Indoor Positioning at Arlanda Airport AHMAD AL RIFAI Master of Science Thesis

Indoor Positioning at Arlanda Airport AHMAD AL RIFAI Master of Science Thesis
Indoor Positioning at Arlanda Airport
AHMAD AL RIFAI
Master of Science Thesis
Stockholm, Sweden 2009
Indoor Positioning at Arlanda Airport
AHMAD AL RIFAI
Master of Science Thesis performed at
the Radio Communication Systems Group, KTH.
June 2009
Examiner: Professor Ben Slimane
KTH School of Information and Communications Technology (ICT)
Radio Communication Systems (RCS)
TRITA-ICT-EX-2009:75
c Ahmad Al Rifai, June 2009
Tryck: Universitetsservice AB
Abstract
Recent years have witnessed remarkable developements in wireless
positioning systems to satisfy the need of the market for real-time services. At Arlanda airport in Stockholm, LFV - department of research
and developement wanted to invest in an indoor positioning system
to deliver services for customers at the correct time and correct place.
In this thesis, three different technologies, WLAN, Bluetooth, and
RFID and their combination are investigated for this purpose. Several
approaches are considered and two searching algorithms are compared,
namely Trilateration and RF fingerprinting. The proposed approaches
should rely on an existing WLAN infrastructure which is already deployed at the airport. The performances of the different considered
solutions in the aforementioned approaches are quantified by means
of simulations.
This thesis work has shown that RF fingerprinting provide more
accurate results than Trilateration algorithm especially in indoor environments, and that infrastructures with a combination of WLAN
and Bluetooth technologies result in lower average error if compared
to infrastructures that adopt only WLAN.
2
Acknowledgements
I owe my deepest gratitude to my supervisor at KTH Luca Stabellini for
his valuable comments and guidance throughout my work. My great thanks
also go to my examiner Ben Slimane who encouraged me for this work and
for his comments as well. I can not forget my friends for all the great times
we had throughout these years.
I am indebted for all my acheivments so far to my family; my sisters, my
father and especially my mother for their endless love, support, and encouragement.
My thanks go as well to Fritjof Andersson and Christine Hiller at LFV.
A heartful thank goes to my fiance Maha who stood by my side all the
time. Your ambition has always pushed me forward.
Contents
1 Introduction
1.1 Background . . . . .
1.2 Related Work . . . .
1.3 Purpose of the Thesis
1.4 Problem Formulation
1.5 Thesis outline . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2 Technologies Involved
2.1 Wireless Local Area Network, WLAN . . . . . .
2.1.1 802.11 Standard . . . . . . . . . . . . . .
2.2 RFID . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Introduction . . . . . . . . . . . . . . . .
2.2.2 Frequency bands and coverage . . . . . .
2.3 Bluetooth . . . . . . . . . . . . . . . . . . . . .
2.3.1 Bluetooth Standard - Physical Layer . .
2.3.2 Bluetooth Standard - The Link Manager
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 System Model
3.1 Approaches . . . . . . . . . . . . . . . . . . . . . . .
3.1.1 Adding WLAN APs . . . . . . . . . . . . . .
3.1.2 Introduce RFID-WLAN tags . . . . . . . . . .
3.1.3 Introducing Bluetooth transceivers . . . . . .
3.1.4 Combining Bluetooth, RFID-WLAN tags, and
3.2 Searching Algorithms . . . . . . . . . . . . . . . . . .
3.2.1 Trilateration . . . . . . . . . . . . . . . . . . .
3.2.2 RF fingerprinting . . . . . . . . . . . . . . . .
3.3 Propagation Model . . . . . . . . . . . . . . . . . . .
3.3.1 Free space Propagation . . . . . . . . . . . . .
3.3.2 Indoor Models . . . . . . . . . . . . . . . . . .
3.4 Methodological Approach . . . . . . . . . . . . . . .
.
.
.
.
.
1
1
2
3
4
4
.
.
.
.
.
.
.
.
5
5
6
8
8
9
10
11
12
. . . . .
. . . . .
. . . . .
. . . . .
WLAN
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
14
14
14
15
15
16
16
16
17
18
19
20
21
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4 Simulation Results
22
4.1 Parameters’ Definition . . . . . . . . . . . . . . . . . . . . . . 22
4.2 Trilateration Results . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Fingerprinting Results . . . . . . . . . . . . . . . . . . . . . . 25
5 Conclusion and Future Work
29
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
i
List of Tables
1
2
3
4
5
6
7
8
9
Different delays for different physical layers . . . . . .
Power classes . . . . . . . . . . . . . . . . . . . . . .
parameters values . . . . . . . . . . . . . . . . . . . .
25, 50 and 90 percentiles for WLAN-only approach . . .
.
.
.
.
.
.
.
.
25, 50 and 90 percentiles for the combined scenario, case 1 .
25, 50 and 90 percentiles for the combined scenario, case 2 .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
11
22
23
24
25
25, 50 and 90 percentiles for the WLAN-only scenario with 6 APs
against the combined one with 4 APs and 20 TRXs . . . . . . . . 26
25, 50 and 90 percentiles for the WLAN-only scenario with 9 APs
against the combined one with 6 APs and 30 TRXs . . . . . . . . 27
25, 50 and 90 percentiles for the WLAN-only scenario with 15 APs
against the combined one with 9 APs and 52 TRXs . . . . . . . . 28
ii
List of Figures
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
WLAN network example . . . . . . . . . . . . . . . . . . . .
Three different RFID tags. . . . . . . . . . . . . . . . . . . .
Different Bluetooth topologies . . . . . . . . . . . . . . . . .
Bluetooth signaling procedure . . . . . . . . . . . . . . . . .
Adding WLAN APs to the existing Network . . . . . . . . .
Adding Bluetooth transceivers to the existing Network . . .
Adding WLAN access points and Bluetooth transceivers to
the existing Network . . . . . . . . . . . . . . . . . . . . . .
Trilateration method with three control points . . . . . . . .
Fingerprinting algorithm: offline and online phases . . . . .
Trilateration algorithm: comparison among four different densities
of WLAN access points . . . . . . . . . . . . . . . . . . . . . .
Trilateration algorithm: comparison among four different densities
of APs and TRXs, case 1 . . . . . . . . . . . . . . . . . . . . .
Trilateration algorithm: comparison among four different densities
of APs and TRXs, case 2 . . . . . . . . . . . . . . . . . . . . .
WLAN-only (6 APs) against the combined scenario (4 APs , 20
TRXs) using Fingerprinting algorithm. . . . . . . . . . . . . .
WLAN-only (9 APs) against the combined scenario (6 APs , 30
TRXs) using Fingerprinting algorithm. . . . . . . . . . . . . .
WLAN-only (15 APs) against the combined scenario (9 APs , 52
TRXs) using Fingerprinting algorithm. . . . . . . . . . . . . .
iii
.
.
.
.
.
.
5
10
12
13
14
15
. 16
. 17
. 18
. 23
. 24
. 25
. 26
. 27
. 28
List of Abbreviations
AFH
Adaptive Frequency Hopping
ACL
Asynchronous Connectionless
ADSL
Asymmetric Digital Subscriber Line
AP
Access point
AOA
Angle Of Arrival
BER
Bit Error Rate
BT
Bluetooth
CCK
Complementary Code Keying
CRC
Cyclic Redunduncy Check
CSMA/CA Carrier Sense Multiple Access with Collision Avoidance
DBPSK Differential Binary Phase Shift Keying
DAB
Digital Audio Broadcasting
DQPSK Differential Quadrature Phase Shift Keying
DSSS
Direct-Sequence Spread Spectrum
EPE
Ekahau Positioning Engine
ERP
Extended Rate Physicals
FCC
Federal Communications Commission
FHSS
Frequency-Hopping Spread Spectrum
GFSK
Gaussian Frequency-Shift Keying
GPS
Global Positioning System
HF
High Frequency
HCI
Host Controller Interface
IEEE
Institute of Electrical and Electronics Engineers
IrDA
Infrared Data Association
ISM
Industrial, Scientific and Medical
LAN
Local Area Network
LF
Low Frequency
LTE
Long Term Evolution
MU
Moving Unit
OFDM
Orthogonal frequency-Division Multiplexing
iv
PBCC
PER
QPSK
RFID
SCO
SIG
SNR
TOA
UHF
WLAN
WSN
Packet Binary Convolutional Code
Packet Error Rate
Quadrature Phase Shift Keying
Radio Frequency Identificaion
Synchronous Connection-Oriented
Special Interest Group
Signal to Noise Ratio
Time Of Arrival
Ultra High Frequency
Wireless Local Area Network
Wireless Sensor Networks
v
1
1.1
Introduction
Background
Mobile communication applications have evolved dramatically in recent
years. Shifting from voice and text messaging services to video calling and
internet surfing, opened the gate for far more complex applications that have
been implemented or are currently being developed. Location based applications such as Global Positioning System (GPS) have also been widely used for
positioning and tracking in outdoor environments. However, this technology is not
efficient in indoor environments due to high signal attenuation, thus alternative
solutions have to be designed.
Several wireless technologies exist now for indoor positioning, namely Bluetooth, Wireless Local Area Network (WLAN), Wireless Sensor Network (WSN),
Infrared Data Association (IrDA), home RF, Radio Frequency Identification (RFID)
etc... . Bluetooth is very popular nowadays and most of the new wireless handsets
or mobile phones are equiped with a Bluetooth interface. Bluetooth is a short
range technology which works in the unlisenced 2.4 GHz Industrial, Scientific, and
Medical (ISM) band and is very useful for localization. Using Bluetooth, position
estimation could be done either at the mobile unit or in a centralized way by the
positioning system that will estimate the position from received data. WLAN is
also another short range technology that works in the 2.4 GHz ISM band and the
5 GHz Unlicensed National Information Infrastructure UN-II band. Since WLAN
access points are linked together through a higher level system, signals from mobile
units could be used to estimate their position. Similar to Bluetooth, localization
could also carried out both at the mobile unit as well as the access point. Apart
from the mentioned technologies, RFID systems play an important role in localization applications. In this case, a tag is attached to an item, person, or animal and a
reader retreives data from these tags on a periodic or demand base. RFID systems
work in differnt frequency bands and this affects their range of operation. WSN
is a promising technology as well for communications in indoor environments and
can be exploited for localization purposes [1]. The latest version of its standard,
IEEE 802.15.4 identifies 6 physical layers working in different frequency bands: 900
MHz, 2450 MHz, 3-5 GHz, and 6-10 GHz supporting different data rates. Home
RF is another technology that could be used for localization [2]. It Works in the
2.4 GHz band and uses Frequency Hopping Spread Spectrum (FHSS) scheme to
support a data rate up to 10 Mbps and range up to 50 meters. However, home RF
is no more under development since 2003 and is rarely used due to high competition from WLAN which supports higher data rates and longer ranges. In a similar
manner, IrDA [3] is a widely used technology in remote controlling tasks such as
in TV sets and earlier versions of laptops in the period 1990 - 2000. However it is
not a good candidate for localization tasks since it requires line of sight between
the involved devices.
1
In this thesis work, I will focus on investigating three of the mentioned technologies; WLAN, RFID, and Bluetooth for localization purposes in indoor environments. In wireless systems, there is no direct metric that can measure the
distance of a mobile unit to some access or reference point. Thus, this distance has
to be measured indirectly through other parameters. Several works [4, 5, 6, 7, 8, 9]
have considered for this purpose the Received Signal Strength (RSS). This metric
has a direct relationship with the distance that can be calculated using proper
propagation model. Other metrics like Angle Of Arrival (AOA) and Time Of Arrival (TOA) have been considered in the literatures, see for instance [10, 11, 12]
and [13, 14, 15] respectively. The later matrics have been investigated thoroughly
separately or in junction to acheive better performance.
1.2
Related Work
Significant effort has been accomplished to optimize localization systems, and
several methods have been used to estimate locations accurately in WLAN, RFID,
and Bluetooth networks. For instance, in [5], a detailed propagation model for
WLAN has been considered where losses due to refraction and wall penetration
have been considered separately. This can potentially result in more accurate estimation of distance. The authors have also implemented two different algorithms
to control the update procedures of the database which holds values of the recent
power levels and their relative estimated distances. [9] introduces a new method
called location fingerprinting. The idea tailored by this method is to collect a
database of signal strength and spatial coordinates for some reference points in
the considered environment. Then an opportune algorithm is used to compare
the signal strength of a moving unit with those saved in the database allowing
to estimate the position of the considered moving terminal. For estimation, the
authors considered both a deterministic approach based on Euclidean or Manhattan distance, as well as a probabilistic one based on Bayesian localization. The
probabilistic approach lead to better estimates due to the fact that more statistical
information could be exploited. In [16], a hetrogenous localization system combining the Ekahau Positioning Engine (EPE), a software solution which utilizes
WLAN infsrastructure, and acoustic localization which is supported by a WSN
was proposed. The authors used radio and acoustic signals and calculated the
difference of their time of arrival to infer the distance between a mobile unit and
a fixed point. Trilateration was also used to estimate the position from at least
three different sources. Due to the fact that sound travels slower than radio waves,
their tests have shown that this hetrogenous system provides better accuracy than
using the EPE system alone especially at boundaries such as walls.
[6] presents a modified version of LANDMARC system [17], for indoor localization using RFID tags. LANDMARC uses some known located reference tags
instead of employing many readers that could increase the application cost. The
2
system uses the Euclidean distance in signal strength as a metric to identify the
closest reference tags to the target. Then the algorithm uses weighted mean to
estimate the location of the target unit among those closest reference points. However, [6] suggests to use only the Euclidean distance for the close reference tags
instead of accounting for all the reference tags available in the database. This will
reduce complexity and allow for quick estimates.
[4] presents an overview study of Bluetooh technology for indoor positioning.
The aim of the work is to develop a Bluetooth service that is not manufacurerspecific. For this purpose, the authors relied on Host Controller Interface (HCI)
commands. Trilateration is used to estimate the location, and signal strength is
related to distance using a logarithmic model. [8] suggests a combination of signal
strength and scene analysis1 to acheive higher accuracy. In this system, the mobile
units measure the signal strength from several Bluetooth access points in the environment and transmit these information to a main server. At the main server, an
RSS map of the environment is built, then the server estimates the position using
trilateration utilizing the RSS map that has been already built. In [7], the authors
used also RSS to estimate distances from power levels, however, they used Least
Square Estimate (LSE) in the trilateration algorithm in order to approximate the
coordinates of the mobile unit.
1.3
Purpose of the Thesis
Arlanda is the biggest airport in Sweden and is located in the north of Stockholm. As part of the developement process, a localizaiton system for indoor environments is thought of as a basis that could be used for many applications by
different departments. Examples of the services that could exploit the availabitity
of location information are:
1. Localization of employees.
2. Localization of moving objects (luggage, vehicles, etc...) within the airport
facility.
3. Advertisement services for customers in the airport.
4. Estimation of population density in several sectors of the airport.
5. More reliable security control.
At Arlanda, a WLAN infrastructure is deployed in all airport sections, and
is mainly used to allow internet access to customers. This infrastructure was
1
scene analysis is the usage of scene features whch have been previously observed to
obtain conclusion about the location of an object.
3
thought to be used in localization for security snd control purposes. The system
was originally implemented using a Cisco location appliance, but its performance
was tremendously weak. Thus, significant modifications and additions are needed
to make the system more accurate. Two major issues arise when thinking about
these modifications; how much accuracy one can acheive and at what cost.
1.4
Problem Formulation
The aim of this thesis work is to compare solutions based on three different technologies, namely WLAN, RFID and Bluetooth, and give recommendation about
which one, or which combination of them, is best suitable for an indoor location
system at Arlanda airport. The decision will be based on the scenario which gives
the lower positioning error with a relative low cost. Futher more, the performance
of two different localization algorithms, trilateration and RF fingerprinting will be
investigated.
The thesis work should answer the question:
• Based on the existing WLAN infrastructure at Arlanda, how could this
network be improved to provide better accuracy?
1.5
Thesis outline
The rest of this report is organized in the following order; chapter 2 presents an
overview of the technologies involved in this study, while chapter 3 shows the system model and the proposed approaches. Chapter 4 shows the simulation results,
and finally chapter 5 includes the conclusions and future work.
4
2
Technologies Involved
In this part, a general overview of the technologies involved is outlined. At
first, we discuss the physical and MAC layers of WLAN which is the deployed
infstracture. Following that, RFID and Bluetooth are presented.
2.1
Wireless Local Area Network, WLAN
Wireless Local Area Network, WLAN, is a wireless communication system used
to replace the existing wired infrastructure at the final link between the end user
and the existing wired network. as shown in Figure 1. This technology has been
widely used recently in public areas, universities, companies, hospitals, etc... and
its benefits are greatly appreciated by the community. The greatest among them
are:
Figure 1: WLAN network example
• Cost efficiency by eliminating the costs of bulky wired networks at the final
end of the network.
• Mobility where clients are able to roam within the range of the network
providing better work efficiency.
• Ease of access to the network where new clients can easily access the network
without the need to exist in special locations such as office, meeting rooms,
etc...
The evolution of WLAN started in mid 1990s where the first IEEE specification,
802.11, about how WLANs in the 2.4 GHz band should work was released in 1997
5
[18]. This version provided a data rate ceiling of 1 or 2 Mbps depending on the
used physical layer. Since then, several improvements of this technology have been
developed in order to provide higher data rates. 802.11b was released in 1999 to
work in the 2.4 GHz and allowing date transfer up to 11 Mbps [19]. In the same
period, another standard was also released, namely 802.11a [20], allowing data
rate up to 54 Mbps and operating in the 5 GHz band. Despite the high data rate
provided, 802.11a faced a major problem of compatibility with older versions of
WLANs. In June 2003, 802.11g was released [21]. This version solved the problem
of compatibility since it worked in the 2.4 GHz and by making use of the 802.11a
OFDM physical layer was able to achieve the same data rate of 54 Mbps. Several
other versions were released and many more are still under developement addressing new problems such as security, spectrum management, region extensions, mesh
networking, etc.... A promising standard, 802.11n is expected to be released in December 2009 claiming to support a data rate of up to 300 - 600 Mbps. However, it
is important to note that the Wi-Fi Alliance has started to certify products based
on IEEE 802.11n Draft 2.0 as of mid-2007.
2.1.1
802.11 Standard
Physical Layer
In this section, I will refer only to 802.11b and 802.11g standards. The 802.11b
standard encloses a radio spread spectrum technique, namely Direct Sequence
Spread Spectrum (DSSS), and one diffuse infrared technique. According to the
ETSI regulations, the DSSS technique splits the 2.412-2.472 2 GHz band into 13
22 MHz channels 3 . 3 non-overlapping channels exist in this scenario while adjacent channels overlap with 5 MHz band difference between 2 consecutive center
frequencies. To supply high data rates, 802.11b defines a new coding algorithm,
Complementary Code Keying CCK, to support 5.5 and 11 Mbps respectively with
the use of Quadrature Phase Shift Keying (QPSK). Binary Phase Shift Keying is
used for low data rates. In addition, Dynamic Rate Shifting is used to adapt the
data rate to the channel conditions and position of the client in the network. High
data rates are used for nearby clients, however for far clients, the system choose
low data rates with proper coding and modulation.
The 802.11g physical layer encloses four different physical sublayers;
ERP-OFDM supports data rate up to 54 Mbps. All devices that are identified
to have this option will use this physical layer to enhance performance unless
an obligation exists to switch to another physical layer.
2
these are the limiting center frequencies according to this regulation.
In USA, France, Spain, and Japan, different regulations apply. In USA, 11 channels
(2.412-2.462 GHz). In Japan, 14 channels (2.412-2.484 GHz). In France, 4 channels
(2.457-2.472 GHz). In Spain, 2 channels (2.457-2.462 GHz).
3
6
Physical Layers
ERP-OFDM
ERP-DSSS/PBCC
ERP-DSSS/CCK
DSSS/OFDM
Preamble and header delays
short µs
long µs
20
20
96
192
96
192
96
192
Table 1: Different delays for different physical layers
ERP-DSSS/PBCC is an updated version of DSSS in 802.11b to support data
rates up to 22 and 33 Mbps. The enhancement is due to a better coding
algorithm PBCC,Packet Binary Convolutional Code, which was approved
as an alternative to CCK.
ERP-DSSS/CCK is the same standard used by 802.11b. It was introduced to
support compatibility.
DSSS/OFDM is a hybrid physical layer which was introduced to support interoperability, where the physical header is transmitted using DSSS where the
payload is transmitted using OFDM, Orthogonal Frequency Division Multiplexing. The reason for this is to make sure that even devices which don’t
work with OFDM system will still be able to sense the channel correctly
since CSMA/CA protocol is used.
OFDM is the enhancement in the physical layer that 802.11g has accomplished
over 802.11b. It aims to transmit several signals on one single link or path by combining different carriers that operate on different frequencies. These frequencies
should be orthogonal so that a receiver which is tuned to a specific frequency will
only detect the message that is transmitted on this frequency. Thus, a key feature
to assure this accuracy in detection is to have orthogonal carriers which inclusively
prevent interference from neighboring frequencies in a tight band.
802.11g obligates communication to be done through short preamble4 rather
than long preamble since it reduces packet overhead in the network. This option
has been also introduced in the 802.11b standard but as an option for devices
which support it. Table 1 shows different delays for different physical layers for
short and long preambles as identified in the 802.11g standard.
MAC Layer
The MAC layer is responsible to associate clients with the access points. Usually
the client has the ability to choose among existing Access Points (APs) according
4
header holds packet information related to thet physical layer, while preamble is used
for synchronization
7
to their power level. Once the client is accepted to the join one specific network,
he will start communicating on the specified channel of this network (22 MHz
wide). If the client is open to several networks, then he will be able to reassociate to another network if he out bounds his own network or due to high traffic
on his network. This procedure is usually termed as load balancing. As mentioned earlier, the 2.4 GHz band is split into 14 overlapping channels where each
network will use one channel to communicate through. Thus, applying ”channel
reuse” pattern as in cellular networks among neighboring APs will increase system
throughput by decreasing the amount of interference. The 802.11b standard also
controls how clients will share the network by using the Carrier Sense Multiple
Access with Collision Avoidance (CSMA/CA) protocol. CSMA/CA protocol allows clients within a network to use the channel without interfering each other; for
instance, a client listens first to the link to make sure it is free then waits a random
period of time and after that transmits if the channel is still free. The receiver
will have to respond back by an acknowledgment ACK to identify the successful
reception of the packet. The need of this acknowledgment is because clients can
not detect if a collision had happened during the transmission since data is sent via
air. The sender will retransmit a packet if an ACK is not received, which means
that extra overhead is assumed to exist in the air due to the missed ACK from
the recipient side. In addition to the ACK mechanism, 802.11b implements also
a Cyclic Redundancy Check (CRC) to compare received data with the supposed
sent one. Another feature is the Packet Fragmentation where long packets are split
into smaller packets to ensure them to get received since long packets are often
more risk to get distorted while in the air.
2.2
2.2.1
RFID
Introduction
Radio Frequency Identification corresponds to an identification technology
which uses radio waves to identify objects or items. The idea of using radio waves
for identification goes back to the 50s of the last century, but it took till the 90s
to start using this technology after it became possible to manufacture feasible size
RFID electronics. An RFID system consists mainly of a reader and a tag. The
basic concept is that a reader transmits a radio wave which in turn induces current
into the circuit which is printed or installed on the tag. The tag use this current
in order to transmit the information that is stored on its internal memory. The
amount of information on the tag and its communication range depend mainly
on the type of the tag. In particular, three types of RFID tags exist; passive,
semi-passive, and active. These are shown in Figure 2.
• Passive tags: this type of tags has no power source. It depends mainly on
the received RF waves to extract the power needed to operate its embedded
8
circuit using the DC component of the received signal. In addition to that,
this tag has no transmitter as conventionaly understood for a device to be
able to transmit electromagnetic waves through the air. It works as follows:
the reader sends RF carrier in the surrounding environment. A tag which
exists in the RF coverage of the reader and has received sufficient energy,
powers up its circuit and start to modulate the recieved signal by shunting
its coil through a transistor according to the data stored in the memory
array of the tag. This modulation of the amplitude of the carrier is detected
by the reader which then decode the information according to the coding
algorithm used.
• Semi-passive tags: this tag is used when more information than those collected by a passive tag need to be stored. These tags use internal batteries
to power up the circuit. They still use the backscattered modulation of the
received signal to send their information to the reader. These tags are more
expensive than passive tags.
• Active tags similar to semi-passive tags, active tags use batteries to power
up their circuitry, however, they also use this power while transmitting their
information. Active tags are equipped with tranceiver to cover larger areas.
Costs increase dramatically if compared to passive tags.
2.2.2
Frequency bands and coverage
RFID systems work in different frequency bands. The most commonly used
are the 125/134 kHz (LF band), 13.56 MHz (HF band), 860-960 MHz (UHF band),
and the 2.4-2.45 GHz,(microwave band), due their unlicensed nature. The LF and
HF systems use inductive coupling while UHF and microwave systems use radiative coupling. Inductive coupling is used in systems where close identification is
needed since coupling falls severely as the distance between the reader and the tag
increases. On the other hand, radiative coupling is used in systems where wider
coverage is needed, but taking into consideration more complex propagation environments where interference increases remarkably than that in LF and HF systems.
Another issue to consider is the range of the RFID system. This metric depends
mainly on the power supply and frequency used. Passive tags are mainly used for
small range communications in the vicinity of centimeters or 1-2 meter range.
However, lately some companies managed to produce some passive tags that can
communicate up to 10 or 12 meters. Semi-passive and active tags are intended
for wider area coverage. Active tags can operate in the vicinity of 100 meters
depending on the environment.
9
Figure 2: Three different RFID tags.
2.3
Bluetooth
The idea of naming the current Bluetooth wireless technology by its name
refers back to a Viking monarch called Heraled Bluetooth who lived in the 10th
century,and was able to unify Denmark and Norway under the same authority.
More than that, according to a myth in [22], a monolith was digged out recently
and showed a self-portrait of Herald Bluetooth holding a cell phone and a laptop
with an antenna. It indicates his forsight though the portrait goes back thousand
years ago.
Away from the name, Ericcson was the first to look for a standard that can
replace the bulky wiring for devices’ connection. Then in 1998, the Bluetooth SIG
was formed to issue the first standard that could be used world wide. This group
consists of five pioneer companies in computer and communications technology;
namely Ericsson, Nokia, Toshiba, IBM and Intel. Later on, several updates of
the standards have been issued promising more advanced technology and better
system performance.
10
Power Class
1
2
3
Maximum
Minimal Output
Power Control
Output Power
Power
100mW (20 dBm) 1mW (0dBm)
Pmin < +4dBm
to Pmax
2.5mW (4dBm)
0.25mW (-6dBm) Optional:
Pmin toPmax
1mW (0dBm)
N/A
Optional:
Pmin toPmax
Range
100m
20m
10m
Table 2: Power classes
2.3.1
Bluetooth Standard - Physical Layer
Bluetooth is another type of short range wireless technology which operates
in the 2.4 GHz free licensed ISM band. This band became free licensed in almost
all countries due to the invention of microwave ovens in earlier years. In order
for bluetooth devices to operate, the standard divides the 2.4 GHz band into 79 1
MHz channels. The band starts at 2401.5 and ends at 2481.5 GHz. The standard
identifies Frequency Hopping Spread Spectrum (FHSS) scheme to trasnmit information among devices using a pseudo random frequency sequence. Three power
classes exist for this standard which are illustrated in table 2. This table has been
taken from the IEEE 802.15.1 standard [23].
The standard also specifies GFSK as the modulation scheme with a bandwidth
time of 0.5, where a binary 1 is represented with a postitive frequency deviation
while a binary 0 is represented by a negative deviation. The data transmission
rate has a symbol rate of 1 Msymbol/sec. In principle, the standard identifies
master and slave or slaves mechanism for devices to communicate in a two-node or
multi-node topologies respectively. In a multi-node network, up to 8 slaves could
be connected to the same master, where multi-node networks could be connected
together to form a bigger network called scatternet. However, slaves within a
scatternet could only communicate through their masters. Figure 3 shows the
three different types.
A pseudo random frequency sequence is generated by the master according to
its physical address and clock. This sequence is communicated to the slave devices
in the piconet. A device could be a master or a slave, which depends basically
on who will start to establish the connection. However, a master and a slave can
exchange roles during operation. A slave responds to the master in the next slot
he receives on only if he is addressed by that particular packet, which is how the
Bluetooth devices transmit data over air. A simple representation of Bluetooth
signaling procedure is shown in Figure 4.
A packet is subdivided into three different blocks; namely access code, header,
and payload.
• The access code is used for timing synchronization, offset compensation, and
11
Figure 3: Different Bluetooth topologies
packet acknowledgment.
• The header contains channel and link control information.
• The payload is intended for information to be sent and some security bits.
A packet may be of different lengths; 1, 3 or 5 time slots, where a time slot is
625 µs. Packets are also of different types depending on what kind of data is
tranmitted; data, voice, or video.
2.3.2
Bluetooth Standard - The Link Manager
The link manager is responsible for establishing and managing the link. The
type of link to be used depends on the type of information that the user wants
to send or recieve. Two types of links exist in the Bluetooth standard; SCO and
ACL.
SCO is a symmetric channel between the master and the slave used mainly for
audio streaming. On this link, a packet may use 3 or 5 time slots, and is not
allowed to be retransmitted. The master can establish an SCO link by a request
to the slave with the appropriate parameters. On the other side, the slave can ask
for an SCO link by a request to the master, then the master sends back a relevant
request with the appropriate parameters.
An ACL link is an assymetric channel that is used for sending data. Only addressed
slaves can reply back in the next slot they receive on. There are 3 different packet
sizes for the ACL link; 1, 2, or 3 time slots. The master decides which size to use.
A slave can ask to change the size of the packet, but that has to be approved by
the master to avoid using long packets on noisy channels where BER is significant.
12
Figure 4: Bluetooth signaling procedure
13
3
System Model
As mentioned earlier, several technologies can take up the role to provide the
system with the needed information about power levels, distances, and locations.
But in this thesis work, the focus will be on WLAN, RFID, and Bluetooth due to
their wide usage; WLAN and Bluetooth radio interfaces are included in almost all
static and portable devices, and RFID tags are easy to configure and implement
in a system and presents a relativly low cost.
In order to reuse the existing infrastructure for localization purposes, some alternative that might potentially improve the accuracy of localization will be described.
3.1
3.1.1
Approaches
Adding WLAN APs
For this option, a number of APs will be added to reduce the areas where a
very weak or no signal is detected. These additional APs will enhance accuracy
by providing higher signal strength levels since the physical distance between any
arbitrary mobile unit and its surrounding APs will be reduced. More than that,
some areas lack the existence of at least 3 APs: these are needed by the trilateration
algorithm to work properly, and the additional APs will compensate for this lack
as well. In Figure 5, the red marked APs indicates the location of some APs that
are thought to be added to the existing network.
Figure 5: Adding WLAN APs to the existing Network
14
3.1.2
Introduce RFID-WLAN tags
To address some of the services mentioned in section 1.3, items such as luggage
trailers and wheel chairs need to be equiped with a device that can be used to
identify their identity. RFID tags are mostly used for such tasks, where tags are
attahed to the items and a reader monitors their presence in its coverage area.
RFID-WLAN tags have also been recently commercialized. These do not require
a reader and can use the WLAN infrastructure to transmit their data.
3.1.3
Introducing Bluetooth transceivers
Bluetooth could be merged with the existing WLAN system by adding Bluetooth transceivers in the gaps where weak signal strength levels from WLAN APs
are recorded. They could also be used to complement the set of 3 APs that are
required for trilateration in areas where only two APs exist. Notice in Figure 6
that the number of Bluetooth transceivers (bluetooth logo) that are needed exceeds the number of APs needed in the Adding WLAN APs scenario. This is due
to the fact even though the transceivers could be of class 1 and may reach up to
100 meters in ideal conditions, many mobile devices uses bluetooth radio interface
of class 3 and allow communication ranges of 10 meters only.
Figure 6: Adding Bluetooth transceivers to the existing Network
15
3.1.4
Combining Bluetooth, RFID-WLAN tags, and WLAN
Combining a WLAN infrastracture with Bluetooth access points or base stations to fill the gaps in coverage, and RFID tags to identify particular items, into
a single platform and using all the data from these sources would help to increase
the accuracy of estimations. Figure 7 sketch the structure of a general system
exploiting this solution.
Figure 7: Adding WLAN access points and Bluetooth transceivers to the
existing Network
3.2
3.2.1
Searching Algorithms
Trilateration
In radio communication networks, Trilateration is one method used for localization [24]. In principle, a signal should be collected at/from three different
control points, base stations or access points, depending whether a centralized
or distributed system is deployed. If the signal is detected by more than one
control point, then a coverage circle is drawn around each of the control points
with a radius proportional to the parameter’s value; signal strength or time of
arrival of signal, and the intersection points among the different circles drawn become the possible locations of the transmitter. However, having a third control
point increases the accuracy by filtering out the incorrect possibility. In case, the
16
third circle doesn’t coincide with any of the intersection points, then an averaging
method is used to estimate the final position. In wireless networks, this is often
the case where the variation in the channel response for each control point doesn’t
guarantee precise distance calculation, in addition to the fact that it is hard to
have more than three or four control points to detect the same transmitter. Figure
8 shows the principle exploited by this method.
Figure 8: Trilateration method with three control points
In this thesis work, the signal strength is used as a parameter to calculate the
relative distance between the control point and the transmitting device, and the
average of the closest intersection points as the final position.
3.2.2
RF fingerprinting
Fingerprinting is another method for localization which can be implemented in
a centralized or distributed manner. This method can be divided in an offline and
an online phase [24]. In the offline phase, data are collected at many predefined
reference points in order to form a database. This database includes parameters
that identify the considered reference points such as their position, the average
signal strength, angle of arrival, and time of arrival of signal or other parameters
that could be important for localization purposes. This might represent a very
demanding procedure because it requires the manual collection of the aforementioned data at each of the reference points.We further remark that in order to
obtain accurate estimates, those data might have to be collected several times in
order to obtain average quantities. For the online phase, the system will try to
locate an arbitrary transmitter by first measuring some of the available parameters
17
(SS, AOA, TOA) and then matching those parameters with the one available in
its database, selecting the reference point that corresponds to the best parameter
match. Weighted mean, see equation 1, gives better results than arithmetic mean
[9], but other averaging methods such as the smallest polygon [25] could also be
used to estimate the final position. The weight is found as the inverse of the difference in signal strength between a reference point and the transmitting device
for the same AP as shown in equation (2). Figure 9 presents an example of a
centralized fingerprinting algorithm where SS and AOA are used as parameters.
N
X
a=
wi ai
i=1
N
X
(1)
wi
i=1
w=
1
|SSref p − SStr |
(2)
Figure 9: Fingerprinting algorithm: offline and online phases
3.3
Propagation Model
To obtain system performance that could be considered reliable, one has to
model the system as close as possible to reality. In context aware systems, accuracy of the system is the first priority. Under this consideration, any type of
18
wireless technology could provide signal strength levels based on the deployed infrastructure. However, the localization algorithm is the key feature which results
in the best estimations. Algorithms which depend on RSS to estimate distances
to mobile units, should rely on a powerful propagation model that takes into
consideration all the factors that causes signal attenuation. According to
[26], propagation mechansims include Reflection, Diffraction, and Scattering.
• Reflection is the action of a wave striking a large sized object as compared to its wavelength.
• Diffraction is the action of a wave hitting a sureface with irregularities,
like the top of a mountain or the corner of a building. The result will be
the creation of new secondary waves behind and around the obstacle.
• Scattering is the action of a wave striking against objects that are small
if compared to its wavelength. Scattered waves are created which will
affect the original wave at the receiver.
Due to its complexity, several propagation models have been proposed by
different authors. In the nect section, a brief introduction about wave propagation is provided, the propagation models that best work for indoor environments will also be outlined.
3.3.1
Free space Propagation
In principle, the free space propagation model describes the power density at the receiver due to a line-of-sight path between the transmitter and
the reciever, where no obstacles are assumed to affect the signal. Assuming
isotropic radiation, equation (3) explains that the power density at the receiver will be equal to the transmitted power uniformly distributed on the
surface of the sphere caused by the isotropic radiation. The power at the receiver is just the same power density multiplied with the area of the receiver
antenna as shown in equation (4).
=r =
Pt
4πr 2
W/m2
Pr = =r Ae =
Pt Ae
4πr 2
W
Pt is the transmitted power
Pr is the received power
=r is the power density at the receiver
Ae is the effective aperture of the receiving antenna
19
(3)
(4)
r is the distance from the transmitter
These equations are just rough approximations that are mostly used for
very long distance communications such as satellites and space detections.
The earth is full of elements that can deteriorate waves. Human beings,
animals, geographical irregularities, trees, buildings, etc..., are just few examples. As we shrink the environment, these factors become more and more
threatening and should not be ignored in calculations. Indoor environments
consist of many objects that are static and others that change dynamically.
In an office, a desk, board, computer desktop, flowers, etc... are some examples of static objects, whereas people, moving chairs, small instruments are
other examples which are dynamically changing. This shows that the channel response will change severely randomly. In the following section, some
comprehensive models are presented.
3.3.2
Indoor Models
In this section, the terminology path loss will be refered to as the ratio
between the transmitted power and the received power, as illustrated in
equation (5).
Pt
(5)
Pr
Some environments require different propagation models according to
their structure. As mentioned earlier, objects within these environments
change location dynamically which inturn changes the channel response accordingly. From these models we mention the Keenan-Motley model which
takes into consideratiion these effects.
However, in this thesis work, the Multi-Wall-and-Floor [27] is used. This
model takes into consideration the nonlinear relationship between the cumulative penetration loss and the number of penetrated floors and walls which
is an affecting factor in determining the power loss. In other words, the penetration loss will decrease with the increase number of penetrated walls and
floors which are of the same type. Equation (6) shows this relation
L=
LM W F = Lo + 10nlog(d) +
K
I X
X
i=1 k=1
Lwik +
J X
M
X
Lf jm
(6)
j=1 m=1
where L0 is the reference path loss at 1m, n is the path-loss exponent, d
is the distance from the transmitter to the receiver, Lwik is the attenuation
from wall type i at the kth penetrated wall of the same type, Lf jm same
20
explanation as Lwik but for floors, I and J are the number of wall and floor
types respectively, and K and M are the number of penetrated walls and
floors of the same type respectively.
3.4
Methodological Approach
The thesis work is based on computer simulations for the different approaches. For this purpose, a Matlab code is written, where the simulations
consist of 1000 snapshots for each case of the different approaches. This number of snapshots is sufficient to provide a clear idea about the percentage of
error for all the considered cases. After defining the network characteristics,
a snapshot consists of randomly placement of a transmitting device within
the considered area, path loss and power calculations, applying the searching
algorithm, and finally estimation of the position. Error calculation is based
on Euclidean distance in two dimensional space between two points. For
instance, if (x, y) are the coordinates of the real position, and if (xo, yo) is
the system’s estimation, then the error is found to be:
p
error = (x − xo)2 + (y − yo)2
(7)
21
4
Simulation Results
This part shows the results of the simulations for the considered approaches. Simulations are based on two Matlab codes that respectively emulate the Trilateration and Fingerprinting algorithms: both took into consideration having one or two technologies involved. Before presenting the
obtained results, in the next section some parameters and values which have
been used will be defined.
4.1
Parameters’ Definition
For both algorithms, the same sample hall with dimensions (200∗200 m2 )
which is very close to the dimensions of all the halls at Arlanda airport has
been considered. Also, the control points (WLAN APs and Bluetooth TRXs)
are distributed symmetrically over the whole area to acheive maximum possible coverage for a specific number of control points. The propagation model
used in these simulations is the MWF model, see equation (6); this takes
into consideration different types of walls and floors. The parameters’ values
used in this model are summarized in Table5 3.
Parameter
Value
WLAN 802.11b trans. power
20 dBm (100 mW)
Bluetooth 802.15.1 trans. power 4 dBm (2.5 mW)
Center frequency
2.45 GHz
distance path loss
2
Lw
3, 5 dB
Lf
18 dB
Table 3: parameters values
4.2
Trilateration Results
The results of the trilateration simulations show that the number of control points and their distribution over the grid is the main factor that allows
to reduce the error. It will be shown that overlapping coverage areas of adjacent APs is important for this algorithm to work properly. In the following
results, the error in distance, see equation (7), is plotted and discussed.
5
for the wall loss coefficient, 3dB corresponds to thin walls and 5dB to thick walls.
22
In Figure 10, the error in distance is plotted for 4 different densities where
only WLAN APs exist as control points. The plots show that as the number
of APs increase, the error will decrease accordingly. The blue and the green
plots correspond to densities of 6 and 13 WLAN APs distributed over the
sample hall, 200*200 m2 . The improvement with 13 APs is not significant
because there are still uncovered gaps and transmitters that couldn’t be
detected by several APs. However with 17 APs, we notice better reduction
in the error reaching up,at 50 percentile level, to 10 meters, and to 17 meters
with 25 APs. In Table 4 the 25, 50 and 90 percentiles are listed for each
density.
1
0.9
0.8
Cumulative probability
0.7
0.6
0.5
0.4
0.3
WLAN: 6
WLAN: 13
WLAN: 17
WLAN: 25
0.2
0.1
0
0
5
10
15
20
Error (m)
25
30
35
40
Figure 10: Trilateration algorithm: comparison among four different densities of
WLAN access points
WLAN density 25% 50% 90%
6 WLAN APs 18.9 27 36.2
13 WLAN APs 16.5 24.8 35.1
17 WLAN APs 8.6 15.1 33.3
25 WLAN APs 5.03 8.6 15.3
Table 4: 25, 50 and 90 percentiles for WLAN-only approach
The second approach is to combine WLAN APs and Bluetooth TRXs in
the same grid. For this approach, an equivalent density of APs and TRXs are
23
deployed instead of having only APs. Many combinations could be deployed,
however, only two cases are considered in this part. The first is shown in
Figure 11 with four different densities which are equivalent to the densities
shown in Figure 10. For all densities, the number of APs is limited to 4 while
the number of TRXs is increased accordingly as the density increase. Table
5 shows the improvements that we can acheive if this case is deployed. The
second case is to increase the number of APs and TRXs, Figure 12, where the
number of APs is increased till 9 while that of TRXs is increased accordingly.
The plots in this figure and values in Table 6 shows that this case has less
accurate results than the first one, but still confirm that the combination
approach is better than using WLAN APs only.
1
0.9
0.8
Cumulative probability
0.7
0.6
0.5
0.4
0.3
WLAN APs: 4, BT TRXs: 90
WLAN APs: 4, BT TRXs: 130
WLAN APs: 4, BT TRXs: 210
WLAN APs: 4, BT TRXs: 20
0.2
0.1
0
0
5
10
15
20
25
Error (m)
30
35
40
Figure 11: Trilateration algorithm: comparison among four different densities of
APs and TRXs, case 1
density
25% 50% 90%
4 WLAN APs, 20 BT TRXs
7.9 20.6 35
4 WLAN APs, 90 BT TRXs 4.26 6.5 26.2
4 WLAN APs, 130 BT TRXs 3.96 6.09 13.6
4 WLAN APs, 210 BT TRXs 2.49 4.24 8.24
Table 5: 25, 50 and 90 percentiles for the combined scenario, case 1
24
1
0.9
0.8
Cumulative probability
0.7
0.6
0.5
0.4
0.3
0.2
WLAN APs: 9, BT TRXs: 40
WLAN APs: 9, BT TRXs: 80
WLAN APs: 9, BT TRXs: 160
WLAN APs: 4, BT TRXs: 20
0.1
0
0
5
10
15
20
25
Error (m)
30
35
40
Figure 12: Trilateration algorithm: comparison among four different densities of
APs and TRXs, case 2
density
25% 50% 90%
4 WLAN APs, 20 BT TRXs
7.9 20.6
35
9 WLAN APs, 40 BT TRXs
6.9 16.6 33.8
9 WLAN APs, 80 BT TRXs
4.8
7.7
30.9
9 WLAN APs, 160 BT TRXs 3.3
5.4 10.66
Table 6: 25, 50 and 90 percentiles for the combined scenario, case 2
4.3
Fingerprinting Results
Similar approaches are considered here as those in the Trilateration simulations, where cases with WLAN APs only are compared against scenarios
with a combination of WLAN APs and Bluetooth TRXs. The difference here
is the efect of the number of reference points on the results as mentioned in
section 3.2.2.
In Figure 13, we start by the smallest density of 6 WLAN APs and its
equivalent of 4 APs and 20 TRXs combined on the same grid. At first, the
number of reference points is set to 63, i.e. on the same area of 200*200 m2
a ref. pt. is collected every 650 m2 . The distribution of these points is fairly
symmetrical over the whole area. Then we increase this number to 170 and
25
6 APs only
4 APs, 20 TRXs
Ref. pts. density
25% 50% 90% 25% 50% 90%
63
7.2 11.3 18.2 5.8
9.1 17.8
170
3.4
5.6 11.3 3.3
5.2
10
325
2.8
4.7
8.8
2.6
3.9
7.6
out coverage conf. int. (95%)
235 ± 5.962
375.8 ± 6.8202
Table 7: 25, 50 and 90 percentiles for the WLAN-only scenario with 6 APs against
the combined one with 4 APs and 20 TRXs
325 which is equivalent to a ref. pt. every 235 and 123 m2 . We see from
this figure that the performance of the combined scenario is always slightly
better than the other. Table 7 shows the error in meters at the 25, 50 and
95 percentile levels for both scenarios and the average of out coveraged with
a 95% confidence interval.
1
0.9
0.8
Cumulative probability
0.7
0.6
0.5
0.4
0.3
WLAN APs: 6, Ref pts: 63
WLAN APs: 6, Ref pts: 170
WLAN APs: 6, Ref pts: 325
WLAN APs: 4, BT TRXs: 20, ref pts: 63
WLAN APs: 4, BT TRXs: 20, ref pts: 170
WLAN APs: 4, BT TRXs: 20, ref pts: 325
0.2
0.1
0
0
5
10
15
Error (m)
20
25
30
Figure 13: WLAN-only (6 APs) against the combined scenario (4 APs , 20 TRXs)
using Fingerprinting algorithm.
Same conclusions can be drawn from Figures 14 and 15 and their corresponding Tables 8 and 9 respectively. However, the out of coverage changes
drastically due to the fact that more areas are covered. We can see that
having only WLAN APs will cover more areas than the combined network,
while counting for accuracy, the combined scenario has showed slightly better
26
results than the WLAN-only one for all densities of reference points.
1
0.9
0.8
Cumulative probability
0.7
0.6
0.5
0.4
0.3
WLAN APs: 9, ref pts: 63
WLAN APs: 9, ref pts: 170
WLAN APs: 9, ref pts: 325
WLAN APs: 6, BT TRXs: 30, ref pts: 63
WLAN APs: 6, BT TRXs: 30, ref pts: 170
WLAN APs: 6, BT TRXs: 30, ref pts: 325
0.2
0.1
0
0
5
10
15
20
Error (m)
25
30
35
Figure 14: WLAN-only (9 APs) against the combined scenario (6 APs , 30 TRXs)
using Fingerprinting algorithm.
9 APs only
6 APs, 30 TRXs
Ref. pts. density
25% 50% 90% 25% 50% 90%
63
6.37 10.35 17.15 5.8
8.8 15.45
170
3.8
6
11.25 2.86 4.7
8.7
325
3.2
5
9.1
1.95 3.3
7.8
out coverage conf. int. (95%)
22.2 ± 1.9424
49.5 ± 2.1067
Table 8: 25, 50 and 90 percentiles for the WLAN-only scenario with 9 APs against
the combined one with 6 APs and 30 TRXs
27
1
0.9
0.8
Cumulative probability
0.7
0.6
0.5
0.4
0.3
WLAN APs: 15, ref pts: 63
WLAN APs: 15, ref pts: 170
WLAN APs: 15, ref pts: 325
WLAN APs: 9, BT TRXs: 52, ref pts: 63
WLAN APs: 9, BT TRXs: 52, ref pts: 170
WLAN APs: 9, BT TRXs: 52, ref pts: 325
0.2
0.1
0
0
5
10
15
20
Error (m)
25
30
35
Figure 15: WLAN-only (15 APs) against the combined scenario (9 APs , 52
TRXs) using Fingerprinting algorithm.
15 APs only
9 APs, 52 TRXs
Ref. pts. density
25% 50% 90% 25% 50% 90%
63
5.4
8.9 16.8 4.75 7.9 15.35
170
2.75 4.4 9.39 2.25 3.5
7.8
325
2.25 3.5
6.6
1.5 2.46
6.2
out coverage conf. int. (95%)
3.2 ± 0.8979
4.4 ± 0.929
Table 9: 25, 50 and 90 percentiles for the WLAN-only scenario with 15 APs
against the combined one with 9 APs and 52 TRXs
28
5
5.1
Conclusion and Future Work
Conclusion
In this thesis work, WLAN, Bluetooth, and RFID technologies have been
evaluated to give recommendation to the department of development and
research at LFV, Arlanda airport, about which technology or a possible combination of them could serve best for positioning services. In addition, two
localization algorithms, namely Trilateration and RF fingerprinting, have
been tested to find the one that give higher accuracy.
A WLAN network already exists at Arlanda to provide internet services
for customer, so based on the existing infrastructure, the work has proposed
three different approaches. The first one is to add more WLAN APs in areas
where the coverage is poor or doesn’t exist. The second is to add Bluetooth
TRXs in these areas and combine them with the WLAN APs into one centralized system that can extract their information and draw results. The
third proposal is to add Bluetooth TRXs and RFID tags following the same
procedures as in the second proposal. It has been shown that according to
the specifications of the company’s needs that RFID tags will only increase
the scope of the services since the company is interested to deploy WLANRFID tags to eliminate the need for RFID readers. Thus, RFID was not
included in the simulations. The first two proposals have been tested using
both Trilateration and RF fingerprinting algorithms.
For the Trilateration algorithm, the simulations have shown that a network of WLAN APs and Bluetooth TRXs has always better performance
in reducing the error between the position of a transmitting device and the
system estimation than having a WLAN-only network. More than that, the
results have also shown that increasing only the number of Bluetooth TRXs
while keeping the number of WLAN APs fixed to its origin will serve better
than increasing both of them.
On the other side, the results of using RF fingerprinting algorithm indicate similar behaviour with the following exceptions; Acheiving better coverage depends mainly on the number of APs and/or TRXs and the combination
of them. More than that, deploying only WLAN APs maintains higher coverage than deploying a combination of WLAN APs and Bluetooth TRXs
for lower densities. However, increasing the accuracy of the system is done
through increasing the number of reference points. Following this, deploying
a combined network of APs and TRXs has shown always slightly better per29
formance than deploying WLAN-only network.
5.2
Future Work
In this thesis work, the main interest was to reduce the error in distance
between the position of a transmitting device and the system’s estimation.
This constitutes a major factor in evaluating the performance of localization
systems and was the main basis to give the recommendations.
But for companies like LFV, a localization system is mainly needed to
locate objects and to offer services for customers. However, the later doesn’t
only need high position accuracy, but good timing to offer these services
especially that indoor environments are relatively small. Thus, it becomes
quite important to test the response time of the system to locate a device in
order to offer a service at that specific location and at the correct time. More
than that, using different radio technologies, will that affect the response
time?
30
References
[1] Katelijne Vandenbussche. Fine-grained Indoor Localization using Wireless Sensor Networks. Master’s thesis, Delft University of Technology,
2005.
[2] HomeRF Working Group. Wireless Networking Choices for the Broadband Internet Home. 2001.
[3] Dave Suvak. IrDA and Bluetooth: A complementary comparison. Technical report, Extended Systems, 2000.
[4] K. Thapa, S. Case. An Indoor Positioning Service for Bluetooth Ad Hoc
Networks. 2003.
[5] E. M. Macias A. Suarez D. Sanchez, S. Afonso. Devices Location in 802.11Infrastructure Networks using Triangulation.
[6] Myong-Soon Park Guang-yao Jin, Xiao-yi Lu. An Indoor Localization Mechanism Using Active RFID Tag. 2006.
[7] A. Zapater Z. Lue S. Feldmann, K. Kyamakya. An indoor Bluetooth-based
positioning system: concept, Implementation and experimental evaluation.
2007.
[8] C. J. Escudero M. Rodriguez, J. P. Pece. In-building location using Bluetooth.
2005.
[9] A. G. Dempster B. Li, J. Salter and C. Rizos. Indoor Positioning Techniques
Based in Wireless LANs. 2006.
[10] R.P. Martin E. Elnahrawy, J. Austen-Francisco. Adding Angle of Arrival
Modality to Basic RSS Location Management Techniques. ISWPC apos;07.
2nd International Symposium, 2007.
[11] Chih-Sheng Hsueh Chin-Der Wann, Yi-Jing Yeh. Hybrid TDOA/AOA Indoor Positioning and Tracking Using Extended Kalman Filters. Vehicular
Technology Conference, VTC 2006-Spring, IEEE 63rd, 2006.
[12] Richard Klukas Carl. M. Wong, Geoffrey G. Messier.
Evaluating
Measurement-based AOA indoor Location using WLAN Infrastructure. Proceedings of the 20th International Technical Meeting of the Satellite Division
of the Institute of Navigation ION GNSS 2007, 2007.
[13] J. Modelski R. Szummy, K. Kurek. Antenna Diversity Impact to Indoor Wireless TOA-based Positioning Systems Accuracy. Radio and Wireless Symposium, IEEE, 2008.
31
[14] F.Barcelo-Arroyo M. Ciurana, S. Cugno. WLAN Indoor Positioning based on
TOA with two refrence points. Positioning, Navigation and Communication,
WPNC 07, 4th Workshop, 2007.
[15] A. S. Omar A. Assie Ali. Time of Arrival Estimation for WLAN Indoor Positioning Systems using Matrix Pencil Super Resolution Algorithm. Proceedings
of the 2nd workshop on Positioning, Navigation, and Communication, WPNC
05, and 1st Ultra-Wideband Expert Talk, UET 05, 2005.
[16] A. Zaslavsky P. Hii. Improving Location accuracy by combining WLAN positioning and Sensor Technology.
[17] Y.C. Lau A.P. Patil L.M. Ni, Y. Liu. LANDMARC: Indoor Location Sensing
Using Active RFID.
[18] IEEE Std 802.11-1997 Information Technology- telecommunications And Information exchange Between Systems-Local And Metropolitan Area Networksspecific Requirements-part 11: Wireless Lan Medium Access Control (MAC)
And Physical Layer (PHY) Specifications.
[19] IEEE 802.11-1999: Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks
- Specific requirements - Part 11: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) Specifications.
[20] IEEE 802.11a-1999: IEEE Standard for Telecommunications and Information Exchange Between Systems - LAN/MAN Specific Requirements - Part
11: Wireless Medium Access Control (MAC) and physical layer (PHY) specifications: High Speed Physical Layer in the 5 GHz band.
[21] IEEE Std 802.11g-2003:Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks
- Specific requirements - Part 11: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) specifications AMENDMENT 4: Further
Higher Data Rate Extension in the 2.4 GHz Band.
[22] Robert Morrow. Bluetooth Operation and Use. McGraw-Hill Professional,
2002.
[23] IEEE 802.15.1-2005 standard part 15.1: Wireless medium access control
(MAC) and physical layer (PHY) specifications for wireless personal area
networks (WPANs).
[24] Johan Hjelm Krzysztof W. Kolodziej. Local positioning systems: LBS applications and services. CRC Press, 2006.
32
[25] Yu-Chee Tseng Shih-Lin Wu. Wireless ad hoc networking: personal-area,
local-area, and the sensory-area networks. CRC Press, 2007.
[26] B. Slimane L. Ahlin, J. Zander. Principles of Wireless Communications.
Studentlitteratur AB, Sweden, 2006.
[27] I. Forkel M. Lott. A Multi-Wall-and-Floor Model for Indoor Radio Propagation.
33
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

Related manuals

Download PDF

advertising