Professional Documents
Culture Documents
Handover ROF
Handover ROF
Proefschrift
Samenstelling promotiecommissie:
Rector Magnificus, voorzitter
Prof. dr. ir. Ignas G. M. M. Niemegeers, Technische Universiteit Delft, promotor
Dr. R. R. Venkatesha Prasad, Technische Universiteit Delft, copromotor
Prof. dr. ir. P. G. M. Baltus, Technische Universiteit Eindhoven
Prof. dr. ir. S.M. Heemstra de Groot, Technische Universiteit Eindhoven
Prof. Bala Natarajan, Kansas State University, USA
Dr. ir. M. J. Bentum, Universiteit Twente
Dr. ir. J. H. Weber, Technische Universiteit Delft
Prof. dr. ir. R. E. Kooij, Technische Universiteit Delft, reservelid
Copyright
2014,
c Quang Van Bien
All rights reserved. No part of this book may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, without prior permission from
the publisher.
ISBN 978-90-6464-830-4
v
vi
Contents
Contents
1 Introduction 1
1.1 Indoor Wireless Networks . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Summary of Main Contributions . . . . . . . . . . . . . . . . . . . . 8
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
vii
Contents
Bibliography 131
Summary 145
Samenvatting 147
Acknowledgements 149
Publications 151
viii
Chapter 1
Introduction
Wireless communication is playing an important role in our daily life since it offers
flexibility and mobility. Currently, there are more than 3.2 billion users and a
hundredfold increase of traffic is expected by the year 2017 [1]. In the future,
the home and office networking environments are predicted to be dominated by a
variety of multimedia services like wireless High-definition and Ultra High-definition
television (HD and UHDTV), wireless home entertainment and virtual wireless
offices [2]. In order to support these applications, the indoor wireless network should
provide the user with a transmission capacity of hundreds of Mbps using short-range
Gbps wireless technology and using techniques like antenna diversity, beam-forming,
sophisticated coding schemes, etc. With 5 GHz of unlicensed spectrum available
at the 60 GHz band, the millimeter wave radio at 60 GHz becomes a promising air
interface for future home networks.
1
1. Introduction
• Peripheral and device sharing - Devices such as scanners and printers are
already accessible from any location in the building through an indoor network.
• File sharing - Calendars, address books, documents etc. can be accessed from
any devices in the home.
• Game, audio and video distribution: real-time or stored audio and video etc.
can be distributed to any device in the home. In [4] this is further grouped
into four categories: (U)HD streaming, (U)HD interactive, (U)HD messaging,
and (U)HD broadcasting. This requires 1 Gb/s data rate and 500 GB storage
per hour.
Table 1.1 illustrates the required data rates for wireless applications in the six
categories. In the six categories, the first four categories are provided in the
traditional Local Area Networks (LAN). The last two categories are becoming the
indispensable part of indoor networks. Most of them require the wireless networks
to provision high data rates of several hundred of Mbps.
2
1.1. Indoor Wireless Networks
One of the radio technologies introduced in the early stage of indoor wireless
networking is Digital Enhanced Cordless Telecommunications (DECT). The DECT
standard was developed by European Telecommunications Standards Institute
(ETSI) for wireless communications over unlicensed spectrum 1880 - 1900 MHz.
This standard can offer upto 2 Mbps data rate and communications range of upto
200 m indoor and upto 6 km using a directional antenna outdoors. In 2006, DECT
Forum introduced a new DECT based home networking technology, named "Cordless
Advanced Technology - internet and quality (CAT-iq)" [5], which aims for VoIP and
other Internet-based services on the standard cordless DECT phones. There are
roughly 800 million DECT devices in the world. DECT is optimized for the new
emerging M2M wireless communications, which is expected to grow exponentially.
Wireless Personal Area Networking (WPAN) is also an indispensable part of
the ubiquitous communication landscape, it is designed for low power and short
ranges. The dominant WPAN technologies are part of the IEEE 802.15 standards
suite [6]. The best known is Bluetooth operating in the 2.4 GHz band. It was
standardized as IEEE 802.15.1. The latest version is Bluetooth v4.0 with low energy,
known as Bluetooth Low Energy (BLE). HomeRF is a wireless technology for home
networking [7]. It is based on the IEEE 802.11 FHSS standard. HomeRF supports
high-quality voice communication and incorporates with DECT. Another standard
released in 2001, IEEE 802.15.4 known as ZigBee, is for low-cost and low power
ubiquitous wireless networking. IEEE 802.15.4 operates at the 868/915 MHz and
3
1. Introduction
2.4 GHz to provide 20/40 Kbps and 250 Kbps. Thus, ZigBee is mainly designed for
command and control, not for video and high quality audio, although it can send
text or voice messages. Moreover, ZigBee has no mechanism for QoS provisioning.
IEEE 802.15.3 specifies MAC/PHY standards to provide high data rate connectivity
among devices for WPAN communication [8]. Sub-task group 802.15.3c formed in
2005 specifies a new millimeter wave based PHY which operates in the unlicensed
57-64 GHz band. It can support high data rate (over 3 Gbps) applications, e.g., high
speed Internet access, HDTV, etc. WPAN and Wireless LAN (WLAN) technologies
are expected to complement each other to provide connectivity to end-users in the
indoor environment.
Table 1.2 presents a summary of some current and emerging wireless standards
and their operating frequencies. Most of the standardized in-home radio technologies
have been concentrating on the 2.4 GHz range (e.g., Bluetooth, IEEE 802.11b/g,
HomeRF, IEEE 802.15.3) and 5 GHz (e.g., IEEE 802.11a, HIPERLAN/2) unlicensed
bands. The limited bandwidth restricts the highest achievable data rate to the order
of 100 Mbps. These radios are designed to support the home networks today, as
they are typically oriented towards sharing data, Internet access and peripheral
devices. However, they are not sufficient for the future of advanced multimedia
applications like UHDTV, IPTV, multi-player gaming, and VoIP. These HD or
UHD multimedia applications require new radio solutions for enabling Gbps high
speed indoor communications. There have been many research initiatives on seeking
solutions from millimeter wave radios. With the huge unlicensed bandwidth of
5 GHz, 60 GHz radio has recently been a promising candidate for providing multi-
Gbps radio links for short range line-of-sight (LOS) communications. Currently, six
standards define physical layer (PHY) specifications operating at the 60 GHz band.
They are IEEE 802.15.3 c, IEEE 802.11 ad, IEEE 802.16 e, Wireless HD, ECMA,
and WiGig. The investigations have focused on using 60 GHz radio as a means
of cable replacement for high data rate communication between HD multimedia
devices; for example, a HDTV set and a Blueray player in close vicinity within a
room. The future home network will require high speed connectivity in the entire
physical dwelling site, e.g., people would like to have their bedroom (U)HDTV
display connected to a media player in the living room; therefore, a 60 GHz indoor
network for connecting all devices in indoor would be a solution.
4
1.1. Indoor Wireless Networks
• Multi-standard support.
• Quality of Service (QoS) support. crucial for interactive and streaming appli-
cations.
• Ease of use: installing and managing the network should be simple. Home
networking products and solutions target a large audience of residential users
who should not be assumed to have technological know-how.
The 60 GHz band offers an opportunity to cope with the increasing demand of
high bandwidth multimedia applications. However, the high attenuation loss of a
60 GHz channel reduces the coverage. Each antenna can normally cover only one
room, thus, a number of antennas is required to cover a whole building, at least one
antenna per room. This means that the complexity of networking and the cost of
the antenna points could become significant issues.
The massive penetration of last mile access technologies such as xDSL and cable are
proposed in the business and household environments. FTTH is one member of the
5
1. Introduction
6
1.2. The Problem Statement
2. The location of an mobile station (MS) can be used for location-based services,
improving the performance of wireless networks including network planning,
network adaptation, resource management and load balancing. In an indoor
environment, a number of location systems have been researched and developed
by universities, research centers and companies. Several artificial intelligent
techniques such as Markov chains, Bayesian networks, hidden Markov models
(HMM), and neural networks, have been proposed to address this. Thus the
challenge is how to choose the right technique and model for the requirements
of RoF networks, and apply it in an indoor environment. The HMM is a
promising prediction model for using both historical location patterns of MSs
and the location of antennas to predict the future location of MSs.
3. Because the 60 GHz cell size is very small, handoff in the proposed indoor
network is frequent. Moreover, the overlapping between two adjacent cells is
small, narrow and directional and usually in open areas such as doors [11]. In
addition, the distribution optical network has an influence on handoff delay due
to its introduced extra delay. Consequently, a handoff should be completely
processed in a small overlapping area and in a short time (of the order of a
few milliseconds). Thus, to solve this problem, a proper handoff algorithm is
7
1. Introduction
4. One of the challenges of mobile networks is the dynamic traffic due to the
mobility of users. It is difficult to guarantee QoS in terms of the call dropping
probability (CDP) and the call blocking probability (CBP). Thus, several
handoff schemes have been proposed to manage bandwidth resources efficiently.
Many of them are operating with the support of the location information of the
MS from GPS. However, GPS cannot work properly in an indoor environment.
Thus, in order to improve QoS, the handoff scheme using the indoor prediction
location system is potentially a good solution.
5. Besides the advantages of optical fibers, the optical distribution network also
introduces additional propagation delay with respect to a non-RoF wireless
network. As the requirements of the future home network, the proposed indoor
network should support multiple MAC (Media Access Control) standards such
as IEEE 802.11, IEEE 802.16, IEEE 802.11 ad and IEEE 802.15.3 c. The
feasibility of each MAC protocol to operate in a RoF network should be
investigated.
1. The dissertation introduces a novel Radio over Fiber 60 GHz Indoor Network
for advanced high speed applications. Handoff research in wireless networks
and in RoF networks is discussed and accessed. Two additional components
prediction model and positioning system are introduced to improve the handoff
performance.
3. Predicting the next location of the MS can help the system to know the direction
information of the MS for improving the handoff performance. A direction
assisted handoff algorithm is designed for RoF networks. The simulation of
this algorithm and conventional handoff algorithms has been performed in this
dissertation.
8
1.4. Thesis Outline
9
1. Introduction
Chapter 2
Radio over Fiber 60 GHz
Indoor Networks
Chapter 7
Chapter 3 Feasibility of IEEE 802.15.3c
Handoff in Wireless Networks and IEEE 802.11ad for RoF
Networks
Chapter 4
Movement Prediction in Indoor
Environments
Chapter 5 Chapter 6
A Direction Assisted Handoff Bandwidth Resource
Algorithm Management in RoF Networks
10
Chapter 2
60 GHz radio is a promising technology that can enable wireless transmission data
rates of the order of Gbps. In this chapter, we introduce the background of 60 GHz
radio, i.e. the worldwide regulations and standardization activities, and the fun-
damental propagation properties, (in Section 2.1). Then Section 2.3 gives a brief
introduction of RoF technology using Optical Frequency Multiplication (OFM).
Thereafter, by combining two technologies, an RoF network for high data rate
multimedia applications is presented in Section 2.3. Finally, Section 2.4 concludes
this chapter.
11
2. Radio over Fiber 60 GHz Indoor Networks
advantages, there are some issues of the 60 GHz band that the network designer
should consider when designing 60 GHz networks. In this section, we discuss the
basic characteristics of this 60 GHz band.
IEEE 802.15.3 c
IEEE 802.15 is a working group of the IEEE which specifies WPAN standards. It
includes seven task groups. Task Group 3 is responsible for high data rate WPAN.
IEEE 802.15.3 defines a standardization of MAC and PHY for such WPANs (11
to 55 Mbit/s, which aims to enable wireless connectivity of high-speed, low-power,
low-cost, and multimedia-capable portable consumer electronic devices [13]. IEEE
802.3 a specifies a high rate UWB PHY, an enhanced amendment to IEEE 802.15.3
for applications of imaging and multimedia. In May, 2006, the IEEE802.15.3 b
enhanced IEEE 802.15.3 to improve implementation and interoperability of the
MAC. However, these standards are still not satisfying the demand for very high
data rate applications. Thus, in March 2005, the IEEE 802.15.3 Task Group 3c [13]
was formed to develop a 60 GHz based alternative PHY for the existing IEEE
802.15.3 WPAN. This standard operates in the range of 57-64 GHz (mmWave), the
mmWave unlicensed band defined by FCC 47 CFR 15.255. It was approved in
September 2009 and called the IEEE 802.15.3c-2009. The 60 GHz based WPAN has
a coupe of valued features.
Firstly, the mmWave–based WPAN can support data rates of at least 1 Gb/s,
making it suitable for applications such as broadband Internet access, video on de-
mand, and optional data rates in excess of 3 Gb/s applications such as uncompressed
HDTV.
Secondly, the mmWave–based WPAN systems can coexist with existing wireless
communication systems such as WiFi, Bluetooth, 2 G/3 G/4 G networks, and UWB
systems due to the large frequency difference.
The IEEE 802.15.3c–2009 specifies three PHY modes for different purposes. The
first is a single carrier mode (SC) for low power and low complexity devices. The
second is a high–speed interface mode for low–latency bidirectional data transfer.
12
2.1. 60 GHz Radio Background
And the third one is an Audio/Video mode for uncompressed high definition video
and audio applications.
WirelessHD
The WirelessHD Consortium was formed by leading technology and consumer
electronics companies, such as Intel Corporation, LG Electronics Inc, Panasonic
Corporation in 2006 to develop the first industry–standard next generation wireless
digital interface specification for consumer electronics, PC, and portable prod-
ucts [14]. Their main focus is to enable wireless HDMI for streaming compressed
and uncompressed Audio/Video at resolutions of up to 1080p. It may be the first
global standard for 60 GHz applications based on the standard IEEE 802.15.3 c. The
first version of Wireless HD specification (version 1.0) was released in January 2008.
The second version, WirelessHD specification version 1.1 published in May 2010,
specifies several capabilities including:
• Supports data rates of at 10 – 28 Gb/s, more than 20x faster than the highest
802.11n data rates,
IEEE 802.11 ad
The IEEE 802.11 is a set of standards to implement WLANs which have been
developed successfully in the commercial market. IEEE 802.11 networks can provide
data rates of up to 150 Mb/s (IEEE 802.11 n). This is less than the requirements of
high speed multimedia applications such as HDTV and definitely UHDTV. In order
to meet the high data rate demand, the IEEE 802.11 ad Task Group [15] was formed
to modify both the IEEE 802.11 PHY and the IEEE 802.11 MAC to enable it to
operate in the 60 GHz band. IEEE 802.11 ad is developed from the IEEE 802.11
VHT (Very High Throughput – IEEE 802.11 ac) study group and should compete
with Gigabit WiFi in some applications in the future. It featured devices that are
expected to be compatible with the existing IEEE 802.11 services, facilities and
network structures since it supports session switching between 2.4 GHz, 5 GHz and
60 GHz bands.
13
2. Radio over Fiber 60 GHz Indoor Networks
The final IEEE 802.11ad-2012 specification was published in December 2012. This
standard is expected to reach the market sometime in 2014.
ECMA
Many countries and regions in the world have their own regulations about the
unlicensed frequency bands around 60 GHz. The unlicensed frequency allocations at
around 60 GHz is not the same in each region, but there is substantial overlap; at
least 3.5 GHz. Currently, the US and Canada, the European Union, Japan, South
14
2.1. 60 GHz Radio Background
Canada 57.05-64
EU 57-66
Japan 59-66
Korea 57-64
Australia 59.4-62.9
57 58 59 60 61 62 63 64 65 66 67
Figure 2.1: The unlicensed frequency bands around 60 GHz in different regions in the world.
Korea, and Australia have approved an unlicensed spectrum allocation in the 60 GHz
region. The situation is illustrated in Figure 2.1.
Transmit power
The emission power is also regulated differently by the different countries similar to
the spectrum allocation. In Europe, the Equivalent Isotropic Radiated Power (EIRP)
of the transmitter is limited to 57 dBm. 500 mW (27 dBm) is the peak transmit power
imposed by the USA and Canada. The average and the peak transmission power
densities at 3 m distance should be controlled below 9 and 18µW/cm2 , equivalent
to EIRP of 40 and 43 dBm. In Japan, the average transmission power should be
less than 10 mW (10 dBm) and the transmit antenna gain is limited to 47 dBi. In
Australia the peak transmission power and EIRP are limited to 10 mW, and 52 dBm,
respectively. In worldwide regulations, the transmission power is limited to 10 dBm
and the EIRP is limited to 25 dBm for portable devices and 37 dBm for mobile
devices.
Oxygen absorption
15
2. Radio over Fiber 60 GHz Indoor Networks
Wavelength
The wavelength of the 60 GHz band is around 5 mm. This is an advantage. 60 GHz
antennas are small and can be easily integrated 60 GHz transceivers for portable
electronic devices. Moreover antenna-arrays for beamforming are also feasible.
Multi-path channel
Why is the multi-path channel considered for the 60 GHz band? The main reason is
the strong multi-path behavior of the 60 GHz band. Typical root mean square delay
spread in an office is 10 to 20 ns and excess delay spread for 30 dB attenuation is
70 ns [19]. This multi-ray propagation model considers the effects of reflected com-
ponents. The propagation losses with the presence of N single reflected components
and M double reflected components are given by the Eq. 2.1:
N M
1 X ri i∆φi
X r .r
ja jb j∆φi
Lp (d) = 68 + 20log + e + e (2.1)
d0 i=1 di di
j=1
where 68 dB is the free path loss at 1 m distance; d0 is the path length of the
direct component and di . ri is the reflection coefficient of i single reflected ray, rja ,
rjb are the reflection coefficients of the j double reflected rays on a and b surfaces.
∆φi = 2π 2π
λ ∆li and ∆φj = λ ∆lj are the phase differentials between the direct ray
and the reflected rays with ∆li and ∆lj the differential path lengths between the
direct and the i single and j double reflected rays, and λ is the wavelength (5 mm).
However, the third and fourth reflected components do not contribute to the
average power at 60 GHz because of high attenuation loss. So, we should reduce the
number of rays by ignore such components in the considered model. The two-ray
model is one case of the multi-ray model and it is usually used for many simulations
due to it is lower complexity. It may be enough to describe the indoor signal
propagation without affecting the accuracy.
Coverage
At 60 GHz there is much more free space loss than at 2.4 GHz or 5 GHz. The Friis
free-space propagation model [20] shows the relationship between the transmit power
16
2.1. 60 GHz Radio Background
where Gt , Gr are the transmit and receiving antenna gain respectively, and d is the
distance between the transmit antenna and the received antenna. λ is the wavelength
(5 × 10−3 m). Therefore, the free-space path loss at 1 m for 2.4 GHz, 5 GHz and
60 GHz are consequently 40.05 dB, 46.42 dB, and 68 dB. In principle, this high path
loss can be compensated by the use of antennas with more directive patterns.
Walls also may attenuate millimeter waves significantly. This attenuation depends
on the type of wall material and its thickness (Table 2.1). The propagation path
attenuation due to glass is in the range from 3 dB to 7 dB, and due to 15 cm thick
concrete it can be 36 dB. This high penetration loss for most materials in the 60 GHz
band is an advantage because it limits interference and isolates from 60 GHz radios
in adjacent rooms. But it is also a disadvantage, because any obstacle can block
a line-of-sight or alternative path and produce strong shadow fading. Now, let us
estimate the range of the 60 GHz channel. At first, the received power should be
greater than the receiver sensitivity, which includes thermal noise power N , shadow
margin M , and the needed signal-to-noise ratio SN R as the following equation:
Pr ≥ N + M + SN R (2.3)
Secondly, considering a signal with bandwidth B = 100 MHz, the noise floor power
is -94 dBm. The thermal noise power N is calculated by the equation: N = F kT B,
where F = 10 is the noise figure of the receiver, k is the Boltzmann constant, and
T = 290K. Shadow-margin M should be taken into account, which is assumed 10 dB
on average. The transmitting power is Pt = 100 mW EIRP (20 dBm) according to
European regulations. Assuming Eb /N0 is higher than 10 dB to achieve a bit error
rate (BER) of 10−3 .
17
2. Radio over Fiber 60 GHz Indoor Networks
Using both equations Eq. 2.2 and Eq. 2.3, with the antenna gains equal to one,
we can estimate the range of the 60 GHz antenna (or access point) is approximately
25 m. Hence it is suitable for short-range indoor communication. Therefore, in order
to deploy a 60 GHz network, it requires a large number of antennas which operate as
access points, which means that the cost of 60 GHz deployment might be a challenge.
Due to the limits imposed on the emitted power, the high temperature noise, and
the high oxygen absorption, the range of 60 GHz system is short. The propagation
of 60 GHz signal is easily obstructed by the movement of people, and the presence
of furniture and walls. This however opens up the opportunity for spatial reuse of
channels. Channels in the indoor or open areas show a strong multipath behavior
because of easy reflection. So 60 GHz is usually envisaged for communication confined
to a room or an open area where LOS signals from the antennas can be expected.
In Fig. 4.6(b), the comparison of the coverage between the 60 GHz band and the
2.4 GHz band has been illustrated using the measured data. The 2.4 GHz system
requires only three antennas or access points (AP) to cover the whole building while
for the 60 GHz band we need at least one antenna per room. In [21] it is shown
that the signal to interference ratio (SIR) can drop from 20 dB to 0 dB within a few
centimeters in the indoor environment (Fig. 2.3).
(a) 2.4(a)
GHz2.4 GHz (b) 60 (b)
GHz60 GHz
(a) 2.4 GHz (b) 60 GHz
Figure 2.2: The coverage of 60 GHz compared with that of 2.4 GHz in the same building
The detailed 60 GHz wireless physical layer and antenna designs are out of the
scope of this dissertation. For more details on 60 GHz PHY, readers please refer
to [8, 22]. In the rest of this thesis, we assume that a non-obstructed room with the
18
2.1. 60 GHz Radio Background
Figure 2.3: The variant SIR along the hallway path [21]
Applications that might benefit from the advantages of 60 GHz are present briefly
bellow.
19
2. Radio over Fiber 60 GHz Indoor Networks
Vehicle-to-highway Communication
The leisure traveler can receive traffic information (such as traffic jams, road ac-
cidents) from audio as well as video in high quality. They also want to access a
limitless library of music and travel information or play on-line games. The use of
millimeter waves also appears to be convenient for inter-vehicle or vehicle-to-fixed-
infrastructure communication [24]. In [25], the proposed infrastructure is providing
high-bandwidth communication services with a uniform user interface independent
of the location or speed of the user.
Railway Communication
In a modern railway transportation system, not only efficient communication be-
tween railway traffic control and the trains in the network is vital, but also the
communication needs of the passengers with the external world is important. Rail-
way communication these days is about high speeds and high density because of a
high number of passengers. During their journey, the demand of entertainment (for
example playing online game, Video on Demand (VoD)) or exchanging information
with the external world is high. They require the telecommunication operators to
provide the large bandwidth (in orders of Gbps) and high QoS. Thus, 60 GHz is
proposed to be used in railway communication [26]. With such proposed architecture,
the passengers should be able to have access high speed Internet.
Aircraft Communication
What do you do on the long journey on the flight? The demand of reading news,
watching films, playing games on the fight is growing higher day by day. An in-
flight entertainment system for passenger aircraft requires a total data rate of the
order of several Gbps with user densities of up to one thousand passengers per
system. A well designed 60 GHz WLAN system is very appropriate for in-flight
entertainment [27, 28]. In this design [28], each antenna serves 36 seats (passengers),
antenna diversity schemes are investigated.
Inter-satellite Communication
The use of the 60 GHz band for inter-satellite links has been investigated by The
National Aeronautics and Space Administration (NASA) and presented in [29]. Since
satellite is totally outside the atmosphere, signals are subject to only free-space
propagation loss. Thus, the 60 GHz link is working well. The proposed application
uses the 60 GHz band for transmission between two or more satellites to relay signals
from two stations on earth which are too far to be served by only one satellite.
20
2.2. RoF technology
The wired broadband services are mainly driven by the platforms of DSLs, cable
and fiber-optic. DSLs have still been the dominant technology for broadband access,
and a variety of improvements of DSLs have been proposed that will increase the
data rate and line-length limitations. However, in recent years, there has been a
remarkable shift from DSLs to optical fiber since the high demand of high speed
Internet multimedia services as well as the rise in services and pricing competition
among operators and service providers (Fig. 2.4).
The point-to-multipoint optical network known as Passive Optical Networks
(PONs) is the dominant broadband network today. The simple PON includes
a telecoms central stattion (CS) connecting to subscribers by using a couple of
wavelengths for downstream and upstream from optical line terminal (OLT) to
optical network units (ONU) and vice versa. The first PON activity was initiated
in the mid-1990s by the Full Service Access Network (FSAN) group [30]. It was
soon standardized by ITU – Telecommunication Standardization Sector, the ITU-T
Recommendation G.983 series of standards. Based on current standards, a PON
provides higher bandwidth for data applications than DSL and Cable Modem, as
well as in a larger area (it can be maximum of 20 km from the OLT to the ONU).
Three drivers that make the large deployment of PONs around the world have higher
bit rates, high service capacity and greater service integration.
Current PON standards such as Broadband PON (B-PON) and Gigabit-capable
PON (G-PON) is bases on time division multiplexing (TDM) with fiber reach up
to 20 km (loss budget of 28 dB with typical split ratio 1 × 32 or smaller. B-PON is
specified in ITU-T Recommendation G.983 series [31] offering numerous broadband
services including Ethernet access and video distribution. G-PON is specified
in ITU-T Recommendation G.984 series [32]. The G-PON is a PON technology
operating at bit-rates of above 1 Gb/s. Its fiber can reach at least 20 km with a
logical support within the protocol of 60 km. The recommendations can support
various bit-rate options using the same protocol, including symmetrical 622 Mb/s,
symmetrical 1.25 Gb/s, 2.5 Gb/s in downstream and 1.25 Gb/s in upstream and
more. The IEEE 802.3 ah (E-PON) is the IEEE standard of Gigabit-Ethernet PON
(1 Gb/s) [33]. In early 2006, work began on an even higher-speed 10 Gbps/second
Ethernet Passive optical network (XEPON or 10G-EPON) standard, ratified in 2009
as IEEE 802.3 av [34].
21
2. Radio over Fiber 60 GHz Indoor Networks
22
Figure 2.4: Fiber-To-The-Home (FTTH) penetration in Dec. 2011 (source: FTTH Council, published in Feb. 2012)
2.2. RoF technology
A simple RoF system includes three main components (depicted in Fig. 2.5): (i) a
CS controls the radio access, generates signal and process signal, (ii) antennas that
contain only RF modules and (iii) the optical distribution network. The CS manages
all signal processing, generating as well as centralized controlling of resources. It can
help in reducing power, optimizing bandwidth utilization and channel allocation.
The antenna is simply to repeat the radio signal and convert signal from electrics
to optical and vice verse. Each antenna has two parts: the first part for downlink
including photo-diode circuity (O/E) and RF power amplifier (PA), the second part
for uplink consisting of laser diode circuity (E/O) and low noise amplifier (LNA) [35].
The radio protocol stack resides in the CS. Then, an end-to-end logical link between
the CS and the MS is established. To communicate with external networks, the CS
is playing a role as a gateway to outside networks such as Internet, Mobile Network,
public switched telephone network (PSTN) and media resources.
In addition to the high capacity of fiber optics and the wireless access flexibility,
in RoF systems, all management algorithms can be carried out in the CS in the
centralized way. This centralized way is simpler to allocate bandwidth, plan cells, and
increase the efficient use of resources. The next advantage of centralized processing
is that the CS can easily update and monitor the status of network load which
is dynamic since the mobility of the MS. Based on that information, the CS can
implement bandwidth allocation algorithms to improve resource efficiency. The
other advantage of the RoF system is its ability to upgrade and maintain network,
firmware, management policies at the CS.
However, when operating at the millimeter-wave bands such as 60 GHz, the
RoF techniques should consider the major challenge of generation and delivery of
the millimeter-wave signals at the AS while keeping the link simple. During the
last few years, several methods have been proposed to address this issue [36–39].
The researches proposed in [9, 36–38] mainly focus on improving the performance
of RoF transmission of microwave carriers and radio signal modulation formats.
However, it is not enough for an RoF system. The RoF system requires a reliable
infrastructure supporting multiple wireless accesses and also having a mechanism to
adapt dynamically the radio link. The work proposed in [39], which is the extension
of the method in [38], the OFM is a promising method to meet the above demands
of the RoF system. The Orthogonal frequency-division multiplexing (OFDM) is
chosen to design the radio part. The following section introduces the principle of
OFM.
23
2. Radio over Fiber 60 GHz Indoor Networks
Central Office
Internet ’ Antenna 1
PSTN
Mobile n Distribution
Controller Tranceivers Fiber
Network ’
Cable TV ’ Antenna 2
... n
’
... ’n
’n
Antenna n
CS AS radio signals
fsw radio signals
fRF = n.fsw ± fsc
fsc
fiber link
IM-
data
MZI BPF
24
2.2. RoF technology
cell capacity allocation, multi-standard support, dynamic radio link adaptation and
remote antenna controlling.
The desired fRF signal can be selected with an adequate bandpass filter. On
the condition that the maximum RF bandwidth, fsw /2, is not exceeded, multiple
wireless signals can be transmitted simultaneously in a sub carrier multiplexing -
SCM scheme [39] (Fig. 2.7). Therefore, the obtained radio signals at the AS can be
selected at the same or at the different value of harmonic bands. It is possible to
increase the cell capacity without the need of installing the additional transceivers
(TRX) which are normally costly.
Multi-standard Support
In [39], the feasibility of the multi-standard wireless access support by one OFM
link is proved. The desired fRF signal can be selected with an adequate bandpass
filter through a proper selection of fsw and fsc at the CS. By this way, different
wireless standards can be simultaneously transmitted to the antenna.
25
2. Radio over Fiber 60 GHz Indoor Networks
• Dynamic Frequency Selection at the CS. The link adaption can be easily
performed from the CS by tuning low frequency sub-carriers.
• Transmit Power Control at the AS. The transmit power can be remotely
adjusted at the AS or at the CS. In order to simplify the AS, the transmit
power adaptation can be done by the CS following the power regulation
procedure. However, the network throughput decreases.
The HCC plays a role as the CS. It has interfaces with the external networks
and resources. The HCC also does all the functions of generating and processing
signals. As mentioned before, the management algorithms, network optimization
and resource managements are performed by the HCC.
Antennas (AT)
These are similar with the AS discussed in Section 2. These ATs are simplified since
all functions of signal processing are transferred into the HCC. In the proposed
system, at least one AT is required to cover a room in the building.
The optical distribution network is used for transmitting signal between the HCC
and the ATs. The length of fiber link is dependent on the Radio over Fiber technique
we choose. With the discussed OFM technique, the fiber can be extended to more
than 20 km [39].
26
2.3. RoF 60 GHz Indoor Networks
Internet
Mobile Networks
Database
Figure 2.8: The proposed Radio over Fiber indoor network operating at 60 GHz
Positioning system
In order to predict the movement of the MS, the network should have the historical
data of the movements of the MS. This can be done by deploying a positioning
system. The positioning system is required to work well in an indoor environment.
The positioning system is presented in the section below.
Movement Prediction
To exploit the historical data of the past movements, the prediction model should be
carried out at the HCC. Up to date, several prediction models have been proposed.
In Chapter 4 of this thesis, the Artificial Intelligent techniques based models are
investigated and the Hidden Markov Model is developed to adapt with the proposed
system.
27
2. Radio over Fiber 60 GHz Indoor Networks
28
2.3. RoF 60 GHz Indoor Networks
each antenna could have more than one pair of wavelengths. Each MS has its own
unique address (M Sq ). To communicate with other MSs and to fix the location
of the MS, HCC has to update the other table containing the address of the MS
and the corresponding antenna to which the MS is connecting to (Table 2.4). The
coordinates of each MS are stored and updated in the database in the server as
Fig. 2.8.
29
2. Radio over Fiber 60 GHz Indoor Networks
Antenna ID MS address
ATi M Sq
ATi M Sp
ATj M So
• The first group is based on the principle of triangulation, using the geometric
properties of triangles to estimate the target location. The methods locating
the target location by measuring the distance from multiple reference points
are lateration methods. These methods do not measure the distance directly.
But the distance is derived by using the received signal strengths (RSS), time
of arrival (TOA) or time difference of arrival (TDOA), and roundtrip time of
flight (RTOF) or received signal phase methods. Other methods in the first
group are computing the angles relative to multiple reference points to locate
the target location, called direction of arrival (DOA).
• The last group locate the location of a target object using proximity technique.
The proximity technique relies on a number of detectors at the known locations.
If the observed object is detected, it is considered to be in the proximity area
covered by the detector.
30
2.3. RoF 60 GHz Indoor Networks
to signal instability, noise from hardware and environmental factors like humans in
motion. The update rate of these methods is in the range of a few seconds. For
example, RADAR positioning system [44] was proposed by a Microsoft research
group. This system uses the fingerprinting technique. In the offline phase, the
RADAR system collects signal strength and signal-to-noise ratio. The alternative
method is using the propagation model with wall and floor attenuation factors. In
the online phase, the multiple nearest neighbors in signal space (NNSS) location
algorithm was proposed. The accuracy of the RADAR system is about 4 m with
about 50% probability. Ekahau1 is the commercial location system. The offline
phase or site survey is done by a software tool, which demonstrates the network
coverage area, signal strength, SNR and the overlapping of the WLAN. A central
location engine is used for the online phase. The Ekahau system can achieve the
accuracy of 1 m if there are three or more overlapping access points to locate a target
object.
In the second category, a new installation of positioning system is required. These
can localize the position of the mobile user with high accuracy and a high degree
of precision. In addition, the update rate of these systems is short, even achieving
real-time responses. The wireless technologies used in the second category can be
UWB, Bluetooth, RFID, IR, or wireless sensor networks. For example, the RFID
positioning system [45] enables flexible and cheap identification of an individual
person or device. In [46], the sensor-based location system is built from a large
number of sensors, which are located in the predefined locations. The problem
with sensors is a limited processing capacity and battery power. Ubisense2 is a
commercial positioning system. The Ubisense system is based on UWB technology
to locate a target object in real-time. The performance of the Ubisense system is
about 99% precision within 30 cm.
In the proposed system (Fig. 2.8), we assume that the positioning system with
a short update rate, high precision and accuracy such as Ubisense is installed to
collect the location information periodically. The information should be collected
daily, and classification is to be made as weekdays and weekend, working days and
vacations. However, this dissertation does not consider this information. Tab. 2.5 is
an example of location information format.
1 Ekahau, http://www.ekahau.com
2 Ubisense, http://www.ubisense.net
31
2. Radio over Fiber 60 GHz Indoor Networks
2.4 Conclusion
In this chapter, the Radio over Fiber 60 GHz indoor network has been proposed. At
first, the characteristics of the 60 GHz band have been presented. The 60 GHz range
offers a wide unlicensed spectrum of 5–7 GHz. It is useful for broadband wireless
networks where the demand of high data rate applications such as (U)HDTV or
online gaming is high. The 60 GHz band also offers the advantage of security of
eavesdropping data since the coverage of one antenna is usually in one room. With
its attractions, several standards of 60 GHz have been published. In order to deploy
60 GHz widely, the combination of wireless and optical communication through RoF
technique is a promising solution by taking advantages of wireless flexibility and the
high capacity of optical communication.
Secondly, the RoF using OFM presented in this chapter can support a fiber length
of up to 20 km. The benefits from the selected techniques are simplifying antennas,
supporting multiple standards, network capacity and dynamic radio link adaptation.
The basic operation of the RoF system is also presented. The OFDM is chosen to
design the radio part. This technique can work well in broadband wireless networks
operating at a high frequency like 60 GHz. It can eliminate the ISI by applying the
cyclic prefix.
However, the limited coverage of 60 GHz antenna and the extra delay caused by the
distribution optical network make a handoff difficult in the proposed network which
is presented in Chapter 3. In order to solve the issues, the proposed RoF network are
employing two additional components: positioning system and movement prediction.
The location information is used for multiple network purposes including network
planning, managing resource and improving handoff performance.
32
Chapter 3
3.1 Introduction
33
3. Handoff in Wireless Networks
3.2.1 General
34
3.2. Handoff in Wireless Networks
and soft handoff - where both the existing connection and the new connections are
used while the handoff takes place:
• The hard handoff that occurs in the current cell is called intra-cell handoff,
and the hard handoff that occurs when a MS moves into another cell is called
inter-cell handoff. This hard handoff is usually used in Frequency division
multiple access (FDMA) and Time division multiple access (TDMA) where
the MS can be connected with at most one base station at a time.
• One can distinguish between two types of soft handoffs: multi-way soft handoff
and the so-called softer handoff, when two BSs exist in two sectors of a
sectorized cell [48].
In principle, soft handoff can be used with any radio technology, however the
cost may be high and the support for soft handoff may not be good for particular
technologies. This is the reason that soft handoff is commonly used only in Code
division multiple access (CDMA). Thus, the soft handoffs will not be discussed
further in this thesis.
35
3. Handoff in Wireless Networks
The vertical handoff process can also be divided in to three phases: i) Handoff
Initiation: in this phase usable networks and available services information is
gathered; ii) Handoff Decision: this is also called network selection. During this
phase, the handoff algorithm determines which network the MS should choose based
on multiple parameters; iii) Handoff Execution: connections are re-routed to the
new network in a seamless way.
Even though vertical handoff has the same basic principles of handoff discussed
until now, it also deals with other factors that must be considered in the handoff
decision phase for effective network usage [51]. The parameters used for handoff
decision can be listed as follows: (1) network parameters including coverage, band-
width, load balancing, latency, RSS, SIR, monetary cost, security and connection
time; (2) terminal parameters: velocity, battery and wireless interfaces; (3) user
preferences: profile, cost (billing plans) and current user conditions; (4) service
capabilities and QoS.
36
3.2. Handoff in Wireless Networks
Measurement
Handoff
Algorithm
No
WLAN Yes
available?
Connected to yes
Registered to Registered to
Yes
Connected to
GSM GPRS WLAN WLAN
No
No
Associate with
Register to GSM
WLAN
• The latency of the handoff must be low. The hard handoff should be completed
in the duration time while the MS passes through the overlap between the
original cell and the target cell. If not, the ongoing call may be interrupted,
and the MS would not be happy, of course. Thus, the handoff should be fast
enough so that the user cannot detect any degradation of service or interruption
during a handoff.
• Minimizing the effect of handoff on QoS. Handoff ensures the mobility of the
MS. However it also effects on the quality of service (QoS) due to it causes call
dropping. For example, the probability of a new call blocking and dropping a
current call should be minimized; and the traffic between the adjacent cells
should be balanced.
37
3. Handoff in Wireless Networks
In order to achieve the desired features of handoffs, the designers must take into
account the factors that affect the handoff process. Let us discuss them here:
• Wireless technology: The wireless technology used in the wireless network also
decides the cell size, the topology and the traffic model. Nowadays networks
may employ multiple wireless technologies simultaneously, for example: Blue-
tooth, WLAN, GSM and UMTS. This leads to a new requirement of handoff
procedures between two different wireless technologies (vertical handoff).
• Cellular structure:There are five types of cells ordered by size: umbrella cell,
macro-cell, micro-cell, pico-cell and femto-cell. When the cell size decreases, for
a given MS mobility scenario, handoff is more frequent and the time constraint
will also be more stringent.
• Mobility: The speed and direction of the user also affects the handoff. When
the user moves fast, time for handoff is short, therefore, a handoff may not be
successful. Hence, the handoff algorithms should deal with this requirement.
They should realize the movement features of the MS and adapt with them.
• QoS: Whenever the QoS decreases such as the BER, bandwidth or packet loss,
the handoff may be requested to connect with another BS that can guarantee
the QoS.
38
3.2. Handoff in Wireless Networks
Received
Signal BS1 BS2
Strength
h
T1
T2
time
A B C D
Figure 8. Received signal strength when MS travels in the handoff area
Figure 3.3: Received signal strength when MS travels in the handoff area
In this method, the RSSs of BSs are measured over time and the BS with the strongest
signal strength is selected to carry out a handoff. For example, when a MS moves to the
point A (Figure 8), the RSS of BS2 exceeds the RSS of BS1 and the MS will be handed Deleted: F
thatover to BS
of BS . If the
1 , 2the MSRSS of BS2 falls
is handed backbelow
to BS that
1 . of BS1to
Due , themultipath
MS is handled back
fading theto signal
BS1.
Due to multipath fading the signal may fluctuate, several handoffs
may fluctuate, several handoffs may be requested while BS1 ’s RSS is strong enough may be requested
while BS1’s RSS is strong enough to communicate with the MS. This results in the
to communicate with the MS. This results in the undesirable so-called ping-pong
undesirable so-called ping-pong effect. Handoff in this case is unnecessary. These
effect. Handoff handoffs
unnecessary in this case
will is unnecessary.
increase These of
the probability unnecessary
a call failinghandoffs
during a will cause
handoff andan
the network
increase in theload.
CDPThe andhandoff techniques
the network to overcome
load. The handoffthis problem haveto
techniques been introduced
overcome this
in [3] have
problem and theybeen are: Relative Signal
introduced in [47]Strength
and theywithare:
Threshold,
Relative Relative
SignalSignal Strength
Strength with
with Hysteresis,
Threshold, Relative Relative
SignalSignal Strength
Strength withwith Hysteresisand
Hysteresis, and Relative
Threshold.Signal
Let usStrength
explain
these techniques:
with Hysteresis and Threshold.
- Relative Signal Strength plus Threshold: The handoff is requested if and only if
The performance
the RSS of BS of handoff algorithms using both absolute signal strength and
1 is lower than the threshold value and the RSS of BS2 is stronger
relative signal
than that of BS1. In Figure 8, T1are
strength measurements explained
is set in [52]. value
as the threshold The authors use isExpected
and handoff carried Deleted: F
out at the
Average Signal point B.of a base station for handoff decisions. This scheme helps
Strength
in avoiding unnecessary handoffs. In [53] a model is proposed to analyze the handoff
This technique helps the network limit handoffs when the signal from the current
algorithmsBSusing
(BS1)both absolute
is strong enough. and
Butrelative signal
the problem strength
is to find the measurements. The
appropriate threshold
benefit of value.
these Iftechniques is, it is simple to implement, thus making them popular
the threshold is too low, the call may drop. If it closes to the point A, the
ping-pong
in the cellular systemseffectsuch
can still persist. UTMS and WLAN.
as: GSM,
The -network
RelativecanSignal Strength
use SIR as a with Hysteresis:
parameter This technique
to decide a handoff. allows
This handoff
method if BS2’s
allows
RSS is stronger than BS
handoff if the SIR of the current BS is lower than the threshold and the SIR ofthis
1 ’s with a hysteresis value h. Figure 8 illustrates the Deleted: F
15
39
3. Handoff in Wireless Networks
proposed Minimum Power Handoff (MPH) [55], the system looks for a pair of BS and
a channel that has a target SIR with minimum transmitted power. The simulation
results show that this algorithm reduces call dropping, but causes a high number of
unnecessary handoffs. In order to reduce the number of handoffs, handoff requests
are delayed by using a handoff timer in MPH (MPHT). If the MPHT finds the best
BS that differs from the currently serving BS, the maximum SIR at the current BS
is calculated. If it is less than the minimum required SIR, handoff is carried out
immediately; otherwise the timer is activated. Handoffs are delayed until the timer
reaches a given value. When the handoff is executed, the algorithm turns off the
timer.
Moreover, the mobility characteristics of a MS, velocity and directional information,
can be useful parameters for improving a handoff. If the user moves fast, the
probability of call drop may be high due to excessive delay during handoff. In [56] a
fast handoff algorithm with velocity adaptation is proposed for urban communication.
Velocity estimators using LCR, ZCR and Covariance methods are used. Corner
detection algorithm is also incorporated into the handoff algorithm to speed up
handoffs in NLOS scenarios. The paper shows that a velocity adaptive handoff
algorithm performs well in a typical NLOS scenario of an urban cellular network
by reducing handoff delay. In [54], the authors propose a velocity adaptive handoff
algorithm using a multi-level threshold. The handoff threshold values are assigned
according to the velocity of the users. Since low speed users spend more time in the
handoff area, they are assigned higher thresholds. High speed users are assigned
lower thresholds. The simulation result in [54] shows that when the number of levels
for a threshold is increased from one to eight levels, the blocking probability and
handoff failure probability is significantly reduced.
Direction biased algorithms are important for high mobility users, especially in
a NLOS handoff [57]. In this algorithm, base stations are divided into two groups.
Group-1 has the base stations that an MS approaches and Group-2 has the base
stations from which the MS moves away. Two hysteresis values are used for handoff
initiation: he for Group-1 and hd for Group-2 where he < h < hd . The direction
biased algorithm improves handoff performance by lowering the mean number of
handoffs while reducing the handoff delay.
40
3.3. Handoff in 60 GHz Network
Single frequency networks (SFNs) is a broadcast network of transmitters that use the
same frequency to transmit information. In [21,61], the authors extend the concept of
Virtual Cellular Network (VCN) for 60 GHz communication with multiple receiving
antennas to form a MIMO system. The VCN architecture utilizes the idea of SFNs
and distributed Access Points (AP) to form an adaptive wireless infrastructure. In
a VCN, there is no conventional Base Station (BS) that manages the channel and
handovers. Instead, the notion of “ports” - essentially simple antennas - has been
introduced. In a network area, all the ports are connected to and controlled by a
Port Server (PS). In this system, a Virtual Cell (VC) is dynamically formed for
each and every MS. It is defined as the area in which the signals sent from the MS
are strong enough to capture a port. Packets destined to an MS are dynamically
routed by the PS to all the ports inside the VC. Since the network operates at a
single frequency channel and a VC is always created to follow an MS, there will
be no conventional handover. Each time the MS moves to a new position, a new
VC is created and the routing table must be updated in the PS. The drawbacks
are twofold, first the whole spectrum is shared by a large number of users; second,
higher traffic overhead to handle the ports dynamically.
Extended Cell
The authors in [23] have proposed the RoF indoor network employing the novel
concept of Extended Cell (EC). ECs are dynamically created by grouping several
adjacent radio cells. This means adjacent antennas (referred as radio cells) are
allowed to transmit the same content over the same frequency channel. Multipath
and shadowing effects in such networks are solved by employing OFDM technology.
41
3. Handoff in Wireless Networks
Now the overlapping area is the overlap area between two adjacent ECs in RoF
indoor networks, so the handoff performance is improved. The simulation results
show the number of handoffs is reduced significantly and the CDP for the case using
EC, to be less than that in the case where EC is not used, is up to 70%.
Two architects for future home networks using 60 GHz have proposed in [2]: cell-
based communication infrastructure using the EC concept and an ad hoc based
home network communication infrastructure. The main challenges in the proposed
architectures are: connectivity, mobility, self-configuration and the architecture of
the cognitive home network.
42
3.3. Handoff in 60 GHz Network
used for uplink transmission and the other M channels (fM +1 , fM +2 ...f2M ) are used
for downlink transmission. Time frame is divided into M time slots ts . Each BS
supports all channels, but using each of them in the specific time slot. Adjacent BSs
are assigned the different FS pattern to avoid co-channel interference. Each MS is
communicated with a pair of frequencies (fi , fM +i ), i = 1, 2, ..., M and a pair of time
slots (tk mod M , tk+1 mod M ), k = 1, 2, ... for uplink and downlink communication,
respectively. Thus, the MS receives downlink for every time frame. In case, the
MS moves into a new BS, it will not receive a downlink packet from the old BS.
Thus within a time, t = M ts , a handoff is confirmed, then the MS requests a uplink
transmission to the new BS. The analysis in [59] shows the proposed MAC scheme
can support QoS requirements and make a fast and easy handoff. The proposed
MAC scheme can also be used for indoor and outdoor communication at 60 GHz.
In [63, 64] the use of directional antennas for 60 GHz systems has been presented. In
these papers, MAC and routing protocol are also described. The enhanced MAC
reduces hidden node problems due to directional antennas. Handoff can take into
account the usage of directional antennas. The signal strength of the beacons sent
out omni-directionally by APs is smaller than that of packets sent out directionally
with the same transmitted power. Normally, handoff is triggered when the RSS of
the associated AP is below a fixed threshold. In a fast handoff approach, handoff is
performed when the RSS of a new AP is greater than that of the current AP plus
with a reasonable hysteresis.
Vertical Handoff
Due to the vulnerable nature of 60 GHz LOS links, vertical handoff from 60 GHz
radio to WLAN is the proposed solution to overcome that problem [65]. The handoff
decision algorithms are designed as a cognitive approach and based on decision
theory with multiple factors such as: user preference, network condition and the
capacity of the terminal. The proposed algorithms are based on the Decision Theory
and Markov Decision Process. These algorithms are to choose one of three naive
algorithms: i) algorithm “r” that chooses randomly networks, ii) algorithm “s” that
always switches to WLAN when LOS is blocked, and iii) algorithm “w” that always
waits for 60 GHz recovery. The simulation results show the proposed algorithms can
make a handoff decision to the backup WLAN well in uncertain situations when
the 60 GHz LOS is blocked by obstacles for a short period (several seconds or even
longer).
43
3. Handoff in Wireless Networks
3.3.2 Handoff
Problem Statements
When compared to the cell size of GSM, UMTS or 802.11 WLAN, the cell size of
60 GHz network is small, i.e., in the order of 10 m. Fig. 4.6(b) shows the difference
in coverage between a 2.4 GHz network and a 60 GHz network. There are only three
base stations to cover the whole area for 2.4 GHz radio, but at 60 GHz it is necessary
to install at least one antenna per room. Due to the small cell size, even a small
movement of the MS may trigger a handoff [23], and this may lead to a large number
of handoffs during a call [26, 66]. The second issue is the smaller overlapping area
between the cells. With GSM, UMTS, 802.11 WLAN, the larger overlapping area
between cells allows the network elements enough time to initiate and complete
a handoff successfully. In 60 GHz networks, the overlapping area is much smaller.
Especially in an indoor environment, the overlapping area usually exists in and
around open areas like windows and doors. When a MS exits a room through a door,
the MS can make a sharp turn (for example: turn left or turn right). In this case,
the overlapping area might be too small to give the MS sufficient time to complete
a handoff if not identified early enough. This phenomenon is termed as the corner
effect [23].
Let us get an impression of the time taken for handoff in 60 GHz systems in the
indoor environment. We assume that the speed of movement of a person is 2 m/s
and it takes 5 s to pass through a cell (room size being 10 m×10 m). This means
that for every 5 s a handoff will be performed. The time allowed for completing a
handoff with a 1 m overlapping area at an open door is thandof f = 0.5 s.
Another issue is the propagation attenuation of the 60 GHz channel by the
surrounding obstacles. Objects such as furniture can cause shadowing, and also
humans can block signals, with attenuation of signals by 20 dB from the link budget
if the LOS path is blocked. Thus handoff is rather challenging in a crowded room.
As mentioned above, SIR can drop from 20 dB to 0 dB and rise up to 15 dB again in
a few centimeters. If a handoff is triggered due to that reason, it is an unnecessary
handoff. A handoff algorithm should consider this situation.
Moreover, the design should consider the introduction of the extra fiber propagation
delay of the optical distribution network. Propagation delay in air is much less than
the propagation delay over fiber. Since fiber limits the speed at which messages
can travel, roughly by a factor of 75% of the speed of light, the propagation delay
in fiber can be estimated to be 5 ns per 1 m of fiber [67]. This delay also have an
influence on handoff delay in an RoF network. If a handoff delay is long, a call may
44
3.3. Handoff in 60 GHz Network
be interrupted.
As mentioned above, the target applications of 60 GHz systems have high bandwidth
capacities with high required-QoS. 60 GHz systems have small overlapping areas
where handoff is carried out. Hence, triggering handoffs, completing handoffs in a
short time and reducing unnecessary handoffs are very important. Since the received
signals from BSs vary irregularly, an MS can switch links with either BS, called
ping-pong effect. This effect results in many unnecessary handoffs between two BSs.
Before discussing handoffs in the 60 GHz band, we will first show the simulation of
conventional handoffs in the RoF network.
The C++ simulation program for 60 GHz in [68] has been modified and enhanced.
In this tool, two propagation models are included: two-ray propagation model
and free-space as well as using the coverage generated by Radiowave Propagation
Simulator (RPS) [69]. The cell range is 20 m, the velocity of an MS is in the range
from 0.25 m/s to 2 m/s. A two-ray propagation model is used. The movement
of a MS is modeled using a random walk mobility model with reflection. In this
simulation study, we evaluate the performance of conventional handoffs in terms of
the number of handoffs, and the number of unnecessary handoffs.
In the first case, the velocity of the MS is 1 m/s and the overlapping area is set
such that it would be enough to complete a handoff. When the hysteresis h is 0 dB,
there are 25,687 initiations of handoffs, out of which there are 15,912 unnecessary
handoffs. This means that more than 61% handoffs are not necessary.
In the second case, the overlapping area is as small as 5 m. Fig. 3.4 and Fig. 3.5
show the percentage of successful handoffs and the percentage of failed handoffs
versus the mobile speed for two hysteresis levels, 0 dB and 5 dB. For the hysteresis of
0 dB and increasing the speed of the MS from 0.25 m/s to 2 m/s, the percentage of
successful handoffs sinks from 99.9% to 97%, and the percentage of failure handoffs
increases from 0.01% to 2.8%. The percentage of successful handoffs significantly
drops from 98.6% to 89.9% in cases where the hysteresis is 5 dB. This means that
the delay in handoff increases. The number of handoffs per call and the percentage
of ping-pong handoffs in case of 0 dB hysteresis are much higher than those for the
hysteresis of 5 dB as shown in Fig. 3.6, and Fig. 3.7.
The simulation results show that if the hysteresis is small, the number of unnec-
essary handoffs is high. If the hysteresis is set high, the number of unnecessary
handoffs is low, but it increases the handoff delay, and reduces the number of suc-
45
3. Handoff in Wireless Networks
100
98
Percentage of Successful Handoff (%)
96
Conv−h=0dB
94
Conv−h=5dB
92
90
88
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
cessful handoffs. Thus traditional handoff algorithms with fixed threshold or fixed
hysteresis are not suitable for 60 GHz indoor networks. In order to keep the number
of handoffs and the handoff delay as small as possible, the adaptive threshold or
hysteresis should be applied.
Next, with the same hysteresis, the number of successful handoffs depends on the
velocity of a MS. A handoff can fail because of a fast moving MS. One thing we should
mention is the corner effect. In cellular communication, a fast handoff algorithm is
proposed to solve this problem [70]. In case of unstable SIR [21], fast handoffs might
not be a good solution due to the high number of handoffs. This requires that the
handoff algorithms must deal with these quick changes, and shadowing problems.
As discussed above, the velocity adaptive handoff algorithms and direction-based
handoff algorithms also perform well in this situation. Especially in the indoor
environment at 60 GHz, the corner effect is more critical since overlapping areas
exist only in open areas such as doors and windows. Thus it is better to use
additional information of a MS such as velocity, direction, and location for making
a handoff decision. Using multiple thresholds according to the velocity of the MS,
handoff algorithms perform better in terms of forced termination and call blocking
probabilities as presented in [54].
46
3.3. Handoff in 60 GHz Network
12
10
Percentage of Failed Handoff (%)
Conv−h=0dB
6
Conv−h=5dB
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
6
Average Number of Handoffs per Call
Conv−h=0dB
4
Conv−h=5dB
1
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
47
3. Handoff in Wireless Networks
60
50
Percentage of Ping−pong Handoffs (%)
40
30
20
10
Conv−h=0dB
Conv−h=5dB
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
Besides the movement of a MS, the cell size and the overlapping area are also
important factors affecting handoffs in 60 GHz systems. Thus, improving the
overlapping area is also a good solution. To do that, “Extended Cell” in [23] and
Virtual Cellular Network [21] can help. However, defining when and how to form an
EC is still a question while designing such networks. Moving Extended Cell [62] and
Moving Cell in [26] can avoid handoffs by moving together with the MS. All above
concepts should be incorporated in the applications; namely the outdoor directional
networks such as highway or train communication and also in the indoor networks
where there are many people moving around.
Next, it is possible to improve handoff algorithms by appropriate handoff schemes
and resource management. Effective resource allocation schemes such as tuning
the MAC protocol are also good ways to solve handoff issues of 60 GHz radio. The
Chess Board MAC protocol proposed in [59] shows that the resource allocation can
help make the handoff process successful and fast. By predicting a handoff of a MS
or using AI technologies, we can know the target base station and reserve resources
for that MS. This would reduce the handoff delay as well as optimize the resource
management.
Finally, vertical handoff could also be carried out to improve handoffs in the
48
3.4. Conclusion
multi standards 60 GHz systems. [2] shows that the future home networks using the
60 GHz band can provide the users with a variety of bandwidth-level services based
on multiple technologies like GSM, WLANs and Bluetooth. This means that the
appropriate vertical handoff schemes between the above technologies can be a key for
solving handoff issues in 60 GHz networks. In [65], a decision algorithm for vertical
handoff from 60 GHz radio to WLAN is a good solution when the 60 GHz LOS link
is blocked. As discussed, vertical handoff algorithm based on IEEE 802.21 MIH may
be applicable in the fifth generation networks [71, 72] where 60 GHz is an obvious
candidate.
3.4 Conclusion
This chapter reviews the state of the art of handoffs in wireless networks such as GSM,
UMTS, LTE, WiMAX, and WLAN. We also discussed vertical handoffs which are
important in future wireless networks (in particular in 5 G) where there are multiple
coexisting networks with different scope, technologies and standards. Next we have
discussed the particular issues in handoffs in 60 GHz networks. Basically, the handoff
has the same procedure of the three steps as in popular cellular wireless networks
and should be performed and completed in the time it takes to cross the overlapping
area of two adjacent cells. However, this becomes more difficult in an RoF network,
because of the small cell size, the narrow, directional, overlapping between two cells,
and the steep signal degradation. In addition, the extra propagation delay via the
optical distribution network may affect handoffs. In particular we observe more
handoffs to be accomplished in less time.
A simple simulation of conventional handoffs in an RoF network at 60 GHz shows
that conventional handoff techniques show poor performance in terms of the number
of handoffs and the successful handoffs. In the first case, the unnecessary handoff
is about 61%. In the second simulation, it is over 10 handoffs per call in case of a
hyteresis of 0 dB. With the hysteresis of 5 dB, the number of handoffs per call is
decreasing but the number of successful handoff is dramatically down from 98.6% to
89.9% when the MS changes its speed from 0.25 m/s to 2 m/s. Since the high number
of handoffs and the low number of successful handoffs, it could result in increasing
the CDP as well as the CBP. Even conventional handoff algorithms are simple and
easy to deploy; however they need to modify to adapt with the requirements of the
RoF network. In this thesis, the recommendation is using additional information
of the MS, such as its direction, velocity and traffic information as well as using
the tools of artificial intelligent, for example, neural network and fuzzy logic, to
support the system making a handoff decision early. These applications can improve
49
3. Handoff in Wireless Networks
the CDP and reduce handoff delay, however these algorithms introduce complexity
in the design. The right tradeoff should be found. In the next chapter, we will
investigate movement prediction in an indoor environment for improving handoff
performance.
50
Chapter 4
4.1 Introduction
“A prediction is a statement about the way things will happen in the future; often,
but not always, based on experience or knowledge"1 . Knowing the future is always
valuable if someone can take advantage of the outcome event to make a profit or
simply to avoid doing something. Now, everyone solves the problem of prediction
with various degrees of accuracy. For example, before going out, people usually
want to know whether it is going to rain or, people want to know the trends of
stock market sor the trend of technologies in the short-term as well as the long-term.
These predictions are made from the study of the tremendous historical data by
using predictors such as prediction models. Sometimes, the predictions are the
opinions of an informed and knowledgeable person. For example, the prediction of
Moore (co-founded Intel Corporation in July 1968) so called Moore’s Law [73], is
still applicable today.
In the telecommunication field, predictions can be applied in many areas such
as: the trends of technologies, services, dynamic subscribers (including potential
customers, churns, dissatisfaction, profitable customers), the network traffic, the
damage of hardware, and network management. To know the next technologies
as well as new services in the future can help telecommunication companies in
1 http://en.wikipedia.org/wiki/Prediction
51
4. Movement Prediction in Indoor Environments
optimizing network design and investment. The traffic prediction is useful for the
network administrators to configure, route the network to avoid the traffic block
and improve the efficiency of using bandwidth. The customers are the most value
assets of the telecommunication service providers. In an open telecommunication
market, the competition among service providers is high. The customers are offered
more choices of classes of service and multiple levels of quality. These reasons lead
to making a new customer, as well as keeping the existing customers, more difficult.
Therefore, the churn prediction is important. It can be made by using the data from
the network; including quality of service, customer care, call blocking probability,
and call dropping probability.
In wireless networks, the traffic is dynamic. Therefore, it is too difficult to monitor
and control network resources: network configuration, network load balancing,
bandwidth allocation and improving handoff performance. Knowing the next location
of the MS is one way to improve network performance. It can reduce the complexity
of network management as well as improve the satisfaction of the MSs.
In the proposed RoF network, network management is even more difficult because
of the two main reasons which are discussed before. The first reason is that handoffs
are frequent and the overlapping between two adjacent cells is small and sharp at
the open areas. The second reason is due to the dynamic 60 GHz radio link, the
MS can lose the signal temporarily, because the SIR can drop significantly, almost
by 20–30 dB, and rise up to 20 dB in a few centimeters [21]. Hence, the handoff
and resource management in the proposed system are critical issues. Based on the
hypothesis that in the indoor environment, people tend to move according to some
patterns depending on their habits and the layout of the building, predicting the
next location of the MS can be possible with high probability. Several prediction
methods have been proposed [74–88]. In this thesis, the hidden Markov model is
selected.
Before introducing the HMM on movement prediction in Section 4.3, the inves-
tigation of other related work is introduced in Section 4.2. The experiment and
results are presented in Section 4.4. Finally, Section 4.5 concludes this chapter.
52
4.2. Movement Prediction
Trace-based Models
The trace models are based on real mobility patterns which are collected (measured)
from a large number of participants over a long time. They are exploited to generate
synthesis traces to refine the existing models. The refined existing models become
more realistic. However, obtaining enough data is difficult. Currently, there are
some repositories of data which are published for research purposes, human mobility
data from Cambridge [89], the Wireless Topology Discovery project at UCSD [90],
the other wireless traces at Dartmouth College [91]. There are several methods to
exploit this history of movement. In [92], a mobility model based on real data from
the campus WLAN at ETH in Zurich is presented. A detailed analysis of the usage
of the WLAN of the Dartmouth College campus is presented in [91].
Synthesis Models
Since the trace-based models (near real models) require the movement history, there
is a trade-off between complexity and usability. In some scenarios, for example a
new network, it is impossible. Thus, the synthesis models have been preferred [93].
Two categories of these synthesis models are proposed.
a) Entity Mobility Models allows changes in the speed and direction in reasonable
time slots. There are several mobility models in this group. Each model has its
advantages and disadvantages. The first two models presented below are usually
used in research for wireless networks.
• Random Walk mobility model (RWMM), the simplest mobility model, based
on the random direction and speed of the mobile nodes and no memory for
last movement.
• Random Waypoint mobility model, allows the mobile node to stay at the
specific location for a specified time (so called pause time) before moving at
the randomly selected direction and speed (which is uniformly distributed in
the specified range [speedmin , speedmax ]
• Random direction mobility model, this model forces the mobile node to travel
to the edge of the simulation area before changing the speed and direction.
53
4. Movement Prediction in Indoor Environments
b) Group Mobility Models. These models are based on the assuming the movement
of the specific node depends on the other mobile nodes in the group. For example,
a group of soldiers is moving to perform a task or a group of vehicles is traveling
along the street.
• Reference Point Group Mobility Model (RPGMM) presents the random motion
of a group mobile nodes as well as of each individual mobile node. The group
movements are based on the path traveled by the logical center of the group.
The other models, Column, Nomadic Community, and Pursue mobility model
can be obtained from the RPGMM as the special cases [93].
• Column mobility model represents a group of mobile nodes that move forward
in a specific direction. It is a slight modification of the RPGMM, when the
center movement is defined as the movement of group behavior in terms of
speed and direction.
• Pursue mobility model is used for target tracking. A group of the mobile nodes
follows a target mobile node.
54
4.2. Movement Prediction
h a
b
g c
f e
Figure 4.1: The sample of movement modeling in a cellular system. The string gf decc is the
movement path.
for every time shift s. The movement history is now the path of V . The simple
Markov model is proposed. Since it is a stationary stochastic process, the parameters
of Markov models are easy to be learned using the movement histories. Using the
entropy for comparing these models, the paper [74] is to build a universal predictor
for this process.
One more step to replace Markovian models with the attempts built by dictionary-
based compression algorithms such as LZ78 [75] and the LeZi-update algorithm has
55
4. Movement Prediction in Indoor Environments
been presented. This LZ78 algorithm divides the string v1 , v2 , ..., vn (vi ∈ ϑ) into c(n)
words w1 , w2 , ..., wc(i) such that for all j > 0, the prefix of the word wj is equal to
some wi for 1, i < j . For example, the movement history “aaababbb” can be phrased
as words “a, aa, b, ab, bb". And the algorithm calculates the occurrence probability of
each word, which is used to build the transition matrix of the probabilistic prediction
of individual symbols (cells). This transition matrix allows one to predict the next
cell.
Another extension of the LZ78, Active LeZi, has been proposed for movement
prediction in an indoor environment; the MavHome project [77]. It also reuses the
framework of the LeZi-update [74]. This algorithm starts with the order-k Markov
model and then employs the LeZi-update to build the probability of the next symbol.
In order to improve the prediction result, a data mining process celled Episode
Discovery (ED) is included. The ED enhances the algorithm with the prediction
accuracy of 100% compared to the original achievement of 47%. With the synthesis
data, Active LeZi can predict with an accuracy of 87%. The other prediction models
using in MavHome are smart home inhabitant prediction (SHIP) and task-based
markov model (TMM). They are also simple in design and able to find the most
likely activities of a smart home inhabitant.
The prediction, based on analyzing the string, is present in the papers [76, 79].
In [76], the information-theoretic framework is proposed for 4 G networks. The
joint residence probability distribution of a mobile node in multiple sub-networks
is considered. Based on the degree of centralized control, the location tracking
strategies are introduced along with the trade-off between location update and the
cost of paging. In [79], the mobility prediction scheme using LeZi data compression
techniques are used for improving the call admission control and bandwidth reserva-
tion scheme. From the mobility attempts, the probability P ri,j (Tk ), the probability
of the MS moving from cell i to cell j in the time slot Tk is calculated; and then the
most likey cell-time is achieved to predict when the MS most likely visits the cell j.
This allows the system to know when the handoff should be taken.
Several machine learning techniques are available to extract patterns from historical
data such as, Markov Chains, Bayesian Networks, Time Series and Neural Networks.
However, the challenge is how to model and apply these algorithms in wireless
networks, especially in an indoor environment. In cellular wireless communication,
use of movement prediction has been proposed by many [80–83]. In [80] authors
argue that the movement profile of an MS repeats daily, and the next movement
56
4.2. Movement Prediction
depends on the present location and the time of day. By using the conventional
location management schemes like zone-based and movement-based, the location
of the MS is collected. The proposed model based on neural networks (Fig. 4.2) is
to learn and find the regular movement patterns and the high probability of each
cell in each time of day the MS may be within; and re-learn if the MS changes its
behaviors.
present time
The other neural network based prediction schemes for wireless network have
presented in [81–83]. These schemes predict the target cell of an MS. The work has
also compared the proposed method with other methods presented in [78]: Bayes’
rule (based on the conditional probability of future direction), Direction Criterion
(based on current direction), Location Criterion (based on current location) and
Time Criterion (based on direction criterion with additional time constraint). The
simulation result shows that the neural-based method gains the highest prediction
accuracy of 72% compared to these other methods. A hybrid method, Bayesian
learning neural network model, is used to predict the user’s movement in cellular
wireless networks [82]. Multilayer perception network (MLP) is suitable to solve
complex nonlinear problems. In Bayesian methods the uncertainty event is quantified
by the probability, and Bayesian learning is the probability distribution to express
how likely the predictions are. This knowledge is incorporated with neural networks
to improve its performance. In this work, the proposed method performs better
than the other MLP techniques: back-propagation, Elman, Resilient, Levenberg-
Marqudat, and One-Step Secant.
The work in [83] compares two Markov model based prediction approaches: Mobile
Host Centric and in Access Point Centric.
• Access Point Centric, this approach allows each AP to build a model of the
movements of the MS passing through. The MSs are only involved in handoff
57
4. Movement Prediction in Indoor Environments
processes when sending the identification of their previous APs. This approach
can get an estimation of the model quickly.
• Mobile Host Centric, in this approach, every MS builds a model using its own
movements. However, this approach is not suitable for the common trend of
wireless networks where the terminal devices are supposed to be small, limited
processing units.
The simple Markov model obtained from [74] is the predictor to compare two
approaches. L = {L1 , L2 , ...} is the set of location history. The next AP is defined
by the distribution probability which is dependent on the last k symbols. The
k-order Markov is modeled as the Eq. 4.2.
And the location of the mth MS, Lm , is defined as the Eq. 4.3. When a model is
predicting, the most probable next AP is chosen. No prediction is performed if the
last k APs of the MS is not available. In such cases, the author in [83] proposes to
use the lower order of Markov model.
OLm m m
P
i−k , Li−1 , l; L
P (Li = l|Li−k , ..., Li−1 ) = Pm∈M m m m
(4.3)
m∈M OLi−k , Li−1 ; L
The experiments in [83] show that the difference in prediction accuracy of MHC and
APC is only a few percent (55% compared to 59%). These results are not improved
when applying the data with time division (such as weekday, weekend). However,
there is no evaluation of the advantages and disadvantages of the two approaches.
The next researches [84–88] have been done for movement prediction (predicting
the next room) in an indoor environment. Predicting the next room by which a
person is passing is applied in several smart home projects. Smart homes are defined
as houses where all the household devices are intelligent and automatic. In the
Adaptive House project [84], the authors proposed and implemented a smart home
environment. A neural network is used to predict the activities of the inhabitants.
Control systems in the residence are based on neural network reinforcement learning
and prediction techniques. The Adaptive House can predict when the occupants
will return home and therefore determine when to start heating the house so that a
comfortable temperature is reached by the time the occupants arrive...
In another smart home project, the Smart Doorplates project [87], the neural
network, Bayesian and hidden Markov models are used to predict the next room of
the inhabitant [85–88]. Neural Network with multilayer perceptron (MLP) and one
58
4.2. Movement Prediction
hidden layer (Fig. 4.3) is applied to predict the next room the user will be in [85].
The people and rooms in the building are coded in bit-stream. The input data
is the code of the person and the code of the last room they visited. The neural
network is trained using a back-propagation in both static and dynamic learning.
The output prediction of a local predictor is greater than one of a global predictor
(89.81% compared to 92%).
Bayesian networks have been used for predicting the next location of a user and
additionally the duration of stay and time when the person is probably moving to a
new location [86]. Each Bayesian is modeled for one person, and consists of different
time slices. In each time slice, the CD depends on the CR of the person, the CR
depends on the last n rooms, and the TD and WD are also in charge of prediction
as illustrated in Fig. 4.4. The predictions for each person’s location and duration
are performed with one room, two rooms and more than two rooms, in two cases of
with training and without training. The Bayesian network predictor achieves high
accuracy of next location and duration; 90% and 80%, respectively.
Applying hidden Markov model in such a project is presented in [88]. The n-order
HMM is used for predicting the next location. The experimental results show that
HMMs outperform other implemented prediction techniques such as Neural Network
and Markov predictors. Surprisingly, the simplest HMM is the most accurate model.
In [87], the five predictors, Dynamic Baysian Network, Markov Predictor, State
Predictor, and two neural network methods: Multi-layer Perception and Elman Net,
59
4. Movement Prediction in Indoor Environments
CR CR CR
TD TD TD
WD WD WD
CD CD CD
t-1 t t+1
Figure 4.4: Bayesian network with dependencies between different time [86]. Note: CR – current
room, CD – current duration, TD – time of day, WD – week day.
are investigated and compared for the same set of scenarios and data. The results
show that the method which gives the highest prediction accuracy requires high
modeling costs and offers lowest stability. Thus, a designer must decide on the most
important criterion to choose the most adaptive prediction technique. The HMM
can use both historical movement patterns and the location of antennas to predict
the next location of the MS. We chose the HMM to develop the prediction model
for the RoF network. The following section presents movement prediction based on
the HMM.
60
4.3. The Movement Prediction Model
M is the number of observation symbols per state. The model has the set of
observation symbols: V = {v1 , v2 , ..., vM }. The observation symbol probability
distribution in state Sj , B = {bj (k)}, where bj (k) = P (Ot = vk |qt = Sj ) for
1 ≤ j ≤ N , and 1 ≤ k ≤ M .
0 ≤ bj (k)
M
X (4.5)
b j (k) = 1
k=1
61
4. Movement Prediction in Indoor Environments
v41 AT16
AT12 AT13 AT14 AT15 AT17 AT18 AT19
v12
O5 v13 v18 v19
v14 v15 v16 v17
v32
O4
v33 v34 v35 v36 v37 v38 v39
S12 S13 S14 S15 AT16
v10 AT17
v8
v30
HCC v31
O3 v11
AT8 v29 AT9 v9 AT10 AT11
O2
S8
O1 S9
v3 v4 v5 v6 v7
AT10
v1 v2 v28
S1 S2 S3 S4
If the next location of the user depends on k previous locations, the set of
movements is the k th -order Markov chain (Eq. 4.8). A k th -order Markov chain may
62
4.3. The Movement Prediction Model
The graph in Fig. 4.5(b) shows the relationship between states {Si }. If the state
Si can transit to the state Sj , they are connected by a line. Otherwise, they are not
connected and the probability of transiting from Si to Sj is zero.
To predict the next location of the user, given the observation sequence O =
{O1 , O2 , ..., OT }, the system determines the next symbol OT +1 . But first, the system
should find the most adaptive HMM model for the collected data.
Training
The parameters of the model λ = {π, A, B} are adjusted to maximize P (O|λ). This
maximization can be done by the Baum-Welch procedure in terms of joint events
ξt (i, j) and state variables γt (i). This procedure is as follows.
3. Let Ā = {āij }, B̄ = b̄i (k) , and π̄ = {π̄i }. The new adjusted model is
¯ .
λ̄ = Ā, B̄, pi
Where γt (i), ξt (i, j) are defined and computed using the forward and backward
variables αt (i), βt (i) [94] as follows.
63
4. Movement Prediction in Indoor Environments
αt (i)βt (i)
γt (i) = P (qt = Si |O, λ) = (4.12)
P (O|λ)
T
X
γt (i) = expected number of transitions from Si (4.13)
t=1
T
X
ξt (i, j) = expected number of transitions from Si to Sj (4.15)
t=1
After training the model, we use the adjusted model λ̄ to predict the next observation
symbol OT +1 , given the observation O = O1 , O2 , ..., OT :
If the system continues predicting the next symbol, let T = T + 1 and repeat the
three above steps again. When the system completes the prediction, the evaluation
is performed. In case of an incorrect prediction, the system has to be adjusted again.
64
4.4. Experiment and Result
4.4.1 Training
At first, the floor is divided into many areas (observable states V ). We use Radiowave
Propagation Simulator (RPS) [69] to generate coverage of antennas. The floor plan
is given in Fig. 4.6(a). Walls in the building are made of concrete and are 10 cm
thick. The outside walls around the building are 30 cm thick. Each antenna (ATi )
operating at 60 GHz with the gain of 0 dB covers only one room at the height of 3 m,
in the large room such as the common room, two antennas are located. A mobile
terminal is 1 m high. A grid with a spacing of 30 cm is defined for the floor. For
each cell of the grid, the signal strength from the neighboring antennas are collected
and stored in the HCC (Fig. 4.6(b)). It is possible to define the overlap areas
between adjacent cells. For example, area v1 is room 1, v20 is created at the door
where the signal strengths of AT1 and AT8 are greater than the threshold of - 80 dB
(Fig. 4.6(c)). Then, the observer provides the system with the observation sequence
O. The hidden states of our model is N = 19 and the number of observation symbols
is M = 41.
Two people are involved in this simulation: i) a guest (we call User 1); ii) a regular
employee (we call User 2), representing two different groups. For User 1, we use
Random Waypoint Mobility Model with reflection [93] to generate 10000 movements
in which 9000 movements are used to train the model and 1000 movements are
used to evaluate the model. For User 2, we use the data of Employee B from data
collections used in [86]. We also modify this data set for matching with our design by
adding more locations such as doors and corridor areas in the movement. In this data
set, we use 945 movements (summer data) for training and 448 movements (fall data)
for evaluating. To generate the antennas, to which a user connects in each position,
a C++ program is developed. It assumes that the user keeps communicating all the
time while moving and handoff delay is zero and the hysteresis is 0 dB.
The initial state probabilities are set equally as 1/19 for User 1. For User 2,
assuming he works in Room 1, the initial probability of room 1 is 100%. The
transition state probability matrix (N × N ) is set as Fig. 4.5(b) in Section 4.3.2.
The number of predictions in our experiment is 1.
4.4.2 Evaluation
In order to evaluate the model, the important parameter is prediction accuracy. It is
the ratio of the number of proper predictions Npp to the total number of predictions
65
4. Movement Prediction in Indoor Environments
AT16
AT12 AT13 AT14 AT15 AT17 AT18 AT19
Optical Distribution
HCC Network
-77 dBm
-83 dBm
-89 dBm
-95 dBm
-101 dBm
(Composite)
Coverage
(b) Coverage
v16 v17
v12 v13 v18 v19
v41
AT16
AT12 AT13 AT14 AT15 AT17 AT18 AT19
v14 v15
O5
v32
O4 v34 v38
v33 v35 v36 v37 v39
v10 v11
v8 v9
v30
HCC v29 v31
O3
AT8 AT9 AT10 AT11
O2
O1
v3 v4 v5 v6 v7
v1 v2
v28
66
4.4. Experiment and Result
Npp
p= (4.16)
Tp
In this experiment, the Ignorant Prediction [74] is used to compare with the
proposed prediction method. Without any information of the previous location, the
mechanism in this method is to randomly choose one of the neighboring areas as
the next location. The probability of selecting one of eight directions is assumed to
be equal (1/8). This method can be alternatively used for the system in cases of the
absence of location prediction.
4.4.3 Result
Whenever the system is requested to predict the next location, it compares the
predicted location with the actual location and returns a correct prediction or an
incorrect prediction. In Fig. 4.7 shows the results of our experiments. For both
users, the proposed system significantly outperforms the system using the Ignorant
Prediction (54.6% vs 18.5%, and 84.1% vs 23.2%).
For User 1, the prediction accuracy of the proposed prediction model is only
54.6%. The reason may be the high degree of randomness selection of the next room.
Therefore, the system needs more historical data of User 1 to improve the prediction
accuracy. For User 2, the proposed prediction model is working well. The prediction
accuracy is high, up to 84.1%. In some cases, User 2 suddenly changes behavior,
the system may predict incorrectly and the system must learn the new movement
habits.
The C++ simulation program in [96] has been modified to simulate handoff in
an RoF network for User 1 in both cases of movement prediction methods. The
system achieves better performance by using a handoff algorithm with HMM based
movement prediction. In cases of using HMM based movement prediction, the
successful handoff rate is 98.25% compared to the successful handoff rate of 94.54%
in case of using the Ignorant Prediction.
67
4. Movement Prediction in Indoor Environments
100
The proposed method
90
Ignorant method
80
successful handoff rate [%]
ccessful handoff rate [%]
The prediction accuracy
he prediction accuracy
70
60
50
40
30
20
10
0
User 1 User 2 Successful handoff rate
(User 1)
Figure 4.7: Experimental result
accurately to 100%. Moreover, two factors, data and predicting time, are known to
effect the accuracy of prediction results.
Data
The collected data must be clean and correct. Incorrect data has an effect on the
accuracy of the prediction model. In the experiment, the data is obtained from the
Doorplate project [87] and the synthesis Random Walk mobility model. In addition,
in the proposed RoF network (in Section 2.3 of Chapter 2), the positioning systems
can causes errors in the location of a MS. For example, UbiSense can achieve the
precision of 99% within 30 cm. The missing data is imputed by discarding the
incomplete data or filling the missing values from the other data.
Moreover, the prediction accuracy could depend on the dimension of data as well
as the data complexity. Recent data is better than older data. In theory, if the data
is big enough, the performance parameters of the predictive model can be tuned
correctly. Therefore, the model is closet to resembling a practical model. However,
it takes time to obtain such a large amount of data.
Latency
68
4.5. Conclusion
4.5 Conclusion
The adaptive architecture for an RoF network is proposed in Section 2.3 of Chapter 2.
The positioning system, such as sensor network and RFID can provide all information
about the location of the user for predicting the next location. Predicting the next
position can help the system in reserving resources, reducing the call drop probability,
decreasing handoff latency and minimizing the number of unnecessary handoffs.
The method based on HMM is developed in this chapter. The proposed prediction
model implemented in the HCC can improve computation time. For particular
data sets we choose, the proposed prediction model can give a good result with the
precision accuracy of 84.1% for regular employees. However, in case a person moves
randomly, it requires more learning by increasing T or processing huge historical data.
Thus, in order to deploy the indoor prediction in an RoF network, it is necessary
to improve the performance of this prediction model in terms of the prediction
accuracy and the process time. In the future, we will study the proposed prediction
model with other data sets. It would be interesting to evaluate the effect of T on
the prediction accuracy as well as estimating the cost (time) of such models.
69
4. Movement Prediction in Indoor Environments
70
Chapter 5
5.1 Introduction
The handoff in the 60 GHz band is worse than that at low frequencies. At first,
the handoff occurs more frequently because of the small coverage of 60 GHz AP.
Secondly, a handoff should be completed in the small and narrow overlapping area
between two adjacent cells. In addition, as mentioned before, the 60 GHz signal can
drop and rise up significantly in a few centimeters [21]. This phenomenon can cause
the handoff to temporarily ping-pong, hence, increasing the number of unnecessary
handoffs. Thus, the handoff algorithm should have the procedure to deal with it.
This chapter proposes the handoff algorithm to use the motion direction of the MS
that can be obtained from the movement prediction model. By this way, the system
can reduce the number of APs (or antennas) taking part in the handoff process. The
proposed algorithm can be improved significantly in terms of reducing the number
of handoffs while increasing the number of successful handoffs.
The remainder of this chapter is organized as follows. Section 5.2 introduces the
classification of handoff algorithms, and the comparison among them. Combination
of the handoff issues are presented in Chapter 3. Section 5.3 introduces the proposed
direction assisted handoff algorithm. The simulation is performed to evaluate the
proposed handoff algorithm in Section 5.4. This chapter is concluded by Section 5.5.
71
5. A Direction Assisted Handoff Algorithm
72
5.3. The Direction Assisted Handoff Algorithm
MS APs
Probe re
quest
sponse
Probe Re Ch A
Probe re
Probe delay
quest
ponse Ch B
Probe Res
X
Probe re
quest
ponse Ch C
Probe Res
Authenti Choose
cation request new AP
Authentication
delay
nse
tion respo
Authentica
Reassoc
iation re
ques t
Reassociation
delay
nse
tion respo
Reassocia
73
5. A Direction Assisted Handoff Algorithm
AP found in the first step; (iii) Association (as illustrated in Fig. 5.1). Therefore,
handoff delay can be defined as the sum of the time of each step. Handoff delay is
calculated as the following equation.
In Eq. 5.1, the time for the discovery step takes approximately 90% of the handoff
delay, which is estimated in the range of 200–600 ms [101]. The total time of the
re-authentication step and association step may be up to 10 ms. Therefore, the
handoff delay depends on the discovery step. Reducing the discovering time is one
way to reduce the handoff delay. Thus, the handoff delay can be reduced if the
number of APs involved in the discovery is less. The following section presents the
direction assisted handoff algorithm that uses the direction of the MS to limit the
number of APs to scan.
74
5.3. The Direction Assisted Handoff Algorithm
ATj ATi
L1
C
AT2 N
MS’
r L2
AT7 AT3
MS
AT1 N’
AT6 AT4
Figure 5.2: The potential antennas from the moving direction of the MS
AT5
in Fig. 5.2), where the angles between the actual direction of the MS and L1 and L2
are ±θ. In the worst case, when the movement of MS is completely random, θ is
90o .
75
5. A Direction Assisted Handoff Algorithm
where (xM S , yM S ),(xATi , yATi ),(x, y), are the coordinates of the MS, the antenna
ATi , and the point N. The L is the distance between the MS and the antenna ATi ,
R is the radius of the coverage of the antenna ATi .
The junction point N is in the coverage of C if the following condition Eq. 5.3 is
satisfied.
θ>ϕ (5.3)
where ϕ is the angle between the line of two points MS and N and the actual path
traveled by the MS.
Assuming that set S consists of all n candidate antennas for the MS, S = {ATi },
with i = 1, ..., n, by using the moving direction, the MS can select a subset of the set
S consisting of m approaching antennas A given by two conditions: i) belongs to set
S, and ii) having an overlap with the coverage of C. Set A is given as: A = {ATj },
where i = 1, ..., m, ATj ∈ S, and m ≤ n. For example, in Fig. 5.2 AT2 and AT3 are
candidate antennas for the MS. The following algorithm Alg. 1 is used by HCC to
select an antenna for set A:
The MS samples the received power from the current antenna for every specific
r
time ∆t. Thus θ can be calculated as, θ = arccos( v∆t ), where r is the error range of
the direction predicting algorithm, and v is the speed of the MS. Whenever the signal
76
5.4. Simulation and Results
strength of the current antenna is below the threshold (in the preparing handoff
area), the MS accesses the strength of the approaching set of antennas from the
HCC and makes a handoff decision. If no antenna in A is selected, the candidate
antenna is chosen from the remaining antennas in S. The handoff algorithm Alg. 2
is proposed as given below.
77
5. A Direction Assisted Handoff Algorithm
Walk Mobility Model with reflection [93]. In order to understand the effect of the
mobile speed on the performance of handoff algorithm, there is a small change in
the mobility model: in each simulation the speed of mobile users is fixed within the
range 0.25 m/s to 2 m/s. The MS samples the received power every ∆t = 50 ms.
The transmitted power of each antenna is 15 dBm. The gain of transmitter and
receiver is set to 0 dB. To measure the received signal strength of the MS, the floor
is divided into equal square cells of 30 cm (Fig. 4.6(b)).
The propagation models investigated are: Friis model and multi-ray propagation
model. In multi-ray propagation model, many components such as single reflected
components and double reflected components are included. In order to simplify the
simulation and reduce the calculation time, a two-ray propagation model is chosen,
given by Eq. 5.4.
Pr (d) = Pt + Gt + Gr − Lp (d) (5.4)
where the path loss in the indoor environment is calculated as Eq. 5.5 [102]:
1 ri j
Lp (d) = 68 + 20log + e ∆φi
(5.5)
d0 di
where, d is the distance between the transmitted antenna and the mobile user; d0 is
the path length of the direct component; di is the path length of the floor reflected
ray; ri is the reflection coefficient of the floor surface ≈ 1.
The following parameters are measured during the simulation: initiation of
handoffs, failed handoffs, false initiation of a handoff, successful handoffs, ping-pong
effect and outage. The initiation of a handoff parameter represents the number of
times a handoff is requested. When a handoff is completed, the location of an MS
is checked. If the location of the MS is in both the old and the new antenna, a
handoff is successful. If the location of the MS is in the old antenna but not in the
new antenna, it is called an initiation failed handoff. Otherwise, the location of the
MS is out of the old antenna and the handoff is failed. Two consequent handoffs
between two antennas within the given ping-pong time tping−pong are considered as
ping-pongs. The tping−pong is calculated as the time required for the MS to come
across the overlap between these two adjacent antennas. Outage is also monitored in
this simulation. All mentioned parameters can increase the CDP. Two parameters,
initiation failed handoff and ping-pong handoff, could increase the overhead load.
The proposed handoff algorithm is evaluated with three cases of the corresponding
error in prediction algorithm, θ: i) the ideal condition: 0o , ii) 45o , and iii) the worst
case: 90o .
78
5.4. Simulation and Results
5.4.2 Results
There is a small number of false initiation of handoffs for every handoff algorithm
studied in our simulation. Thus the percentage of successful handoffs, the percentage
of failed handoffs, the percentage of ping-pong, the number of handoffs per call,
and the outage probability are determined based on the measured parameters. In
Fig. 5.3, the number of successful handoffs sinks significantly when the hysteresis is
increased from 0 dB to 5 dB. When the speed of the mobile user is 0.25 m/s, this is
99% and 98%. In case of the speed of 2 m/s, there is a huge difference of 95% to 87%.
Thus handoff in 60 GHz needs an adaptive hysteresis with the mobile speed. With
the direction assisted handoff algorithm (DHO) it is above 99% when the mobile
speed is between 0.25 m/s to 2 m/s. In the worst case, it is slightly higher than that
of a conventional handoff when no hysteresis (conv-h = 0 dB) is used.
100
98
Percentage of Successful Handoff (%)
96
Conv−h=0dB
Conv−h=5dB
94 DHO−Theta=0
DHO−Theta=45
DHO−Theta=90
92
90
88
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
In Fig. 5.4, the percentage of failed handoff of the conventional handoff with
hysteresis of 5 dB (conv-h = 5 dB) is worst, as it goes down to 10% at two speeds
of the mobile, 1.75 m/s and 2 m/s. Meanwhile, the percentage of failed handoff of
DHO is less than 1% in the case of θ = 0o . Even in the worst case, θ = 90o , its
79
5. A Direction Assisted Handoff Algorithm
performance is slightly better than that of the conv-h = 0 dB. This is due to the
proposed algorithm possibly saving time by scanning only candidate antennas.
12
10
Percentage of Failed Handoff (%)
Conv−h=0dB
Conv−h=5dB
6 DHO−Theta=0
DHO−Theta=45
DHO−Theta=90
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
80
5.5. Conclusion
6
Number of Handoffs per Call
Conv−h=0dB
Conv−h=5dB
4 DHO−Theta=0
DHO−Theta=45
DHO−Theta=90
1
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
Fig. 5.7 shows the outage probability of handoff algorithms. The outage probability
of the conv-h = 0 dB is very small; less than 0.01. Three cases of DHO have nearly
the same values, with the peak being slightly higher than 0.05 at the mobile speed
of 1.75 m/s. The conv-h = 5 dB is the worst one. Its outage probability is much
higher than that of DHOs, and its peak value is nearly 0.15 for mobile speed between
1.5 m/s and 1.75 m/s.
5.5 Conclusion
Handoff becomes more critical in an RoF network; the promising architecture
for providing broadband multimedia applications in indoor environments. In this
network, conventional handoff algorithms with fixed hysteresis or threshold seem to
be impractical due to their poor performance. Thus, the direction assisted handoff
algorithm is introduced with the proposed RoF network. By utilizing the moving
direction and location of the mobile user, there are a limited number of antennas
involved in handoff; hence the scanning time can be reduced. The simulation results
show that the proposed algorithm reduces the number of handoffs and increases
81
5. A Direction Assisted Handoff Algorithm
60
50
Percentage of Ping−pong Handoffs (%)
40
30
20
Conv−h=0dB
10 Conv−h=5dB
DHO−Theta=0
DHO−Theta=45
DHO−Theta=90
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
the number of successful handoffs over a wide velocity range of a mobile user. The
effect of the proposed handoff algorithm on the handoff delay is the next issue to be
evaluated.
Since the proposed handoff algorithm depends on the direction prediction infor-
mation, the system needs to know whether the prediction is usable or not. In case of
an inaccuracy prediction, the handoff is longer due to the system taking more time
to scan the other cells to find a suitable cell. Thus, in order to improve the direction
assisted handoff algorithm, increasing the accuracy of the prediction and reducing
the cost of the prediction model (the prediction time) should be considered.
82
5.5. Conclusion
0.15
0.12
Outage Probability (%)
0.09 Conv−h=0dB
Conv−h=5dB
DHO−Theta=0
DHO−Theta=45
0.06 DHO−Theta=90
0.03
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Mobile Speed (m/s)
83
5. A Direction Assisted Handoff Algorithm
84
Chapter 6
Bandwidth Resource
Management in RoF Networks
6.1 Introduction
Wireless indoor networks are highly sought-after to connect all electronic devices
a person can own: Laptops, Netbooks, Tablet-PCs, Cameras, Mobile phones, and
Personal Digital Assistants. In addition, the increasing demand for broadband
multimedia applications such as VoIP, video conferencing, online games, VoD,
HDTV, and even uncompressed video, requires wireless networks to provide mobile
users with the data rate of hundreds of Mbps and a high QoS.
QoS in wireless networks is difficult to guarantee due to user mobility. When a
user moves from one cell to another cell, handoff must be supported to provide the
session continuity. If the new cell lacks resources to allocate to the new user, his call
will be disconnected. It even becomes a critical problem in an RoF network due to
the small overlapping area between adjacent cells. Moreover, in the RoF network, the
handoff events occur more frequently than in traditional wireless cellular networks
such as GSM, UMTS, WLAN. Therefore, providing QoS in the RoF network is an
arduous task.
Two important aspects of QoS are investigated in this chapter; they are CDP and
CBP. Normally, it is impossible to keep both of them low. From the users’ point of
view, the user might be less annoyed in a case when he is denied a new call than in
a case when his ongoing call (called “handoff call") is dropped due to the shortage
85
6. Bandwidth Resource Management in RoF Networks
of resources in the target cell. Thus, avoiding the dropping of a handoff call is one
solution to improve QoS.
In order to do that, wireless cellular networks can assign handoff calls higher
priority over new calls. Several handoff prioritization schemes have been proposed.
By intelligently exploiting the history, and coupling it with predicting the next
location of an MS, can help in reserving the required amount of bandwidth for a
handoff request of the MS in advance. Hidden Markov Model is used for learning
the historical data as presented in Chapter 4. Based on these concepts, a handoff
scheme using movement prediction is introduced. This proposed scheme is also
compared with no reservation [103] and fixed reservation schemes [103, 104].
The rest of this chapter is organized as follows. Section II presents recently related
work. The proposed handoff scheme is introduced in Section III. Section IV shows
the simulation and results of the proposed handoff scheme in the RoF network.
Section V concludes this chapter.
86
6.2. Related Work
The bandwidth in a wireless network may be the most precious and important
resource. When a bandwidth reservation is requested or when a channel is available,
a handoff request can be carried out. A simple solution is that each cell should
reserve fractional bandwidths of its capacity and this reserved bandwidth should
be used only for handoffs and not for the new call request. However, the open
question is to find how much bandwidth is sufficient, while the network also maintains
the maximum bandwidth utilization and keeps the maximum rate of unsuccessful
incoming handoffs below the acceptance level. Numerous schemes have been proposed
to dynamically manage the allocation of bandwidth resources such as Complete
Sharing; all traffic classes share the entire bandwidth, and Complete Partitioning;
bandwidth is divided into distinct portions with each portion corresponding to a
particular traffic class [105].
Power control is a necessary mechanism in all mobile systems because of an
important role in spectrum and resource allocation as well as for battery life and
safety reasons. Without interference, high transmission power achieves high SINR
(lower BER and high transmission rate). Power control schemes can be used to
achieve the required CIR level (Carrier-Interference Ratio). They try to reduce
the overall CIR in the system by measuring the received power and increasing or
decreasing the transmitted power in order to maximize the minimum CIR in a
given channel allocation of the system. This helps in increasing the capacity of the
network in terms of the number of mobile terminals that can be supported. Power
control can be done in an either centralized or a distributed way. The power used in
SIR based handoff algorithms has been proposed in [55, 106]. The power is also a
parameter of the input measurements that help in carrying out handoff decisions,
especially in a vertical handoff
Figure 9.of UMTS
Handoff with
between other
WLAN andsystems.
GPRS
Resource Management
Figure
Figure 6.1:10.Handoff-related
Handoff-related resource management
resource management
Some issues of the resource management related handoffs include admission control,
bandwidth reservation, and power control (Figure 10). With admission control, new calls Deleted: Figure 10
and ongoing calls can be treated differently. It helps to keep the system from being
overloaded. New calls may be queued and handoffs may be prioritized.
87
When a bandwidth reservation is done or when a channel is available a handoff
request can be carried out. A simple solution is that each cell should reserve fractional
bandwidths of its capacity and this reserved bandwidth should be used only for handoffs
and not for the new call requests. However, the open question is to find how much
bandwidth is sufficient.
6. Bandwidth Resource Management in RoF Networks
88
6.2. Related Work
On the other hand, the prioritized handoff schemes based on reservation schemes
(known as early blocking schemes [49]) will reserve handoff resources and admission
control to be used for handoff requests. New calls can not be allocated in those
reserved resources. Thus, these schemes trade a lower possibility of CDP for a higher
value of CBP. These schemes reuse some components of the QoS control plane such
as admission control combined with other concepts: handoff resource reservation
and mobility prediction. Two sub-types are distinguished in these scheme classes:
• Static schemes are simple since there is no mobility prediction and no signaling
among BSs to reserve handoff resources. The Guard Channel Scheme also
referred to as Cut-Off scheme in [104, 110] reserves a number of channels for
handoff requests; the remaining channels are shared equally by new calls and
handoff requests. However, networks employing static schemes can have a
low resource utilization or handoff resource shortage in the case of too many
handoff requests or too few handoff reservation resources. Thus, static schemes
should be considered more for application in microcell mobile networks.
• Dynamic schemes are more complex than static; they determine automatically
handoff resources depending on the actual handoff resource demand. They
need to predict the user mobility to adapt the amount of handoff resource
dynamically. In [105], dynamic schemes are grouped into three: (1) local
schemes require no signaling among BSs to retrieve possible handoff informa-
tion and are easy to deploy with low cost. Dynamic Channel Pre-Reservation
Scheme [111, 112] requires a few parameters. The more complex local scheme
in [113] needs to adapt the amount of handoff resources every two hours with
manual configuration of designed parameters. However, local schemes are
suitable for networks with traffic pattern which does not change too quickly;
(2) Collaborative per-mobile schemes can provide a higher accuracy of the
estimated necessary handoff resources than static schemes and local schemes.
Each mobile requires its own information storage to pre-reserve handoff re-
sources in neighboring cells. Predictive Channel Reservation scheme [114] and
Adaptive Channel Reservation [115] use Global Positioning System (GPS) as a
location system. Variable Reservation Scheme in [116] sums up the number of
handoffs arriving from neighbor cells to determine how much resources should
be reserved. It uses signal strength and speed as a location system. In small
cell networks like 60 GHz, these schemes may be considered more often since
there are many more terminals and handoff requests; and (3) Collaborative
aggregate schemes are less complex than per-mobile schemes because they
use aggregated information instead of per-mobile information to pre-reserve
89
6. Bandwidth Resource Management in RoF Networks
90
6.3. The Proposed Handoff Scheme
Monitoring Movement
location of MS, Prediction
RSS and other (A set of potential
metrics cells)
Y Bandwidth
Call finished Finish call? reservation in
potential cells
N
Request handoff
to the target cell Cj
N
Y
Call finished Finish call?
Dropped Y RSS<Threshold N
handoff call or excess time T
91
6. Bandwidth Resource Management in RoF Networks
bandwidth after the predefined time T if there is no handoff call requested to it. In
case the prediction cells are not correct, the no reservation scheme is performed.
6.4.1 Simulation
To examine our proposed algorithm, we developed a simulator written in C++
similar to [96]. The floor plan used in the simulation is given in Fig. 4.6(a). Walls
in the buildings are made of concrete and are 10 cm thick. The outside walls of the
building are 30 cm thick. Each antenna (ATi ) operating at 60 GHz with the gain
of 0 dB covers only a single room of height 3 m; in a large room such as a common
92
6.4. Simulation and Results
room, two antennas are located. Assuming that each antenna offers its bandwidth
of 2 Gb and 20 channels. A mobile terminal is 1 m high. Radiowave propagation
simulator (RPS) [69] is used to define areas between cells as observation symbols
{vi }, as well as to calculate the received signal strengths of the MS at the specific
location from antennas (Fig. 4.6(b)). We use Random Waypoint Mobility model
(RWP) with reflection [93] in our simulation; we use speed uniformly selected from
the range (0.25 m, 2 m).
We use the RWP model with reflections [93] to generate 10000 movements. These
movements are used as movement history for the prediction model to predict the
next location of the mobile user. The prediction model is modeled with hidden
states as antennas, N = 19.
Two classes of service are, s1 (5 Kbps – 200 Mbps) and high definition video on
demand s2 (100 Mbps – 200 Mbps). The system can reduce the required bandwidth
of a s1 call to provide a s2 call with more bandwidth. The call duration of class s1
is chosen randomly in the range (10 s, 180 s), and the call duration of class s2 call is
chosen randomly in the range (60 s, 600 s).
In our simulation, the bandwidth reservation is set to 20% of the whole bandwidth
for every cell of the FR scheme. Three schemes are simulated with the same
movement pattern of mobile users. Call arrival rate, or the rate of new calls, is
93
6. Bandwidth Resource Management in RoF Networks
calculated as the average number of new calls per second per cell (call/cell/s). In
the simulation scenarios, call arrival rate is in the range of 0.01 - 0.1 call/cell/s. A
new call is initiated from anywhere within a cell with equal probability.
The parameters calculated for handoff scheme performance evaluation are: CDP,
CBP, bandwidth utilization and the distribution of dropped call duration. The CBP
is the probability that a new call is blocked due to the shortage of resources of the
original cell. The CDP is the probability that a handoff call is dropped due to the
shortage of resources of the target cell. Bandwidth utilization (BU) is the ratio of
the used bandwidth in a cell to the whole bandwidth given as follows.
X
used bandwidth
for each cell
BU = X (6.1)
maximum bandwidth of each cell
for each cell
The last parameter presents the density of the dropped call duration in different
handoff schemes. This shows the probability density functions of the holding time
of calls before they are dropped due to lack of resources.
6.4.3 Results
Both the CBP and the CDP in the three schemes increase when the call arrival
rate increases from 0.01 to 0.1 call/cell/s as shown in Fig. 6.3, Fig. 6.6 only for
calls of class s2 , and Fig. 6.4, Fig. 6.5 for all calls of both classes s1 , s2 . In each
scheme, both the CBP and the CDP for class s2 is less than these for both classes
94
6.4. Simulation and Results
s1 , s2 since the first priority is given to the class s2 . For example, with the MPR
scheme, at the arrival rate of 0.05, the CBP and the CDP for class s2 are 10.6%
and 3.08% compared to 12.6% and 4.08% of the CBP and the CDP for both class
s1 , s2 , respectively.
0.5
0.45
0.4
0.35
Call blocking probability
0.3
0.25
0.2
0.15
0.1
NR−s2
0.05 FR−s2
MPR−s2
0
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Call arrival rate (call/cell/s)
In Fig. 6.3 and Fig. 6.4, as expected, due to equal treatment with a handoff call
and a new call and no reserved resource, the NR offers the lowest CBP and the
highest bandwidth utilization. The NR performs better than the MPR and the FR.
The amount of bandwidth for new calls in the NR may be greater than those of the
other schemes. The MPR is more effective than the FR because only the potential
cells reserve the required amount of bandwidth for the predicted handoff calls. For
example, in Fig. 6.4, at the arrival rate of 0.01, the CBP of the NR, the FR and the
MPR are 0.9%, 10.8%, and 1.08% respectively; at the arrival rate of 0.1, the CBP
of the NR is 25.2% compared with 27.9% of the MPR and 44.6% of the FR.
On the other hand, the NR causes the highest CDP while the MPR offers the
lowest CDP (Fig. 6.6, and Fig. 6.5). In Fig. 6.6, at the arrival rate of 0.01, the CDP
of the NR, the FR and the MPR are 3%, 0.85% and 1.36%, respectively. In this
case, the FR is slightly better than the MPR. When the arrival rate is increasing,
the CDP of the MPR is lower than that of the FR. At the arrival rate of 0.1, they
95
6. Bandwidth Resource Management in RoF Networks
0.5
0.45
0.4
0.35
Call blocking probability
0.3
0.25
0.2
0.15
0.1
NR
0.05 FR
MPR
0
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Call arrival rate (call/cell/s)
Figure 6.4: Call blocking probability for both classes s1 , s2 in three schemes
0.25
NR−s2
FR−s2
MPR−s2
0.2
Call dropping probability
0.15
0.1
0.05
0
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Call arrival rate (call/cell/s)
96
6.4. Simulation and Results
0.25
NR
FR
MPR
0.2
Call dropping probability
0.15
0.1
0.05
0
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Call arrival rate (call/cell/s)
Figure 6.6: Call dropping probability for both classes s1 , s2 in three schemes
1
NR
0.9 FR
MPR
0.8
0.7
Bandwidth utilization
0.6
0.5
0.4
0.3
0.2
0.1
0
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Call arrival rate (call/cell/s)
97
6. Bandwidth Resource Management in RoF Networks
are 22.95%, 13.2%, and 7.4% (Fig. 6.6). It can be so, because in the NR scheme,
the amount of bandwidth for handoff calls is lower than that of the other schemes;
the FR and the MPR. When the call arrival rate increases, the number of handoff
calls is also increasing. Thus, the fixed amount of bandwidth for handoff calls in the
FR is not enough. Meanwhile, the MPR could adjust the amount of bandwidth for
handoff calls on demand.
The bandwidth utilization of the NR is the highest at every arrival rate as
illustrated in Fig. 6.7. The unused amount of bandwidth in the MPR is lower than
that in the FR, so that bandwidth utilization of the FR is lowest. In Fig. 6.7, at the
arrival rate of 0.1, the bandwidth utilization of the NR, the FR, and the MPR are
77.67%, 70.2%, and 74.7%, respectively.
To analysis the distribution of dropped call duration, in our simulation, the call
holding times are predefined values of 180 s, 300 s, and 600 s. Fig. 6.8, 6.9 and 6.10
illustrate the distributions of dropped call duration of three schemes in three cases
of the predetermined call holding times (180 s, 300 s, and 600 s) at a call arrival rate
of 0.08. The results show that on average, dropped calls in the MPR have a longer
time duration (holding time) than the two other schemes. This is evident from the
distribution graphs which show that dropped calls in the MPR are skewed towards
the right of the center of the graphs. For example, in Fig. 6.9, the density of long
dropped call of the MPR (e.g, from 250 s to 300 s) is higher than that of the NR
and the FR. Meanwhile, the density of dropped call duration in the range of 10 s to
100 s is over 50% of the total number of dropped calls in case of using the NR in
Fig 6.10. When call arrival rates are 0.05 and 0.1, the distributions of dropped call
duration of the three schemes presented in Fig. 6.11, 6.12, 6.13, 6.14, 6.15, 6.16 also
show that the MPR not only results in minimum call dropping probability, but the
calls which are dropped, have longer holding time compared to the calls dropped
in the other two schemes. Therefore, the MPR scheme certainly helps to provide a
better quality of service and an improved user experience.
6.5 Conclusion
Managing resources in RoF networks is difficult because handoff occurs frequently due
to the limited coverage area of 60 GHz transmitters and as such smaller overlapping
area between neighboring cells. The handoff scheme exploiting movement prediction
of users has been proposed to manage resources effectively. The proposed scheme
performs better than no reservation and the fixed reservation scheme in terms of the
CDP. With the lowest CDP, the MPR can provide mobile users a better quality of
service experience. Moreover, the proposed scheme also results in a longer holding
98
6.5. Conclusion
0.14
FR
NR
0.12 MPR
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 20 40 60 80 100 120 140 160 180
Dropped Call Duration (s)
Figure 6.8: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.08 and the call holding time of 180 s.
time for dropped calls. This implies that even if a call is dropped, the average
holding time of a call before it is dropped, is higher in the proposed scheme.
However, this method depends on the additional modules, the movement prediction
system and the memory elements in the HCC. If the movement prediction is not
precise, the performance of the MPR may be affected. User behaviour and the
characteristics of environment have a very high influence on the prediction accuracy.
Thus, the next step is to increase the prediction accuracy of the prediction model,
keeping in mind the different indoor scenarios and user behaviors.
99
6. Bandwidth Resource Management in RoF Networks
0.16
FR
0.14 NR
MPR
0.12
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 50 100 150 200 250 300
Dropped Call Duration (s)
Figure 6.9: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.08 and the call holding time of 300 s.
0.5 FR
NR
MPR
0.45
0.4
0.35
Frequency (%)
0.3
0.25
0.2
0.15
0.1
0.05
0
0 100 200 300 400 500 600
Dropped Call Duration (s)
Figure 6.10: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.08 and the call holding time of 600 s.
100
6.5. Conclusion
0.16
FR
NR
0.14 MPR
0.12
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 20 40 60 80 100 120 140 160 180
Dropped Call Duration (s)
Figure 6.11: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.05 and the call holding time of 180 s.
0.16
FR
NR
0.14
MPR
0.12
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 50 100 150 200 250 300
Dropped Call Duration (s)
Figure 6.12: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.05 and the call holding time of 300 s.
101
6. Bandwidth Resource Management in RoF Networks
0.16
FR
NR
0.14
MPR
0.12
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 100 200 300 400 500 600
Dropped Call Duration (s)
Figure 6.13: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.05 and the call holding time of 600 s.
0.14
FR
NR
0.12 MPR
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 20 40 60 80 100 120 140 160 180
Dropped Call Duration (s)
Figure 6.14: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.1 and the call holding time of 180 s.
102
6.5. Conclusion
0.16
FR
NR
0.14
MPR
0.12
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 50 100 150 200 250 300
Dropped Call Duration (s)
Figure 6.15: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.1 and the call holding time of 300 s.
0.16
FR
NR
MPR
0.14
0.12
0.1
Frequency (%)
0.08
0.06
0.04
0.02
0
0 100 200 300 400 500 600
Dropped Call Duration (s)
Figure 6.16: The distribution of dropped call duration of the three schemes at a call arrival rate
of 0.1 and the call holding time of 600 s.
103
6. Bandwidth Resource Management in RoF Networks
104
Chapter 7
7.1 Introduction
MAC protocols play an important role in any communication system. These allow
multiple devices access to the medium at the same time. Basically, a MAC layer
protocol is a sublayer of Layer 2 in the OSI model as well as in the TCP/IP model (as
illustrated in Fig. 7.1). The MAC protocol defines the rules for the channel resource
management and coordination. There are hundreds of MAC protocols proposed
for wireless communication systems. They can be divided into two categories: (i)
distributed MAC protocols and (ii) centralized MAC protocols.
7 Application
Upper
layers
6 Presentation 5 Application
Receive from Network
Send to Network
5 Session
4 Transport
4 Transport
(TCP/UDP)
3 Network
3 Network
Lower
layers
(IP)
MAC
2 Data Link 2 Data Link
LLC
1 Physical 1 Physical
Figure 7.1: The sublayer MAC in OSI model and TCP/IP model.
105
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
The distributed MAC protocols are based on random access. In these protocols,
every device is equal in controlling the medium. The first MAC protocol, called
ALOHA [119], is a distributed protocol or contention-based MAC protocol. Since
it is a random access control, this protocol does not check whether the channel is
busy before transmitting; it sends data immediately if demanded. In case there is a
collision with another transmission, it tries to send the data again. The ALOHA
is simple but achieves low throughput and channel utilization. To increase the
maximum throughput, an improvement of ALOHA is proposed, Slotted-ALOHA (S-
ALOHA) [119]. In this protocol, time is divided into timeslots, and a station can send
data only at the beginning of a timeslot, so it reduces collisions. Using the random
access channel of ALOHA, carrier sensing multiple access (CSMA) is developed [120].
The basics of this protocol is “listen before send”. Before transmitting, a station
listens for other traffic on the shared medium. If there is another node transmitting,
it waits for this progress transmission to finish and then starts sending the data.
However, it is still possible that when multiple stations listen, the channel is idle
and they start sending data simultaneously. Two improvements of this protocol are
CSMA with Collision Detect (CSMA/CD) [121] and CSMA with Collision Avoidance
(CSMA/CA) [122].
Two major problems in wireless networks are hidden and exposed terminal prob-
lems. The hidden terminal problem is depicted in Fig. 7.2(a). The terminal A wants
to transmit data to terminal B. By only sensing the medium, A will not know the
transmissions by any terminals C in the grey area, and then the collisions occur
at terminal B when A starts sending data. The problem is solved by using the
“request to send” and “clear to send” (RTS/CTS) handshake mechanism. When
using RTS/CTS, the other problem, called exposed terminal problem, should be
considered (Fig. 7.2(b)). Terminal A sends an RTS and waits for a CTS from B.
Suppose a terminal D wants to send data to C, and it sends an RTS to C before A
sends the RTS to B. B senses the CTS sent by C to D, hence B will delay sending
a CTS to A in a backoff period. This means that the communication between D and
C prevent the communication between A and B, although these communications can
start simultaneously because a collision is detected in the transmitter in CSMA/CD.
It is not efficient in wireless networks because of the hidden terminal problem. An
other modification of CSMA is CSMA/CA [122] which allows the receiver to detect
the collisions.
Due to its simple and robust, the distributed MAC protocols are used in IEEE
802.11 standard and HIPERLAN standard.
Different from the distributed MAC protocols, the centralized MAC protocols
106
7.1. Introduction
Transmission range of
terminal A
Hidden terminal for pair A/B
A B
Transmission range of
terminal B
Transmission range of D
terminal A
Transmission range of
A B terminal C
Transmission range of
terminal B
Figure 7.2: The hidden and exposed terminal problems in wireless networks.
require a central unit, for example, a BS, which controls the channel access for all
stations. Thus, hidden-terminal problem and exposed terminal problem are not
apparent in centralized MAC protocols. In addition to the random access mechanism,
the centralized MAC protocols support guaranteed and hybrid mechanisms.
In guaranteed mechanisms, the station is arranged by a round-robin algorithm
to access the medium. Two groups in guaranteed mechanisms are token-passing
protocols and polling protocols. In the first group, only the station with the token
is allowed to transmit data. After finishing transmitting its data, the token is
exchanged to the other station. Since the token could easily be lost if the station
loses its wireless connectivity, the token-passing protocols are rarely used in wireless
107
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
networks. In the second group, the polling-based protocols are an important part
of WLAN. These protocols require a central unit, called master. The master sends
a poll to trigger the transmission by the station. Each station, which wants to
transmit data, must wait to receive a poll. However, QoS is not considered in these
protocols.
Hybrid mechanisms are developed from the best features of random and guaranteed
mechanisms. They are working as request-grant mechanism. The request is sent
in a random access protocol to reserve time or bandwidth which is required to
transmit the data (guaranteed). Once receiving the request, the BS allocates the
timeslot and sends a grant back to the station with the timeslot information. The
popular standard using hybrid mechanisms are IEEE 802.16, IEEE 802.15.3 c, IEEE
802.11 ad.
Both IEEE 802.15.3 c and IEEE 802.11 ad support hybrid multiple access CSMA/CA
and TDMA. CSMA/CA is suitable for data applications such as web browsing be-
cause of its lower average latency. While TDMA is suitable for realtime applications
such as VoD since it offers significantly higher QoS. In [123], a dynamic allocation
polling mechanism is proposed to be used on the top of other accesses such as
CSMA/CA and TDMA.
As introduced in Chapter 2, there are several MAC standards of 60 GHz are IEEE
802.15.3 c, WirelessHD, IEEE 802.11 ad, Wireless Gigabit Alliance, and ECMA.
However, MAC protocols in an RoF network should consider the extra delay intro-
duced by the optical distribution network. Moreover, in [124, 125], the authors have
proposed that the centralized MAC protocols are better suited for an RoF network.
Thus, this chapter investigates the feasibility of both centralized protocols, IEEE
802.15.3 c and IEEE 802.11 ad, in an RoF network. It should help the designer in
choosing the proper MAC protocol.
The remainder of this chapter has been organized as follows. Section 2 presents
the related work. Section 3 introduces the features of two IEEE standards working at
60 GHz which are IEEE 802.15.3 c and IEEE 802.11 ad. The analytical model of the
IEEE 802.15.3 c is explained in Section 4. The numerical results of the performance
of IEEE 802.15.3 c are present in Section 5. Section 6 concludes this chapter.
108
7.3. Fundamentals of MAC protocols
publications, the network throughput in both protocols does not drop significantly
when the fiber link is extended to 4 km. The effects of the hidden-terminal problem is
also presented. The simulation result shows that the network throughput decreases
to the unacceptable value when the number of hidden terminals increases [124, 125].
The paper [126] proposes the solution to extend the fiber length based on the
manipulation of RTS/CTS timeout. The analytical approximations, simulations,
and test-bed experiments are performed. These results show that the network can
apply fiber length up to 8 km while the throughput decreases less than 15%.
The impact of RoF on IEEE 802.16 WiMAX is presented in [23, 35]. Differing
from the distributed MAC protocols, the centralized IEEE 802.16 MAC protocol
can be adjusted to support RoF. The hidden-terminal problem does not affect the
contention mechanism of a IEEE 802.16 network. Accommodating the fiber delay in
TDD system, the authors in [23] present the simulation results for throughput and
MAC efficiency with different fiber lengths and packet sizes.
In [35], the authors presented the analysis of the use of RoF in WiMAX networks.
Both analytical result and simulation result are carried out for both indoor and
outdoor environments when the fiber link is setting up to in the order of a hundred
kilometers. The throughput drops to 20% as the fiber link gets to 115 km.
109
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
the specification defines the new PHY and enhances MAC from the existing MAC
specifications of IEEE 802.15.3 and IEEE 802.15.3 b.
Three PHYs are specified in the IEEE 802.15.3 c standard [8], single carrier (SC)
PHY, high speed interface (HSI) OFDM PHY and audio video (AV) OFDM PHY.
Each PHYs are designed for a different target application. The SC PHY supports
the low-power low-cost mobile market. HSI PHY is used for low-latency, high-speed
bidirectional data transmission (the data rate can be excess of 6 Gb/s), while AV
PHY is optimized for AV specific applications. Four different modulation and coding
schemes (MCSs) are proposed for different applications. The achievable data rate
using different modulation schemes is listed in Tab. 7.1
The 802.15.3 c network, called the piconet, is composed of a number independent
devices (DEVs) (Fig. 7.3). One of them is selected as the piconet coordinator (PNC).
In the MAC of the IEEE 802.15.3 c, the time is divided into multiple superframes.
The superframe consists of three parts (as Fig. 7.4):
• Beacon: Broadcasts the timeslot information for each DEV and the manage-
ment information for the piconet.
• Channel Time Allocation Period (CTAP): This part concludes time slots called
Channel Time Allocations (CTAs) which are used for commands, isochronous
streams and asynchronous data based on TDMA protocol. The PNC is
responsible for allocating CTAs.
110
7.3. Fundamentals of MAC protocols
ta
Da
con
Be
Bea
Da
Dat
ac
a
ta
on
Beacon Beacon
PICONET
Be
on COORDINATOR
ac
ac
Be
on
Data
CTAs
Beacon
CAP MCTA MCTA CTA
#n CTA 1 CTA 2 ... CTA n
1 2 n-1
ACK policies, in which MIFS stands for Minimum Inter-frame Space and SIFS is
Short Inter-frame Space. MIFS is usually less than SIFS. These acknowledgment
mechanisms are explained as follows.
• No-ACK: In this mechanism, the ACK is not sent after receiving the packet.
It is useful for frames that do not require guaranteed delivery or in the
case acknowledgment is performed by a higher layer or by other methods
(Fig. 7.5(a)).
111
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
This mechanism is more stable and reliable than the No-ACK but the data
rate is reduced.
MIFS
MIFS
Frame #1 Frame #2 Frame #3
• Blk-ACK: In the Blk-ACK mechanism, a number of sub-frames are combined
into an aggregated frame and the receiver only sends a block ACK after
MIFS
MIFS
MIFS
receiving
Frame #1 the aggregated
Frame #2 frame SIFS
Frame (Fig.
#3 7.5(d)).
F
MIFS
MIFS
SIFS
SIFS
SIFS
SIFS
SIFS
ACK
ACK
ACK
Frame #1 Frame #2 Frame #3
(a) No ACK.
SIFS
SIFS
SIFS
SIFS
SIFS
SIFS
ACK
ACK
ACK
SIFS
SIFS
SIFS
SIFS
SIFS
SIFS
ACK
ACK
ACK
Frame
Frame#1#1 Frame
Frame #2#2 FrameFrame
#3 #3
F
F
(b) Immediate ACK.
One Burst (K=3)
MIFS
MIFS
SIFS
SIFS
ACK
MIFS
SIFS
SIFS
ACK
F
SIFS
SIFS
ACK
SIFS
ACK
Aggregated Frame
7.3.2 IEEE 802.11 ad
SIFS
SIFS
ACK
The IEEE 802.11 Task Group n specifies the standard to provide a maximum
throughput up to 400 Mb/s [128]. Another group, the IEEE 802.11 Very High
Throughput (VHT) Study Group has enhanced medium access control (MAC)
with higher PHY [129]. The enhancements of VHT WLAN are proposed in IEEE
802.11 ac. This standard adds higher density modulation schemes, wider bandwidths,
112
7.3. Fundamentals of MAC protocols
Beacon Interval
Data Transmission Interval
113
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
to the interfaces. This feature makes IEEE 802.11 ad compatible with backward
standards.
Figure 7.7: Fast session transfer protocol stack in IEEE 802.11 ad.
Table 7.2 shows the comparison of the main features of the two MAC standards, IEEE
802.15.3 c and IEEE 802.11 ad. The purpose of IEEE 802.15.3 c is for communication
of WPAN. IEEE 802.11 ad is for WLAN. Both standards are working at the same
frequency band in the range of 57-66 GHz. However, the spectrum is dependent on
the regulations of each country, or region. For example, in the United States and
Canada, it is from 57.05-64 GHz. Meanwhile, the EU allows that the unlicensed
band to be in the range of 57-66 GHz. The limited transmitted power is 43 dBi
for IEEE 802.11 ad and 40 dBi (outdoor) and 27 dBi (indoor) for IEEE 802.15.3 c.
The basic cell of IEEE 802.15.3 c is piconet, while for IEEE 802.11 ad, it is PBSS.
However, MAC access protocol is centralized and controlled by PNC or PCP/AP.
Both standards are designed for short-range communication in the range 10 m
coverage. The only significant difference of both standards is that IEEE 802.11 ad is
designed to work with the forthcoming standards and IEEE 802.15.3 c does not.
The performance analysis of delay and throughput of both IEEE 802.15.3 c and
IEEE 802.11 ad is shown in [130]. The analysis shows that both standards have
almost the same throughput and delay. Thus, in this chapter, the feasibility of
IEEE 802.15.3 c is analytically investigated, but the results can also apply to IEEE
802.11 ad.
114
7.4. Analysis and Results
To increase the transmission reliability, two different signal rates for control
signals Rc and data Rd are considered. Thus, the packet preamble, header, ACKs
are transmitted using a low control rate and data frames are transmitted using a
higher rate.
The probability for a successful transmission of a l bit length frame can be
calculated as:
ps = (1 − pe )l (7.3)
In which, pe is the probability of transmission error for a frame.
115
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
where tsucc is the duration of the successful transmission of a data frame. tf ail is
the duration of the retransmission of a data frame.
The maximum throughput C within CTAP can be approximated as the follow-
ing [131].
ρl
C = (1 − (1 − ps )(nr +1) ) (7.5)
t̄
where ρ is the proportion of the CTAP within a superframe.
Applying for all four ACK mechanisms of IEEE 802.15.3 c, the probability of the
successful transmission of a frame for each mechanism is calculated as follows.
ps (N o − ACK) = (1 − pRd ld Rc lh
e ) (1 − pe )
ps (Imm − ACK) = (1 − pR d ld Rc lh +lack
e ) (1 − pe )
d ld ∗K
ps (Dly − ACK) = (1 − pR
e ) (1 − pR c (K∗lh +ldly−ack )
e ) (7.6)
d ld ∗K
ps (Blk − ACK) = (1 − pR
e ) (1 − pRc (lh +lblk−ack )
e )
where pR Rc
e and pe are the probability of transmission error for a data frame and
d
an ACK, respectively. ld and lh are the length of data and ACK, respectively. K
represents the burst size (number of subframes) in Dly-ACK (Blk-ACK).
The time for successful and failed transmissions of a data frame corresponding to
each ACK mechanism are calculated as follows:
tN o−ACKsucc = tf rame + M IF S
tN o−ACKf ail = tf rame + M IF S
tImm−ACKsucc = tf rame + tImm−ACK + 2SIF S
tImm−ACKf ail = tf rame + RIF S (7.7)
tDly−ACKsucc = tf rame × K + tDly−ACK + (K − 1)M IF S
+ 2SIF S
tDly−ACKf ail = tf rame × K + (K − 1)M IF S + RIF S
tBlk−ACKsucc = tf rame × K + tBlk−ACK + 2SIF S
tBlk−ACKf ail = tf rame × K + RIF S,
116
7.4. Analysis and Results
where tP reamble is the duration of the PHY layer convergence procedure (PLCP)
preamble, tHeader is the duration of the PLCP header, and tP ayload is the duration
of the PLCP payload.
The ACK of the Imm-ACK has only a MAC header and no payload since every
packet is expected to be acknowledged immediately. Thus, the transmission time of
an ACK in the Imm-ACK is,
For Dly-ACK and Blk-ACK, the ACK consists of a MAC header and payload.
Thus, the transmission time of an ACK is given by the following equation.
MIFS
MIFS
(a) No ACK
SIFS
SIFS
SIFS
SIFS
SIFS
SIFS
ACK
ACK
ACK
MIFS
SIFS
SIFS
ACK
SIFS
ACK
117
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
Therefore, in the RoF network the successful and failed transmission of a data
frame for each type of ACK mechanisms can be found as Eq. 7.11.
tN o−ACKsucc = tf rame + fd + M IF S
tN o−ACKf ail = tf rame + fd + RIF S
tImm−ACKsucc = tf rame + tImm−ACK + 2fd + 2SIF S
tImm−ACKf ail = tf rame + fd + RIF S (7.11)
tDly−ACKsucc = tf rame × K + tDly−ACK + (K − 1)M IF S
+ 2fd + 2SIF S
tDly−ACKf ail = tf rame × K + (K − 1)M IF S
+ fd + RIF S
tBlk−ACKsucc = tf rame × K + tBlk−ACK + 2fd + 2SIF S
tBlk−ACKf ail = tf rame × K + fd + RIF S
packet is transmitted without experiencing any error, it is received by all the receivers in the range
with probability 1.
118
7.4. Analysis and Results
Parameters Value
Beacon frame 832 bits
MAC subheader length 38 bytes
Imm-ACK frame 25 bytes
Dly-ACK frame 40+2K bytes
Blk-ACK frame 56+2K bytes
Preamble length 3.26µs
SIFS 2.5µs
MIFS 1µs
RIFS 1.8µs
layers take care of the acknowledgement and thus saving the efforts and bandwidth
at the MAC. Imm-ACK always starts at a lower side for smaller frame sizes, due to
wastage of bandwidth in immediately acknowledging packets, but improves when
the frame size increases. This trend is seen throughout.
The performance of the Blk-ACK and Dly-ACK are satisfactory at low BERs, and
they degrade quickly in the case of higher BERs. At BER=10−7 , and BER=10−6 ,
the throughput reaches maximum at a specific value of frame size. The maximum
value for No-ACK, Imm-ACK, Dly-ACK, Blk-ACK are 3.1 Gb/s (at frame size
of 2.5 Kbytes), 3.025 Gb/s (at frame size of 4 Kbytes), 2.82 Gb/s (at frame size
of 1 Kbytes),2.84 Gb/s (at frame size of 1 Kbytes), respectively. However, after
reaching maximum the throughput decreases slowly at BER=10−7 for Dly-ACK and
Blk-ACK, where as for No-ACK and Imm-ACK it remains constant. This is due
to the fact that Dly-ACK and BlK-ACK see increases in throughput because of a
reduced number of ACK frames. However, it loses its advantage when the frame size
increases at a slightly higher BER due to the possible higher packet loss (Fig. 7.9(b)).
We can see this trend very clearly when BER=10−6 . While all the schemes see a
reduction in throughput, the Dly-ACK and Blk-ACK see a much higher fall in their
throughput (Fig. 7.9(c)). The advantage of aggregation is lost in this case.
The bandwidth efficiency with respect to various frame sizes is similar to the
throughput curves (Fig. 7.10). At BER=10−6 , the maximum bandwidth efficiency
for No-ACK, Imm-ACK, Dly-ACK, Blk-ACK are 95.9% (at frame size of 2.5 Kbytes),
93% (4 Kbytes), 87% (1 Kbytes), 88% (1 Kbytes), respectively. The lowest bandwidth
efficiency of 70% is found for 10 Kbytes frame sizes for Blk-ACK corresponding to
2.2Gbps.
We now present the effect of fiber length on the throughput. Fig. 7.11 shows a
119
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
3.2
3.1
Throughput [Mbps]
2.9
2.8
2.7
2.6
No-ACK
2.5 Imm-ACK
Dly-ACK
Blk-ACK
0 1 2 3 4 5 6 7 8 9 10
Frame Sizes [Kbytes]
(a) BER=10−9
Throughput vs. Frame Sizes (BER=10 -7)
3.2
3.1
2.9
Throughput [Mbps]
2.8
2.7
2.6
2.5 No-ACK
Imm-ACK
Dly-ACK
Blk-ACK
0 1 2 3 4 5 6 7 8 9 10
Frame Sizes [Kbytes]
(b) BER=10−7
Throughput vs. Frame Sizes (BER=10 -6)
3.1
2.9
2.8
Throughput [Mbps]
2.7 No-ACK
Imm-ACK
Dly-ACK
2.6 Blk-ACK
2.5
2.4
2.3
2.2
0 1 2 3 4 5 6 7 8 9 10
Frame Sizes [Kbytes]
(c) BER=10−6
Figure 7.9: Throughput for different frame sizes at a fiber length of 1000 m
120
7.5. Conclusion
95
95
90
Bandwidth Efficiency [%]
75
80
No-ACK 70
Imm-ACK
Dly-ACK
Blk-ACK
75
0 1 2 3 4 5 6 7 8 9 10 65
0 1 2 3 4 5 6 7 8 9 10
Frame Sizes [Kbytes]
Frame Sizes [Kbytes]
Figure 7.10: Bandwidth Efficiency for different frame sizes at a fiber length of 1000 m.
steady decrease in the throughput for different fiber lengths at BER 10−9 and 10−6 .
At BER=10−9 , the throughput of the Blk-ACK is greater than the Dly-ACK and the
Imm-ACK; even better than the No-ACK when the fiber length is greater than 8 km.
The drop in throughput when increasing the fiber length from 10 m to 10000 m for
No-ACK and Imm-ACK are 1%, 1.6%, respectively. For example, the throughput of
Imm-ACK at 1000 m fiber length is 3.09 Gb/s, downgraded to 2.9 Gb/s at 9 km fiber
length. For the Dly-ACK mechanisms and the Blk-ACK, the drop in throughput is
less than 1%. At BER=10−6 , the throughput of five different ACK mechanisms are
lesser, than they are at BER=10−9 . It decreases when the fiber length increases.
When frame size is large (at 2048 Bytes), the Blk-ACK performance is lower. This
shows that the effect of fiber while using IEEE 802.15.3 c is tolerable and is a good
candidate for RoF systems. This effect of fiber length on the bandwidth efficiency is
similar to the throughput. When the BER decreases, the bandwidth efficiency also
decreases as illustrated in Fig. 7.12.
7.5 Conclusion
This chapter has introduced two 60 GHz MAC protocols, IEEE 802.15.3 c, IEEE
802.11 ad, and given a brief comparison. The performance of IEEE 802.15.3c is
analysed when it is used in an RoF network in terms of throughput and bandwidth
efficiency for different ACK mechanisms. Since the two standards have many similar
features, IEEE 802.11 ad is analyzed in a like manner. The achieved numerical results
prove that there is a trade-off between throughput and the fiber length. However,
the drop in throughput when increasing the fiber length is acceptable. This is in
121
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
Throughput vs. Fiber Length (BER=10 -9) Throughput vs. Fiber Length (BER=10 -9)
3
3.1
2.9
3.05
2.8
3
2.7
Throughput [Gb/s]
Throughput [Gb/s]
No-ACK 2.95 No-ACK
Imm-ACK Imm-ACK
2.6
Dly-ACK Dly-ACK
Blk-ACK Blk-ACK
2.9
2.5
2.85
2.4
2.3 2.8
2.2 2.75
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Fiber Length [m] Fiber Length [m]
(a) Frame of 512 bytes, BER=10−9 (b) Frame of 1024 bytes, BER=10−9
Throughput vs. Fiber Length (BER=10 -9) Throughput vs. Fiber Length (BER=10 -6)
3.15
No-ACK
2.9 Imm-ACK
Dly-ACK
Blk-ACK
2.8
3.1
2.7
Throughput [Gb/s]
Throughput [Gb/s]
No-ACK 2.6
Imm-ACK
3.05
Dly-ACK
Blk-ACK
2.5
2.4
3
2.3
2.95 2.2
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Fiber Length [m] Fiber Length [m]
(c) Frame of 2048 bytes, BER=10−9 (d) Frame of 512 bytes, BER=10−6
Throughput vs. Fiber Length (BER=10 -6)
Throughput vs. Fiber Length (BER=10 -6) 3.1
3.05
3.05
3
3
2.95
2.95
Throughput [Gb/s]
Throughput [Gb/s]
2.9 No-ACK
No-ACK Imm-ACK
Imm-ACK 2.9
Dly-ACK
Dly-ACK Blk-ACK
Blk-ACK
2.85
2.85
2.8
2.8
2.75 2.75
2.7
2.7 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Fiber Length [m]
Fiber Length [m]
(e) Frame of 1024 bytes, BER=10−6 (f) Frame of 2048 bytes, BER=10−6
Figure 7.11: Throughput for different fiber lengths and different frame sizes in case of BER=10−9
and BER=10−6
122
7.5. Conclusion
85
80
70
65
60
55 NoACK
Imm-ACK
Dly-ACK
Blk-ACK
50
-9 -8 -7 -6 -5 -4
10 10 10 10 10 10
BER
85
Bandwidth Efficiency [%]
80
75
70
65
60
55
-9 -8 -7 -6 -5 -4
10 10 10 10 10 10
BER
90
Bandwidth Efficiency [%]
85
80
75
70
65
60
-9 -8 -7 -6 -5 -4
10 10 10 10 10 10
BER
Figure 7.12: Bandwidth Efficiency for different fiber lengths and different frame sizes in the case
of BER= 10−9 and BER= 10−6
123
7. Feasibility of IEEE 802.15.3c and IEEE 802.11ad for RoF Networks
contrast with the previous results [11, 23, 124], where IEEE 802.11, HiperLAN/2,
and IEEE 802.16, etc., were considered in a RoF system. In the analysis, even when
the fiber length is increased to 10 km, the drop is less than 2%. The results also
show that, the throughput increases while increasing the frame size, the Blk-ACK
and the Dly-ACK performs better than the Imm-ACK in terms of both throughput
and bandwidth efficiency. However, in case of high channel error rate and large
frame size throughput degrades. Thus we plan to study next an adaptive ACK
mechanism that takes into account various error probability, length of fiber, including
the requirements of the applications.
124
Chapter 8
In this chapter, we will review the contributions of the thesis, as well as discussing
about the possible directions of future research.
8.1 Summary
Indoor networks are now used for providing advanced high speed multimedia ap-
plications such as HD and UHDTV, online games, etc. These applications require
a high data rate in the realm of Gb/s. Moreover, wireless technology allows users
to move around in their home with their mobile devices in the most convenient
way. Thus, it requires indoor networks to provide Gbps wireless transmissions.
With the availability of the 5 GHz unlicensed spectrum, the 60 GHz band becomes a
promising air interface for future home networks. However, the high attenuation
loss of the 60 GHz channel makes it that one antenna can only cover one room
(less than 10 m). Hence, to cover a whole building, a high number of antennas
is required. An RoF technique using OFM is employed to extend radio domain
in indoor environments. This technique also helps to transfer the complex signal
processing unit from antennas, into the HCC. Thus, these antennas are simplified.
The network operating at 60 GHz is described in Chapter 2 with two additional
components: movement prediction unit and positioning system. The proposed
system can utilize the advantages of the centralized control architecture to manage
resource management and mobility algorithms easily at the HCC. However, the
extra delay introduced by the distribution optical network is an issue that should
be considered in handoff and MAC performance.
125
8. Summary and Outlook
One of the major issues with popular wireless networks such as GSM, WLAN
and UMTS is handoff. Handoff is even more critical in an RoF network due to
the small cell size, the narrow overlap between two adjacent cells and the delay
introduced by the distribution optical network. Thus a handoff must be completed
in a short time. In Chapter 3, the simulation results show that the conventional
handoff algorithms perform poorly with the high number of handoffs as well as the
high number of unnecessary handoffs. Thus, the system requirements demand to
propose a proper handoff algorithm. Vertical handoffs are also briefly discussed in
this chapter because it is important in future wireless indoor networks (for example
5G network) where multiple networks with different technologies, standards, and
scopes are coexisting.
Based on the evidence that a person tends to move in some pattern in an indoor
environment, the next location of the MS can be predicted with reasonable probability.
The next location of the MS can help the system easily to find the appropriate
cell for performing handoff and effectively manage bandwidth resources. Several
movement prediction techniques have been proposed, for example synthesis mobility
models, using neural networks, Bayesian network and HMMs. Among them, the
HMM is chosen as the prediction model for an RoF network due to its robust and
accuracy (in Chapter 4). In this model, the floor is divided into several areas called
symbols. The antennas are modeled as the hidden states. To predict the location of
the MS, the model is trained by using the Baum-Welch algorithm. The model is
tested with two data sets, the real data of the regular employee and the simulated
data for the guest. It is observed that the predictions from the model are acceptable
in the two cases of employee and guest. The accuracy in the case of the regular
employee is of 84.1%, and more than 54.6% for the guest.
In Chapter 3, it shows that handoff in an RoF network is occurring frequently
and shall be completed in a short time. In the three steps of the handoff process,
it is investigated that the discovery phase is taking about 90% of the total time
of a handoff. Thus, reducing the discovery time is the way to improve handoff
performance. In Chapter 5, the handoff with the assistance of direction and location
information of the MS is introduced. The HCC uses the location information of the
MS to define the potential cells to which the MS will move to next. Thus, the cells
involved in the handoff process are limited. By this way, the time for discovery in a
handoff is reduced. The simulation results show the significant improvement of the
proposed handoff algorithm with the conventional handoff algorithm.
Similar to handoff, resource management is an arduous task in wireless networks
since the MS frequently moves between cells. The small cell size and the extra delay
126
8.1. Summary
make this work even more difficult in an RoF network, since the bandwidth of a cell
is limited. Several calls are dropped if there are more new calls and handoff calls
requesting the cell to allocate their required bandwidth. Using movement prediction,
the proposed bandwidth resource algorithm is performed by the HCC (in Chapter
6). The HCC calculates and reserves the predicted bandwidth for handoff calls in
all cells. The simulation shows that there is a tradeoff between the CDP and the
CBP; but it is possible. The proposed algorithm allows more handoff calls (the lower
CDP) to continue to satisfy the users’ demands. This means that QoS is improved.
Due to the attractive features of 60 GHz, many standards have been introduced
during last few years, for instance, IEEE 802.11 ad, IEEE 802.15.3 c, Wireless HD,
etc. However, the question is could they be applied in an RoF network where the
introduced delay of the optical distribution network may effect MAC protocols. In
Chapter 7, we present two centralized standards for 60 GHz: IEEE 802.11 ad and
IEEE 802.15.3 c. In theory, the two standards have many similar features. Thus,
only IEEE 802.15.3 c is modeled and analyzed with different ACK mechanisms.
The numerical results can be used for IEEE 802.11 ad in a like manner. In an
indoor environment, the drop in throughput with an increase in the fiber length is
acceptable. The Blk-ACK and the Dly-ACK performs better than the Imm-ACK in
terms of both throughput and bandwidth efficiency. Thus, the HCC needs to choose
an adaptive ACK mechanism that takes into account various error probabilities and
lengths of fiber, as well as the requirements of the applications.
The proposed RoF network is a promising candidate for indoor optical wireless
communications. However, it is also challenging at this time. This dissertation is to
solve the challenges that are raised by this network. The overview of contributions
of each chapter is presented in Fig. 8.1.
127
8. Summary and Outlook
8.2 Outlook
Radio over Fiber 60 GHz indoor network is the promising candidate for future home
networks. However, we believe that there are still gaps to be filled when deploying
the proposed system in the real world. Therefore, we raise a number of issues which
will be further addressed in the future.
• Improving the prediction model: The HMM is a very powerful statistical tool
given its modularity and transparency. Moreover, it is incorporating prior
knowledge into its model structure. However, the greatest disadvantages of
HMM are overfitting and the high cost of memory and running time, therefore,
128
8.2. Outlook
it is necessary to look for other training approaches. Getting more data is also
helpful in avoiding overfitting. However, the size of data is constrained by the
training time and the number of iterations. Thus, the solution is to find a
reasonable data size which can help the model to avoid overfitting, reduce the
cost, but still maintaining the high accuracy of the prediction model.
129
8. Summary and Outlook
130
Bibliography
Bibliography
[1] Cisco. (2013, Feb.) Cisco visual networking index: Global mobile data traffic forecast
update, 2012 – 2017. Whitepaper.
[2] Z. Genc, B. L. Dang, J. Wang, and I. G. Niemegeers, “Home networking at 60 GHz:
challenges and research issues,” Annales des Telecommunications, pp. 501–509, 2008.
[3] G. T. Edens, “Home Networking and the CableHome Project at CableLabs,” IEEE
Communications Magazine, vol. 39, no. 6, pp. 112–121, 2001.
[4] J. Wang, “Networking Technologies for Future Home Networks Using 60 GHz Radio,”
Ph.D. dissertation, Technology University of Delft, May 2010.
[5] Cordless advanced technology - internet and quality. [online]. available:
http://www.cat-iq.org/.
[6] IEEE 802.15 Working Groups. [Online]. http://www.ieee802.org/15/.
[7] Homerf resource center. [online]. available: http://www.palowireless.com/homerf/homerf2.asp.
[8] IEEE Standard for Information Technology - Telecommunications and Information
Exchange Between Systems - Local and Metropolitan Area Networks - Specific Re-
quirements. - Part 15.1: Wireless Medium Access Control (MAC) and Physical Layer
(PHY) Specifications for Wireless Personal Area Networks (WPANs), IEEE Std.
802.15.1-2005 (Revision of 802.15.1-2002), 2005.
[9] A. Koonen and E. Tangdiongga, “Photonic home area networks,” Lightwave Technol-
ogy, Journal of, vol. 32, no. 4, pp. 591–604, Feb 2014.
[10] G. Filippo, L. Marco, and R. Ruggero, “Mobile and Personal Communications in the
60 GHz Band: A Survey,” Wirel. Pers. Commun., vol. 10, no. 2, pp. 207–243–, 1999.
[11] B. L. Dang, R. V. Prasad, I. Niemegeers, M. G. Larrode, and A. M. J. Koonen,
“Toward a Seamless Communication Architecture for In-building Networks at the 60
GHz band,” in Proc. 31st IEEE Conf. Local Computer Networks, 2006, pp. 300–307.
[12] D. T. Emerson, “The work of Jagadis Chandra Bose: 100 years of mm-wave research,”
IEEE Transactions on Microwave Theory and Techniques, vol. 45, pp. 2267–2273,
1997.
[13] IEEE 802.15.3c Task Group, IEEE. [Online]. Available:
http://www.ieee802.org/15/pub/TG3c.html
[14] WirelessHD Consortium. [online at] www.wirelesshd.org.
[15] (2011) IEEE 802.11 ad Task Group - Status of Project IEEE 802.11ad [online at]
http://grouper.ieee.org/groups/802/11/Reports/tgad_update.htm.
[16] Wireless Gigabit Alliance. [online at] http://wirelessgigabitalliance.org/.
131
[17] ECMA Internaltion [online at] http://www.ecma-international.org/default.htm.
[18] P. F. M. Smulders, “Broadband wireless LANs: A feasibility study,” Ph.D. dissertation,
Eindhoven Univ., Netherlands, Dec. 1995.
[19] J. Hubner, S. Zeisberg, K. Koora, J. Borowski, and A. Finger, “Simple channel model
for 60 ghz indoor wireless lan design based on complex wideband measurements,” in
Proc. IEEE 47th Vehicular Technology Conf., vol. 2, 1997, pp. 1004–1008.
[20] H. T. Friis, “A Note on a Simple Transmission Formula,” Proceedings of the IRE,
vol. 34, no. 5, pp. 254–256, 1946.
[21] M. Flament, “Propagation and Interference Issues in a 60 GHz Mobile Network,” in
2nd Personal Computing and Commun. Workshop, 1999.
[22] Wireless LAN MAC and PHY Specifications-Enhancements for very high throughput
in the 60 GHz Band,, IEEE 802.11/D9.0 draft specification. Std., July, 2012.
[23] B. L. Dang, M. G. Larrode, R. V. Prasad, I. Niemegeers, and A. M. J. Koonen,
“Radio-over-Fiber based architecture for seamless wireless indoor communication
in the 60GHz band,” Comput. Commun., vol. 30, pp. 3598–3613, December 2007.
[Online]. Available: http://dl.acm.org/citation.cfm?id=1323733.1323773
[24] O. Andrisano, M. Chiani, V. Tralli, and R. Verdone, “Impact of cochannel interfer-
ence on vehicle-to-vehicle communications at millimeter waves,” in Proc. Singapore
ICCS/ISITA ’92. ’Communications the Move’, 1992, pp. 924–928.
[25] J. Gavrilovich, C. D., “Broadband communication on the highways of tomorrow,”
IEEE Communications Magazine, vol. 39, no. 4, pp. 146–154, 2001.
[26] B. Lannoo, D. Colle, M. Pickavet, and P. Demeester, “Radio-over-Fiber-based Solu-
tion to Provide Broadband Internet Access to Train Passengers [Topics in Optical
Communications],” IEEE Communications Magazine, vol. 45, no. 2, pp. 56–62, 2007.
[27] J. Luo, W. Keusgen, A. Kortke, and M. Peter, “A design concept for a 60 ghz wireless
in-flight entertainment system,” in Proc. VTC 2008-Fall Vehicular Technology Conf.
IEEE 68th, 2008, pp. 1–5.
[28] M. Peter, W. Keusgen, and J. Luo, “A Survey on 60 GHz Broadband Communication:
Capability, Applications and System Design,” in Proc. European Microwave Integrated
Circuit Conf. EuMIC 2008, 2008, pp. 1–4.
[29] G. Anzic, D. Connolly, E. Haugland, H. Kosmahl, and J. Chitwood, “A Study of 60
GHz Intersatellite Link Applications,” in NASA TM-83337, 1983.
[30] FSAN group [online at] www.fsanweb.org.
[31] ITU, “ITU-T Recommendation G.983 series: Broadband optical access systems
based on Passive Optical Network (PON). [online at] http://www.itu.int/rec/T-
REC-G.983/en,” International Telecommunication Union, Tech. Rep., 2005. [Online].
Available: http://www.itu.int/rec/T-REC-G.983/en
[32] ——, “ITU-T. G.984 series: Gigabit-capable Passive Optical Networks (GPON. [On-
line at] http://www.itu.int/rec/T-REC-G.983/en,” International Telecommunication
Union, Tech. Rep., 2010. [Online]. Available: http://www.itu.int/rec/T-REC-
G.983/en
[33] IEEE Standard for Information Technology- Telecommunications and Information
Exchange Between Systems- Local and Metropolitan Area Networks- Specific Require-
ments Part 3: Carrier Sense Multiple Access With Collision Detection (CSMA/CD)
Access Method and Physical Layer Specifications Amendment: Media Access Control
Parameters, Physical Layers, and Management Parameters for Subscriber Access
Networks, IEEE Std. 802.3ah-2004, 2004.
[34] IEEE Standard for Information technology - Telecommunications and information
exchange between systems - Local and metropolitan area networks - Specific require-
ments Part 3: Carrier Sense Multiple Access with Collision Detection (CSMA/CD)
Access Method and Physical Layer Specifications Amendment 1: Physical Layer Spec-
ifications and Management Parameters for 10 Gb/s Passive Optical Networks, IEEE
Std. 802.3av-2009 (Amendment to 802.3-2008), 2009.
[35] P. H. Gomes, N. L. S. da Fonseca, and O. C. Branquinho, “LatinCon04 - Analysis
of Performance Degradation in Radio-Over-Fiber Systems Based on IEEE 802.16
Protocol,” IEEE (Revista IEEE America Latina) Latin America Transactions, vol. 8,
no. 4, pp. 377–384, 2010.
[36] J. E. Mitchell, “Performance of OFDM at 5.8 GHz using radio over fibre link,”
Electronics Letters, vol. 40, no. 21, pp. 1353–1354, 2004.
[37] C. Lim, A. Nirmalathas, M. Attygalle, D. Novak, and R. Waterhouse, “On the
merging of millimeter-wave fiber-radio backbone with 25-GHz WDM ring networks,”
IEEE/OSA Journal of Lightwave Technology, vol. 21, no. 10, pp. 2203–2210, 2003.
[38] T. Koonen, A. Ng’oma, P. Smulders, H. van den Boom, I. T. Monroy,
and G.-D. Khoe, “In-House Networks Using Multimode Polymer Optical
Fiber for Broadband Wireless Services,” Photonic Network Communications,
vol. 5, pp. 177–187, 2003, 10.1023/A:1022172511450. [Online]. Available:
http://dx.doi.org/10.1023/A:1022172511450
[39] M. G. Larrode, A. M. J. Koonen, and J. J. V. Olmos, “Fiber-based broadband wireless
access employing optical frequency multiplication,” IEEE Journal of Selected Topics
in Quantum Electronics, vol. 12, no. 4, pp. 875–881, 2006.
[40] M. Flament and A. Svensson, “Virtual Cellular Networks for 60 GHz Wireless
Infrastructure,” in Proc. IEEE Int. Conf. Communications ICC ’03, vol. 2, 2003, pp.
1223–1227.
[41] M. Vossiek, L. Wiebking, P. Gulden, J. Wieghardt, C. Hoffmann, and P. Heide,
“Wireless local positioning,” IEEE Microwave Magazine, vol. 4, no. 4, pp. 77–86,
2003.
[42] H. Liu, H. Darabi, P. Banerjee, and J. Liu, “Survey of Wireless Indoor Positioning
Techniques and Systems,” IEEE Transactions on Systems, Man, and Cybernetics,
Part C: Applications and Reviews, vol. 37, no. 6, pp. 1067–1080, 2007.
[43] Y. Gu, A. Lo, and I. Niemegeers, “A survey of indoor positioning systems for wireless
personal networks,” IEEE Communications Surveys & Tutorials, vol. 11, no. 1, pp.
13–32, 2009.
[45] H. D. Don, S. Jun, H. Jung, and S. W. An, “Using rfid for accurate positioning,” in
Proc. Int. Symp. on GNSS, Australia, 2004.
[47] Handbook of wireless networks and mobile computing. John Willey & Sons Inc.,
USA, 2002.
[48] S.-W. Wang and I. Wang, “Effects of soft handoff, frequency reuse and non-ideal
antenna sectorization on CDMA system capacity,” in Proc. IEEE 43rd Vehicular
Technology Conf, 1993, pp. 850–854.
[51] P. Goyal and S. K. Saxena, “A dynamic decision model for vertical handoffs across
heterogeneous wireless networks,” in Proc. World Academy of Science, Engineering
and Technology 41, July 2008.
[52] P. Marichamy and S. Chakrabarti, “On threshold setting and hysteresis issues of
handoff algorithms,” in Proc. IEEE Int Personal Wireless Communication Conf,
1999, pp. 436–440.
[53] N. Zhang and J. M. Holtzman, “Analysis of Handoff Algorithms using both Absolute
and Relative Measurements,” IEEE Transactions on Vehicular Technology, vol. 45,
no. 1, pp. 174–179, 1996.
[54] Y.-I. Kim, K.-J. Lee, and Y.-O. Chin, “Analysis of multi-level threshold handoff
algorithm,” in Proc. ’Communications: The Key to Global Prosperity Global Telecom-
munications Conf. GLOBECOM ’96, vol. 2, 1996, pp. 1141–1145.
[55] C.-N. Chuah and R. D. Yates, “Evaluation of a Minimum Power Handoff Algo-
rithm,” in Proc. Sixth IEEE Int Personal, Indoor and Mobile Radio Communications
PIMRC’95. ’Wireless: Merging to the Information Superhighway’. Symp, vol. 2, 1995,
pp. 814–818.
[56] M. D. Austin and G. L. Stuber, “Velocity Adaptive Handoff Algorithms for Micro-
cellular Systems,” IEEE Transactions on Vehicular Technology, vol. 43, no. 3, pp.
549–561, 1994.
[57] ——, “Direction biased handoff algorithms for urban microcells,” in Proc. IEEE 44th
Vehicular Technology Conf, 1994, pp. 101–105.
[58] H.-C. Ji, H. Kim, and Y. C. Chung, “Cost-effective radio-over-fiber systems employing
phase-modulated downlink and intensity-modulated uplink,” in Proc. Conf. Optical
Fiber Communication - incudes post deadline papers OFC 2009, 2009, pp. 1–3.
[59] H. B. Kim and A. Wolisz, “Performance evaluation of a MAC protocol for radio
over fiber wireless LAN operating in the 60-GHz band,” in Proc. IEEE Global
Telecommunications Conf. GLOBECOM ’03, vol. 5, 2003, pp. 2659–2663.
[60] H. B. Kim, M. Emmelmann, B. Rathke, and A. Wolisz, “A radio over fiber network
architecture for road vehicle communication systems,” in Proc. VTC 2005-Spring
Vehicular Technology Conf. 2005 IEEE 61st, vol. 5, 2005, pp. 2920–2924.
[61] H. J. Kim and J.-P. Linnartz, “Virtual cellular network: a new wireless commu-
nications architecture with multiple access ports,” in Proc. IEEE 44th Vehicular
Technology Conf, 1994, pp. 1055–1059.
[62] N. Pleros, K. Tsagkaris, and N. Tselikas, “A moving extended cell concept for seamless
communication in 60 GHz radio-over-fiber networks,” IEEE Communications Letters,
vol. 12, no. 11, pp. 852–854, 2008.
[63] Z. Fan, “Wireless networking with directional antennas for 60 GHz systems,” in Proc.
14th European Wireless Conf. EW 2008, 2008, pp. 1–7.
[81] S.-C. Liou and H.-C. Lu, “Applied neural network for location prediction and re-
sources reservation scheme in wireless networks,” in Proc. Int. Conf. Communication
Technology ICCT 2003, vol. 2, 2003, pp. 958–961.
[82] S. Akoush and A. Sameh, “Movement Prediction Using Bayesian Learning for Neural
Networks,” in Proc. Second Int. Conf. Systems and Networks Communications ICSNC
2007, 2007.
[83] J.-M. François and G. Leduc, “AP and MN-centric mobility prediction: a comparative
study based on wireless traces,” in Proceedings of the 6th international IFIP-TC6
conference on Ad Hoc and sensor networks, wireless networks, next generation
internet, ser. NETWORKING’07. Berlin, Heidelberg: Springer-Verlag, 2007, pp.
322–332. [Online]. Available: http://dl.acm.org/citation.cfm?id=1772322.1772358
[84] M. Mozer, “The neural network house: An environment that adapts to its inhabitants,”
in Proc. of the American Association for Arti?cial Intelligence Spring Symposium on
Intelligent Environments . Menlo Park, CA: AAAI Press., 1998, pp. (pp. 110–114).
[86] J. Petzold and et. al., “Prediction of Indoor Movements Using Bayesian Networks,” in
Proc. of Location and Context Awareness (LoCA2005) Germany, pp. 211–222, 2005.
[87] ——, “Comparison of Different Methods for Next Location Prediction,” Euro-Par
2006, LNCS4128, vol. pp.909–918, 2006.
[88] A. Gellert and L. Vintan, “Person Movement Prediction Using Hidden Markov
Models,” in Proc. of Studies in Informatics and Control, Vol. 15, No.1, March., 2006.
[89] P. H. Augustin Chaintreau, “Pocket Switched Networks: Real-world mobility and its
consequences for opportunistic forwarding,” 2006 Computer Laboratory, University
of Cambridge, Tech. Rep., Feb. 2005.
[90] M. McNett and G. M. Voelker, “Access and mobility of wireless PDA users,”
SIGMOBILE Mob. Comput. Commun. Rev., vol. 9, pp. 40–55, April 2005. [Online].
Available: http://doi.acm.org/10.1145/1072989.1072995
[91] M. Kim, D. Kotz, and S. Kim, “Extracting a Mobility Model from Real User Traces,”
in Proc. 25th IEEE Int. Conf. Computer Communications INFOCOM 2006, 2006,
pp. 1–13.
[92] C. Tuduce and T. Gross, “A mobility model based on WLAN traces and its validation,”
in Proc. IEEE 24th Annual Joint Conf. of the IEEE Computer and Communications
Societies INFOCOM 2005, vol. 1, 2005, pp. 664–674.
[93] T. Camp, J. Boleng, and V. Davies, “A survey of mobility models for ad hoc
network research,” Wireless Communications and Mobile Computing, vol. 2, no. 5,
pp. 483–502, 2002. [Online]. Available: http://dx.doi.org/10.1002/wcm.72
[94] L. R. Rabiner, “A tutorial on hidden markov models and selected applications in
speech recognition,” Proceedings of the IEEE, vol. 77, no. 2, pp. 257–286, 1989.
[95] F. Jelinek, Statistical methods for speech recognition. MIT Press, 1997.
[96] V. Q. Bien, R. V. Prasad, and I. Niemegeers, “A direction assisted handoff
algorithm for radio over fiber indoor networks at 60 GHz,” in Proceedings of the
6th International Wireless Communications and Mobile Computing Conference, ser.
IWCMC ’10. New York, NY, USA: ACM, 2010, pp. 864–869. [Online]. Available:
http://doi.acm.org/10.1145/1815396.1815595
[97] L. Zheng, W. Yu, and W. Weiling, “Position Location and Direction Assisted Handoff
Algorithm in DWCS,” in Proc. 15th IEEE Int. Symp. Personal, Indoor and Mobile
Radio Communications PIMRC 2004, vol. 1, 2004, pp. 663–666.
[98] D. Munoz, Rodriguez and K. W. Cattermole, “Multiple criteria for hand-off in cellular
mobile radio,” IEE Proceedings F Communications, Radar and Signal Processing, vol.
134, no. 1, pp. 85–88, 1987.
[99] G. Edwards and R. Sankar, “Hand-off using fuzzy logic,” in Proc. IEEE Global
Telecommunications Conf. GLOBECOM ’95, vol. 1, 1995, pp. 524–528.
[100] B. V. Quang, R. V. Prasad, and I. Niemegeers, “A Survey on Handoffs — Lessons
for 60 GHz Based Wireless Systems,” IEEE Communications Surveys & Tutorials,
vol. 14, no. 1, pp. 64–86, 2012.
[101] A. Mishra, M. Shin, and W. Arbaugh, “An empirical analysis of the ieee 802.11
mac layer handoff process,” SIGCOMM Comput. Commun. Rev., vol. 33, no. 2, pp.
93–102, Apr. 2003. [Online]. Available: http://doi.acm.org/10.1145/956981.956990
[102] W. Schafer and E. Lutz, “Propagation characteristics of short-range radio links at
60 ghz for mobile intervehicle communication,” in Telecommunications Symposium,
1990. ITS ’90 Symposium Record., SBT/IEEE International, 1990, pp. 212–216.
[103] H. ki Kim and J. il Jung, “A Mobility Prediction Handover Algorithm for Effective
Channel Assignment in Wireless ATM,” in Proc. IEEE Global Telecommunications
Conf. GLOBECOM ’01, vol. 6, 2001, pp. 3673–3680.
[105] J. Diederich and M. Zitterbart, “Handoff prioritization schemes using early blocking,”
IEEE Communications Surveys & Tutorials, vol. 7, no. 2, pp. 26–45, 2005.
[108] H. Wu, L. Li, B. Li, L. Yin, I. Chlamtac, and B. Li, “On handoff performance for an
integrated voice/data cellular system,” in Proc. 13th IEEE Int Personal, Indoor and
Mobile Radio Communications Symp, vol. 5, 2002, pp. 2180–2184.
[109] Y.-R. Haung, Y.-B. Lin, and J.-M. Ho, “Performance analysis for voice/data integra-
tion on a finite-buffer mobile system,” IEEE Transactions on Vehicular Technology,
vol. 49, no. 2, pp. 367–378, 2000.
[110] D. Hong and S. S. Rappaport, “Traffic Model and Performance Analysis for Cellu-
lar Mobile Radio Telephone Systems with Prioritized and Nonprioritized Handoff
Procedures,” IEEE Transactions on Vehicular Technology, vol. 35, no. 3, pp. 77–92,
1986.
[111] X. Luo, I. Thng, and W. Zhuang, “A dynamic channel pre-reservation scheme for
handoffs with GoS guarantee in mobile networks,” in Proc. IEEE Int Computers and
Communications Symp, 1999, pp. 404–408.
[113] Y. Zhang and D. Liu, “An adaptive algorithm for call admission control in wireless
networks,” in Proc. IEEE Global Telecommunications Conf. GLOBECOM ’01, vol. 6,
2001, pp. 3628–3632.
[114] M.-H. Chiu and M. A. Bassiouni, “Predictive schemes for handoff prioritization in
cellular networks based on mobile positioning,” IEEE Journal on Selected Areas in
Communications, vol. 18, no. 3, pp. 510–522, 2000.
[115] Z. Xu, Z. Ye, S. V. Krishnamurthy, S. K. Tripathi, and M. Molle, “A New
Adaptive Channel Reservation Scheme for Handoff Calls in Wireless Cellular
Networks,” in Proceedings of the Second International IFIP-TC6 Networking
Conference on Networking Technologies, Services, and Protocols; Performance of
Computer and Communication Networks; and Mobile and Wireless Communications,
ser. NETWORKING ’02. London, UK, UK: Springer-Verlag, 2002, pp. 672–684.
[Online]. Available: http://dl.acm.org/citation.cfm?id=647077.714641
[116] M. Oliver and J. Paradells, “Adaptative Reservation Mechanism for Wireless Net-
works,” in Proc. of the IEEE International Workshop on Mobile Multimedia Commu-
nications (MoMuC), Seoul, Korea, September 1997, pp. 56–59.
[117] S. Choi and K. G. Shin, “A comparative study of bandwidth reservation and admission
control schemes in QoS-sensitive cellular networks,” Wirel. Netw., vol. 6, no. 4, pp.
289–305, Jul. 2000. [Online]. Available: http://dx.doi.org/10.1023/A:1019154001580
[118] W.-S. Soh and H. S. Kim, “Dynamic guard bandwidth scheme for wireless broadband
networks,” in Proc. IEEE Twentieth Annual Joint Conf. of the IEEE Computer and
Communications Societies INFOCOM 2001, vol. 1, 2001, pp. 572–581.
[120] L. Kleinrock and F. Tobagi, “Packet switching in radio channels: Part i–carrier sense
multiple-access modes and their throughput-delay characteristics,” IEEE Transactions
on Communications, vol. 23, no. 12, pp. 1400–1416, 1975.
[121] S. S. Lam, “A carrier sense multiple access protocol for local networks,” Austin, TX,
USA, Tech. Rep., 1979.
[123] e. a. Nandagopalan S., Cordeiro C., “MAC Channel Access in 60 GHz,” IEEE, Tech.
Rep., 2009. [Online]. Available: https://mentor.ieee.org/802.11/dcn/09/11-09-0572-
00-00ad-mac-channel-access-in-60-ghz.ppt
[124] B. L. Dang and I. G. Niemegeers, “Analysis of IEEE 802.11 in Radio over Fiber
Home Networks.” in LCN’05, 2005, pp. 744–747.
[125] B. L. Dang, R. Venkatesha Prasad, and I. Niemegeers, “On the MAC protocols for
Radio over Fiber indoor networks,” in Proc. First Int. Conf. Communications and
Electronics ICCE ’06, 2006, pp. 112–117.
[126] B. Kalantari-Sabet, M. Mjeku, N. J. Gomes, and J. E. Mitchell, “Performance
Impairments in Single-Mode Radio-Over-Fiber Systems Due to MAC Constraints,”
IEEE/OSA Journal of Lightwave Technology, vol. 26, no. 15, pp. 2540–2548, 2008.
142
List of Abbreviations
143
Imm-ACK Immediate ACK
ISI Inter-symbol Interference
ITU International Telecommunication Union
LAN Local Area Network
LLC Link Layer Control
LOS Light of Sight
LTE Long-Term Evolution
M2M Machine to Machine technologies
MAC Media Access Control
MIFS Minimum Inter-frame Space
MS Mobile Station - it refers to a person who carries his MS.
NASA The National Aeronautics and Space Administration, US
NNSS Nearest Neighbors in Signal Space
OFDM Orthogonal frequency-division multiplexing
OFM Optical Frequency Multiplication
PLCP Physical Layer Convergence Procedure
PON Passive Optical Networks
PSTN Public Switched Telephone Network
QAM Quadrature Amplitude Modulation
QoS Quality of Service
QPSK Quadrature Phase Shift Keying
RFID Radio-frequency Identification
RIFS Retransmission Inter-Frame Space
RoF Radio over Fiber
RSS Relative Signal Strength
SIFS Short Inter-frame Space
SIR Signal to Interference Ratio
SNR Signal to Noise Ratio
TDM Time Division Multiplexing
TDMA Time Division Multiple Access
UMTS Universal Mobile Telecommunications System
UWB Ultra Wideband
VCN Virtual Cellular Network
VoD Video on Demand
WDM Wavelength-division multiplexing
WiMAX Worldwide Interoperability for Microwave Access
WLAN Wireless Local Area Network
WPAN Wireless Personal Area Network
xDSL Digital Subscriber Line technologies
Summary
Because of high data rate multimedia applications such as HD and UHDTV, online
games, etc., the future home networks are expected to support short-range gigabit
transmission. With the worldwide availability of 5 GHz spectrum at the 60 GHz band,
it creates the opportunity for a promising air interface for future home networks.
This dissertation is concerned with the application of the 60 GHz band in an indoor
networking environment.
First, we find the solution for the problem of the smaller size of 60 GHz cells.
Indoor networks are more complex since they consist of a high number of access
points. To reduce the complexity of the networks and extend the coverage, a Radio
over Fiber (RoF) technique using OFM, which combines 60 GHz wireless and optical
communication, has been employed. In the proposed network, all signal processing
functions in antennas are now transferred to a central controller. In addition, QoS
and mobility control algorithms can be performed easily.
In indoor environments, the 60 GHz radio link is highly obstructed by objects.
Moreover, the overlapping area between two adjacent cells is small, and the optical
network introduces an extra delay. Both the small overlapping area and the extra
delay make handoff in the proposed network more difficult. To overcome this
problem, a prediction model and a positioning system could be added to an RoF
network. The indoor prediction model is based on a Hidden Markov Model. We
show that this model is working well in an indoor environment in two simulation
cases.
Utilizing the location information of the mobile station (MS), a handoff algorithm
is proposed. This algorithm reduces the number of cells to be considered for
scanning in the handoff process. Thus, the handoff delay is reduced. Bandwidth
resource management is also addressed by reserving bandwidth for handoff calls
only in candidate cells. In this mechanism, only handoff calls can use the reserved
bandwidth in each cell, new calls are not allowed to use it. In that manner, QoS
can be maintained since the call dropping probability is decreasing.
Finally, the feasibility of two centralized 60 GHz MAC protocols, IEEE 802.15.3 c
and IEEE 802.11 ad, is investigated in an RoF network. The major effect on MAC
protocols is the extra delay introduced by the optical links. The analysis indicates
that the drop in the throughput when increasing the fiber length is not significant.
Based on network parameters and the requirements of applications, the system
should apply an adaptive ACK mechanism.
Quang Van Bien
145
Summary
146
Samenvatting
147
Samenvatting
148
Acknowledgements
The research in this thesis has been carried out at Delft University of Technology,
in the Embedded Software group (and the Wireless and Mobile Communications
group) of the Faculty of Electrical Engineering, Mathematics and Computer Science.
Funding for this research was provided by the Vietnamese 322 project and the Dutch
IOPGenCOM project. I gratefully acknowledge their support.
I would like to express my deepest gratitude to my promotor, Prof. Ignas G.M.M.
Niemegeers, for giving me the opportunity to pursue this research. I am grateful for
his support and guidance, and for his recommendations in writing this dissertation.
This dissertation would not be done without the precious time and efforts from
my co-promotor, Dr. R. R. Venkatesha Prasad. His guidance, experience, and
encouragement have been valuable factors to give me the confidence and keep me
on the right track despite the distance in time and space.
I thank my supervisor in Vietnam, Prof. Huong T.V. Nguyen, for her support
during my study time in Vietnam. I would like to heartily express my appreciation
to the members of my PhD committee for their time and effort spent on this thesis.
I am indebted to their helpful comments on the thesis.
I would like to thank my friendly and helpful colleagues: Vijay, Kishor, Xueli,
Shi, Linh, other colleagues from our past Wireless Mobile Communication group
and Embedded System group. Special thanks goes to Linh for introducing the
WMC group to me and his ideas for selecting the research topic; to Kishor for his
collaboration resulting in a section in this dissertation and his kind assistance in the
finalizing stage of my PhD research.
I thank the people who made my living in Delft easier. My sincere appreciation
goes to Ms. Veronique van de Varst from CICAT, for her consideration and support
during my stay in Delft. My special thanks go to the secretaries, Marjon, Laura,
Wendy, Stefanie (former WMC group), and Rina (Embedded Software group) for
their constant help in administrative matters. I am also grateful to Ms. Lien (the
322 project - MOET) and Ms. Trang (HUT).
Many thanks go to the Vietnamese Community in Delft (VCID): Khoai’s family,
Linh, Triết, Toàn (TUD), Toàn (IHE), Hải, Chí, Trung, Việt, Tú, Dương, Dũng,
Nghi., Tuấn Dung, Tuấn Rang, Vân Anh, Hùng and many other people whom I
cannot name here, for their precious kindness and friendliness. Your activities,
football matches, playing AOE games and BBQ at the weekend are forever good
149
memories. I am also grateful to my colleagues from VNPT for their help during the
time of writing this dissertation.
This dissertation is dedicated to my father and mother, who have always un-
conditionally given me all their support. I am also grateful to my sister and my
brother-in-law, Hương and Quân, for always being there for me. Finally, I owe this
dissertation to my beloved children, Minh and Vy, who filled me with energy and
motivation to write it, and to Quỳnh, our excellent chef.
Quang Van Bien
Hanoi, May 2014
Publications
Journals
[J1] Q. Bien, R. V. Prasad, K. Chandra, I. Niemegeers, and H. Nguyen,“Resource
Management in Fi-Wi network", to appear Transactions on Emerging Telecommuni-
cations Technologies, Wiley Online Library, 2014.
[J2] Quang Bien, R. Venkatesha Prasad, and Ignas Niemegeers,“A survey on handoffs
– lessons for 60 GHz based wireless systems." IEEE Communications Surveys &
Tutorials, 14(1):64–86, 2012.
[J3] Quang Bien, R.Venkatesha Prasad, and Ignas Niemieegers, “Handoff in Radio
over Fiber Indoor Networks at 60 GHz", J. of Wireless Mobile Networks, Ubiquitous
Computing, and Dependable Applications, 2010.
Conferences
[C1] Quang Bien, R. Venkatesha Prasad, and Ignas Niemegeers. “A direction assisted
handoff algorithm for radio over fiber indoor networks at 60 GHz." In Proc. of the
6th International Wireless Communications and Mobile Computing Conference,
IWCMC ’10, pages 864–869, New York, NY, USA, 2010. ACM.
[C2] Quang Bien, R. Venkatesha Prasad, Ignas Niemieegers, and N. T. V. Huong.
“A study on handoff issues in radio over fiber network at 60 GHz." in Proc. Third
Int Communications and Electronics (ICCE) Conf, pages 50–54, Vietnam 2010.
[C3] Quang Bien, R. Venkatesha Prasad, Ignas Niemegeers, and Thi Viet Huong
Nguyen. “An approach for movement prediction in radio over fiber indoor network
at 60 GHz." In Proc. Int Communications and Signal Processing (ICCSP) Conf.,
pages 389–393, 2011.
[C4] Quang Bien, R. Venkatesha Prasad, and Ignas Niemegeers,“A handoff scheme
using movement prediction in radio over fiber indoor network at 60 GHz." In
Proc. IEEE Workshops of Int Advanced Information Networking and Applications
(WAINA) Conf, pages 664–669, 2011.
[C5] Quang Bien,“Handoff Management in Radio over Fiber Networks at 60 GHz,"
In Proc. of The Asian Forum on Information and Communications Technology,
AFICT’11, Vietnam 2011.
[C6] R. Venkatesha Prasad, Quang Bien, Kishor Chandra, Xueli An, Ignas G.M.M.
151
Niemegeers, Huong Nguyen,“Analysing IEEE 802.15.3 c protocol in Fi-Wi hybrid net-
works," Proc. of 2013 IEEE Consumer Communications and Networking Conference
(CCNC 2013), pp.749–752, 11-14 Jan. 2013.
[C7] Kishor Chandra, R Venkatesha Prasad, Quang Bien, Ignas G.M.M. Niemegeers,
MD Abdur Rahim, “Analysis of Fi-Wi Indoor Network Architecture based on IEEE
802.15.3 c," In Proc. of 2014 IEEE 11th Consumer Communications and Networking
Conference (CCNC)(CCNC 2014), pp. 113–118.
Table P1. The relation between publications by the author and chapters of this dissertation.
Van Quang Bien was born on May 23rd , 1978, in Thai Binh, Vietnam. He received the
B.S. degree and the M.Sc. degree in Electronics and Telecommunications from Hanoi
University of Technology, Vietnam, in 2001 and 2004, respectively. In August 2008,
he started working toward his Ph.D. degree in Wireless and Mobile Communications
(WMC) group, Delft University of Technology. His research interests include billing,
mobility, home networking, performance analysis of various wireless technologies,
e.g., IEEE 802.15.3 and 802.11. Currently, he is also involving in designing high
data rate wireless local/personal networks.
153