion is
presented in (Kaplan, 1996, p. 44-46). Using an approximate position location
( xˆu , yˆ u , zˆu ) and time bias estimate tˆu an approximate position can be calculated
ˆ i  ( xi  xˆu ) 2  ( yi  yˆ u ) 2  ( zi  zˆu ) 2  ctˆu  f ( xˆu , yˆ u , zˆu , tˆu )
(5.3)
The relationship between the unknown position, the approximate position and the
displacement between them is (Kaplan, 1996, p. 44):
x u  xˆ u  xu
y u  yˆ u  xu
(5.4)
z u  zˆu  z u
t u  tˆu  t u
Therefore, we can write
f ( xu , yu , zu , t u )  f ( xˆu  xu , yˆ  yu , zˆu  zu , tˆu  t u )
(5.5)
This function can be expanded about the approximate point and associated predicted
clock offset ( xˆu , yˆ u , zˆu , tˆy ) using a Taylor series (Kaplan, 1996, p. 45):
f ( xˆ u  xu , yˆ  y u , zˆ u  z u , tˆu  t u )  f ( xˆ u , yˆ u , zˆ u , tˆu ) 
f ( xˆ u , yˆ u , zˆ u , tˆu )
f ( xˆ u , yˆ u , zˆ u , tˆu )
xu 
y u 
xˆ u
yˆ u
(5.6)
f ( xˆ u , yˆ u , zˆ u , tˆu )
f ( xˆ u , yˆ u , zˆ u , tˆu )
z u 
t u  ...
zˆ u
tˆu
The expansion has been truncated after the first order partial derivatives to eliminate
non-linear terms. Taking partial derivatives and substituting (5.3) into (5.6) yields
(Kaplan, 1996, p. 45)
 i  ˆ i   i 
xi  xˆ u
y  yˆ u
z  zˆu
xu  i
yu  i
z u  ctˆu
rˆi
rˆi
rˆi
(5.7)
where ( xˆu , yˆ u , zˆu ) is an approximate position location, and tˆu is a time bias estimate
and ̂ i is an approximate pseudorange, and (xu , yu , zu , t u ) is the displacement
from the approximate position to the true position, and
54
rˆi  ( xi  xˆu ) 2  ( yi  yˆu ) 2  ( zi  zˆu ) 2
(5.8)
The above equation is simplified by introducing new variables where (Kaplan, 1996, p.
46)
 i  ˆ i   i
x  xˆ u
a xi  i
rˆi
y  yˆ u
a yi  i
rˆi
z  zˆ u
a zi  i
rˆi
(5.9)
Equation (5.7) can be written more simply as (Kaplan, 1996, p. 46)
 i  a xi xu  a yi yu  a zi zu  ct u
(5.10)
In Equation (5.10) there are four unknowns (xu , yu , zu , t u ) which can be solved
by making ranging measurements to four base stations. The unknown displacement
can be determined from Equation (5.11) shown in matrix form
Δρ  Hx
(5.11)
where
 xu 
 y 
u 
x  
 z u 


 ct u 
 a x1
a
x2
H
a x3

a x 4
a y1
a y2
a y3
a y4
a z1 1
 1 

 
a z 2 1
ρ   2 
  3 
a z 3 1



a z 4 1
 4 
(5.12)
In Equation (5.12)  i is the difference of approximate (predicted) and measured
pseudorange.
Equation (5.12) has the solution
x  H 1 Δρ
(5.13)
55
where H is the direction cosine matrix containing unit vectors pointing from the
linearization point to the location of the ith base station, x is the offset of the user
from the linearization point, and ρ contains differences between the predicted and
observed ranges (Kaplan, 1996, p. 47).
User’s position ( xu , yu , z u ) and the clock offset t u are calculated by using an iterative
process. At the system startup the approximated location and time offset
( xˆu , yˆ u , zˆu , tˆy ) are set to some initial value. In case of local positioning systems, the
center of the service area is usually a good choice. Next the direction cosine matrix H
and the predicted-minus-observed pseudorange vector ρ are computed. After that,
the unknown displacement vector (xu , yu , zu , t u ) is calculated by using Equation
(5.13). A new approximate position and clock offset are obtained by using Equation
(5.4). The iteration process is repeated until the length of the displacement vector does
not become smaller any more.
Least Square Solution
In reality, the range or pseudorange estimates are corrupted by errors, such as
measurement noise and multipath. These measurements errors translate to errors in the
components of vector x , as presented below (Kaplan, 1996, p. 47)
ε x  H 1ε meas
(5.14)
where ε meas is the vector containing range or pseudorange measurement errors and ε x
is the vector representing errors in the user position and receiver clock offset.
The error contribution ε x can be minimized by using measurements of more base
stations than the number of unknown variables in the computation. Using redundant
measurements will result an over-determined solution, which can be solved by
employing least squares estimation techniques.
The residual vector r is defined (Kaplan, 1996, p. 520)
r  Hx  ρ
(5.15)
The ordinary least squares solution is defined as the value of x that minimizes the
square of the residual. This is equivalent to minimizing the sum of squares of the
components in the residual vector. The square of the residual RSE is given as a function
of x (Kaplan, 1996, p. 520).
56
RSE (x)  (Hx  ρ) 2
(5.16)
The following solution minimizes the residual RSE (Kaplan, 1996, p. 520)
x  (H T H) 1 H T Δρ
5.2
(5.17)
Time Difference of Arrival Positioning Techniques
In hyperbolic positioning method the location of an object is calculated from the time
difference of arrival (TDOA) measurements. The position of the object is determined
from the intersection of a set of hyperbolas defined by the TDOA estimates, as
illustrated in Figure 13.
Distance between the mobile terminal and ith base station is
ri  ( xi  xu ) 2  ( yi  yu ) 2  ( zi  zu ) 2
(5.18)
where i is the index of the base station, ( xu , yu , z u ) is the unknown position of the
mobile terminal and (xi, yi, zi) is the known location of the ith base station. The
distance difference between the mobile terminal with respect to the base station where
the signal arrives first, is (Aatique, 1997, p. 23).
ri ,1  cd i ,1  ri  r1
 ( xi  xu ) 2  ( yi  yu ) 2  ( z i  z u ) 2  ( x1  xu ) 2  ( y1  y u ) 2  ( z1  z u ) 2
 f ( xu , y u , z u )
(5.19)
where c is the signal propagation speed, ri,1 is the distance difference between the first
base station and the ith base station, r1 is the distance between the first base station and
the mobile terminal, and di,1 is the measured TDOA between the first base station and
the ith base station. Equation (5.19) defines a set of nonlinear hyperbolic equations
whose solution provides the three-dimensional coordinates of the mobile terminal.
57
Figure 13. Hyperbolic positioning
Solving the set of equations (5.19) using Taylor series expansion is presented in the
following subsection. In the literature, there are also several other methods to solve the
hyperbolic equations. These methods are discussed later on this chapter.
5.2.1 Taylor Series Method for Hyperbolic Equations
The Taylor series method used with the TDOA measurements is almost the same as
the Taylor series method used with the TOA measurements. In GPS calculation (TOA)
the Taylor series method has been proven accurate and robust.
The linearization process presented in (Kaplan, 1996, p. 44-47) for TOA
measurements is applied here for TDOA measurements. The concept of the
approximate position is used again. The unknown user position is considered to consist
of an approximate component and an incremental component as stated below
x u  xˆ u  xu
y u  yˆ u  xu
(5.20)
z u  zˆ u  z u
Therefore we can write (Kaplan, 1996, p. 45)
f ( xu , yu , zu )  f ( xˆu  xu , yˆ  yu , zˆu  zu )
(5.21)
The time difference of arrival di,1 can be expanded about the approximate point
( xˆu , yˆ u , zˆu ) using Taylor series as presented in Equation (5.22).
58
f ( xˆ u  xu , yˆ  yu , zˆu  z u )  f ( xˆ u , yˆ u , zˆu ) 
f ( xˆ u , yˆ u , zˆu )
xu 
xˆ u
f ( xˆ u , yˆ u , zˆu )
f ( xˆ u , yˆ u , zˆu )
yu 
z u  ...
yˆ u
zˆu
(5.22)
Like with the TOA calculation, only the first order partial derivatives are taken into
account. Taking partial derivatives yields
ri ,1  ri ,1  (rˆi  rˆ1 )
 x  xˆ u xi  xˆ u
  1

rˆi
 rˆ1

 y  yˆ u yi  yˆ u
xu   1

rˆi

 rˆ1

 z  zˆ
z  zˆu
yu   1 u  i
rˆi

 rˆ1

z u

(5.23)
where
rˆi  ( xi  xˆu ) 2  ( yi  yˆu ) 2  ( zi  zˆu ) 2
(5.24)
Equation 5.23 is simplified by introducing new variables where
ri ,1  ri ,1  (rˆi  rˆ1 )
x1  xˆ u xi  xˆ u

rˆ1
rˆi
y  yˆ u y i  yˆ u
a yi  1

rˆ1
rˆi
z  zˆu z i  zˆu
a zi  1

rˆ1
rˆi
a xi 
(5.25)
Equation 5.23 can be written more simply as
ri ,1  ri ,1  (rˆi  rˆ1 )  axi xu  a yi yu  azi zu
(5.26)
In the Equation (5.26) there are now three unknowns (xu , yu , zu ) which can be
solved for by using three pair of TDOA measurements. In other words, measurements
of four base stations are needed. The unknown displacement can be determined by
solving the set of linear equations below shown in matrix form
59
Δr  Gx
(5.27)
where
a x 2
xu 



x  y u  G   a x 3
a x 4
 z u 

a y2
a y3
a y4
r2,1 
az2 



a z 3  r   r3,1 
r4,1 
a z 4 
(5.28)
In equations (5.27) and (5.28) x is the offset of the user from the linearization point,
and r contains the difference between the measured TDOA and predicted TDOA.
Like with the TOA method, In TDOA method the user’s position ( xu , yu , z u ) is
calculated by using an iterative process. The positioning process starts with an initial
guess, and improves the estimate at each iteration step by determining the local linear
least squares solution.
If there are more observations than unknown variables, more accurate position
estimation can be obtained by using the least squares method. In this case x is
calculated using the least squares approach as shown in Equation (5.29).
x  (H T H) 1 H T Δr
(5.29)
5.2.2 Comparing Taylor Series Method to Other Methods
In addition to Taylor series method, there are several methods proposed to solve the set
of hyperbolic Equations (5.19) (Aatique, 1997, p. 27).

Fang’s method

Friedlander’s method

Spherical-intersection method

Spherical-interpolation method

Divide-and-Conquer method

Chan’s method
These methods are discussed here briefly, but they are not presented in more detail.
Fang’s method provides an exact solution, when the number of TDOA measurements
equals to the number of unknowns. However, Fang’s method (Fang, 1990) cannot
make use of redundant measurements from extra base stations, which would improve
60
the positioning accuracy and reliability (Chan & Ho, 1994). This defect makes the use
of Fang’s method impractical in most of the positioning systems. Friedlander’s
method, spherical-intersection method and spherical-interpolation method utilize the
redundant measurements, but their estimators are not optimum (Chan & Ho, 1994).
The divide-and-conquer method can achieve the optimum performance when the errors
are small, but the Fischer information matrix is large (Aatique, 1997, p. 27).
Chan’s method is the preferred method in many papers (Chan & Ho, 1994; Dickerson
& al., 2005; Shin & Sung, 2002). Chan’s method is a closed-form solution (noniterative) and it can make use of redundant measurements. With low noise data, the
performance of the Chan’s method is excellent, but when the noise increases, it has
been reported to have some difficulties to provide the location (Li & al. 2005;
Mizusawa, 1996).
The Taylor series method has been proven reliable and accurate, and it works well
even in the presence of noise and NLOS error (Li & al., 2005; Mizusawa, 1996).
However, Chan & Ho (1994) criticize the Taylor series method to be converging
towards a local minimum, if the initial guess is not close enough to the true position.
Chan & Ho also state that Taylor series method may be computationally intensive.
However, in GPS calculation the Taylor series method goes to the local minima very
seldom. Even in the case when the initial guess is in the centre of the earth, the Taylor
series method converges towards the true position with four or five iterations only. If
the solution goes to a local minimum, a new initial guess can be selected randomly,
and the iteration can be tried again. In case of local positioning systems, the initial
guess can be set to the middle of the search area surrounded by the base stations. In
this case it is not probable that the Taylor series method would provide an incorrect
local minimum as a result.
According to Aatique (1997, p 27) the linearization process can introduce significant
errors when DOP value is bad. With a bad DOP value relative small ranging error can
result a large positioning error. It has been shown that eliminating the second order
terms of Taylor series expansion can provide significant errors in this situation.
However, in practical local (or global) positioning systems the computation should
only be done with relative good DOP values. The inaccuracy due to eliminating the
higher order terms of Taylor series expansion is very small when compared to the
typical ranging accuracy.
Taylor series method can be easily improved so that the erroneous measurements due
to NLOS conditions are detected and identified. In case of over-determined solution,
61
the erroneous observations can be identified and removed by using Receiver
Autonomous Integrity Monitoring (RAIM) algorithm (Kaplan, 1996, p. 306).
5.3
Angle of Arrival Techniques
In AOA technique the position of the mobile terminal is calculated from the angle of
arrival measurements of the received signals. In two dimensional systems the azimuth
angle from the base station (BS) to the mobile terminal (MT) is measured. In three
dimensional systems also the elevation angle  is measured. For simplicity, this
chapter presents only the two-dimensional computation. The three-dimensional
computation is explained in (Du & Lee, 2004).
The geometry of the AOA location technique is shown in Figure 14.
Figure 14. AOA Positioning technique
The measured angle of arrival i from base station i and the measurement error ni
satisfy (Vidal & al., 2001, p. 25),
 i   i 0  ni
(5.30)
with
 yu  yi 

 xu  xi 
 i 0  arctan
(5.31)
62
where (xu, yu) are the mobile station coordinates and (xi, yi) are the coordinates of the
ith base station. Equation 2.4 can be rewritten as (Du & Lee, 2004)
ni   i   i 0
(5.32)
Taking the sine of both sides of the equation 2.7 and multiplying by ri we have (Du &
Lee, 2004)
ri sin ni  ri sin( i   i 0 )
(5.33)
ri sin ni  ri sin  i cos  i 0  ri cos  i sin  i 0
(5.34)
or
where ri  ( xu  xi ) 2  ( yu  yi ) 2 is the distance from the mobile terminal to the ith
base station.
Figure 15. The geometry of AOA location method
From Figure 15 we obtain (Du & Lee, 2004)
xu  xi  ri cos  i 0
yu  yi  ri sin  i 0
Combining the equations (5.34) and (5.35) we get (Du & Lee, 2004)
(5.35)
63
ri sin ni  ( xu  xi ) sin  i  ( yu  yi ) cos  i
(5.36)
To simplify the equation, it is assumed that the measurement error of the angle is very
small, i.e., |n | << 1 and therefore sin ni  ni (Du & Lee, 2004). Based on this
assumption, equation 2.10 can now be written in linear form
0  ( xu  xi ) sin  i  ( yu  yi ) cos  i  ri ni
(5.37)
Arranging the unknown variables to the left hand side of the equation we obtain
 xu sin  i  yu cos  i  xi sin  i  yi cos  i  ri ni
(5.38)
Equation (5.38) is linear with respect to x and y. The position of the target object (x, y)
can be solved now, if there are two equations of form (5.38). The equation 2.14 can be
expressed in matrix form as
h  Fx
(5.39)
where
  sin 1
F
 sin  2
cos 1 
cos  2 
 x
x 
 y
  x sin 1  y1 cos 1 
h 1

 x2 sin  2  y 2 cos  2 
(5.40)
Equation (5.39) has a solution
x  F 1h
(5.41)
The over-determined solution is
x  (F T F) 1 F T h
5.4
(5.42)
Location Fingerprinting
Location fingerprinting is a position estimation algorithm based on pattern-matching.
Usually the systems employing the location fingerprinting method use RSS
measurements to obtain the location of the target device. However, the RSS
64
measurements are not used to estimate the distances between the base stations and the
target device, neither trilateration is employed for position estimation. Instead, the
position of the target device is obtained by finding the best match between a vector of
measured RSS values and a vector of known ―fingerprint‖ stored earlier in database.
The fingerprinting technique is relative simple to deploy, when compared to AOA or
TOA/TDOA range estimation techniques. No specialized hardware is needed at the
mobile device or base stations, and any existing wireless LAN infrastructure can be
reused for the positioning system. Currently, the location fingerprinting is one of the
most common methods used in indoor positioning systems. Example of a commercial
product utilizing the location fingerprinting algorithm is Ekahau (Ekahau, [Cited
January 24, 2008]).
The location fingerprinting technique fundamentally assumes that each potential
location in the target area ideally possesses a distinct unique RF ―signature‖. The
closer the reality is to this ideal, the better the performance of the location
fingerprinting solution (CISCO Systems Inc., 2006).
In principle, the location fingerprinting technique can be used both with the selfpositioning systems and network centric systems. In Ekahau’s system, the RSS is
estimated at the mobile terminal. The measured RSS is then sent to the Ekahau
positioning engine running on the server. Ekahau’s system is thus a network centric
approach. In this text it is assumed that the RSS values of the signals transmitted by
the base stations are estimated at the mobile terminal.
The location fingerprinting process contains two phases: the offline phase and the
online phase. In the offline phase the RSS values are measured from multiple access
points at each point in the positioning area. This vector of RSS measurements, called
location fingerprint, is stored to the database along with the position coordinates of that
point. In the on-line phase the mobile device will measure the RSS from multiple
access points. The vector of measured RSS samples is then compared to the location
fingerprints stored in the database. The position associated with the location fingerprint
having the closest match to the measured RSS vector is selected as the position
estimate.
There are two methods to build the database. In the first method based on the empirical
measurements the user of the mobile device walks around in the building. At different
locations the user records the physical coordinates and the signal strengths from each
65
access point in the range. In the second method the database is constructed using a
mathematical model of indoor radio propagation (Bahl & Padmanabhan, 2000b).
5.4.1 Offline Phase
In the offline (calibration) phase the entire area is first divided to a rectangular grid of
points. The location fingerprints are then collected by performing a walk-around of the
target environment with a mobile device. At each point of the grid, the mobile terminal
measures the received signal strength (RSS) from multiple access points. This vector
of RSS values is called location fingerprint, and the size of the vector for this sample
point equals to the number of base stations that can be detected by the mobile device.
The vector of RSS values at that point is stored to the database with the position
information of that point. Two location fingerprints and their respective locations are
shown in Figure 16.
Figure 16. Example of location fingerprints estimated for two locations using four
base stations (BS) (CISCO Systems Inc., 2006)
During the offline phase the RSS values are measured with enough statistics before the
location fingerprint is stored to the database.
66
5.4.2 Online Phase
In the on-line phase the mobile device will measure the RSS from multiple access
points. The vector of measured RSS samples is sent to the central server for position
computation. Another technique is that a group of access points collect the RSS
measurements from a mobile device and send them to the central server. The server
estimates the location of the mobile device and reports the estimate back to the mobile
device or to another application requesting the positioning information.
A simple approach to estimate the location of the mobile device is to compare the
measured RSS vector to each fingerprint in the database. The location of the mobile
device is the point in the grid having the fingerprint closest to the measured sample
vector. The most common algorithm to find the best match between the measurement
and the fingerprint is to compute the Euclidean distance between these vectors. In
location fingerprinting the Euclidian distance is also called signal distance
(Kaemarungsi & Krishnamurthy, 2004). Two vectors are normally used in the location
fingerprinting technique to estimate the position of the mobile device. The first vector
called sample RSS vector consists of sample RSS measured at the mobile device from
N access points. Sample RSS vector is denoted as: R  1 ,  2 ,  3 ,...,  N . The second
vector called location fingerprint vector consists of the true means of all the received
signal strength random variables at a particular location from N access points. The
values are measured at the off-line phase and recorded in the location database. This
~
vector is denoted as: R  r1 , r2 , r3 ,..., rN  .
The signal distance between the sample RSS vector and the average RSS vectors is
calculated as follows (Kaemarungsi & Krsihnamurthy, 2004):
1
1
N
2  N
2
Z   (  i  ri ) 2    (qi ) 2 
 i 1

 i 1

(5.43)
The location fingerprint vector that minimizes the distance in signal space correspond
the physical coordinates as its estimate of the user’s location. This simple method is
called nearest neighbor in signal space (NNSS) (Bahl & Padmanabhan 2000a).
Several improvements have been developed for the NNSS method. There are also
other algorithms which use neural networks (Small & al., 2000) or Bayesian modeling
(Castro & al., 2001) to relate the sample RSS vector to the fingerprint in the database.
Ekahau system is based on Bayesian modeling approach (Eissfeller & al., 2004).
67
5.4.3 Improvements to the NNSS Method
In practice, the actual location is not exactly in the same place as the locations stored
in the radio map. As a consequence, there can be several location fingerprints whose
signal distance is very close to the fingerprint having the minimum signal distance.
There is no reason to pick just the closest one and discard others that are almost as
close. The NNSS algorithm can be modified so that a small number of closely
matching fingerprints is used instead of using only the closest one. The estimate of the
user’s position is obtained by averaging the physical coordinates of these fingerprints.
Bahl and Padmanabhan (2000b) call this method as NNSS-AVG. They found that for
small number of nearest neighbors there is a small improvement over the single nearest
neighbor approach. For large number (more than 8) of nearest neighbors the location
estimation error is increased.
Another method to improve the NNSS method is to take the user’s dynamics into
account. Bahl and Padmanabhan (2000b) have developed a method which they call
continuous user tracking. The idea behind the continuous user tracking is to utilize the
position information from the past. The user’s position at a given time instant is likely
to be near that at the previous time instant. In this method the nearest k NNSS values
are stored to a buffer of depth h. A Viterbi-like algorithm searches a path between
NNSS values which minimizes the path between the oldest and newest value in the
buffer. Another possibility to utilize the continuous user data is to use an Extended
Kalman Filter to smooth the position data. Ekahau system utilizes also the digital map
of the building so that impossible paths of the mobile terminal (those going through
walls) are rejected (Eissfeller & al, 2004).
5.4.4 Effect of the Environment and Infrastructure on Performance
Various parameters of the environment and the infrastructure of the system affect to
the performance of the location-based system. The performance of location
fingerprinting algorithm depends on the user’s orientation, number of base stations,
grid spacing used, and number of samples collected during the offline and online
phase. These factors are studied in the following sections.
Impact of the User Orientation
Bahl & Padmanabhan (2000a) discovered that signal strength at a given location varies
significantly depending on the direction the user is facing. In one orientation, the
antenna of the mobile device may have LOS connectivity to the antenna of the access
68
point, while in the opposite orientation the user’s body may form an obstruction.
Because human body consists of 70% water and the resonance frequency of the water
is 2.4 GHz, the signal is absorbed when the user obstructs the signal path (Ladd & al.,
2002). Kaemarungsi & Krishnamurthy (2004) found that received signal strength may
attenuate more than 9 dBm due to the obstruction of human body. The attenuation of
the body may completely block the signal from an access point. Similar results were
obtained by Bahl & Padmanabhan (2000a). These results suggest that user’s
orientation is crucial and should be taken into account in the computation.
Bahl and Padmanabhan (2000a) developed the location fingerprinting method so that
the user’s orientation is also recorded during the off-line phase. For each physical
location the received signal strength is measured in east, north, south and west
directions of the building. All of these four signal strength measurements are stored in
the database separately, so that there are four individual signal strength values for each
physical location. During the online phase the measured signal strength is compared to
the fingerprints of each location and cardinal direction. This approach will need four
times as much data to be stored in the database as the original approach, but would
improve the accuracy of the positioning.
Effect of the Number of Access Points
Intuitively, increasing number of access points improves the accuracy and precision of
the location estimated by the location fingerprinting technique. This holds also to the
navigation systems using TOA or AOA measurements. Increasing the number of base
stations improves the geometry. Additionally, the extra observations make the effect of
noisy or incorrect observations smaller.
In location fingerprinting increasing the number of access points improves the
probability of returning the correct location. According to the tests a higher number of
access points improves the precision, but the probability does not increase significantly
if the number of access points is bigger than five (Kaemarungsi 2005; Bahl &
Padmanabhan 2000b). If the RSS measurements have large standard deviation, the
high number of access points has more effect on the accuracy and the precision.
Kaemarungsi (2005, p. 138) recommends that at each data point the receiver is able to
track at least four access points.
In principle, all the points in the area may have unique location fingerprints even if
there is only one access point. However, probability that two or more points share a
fingerprint with values close to each other is high. In practice, at least three access
69
points are needed in most of cases. The mutual locations of the access points also
affect to the performance. If the access points are far away in the same direction from
the research area, the fingerprints of the points in this area are probably close to each
other.
Impact of the Grid Spacing
In location fingerprinting method the research area is covered by a rectangular grid of
points. The distance between two adjacent physical locations is called grid spacing and
reported in meters or feet (Kaemarungsi, 2005, p. 6). The grid spacing is selected prior
to the site-survey. Large grid spacing provides poor accuracy or granularity.
Kaemarungsi (2005, p. 110) examined the probability of returning the correct location
when the grid spacing goes smaller (the area of the point reduces). He found an
obvious result: the probability of returning the correct location reduces with small grid
spacing. Bahl & Padmanabhan (2000a) measured the error distance (accuracy)
between the estimated and true locations while increasing the number of data points.
They found that increasing the data points first diminished the error distance rapidly,
but with large number of data points the error distance did not diminish any more.
Kaemarungsi (2005, p. 139) suggests that grid spacing should be greater than 1.25 m.
Impact of the Number of Samples
Collecting enough statistics during the off-line phase is essential to achieve good
performance. If the positioning system only uses the mean values of RSS to create
fingerprints, a small number of samples is sufficient. In RADAR system Bahl &
Padmanabhan (2000a) found that order of 20 samples is sufficient. Even with a smaller
number of real-time samples the accuracy starts to approach the accuracy obtained
using all of the samples. They found that with 1 real-time sample, the error distance
was 30% worse than with all of the samples. With 2 samples, it was 11% worse and
with 3 samples it was under 4% worse. Larger number of samples (order of 300) is
needed for the probabilistic approach to create accurate diagrams (Kaemarungsi, 2005,
p. 72). Also the sampling period of the device affects how many samples are needed.
According to the results of Kaemarungsi (2005, p. 73) only 30 to 50 samples would be
sufficient for a location fingerprint on the mean values only. If the distribution of the
RSS values is needed, the distribution can be completely described by the mean and
variance. Therefore, the convergence of standard deviation can be used as a condition
to stop collecting new samples. According to the results of Kaemarungsi research, a
number between 150 and 200 should be sufficient.
70
Kaemarungsi (2005, p. 123) indicated that during the online phase he used only one
sample to detect the location.
5.4.5 Radio Propagation Model
Instead of building the radio map empirically, i.e. measuring the RSS manually from
multiple access points at each point in the grid, the radio map can be built by modeling
the radio wave propagation. Bahl & Padmanabhan (2000b) have studied using the
radio propagation model in location fingerprinting. The purpose of using radio
propagation model was to decrease the amount of time needed to take measurements in
the building. At the offline phase the theoretically computed signal strengths are
calculated for each grid point on the radio map. At the online phase the NNSS
algorithm or its variants are used to determine the user’s location by the same way as
with the empirically constructed radio map. Obviously the performance of this method
is directly impacted by the quality of the propagation model.
At indoor environment the signal propagation is dominated by reflections, diffraction
and scattering of radio waves caused by the walls and other obstacles within the
building. Generally the transmitted signal reaches the receiver via multiple paths.
Signal components arriving from direct and indirect paths combine to produce a
distorted version of the transmitted signal. This affects also the received signal
strength. Because of the multipath, modeling the radio wave propagation indoors is
very challenging. Bahl & Padmanabhan (2000b) examined three different models and
chose the Wall Attenuation Factor (WAF) propagation model for further examination.
In this method the theoretical signal strengths were estimated as a function of distance
and the number of walls in the path from the access points to the grid points. A few
measurements are beneficial to determine the actual attenuation caused by each wall.
The reported results of the radio propagation model were worse than the results of the
empirical method.
5.4.6 Aliasing
One problem in location fingerprinting is a phenomenon which Bahl & Padmanabhan
(2000b) call aliasing. Large error in position may occur when two points are far apart
physically but very close together in signal space. Such aliasing may happen indoors
due to complex radio signal propagation environment. A point close to an access point
may have similar signal strength as another point far away. This may happen when the
signal strength measured close to the access point is attenuated by an obstruction and
71
the far apart point has an unobstructed signal. How often the aliasing can occur,
depends on the building layout and the number of access points and their location.
72
73
6 Indoor GPS
This chapter presents an overview of GPS-based indoor positioning technologies and
compares these technologies to the other indoor location techniques and to the local
positioning systems.
6.1
Introduction
Positioning technologies intended specially for indoor use have many advantages over
the Global Positioning System. For example, systems based on WLAN and UWB
provide superior indoor positioning performance when compared to the satellite-based
navigation systems. However, GPS is still an attractive method for the indoor
positioning in some applications, where the positioning accuracy is not the most
important requirement. Today GPS is the most commonly used navigation system,
which makes it a potential alternative also for indoor use.
Navigation systems based on a local radio network have the drawback that they can
only be used locally, on the coverage of the local antennas or beacons. Most of the
local positioning systems are also network-centric: the position of the moving targets
must be calculated in the server from the measurements made by the fixed receivers. In
centralized systems the moving object cannot measure the distances to the beacons and
calculate the position by itself. If the objects to be located are persons, the centralized
architecture may hurt the individuals’ privacy.
The need for the GPS indoor location ability was first found in the mid-1990s when
U.S. FCC decided to require ability to locate the mobile phone in case of emergency
call (E911). Many users require continuous outdoor and indoor availability, and using
a GPS receiver indoors has some advantages. The GPS module receives the satellite
signals and computes the position of the object, and no one else needs to know the
user’s location. The same GPS receiver can be used everywhere if the coverage of the
GPS is extended to the environments where the signal is attenuated.
The GPS signal received at the Earth’s surface is very weak. According to IS-GPS-200
(2004) specification, the power level of the GPS coarse acquisition (C/A) code on the
earth’s surface is about -158.5 dBW (i.e. -128.5 dBm). At indoor use the structures of
74
the building attenuate the weak GPS signal further. According to Dedes and Dempster
(2005) the GPS receiver needs to be able to track signals with power levels ranging
from -160 dBW to -200dBW to operate indoors. Standard GPS receivers designed
during 1990’s are not able to track such weak signals. However, due to the recent
development in GPS technology, most of the modern GPS receivers are able to track
the GPS satellites also indoor buildings, but with lower accuracy.
In the following sections three methods are discussed to help the GPS receiver operate
better in weak signal conditions. First, the GPS receiver sensitivity can be improved by
integrating the GPS signal for a longer time and by assisting the receiver with external
aiding information. Second, pseudolites transmitting a GPS-like signal can be installed
inside building. Third, GPS signal received at a roof antenna can be radiated inside
building.
6.2
Assisted GPS and High Sensitivity GPS
Most of the GPS receivers have separate acquisition and tracking modes. GPS signal
acquisition has traditionally been a slow process. The acquisition process can be
speeded up by using external aiding information, such as satellite orbit parameters,
GPS time and approximate position. This concept of using the aiding information is
called assisted GPS. According to Diggelen & Abraham (2001) assisted GPS provide
10 dB increase in sensitivity, which is not enough for indoor operation. One would
need at least 30 dB improvement over the traditional GPS receivers. In addition to the
GPS aiding, a technology called high sensitivity GPS (HSGPS) is needed to enable
GPS operation indoors. In HSGPS technology the GPS signal is integrated for a long
period, which improves the signal to noise ratio.
6.2.1 GPS Signal Acquisition
GPS signal acquisition is a time-consuming process for traditional GPS receivers.
Usually GPS receivers search the incoming signal in two dimensions: code phase and
Doppler frequency. In the code phase dimension the GPS receiver acquires the satellite
signal as follows: The GPS receiver generates a replica of the PRN code that is
transmitted by the satellite. The phase of the replica code is shifted until it correlates
with the satellite PRN code. When the phase of the receiver generated replica code
matches the phase of the incoming satellite code, there is maximum correlation. When
the phase of the replica code is offset by more than one chip of the incoming satellite
signal, there is minimum correlation. The GPS receiver must also detect the satellite in
carrier phase dimension by replicating the carrier frequency and Doppler. The Doppler
75
is due to the relative dynamics between the satellite and the receiver. Also the
receiver’s reference oscillator instability affects the frequency uncertainty. Thus, the
GPS signal acquisition and tracking is a two dimensional (code and carrier) process
(Kaplan, 1996, p. 119). This process is illustrated in Figure 17.
Figure 17. Two-dimensional acquisition search space
The two-dimensional search space could have a Doppler of ± 4.5 kHz and a 0-1022
chip C/A code phase. Correlation is performed in each cell (called bin) by using
predetection integration and comparing the correlation value with the detection
threshold. The search goes on to the next bin until the correlation value is more than
the threshold. Usually two samples per chip are used in searching in the code phase.
Because there are two channels per satellite (In-phase and Quadrature-phase), there is a
total of 4092 samples. The width of the bin fc in the frequency dimension is obtained
from the equation (Singh, 2006b, p. 19)
fc 
2
3N
(6.1)
where N is the signal integration time.
For strong GPS signals, an integration time of C/A code repetition time (1 ms) is
sufficient (Dedes & Dempster 2005). This results 667 Hz frequency width. In this case
the total number of frequency bins would be 4092 code samples * 9.0 kHz / 667 Hz =
76
55214 bins. Increasing the integration period increases the number of search bin
needed. Long integration time is used in HSGPS receivers when the signal weak. For
20 ms integration time the frequency width is 33 Hz and the number of search bins is
549672.
6.2.2 Assisted GPS
Typically GPS receivers have been designed to dwell for about one millisecond in
each search bin (Diggelen & Abraham, 2001). If the number of bins is 50000, then it
would take 50 seconds to search the entire frequency/code delay space. The need to
make the acquisition process faster has led to an idea of assisted GPS. The search area
can be reduced significantly if the visible satellites, their Dopplers and relative code
delay offsets are known in advance. In assisted GPS the initial position estimate, time
estimate and satellite ephemeris can be provided from the external source. The receiver
can then estimate the satellite Dopplers and relative code delay offsets. As a result the
frequency/code delay space to be searched is reduced significantly.
According to Diggelen & Abraham (2001) the range of possible frequencies can be
reduced by a factor of ten by using GPS aiding information. As a result the acquisition
time is reduced to 10 percent of the original value. It is interesting that instead of
speeding acquisition, one could make use of the aiding to increase the dwell time in
each bin, thereby increasing the sensitivity. However, Diggelen & Abraham calculate
that this does not improve sensitivity enough to allow indoor operation. If the search
space is reduced by a factor of 10, then the dwell time in each bin could be increased
from 1 to 10 ms, without increasing the total search time. According to Diggelen &
Abraham increasing the integration time from 1 ms to 10 ms yields a signal to noise
ratio gain that approaches 10 dB. However, this is not enough since GPS signal levels
indoors are from 20 to 30 dB below the signal levels outdoors. As a result, aiding
information provides significant improvement to the signal acquisition or receiver
sensitivity, but it is not alone sufficient method to make a receiver work indoors.
Changes in the hardware architecture of the GPS receiver are also needed.
6.2.3 Weak GPS Signal Processing
When the signal is attenuated, integration can be done over a longer interval in order to
increase the processing gain. Coherent integration interval can be up to 20 ms when
GPS signal navigation message bit transition occurs. If integration is performed over a
longer period, the signal must be squared. Squaring also squares the noise, thus
diminishing the net gain of non-coherent integration. Lachapelle (2004) claims that
77
coherent integration is often the only method used in the tracking mode. In signal
acquisition, non-coherent integration of a few hundred milliseconds is often used.
However, Signav has demonstrated acquisition down to 183 dBW and tracking down
to -185 dBW using 256 ms integration periods without external assistance (Bryant & al
2001).
When the integration period is long, the Doppler error may degrade the tracking
performance. The receiver clock drift and the user antenna motion can cause
significant phase changes and effectively limit the integration time. The signal
integration in conventional receivers is not a real time process. In other words, there is
no feedback from the GPS signals to the local oscillator during the integration period.
Lachapelle (2004) proposes that using a higher grade clock reduces the effect of the
clock drift component. He also suggests that the user motion component may be
eliminated by using inertial measurement unit. However, both the high grade clock and
the inertial measurement unit are too expensive components for a cell phone and other
portable devices.
Diggelen & Abraham (2001) present a hardware approach which enables long
integration periods without the need of high precision clock or inertial measurement
unit. This approach contains a convolution processor which performs a real-time
convolution of GPS signals over the entire range of possible code delays. The
convolution processor contains over 2000 correlators per satellite, and it can compute
all possible correlation delays in real-time. According to Diggelen and Abraham, this
design makes the separate acquisition and tracking modes obsolete. Because the
hardware processing approach is a real-time implementation, it supports GPS-based
frequency adjustment. The feedback loop from the convolution processor drives the
local oscillator to produce the correct frequency.
Assisted GPS is not used in all cellular network techniques. Also concrete walls of the
buildings block the GPS signal completely. Supersensitive assisted GPS may be too
expensive and it is still not able to navigate in all places.
6.3
Pseudolites
Pseudolites are ground-based GPS signal transmitters which can improve the GPS
signal availability. Pseudolites can be installed inside buildings or other places where
the GPS satellite signal coverage is poor or completely missing. Pseudolites use low
cost temperature controlled oscillators while the GPS satellites use high precision
atomic clocks. Usually the local navigation system consists of at least four pseudolites.
78
Sometimes the signals from the GPS satellites can be used together with the
pseudolites. Usually the clocks of the pseudolites are synchronized to each other and
also the GPS (satellite) time.
The pseudolites are still too expensive for many purposes, but there are plans to use the
pseudolites to help airplane landing system, which requires CAT-III. Another
application is to augment the GPS navigation in deep open mines, where only the
highest elevation GPS satellites are visible.
A problem with pseudolites is so called near-far problem. The received signal
strengths of the satellites and the pseudolites have to be on the same level. All the
satellites are within the distance of 20000 - 26000 km from the user. The received
signal level is always on the same range. However, the pseudolites are quite close to
the user, probably within tens of meters, and the relative distance to the pseudolites
varies quickly when the user moves. When the user is near the pseudolite, the receiver
cannot track the other pseudolites or GPS satellites. Some pseudolite systems use time
switching, which enables only one pseudolite to transmit the signal at time. Usually the
standard GPS receiver hardware works with pseudolites, but the GPS receiver
firmware requires some modifications due to the different PRN codes and navigation
message.
The L1 band is protected globally by legislation. In general, it is illegal to transmit on
L1 frequency band without an approved radio licence.
6.4
GPS Repeater
GPS repeater consists of a GPS antenna located at the roof of the building, a GPS
reference receiver and an amplifier that amplifies the signal to be radiated inside the
building. The GPS signal can then be tracked inside the building with a GPS receiver.
It is well known that in this case the GPS receiver measures the roof antenna position.
However, the fourth unknown component of the navigation solution, the clock bias,
changes when the distance from the user to the radiator changes. This information can
be used to solve the position of the indoor receiver. Four repeaters are needed to
calculate the user position. Because the receiver can track signals only from one
repeater at time, the repeaters must be synchronized so that only one repeater transmits
at time. Standard GPS receiver cannot be used without modifications to the firmware
(Bu & al., 2003).
79
7 Local Positioning Systems
This chapter presents examples of local positioning systems based on different signal
technologies and computation methods. The local positioning systems discussed in this
chapter utilize ultrasound pulses, and WLAN and UWB radio signals.
7.1
Ultrasound-Based Positioning Systems
Ultrasound technology provides great ranging accuracy. Because the speed of sound is
relative low (340 m/s), the accuracy requirement for measuring the propagation time is
also low. For example, only 3 ms timing accuracy is needed to measure the distance
within one meter accuracy. In contrast, using RF signal propagating at the speed of
light, 3 ns timing accuracy would be needed.
The TOA method is most often used in ultrasonic positioning. In TOA positioning the
transmit time of the ultrasonic pulse is needed. In many ultrasonic systems, the
originator device transmits a RF pulse and an ultrasonic at the same time. The recipient
device measures the difference between the arrived RF and ultrasonic pulses, which
gives an estimate for the ultrasonic pulse propagation time. The target object is located
at a circle whose center is at the base station and whose radius is the product of the
measured propagation time and the speed of sound. The position of the object can be
solved by the same way as in GPS. Only three measurements are needed for threedimensional positioning.
In spite of high accuracy of the ultrasound technology, it is not the most common
indoor positioning technique. Ultrasound sensors react to ambient ultrasonic pulse and
high-energy sound pulses, which makes ultrasound positioning technique vulnerable in
many indoor environments.
The most well known positioning systems in research community are MIT Cricket
(Priyantha, 2005) and Active Bat (Ward & al., 1997). Both systems obtain the
propagation time of the ultrasonic pulse by measuring the time difference between
incoming RF and ultrasonic pulse. The MIT Cricket system is a user centric system.
The fixed base stations work as transmitters and the mobile node receives the signal.
The Active Bat system is network centric. The mobile node transmits the RF and
80
ultrasonic pulses and the base stations receive the signals. The position of the mobile
node is calculated at the network. Another example of a network-based ultrasound
positioning system is Sonitor ([Cited January 24, 2008]), which is primarily intended
for tracking patients and objects in hospitals. MIT Cricket system and Active Bat
systems are discussed in the following sections.
7.1.1 MIT Cricket System
MIT Cricket is an indoor positioning system based on ultrasound technology. MIT
Cricket provides two kinds of positioning information: Cartesian position coordinates
and symbolic coordinates called space identifiers. Space identifiers are application
specific names associated with spaces such as rooms (Cricket, 2008).
The Cricket system consists of the set of base stations, called beacons, and mobile
devices called listeners whose location needs to be obtained. The beacons are deployed
on walls or ceilings. A listener is attached to a host device, like handheld or laptop
computer. At the system setup, the administrator configures the position coordinates
and the space identifiers of the beacons. Each beacon periodically broadcasts its space
identifier and position coordinates on a RF channel. Each beacon also broadcasts an
ultrasonic pulse at the same time as the RF message.
To infer the distance from the beacon, the listener measures the time difference
between arrived RF and ultrasonic pulses. The listener sends the measured distance
and the beacon position information obtained from the RF message to the host
computer. The host computer calculates the listener position coordinates based on
distances from multiple beacons. The computation algorithm associates also the
listener with the space corresponding to the nearest beacon. The spaces may be
separated by physical boundaries like walls, or they may be virtual. Because
ultrasound does not travel through walls, Cricket can easily demarcate spaces
separated by walls.
Because the position is calculated at the target device, the system administrator cannot
track the users. This is a good issue if privacy is a desired goal. In network centric
ultrasonic system the location is calculated at the network base station, and ability to
track users is thus inherent.
81
7.1.2 Active Bat System
Like MIT Cricket, the Active Bat system developed by AT&T Laboratories at
Cambridge is intended for indoor positioning. Unlike MIT Cricket, Active Bat is
network centric, which means that the objects to be located, called bats, are active. The
bat transmits ultrasound pulse, which is received by the receivers located at fixed
coordinates. Like MIT Cricket the Active Bat system uses radio signal for transmit
time synchronization.
The Active Bat system consists of a bat which is attached to an object to be located
and a set of receivers deployed to the roof of a room. The system also has a base
station which sends a synchronization signal to the receivers and the bats. The base
station is responsible of calculating the position coordinates of the bats (Ward & al.,
1997).
The base station sends simultaneously a timing signal to the fixed ultrasound receivers
and a radio signal to the bat. When the bat receives the radio signal it transmits
ultrasound pulse, which is received by multiple ultrasound receivers at the roof. The
ultrasound receivers measure the time difference between the timing signal and the
ultrasonic pulse and send times of flight of the ultrasound pulses to the base station.
The base station computes the distances between the bat and the receivers and the bat
position coordinates using TOA algorithm (Ward & al., 1997).
7.2
WLAN-Based Positioning Systems
It is often beneficial to utilize the existing communications infrastructure for indoor
positioning. Many offices, hotels and airports have a wireless LAN system installed to
serve the communication needs of people. In addition to wireless communication,
IEEE 802.11b or g WLAN systems can be used for indoor navigation. Most new
laptop computers and many smart phones are equipped with WLAN device to provide
access to the wireless network. Positioning infrastructure can be built for example to
an office building by using only the existing WLAN infrastructure and the user devices
without adding any new hardware to the system. The positioning function is added to
the communication system by adding server software responsible of position
estimation and client software to be installed to the mobile terminal. Low-cost WLAN
tags having the client software can also be attached to people and assets.
Positioning systems utilizing standard WLAN access points and other components use
received signal strength (RSS) as the positioning observable. Usually, the RSS
82
measurements are not used directly to form range estimates. Instead, the position of the
user is estimated by location fingerprinting, where the observed RSS values are
compared to the table of previously measured RSS values. Examples of WLAN-based
positioning systems utilizing the location fingerprinting are Microsoft RADAR (Bahl
& Padmanabhan 2000a, 2000b) and Ekahau (2008).
Also the time difference of arrival (TDOA) can be measured from WLAN signal. If
the TDOA observation is obtained, the position of the object is calculated using
hyperbolic positioning (trilateration). TDOA cannot be measured with standard
WLAN access points, and additional hardware and signal processing is needed to
synchronize the base stations to common time and to estimate the time difference of
arrival.
Cell of origin technique is not suitable for indoor positioning with WLAN because
cells are too wide. Also, the WLAN standard devices keep connections to access point
even if a stronger signal from other access point is available (Rerrer & Kao, 2005).
7.2.1 Aeroscout Positioning System
An example of WLAN-based positioning system using TDOA observables and
hyperbolic positioning is Aeroscout (2008, [Cited January 24, 2008]). Aeroscout
system uses WLAN to locate any standard 802.11b and g device, such as cellular
phones, PDAs and laptop computers. In addition, the Aeroscout system locates battery
powered Aeroscout tags attached to people or assets. Seinäjoki University of Applied
Sciences has used Aeroscout positioning system to track cattle inside a building.
The Aeroscout system is able to locate the object using either TDOA or RSS
measurements. If RSS observable is used, the Aeroscout system is able to utilize
standard WLAN access points as positioning receivers. In this case the existing
WLAN installation can be utilized. However, only the Aeroscout’s location receivers
are able to provide TDOA measurements. In addition, the location receivers have to be
synchronized accurately to common time. TDOA is used for large open indoor
environments and outdoors while RSS may perform better in a tight indoor
environment. RSS positioning is based on trilateration, not on location fingerprinting.
Aeroscout System Architecture
The architecture of Aeroscout positioning system is shown in Figure 18 (Aeroscout,
[Cited January 24, 2008]).
83
Figure 18. Aeroscout system components (Aeroscout, [Cited January 24, 2008])
Aeroscout system’s hardware components are:

IEEE 802.11 wireless local area network

Standard WLAN access points

Aeroscout location receivers

Server computer running the Aeroscout engine

Client WLAN devices (standard WLAN devices or Aeroscout tags)
Aeroscout’s software environment consists of the following components:

Aeroscout engine

Aeroscout mobile view

Third party location-based applications
The Aeroscout tags are active RFID tags using the WLAN standard. The tags enable
the Aeroscout positioning system to locate people and assets otherwise not connected
to a wireless network. Aeroscout T3 tag contains motion and temperature sensors and
two call buttons. Unlike in the Ekahau system, no client software is needed in the
standard WLAN devices.
The location receiver measures the arrival time of standard 802.11b and g messages
and sends that information to the Aeroscout engine for further processing. Each
84
location receiver can process over 300 location measurements per second. Location
receivers can also process presence-based location for areas where there are not
enough receivers available for trilateration (Aeroscout, [Cited January 24, 2008]).
In Aeroscout system, like in all TDOA systems, the location receivers are
synchronized to common time. In Seinäjoki research laboratory installation one of the
location receivers acts as a master. The time difference between the master location
receiver and ordinary location receiver is obtained by subtracting the measured
distance from the theoretically calculated distance between the fixed receivers at
known locations (Aljadeff & Granot, 2005).
Aeroscout Engine processes the information received from the location receivers and
standard WLAN access points. To calculate the location of the user Aeroscout engine
uses TDOA or RSSI method. Aeroscout engine contains positioning server software,
which provides location or presence data to the MobileView software or third party
applications. Third party applications can access the position data through XML API
using TCP/IP sockets.
Aeroscout Exciters provide RFID detection capabilities, using the same tags that can
also be used for positioning. The exciters provide RFID choke point functionality by
triggering transmissions from the tags as they pass through doors, gates and other
choke points (Kolodziej & Hjelm, 2006, p. 260).
7.2.2 Ekahau System
The Ekahau (2008) positioning system is based on IEEE 802.11 wireless LAN
infrastructure. Ekahau system measures the received signal strengths of standard
WLAN signals and computes the location using the location fingerprinting technique.
Standard WLAN access points of any mark can be used and no proprietary
infrastructure is needed. Ekahau system tracks Ekahau WLAN tags, laptops, PDAs,
smart phones and other WLAN enabled devices (Ekahau, [Cited January 24, 2008]).
Examples of applications where Ekahau system has been used are patient and asset
tracking in hospitals, industry and warehouses.
System Architecture
The system architecture of Ekahau Real-Time Location System (RTLS) is shown in
Figure 19 (Ekahau, [Cited January 24, 2008]).
85
Figure 19. Ekahau system components (Ekahau, 2008)
Ekahau system’s hardware components are:

IEEE 802.11 wireless local area network

Server computer running the Ekahau positioning engine (EPE)

Client WLAN devices running the Ekahau client software
Ekahau’s software environment consists of the following components:

Ekahau Positiong Engine (EPE)

Ekahau Client

Ekahau Site Survey
Ekahau Client is a software component that runs on a client device (laptop, smart
phone, PDA, WLAN tag). Ekahau Client reads the signal strength measurement data
from the network adapter and sends that data to the positioning engine. Only tens of
bytes data is transferred per location request (Kolodziej & Hjelm, 2006, p. 237).
Ekahau supports plenty of WLAN adapters. A problem is that different WLAN
adapters measure the signal strengths differently and thus the measurements are not
compatible. Ekahau tries to make different adapter measurements compatible with the
measurements of the adapter used for building the model (Salaur, 2005, p. 35).
Ekahau tags obviously contain the functionality of the client software. There are three
generations of the tags, named as T101, T201 and T301, and the tags can be attached
to people or assets. According to Ekahau (2008), the tags have battery life of up to 5
86
years. The tags have a motion sensor that activates the tag only when the tracked
object is moving. The tags can be managed remotely within the WLAN coverage area.
The remote management automates the configuration changes, battery level
monitoring and tag software updates. The tag contains two buttons, which can be
configured to send alarms or other messages to the applications (Ekahau, [Cited
January 24, 2008]).
Ekahau Positiong Engine (EPE) runs on the server computer. EPE stores the radio
map, reads the signal strength data send by the Ekahau Clients (and tags) and
calculates the clients’ location. User applications can be integrated to EPE via HTTP
and XML-based Application Programming Interface (API) and Software Development
Toolkit (SDK). EPE contains also a web-based management and tracking console,
which can be used for system configuration and remote configuration of the location
tags. The radio map is created during the off-line phase with assistance of Ekahau
Location Survey tool or with Ekahau’s standalone Site Survey application, which is
intended for network planning. (Ekahau, [Cited January 24, 2008].)
7.3
UWB-Based Positioning Systems
UWB is a radio technology which can be used for short-range and high speed
communications by using a large portion of the radio bandwidth. UWB can be used to
refer any radio technology having relative bandwidth larger than 20% or absolute
bandwidth more than 500 MHz as illustrated in Figure 20 (Reed & al., 2008).
Figure 20. FCC definition of UWB. Relative bandwidth (fH-fL)/fC > 20% or total
bandwidth is greater than 500 MHz (Reed & al., 2008).
In UWB technology extremely short duration pulses (sub-nanosecond) are used instead
of continuous waves to transmit information. The short pulse generates directly a very
87
wide bandwidth signal. The spread bandwidth of UWB waveform is generated directly
without individual bit modulation. The conventional spread spectrum waveforms
(direct sequence DSSS or frequency hopping FHSS) typically have 100% duty cycle
and peak and average power levels are equal. In UWB the waveform duty cycles are
typically less than one percent and peak-to-average can be quite large (Reed & al.,
2008).
7.3.1 Advantages of UWB
Wide bandwidth of UWB offers advantages for communications, navigation and radar
applications. In all of these applications the large bandwidth improves reliability as the
signal contains different frequency components. This increases the probability that the
signal will go through in the presence of obstacles or interfering signals. In navigation
and radar applications a large absolute bandwidth provides higher ranging accuracy.
For communications both large relative and large absolute bandwidth decrease the
power spectral density, thus reducing interference to existing narrowband systems
(Gezisi & al., 2005). Extremely wide bandwidth offers obviously high data rate, even
though the duty cycle is low. One important advantage in indoor navigation and
communication is that UWB is relative immune to multipath cancellation effects (Reed
& al., 2008). UWB technology offers also low power consumption with low cost.
UWB transceiver can be implemented nearly all-digital, with minimal RF electronics.
7.3.2 Applications
UWB radars have been developed since the late 1960’s (Barrett, 2000). UWB has been
used for short-range radio communications only in the early 1990’s, but has received
much interest after the FCC allowed the use of unlicensed communications. The first
commercial systems are intended for short range personal area networks (PANs)
having a high data rate (Gezici & al., 2005). Also the first indoor positioning systems
based on UWB are introduced recently (Multispectral, [Cited January 24, 2008]).
UWB offers much higher data rate than Bluetooth or 802.11 for personal area
networking (PAN). Data can be transferred very fast between cellular phones, PDAs,
cameras, MP3 players and laptops. UWB communications can also be integrated into
automotive in-car services and entertainment. Driving directions, music and videos can
be downloaded to car’s navigation and entertainment systems from PDAs or laptops or
road side UWB stations.
88
UWB is suitable for sensor networks as well. Typically, in sensor networks there is no
need for high data rate communication, but the positioning capability is important.
UWB signaling is suitable for sensor networks because it allows centimeter accuracy
and does not suffer from multipath effect. Positioning capability offers new markets
for sensor networks, like in area of logistics, security, manufacturing and health care.
Examples of companies developing UWB technology for positioning applications are
Aetherwire ([Cited January 24, 2008]), Multispectral ([Cited January 24, 2008]) and
Ubisense ([Cited January 24, 2008]). In addition to radar applications, communications
and location finding the ultra-wideband technology can be used to through wall
imaging and medical imaging applications (Reed & al., 2007).
7.3.3 UWB in Positioning
UWB technology provides unique advantages for local positioning systems and indoor
navigation. The use of short pulse RF waveforms provides inherent precision for TOA
or TDOA measurements. The time of arrival of the sub-nanosecond duration pulse can
be accurately determined with a sensitive and high speed receiver. The very wide
bandwidth of the UWB signal increases the probability of some frequency components
penetrating through or going around obstacles. In other words, both high time
resolution and penetration capability make UWB signals suitable for ranging purposes
(Sahinoglu & Gezisi, 2006).
Figure 21. Multipath effect of narrowband signal (left) and ultra-wideband signal.
A unique advantage of the UWB is its ranging performance at indoor environments,
where the main sources of error are the multipath propagation and non-line-of-sight
(NLOS) propagation. At indoor environment multiple copies of transmitted signal with
89
various attenuation levels and time delay arrive at the receiver. In narrowband systems
like GPS the reflecting components may distort the direct path signal and make
accurate timing difficult. In UWB the direct path signal can be distinguished from the
reflections, making the pulse timing easier, as illustrated in Figure 21.
However, the first peak which indicates the shortest path and the best range estimate
may not be the strongest peak. If the direct line-of-sight (LOS) is obstructed, a
reflected signal may provide a stronger peak. Multipath mitigation techniques for
UWB are still needed and they are presented in (Lee & Scholtz, 2002).
In sensor networks and other applications where the signal is used also for positioning,
the data rate is usually relatively low. From regulatory perspective, the UWB
positioning systems having low data rates are allowed to have significantly higher peak
powers than the high data rate systems designed mainly for communications (Fontana
& al., 2003). Thus, a UWB positioning system can operate over significantly longer
distances, as compared to a system intended for high data rate communications. Also
the very low duty cycle waveform needed in sensor network and positioning system
reduces the energy consumption. The sensor network nodes can be equipped with very
small batteries and still have years of operational lifetime.
7.3.4 Standards and Regulations
This section discusses the European and American standardization work for UWB.
UWB Regulations in the United States
The United States Federal Communications Commission (FCC) allowed the use of
unlicensed UWB communications in April 2002 (FCC 02-48, 2002). The devices
intended for communication must operate with their -10 dB bandwidth in the
frequency band 3.1-10.6 GHz. According to FCC rules, ―The equipment must be
designed to ensure that operation can only occur indoors or it consists of hand held
devices that may be employed for such activities as peer-to-peer operation‖. (FCC 0248, 2002, p. 12)
FCC has adopted a very conservative out of band emission mask to address concerns
of companies which make or market indoor electronic equipment. The limits on
unwanted emissions are even more stringent for hand held (outdoor) devices (FCC 0248, 2002, p. 27). The emission masks of indoor and hand held devices is shown in
Figure 22 (Reed & al., 2007).
90
In addition, FCC requires that the hand held devices transmit only in communication
with an associated receiver. The transmitter must stop transmission within 10 seconds
unless it receives an appropriate acknowledgement from the associated receiver, and
the acknowledgement will continue during the transmission at 10 second intervals.
This rule is a restriction for some wireless sensor networks, because two way
communications is required.
Figure 22. FCC Emission masks for indoor and hand held (outdoor) devices (Reed &
al., 2007)
Other UWB applications, such as vehicular radars, through wall imaging systems and
surveillance systems have partly or completely different regulations for frequency
band end emission levels.
UWB Regulations in Europe
During writing this thesis (Autumn 2007) UWB can be used legally only in the United
States, with the exception of Singapore, which allows experimental use. During recent
years Europe has made significant progress towards regulation of generic UWB
applications. The current European UWB regulation is defined in European
Commission (EC) decision 2007/131/EC (EC, 2007) and CEPT decisions
ECC/DEC/(06)04 (2007) and ECC/DEC/(06)12 (2006). Before UWB equipment can
91
be marketed, the national regulations or legislation must be made in member states to
reflect the EC decision.
US FCC regulation allows -41.3 dBm/MHz EIRP emissions in the frequency band 3.110.6 GHz. European regulation allows -41.3 dBm/MHz EIRP emissions only in
frequency band 6-8.5 GHz and with limitations in frequency bands 4.2-4.8 GHz and
3.4-4.8 GHz. The draft of the European spectrum mask is shown in Figure 23
(Europcom, 2007).
Figure 23. ECC draft spectrum mask. The blue solid line shows the emission mask
according to general decision for devices under 10.6 GHz. The blue dashed line
shows the allowed power level if LDC is use. The red solid line shows the allowed
power level if mitigation technique is used (Europcom, 2007).
CEPT decision ECC/DEC/(06)12 (2006) allows -41.3 dBm/MHz emissions in the
frequency band 3.4-4.8 GHz for devices without low duty cycle (LDC) before 31
December 2010. After that only UWB devices implementing LDC will be permitted to
operate at the power level -41.3 dBm/MHz. Until 31 December 2010 UWB devices
are permitted with maximum mean EIRP density of 41.3 dBm/MHz (EC, 2007). After
that appropriate mitigation technique like ―Detect and Avoid‖ DAA must be used. The
motivation to allow the use of 4.2-4.8 GHz band temporarily is to allow FCCcompliant devices to be used in Europe for few years, until new standards are adopted
(Europcom, 2007).
Use of equipment using UWB is not permitted if the device is at fixed outdoor location
or it is connected to a fixed outdoor antenna. The use of UWB in vehicles is not
92
permitted either. According to EC decision (EC, 2007) the potential interference
caused by such uses requires further study.
In addition to the regulations defined in 2007/131/EC (EC, 2007) standardization is
needed. European Telecommunications Standards Institute (ETSI) is defining
standards for positioning systems based on UWB. ETSI standard EN 302 500 is
intended for indoor and EN 302 501 for outdoor location tracking. These standards are
not published yet.
IEEE Standard 802.15.4
IEEE standard 802.15.4 (2007) defines the protocol and compatible interconnection for
data communication devices using low data rate radio frequency transmissions in a
wireless personal area network (WPAN). Other targets in developing the 802.15.4
standard have been multi-year battery life and very low complexity. Low rate personal
area networks (LR-WPANs) are used to transmit information over relative short
distances. Unlike WLANs, connections effected via WPANs involve little or no
infrastructure, which makes WPANs suitable for wireless sensor networks. Other
potential applications are interactive toys, smart badges, remote controls and home
automation. IEEE standard 802.15.4 (2007) enumerates the following features for LRWPAN:

Over the air data rates of 250 kb/s, 100 kb/s, and 20 kb/s

Two addressing modes: 16-bit short and 64-bit IEEE addressing

Support for critical latency devices

CSMA-CA channel access

Automatic network establishment by the coordinator

Fully handshake protocol for transfer reliability

Power management to ensure low power consumption

16 channels in the 2.4 GHz ISM band, 10 channel in the 915 MHz band and
one channel in the 868 MHz band.
The ZigBee specification is based on the IEEE 802.15.4 standard.
IEEE Draft Standard 802.15.4a
The IEEE 802.15.4a Task Group (TG4a) has defined an alternative physical layer
(PHY) standard (IEEE Draft Standard 802.15.4a/D4, 2006) to the 802.15.4 standard.
93
The main interest of TG4a has been in providing the following capabilities over the
existing 802.15.4 standard:

high precision ranging and positioning (1 meter accuracy or better)

higher data rate than 802.15.4

ultra low power consumption

low cost

scalability to different data rates

longer range than in 802.15.4
The IEEE standard 802.15.4a (2006) specifies two optional signaling formats. The first
signaling format, which is based on impulse radio UWB, can use 250-750 MHz,
3.244-4.742 GHz or 5.944-10.234 GHz frequency bands. The other signaling option,
which is based on chirp spread spectrum (CSS), uses the 2.4-2.4835 GHz band. Only
the UWB option contains an optional ranging capability, whereas the CSS signals can
be used only for communication (Sahinoglu & Gezisi, 2006).
Ranging Protocols in 802.15.4a
There are three different ranging protocols defined in the IEEE 802.15.4a standard
(Sahinoglu & Gezisi, 2006):

mandatory two-way ranging protocol

optional symmetric double-sided two-way ranging protocol

optional private ranging protocol
These ranging protocols are illustrated in Figure 24.
The mandatory ranging protocol is based on the simple two-way ranging concept
illustrated in section 4.4.2. The mandatory ranging protocol consists only of the
transmission of the messages D2, A2, D4 and A4 in Figure 24. First, the originator
device sends a range request message D2 and the recipient device replies with the
acknowledgement A2. The recipient also transmits a timestamp message D4. Finally,
the originator sends an acknowledgement, A4, for the timestamp (Sahinoglu & Gezisi,
2006).
The optional symmetric double-sided two-way ranging protocol is illustrated with
messages D2, A2, and D3 in Figure 24. Addition of D3 to the simple double-sided two
protocol implements the symmetric protocol. As explained in section 4.4.2, performing
94
the two-way ranging symmetrically reduces the accuracy requirements of the crystals
in the originator and recipient devices.
Figure 24. Illustration of the ranging protocols supported by the IEEE 802.15.4a
standard (Sahinoglu & Gezisi, 2006)
The optional private ranging protocol makes the UWB ranging more secure against
hostile attacks. The private ranging is implemented by adding the authentication step
consisting of the messages D1 and A1 before the ranging steps, as illustrated in Figure
24. The authentication process and encrypting the contents of the messages are
explained in (Sahinoglu & Gezisi, 2006).
7.4
Fusing Positioning Technologies
There is no single location technology that may be relied upon in all environments to
provide accurate location information. Some technologies are designed to work best
indoors while some technologies have better outdoor accuracy and global availability.
However, there is a need for a location-based service which would provide position
95
information both indoors and outdoors. In other words, the positioning device or
service would change the underlying positioning technology seamlessly when the user
moves between the outdoor and indoor environments. As an example, the user’s
handset uses UWB or WLAN signals indoors for position determination. On the other
hand, when the user walks out of the building the handset switches to use GPS signals.
In addition to combining outdoor and indoor positioning systems, there is also a need
to fuse different indoor positioning technologies. For example, Aeroscout recently
introduced a positioning system which combines the WLAN-based positioning system
with UWB positioning. In general this system relies on WLAN positioning, but certain
areas that call for very high accuracy can be equipped with UWB receivers. When the
target device is in the coverage of UWB network, both WLAN and UWB signals are
utilized in position determination and improved positioning accuracy is achieved.
(Aeroscout, [Cited April 23, 2008].)
Positioning systems provide their information in very different formats and offer
different level of accuracy. Borriello & Deshpande (2002) present a concept called
Universal Location Framework, which fuses location technologies together. In this
concept positioning technologies can be used either independently from each other, or
the measurements obtained from them can be combined with each other. To make the
position technology fusion easier, a layered software engineering model has been
proposed. This model utilizes the common design principles identified in all
positioning systems (Hightower & al., 2002).
96
97
8 Conclusions and Discussion
This chapter discusses the results of the literature research made in this thesis and
presents the conclusions. First, an insight to the markets of the local positioning
systems is given. Second, the accuracy and availability requirements of different
location-based services are discussed. Third, the signal technologies investigated in
this thesis are compared. Finally, the conclusions of the thesis are summarized and
future research topics are suggested.
8.1
Cost and Market Issues
The current local positioning systems are still relative expensive to purchase and
maintain. For example, in Legoland Billund the parents can track the movement of
their children inside the amusement park by using location service based on Aeroscout
TDOA estimation technology. The investment for a location system similar to
Legoland amusement park is between 100 000 € and 150 000 € (InformationWeek,
2004). Also the software-based systems may be expensive. A location fingerprinting
system using WLAN RSS measurements for a hospital may cost 300 000 $ (Helsingin
Sanomat, 2008).
Because the current local positioning systems are quite expensive, the potential
applications are still limited to the markets where relative valuable assets or wealthy
people are tracked. Today, healthcare is the most important market segment of the
indoor positioning systems (Helsingin Sanomat, 2008). Most of the current indoor
positioning systems are used to track people and assets in hospitals, especially in the
United States. USA spends more money per capita on healthcare than any other
nations (Medical News Today, 2005). Because of the huge amount of money in the
healthcare segment, USA produces and consumes over 50% of world’s output medical
technology (US Department of Commerce, 2007). Obviously, the American hospitals
are good business for the location services and positioning equipment vendors.
Investing in the local positioning system may be financially justified also in
amusement parks, because of the huge amount of potential users. There are 1.6 million
visitors annually in Legoland Billund. The investment is paid back within a year if 2 %
of the visitors rent a WLAN tag and pay 5 € rent. However, tracking children with
98
location-based systems at kindergarten or elementary school for safety reasons is
probably still too expensive for the society. Even though human life is irreplaceable
and any investment for the safety can be considered justified. However, in
kindergarten and school environment the safety of the kids is improved better if the
money is spent for hiring new personnel instead of purchasing child tracking
technology.
In addition to the hospitals and amusement parks, there are some other potential
market segments for the local positioning systems. Shopping malls, warehouses and
industrial automation have been mentioned as examples of market areas where the
local positioning technology can be applied (Aeroscout, [Cited January 24, 2008]).
According to (Helsingin Sanomat, 2008), improvement of safety in dangerous
environments by utilizing the position information, is one potential application.
There are also outdoor applications, where the use of the local positioning technology
is beneficial. Examples of these applications are various kinds of systems measuring
the environment. For example, nodes of the wireless sensor networks must be very
inexpensive and energy efficient. On the other hand, wireless sensor networks are used
in military applications and the use of the C/A code receivers is not allowed for US
military (CJCSI 6130.01C, 2003). Use of P(Y) code receivers in sensor network nodes
is too expensive. In some systems, the position of the sensor node is tracked by
measuring the AOA or TDOA of the sensor node transmission at the base stations and
no extra hardware is needed at the sensor node. Using a local positioning system is
financially justified, if a huge number of expendable sensor nodes are being used.
However, it can be anticipated that the cost of the local positioning system will
become smaller in the future, when the amount of sold systems will be increased.
Most of the local positioning systems used today are network centric. If networkcentric systems are intended to locate persons, the privacy issues are raised. When
position of the people is tracked, consent is always required. The markets of the indoor
and outdoor location-based systems will not develop, if people do not have trust in the
privacy of the location information.
8.2
Accuracy and Availability Requirements of Location-Based Services
The driving factor for the development of indoor GPS positioning techniques and
cellular network-based positioning methods has been the United States Federal
Communications Commission (FCC) E911 docket on emergency call positioning in
the USA (Dedes & Dempster, 2005). The FCC-E911 docket mandates that the
99
accuracy requirement for the handset-based positioning is 67% circular error
probability (CERP) within 50 m and 95% CERP within 150 m. For the cellular
network-based positioning the accuracy requirements are 67% CERP within 100 m
and 95% CERP within 300 m.
These accuracy requirements are rather challenging both for the cellular positioning
techniques and for the indoor GPS techniques. According to the tests made in (Singh,
2006a) the accuracy of current HSGPS receivers can be worse than 50 m in many
indoor facilities. Also the accuracy of the cellular network-based positioning systems
is usually worse than 100 m (Syrjärinne, 2001).
In spite of the extensive research on cellular network-based positioning techniques,
these systems are mainly used on applications used by the emergency and law
enforcement authorities. There are only few widely used commercial location-based
services using the cellular network-based positioning technologies. The most important
reason for this is the insufficient accuracy of current technologies. Other reasons are
the lack of privacy due to the network centric approach and fees to be paid to the
telecommunications operators.
The accuracy requirements for the most location-based services are higher than those
set by USA and European emergency call mandates. For vehicle navigation or
positioning people and objects inside buildings 1-10 meters accuracy is usually
required to guarantee correct guidance and information to the user. Positioning
accuracy of 50 meters is not enough to help a user to find a way to a correct address in
urban areas (Syrjärinne, 2001, p. 10) or locate assets in hospitals or warehouses.
Typically, the accuracy level of GPS in good conditions is required for most of the
outdoor location-based services. At indoor applications, granularity of room level or
better is typically required.
In addition to accuracy, the availability of the positioning service is also an important
property for a location-based application. Figure 25 shows the availability and
accuracy of various positioning technologies. The GPS signals are available globally,
but their accuracy starts to diminish at indoor facilities. On the other hand, the cellular
network-based technologies can provide positioning capability for both indoor
environments and very large geographical areas, but the accuracy is not sufficient for
most of the applications. On the other hand, WLAN-based systems provide moderate
accuracy in places where the access point network is dense, such as inside office
buildings, but the coverage is limited to the campus areas or city centers. UWB
technology would provide the best accuracy, but its outdoor use is limited.
100
Figure 25. Comparison of positioning technologies in terms of accuracy and
availability. Accuracy and availability numbers for GPS and cellular network-based
systems are obtained from (Syrjärinne, 2001, p. 46).
Figure 25 shows that there is no positioning technology which would work both inside
buildings and in large geographic areas and provide 1-10 meters positioning accuracy.
When accuracy and availability are considered, there is a market segment for local
position systems in environments where the location finding is performed at least
partially indoors.
8.3
Comparison of the Signal Technologies
In this thesis the most common signal technologies were investigated. These signal
technologies and their characteristics are summarized in Table 2.
The most promising positioning technologies used in the local positioning systems use
radio frequency (RF) electro-magnetic waves. The RF signal is used for the
positioning from the same reasons as it is used for communications. The RF signal can
penetrate walls and it has a wider communications bandwidth than ultrasound or
infrared signal. Most of the current and emerging indoor positioning technologies are
based on WLAN or UWB signals or proximity measurements using RFID technology.
All the other signal technologies have significant drawbacks when applied to local
positioning systems. The infrared-based systems suffer from short range transmitters
101
and the huge amount of additional hardware (Rerrer & Kao, 2005). Also the
ultrasound-based systems have drawbacks. Ultrasound pulses do not penetrate the
walls and the pulses are easily interfered. In addition, a lot of hardware is needed to
build the positioning infrastructure. Sensors used in dead-reckoning applications, such
as accelerometer, gyroscope and magnetic compass, are practical in applications where
the calibration to the local frame of reference is not a problem. In some applications,
utilizing the motion tracking sensors may be justified when they are used in
conjunction with radio frequency range observables.
Computer vision systems can determine the location of an object with sub-centimeter
accuracy from analysis of video images. They are designed to capture position well in
a single room immersive environment. However, computer vision systems are not
usually designed to be scalable, not even to a building level.
Table 2.
Summary of positioning technologies. Accuracies are obtained from
(Muthukrihnan & al., 2005; Hightower & Borriello, 2001)
Technology
Accuracy
Availability
Location estimation
Examples
Infrared-based
5-10 m
Indoors/
building
Proximity
Active Badge
Vision-based
1-100 cm
Indoors/
room
Scene analysis
Easy Living
Ultrasound-based
1-10 cm
Indoors/
building
TOA trilateration, proximity
Active Bat, MIT
Cricket
Satellite-based
5-10 m
(20-50 m
indoors)
Global
TOA/
trilateration
GPS,
Galileo,
GLONASS
Cellular networkbased
50-300 m
Rural
RSS, AOA, TDOA/
proximity, triangulation
Cellular phone
localization
WLAN-based
2-100 m
Campus/
indoors/
outdoors
TDOA lateration,
location fingerprinting
Aeroscout,
Ekahau
UWB-based
6-10 cm
Indoors/
building
TOA/TDOA trilateration
Ubisense
8.4
Conclusions
In this thesis various technologies for local positioning systems were investigated.
Table 3 summarizes the merits and drawbacks of the positioning technologies
discussed in this thesis. The positioning technologies based on radio frequency signals
were found most promising.
102
Table 3.
Merits and drawbacks of the positioning technologies in local positioning
system context (Muthukrishnan & al., 2005).
Technology
Merits
Drawbacks/remarks
Infrared-based
Compact
Range is typically less than 5 m
Low power
Restriction to line of sight conditions
Simple
Unusable in direct sunlight
High accuracy
Restriction to line of sight conditions
Vision-based
Unusable in direct sunlight
Ultrasound-based
Satellite-based
(HSGPS, AGPS)
Measurement of TOA is easy (sound
propagates slowly)
Range is less than 10 m
High accuracy
Vulnerable due to external
interference
Simple
Multipath
No need to build the positioning
infrastructure
Low accuracy and availability at
indoor facilities
Global scale
Mature technology
Cellular networkbased
Existing infrastructure can be used for
positioning and communications
Low accuracy
Operator fees
Very wide scale
WLAN-based
Moderate accuracy
Range is between 50 and 100 m
High bandwidth for communications
Multipath is a problem in TDOA
systems
Existing communications infrastructure
can be used for positioning
Based on mature technology
UWB-based
Radio map has to be calibrated in LF
systems
High accuracy
Not mature technology yet
Less affected by multipath
Smaller range than in WLAN
Currently, WLAN-based systems are very popular in indoor positioning. It has been
estimated that the market share of the WLAN-based systems will be at least one third
of the indoor positioning markets (Helsingin Sanomat, 2008). One of the most
important advantages of WLAN-based positioning techniques is that WLAN
communications technology is mature and well-developed. WLAN devices are
standard and rather inexpensive. Existing infrastructure built for communications
purposes can be used also for positioning purposes, if the system uses RSS
measurements for location estimation. Thus no deployment of external hardware is
needed. Also the existing communication channel can be used for delivering the
measurement data and position solutions between the users and network servers.
There are two commercial positioning technologies based on the WLAN signals. The
other systems measure the RSS and estimate the location of the mobile device by the
103
location fingerprinting method (Ekahau, [Cited January 24, 2008]), while the other
systems measure the TDOA and compute location by trilateration (Aeroscout, [Cited
January 24, 2008]). The RSS-based systems can utilize the existing wireless
communications system built using standard WLAN equipment. However, these
systems based on location fingerprinting have the drawback that the radio map has to
be built before use by measuring the signal strengths from the whole target area. On
the other hand, the offline phase calibration is not needed in the TDOA-based systems,
but special receivers capable of TDOA measurements are needed. Also, the base
stations have to be synchronized precisely in time. The TDOA approach works well in
open areas where the dominant direct path (DDP) channel condition is typical. On the
other hand, in heavily multipath environments where the undetected direct path (UDP)
channel condition is common, the location fingerprinting approach performs better
than TDOA approach (Hatami & al., 2006). Thus, the location fingerprinting approach
may be a better choice for office and hospital environment and TDOA approach for
open areas, such as sport halls. Neither of these WLAN-based techniques is feasible, if
better than 1–2 meters accuracy is required.
When the accuracy is important UWB is the most interesting indoor positioning
technology. UWB provides a great ranging accuracy due to the very short pulse from
which the arrival time of the signal is obtained. In addition, UWB is less affected by
the multipath than WLAN or indoor GPS. On the other hand, limited range of the
signal is a problem of current UWB. Opposite to WLAN, UWB is not a mature
technology yet. One of the main problems in UWB technology is the complexity of the
current arrival time measurement schemes due to the high sampling rate needed. A lot
of development will be still needed before competitive UWB-based indoor positioning
systems appear to the market. USA has recently approved using of UWB for public
applications, while in Europe the legislation and standardization are not yet ready.
UWB transceivers are still quite expensive, but their prices may come down when the
volumes of manufactured components rise.
Obviously GPS (or another GNSS) is the best candidate for positioning in outdoor
environment. Use of GPS should be considered also in combined indoor and outdoor
positioning applications where only building-level accuracy is needed when the user is
indoors. Thus, high sensitivity GPS receiver with assistance information is a good
choice for cellular phones. GPS is also the best technology for many wireless sensor
applications measuring the environmental parameters outdoors. However, if the
number of sensor units is very high, deploying a local position system may be
beneficial.
104
8.5
Future Research
The market of the local positioning systems is still quite small, mainly because the
purchasing prices and the maintenance costs of these systems are relatively high. In
addition, the current local positioning systems should be more accurate, more scalable
and more reliable before they could be used more widely. A lot of research is still
needed to improve the existing positioning technologies and to develop new ones,
before the consumer and industrial markets will open for the local positioning systems.
Two potential future research topics are proposed in this section. First, utilization of
the redundant measurements in location fingerprinting method is discussed. Second,
the importance of fusing local positioning techniques with GPS is emphasized.
Improvements to the Location Fingerprinting Method
According to Kaemarungsi (2005) increasing the number of access points in location
fingerprinting method improves the probability of returning the correct location. This
is intuitive and it holds also to the navigation systems using TOA or AOA
measurements. Increasing the number of base stations improves the geometry.
Additionally, the extra observations make the effect of noisy or incorrect observations
smaller.
Not only has the number of access points affected to the performance, but also their
mutual locations. If the access points are far away in the same direction from the
research area, the fingerprints of the points in this area are probably close to each
other. In spite of that the concept of dilution of precision (DOP) cannot be applied
directly to the location fingerprinting, the geometry of the access points is important.
This has not yet been discussed in the research papers. Intuitively, the DOP value
should be calculated for each point in the research area while examining the optimal
locations of the access points.
Another issue related to the number of access points and geometry is exclusion of the
faulty measurements. In GPS redundant measurements can be utilized to detect and
exclude the faulty measurements from the navigation solution. RAIM algorithm is able
to detect the presence of a faulty measurement if there are at least five measurements
(one measurement more than unknown variables). If there are at least six
measurements, one faulty measurement can be identified and isolated from the solution
(Kaplan, 1996, p. 306). RAIM algorithm is based on linear algebra and thus not
applicable in location fingerprinting. In location fingerprinting increasing number of
105
access points improves the redundancy, which should enable to detect and isolate
faulty measurements. However, no papers were found where redundancy is utilized in
excluding the faulty measurements. Utilizing the redundant measurements and
applying the geometry (DOP) concept to the location fingerprinting method are
suggested for a future research topic.
Fusing Positioning Technologies
None of the positioning technologies discussed above would have optimal
performance for all possible applications and environments. Some technologies are
designed to work best indoors while some technologies have better outdoor accuracy
and global availability. However, there is a need for a location-based service which
would provide position information both indoors and outdoors. Besides this,
integrating different indoor positioning technologies together, such as WLAN and
UWB, would also open new interesting markets.
A service or a device which integrates GPS and some indoor positioning technology
together would have the highest market potential. Obviously, cellular phone would be
the most interesting user terminal for such positioning service. The important question
is which indoor positioning technology should be integrated with GPS. Based on the
conclusions above, the most potential alternative technologies would be based on
UWB or WLAN.
It is obvious that the indoor positioning technology to be integrated with satellite
navigation should be based on open standards before it can be widely accepted in
consumer and industrial markets. In practice this means that the base stations of the
indoor positioning systems must support mobile terminals of different vendors and
vice-versa. In other words, there is a similar need for standardization of the
communication and ranging protocols in the indoor part of the fused positioning
system as in cellular network positioning systems and in GPS. In addition, it is
favorable that also the indoor positioning part of the fused system would be based on
self positioning.
In spite of that the location fingerprinting of WLAN signals is currently the most
potential technology used, it is more probable that the indoor component of the fused
system will be based on TDOA or two way TOA schemas. Opposite to location
fingerprinting, the TDOA and TOA systems can be designed as self-positioning
systems. In the self-positioning approach the mobile terminal can make the decision
whether to use GPS signals or indoor positioning system signals or both.
106
At the moment it is not clear which local positioning technique would be the best
alternative for the indoor component of the fused positioning system. WLAN is
designed for communications purposes and there are no standards defined for ranging
purposes. In principle it would be possible to define ranging standard also for WLAN.
However, arrival time measurements of WLAN signals suffer from multipath effect in
many indoor conditions. On the other hand, UWB is more accurate in multipath
environment and there is already IEEE draft standard 802.15.4a for high precision
ranging and positioning. It is a drawback for UWB is that it has smaller range than
WLAN and it is intended rather for personal area than local area communications.
Which local positioning technology will be integrated with global positioning
technologies, remains to be seen in the future.
107
9 Summary
Markets of the location-based services, such as map-based guidance and navigation,
intelligent transport services and tracking of people and valuable assets, are growing
rapidly. This creates a promising area to be investigated further. Today most of these
location-based services depend on GPS as a position data provider, but there is also a
growing interest in positioning applications which are independent from GPS or other
satellite-based navigation technologies. These positioning systems, which are designed
to operate inside a limited geographical area, are called local positioning systems.
This research identified and analyzed state-of-the-art techniques for local positioning
systems. The research also aimed to match the characteristics of the identified local
positioning techniques to the needs of various location-based applications. At this
stage it can be concluded that no single technology was found, which would have
provided optimal performance for all possible applications and environments.
However, there clearly are a few promising methods which outperform certain others,
depending on what the purpose of usage is.
The most widely used local positioning systems today are based on the use of WLAN
signals. There are two WLAN-based methods that can be recommended – depending
on where it is used. The technology based on location fingerprinting method was found
the most promising technology in heavily multipath environment, such as office
buildings. On the other hand, the technology based on measuring the WLAN signal
propagation time performs better in large open areas, such as sport halls. Neither of
these WLAN-based technologies is feasible, if better than 1–2 meters accuracy is
required.
When it comes to accuracy instead, different techniques are recommended to be used,
some of which are not perfect but will do. Based on this study, utilizing UWB signals
for positioning would be an interesting alternative, especially when accuracy is
important. UWB provides a great accuracy due to the very short pulse, from which the
arrival time of the signal is obtained. However, opposite to WLAN, UWB is not a
perfectly mature technology yet and the legislation process of UWB signals is still
going on in Europe. As it develops, in the future UWB should be taken into account as
a reasonably useful option in indoor positioning.
108
If only building level accuracy is required at indoor environment, use of the assisted
GPS is recommended. However, due to the multipath problems present in narrowband
signal TOA estimation the satellite navigation systems cannot compete with local
positioning technologies in navigation or asset tracking applications inside the
buildings.
Several other positioning technologies have also been identified, which are not based
on using radio frequency electro-magnetic waves like WLAN and UWB-based
systems. For example in computer vision systems the location of an object can be
derived with sub-centimeter accuracy from analysis of video images. It is obvious that
this kind of technology would be most suitable for such applications where optical line
of sight between the video camera and the target object can be guaranteed. One of
considerable technologies is dead-reckoning method, which can only be recommended
in applications where the calibration to the local frame of reference is not a problem.
Compared to computer vision and dead-reckoning systems, the RF-based local
positioning systems are still more popular because of the better scalability and
coverage.
Despite the significant research in the field of positioning, there are still numerous big
challenges to be solved in local positioning systems and indoor navigation. Based on
this study it can be anticipated that one of the most important research topics in the
future would be fusing indoor positioning techniques with the satellite-based
navigation systems. There will be remarkable market opportunities for a locationbased service, which can switch between the positioning technologies when the user
moves between outdoor and indoor environments. In this research it is considered that
the most probable technology to be integrated with GPS will be based on ranging
approach of UWB or WLAN signals. Finally, another interesting future research topic
will be better utilization of redundant measurements in location fingerprinting systems.
109
References
Aatique, M. 1997. Evaluation of TDOA Techniques for Position Location in CDMA
Systems. Master's Thesis, Virginia Polytechnic Institute and State University.
Aeroscout, Inc. [Online]. [Cited January 24, 2008]. Available at:
http://www.aeroscout.com
Aeroscout, Inc. 2008. Aeroscout and Time Domain Introduce the Industry's First
Combined WiFi / Ultra Wideband Tracking System [Online]. [Cited April 23, 2008].
Available at: http://www.aeroscout.com/viewItem.asp?type=press&itemId=65
Aether Wire & Location, Inc [Online]. [Cited January 24, 2008]. Available at:
http://www.aetherwire.com
Aitenbichler, E. & Muhlhauser, M. 2003. An IR Local Positioning System for Smart
Items and Devices. Proceedings of the 23rd IEEE International Conference on
Distributed Computing Systems Workshops (IWSAWC03), Los Alamitos, CA, pp.
334–339.
Aljadeff, D. & Granot, Y. 2005. Patent No. 6,968,194. US Patent.
Ascension Technology Corporation [Online]. [Cited January 24, 2008]. Available at:
http://www.ascension-tech.com
Bahl, P. & Padmanabhan, V. 2000. Enhancements to the RADAR User Location and
Tracking System. Microsoft Corporation.
Bahl, P. & Padmanabhan, V. 2000. RADAR: An In-Building RF-based User Location
and Tracking System. Proceedings of 19th Annual Joint Conference of the IEEE
Computer and Communications Societies (INFOCOM '00), Tel Aviv, Israel. pp.
775–784.
Barrett, W. 2000. History of Ultra WideBand (UWB) Radar & Communications:
Pioneers and Innovators. Proc. Progress in Electromagnetics Symposium 2000
(PIERS2000), Cambridge, MA, 29 p.
110
Baumann, S., Collomb, F., Dien, H., Pilloni, P. & Casal, C. 2002. Implementation
Options for Enhanced 112 Emergency Services. Technical report, Information
Society Technologies, IST-1999-14093 LOCUS, Deliverable D3.
Borriello, G. & Deshpande, N. 2002. Location-Aware Computing: Creating Innovative
and Profitable Applications and Services. Intel Developer Update Magazine, pp. 1–
6.
Brännström, F. 2002. Positioning Techniques Alternative to GPS. Master's Thesis,
Luleå University of Technology, Sweden.
Bu, S.-C., Choi, J.-H., Jee, G.-I. & Kim, H. 2003. An Indoor Positioning Using GPS
Repeater. Proceedings of the 16th International Technical Meeting of the Satellite
Division of the Institute of Navigation ION GPS/GNSS 2003, Portland, Oregon.
pp. 1129–1134.
Castro, P., Chiu, P., Kremenek, T. & Muntz, R. 2001. A Probabilistic Room Location
Service for Wireless Networked Environments. Proceedings of Ubicom 2001,
Atlanta, Georgia. pp. 18–34.
CDG. 2000. CDG Test Plan Document for Location Determination Technologies
Evaluation. CDMA Development Group (CDG) Document .
Chan, Y. & Ho, K. 1994. A Simple and Efficient Estimator for Hyperbolic Location.
IEEE Transactions on Signal Processing, 42 (8), pp. 1905–1915.
CISCO Systems Inc. 2006. Wi-Fi Location-Based Services — Design and Deployment
Considerations [Online]. [Cited January 24, 2008]. Available at:
http://www.cisco.com/univercd/cc/td/doc/solution/wifidesi.pdf
CJCSI 6130.01C. 2003. 2003 CJCS Master Positioning, Navigation, and Timing Plan.
US Joint Chiefs of Staff.
Cricket. [Online]. The Cricket Indoor Location System. [Cited January 24, 2008].
Available at: http://cricket.csail.mit.edu
Dedes, G. & Dempster, A. 2005. Indoor GPS: Positioning Challenges and
Opportunities. IEEE 62nd Semiannual Vehicular Technology Conference. Dallas,
TX, pp. 412–415.
111
Dempster, A. 2006. Dilution of Precision in Angle-of-Arrival Positioning Systems.
Electronics Letters, 42 (5), pp. 291–292.
Dickerson, E., Arndt, D. & Ni, J. 2005. UWB Tracking System Design with TDOA
Algorithm for Space Applications [Online]. [Cited January 24, 2008]. Available at:
http://www.isso.uh.edu/publications/A2005/2005_019_dickerson.htm
Diggelen, F. & Abraham, C. 2001. Indoor GPS Technology [Online].
[Cited January 24, 2008]. Available at:
http://www.globallocate.com/SEMICONDUCTORS/Semi_Libr_Piece/IndoorGPS
Technology.pdf
Du, H.-J. & Lee, J. 2004. Simulation of Multi-Platform Geolocation using a Hybrid
TDOA/AOA Method. Technical report, Ottawa, Canada: Defence T&D Canada.
Dumont, L., Fattouche, M. & Morrison, G. 1994. Super-Resolution of Discrete
Arrivals in a Spread Spectrum System. Wireless 94 Conference, Calgary, Canada,
pp. 709–728.
ECC/DEC/(06)04. 2007. ECC Decision of 24 March 2006 amended 6 July 2007 at
Constanta on the harmonised conditions for devices using Ultra-Wideband (UWB)
technology in bands below 10.6 GHz.
ECC/DEC/(06)12. 2006. ECC Decision of 1 December 2006 on the harmonised
conditions for devices using Ultra-Wideband (UWB) technology with Low Duty
Cycle (LDC) in the frequency band 3.4–4.8 GHz.
Eissfeller, B., Gänsch, D., Müller, S. & Teuber, A. 2004. Indoor Positioning Using
Wireless LAN Radio Signals. Proceedings of ION-GNSS 2004, Long Beach, CA,
pp. 1936–1947.
Ekahau, Inc. [Online]. [Cited January 24, 2008]. Available at: http://www.ekahau.com/
EUROPCOM. 2004. Emergency Ultrawideband RadiO for Positioning and
COMmunications, Summary of Regulatory Position (Issue 3) [Online]. [Cited
January 24, 2008]. Available at: http://cordis.europa.eu/fp7/ict/sustainablegrowth/previous-fp6_en.html
European Commission. 2007. On allowing the Use of the Radio Spectrum for
Equipment Using Ultra-Wideband Technology in a Harmonized Manner in the
Community. Official Journal of the European Union, L55/33, 2007/131/EC .
112
Fang, B. 1990. Simple Solutions for Hyperbolic and Related Fixes. IEEE Transactions
on Aerospace and Electronic Systems, 26 (5), pp. 748–753.
FCC. 2002. Revision of Part 15 of the Commission’s Rules Regarding Ultra-Wideband
Transmission Systems. First Report and Order. Washington DC: FCC 02-48.
Fikouras, I., Peters, O., Wunram, M., Fikouras, N. & Görg, C. 2001. Positioning
Information in Service Discovery Protocols for Integrated Network Platforms
[Online]. [Cited January 24, 2008]. Available at: http://www.istnomad.net/publications.php.
Fontana, R., Richley, E. & Barney, J. 2003. Commercialization of an Ultra-Wideband
Precision Asset Location System. Proc. IEEE Conference on Ultra Wideband
Systems and Technologies, Reston, VA, pp. 369–373.
Gadzheva, M. 2007. Privacy Concerns Pertaining to Location-Based Services.
International Journal of Intercultural Information Management, 1, pp. 49–57.
Gezici, S., Tian, Z., Giannakis, G., Kobayashi, H., Molisch, A., Poor, H., et al. 2005.
Localization via Ultra-Wideband Radios: A Look at Positioning Aspects for Future
Sensor Networks. IEEE Signal Processing Magazine, 22 (4), 70–84.
Hach, R. 2005. Symmetric Double Sided Two-way Ranging [Online]. [Cited January
24, 2008]. Available at: http://www.ieee802.org/15/pub/05/15-05-0334-00-004asymetric-double-sided-two-way-ranging.ppt
Hatami, A., Pahlavan, K., Heidari, M. & Akgul, F. 2006. On RSS and TOA based
Indoor Geolocation − A Comparative Performance Evaluation. Proceedings of
IEEE Wireless Communications and Networking Conference, Las Vegas, NV, pp.
2267–2272.
Helsingin Sanomat. 2008, January 5. Positioning is Useful also at Indoor Environment.
(J. Raivio, Ed.). In Finnish.
Hightower, J. & Borriello, G. 2001. Location Systems for Ubiquitous Computing.
IEEE Computer , 34 (8), pp. 57–66.
Hightower, J., Brumitt, B. & Borriello, G. 2002. The Location Stack: A Layered
Model for Location in Ubiquitous Computing. Proceedings of the 4th IEEE
Workshop on Mobile Computing Systems & Applications (WMCSA 2002),
Callicoon, NY, pp. 22–28.
113
IEEE Draft Standard 802.15.4a/D4. 2006. Part 15.4: Wireless Medium Access Control
(MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal
Area Neworks (WPANs), Amendment 1: Add Alternate PHYs.
IEEE Standard 802.15.4. 2007. Wireless Medium Access Control (MAC) and Physical
Layer (PHY) Specifications for Low-Rate Wireless Personal Area Neworks
(WPANs).
InformationWeek. 2004. Legoland Uses Wireless And RFID For Child Security
[Online]. [Cited January 24, 2008]. Available at:
http://www.informationweek.com/showArticle.jhtml?articleID=19202099
Kaemarungsi, K. 2005. Design of Indoor Positioning Systems Based on Location
Fingerprinting Technique. Doctoral dissertation, University of Pittsburgh,
Pittsburgh.
Kaemarungsi, K. & Krishnamurthy, P. 2004. Modeling of Indoor Positioning Systems
Based on Location Fingerprinting. Proceedings of 23rd Annual Joint Conference of
the IEEE Computer and Communications Societies (INFOCOM '04), Hong Kong.
pp. 1012–1022.
Kaplan, E. 1996. Understanding GPS: Principles and Applications. Artech House, Inc.
Kitasuka, T., Nakanishi, T. & Fukuda, A. 2005. Design of WiPS: WLAN-Based
Indoor Positioning System. Third International Conference on Information
Technology and Applications (ICITA 2005). pp. 346–349.
Kolodziej, K. & Hjelm, J. 2006. Local Positioning Systems: LBS Applications and
Services. Boca Raton, Florida, CRC Press.
Lachapelle, G. 2004. GNSS Indoor Location Technologies. Journal of Positioning
Systems, pp. 2–11.
Ladd, A., Bekris, K., Marceau, G., Rudys, A., Kavraki, L. & Wallach, D. 2002.
Robotics-Based Location Sensing Using Wireless Ethernet. Proceedings of ACM
International Conference on Mobile Computing and Networking (MOBICOM'02),
Atlanta, Georgia. pp. 227–238.
Lee, J. & Scholtz, A. 2002. Ranging in a Dense Multipath Environments Using UWB
Ranging. IEEE Journal on Selected Areas in Communications, 20 (9), pp. 1677–
1683.
114
Li, B., Dempster, A., Rizos, C. & Barnes, J. 2005. Hybrid Method for Localization
Using WLAN. Proceedings of the Spatial Science Institute Conference: Spatial
Intelligence, Innovation and Praxis .
Li, X. 2003. Super-Resolution TOA Estimation with Diversity Techniques for Indoor
Geolocation Applications. Doctoral dissertation, Worcester Polytechnic Institute,
Worcester, MA.
Medical News Today. 2005. USA Spends More Per Capita on Health Care Than Other
Nations [Online]. [Cited January 24, 2008]. Available at:
http://www.medicalnewstoday.com/articles/27348.php
Microsoft Research. 2008. Easy Living [Online]. [Cited January 24, 2008]. Available
at: http://research.microsoft.com/easyliving
Mizusawa, G. 1996. Performance of Hyperbolic Position Location Techniques for
Code Division Multiple Access. Virginia Polytechnic Institute and State
University.
Multispectral Solutions, Inc. [Online].[Cited January 24, 2008]. Available at
http://www.multispectral.com/
Muthukrishnan, K., Lijding, M. & Havinga, P. 2005. Towards Smart Surroundings:
Enabling Techniques and Technologies for Localization. First Intenational
Workshop on Location- and Context-Awareness (LoCA), Oberpfaffenhofen,
Germany, Springer-Verlag. pp. 350–362.
Nokia. 2007. Nokia to acquire NAVTEQ [Online]. [Cited January 24, 2008]. Available
at: http://www.nokia.com/A4136001?newsid=1157198
Pahlavan, K., Krishnamurthy, P. & Beneat, J. 1998. Wideband Radio Propagation
Modeling for Indoor Geolocation Applications. IEEE Communications Magazine,
pp. 60–65.
Pahlavan, K., Li, X. & Mäkelä, J.-P. 2002. Indoor Geolocation Science and
Technology. IEEE Communications Magazine. pp. 112–118.
Priyantha, N. 2005. The Cricket Indoor Location System. Doctoral dissertation.
Massachusetts Institute of Technology.
115
Ramirez-Mireles, F. 2001. On the Performance of Ultra-Wideband Signals in Gaussian
Noise and Dense Multipath. IEEE Transactions on Vehicular Tecnology, 50 (1),
pp. 244–249.
Reed, U., Buehrer, M. & Ha, D. 2008. Introduction to UWB: Impulse Radio for Radar
and Wireless Communications [Online]. [Cited January 24, 2008]. Available at:
http:// www.mprg.org
Rerrer, U. & Kao, O. 2005. Suitability of Positioning Techniques for Location-based
Services in Wireless LANs. Proceedings of the Workshop on Positioning,
Navigation and Communication (WPNC 2005), Hannover, Germany, pp. 51–56.
Rintala, M. 2007. UWB Frequency Allocation [Online]. [Cited January 24, 2008].
Avaialble at: http://www.ficora.fi/attachments/suomi_R_Y/5n8IuJWNH/Files/CurrentFile/RTP_Rintala.PPT. In Finnish.
Sahinoglu, Z. & Gezisi, S. 2006. Ranging in the IEEE 802.15.4a Standard. IEEE
Wireless and Microwave Technology Conference (WAMICON), Clearwater
Beach, FL, pp. 1-5.
Salaur, L. 2005. Building a Commodity Location-based Service at Botanic Garden of
University of Freiburg. Master's Thesis, University of Freiburg, Germany.
Saukoski, M. 2008. System and Circuit Design for a Capacitive MEMS Gyroscope.
Doctoral Dissertation, Helsinki University of Technology, Finland.
Shin, D. & Sung, T. 2002. Comparisons of Error Characteristics between TOA and
TDOA Positioning. IEEE Transactions on Aerospace Electronic Systems, 38 (1),
pp. 307–311.
Simojoki, S. 2003. Location-Based Services and General Privacy and Data Protection
Principles. Project report, Helsinki Institute of Information Technology (HIIT),
Finland.
Singh, S. 2006. Evaluation of Assisted GPS (AGPS) Performance using Simulator and
Field Tests. Proceedings of ION GNSS 2006, Fort Worth, TX, pp. 2956-2967.
Singh, S. 2006. Comparison of Assisted GPS and High Sensitivity GPS in Weak
Signal Conditions. Master's Thesis, University of Calgary, Canada.
116
Small, J., Smailagic, A. & Siewiorek, D. 2000. Determining User Location For
Context Aware Computing Through the Use of a Wireless LAN Infrastructure
[Online]. [Cited January 24, 2008]. Available at: http://www.cs.cmu.edu/~aura/docdir/small00.pdf
Sonitor Technologies, Inc. [Online]. [Cited January 24, 2008]. Available at:
http://www.sonitor.com
Syrjärinne, J. 2001. Studies of Modern Techniques for Personal Positioning. Doctoral
Dissertation, Tampere University of Technology, Finland.
Szewczyk, R. 2004. UWB: Technology and Implications for Sensor Networks
[Online]. [Cited January 24, 2008]. Available at: www.cs.berkeley.edu/~binetude/NEST/UWB.ppt
Ubisense Ltd. [Online]. [Cited January 24, 2008]. Available at:
http://www.ubisense.net/
US Department of Commerce. 2007. Investing in World’s Largest Market for Medical
Technology – the USA [Online]. [Cited January 24, 2008]. Available at:
http://www.buyusa.gov/germany/en/medicainvest.pdf
Want, R., Hopper, A., Falcao, V. & Gibbons, J. 1992. The Active Badge Location
System. ACM Transactions on Information Systems, 10 (1), pp. 91–102.
Ward, A., Jones, A. & Hopper, A. 1997. A New Location Technique for the Active
Office. IEEE Personal Communications, 4 (5), pp. 35–41.
Vidal, J., Cabrera, M., Játiva, R., Nájar, M., Pagès, A. & Simon, C. 2001. Location
Based Services Performance Evaluation, IST-1999-10322 SATURN. Technical
report, Information Society Technologies, European Commission.
Wikipedia. Inertial Navigation System [Online]. [Cited January 24, 2008]. Available
at: http://en.wikipedia.org/wiki/Inertial_guidance_system
Zeimpekis, V., Giaglis, G. & Lekaros, G. 2003. A Taxonomy of Indoor and Outdoor
Positioning Techniques for Mobile Location Services. Journal of ACM SIGecom
Exchanges, 3 (4), pp. 19–27.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising