Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Performance Enhancement Schemes with
Packet-level Coding in Wireless Networks
by
Kyu-Hwan Lee
Department of Electrical and Computer Engineering
Graduate School
Ajou University
August, 2015
Performance Enhancement Schemes with
Packet-level Coding in Wireless Networks
Principal Advisor: Jae-Hyun Kim
by
Kyu-Hwan Lee
A Dissertation Submitted to the Graduate School of Ajou University
in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
Department of Electrical and Computer Engineering Graduate School Ajou University
August, 2015
The doctoral dissertation of Kyu-Hwan Lee
is hereby approved.
Graduate School Ajou University June 16th, 2015
i
감사의 글
제가 인턴대학원생을 시작으로 대학원생활을 한지 거의
10년이라는 시간이 지났습니다. 처음엔 걱정도 두려움도 많
았지만, 중간엔 많은 고난과 어려움이 있었지만 포기하지
않고 헤쳐 나와 이제 대학원이라는 둥지에서 알을 깨고 나
와서 박사라는 깃털이 달린 큰 날개를 달아 세상 속으로 높
게 비상하려 합니다.
먼저, 투박한 돌멩이에 불과했던 저를 빛나는 보석으로
만들어 주신 김재현 교수님께 감사드립니다. 항상 모든 일
에 긍정적이며 적극적으로 임하고, 배우는 자세로 겸손하게
살아가라는 교수님의 가르침 덕택에 제가 많은 것을 배울
수 있었고, 그만큼 성장할 수 있었습니다. 앞으로도 교수님
의 가르침을 가슴 속에 새기며 교수님께 자랑스런 제자가
되도록 계속 노력하면서 살아가겠습니다. 그리고 저의 학위
논문을 심사해 주신 김영길 교수님, 노병희 교수님, 조성현
교수님, 최영준 교수님께도 감사 드립니다. 교수님들께서 바
쁘신 가운데도 세심하게 지도해 주셔서 저의 박사학위논문
을 잘 완성 할 수 있었습니다.
무선인터넷 연구실 일원들에게도 감사의 뜻을 전합니다.
특히, 제가 방황하고 있을 때 언제나 저의 멘토가 되어 준
재룡형과 저를 잘 이끌어 준 성민형, 현진형, 지수형에게 감
ii
사드립니다. 그리고 동광빌라에서 동거동락했던 신헌형과
승환형에게도 감사드립니다. 동기 성형이와 같이 졸업하는
광춘이, 정호형, 그리고 잘 따라와 준 Nathy에게도 모두 감
사하다는 마음을 전하고 싶습니다.
항상 옆에서 있어주면서 내게 힘이 되어준 친구들에게도
감사의 뜻을 전합니다. 특히, 내가 힘들 때나 기쁠 때나 언
제나 옆에 있어준 친구이자 또 하나의 가족인 강일과 진희
에게 감사의 마음을 전하고 싶습니다. 그리고, 제가 실의에
빠져 있을 때 먼저 다가와 친구가 되어주고 잘 챙겨준 빨래
방 친구들에게 감사의 마음을 전합니다. 또한, 잊지 못할 생
에 가장 좋은 추억을 남겨 준 미스터마우스와 아이다 공연
을 같이한 친구들에게도 고맙다고 말하고 싶습니다.
마지막으로 항상 아들 잘되길 기도하시면서, 뒷바라지 해
주신 어머니와 아버지, 할머니, 동생에게 고마운 마음을 전
하고 싶습니다. 앞으로 그 믿음에 보답하며 효도 잘하고 동
생 잘 챙기는 사람이 되겠습니다.
제가 박사가 될 수 있도록 도움을 주신 모두에게 이 논문
을 바칩니다.
iii
Abstract
Today, wireless communications have been widely used in many
fields. Wireless communication devices have become an important
part of everyday life for the people. However, there are technical
problems such as the power consumption, the network traffic
optimization, and the reliable data transmission in designing the
wireless network efficiently. One of the possible solutions is a
packet-level coding in the wireless network. Therefore, this
dissertation presents the performance enhancement schemes with
packet-level coding in various wireless networks to overcome
technical challenges.
The first proposed scheme is a power saving mechanism using
network coding (NC) and duty cycling in the bottleneck zone of a
wireless sensor networks (WSNs) to prolong the lifetime of WSNs.
The lifetime of a WSN depends on the power consumption of the
nodes in the bottleneck zone near each sink node, where all sensing
data is collected via the nodes in the bottleneck. However, these
nodes' energy is depleted very quickly because of the heavy traffic
imposed on them. Thus, we propose duty cycling, packet
forwarding, and role switching schemes for nodes in the bottleneck
iv
zone. In our proposed approach, the packet forwarding in the coder
nodes employs random linear network coding (RLNC) to enhance
energy efficiency and reliabili ty of the packet delivery in the
bottleneck zone. We evaluate the performance to show that the
proposed protocol outperforms the conventional system in terms of
the lifetime of WSNs, without reducing the reliability of packet
delivery in the bottleneck zone. In a grid topology network, the
lifetime achieved with the proposed protocol is enhanced as
compared with the conventional system.
The second proposed scheme is a multi-way relay system with
NC in multi-spot beam satellite networks. In particular, we focus
on multiparty video conferencing via a satelli te. Our proposed
protocol uses the multicasting routing information and number of
video frame packets to generate coded packets. The proposed
protocol ensures the reliable transmission of multicasting data for
mobile users using the decoding error rate for the RLNC batch. To
minimize the delay in the link layer, we propose a resource
allocation scheme for multiparty video conferencing with NC in
satellite communications. For the resource allocation, we use
application information acquired by a performance enhancing
proxy. The simulation results show that the achievable rate can be
v
increased by the proposed protocol. The proposed protocol can
also reduce the number of packet transmissions, resulting in the
efficient usage of satellite radio resources. Furthermore, it is
shown that the proposed protocol ensures the reliable transmission
of multicasting data for mobile users by using resources saved by
NC. The average peak signal-to-noise of the video streaming for
mobile users is better than that of the conventional system. As a
result, the visual quality of video streaming services is improved.
The third proposed scheme is a fully reliable file transfer
framework with application layer forward error correction
(AL-FEC) for satellite communications on the move (SOTM)
systems to enhance the network throughput. In particular, we
propose an acknowledgement exchange protocol to ensure the
reliability of the end-to-end data transfer as well as a transmission
control scheme aided by navigation systems to enhance the
resource efficiency in the file transfer framework. The proposed
file transfer framework can predict channel blockage by utilizing
navigation systems. The proposed mechanism then makes it
possible to suspend the data transmission for the duration of the
channel blockage. We also theoretically derive the file transfer
time, the goodput, and the resource efficiency to justify the
vi
effectiveness of the proposed file transfer framework. The
performance results show that the proposed file transfer framework
can significantly enhance the goodput as compared with that of
TCP. Furthermore, the resource efficiency is improved with the aid
of the navigation systems.
Contents
List of Figures xi
List of Tables xv
Abbreviation xvii
1 Introduction 1
1.1 Background and Motivation . . . . . . . . . . . . . . . . 1
1.2 NC in Wireless Networks . . . . . . . . . . . . . . . . . . 5
1.3 AL-FEC in Wireless Networks . . . . . . . . . . . . . . . 6
1.4 Contributions . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Related Work 11
2.1 Sensor Networks with NC . . . . . . . . . . . . . . . . . 11
2.2 MRS and Satellite Networks with NC . . . . . . . . . . . 12
2.3 AL-FEC in Various Networks . . . . . . . . . . . . . . . 14
3 Power Saving Mechanism with NC in WSNs 19
3.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 19
vii
3.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Proposed Power Saving Mechanism . . . . . . . . . . . . 24
3.3.1 Basic Ideas . . . . . . . . . . . . . . . . . . . . . 24
3.3.2 Proposed Node Initiation and Duty Cycling Scheme 26
3.3.3 Proposed Packet Forwarding Scheme . . . . . . . 30
3.3.4 Proposed Role Switching Scheme . . . . . . . . . 35
3.4 Performance Analysis . . . . . . . . . . . . . . . . . . . . 35
3.5 Performance Evaluation . . . . . . . . . . . . . . . . . . 40
3.5.1 Simple Topology . . . . . . . . . . . . . . . . . . 41
3.5.2 Grid Topology Networks . . . . . . . . . . . . . . 48
3.5.3 Ratio of Coder Nodes in Networks . . . . . . . . 56
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4 Multi-way Relay System with NC in MBSNs 61
4.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 Proposed MRS with NC . . . . . . . . . . . . . . . . . . 64
4.2.1 System Model . . . . . . . . . . . . . . . . . . . . 64
4.2.2 Operation of the Proposed NC System . . . . . . 66
4.2.3 Reliability Mode of the Proposed NC System . . 67
4.2.4 Resource Allocation for the Proposed NC System 69
4.2.5 Coefficient Matrix . . . . . . . . . . . . . . . . . . 71
4.3 Theoretical Analysis for MRS with NC . . . . . . . . . . 74
4.4 Performance Evaluation . . . . . . . . . . . . . . . . . . 77
4.4.1 Interested Performance Metrics . . . . . . . . . . 78
4.4.2 Simulation Results . . . . . . . . . . . . . . . . . 81
viii
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5 File Transfer Framework with AL-FEC Aided by Naviga-
tion Systems in SOTM Systems 95
5.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2 Proposed File Transfer Framework . . . . . . . . . . . . 98
5.2.1 System Model . . . . . . . . . . . . . . . . . . . . 98
5.2.2 ACK Exchange Procedure . . . . . . . . . . . . . 100
5.2.3 Transmission Control Aided by Navigation Systems 103
5.2.4 Benefit and Overhead of Proposed Protocol . . . 112
5.3 Theoretical Analysis . . . . . . . . . . . . . . . . . . . . 114
5.3.1 Transfer Time and Goodput . . . . . . . . . . . . 114
5.3.2 Resource Efficiency . . . . . . . . . . . . . . . . . 117
5.3.3 Performance Analysis with Navigation Systems . 118
5.4 Performance Evaluation . . . . . . . . . . . . . . . . . . 120
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6 Conclusion 145
References 147
ix
x
List of Figures
1.1 Classification of performance enhancement schemes with
packet level coding. . . . . . . . . . . . . . . . . . . . . . 4
1.2 Example of NC. . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Example for benefits of AL-FEC in channel blockage. . . 6
3.1 System model of the proposed protocol. . . . . . . . . . . 22
3.2 Example of PSM in an IEEE 802.11 system. . . . . . . . 23
3.3 Overall framework of the proposed protocol. . . . . . . . 25
3.4 PSM with priority in the proposed protocol. . . . . . . . 27
3.5 Reference network model for the simple topology. . . . . 36
3.6 Average packet delivery ratio in the bottleneck zone (R=
50%). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.7 Average packet delivery ratio in the bottleneck zone ac-
cording to R (k= 16). . . . . . . . . . . . . . . . . . . . . 39
3.8 Power consumption in nodes (reference network model). . 43
3.9 Energy efficiency in nodes (reference network model). . . 44
3.10 Network lifetime (reference network model). . . . . . . . 45
3.11 Packet delivery ratio (reference network model). . . . . . 46
xi
3.12 Power consumption according to NC cycling constant (ref-
erence network model). . . . . . . . . . . . . . . . . . . . 47
3.13 Average power consumption of nodes in the bottleneck
zone (grid topology network). . . . . . . . . . . . . . . . 50
3.14 Average energy efficiency of nodes in the bottleneck zone
(grid topology network). . . . . . . . . . . . . . . . . . . 51
3.15 Average traffic load of nodes in the bottleneck zone (grid
topology network). . . . . . . . . . . . . . . . . . . . . . 52
3.16 Average network lifetime (grid topology network). . . . . 53
3.17 Average packet delivery ratio in the bottleneck zone (Grid
topology network). . . . . . . . . . . . . . . . . . . . . . 54
3.18 Average power consumption by RLNC encoding in the
coder node (Grid topology network). . . . . . . . . . . . 55
3.19 Average lifetime of nodes in the bottleneck zone according
to the ratio of coder nodes (grid topology network). . . . 57
3.20 Average packet delivery ratio in the bottleneck zone ac-
cording to the ratio of coder nodes (grid topology network). 58
4.1 System model of the proposed NC system in MBSNs. . . 64
4.2 System architecture of the proposed NC system. . . . . . 66
4.3 System model of resource allocation request. . . . . . . . 70
4.4 Resource allocation example. . . . . . . . . . . . . . . . . 70
4.5 MRC model with S spot-beams, each of which is composed
of K distinct terminals. . . . . . . . . . . . . . . . . . . . 73
4.6 Achievable symmetric rate. . . . . . . . . . . . . . . . . . 83
xii
4.7 NC gain in a single-spot beam. . . . . . . . . . . . . . . 84
4.8 NC gain in multi-spot beams (N = 8). . . . . . . . . . . 85
4.9 Average power consumption by RLNC encoding in the
satellite . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.10 Average frame loss rate for a mobile user (S = 1, N = 4). 89
4.11 Average PSNR for a mobile user. . . . . . . . . . . . . . 90
4.12 Visual quality of video streaming service (Conv.). . . . . 91
4.13 Visual quality of video streaming service (Prop.). . . . . 92
4.14 NC gain with a mobile user (S = 1, N = 4). . . . . . . . 93
5.1 System model of the proposed reliable file transfer frame-
work. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.2 Channel model of SOTM. . . . . . . . . . . . . . . . . . 101
5.3 System architecture with navigation systems. . . . . . . . 110
5.4 Example of proposed framework with navigation systems
(SPCB). . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.5 Example of the overhead for the proposed protocol. . . . 112
5.6 Markov chain model for average file transfer time of the
proposed framework. . . . . . . . . . . . . . . . . . . . . 115
5.7 Packet delivery ratio in UDP with and without prop. frame-
work. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.8 Average file transfer time in the city environment. . . . . 126
5.9 Average goodput in the city environment. . . . . . . . . 127
5.10 Average resource efficiency in the city environment. . . . 128
xiii
5.11 Average file transfer time in the SOTM environment (Sce-
nario 1). . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.12 Average goodput in the SOTM environment (Scenario 1). 131
5.13 Average resource efficiency in the SOTM environment (Sce-
nario 1). . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.14 Average goodput in the SOTM environment (Scenario 2). 135
5.15 Average resource efficiency in the SOTM environment (Sce-
nario 2). . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.16 Average goodput in the SOTM environment (Scenario 3,
blockage rate = 20%). . . . . . . . . . . . . . . . . . . . 137
5.17 Average resource efficiency in the SOTM environment (Sce-
nario 3, blockage rate = 20%). . . . . . . . . . . . . . . . 138
5.18 Average goodput in the SOTM environment (Scenario 4). 140
5.19 Average goodput in the SOTM environment (Scenario 5). 141
5.20 Average resource efficiency in the SOTM environment (Sce-
nario 5). . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
xiv
List of Tables
2.1 Comparison of conventional works: Sensor networks with
NC for the power saving . . . . . . . . . . . . . . . . . . 16
2.2 Comparison of conventional works: MRS and satellite net-
works with NC . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Comparison of conventional works: AL-FEC in various
networks . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1 Parameters in the performance evaluation. . . . . . . . . 40
4.1 Specifications for the video sources in the simulation. . . 78
5.1 Parameters in the performance analysis. . . . . . . . . . 121
5.2 Scenario for simulation. . . . . . . . . . . . . . . . . . . . 124
5.3 Goodput and resource efficiency(Scenario 3, Blockage rate
= 40%, α = 60%, LF = 900Mbits). . . . . . . . . . . . . 134
xv
xvi
Abbreviation
3GPP 3rd generation partnership project
ACK acknowledgement
ACM adaptive coding and modulation
AL-FEC application-layer forward error correction
ATIM announcement traffic indication message
B frame bi-predictive frame
CBP channel blockage predictor
CBR constant bit rate
CSMA/CA carrier sense multiple access with collision avoidance
DAMA demand assignment multiple access
DF decoding and forward
DVB-RCS digital video broadcasting-return channel via satellite
DVB-S digital video broadcasting-satellite
xvii
FEC forward error correction
GEO geosynchronous earth orbit
GIS geographic information system
GOP group of pictures
GPS global positioning system
I frame intra frame
IoT internet of things
IP internet protocol
IRIS internet router in space
LDPC low-density parity-check
LTE-A long term evolution advanced
MBMS multimedia broadcast multicast service
MBSN multi-spot beams satellite network
MF-TDMA multi-frequency time division multiple access
MRC multi-way relay channel
MRS multi-way relay system
NACK negative acknowledgement
NC network coding
NCC network control center
xviii
P frame predicted frame
P2P peer-to-peer network
PCS predicted channel state
PEP performance enhancing proxy
PL-FEC packet-level FEC
PSM power saving mechanism
PSNR peak signal-to-noise ratio
R-ACK receiver-ACK
RCST return channel satellite terminal
RLNC random linear network coding
RTT round trip time
S-ACK sender-ACK
SF super-frame
SNR signal-to-noise ratio
SOTM satellite communications on the move
SPCB static predicted channel blockage
TCP transmission control protocol
TX transmission
UDP user datagram protocol
xix
VBR variable bit rate
VPCB variable predicted channel blockage
WiFi wireless fidelity
WiMAX worldwide interoperability for microwave access
WoT web of things
WSN wireless sensor network
XOR exclusive OR
xx
1
Introduction
1.1 Background and Motivation
With the development of wireless networking technologies such as LTE-
A, WiFi, Bluetooth, ZigBee and DVB-S/RCS, wireless communications
have been widely used in both commercial and military fields. Indeed,
wireless communication devices such as smart phones, tablet PCs, wire-
less sensors and potable satellite terminals have become a crucial part
of everyday life. Through wireless communication devices, users are pro-
vided with various services such as voice, video streaming, file download
and e-mail anytime anywhere [1]. However, many technical challenges
remain in designing wireless networks to support these wireless commu-
nication devices as follows.
� Power consumption: Basically, the power consumption in the com-
munication module is one of the most important issues for the bat-
1
tery life of wireless communication devices [2]. Furthermore, green
IT in wireless networks has been issued to reduce the CO2 emission
recently [3]. Therefore, the power consumption should be consid-
ered in designing wireless networks.
� Network traffic optimization: Recently, demand for multimedia ser-
vices that consume heavy bandwidth rapidly increases. However,
users can experience the poor quality of services because limited
radio resource can be allocated to wireless communication devices
[4]. Therefore, through the network traffic optimization, the radio
resources in wireless networks should be efficiently used.
� Reliable data transmission: In wireless communications, channel
coding is important because it ensures the reliability of data trans-
mission protecting it from the data corruption by noise and in-
terference. However, for mobile users, channel blockage due to
intermittent shadowing can cause packet loss even though channel
coding is applied [5]. Thus, the reliable transmission for mobile
user should be considered in designing wireless networks.
Recently, a new paradigm named “packet level coding” has been intro-
duced in many literatures for wireless networks to solve these problems.
Packet level coding is coding across packets in upper layers such as link,
network, transport, and application layer. Generally, NC and AL-FEC
as the packet level coding are widely used in wireless networks [6–13]. It
can cooperate with various communication protocols. The performance
enhancement scheme with packet level coding is able to reduce the net-
2
work traffic and the power consumption. Furthermore, it can enhance
the reliability of data transfer in severe fading such as channel blockage
caused by the user mobility. It also provides the secure transmission in
the wiretap network. The classification of the performance enhancement
scheme with packet level coding is shown in Fig. 1.1. This dissertation
considers some applications for the NC and AL-FEC as shown in Fig.
1.1.
3
Pac
ket l
evel
co
ding
AL
-FE
CT
ype
App
.
Tar
get
syst
em
Sec
ure
tran
smis
sion
Rel
iabl
e tr
ansm
issi
on
Cen
tral
ized
ne
twor
kD
istr
ibut
ed
netw
ork
Uni
cast
Mut
icas
tU
nica
stM
ulti
cast
Uni
cast
Mul
tica
stW
iret
ap
netw
ork
Ch
ap. 3
. Pow
er S
avin
g M
echa
nis
m w
ith
NC
in W
SN
sC
hap
. 4. M
ult
i-w
ay R
elay
Sys
tem
w
ith
NC
in M
BS
Ns
Ch
ap. 5
. Fil
e T
ran
sfer
F
ram
ewor
k w
ith
AL
-FE
C
Aid
ed b
y N
avig
atio
n S
yste
m
in S
OT
M S
yste
m
Rel
iabl
e tr
ansm
issi
onT
raff
ic
opti
miz
atio
nP
ower
sa
ving
Net
wor
k co
ding
Fig. 1.1. Classification of performance enhancement schemes with packet
level coding.
4
1 2X X 1 2X X
2X1X
Fig. 1.2. Example of NC.
1.2 NC in Wireless Networks
Recently, NC has attracted a great deal of research interest in wired and
wireless network systems since its introduction in the area of information
theory [6, 7]. Generally, it is known that NC has the potential to yield
better throughput and reliability for networks of both unicast and mul-
ticast applications. In particular, the number of packets sent by nodes
can be reduced in a network where there exist coding structures such as
chain as shown in Fig. 1.2 [6]. In this figure, node Relay with NC receives
packets (X1, X2) from both node Alice and Bob. node Relay relays a
coded packet X1⊕X2 to both nodes using the broadcasting. Node Alice
can decode the data received from node Relay using the information in
the packet it transmitted to node Relay (i.e., X1 ⊕ (X1 ⊕X2) = X2).
Similarly, node Bob can also decode the data. In the absence of net-
work coding, node Relay has to transmit X1 and X2 separately using
two transmissions. Thus, NC saves the bandwidth and the power by
reducing the number of transmissions. In NC, RLNC is useful to select
code. Generally, code construction algorithms that are deterministic al-
gorithms require a centralized design based on information on all nodes
in networks. However, the implementation of the centralized design is
uneasy in networks with many nodes. In RLNC, it can be able to design
5
Packet loss = 20%(Channel coding)
Packet loss = 0%(Channel coding
+ AL-FEC)
AL-FECDecoding
Source block
Channel blockage
Source block Repair block
Channel coding works
Channel coding fails
Fig. 1.3. Example for benefits of AL-FEC in channel blockage.
practical protocol because every coding coefficient is randomly selected
[8].
1.3 AL-FEC in Wireless Networks
As mentioned above, channel blockage can cause packet loss even though
channel coding is applied. To solve this problem, many studies have in-
vestigated AL-FEC in many communication systems [10–13]. AL-FEC
covers the packet loss not recovered by channel coding because it is ap-
plied above layer 2 and uses the fountain code known as the rateless
erasure code. Recently, Raptor code has been commercially used as the
fountain code because of its dynamic packet loss protection, exception-
ally high computational efficiency, and low transmission and reception
overheads [10, 11]. Therefore, its advantages allow a software implemen-
6
tation and also provide end-to-end error correction without requiring any
change in legacy standards, resulting in ease of deployment in the net-
work [10]. Fig. 1.3 shows the example for the benefit of AL-FEC in the
channel blockage [10]. It is shown that AL-FEC successfully decodes the
original source block in the environment with channel blockages because
AL-FEC interleaves and corrects the packet loss thanks to longer source
block and rateless coding [10, 12, 13]. Actually, long time interleaving can
be implemented in channel coding of DVB-SH and DVB-NGH systems
[14, 15]. However, it implies longer latency and hardware memory at the
receiver. It only supports the static coding rate. Therefore, AL-FEC can
be one of promising solutions for SOTM systems.
1.4 Contributions
The goal of this dissertation is overcoming technical challenges in various
wireless networks through the performance enhancement schemes with
packet-level coding. As mentioned above, the performance enhancement
schemes with packet level coding is not only able to reduce the network
traffic and the power consumption but also can enhance the reliability of
data transfer in severe fading. The followings are the principal contribu-
tions of this dissertation:
� Power saving mechanism with NC in the bottleneck zone of multi-
media sensor networks was presented. Both NC and duty cycling
are considered in the proposed protocol for the efficient power sav-
ing. The proposed protocol outperforms the conventional system
7
in terms of the lifetime of WSNs, without reducing the reliability
of packet delivery in the bottleneck zone.
� MRS with NC in MBSNs was presented. The proposed proto-
col uses the multicasting routing information and number of video
frame packets to generate coded packets. The proposed protocol
ensures the reliable transmission of multicasting data for mobile
users using the decoding error rate for the random linear network
coding batch. The proposed protocol can reduce the number of
packet transmissions, resulting in the efficient usage of satellite ra-
dio resources.
� Transfer time analysis of the file transfer scheme with AL-FEC in
SOTM networks was presented. The proposed file transfer scheme
uses the two-way ACK exchange mechanism. The file transfer time
of the reliable file transfer scheme with AL-FEC by Markov chain
was analyzed.
� Efficient file transfer framework with AL-FEC aided by navigation
systems for SOTM systems was presented. A transmission control
scheme aided by navigation systems is proposed to enhance the re-
source efficiency. The proposed file transfer framework can predict
channel blockage by utilizing navigation systems. The proposed file
transfer framework can not only significantly enhance the goodput
as compared with that of TCP but also improve the resource effi-
ciency with the aid of the navigation systems.
8
This dissertation concentrates on some research areas of the perfor-
mance enhancement schemes with packet level coding as shown in Fig.
1.1. Therefore, in this dissertation, performance enhancement schemes
using NC and AL-FEC are proposed for WSNs, MRS, and SOTM net-
works.
9
10
2
Related Work
2.1 Sensor Networks with NC
NC is a promising technique that reduces network congestion by com-
bining packets for distinct destinations. Many studies have investigated
WSNs with NC to enhance their energy efficiency [16–27]. Platz et al.
studied the energy efficiency of NC in all-to-all broadcast applications
[16]. Shwe et al. enhanced the energy efficiency of the NC scheme in
WSNs by using neighbor discovery [17]. In [18, 19], researchers solved the
optimization problem in WSNs with NC, based on theory. In [20, 21], the
authors proposed NC-based multipath routing for improved energy effi-
ciency in WSNs. In [22], the authors proposed NC-based energy-efficient
data fusion and transmission in WSNs with heterogeneous receivers.
Glatz et al. implemented energy-harvesting aware routing and oppor-
tunistic NC in WSNs using TinyOS [23]. In [24], the authors solved the
optimization problem for network lifetime and video distortion in mul-
11
timedia sensor networks. These previous works improved the efficiency
of power consumption in WSNs by using NC. However, most previous
works do not consider the bottleneck zone. They also do not take into
account interworking with duty cycling, which is a major technique for
power saving in WSNs. Recently, power saving mechanisms considering
both NC and duty cycling have been studied [25–27]. In [25], Chandanala
et al. proposed DutyCode, combining NC with duty cycling by using in-
formation on packet streaming in flooding-based WSNs. However, they
did not exploit power saving in the nodes of a bottleneck zone around a
sink node. In [26, 27], researchers derived the upper bound of network
lifetime in WSNs using random duty cycling and NC. They showed that
the lifetimes of coder nodes were prolonged because nodes in the bot-
tleneck zone encoded packets using XOR NC. However, XOR NC can
reduce the reliability of the sensing data delivery. Furthermore, nodes
other than coder nodes in the bottleneck zone do not benefit from NC
in terms of power consumption. Table 2.1 summarizes the comparison of
conventional works for sensor network with NC.
2.2 MRS and Satellite Networks with NC
An MRS such as multiparty video conferencing in satellite networks is
a system where N users (N ≥ 2) exchange data via a relay and each
user sends its data to all other users [28, 29]. NC is a potential tech-
nique to reduce bandwidth in MRS. There are many theoretical stud-
ies investigating MRS with NC [28–34]. The study in [28] showed the
12
achievable rate region and optimal diversity multiplexing tradeoff of sev-
eral strategies in a half-duplex MRS. The authors of [29] focused on the
achievable rate region in a full-duplex MRS. The study in [30] proposed
a functional-decode-forward coding strategy with rate splitting and joint
source-channel decoding in an MRS over finite fields. The study in [31]
focused on an MRS with three users and proposed code design. The
authors of [32] proposed resource allocation for asymmetric MRS over
an orthogonal channel. Other works on MRS was done in [33, 34] where
the authors considered regenerative relaying and multi-group multi-way
relaying.
There are some studies that investigated NC in satellite communi-
cations [35–39]. The studies focused on reliable multicasting in satellite
communications with NC [35–37]. The authors of [35] and [36] used NC
as a forward error correction method for multicasting in a lossy environ-
ment. In the multicasting with NC, multicasting data is more minimized
than in traditional reliable multicasting because each node only needs to
receive a sufficient number of coded packets to successfully decode the
data without packet ordering. The study in [37] proposed an NC proto-
col for broadcast streaming applications over hybrid satellite systems. It
used proactive retransmissions without prior knowledge of the lost pack-
ets and reactive network coded retransmissions from NACK messages,
resulting in a lower delay and higher throughput in a lossy environment.
Other studies focused on the load balancing in a multi-beam satellite
system with NC [38, 39]. In case of handover between beams, this pro-
tocol applies NC to the multiple routes created by overlapping narrow
13
beams. Thus, this protocol can improve the system’s stable throughput
because of the cooperation among neighboring beams. Furthermore, the
use of random linear network coded packets in the protocol can enhance
reliability and simplify the allocation of packets to the different beams.
Table 2.2 summarizes the comparison of conventional works for MRS and
satellite networks with NC.
2.3 AL-FEC in Various Networks
Since Raptor code is considered in the 3GPP standardization, many stud-
ies have focused on AL-FEC for streaming and the download delivery ser-
vices in LTE MBMS [40–45]. In [40], the file download time is analyzed
in an adaptive LDPC AL-FEC system for multicast content distribu-
tion with loss rate information feedback. They showed that the LDPC
AL-FEC codes achieve a download time that is almost similar to that ob-
tained with ideal rateless codes with less coding complexity. The authors
of [41–45] presented various simulation results to verify the enhanced per-
formance of Raptor code in the LTE MBMS system in terms of AL-FEC
transmission overhead, physical layer parameters, service coverage, and
tune-in delay.
There are some literatures on AL-FEC in the various applications
and networks [46–50]. The authors of [46] analyzed the performance of
fountain codes in multi-hop relay networks. In [47], the authors proposed
a file exchange scheme with AL-FEC in vehicular ad hoc networks to
improve the network throughput. The authors of [48] applied AL-FEC
14
to WiFi multicast streaming inside the high-speed trains. The authors
of [49] presented an analysis of the application of AL-FEC to the domain
of P2P streaming. In [50], authors proposed the AL-FEC scheme based
on the Chinese remainder theorem applied to streaming over WiMAX
networks for high speed rail reception.
Some studies focused on AL-FEC for the streaming service through
broadcasting systems such as satellite broadcasting systems [51–54]. The
authors of [51] applied AL-FEC protection to digital video broadcasting-
terrestrial services. In [52], the authors investigated the potential gain
that can be obtained in digital video broadcasting-handheld using AL-
FEC for mobile terminals. They also analyzed the overhead and the
latency of AL-FEC. The authors of [53] proposed AL-FEC with a long
time inter-leaver and fast tune-in for mobile satellite television services.
In [54], the authors implemented an inter-layer protection scheme with
AL-FEC to protect video streaming services of scalable video coding
based on real-time transport protocol. Table 2.3 summarizes the com-
parison of conventional works for AL-FEC in various networks
15
Table
2.1.
Com
pariso
nof
conven
tional
work
s:S
ensor
netw
orks
with
NC
forth
ep
ower
savin
g
Work
sC
onsid
erationof
Con
sideration
ofO
bjective
Duty
cyclin
gB
ottleneck
zone
[16]N
oN
oD
esignof
NC
schem
efor
all-to-allbroad
castap
plication
s
[17]N
oN
oD
esignof
NC
schem
eusin
gneigh
bor
discovery
[18,19]
No
No
Optim
izationof
the
netw
orklifetim
ein
WSN
sw
ithN
C
[20,21]
No
No
Design
ofN
C-b
asedm
ultip
athrou
ting
[22]N
oN
oD
esignof
NC
-based
energy
-efficien
tdata
fusion
[23]N
oN
oIm
plem
entation
ofop
portu
nistic
NC
schem
eusin
gT
inyO
S
[24]N
oN
oO
ptim
izationof
the
netw
orklifetim
ean
dth
evid
eoquality
[25]Y
esN
oD
esignof
packet
floodin
gcom
bin
ing
NC
with
duty
cyclin
g
[24]Y
esY
esA
naly
sisof
the
netw
orklifetim
ein
bottlen
eckzon
e
usin
gran
dom
duty
cyclin
gan
dN
C
16
Table
2.2.
Com
pari
son
ofco
nve
nti
onal
wor
ks:
MR
San
dsa
tell
ite
net
wor
ks
wit
hN
C
Wor
ks
Tar
get
syst
emO
bje
ctiv
e
[28]
MR
SD
evia
tion
ofth
eac
hie
vable
rate
ina
hal
f-duple
xM
RS
[29]
MR
SD
evia
tion
ofth
eac
hie
vable
rate
ina
full-d
uple
xM
RS
[30]
MR
SD
esig
nof
afu
nct
ional
-dec
ode-
forw
ard
codin
gst
rate
gyin
MR
S
[31]
MR
SC
ode
des
ign
inM
RS
wit
hth
ree
use
rs
[33,
34]
MR
SD
esig
nof
MR
Sw
ith
rege
ner
ativ
ere
layin
gan
dm
ult
i-gr
oup
mult
i-w
ayre
layin
g
[35–
37]
Bro
adca
stvia
sate
llit
eD
esig
nof
forw
ard
erro
rco
rrec
tion
met
hod
usi
ng
NC
for
bro
adca
stin
g
[38,
39]
MB
SN
Des
ign
ofdat
atr
ansm
issi
onw
ith
NC
for
mult
iple
route
s
17
Table
2.3.
Com
parison
ofcon
vention
alw
orks:
AL
-FE
Cin
various
netw
orks
Work
sT
argetsy
stemO
bjective
[40]LT
EM
BM
SA
naly
sisof
file
dow
nload
time
[41–45]LT
EM
BM
SP
erforman
ceevalu
ationof
Rap
torco
de
inLT
EM
BM
S
[46]M
ulti-h
oprelay
netw
orkP
erforman
cean
alysis
offou
ntain
codes
[47]veh
icular
ad-h
oc
netw
orks
Design
ofa
file
exch
ange
schem
ew
ithA
L-F
EC
[48]M
ulticast
via
WiF
iD
esignof
the
AL
-FE
Csch
eme
inm
ulticast
insid
eth
ehigh
speed
train
[49]P
2Pstream
ing
Design
ofth
eA
L-F
EC
schem
efor
P2P
streamin
g
[50]W
iMA
Xnetw
orks
Design
ofth
eA
L-F
EC
schem
ebased
onC
hin
eserem
ainder
theorem
[51]B
roadcastin
gsy
stemD
esignof
AL
-FE
Cprotection
forvid
eobroad
casting-terrestrial
services
[52]D
VB
-Ssy
stemD
esignof
AL
-FE
Csch
eme
form
obile
termin
al
[53]D
VB
-Ssy
stemD
esignof
AL
-FE
Csch
me
with
alon
gtim
ein
ter-leaveran
dfast
tunin
g
[54]V
ideo
streamin
gsy
stemIm
plem
entation
ofin
ter-layerprotection
with
AL
-FE
C
18
3
Power Saving Mechanism
with NC in WSNs
3.1 Motivation
WSNs have been used for many applications, such as the military surveil-
lance, disaster monitoring, and environmental monitoring [55–57]. In the
future Internet, WSNs are expected to be major components in the IoT
and the WoT [58, 59]. In a sensor network, thanks to low-cost hardware
miniaturization and advances in wireless communication technologies, a
large number of sensors can be deployed to monitor target areas and
acquire sensing data that is autonomously collected in sink nodes [57].
However, WSN designs have many restrictions, such as fault tolerance,
scalability, network topology, hardware constraints, and power consump-
tion. Among these, power consumption is one of the most important
19
issues related to the lifetime of a WSN. The sensor nodes in a bottleneck
zone, which is the area near a sink node, can experience node failure
quickly because all sensing data is collected in sink nodes from the other
nodes in the bottleneck zone, which imposes heavy traffic on those nodes.
Thus, the lifetime of a WSN is mainly determined by the lifetime of the
sensor nodes in bottleneck zones [2]. For this reason, this chapter focuses
on a power saving mechanism for the bottleneck zones of WSNs.
In this chapter, we propose a power saving mechanism using NC and
duty cycling in the bottleneck zone of WSNs to enhance the lifetime of
WSNs. In particular, we propose a duty cycling scheme that yields more
sleeping opportunities for nodes in the bottleneck zone. Furthermore,
our power saving mechanism uses RLNC for packet encoding to enhance
energy efficiency without reducing the reliability of the packet delivery
in the bottleneck zone. Nodes in the bottleneck zone periodically switch
roles to prolong the lifetime of all nodes in the bottleneck zone by using
NC. The main contributions of the chapter are as follows:
� A duty cycling scheme for power saving in nodes of the bottleneck
zone.
� A packet forwarding scheme using RLNC in coder nodes to enhance
the energy efficiency and the reliability in the bottleneck zone.
� A role switching scheme for nodes of the bottleneck zone to prolong
the lifetime of WSNs.
20
3.2 System Model
The system model is composed of a sink node and sensor nodes, as shown
in Fig. 3.1. All sensor nodes sense data periodically, and then the sensing
data generated by all sensor nodes are aggregated at a sink node. To
improve reliability of packet delivery in WSNs, sensing data travel via
multi-path forwarding from a sensor to a sink [60, 61]. Thus, a node
broadcasts the data for multi-path forwarding. Around a sink node,
there is a bottleneck zone (B) as shown in Fig. 3.1. The bottleneck
zone B is defined as the area within distance D from the sink node,
where D is the transmission range of the sensor nodes [2, 62]. Thus,
nodes in B consume more power than nodes outside B because all data
are relayed through the nodes in B. In this dissertation, we consider a
fixed multimedia sensor network used in applications such as multimedia
surveillance, traffic monitoring and control systems, and environmental
monitoring [55–57]. In the multimedia sensor network, the high data rate
the low latency is needed in physical and link layers. Therefore, an IEEE
802.11 system is used to convey the sensing data [57].
In a WSN, sensor nodes have constrained energy resources, although
the sink has no such limitation. Therefore, power saving is very impor-
tant to the network lifetime of WSNs. In an IEEE 802.11 system, a PSM
exists that uses an ATIM [63]. The procedure of the PSM is shown in
Fig. 3.2. In an IEEE 802.11 system, time is divided into beacon in-
tervals by means of a distributed protocol using beacon transmission.
At the start of each beacon interval, all nodes stay awake to announce
21
Sin
k
Bot
tlen
eck
Zon
e (B
)
D
Fig. 3.1. System model of the proposed protocol.
22
ATIM window
Beacon interval Beacon interval
Sleeping
Sleeping
SleepingSleeping
ATIM window
Node A
Node B
Node C
ATIM
ATIM-ACK
Data
ACK
Fig. 3.2. Example of PSM in an IEEE 802.11 system.
the packet transmission for an ATIM window. For example, node A an-
nounces packets destined for node B by transmitting an ATIM frame
during the ATIM window. Upon receiving the ATIM frame, node B re-
sponds by sending an ATIM-ACK message. This message is transmitted
using CSMA/CA in IEEE 802.11. When the node has sent an ATIM
frame or ATIM-ACK message to another node, such as node A or B
in Fig. 3.2, the node remains awake for the entire beacon interval to
transmit packets. If a node has not received an ATIM frame and has no
data packet to be transmitted, it can go into a sleeping state, like node
C in Fig. 3.2, resulting in power saving. All sleeping nodes wake up
again in the ATIM window at the start of the next beacon interval. For
multicasting and broadcasting, ATIM-ACK is not transmitted.
23
3.3 Proposed Power Saving Mechanism
3.3.1 Basic Ideas
To enhance the lifetime of nodes in a bottleneck zone, we propose a
power saving mechanism using NC and duty cycling. The basic ideas of
the proposed protocol are as follows.
� Duty cycling: To enhance the network lifetime, it is important to
reduce the power consumption of the nodes in the bottleneck zone.
In the proposed duty cycling scheme, we thus consider the role of
the nodes in packet reception and forwarding to reduce their power
consumption.
� Packet forwarding: Coder nodes with NC reduce the network load
thanks to packet encoding, resulting in power saving in the nodes.
However, XOR NC can reduce the reliability of data delivery. Fur-
thermore, if the packet forwarding with NC does not cooperate
with duty cycling, significant power saving cannot be achieved in
the WSN. Therefore, in the proposed packet forwarding scheme,
we use RLNC and the gathering of data packets in coder nodes to
enhance the reliability of data delivery and the efficiency of power
consumption of nodes in the bottleneck zone.
� Role switching: Only coder nodes benefit from NC in terms of
power consumption. Thus, we propose a role switching scheme
that periodically executes role switching among the nodes in the
bottleneck zone to prolong their lifetime.
24
Tim
er M
anag
emen
t
Rol
e S
elec
tion
Pow
er S
avin
g M
anag
emen
t
Pac
ket R
ecep
tion
Pac
ket F
orw
ardi
ng
NC
Enc
oder
Pac
ket q
ueui
ng
Dut
y cy
clin
g In
fo.
Info
. on
Com
plet
ion
of
Pac
ket R
ecep
tion
Rol
e In
fo. In
fo. o
n C
ompl
etio
n of
P
acke
t For
war
ding
Dut
y C
ycli
ng T
imer
Rol
e S
wit
chin
g T
imer
Cod
erT
imer
Pac
kets
fro
m
Nei
ghbo
r N
odes
Nat
ive
Pac
ket
For
war
ding
or
Cod
ed P
acke
t F
orw
ardi
ng
Fig. 3.3. Overall framework of the proposed protocol.
25
We will explain the details of the algorithms in Sections 3.3.2, 3.3.3,
and 3.3.4. The overall framework of the proposed protocol is shown in
Fig. 3.3. The framework consists of five parts: packet reception, packet
forwarding, role selection, power saving management, and timer man-
agement. In packet reception and forwarding, packets from neighbor
nodes are inserted into a received queue, and then packet forwarding
is conducted according to the roles of the nodes. Packets gathered in
the queue are encoded in coder nodes. In role selection, information
about the role of each node is reported to the algorithms performing
packet reception and forwarding and power saving management. The
roles of nodes are decided in the role selection algorithm, and they are
periodically switched. The power saving management algorithm reports
information about duty cycling to the packet reception and forwarding al-
gorithm. It also decides whether a node goes into the sleeping state based
on information about completion of packet reception and forwarding in
the node. As shown in Fig. 3.3, in the timer management algorithm, the
events of role switching, duty cycling, and packet encoding in the nodes
are managed by the role switching timer, the duty cycling timer, and the
coder timer, respectively.
3.3.2 Proposed Node Initiation and Duty Cycling
Scheme
In the proposed protocol, we modify the conventional PSM of the IEEE
802.11 system to enhance the energy efficiency of nodes in a bottleneck
26
ATIMwindow
Beacon interval
TXI1 TXI2 TXI3
Fig. 3.4. PSM with priority in the proposed protocol.
zone. In conventional PSM, nodes involved in packet transmission and
reception remain awake for the entire beacon interval. This procedure
consumes unnecessary energy when nodes remain in the active state for
the entire beacon interval even after completing packet transmission and
reception. In the proposed protocol, nodes enter the sleeping state after
completing packet transmission and reception explicitly announced in the
ATIM window. In this way, both coder and relay nodes in the bottleneck
zone spend more time in the sleeping state. Algorithms 3.1 and 3.2
show the pseudo codes for the proposed node initiation and duty cycling
scheme.
Initially, a sink node sends the B Indication message to nodes within
a 1-hop range, and then the nodes receiving the B Indication message
broadcast it to the nodes within a 2-hop range from the sink node. Con-
sequently, the nodes recognize the number of hop counts from a sink
node, thus defining the bottleneck zone, as shown in Algorithm 3.1. In
Algorithm 3.1, NB1 and NB2 are the group of nodes within 1 hop and 2
hop ranges from a sink node, respectively. The nodes in the bottleneck
zone are included in NB1. NCG and NRG are the groups of the coder and
relay nodes in the bottleneck zone, respectively. pC is the probability
27
Algorithm 3.1 Initiation algorithm for nodes in the bottleneck zone
Require: Node n receives B Indication message from a sink node or
nodes ∈ NB1.
Ensure: Node n is included in the NB1 or NB2 set. If node n ∈ NB1, it
is included in the NCG or NRG subset.
1: if B Indication message from a sink node then
2: Node n gets Parameter R from B Indication message, Node n
∈ NB1.
3: Node n randomly selects a subset of NCG with probability pC.
4: if subset = NCG then
5: Node n ∈ NCG.
6: else
7: Node n ∈ NRG.
8: end if
9: Broadcast B Indication message.
10: else
11: if Node n ∈ NB1 then
12: Discard the B Indication message.
13: else
14: Node n ∈ NB2.
15: end if
16: end if
28
Algorithm 3.2 Algorithm of preparation operation for power saving in
nodes
Require: : All nodes support the power saving mechanism using ATIM.
A sink node is always on the active state. Node n can enter the sleep-
ing state when Qtx is empty and packet receptions that are explicitly
announced in the ATIM window is completed.
Ensure: : Node n selects the interval where packets is transmitted.
1: if Node n ∈ NB1 then
2: When Node n has packets to transmit a sink node, it does not
send ATIM frame.
3: Node n selects TXI2 interval.
4: else if Node n ∈ NB2 then
5: Node n selects TXI1 interval.
6: else
7: Node n selects TXI3 interval.
8: end if
29
of selecting the coder group. Nodes in NB2 aid the power saving in the
nodes of NB1. In the B Indication message, the sending rate R of the
coded packets is included, and thus nodes receiving this message from
a sink node know information on R. To guarantee decoding at the sink
node, at least 50% of the nodes in the bottleneck zone must be a relay
node [26]. In this dissertation, a sink node selects 50% of the nodes in
the bottleneck as coder nodes, and the others as the relay nodes. In
Algorithm 3.2, nodes ∈ NB2 have the highest priority to access the chan-
nel. Therefore, when nodes ∈ NB1 have packets to send, they are sent in
TXI2 as shown in Fig. 3.4. When nodes ∈ NB1 have no packets to send,
they can enter the sleeping state after TXI1. Even if nodes ∈ NB1 have
packets to be transmitted, they can go into the sleeping state after TXI2
as shown in Fig. 3.4.
3.3.3 Proposed Packet Forwarding Scheme
In WSNs, multi-path forwarding provides better packet delivery relia-
bility than a routing-based system [60, 61]. In this environment, the
NC-based approach can not only reduce the network traffic but also pro-
vide the reliability of the packet delivery similar to multi-path forwarding
with duplicated packets. In the proposed protocol, to reduce traffic and
enhance reliability in the bottleneck zone, the coder nodes encode pack-
ets from multiple paths by using RLNC. Algorithms 3.3 and 3.4 show
the pseudo codes for the packet forwarding in the proposed protocol.
In Algorithm 3.3, Qrecv is a received queue and Qforw is a queue that
30
Algorithm 3.3 Algorithm for packet reception in nodes
Require: : Node receives a data packet pi from neighbor nodes.
Ensure: : In Node n, pi is inserted into Qrecv or discarded.
1: Node n gets Parameter IP from the packet header.
2: if Packet pi ∈ Qrecv or ∈ Qforw then
3: Discard a pi.
4: else
5: if IP = Coded packet then
6: Discard a pi.
7: else
8: Insert a pi into Qrecv.
9: end if
10: end if
31
Algorithm 3.4 Packet forwarding algorithm of nodes in the bottleneck
zone
Require: Node n receives data packets from neighbor nodes and re-
ceived packets are inserted into Qrecv. Beacon interval is started a
new.
Ensure: Node n inserts native or coded packets to Qtx.
1: if Node n ∈ NB1 then
2: if Node n ∈ NCG then
3: if Timer Tcoder expires then
4: Generate dkRe coded packets from all packets in Qrecv.
5: Set IP to “Coded packet”.
6: Insert all packets of Qrecv into Qforw.
7: Insert dkRe coded packets into Qtx.
8: end if
9: else
10: Pick a packet pi from Qrecv.
11: Insert pi into Qtx.
12: Insert pi into Qforw.
13: end if
14: end if
15: if Qrecv 6= empty then
16: goto step 1.
17: end if
32
stores the forwarded packets and restricts the transmission of duplicated
packets. IP is information about the packet type. Upon receiving a data
packet, the node inserts the packet into Qrecv after checking the packet
freshness. In addition, coded packets are discarded in nodes except a sink
node because the packets are not decoded in these node. In Algorithm
3.4, packets in the Qrecv of the relay nodes are forwarded to a sink node
when it has an opportunity to access the channel. The coder nodes
generate dkRe coded packets using RLNC to reduce the traffic. d∗e is
the ceiling function. The packet encoding process is conducted at an
interval of NC cycle TC. Thus, data packets are gathered in the coder
nodes during TC. TC is
αTB, α ≥ 1, (3.1)
where TB is the beacon interval and α is the NC cycle coefficient con-
stant. In Algorithm 3.4, Tcoder is the timer that checks whether time
of TC elapsed. k is the number of packets in Qrecv of the coder node.
Gathering of data packets at coder nodes can reduce power consump-
tion because coder nodes reduce traffic through packet encoding as well
as having opportunities to go into the sleeping state before the end of
TXI1. Furthermore, gathering data packets improves reliability. Using
RLNC, coder nodes can enhance reliability by encoding more packets into
a coded packet [64]. Packets in transmission queue Qtx are transmitted
to the sink node when it has a chance to access the channel.
33
Algorithm 3.5 Algorithm for role switching of nodes in the bottleneck
zone
Require: ATIM window in the current beacon interval ends.
Ensure: Node n keeps its role or changes its role.
1: if Node n ∈ NB1 then
2: if Timer Trole expires then
3: Node n randomly selects a subset of NCG with probability pC.
4: if subset = NCG then
5: Node n ∈ NCG.
6: else
7: Node n ∈ NRG.
8: end if
9: end if
10: end if
34
3.3.4 Proposed Role Switching Scheme
In the proposed protocol, if the role of each node in the bottleneck zone is
static, the relay nodes deplete their energy very quickly. A sink node can-
not decode coded packets without the native packet transmission of the
relay node. So that the energy is evenly consumed among the nodes in the
bottleneck zone, in the proposed protocol, the nodes’ roles are changed
periodically, which prolongs the lifetime of all nodes in the bottleneck
zone and, thus, the lifetime of the WSN. Role switching is conducted at
the interval of the role change cycle TR as shown in Algorithm 3.5. TR is
βTC, β ≥ 1, (3.2)
where β is the coefficient constant of the role change cycle. Nodes in the
bottleneck zone randomly select their role with probability pC. Trole is
the timer that checks whether time of TR elapsed.
3.4 Performance Analysis
In this section, we theoretically derive the packet delivery ratio of the
system for three approaches-without NC, with XOR NC, and with the
proposed protocol-in a simple topology of three sensor nodes and one
sink node, as shown in Fig. 3.5. A sensor node out of the bottleneck
zone senses the data periodically and then forwards the sensing data to
the sink node via two sensor nodes in the bottleneck zone. One of these
nodes is the relay node, and the other is the coder node. In the system
without NC, all nodes in the bottleneck zone are relay nodes. With XOR
35
Sink
CoderRelay
Sensor
Bottleneck Zone
Fig. 3.5. Reference network model for the simple topology.
NC, two packets are encoded into a coded packet at the coder node.
With the proposed protocol, k packets in the Qrecv of the coder node are
encoded into dkRe coded packets. Each coded packet has independent
information [65]. The packet delivery ratio of the system without NC
can be calculated as
PDRCONV = 1− Le2, (3.3)
where Le is the packet loss rate in the bottleneck zone. The packet
delivery ratio of XOR NC can be calculated as
PDRXOR = 1− Le + Le(1− Le)2. (3.4)
The packet delivery ratio of the proposed protocol is
PDRPROP = 1− Le +N−1∑i=k
N − 1
i
(1− Le)iLN−ie , (3.5)
where N = k + dkRe. Fig. 3.6 shows the packet delivery ratio when R
= 0.5. It is shown that the proposed protocol outperforms the protocol
36
without NC and the protocol with XOR NC with respect to the packet
delivery ratio for the certain packet loss rate. Furthermore, the packet
delivery ratio of the proposed protocol increases with increasing k as more
packets are encoded into a coded packet [64]. However, in a severely lossy
environment, the packet delivery ratio of the proposed protocol is sharply
reduced because a sink node cannot receive more than k packets in this
environment. Fig. 3.7 shows the packet delivery ratio of the proposed
protocol according to R. When R increases, the packet delivery ratio of
the proposed protocol is enhanced. However, the power saving realized
through NC is reduced because traffic increases with increasing R.
37
0 0.05 0.1 0.15 0.2 0.25 0.30.8
0.85
0.9
0.95
1
Packet loss rate in a sink
Pack
et d
eliv
ery
ratio
System w/o NCXORProp. (k = 4)Prop. (k = 8)Prop. (k = 16)
Fig. 3.6. Average packet delivery ratio in the bottleneck zone (R= 50%).
38
0 0.05 0.1 0.15 0.2 0.25 0.30.88
0.9
0.92
0.94
0.96
0.98
1
Packet loss rate in a sink
Pack
et d
eliv
ery
ratio
Prop. (R = 50%)Prop. (R = 60%)Prop. (R = 70%)Prop. (R = 80%)Prop. (R = 90%)Prop. (R = 100%)
Fig. 3.7. Average packet delivery ratio in the bottleneck zone according
to R (k= 16).
39
Table 3.1. Parameters in the performance evaluation.
Parameter Value
Data size 1024bytes
WLAN TX rate 65Mbps
TX range 100m
Hop limit of broadcasting 4 hop
TXI1 0.2TB
TXI2 0.1TB
pC 0.5
R 50%
3.5 Performance Evaluation
We compared the performance of the proposed protocol with that of
conventional systems for the parameters listed in Table 3.1 [26, 63]. We
implemented an event-driven simulator in the Riverbed modeler formerly
known as OPNET [66]. In the simulation, we used UDP traffic sources.
Each flow enters the network at a different time with a uniform distri-
bution U (0, ρ). ρ is the packet inter-arrival time. All flows have a CBR
with a fixed packet size. We use the power consumption model in [67].
This model is based on the measurements for a real IEEE 802.11n system
[67]. In our simulation, we do not consider the processing energy required
to perform NC because it is insignificant compared with the power con-
40
sumption involved in communication [68]. In [68], the encoder consumes
22.15µW at 0.4 V, achieving a processing throughput of 80 MB/s. In
the simulation, TGn channel model is used as the wireless channel model
[69]. In this dissertation, similar to the bits-per-joule capacity described
in [70], we define the energy efficiency in the bottleneck zone (bits/J) as
η =S
PB
, (3.6)
where S is the total generation rate of sensing data in the WSN (bits/s)
and PB is the total power consumption of the nodes in the bottleneck
zone (W).
3.5.1 Simple Topology
To analyze the basic performance of the proposed protocol, we used the
reference network model as shown in Fig. 3.5.
Figs. 3.8, 3.9, 3.10, and 3.11 show the power consumption, the energy
efficiency, the network lifetime and the packet delivery ratio according
to the packet inter-arrival time in the reference network model, respec-
tively. The baseline indicates the conventional system without NC. In
static role, the role of nodes is static. On the other hand, the role of
nodes is periodically switched in role switching. Figs. 3.8 and 3.9 show
that the proposed protocol not only reduces power consumption, but
also efficiently uses energy to deliver sensing data in the bottleneck zone,
compared with the baseline, thanks to the proposed duty cycling scheme.
More power saving is achieved by the coder nodes through packet encod-
ing in the proposed packet forwarding scheme. For a packet inter-arrival
41
time of 20 ms, the energy efficiency of the coder nodes is improved by
three times compared with that of the baseline, as shown in Fig. 3.9.
With role switching, the nodes consume less power than relay nodes and
more than coder nodes in the static role. Overall, the proposed proto-
col with role switching provides the longest network lifetime, as shown
in Fig. 3.10, because the energy of all nodes in the bottleneck zone is
evenly consumed. The network lifetime of the proposed protocol with
static roles is limited by the lifetime of the relay nodes because a sink
node cannot decode coded packets without the native packet transmis-
sion from the relay node. Because the relay nodes do not benefit from
NC in terms of power consumption, the relay nodes deplete their energy
more quickly than the coder nodes. As Fig. 3.11 shows, the packet de-
livery ratio is not reduced by the packet encoding in the proposed packet
forwarding scheme.
Fig. 3.12 shows the power consumption according to the NC cycle
constant α. The power saving achieved through packet encoding and
gathering in coder nodes increases with increasing NC cycle constant. In
addition, packet gathering with a small value of the NC cycle constant
in coder nodes can achieve substantial power saving.
42
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.090.2
0.4
0.6
0.8
1
Packet inter−arrival time (s)
Pow
er c
onsu
mpt
ion
(W)
BaselineProp. (Static role−Relay)Prop. (Static role−Coder)Prop. (Role switching)
Fig. 3.8. Power consumption in nodes (reference network model).
43
0.5 1 1.5 20
50
100
150
200
250
Packet inter−arrival time (s)
Ene
rgy
effic
ienc
y (K
bits
/J)
BaselineProp.Conv. NC scheme
Fig. 3.9. Energy efficiency in nodes (reference network model).
44
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.090
100
200
300
400
Packet inter−arrival time (s)
Net
wor
k lif
etim
e (s
)
BaselineProp. (Static role)Prop. (Role switching)
Fig. 3.10. Network lifetime (reference network model).
45
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.090.975
0.98
0.985
0.99
0.995
1
Packet inter−arrival time (s)
Pac
ket d
eliv
ery
ratio
BaselineProp. (Static role)Prop. (Role switching)
Fig. 3.11. Packet delivery ratio (reference network model).
46
2 3 4 5 6 7 80.2
0.25
0.3
0.35
NC cycle coefficient, α
Pow
er c
onsu
mpt
ion
(W)
Prop. (Static role−Relay)Prop. (Static role−Coder)Prop. (Role switching)
Fig. 3.12. Power consumption according to NC cycling constant (refer-
ence network model).
47
3.5.2 Grid Topology Networks
In this performance evaluation, we compare the performances in a large
7×7 grid topology network. In the network, the vertical and horizontal
distances between adjacent nodes are 50m. A sink node is at the center
of the network.
Figs. 3.13, 3.14, 3.15, 3.16, and 3.17 show average power consump-
tion, average energy efficiency, average traffic load, network lifetime, and
average packet delivery ratio in the bottleneck zone of the network, re-
spectively. In the conventional NC scheme, with the proposed duty cy-
cling scheme, XOR NC packet forwarding is used [26]. However, the pro-
posed role switching scheme is not used in the conventional NC scheme.
In Figs. 3.13 and 3.14, it is shown that the proposed protocol outperforms
the baseline and the conventional NC scheme in terms of the power saving
and the energy efficiency for the light traffic. Furthermore, the proposed
protocol reduces the traffic load in the bottleneck zone as compared with
the baseline due to the packet forwarding with NC as shown in Fig. 3.15.
However, in the heavy traffic, the conventional NC scheme consumes less
power than the proposed protocol. With the proposed protocol, coder
nodes have opportunities to go into the sleeping state before the end
of TXI1 due to the packet gathering during TC. In heavy traffic, they
have few sleeping opportunities because they continue to receive packets
from neighbor nodes until almost the end of TXI1. Furthermore, the
traffic load in the bottleneck zone is higher with the proposed protocol
than with the conventional NC scheme, as shown in Fig. 3.15. With the
48
proposed protocol, the traffic load in the bottleneck zone is higher than
with the conventional NC scheme. Because packet gathering in the coder
nodes reduces network congestion, the relay nodes in the bottleneck zone
can exchange more data packets with each other. However, the proposed
protocol outperforms the baseline and conventional NC schemes in terms
of network lifetime regardless of the traffic load, thanks to the proposed
role switching scheme as shown in Fig. 3.16. Furthermore, the packet
delivery ratio in the proposed protocol is not reduced in the heavy traffic
because, in the proposed protocol, more packets are encoded into a coded
packet using RLNC as shown in Fig. 3.17 [64]. In the conventional NC
scheme, the packet delivery ratio can be reduced in lossy environment
because it only encodes two packets by XOR NC.
Fig. 3.18 shows the overhead of RLNC processing in terms of power
consumption. It is shown that this overhead is insignificant for the life-
time of coder nodes if RLNC encoding is conducted by the optimal de-
signed hardware[68]. However, it can increase the cost of sensor devices.
49
0.5 1 1.5 20
1
2
3
4
5
6
7
Packet inter−arrival time (s)
Pow
er c
onsu
mpt
ion
(W)
BaselineProp.Conv. NC scheme
Fig. 3.13. Average power consumption of nodes in the bottleneck zone
(grid topology network).
50
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.090
0.5
1
1.5
Packet inter−arrival time (s)
Ene
rgy
effic
ienc
y (G
bits
/J)
BaselineProp. (Static role−Relay)Prop. (Static role−Coder)Prop. (Role switching)
Fig. 3.14. Average energy efficiency of nodes in the bottleneck zone (grid
topology network).
51
0.5 1 1.5 20
1000
2000
3000
4000
5000
Packet inter−arrival time (s)
Tra
ffic
load
(K
bps)
BaselineProp.Conv. NC scheme
Fig. 3.15. Average traffic load of nodes in the bottleneck zone (grid
topology network).
52
0.5 1 1.5 2100
150
200
250
300
350
400
Packet inter−arrival time (s)
Net
wor
k lif
etim
e (s
)
BaselineProp.Conv. NC scheme
Fig. 3.16. Average network lifetime (grid topology network).
53
0.5 1 1.5 20.9
0.92
0.94
0.96
0.98
1
Packet inter−arrival time (s)
Pac
ket d
eliv
ery
ratio
BaselineProp.Conv. NC scheme
Fig. 3.17. Average packet delivery ratio in the bottleneck zone (Grid
topology network).
54
0.5 1 1.5 20
0.005
0.01
0.015
0.02
Packet inter−arrival time (s)
Pow
er c
onsu
mpt
ion
(µW)
Fig. 3.18. Average power consumption by RLNC encoding in the coder
node (Grid topology network).
55
3.5.3 Ratio of Coder Nodes in Networks
In this performance evaluation, we analyze the performance of the pro-
posed protocol in a large grid topology network according to the ratio of
coder nodes, pC. In the simulation, the grid topology network has the
same environment as in Section 3.5.2 and the packet inter-arrival time of
the sensor nodes is 0.4 seconds.
Figs. 3.19 and 3.20 show the average lifetime of the nodes and the
average packet delivery ratio in the bottleneck zone. In Fig. 3.19, it is
shown that the lifetime of nodes increases with increasing pC because the
number of coder nodes is increased in the bottleneck zone. However, the
packet delivery ratio is rapidly reduced when pC is above 60% as shown
in Fig. 3.20. About 50% of the nodes in the bottleneck zone should be
relay nodes as explained in Section 3.3.2 [26].
56
0.2 0.4 0.6 0.8 1100
150
200
250
300
Probability pC
Lif
etim
e of
nod
es (
s)
BaselineProp.
Fig. 3.19. Average lifetime of nodes in the bottleneck zone according to
the ratio of coder nodes (grid topology network).
57
0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
Probability pC
Pack
et d
eliv
ery
ratio
BaselineProp.
Fig. 3.20. Average packet delivery ratio in the bottleneck zone according
to the ratio of coder nodes (grid topology network).
58
3.6 Summary
In this chapter, we considered a power saving mechanism for the bot-
tleneck zones of WSNs to enhance the lifetime of WSNs. In particular,
to improve the energy efficiency without reducing the reliability of the
packet delivery in the bottleneck zone, we proposed a duty cycling scheme
and a packet forwarding scheme with RLNC for the nodes in the bottle-
neck zone. In addition, we proposed a role switching scheme to evenly
prolong the lifetime of all nodes in the bottleneck zone. The results of
performance evaluation for simple and grid topology networks show that
the proposed protocol reduces the power consumption and the traffic
load, compared with the conventional system. Furthermore, the results
show that the proposed protocol outperforms the conventional system in
terms of the lifetime of WSNs, without reducing the reliability of packet
delivery in the bottleneck zone.
59
60
4
Multi-way Relay System with
NC in MBSNs
4.1 Motivation
Because of its large coverage areas without geographical limits, GEO
satellite communication is useful for military purposes and regions that
do not have available terrestrial infrastructure. Furthermore, multicas-
ting data via GEO satellite communication can reliably deliver data
because the GEO satellite can broadcast data to large coverage areas
without traversing many congested routers [71]. In the near future, it
is expected that IP multicasting will be efficiently provided by satellite
communication because an IRIS that is an IP router in the satellite has
been commercialized [72].
In multiparty video conferencing, the video traffic of each user needs
61
to be multicasted to the other users. Therefore, satellite communication
is useful to delivery multicasting data for video conferencing because of its
large coverage areas and efficient multicasting capability. In addition, a
real-time service such as video conferencing may be sufficiently supported
in the satellite network because the propagation delay in the satellite net-
work with an IRIS can be minimized from 500 to 250 ms, as traffic no
longer traverses the terrestrial gateway for the routing [72]. However, re-
search on the minimization of resource usage in satellite communication
is needed before multiparty video conferencing via the satellite network
can be provided because the unit cost of satellite radio resources is ex-
pensive [73].
NC is a potential technique to reduce bandwidth in the satellite com-
munications [73, 74]. However, to the best of our knowledge, a MRS with
NC in MBSNs has not been explored before. An MRS is a system where
N users (N ≥ 2) exchange data via a relay and each user sends its data
to all other users [29]. A spot beam is a satellite signal concentrated
in power for covering a limited geographic area to increase transmission
capacity and abide by the difference in regulation among countries. A
typical spot beam has a coverage radius of approximately 200 – 500 km
[75, 76].
Therefore, we consider an MRS with NC in MBSNs in this chapter.
In particular, we focus on an NC system for MRS in MBSNs to minimize
the multicasting data for multiparty video conferencing. We also propose
a reliable NC scheme to provide reliable multicasting data transmission
for mobile users. For reliable data transfer, additional coded packets are
62
transmitted on the resources saved by the proposed NC system. Fur-
thermore, we propose a resource allocation scheme for multiparty video
conferencing with NC in satellite communications to minimize the delay
in the link layer. The main contributions of the chapter are as follows:
� A proposed algorithm to calculate the number of coded packets
using the IP routing table for multicasting and information on the
video frame
� A proposed reliable NC scheme for mobile users using the decoding
error rate
� A proposed resource allocation scheme for multiparty video confer-
encing with NC
� A theoretical analysis of the achievable rate and the NC gain for
the MRS with NC in multi-spot beam satellite networks
63
Fig. 4.1. System model of the proposed NC system in MBSNs.
4.2 Proposed MRS with NC
4.2.1 System Model
The system model is composed of a satellite with an IRIS, RCSTs, and
user nodes as shown in Fig. 4.1. The satellite, operating as an MRS
relay, has multi-spot beams. The IRIS can generate coded packets of
multicasting data received from the RCSTs. In the proposed NC system,
we use RLNC to generate coded packets [8]. The RCST can decode
multicasting data from both its own packets and the coded packets. A
user node is connected to an RCST and generates the video conference
multicasting data. We make the following assumptions in the design of
our proposed NC system.
64
� For each user connected to the RCST, the video conference multi-
casting data are generated from a unit of video frames. In general,
a video frame is generated by a video compression codec such as
MPEG-4 or H.264. The compressed video stream consists of suc-
cessive GOPs. A GOP is composed of I, P, and B frames [77].
We also assume that the generation of video frames for all users
is almost synchronized. All GOPs for video streaming generated
by users have an identical structure. Thus, all packets of frames
generated by users arrive at each RCST simultaneously and the
proposed NC system encodes packets of frames of the same type.
� In the fixed user node, packet loss does not occur in the satellite link
because of power control and ACM. However, packet loss can occur
in the satellite link for the mobile user node because of temporary
link blockage caused by shadowing [5]. Thus, the proposed protocol
provides a reliability mode to enhance the reliability of the data
transmission for mobile users in Section 4.2.3.
� In the link layer of satellite communication, resources are suffi-
cient allocated for all RCSTs; thus, the queuing delay is negligible.
However, in dynamic resource allocation, delays can occur [78–80].
Therefore, we consider dynamic resource allocation for the pro-
posed NC system in Section 4.2.4.
65
Fig. 4.2. System architecture of the proposed NC system.
4.2.2 Operation of the Proposed NC System
The system architecture of the proposed NC system is shown in Fig. 4.2.
The proposed NC system consists of a multiparty detector, a manager
of mobile users and a selector of the number of coded packets, and the
generator of coded packets. The proposed NC system is implemented
in the NC module of the IRIS. The basic operation of the proposed NC
system considers fixed user nodes. Thus, there is no packet loss. The
operation of the proposed NC system is as follows. First, when packets
are received from the RCSTs, the NC module checks whether there are
packets of frames for multiparty video conferencing using the multicasting
routing table. The NC module then calculates the required number of
coded packets in spot beam k, NC,k, to be decoded in all RCSTs, written
66
as
NC,k =N∑u=1
npk, u − nmin, k, (4.1)
where npk,u is the number of packets for a video frame transmitted from
user u and packets are generated from user node u for 1 ≤ u ≤ N .
N denotes the number of users participating in the multiparty video
conferencing, and nmin, k is
minu∈ k{npk, u} . (4.2)
In RLNC, all the RCSTs should receive a number of coded packets that
are the same size as an RLNC batch NB to decode the coded packets in
all the RCSTs. In the proposed NC system, NB is
N∑u=1
npk, u . (4.3)
Coded packets as well as the packets of each RCST are used for decoding.
Therefore, NC,k packets are needed to decode the packets of all the RCSTs
of spot beam k. Finally, the NC module generatesNC,k coded packets and
appends the coefficient set in the packet header to generate the decoding
matrix in the RCSTs [7]. The decoding matrix is described in detail in
Section 4.2.5.
4.2.3 Reliability Mode of the Proposed NC System
Even though there are power control and ACM in satellite communica-
tion, temporary link blockages caused by shadowing can occur for mobile
67
users [5]. To provide reliable data transmission, additive transmission of
coded packets is needed in a multicasting system with NC [81]. There-
fore, our NC system provides an enhanced reliability mode of data trans-
mission for mobile users. In the proposed NC system, additional coded
packets are transmitted to the spot beam where there are mobile users.
The resources saved by the proposed NC systems are used to transmit
additional coded packets. The detailed procedure is as follows.
Once the user node realizes that it is mobile, it notifies the manager
of mobile users in the satellite as shown in Fig. 4.2. The NC system then
calculates the NC,k, as shown in Algorithm 4.1. The NC system initially
Algorithm 4.1 Algorithm to calculate NC,k
1: if mobile user in spot k then
2: Set NC,k ← NB.
3: Calculate pD,k.
4: while pD,k > ϕ do
5: Set NC,k ← NC,k + 1.
6: Calculate pD,k.
7: end while
8: else
9: NC,k =N∑u=1
npk, u − nmin, k.
10: end if
changes NC,k to NB for the mobile user. The mobile user periodically
reports the packet loss rate to the manager of mobile users in the satellite.
If NC,k cannot cover the packet loss rate reported by the mobile user, the
68
satellite increases NC,k until the following condition is satisfied.
pD,k < ϕ, (4.4)
where pD,k is the decoding error rate for the RLNC batch and ϕ is the
threshold of the required pD,k. A description of pD,k is detailed in Section
4.4.1. In satellite communication, the channel condition cannot be ex-
actly estimated because of the long propagation delay. Thus, NC,k does
not fall below NB even if the packet loss rate reported by the mobile
user is lower than expected value. Consequently, the mobile user can
receive the multicasting data of other users without loss in the enhanced
reliability mode. At a certain rate of packet loss, the loss can be covered
by the transmission of NB coded packets alone. However, in a severely
lossy environment, more resources would be consumed by the proposed
NC system than a conventional system.
4.2.4 Resource Allocation for the Proposed NC Sys-
tem
We propose a resource allocation scheme for multiparty video conferenc-
ing with NC to minimize delay in the link layer. We consider that all
RCSTs are provided with a PEP [82, 83]. In our system, the PEP can
analyze application information. A DAMA agent of the RCST requests
a resource from a DAMA controller in a NCC using application informa-
tion from the PEP, as shown in Fig 4.3 [78–80]. The detailed procedure
is as follows.
69
Fig. 4.3. System model of resource allocation request.
Fig. 4.4. Resource allocation example.
� In the session initiation of a multiparty video conference, the PEP
of the RCST intercepts the control packets and extracts the appli-
cation information such as frames per second and the maximum
frame size [82, 83]. The PEP sends this application information to
the DAMA agent. Upon receiving the application information, the
DAMA agent requests the resource from the DAMA controller in
70
the NCC. Upon receiving the requests from the RCSTs participat-
ing in the multiparty video conference, the NCC allocates the peri-
odic and synchronized resource needed to convey the maximum-size
frame in the SFs for these RCSTs. For example, when the frame
generation interval is 40 ms, the size of the SFs is 160 ms, the size
of a time slot is 10 ms, one time slot can convey a frame of maxi-
mum size, and two users are participating in the multiparty video
conference, the NCC allocates the resource as shown in Fig. 4.4.
� During the session closure, the PEP of the RCST also intercepts the
control packets. If the PEP detects session closure, it commands
the DAMA agent to release the resource. The DAMA agent sends
the request of the resource release to the DAMA controller and it
then releases the resource.
In the forward link of the satellite, data is transmitted through time
division multiplexing. Thus, the transmission of video conferencing data
is given high priority to minimize queueing delay in the satellite.
4.2.5 Coefficient Matrix
In the RCST connected to user node u in spot beam k, the decoding
process is X1
...
XN
= D
Y1:Iu−1
Xu
YIu:NC,k
, (4.5)
71
D =
A1:Iu−1, 1:Iu−1 A1:Iu−1, Iu:Iu+npk,i−1 A1:Iu−1, Iu+npk,i:NB
0 I 0
AIu:NC,k, 1:Iu−1 AIu:NC,k, Iu:Iu+npk,i−1 AIu:NC,k, Iu+npk,i:NB
−1
,
(4.6)
where Xu =[x1 . . . xnpk,u
]Tand Yq:r =
[yq . . . yr
]Tfor Yq:r ⊂ Y,
Y =[y1 . . . yNC,k
]T, xα is the α-th packet of the frame generated by
user u, yβ is the β-th coded packet generated by the NC module in the
satellite for spot beam k, and D is the decoding matrix. For user node
u, Iu is
u−1∑l=1
npk, l + 1. (4.7)
A is the coefficient matrix for coded packets transmitted to spot beam
k. An:m,g:h is a sub-matrix of A, and αi,j is the coefficient value for
1 ≤ i ≤ NC,k, 1 ≤ j ≤ NB.
A =
α1,1 . . . α1,NB
.... . .
...
αNC,k,1 . . . αNC,k,NB
, (4.8)
An:m,g:h =
αn,g . . . αn,h
.... . .
...
αm,g . . . αm,h
. (4.9)
72
Mul
ti-w
ayR
elay
Cha
nnel
T 11
T 1K
T 11
T 1K
Rel
ay
...
...
rXrY
Spo
t-be
am 1
Spo
t-be
am S
1111
1111
111
1ˆ
ˆˆ
ˆ,..
.,,..
.,,..
.,K
SSK
WW
WW
11W 1KW
11
11
111
1ˆ
ˆˆ
ˆ,..
.,,..
.,,..
.,K
KK
KK
SSK
WW
WW
1SW
SKW
1
11
111
11
ˆˆ
ˆˆ
,...,
,...,
,...,
SS
SS
KS
SKW
WW
W
111
1ˆ
ˆˆ
ˆ,..
.,,..
.,,..
.,SK
SKSK
SKK
SSK
WW
WW
Fig. 4.5. MRC model with S spot-beams, each of which is composed of
K distinct terminals.
73
4.3 Theoretical Analysis for MRS with NC
In this section, we theoretically analyze the achievable rate and NC gain
for the MRS with NC in MBSNs. We denote the sets 1, ..., K and 1, ..., S
by IK and IS for positive integers K and S, respectively. There are S ≥ 1
spot-beams in the satellite network, where each spot-beam has K ≥ 1
users. Users in spot-beam j, j ∈ IS , are denoted by Tj1, ..., TjK as shown
in Fig. 4.5. Let Wji ∈ Wji be the message of user Tji, and Tji wants
to decode the messages (W11, ...W1K , ...,WS1, ...WSK) for j ∈ IS, i ∈ IK
that are the messages of all the users in the satellite network. We denote
the set of users in spot-beam j by Tj, and the set of all users by T. The
Gaussian MRC is modeled [29] as
Yr [t] =L∑j=1
K∑i=1
Xji [t] + Zr [t] , (4.10)
Yji [t] = Xr [t] + Zji [t] , j ∈ IS, i ∈ IK , (4.11)
where Xji[t] and Yji[t] are the input and the output of user Tji at time
t, respectively; Xr[t] and Yr[t] are the input and the output of the relay,
respectively; Zr is the zero-mean Gaussian noise term at the relay with
variance Nr; and Zji is the Gaussian noise at the user Tji with variance
Nji [29]. The constraints of the transmission power at the relay and user
Tji are
1
nE
[n∑t=1
|Xr [t]|2]≤ Pr, (4.12)
74
1
nE
[n∑t=1
|Xji [t]|2]≤ Pji. (4.13)
A(2nR11 , ..., 2nR1K , ..., 2nRS1 , ..., 2nRSK , n
)code for the MRC is com-
posed of SK sets of integers Wji = {1, 2, ..., 2nRji} for j ∈ IS, i ∈ IK .
Encoding functions at the users are
Xnji = fji (Wji) . (4.14)
The set of encoding functions {f tr}nt=1 at the relay are
Xr [t] = f tr(Y t−1
r
), 1 ≤ t ≤ n. (4.15)
Decoding functions at the users are
gji(Wji, Y
nji
)=(W ji
11, ..., Wji1K , ..., W
jiS1, ..., W
jiSK
). (4.16)
The average probability of decoding error is defined as
P ne = Pr
⋃j∈IL,i∈IS
{gji(Wji, Y
nji
)6= (W11, ...,W1K , ...,WS1, ...,WSK)
}.
(4.17)
A rate tuple (R11, ..., R1K , ..., RS1, ...RSK) is said to be achievable for an
MRC with S spot beams of K users if there exists a sequence (2nR11 , ...,
2nR1K , ..., 2nRS1 , ..., 2nRSK , n) codes such that P ne → 0 as n → ∞. When
Rji = R for j, j ∈ IS and i, i ∈ IK , the symmetric rate is defined as [29]
CS,Ksym
∆= sup {R : (R, ..., R) is achievable} . (4.18)
The symmetric rate is useful in systems where all users have the same
amount of information to send, or in fair systems where every user is to
75
be given the same guaranteed uplink bandwidth, i.e., each user can send
data up to a certain rate [29]. To focus on the fundamental behavior of
the proposed protocol, we consider the symmetric network where Pji = P
and Nr = Nji = 1 for all j ∈ IS, i ∈ IK .
Because DF relaying is used in the satellite in the proposed proto-
col, we only derive the symmetric rate of DF relaying. DF consists of
two phases: data transmission from users to relay and the packet broad-
casting from the relay to users. In the second phase, we consider the
frequency division transmission among the spot beams [84]. We assume
that the constraints of the transmission power for all spot beams is Pr.
The relay broadcasts (W11, ...W1K , ...,WS1, ...WSK) to each spot beam
using the transmission scheme with NC [8]. Therefore, the symmetric
rate of DF relaying is derived as
RDF = min
{log2 (1 + SKP )
SK,log2 (1 + Pr)
SK − 1
}. (4.19)
It is indicated that the user can achieve a greater rate than when using
MRS without NC if the relay power is bottlenecked, i.e, Pr ≤ (1 + SKP )1− 1SK
− 1.
When rates of the MRS with and without NC are identical, the num-
ber of resources required to transmit data is less in the MRS with NC
than MRS without NC [6, 85]. Therefore, we can achieve an NC gain
defined as
G =SK (S + 1)
SK (S + 1)− S. (4.20)
In a MRS with NC in a satellite with multi-spot beams, there are
two factors that affect the NC gain: number of users and number of spot
76
beams. For satellite communication without using overheard packets in
the return link, the NC gain decreases with an increasing number of users,
because only one transmission may be reduced. On the other hand, the
NC gain can be enhanced when the number of spot beams increases. The
NC gain is increased in this case because the data are multicasted in all
spot beams when users are located in different spot beams, as shown
in Fig. 4.1. Four users located in spot beams 1 and 2 transmit four
packets to the satellite. The satellite then generates three coded packets
and broadcasts them to spot beams 1 and 2. Therefore, the number
of transmissions is reduced from 12 to 10. Thus, the NC gain becomes
12/10 = 1.2. When there are four users in single spot beam, the NC gain
is 8/7 = 1.14.
4.4 Performance Evaluation
In the performance evaluation, we evaluated the proposed NC system for
MRS in MBSNs in terms of the achievable symmetric rate, NC gain, de-
coding error rate, and PSNR. We implemented an event-driven simulator
in MATLAB. In the simulation, we used the real video sources encoded
by the H.264 codec listed in Table 4.1 [86]. To evaluate a more realistic
NC gain, we considered two encoding types: CBR and VBR encoding.
Each generated video frame in users is segmented by IP packets of 1500
bytes and is sent to RCST. To analyze the maximum NC gain, we used
the CBR encoding in the simulation and a video source for evaluation.
To analyze the reduction of the NC gain due to asymmetric video traffic
77
Table 4.1. Specifications for the video sources in the simulation.
Video source name Bit rate (kbps)
Akiyo 254.90
Bowing 251.13
Foreman 591.27
Hall 442.13
News 394.88
Mother-daughter 288.22
Pamphlet 304.65
Silent 558.58
generated by the users, we used the VBR encoding in the simulation.
For VBR encoding, we make each user randomly select a video source
from Table 4.1. In the simulation, there is no packet loss for fixed users.
In the other hand, for mobile users, the packet loss occurs by two-state
Markov channel model [87].
4.4.1 Interested Performance Metrics
Achievable symmetric rate : Achievable symmetric rate means the
rate achievable by a user when the rates of all users are equivalent as
mention above in Section 4.3.
78
NC Gain : The NC gain G without mobile users is
G =TM
TCM ∗ (1 + ε), (4.21)
where TM and TCM are the total number of transmissions required to
deliver the multicasting data in a satellite network with and without the
proposed NC system, respectively. The total number of transmissions
is the sum of the number of unicast transmissions from all the RCSTs
to the satellite and the number of multicasting transmissions from the
satellite to all the RCSTs in each spot beam. Therefore, TM and TCM
can be expressed as
TM =N∑u=1
npk, u +S∑k=1
{N∑u=1
npk, u − δ (k)nmin, k
}, (4.22)
TCM =N∑u=1
npk, u +S∑k=1
{N∑u=1
npk, u − nmin, k
}, (4.23)
where S is the number of spot beams and δ (k) is an indicator function
that specifies if the number users in spot beam k is one. When there is
a user in spot beam k, the satellite only multicasts the video data of the
other users excluding the user in spot beam k. Therefore, there is no NC
gain in a spot beam with a single user. ε denotes the RLNC overhead
such as the coefficient set. ε is derived as
ε =o
L, (4.24)
where L and o denote the packet size (in bits) and overhead appended to
a packet (in bits), respectively. In this dissertation, we use the minimized
overhead value as ε [88].
79
Decoding Error Rate for Mobile Users: For a mobile user, the
decoding error rate for the RLNC batch is
pD,k = maxu∈Mk
pD (k, u) , (4.25)
pD (k, u) = 1−NC,k∑
i=NB−(1−pu)npk,u
NC,k
i
(1− pu)ipNC,k−iu , (4.26)
where Mk is the set of mobile users in spot beam k and pu is the packet
loss rate in the link between a satellite and mobile user u. If the decoding
for the batch of RLNC fails, frames within the RLNC batch are also lost.
Thus, pD,k also indicates the average frame loss rate in spot beam k.
NC gain for mobile users: With mobile users, packet loss oc-
curs and TCM increases because of the additional transmission of coded
packets. Therefore, TM and TCM for mobile users are defined as
TM =N∑u=1
(1− pu)npk,u +S∑k=1
{N∑u=1
(1− pu)npk,u − δ (k)nmin, k
},
(4.27)
TCM =N∑u=1
(1− pu)npk,u
+S∑k=1
{γ (k)NRC,k + (1− γ (k))
(N∑u=1
(1− pu)npk,u − nmin,k
)},
(4.28)
where NRC,k is NC,k to cover the packet loss in spot beam k and γ (k) is
an indicator function that specifies if there are mobile users in spot beam
k.
80
PSNR: The PSNR between original and erroneous images is the most
widespread method to evaluate video quality. It uses the well-known
SNR that compares signal energy to error energy. PSNR compares the
maximum possible signal energy to the noise energy that has shown to
result in a higher correlation with subjective quality perception than
conventional SNR [89].
4.4.2 Simulation Results
Initially, we evaluated the achievable symmetric rate for MRS with NC
in multi-spot beam satellite networks. Fig. 4.6 shows the achievable
symmetric rate in MRS with and without NC when the relay power is
bottlenecked and Pr is 0 dB. It is clear that the simulation results are
slightly worse than the theoretical results because of RLNC overheads
such as the coefficient set. The results also indicate that MRS with NC
outperforms MRS without NC in terms of the achievable rate for a small
number of users.
For an environment without mobile users, we evaluated the perfor-
mance of the proposed NC system in terms of the NC gain considering
various factors such as the RLNC overhead, encoding types, number of
spot beams and deployment of nodes. Fig. 4.7 shows the NC gain in
a single-spot beam with and without RLNC overhead according to the
number of users. The NC gain decreases with increasing the number of
users as mentioned in Section 4.3. In addition, because of RLNC over-
head such as the coefficient set, the NC gain is less than the theoretical
81
NC gain. Because the NC gain is determined by nmin, k, the NC gain
in the VBR environment is less than the theoretical NC gain. Fig. 4.8
shows the NC gain for multi-spot beams when there are eight users. In
this scenario, we used two methods of user deployment: even and random
deployment. To analyze the maximum NC gain according to the num-
ber of spot beams, we used even deployment in the simulation where at
least two users are deployed in as many spot beams as possible. NC gain
can be achieved when the number of users in the spot beam is greater
than two. Therefore, the NC gain increases with an increasing number
of spot beams until N/2 spot beams although it is less than the theo-
retical NC gain (N = 8, S = 4). However, when S > N/2, the NC
gain decreases because the number of spot beams where only one user
is deployed increases. On the other hand, the NC gain can decrease de-
spite an increasing number of spots when random deployment is used.
With random user deployment, only one user can be deployed in the spot
beam. Therefore, the NC gain can decrease in a more realistic situation
such as random deployment.
Fig. 4.9 shows the overhead of RLNC processing in terms of power
consumption. In Fig. 4.9, we use the power consumption mode of RLNC
encoder in [68]. It is shown that this overhead is insignificant in the
satellite with increasing the number of users and spot beams.
82
2 4 6 8 10 12 140
0.2
0.4
0.6
0.8
1
Number of users
Rat
e
Theo. analy. (MRS with NC)Simul. (MRS with NC)MRS w/o NC
Fig. 4.6. Achievable symmetric rate.
83
2 3 4 5 6 7 81
1.05
1.1
1.15
1.2
1.25
1.3
1.35
Number of users
NC
gai
n
Theo. analy.Simul. (CBR)Simul. (VBR)
Fig. 4.7. NC gain in a single-spot beam.
84
1 2 3 4 5 6 7 80.96
0.98
1
1.02
1.04
1.06
1.08
1.1
1.12
Number of spots
NC
gai
n
Theo. analy.(N=8, S=4)Evenly deployed with CBREvenly deployed with VBRRandomly deployed with CBRRandomly deployed with VBR
Fig. 4.8. NC gain in multi-spot beams (N = 8).
85
2 3 4 5 6 7 80
0.1
0.2
0.3
0.4
0.5
Number of users
Pow
er c
onsu
mpt
ion
(µW
)
S = 1S = 2S = 3S = 4
Fig. 4.9. Average power consumption by RLNC encoding in the satellite
86
For the reliability mode in an environment with mobile users, we
evaluated the performance of the proposed NC system in terms of frame
loss rate, NC gain and PSNR. In the simulation, we consider a single-spot
beam, four users, and ϕ of 0.1%. Fig. 4.10 shows the average frame loss
rate in the multicasting system with and without the proposed NC system
according to the packet loss rate for mobile users. In the multicasting
system without the proposed NC system, the loss rate of I frames is
the largest among the I, P, and B frames because of its size. I frame
are encoded and decoded independently of any other frame. On the
other hand, P frames depend on the previous I or P frame. B frames
also depend on the previous I or P frame as well as following an I or
P frame. The size of the I frame is the largest [77] and followed by
the P frame. However, in the multicasting system with the proposed
NC system, the frame loss is less than 0.1% because of the transmission
of additive coded packets. Fig. 4.11 shows the average PSNR in the
multicasting system with and without the proposed NC system. In the
multicasting system without the proposed NC system, the average PSNR
is reduced with increased the packet loss rates. However, the average
PSNR is maintained in the multicasting system with the proposed NC
system. Figs 4.12 and 4.13 show the visual quality of video streaming in
the multicasting system with and without the proposed NC system when
the packet loss rate is 4%. In the conventional multicasting system, the
visual quality of video streaming service is severely deteriorated. On the
other hand, in the multicasting system with the proposed NC system,
there is no distortion of the video stream. Fig. 4.14 shows the NC gain
87
with a mobile user. It is shown that the NC gain is almost one. This
indicates that a certain rate of packet loss can be compensated for by
the transmission of NB coded packets. Consequently, it is shown that
the proposed NC system can cover the certain amount of packet loss by
using resource saved by NC.
88
10−5
10−4
10−3
10−2
0
0.1
0.2
0.3
0.4
0.5
Packet loss rate
Ave
. fra
me
loss
rat
e
I frame (w/o NC)P frame (w/o NC)B frame (w/o NC)I frame (with NC)P frame (with NC)B frame (with NC)
Fig. 4.10. Average frame loss rate for a mobile user (S = 1, N = 4).
89
10−5
10−4
10−3
10−2
24
26
28
30
32
34
36
38
40
42
Packet loss rate
Ave
. PS
NR
w/o NCwith NC
Fig. 4.11. Average PSNR for a mobile user.
90
Fig. 4.12. Visual quality of video streaming service (Conv.).
91
Fig. 4.13. Visual quality of video streaming service (Prop.).
92
10−5
10−4
10−3
10−2
0
0.2
0.4
0.6
0.8
1
1.2
Packet loss rate
NC
gai
n
w/o mobile userwith mobile user
Fig. 4.14. NC gain with a mobile user (S = 1, N = 4).
93
4.5 Summary
In this chapter, we considered an NC system for MRS in MBSNs. In par-
ticular, we proposed an NC system for video conference multicasting data
in a satellite network that uses satellite radio resources. Furthermore, we
proposed not only an NC system to reliably transmit multicasting data
for mobile users but also a resource allocation scheme for multiparty
video conferencing with NC to minimize the delay of the satellite link
layer. We also evaluated the performance of the proposed NC proto-
col. Simulation results indicate that the achievable rate can be increased
by the proposed NC system. In addition, the proposed NC system can
achieve NC gain because of a reduced number of transmissions by NC
in the satellite network. As a result, multicasting with the proposed NC
system outperformed conventional multicasting in terms of resource effi-
ciency. In particular, NC gain can be achieved in an environment where
the satellite has multi-spot beams. However, NC gain in a VBR envi-
ronment can decrease because of the different number of packets for the
frames of each user. For a spot beam with only one user, there is no NC
gain in the spot beam. Furthermore, the large size of the coefficient set
can reduce the NC gain. Therefore, we should consider these overheads
when applying the proposed NC system. For mobile users, it is shown
that the average frame loss rate is reduced below the required frame loss
rate by additional transmission of coded packets using resources saved by
proposed protocol. It results in an enhanced PSNR of the video stream
and good visual quality without performance degradation.
94
5
File Transfer Framework with
AL-FEC Aided by Navigation
Systems in SOTM Systems
5.1 Motivation
Thanks to SOTM systems, many commercial and military applications
are made available to mobile platforms such as airborne vehicles, trains,
ships, etc. through a satellite. In an SOTM system, the antenna of
the SOTM terminal, which is equipped with an active control system
and an inertial navigation system, is pointed toward the satellite [90].
Thus, the wireless link between the SOTM terminal and the satellite
can be continuously maintained, resulting in providing a high data rate
for mobile platforms. However, the link between the SOTM terminal
95
and the satellite can experience a temporary outage owing to channel
blockage due to pointing errors of the antenna on account of the rugged
terrain and obstructions such as tunnels, high buildings and trees. This
can cause packet loss in the satellite link [5, 91–94].
One of the solutions to this problem that can ensure reliable commu-
nication in SOTM environments is an AL-FEC. In wireless communica-
tions, channel coding is important because it ensures the reliability of
data transmission protecting it from data corruption by noise and inter-
ference. However, in mobile networks, channel blockage due to intermit-
tent shadowing can cause packet loss even though channel coding is ap-
plied. In this environment, lost packets should be retransmitted by TCP
with a retransmission delay. As a result, the network throughput may be
reduced. To solve this problem, many studies have investigated AL-FEC
in various communication systems [10, 12, 13, 95]. AL-FEC covers the
packet loss not recovered by channel coding because it is applied above
layer 2 and uses the fountain code known as the rateless erasure code
without the retransmission delay, resulting in achieving a high through-
put. Recently, Raptor code has been commercially used in the AL-FEC
systems because of its dynamic packet loss protection, exceptionally high
computational efficiency, and low transmission and reception overheads
[11, 12]. Its advantages allow a software implementation and also pro-
vide end-to-end error correction without requiring any change in legacy
standards, resulting in ease of deployment in the network [12]. For these
reasons, AL-FEC can be applied to satellite communications [96, 97].
In SOTM environments, channel blockage is unpredictable. Thus, the
96
AL-FEC system should use the on-the-fly manner to achieve the fully
reliable file transfer [12]. In the on-the-fly manner, the sender continu-
ously transmits the encoded packets until the receiver decodes a file from
encoded packets. However, it requires more resource consumption to
transfer a file completely. Therefore, a study on enhancing the resource
efficiency in AL-FEC systems in SOTM environments is needed. The
navigation system is an essential component in the intelligent transport
systems [98]. In the navigation system, a GPS, a GIS-based road map,
and a map-matching algorithm are used to determine the vehicle position
on the road [99]. To enhance the accuracy of the positioning informa-
tion, many researches on enhancing GPS accuracy have been addressed
[99–101]. If the navigation system cooperates with the communication
systems of SOTM environments, channel blockages caused by tunnels,
overpasses, and underpasses on the route can be predicted [102–105].
In this chapter, we focus on a fully reliable file transfer framework
with AL-FEC. Therefore, we propose an ACK exchange protocol in the
proposed framework to make the end-to-end data transfer reliable. To
enhance the resource efficiency, we also propose a transmission control
scheme aided by navigation systems. In particular, we use the naviga-
tion system to predict channel blockages. During the predicted channel
blockage, the packet transmission of the AL-FEC is paused, resulting in
reduction in usage of the satellite resource. The main contributions of
the chapter are as follows:
� An ACK exchange protocol for a fully reliable file transfer frame-
work with AL-FEC in SOTM networks.
97
� A transmission control scheme aided by navigation systems in the
proposed framework.
� A theoretical analysis model to justify the effectiveness of the pro-
posed framework.
5.2 Proposed File Transfer Framework
5.2.1 System Model
The system model consists of a ground station, a SOTM node, a satellite,
and file servers as shown in Fig. 5.1. The SOTM node is connected to the
ground station by satellite links. Because SOTM nodes have mobility,
the satellite link can be intermittently disconnected owing to the rugged
terrain and obstructions such as high buildings and trees. The ground
station is connected to file servers by high-speed wired links, and a PEP
with the splitting connection is implemented in it [82, 83]. The ground
station that received files from file servers transmits data files to the
SOTM nodes. SOTM nodes can transmit ACK messages through the
return link. A file is segmented into k native packets. The native packet
is the original data segmented from a file. k is⌈LF
LS
⌉, (5.1)
where LF and LS are the size of a file and a native packet, respectively.
Repair packets are generated from the k native packets by RaptorQ code.
RaptorQ code is the advanced version of Raptor code [12]. They are
98
App
lica
tion
(F
TP
)
UD
P
IP
MA
C
PH
Y
FE
C F
ram
ewor
kFE
C
Gen
erat
ion
Con
stan
tT
X R
ate
UD
P
IP
MA
C
PH
Y
TC
P
IP
MA
C
PH
Y
FE
C F
ram
ewor
kF
EC
G
ener
atio
nC
onst
ant
TX
Rat
e
TC
P
IP
MA
C
PH
Y
App
lica
tion
(F
TP
)
PE
P
SO
TM
Lin
kH
igh-
spee
d W
ired
Lin
k
SO
TM
N
ode
Sat
elli
teG
roun
dS
tati
onF
ile
Ser
vers
Fig. 5.1. System model of the proposed reliable file transfer framework.
99
transmitted to recover packets lost owing to channel blockage. The size
of the repair packet is LS. We make the following assumptions regarding
our proposed framework:
� A two-state Markov chain model is used as the channel model [91–
93] because the antenna of the SOTM terminal is pointed toward
the satellite and high-efficiency parabolic antennas or phased-array
antennas [5]. The channel model has channel open (o) and blockage
(b) states as shown in Fig. 5.2.
� In the link layer of satellite communications, the satellite resource
is enough to transmit a file and ACK messages. Thus, the queuing
delay in the link layer is negligible.
� If the SOTM node receives k + 2 packets generated from a file,
it can decode the file without an error. In RaptorQ code [12], the
decoding probability is 99.9999% for all k values when k + 2 packets
are received. The maximum size of k is 56,403.
� The processing time for AL-FEC is negligible [12].
� Clocks in the SOTM node and the ground station are synchronized.
5.2.2 ACK Exchange Procedure
In this paper, we consider applications such as urgent message trans-
mission and file transfer services. Thus, the fully reliable transmission
100
o b
poo pbb
pob
pbo
Fig. 5.2. Channel model of SOTM.
is needed. The proposed framework is implemented in the FEC frame-
work, which lies between the application layer and the transport layer.
All packets generated from the FEC framework are transmitted to the
UDP layer at a constant transmission rate, RTX. The detailed procedures
of the proposed framework in the sender and receiver are as follows:
� Sender: Upon receiving the data of a file from the file server, the
FEC framework of the sender segments it into native packets by
the size of LS. The FEC framework then inserts native packets into
both transmission and encoding queues. Packets in the transmis-
sion queue are transmitted to UDP layer by constant TX rate of
RTX. Next, when the FEC framework receives a file completely, it
generates repair packets from the k native packets in the encoding
queue with RaptorQ code. After the repair packet generation, all
the packets are inserted into the transmission queue. When receiv-
ing the receiver-ACK (R-ACK) message that indicates the com-
pletion of the file reception from the receiver, the FEC framework
removes all packets from the transmission queue and terminates the
file transfer to the receiver. The sender then sends the sender-ACK
101
(S-ACK) that indicates the reception of R-ACK to the receiver. If
the sender receives duplicated R-ACK, the sender retransmits the
S-ACK.
� Receiver: Upon receiving packets from the sender, the FEC frame-
work of the receiver inserts them into the reception queue. The
FEC framework checks whether the packets in the reception queue
can be decoded to a file by Raptor code. If the decoding is com-
pleted, the receiver sends an R-ACK message to the sender (ACK-
based mechanism) and the file is forwarded to the application layer,
immediately. If the receiver does not receive the S-ACK message
within the RTT, it retransmits the R-ACK message to the sender.
In the proposed framework, the two-way ACK exchange is used to en-
sure the reliability of the data transfer because the S-ACK and R-ACK
messages could be lost due to channel blockage in the satellite communi-
cation environment. At the receiver, there is no extra delay causing by
the ACK exchange because a file is forwarded to the application layer
immediately just after the completed decoding. However, the proposed
ACK exchange can incur the additional usage of satellite resource at the
sender because the sender continuously transmits the packets to receiver
until receiving the R-ACK. The detailed analysis of this overhead is dis-
cussed in Section and 5.3.2 and 5.4.
Even though the AL-FEC system is easily applied to networks by
software updates, the high cost can be incurred to deploy it in most of
file servers. To reduce the cost, the PEP can be used in the ground
102
station for the proposed framework [82, 83]. Because PEPs use the
splitting connection[82, 83], the proposed framework is only deployed in
SOTM nodes and the ground station as shown in Fig. 5.1. Furthermore,
by PEPs with the splitting connection, the proposed framework can be
compliant with standard applications. For example, the legacy protocol
such as TCP is used between the ground station and file servers. The
proposed framework is applied between the ground station and SOTM
nodes. In the proposed framework with PEPs, the file download should
be completed between the ground station and the server to generate re-
pair packets. However, the additional delay by the splitting connection is
not incurred because the capacity of the wired link between the ground
station and the file server is much greater than that of the satellite link.
Therefore, the file download from the server to the ground station can
be finished while transmitting native packets from the ground station to
the SOTM node.
5.2.3 Transmission Control Aided by Navigation Sys-
tems
Unlike other communication systems, in the SOTM system, it is hard to
estimate the channel blockage in real-time because the real-time channel
state cannot be applied to the SOTM system owing to the long time
propagation delay. Furthermore, the channel blockage occurs during
long time by obstacles because the SOTM terminal should maintains
the line of sight between its antenna and a satellite for the communica-
103
tion. Therefore, unnecessary resource is consumed in the proposed file
transfer framework with AL-FEC. To enhance the resource efficiency,
the data transmission should be paused for the duration of the channel
blockage in the proposed framework. There are two types of channel
blockages: unpredicted and predicted channel blockages. Channel block-
ages caused by the rugged terrain are hard to predict. However, channel
blockages caused by obstacles on the route can be predicted by a naviga-
tion system since it uses the GIS-based road map. In this dissertation,
we define two predicted channel blockages: SPCBs and VPCBs. SPCBs
are channel blockages caused by a tunnel, an underpass, etc. on the road.
In SPCBs, the channel blockage duration is static because there are no
traffic signal and variation of the vehicular traffic load. It is easily pre-
dicted by using map information. VPCBs are channel blockages caused
by tall building, tree, etc. on the roadside. In VPCBs, the channel
blockage duration can be variable due to traffic signal and variation of
the vehicular traffic load. To predict this channel blockages, predefined
location-based information is needed. This information is generated by
empirical measurement of channel blockage on the road [91–93]. For ex-
ample, similar to OpenSignal [106], data for channel blockages is collected
by many users of the proposed framework in real-time. Channel status
information with GIS-based road map is then uploaded to servers of the
proposed framework. Finally, servers share channel status information
to all users by periodic updates. This blockage information can be mea-
sured by users who exploit the SOTM system in real-time. Thus, the
impact on seasonal variations and new constructions can be reflected in
104
information on the channel blockage. The locations such as mountainous
roads and a financial district where there are tall, closely spaced high-rise
office buildings can be defined as the district with VPCBs. Therefore, we
use a navigation system in the proposed framework to predict channel
blockage.
The system architecture for the proposed framework aided by a navi-
gation system is shown in Fig. 5.3. When the navigation system detects a
change in the PCS on the road, it sends the predicted information on the
type of blockages, the current position, the velocity of the vehicle, and the
distance from the obstacle that caused the channel blockage to the CBP
of the SOTM node. If the predicted channel blockage is SPCB, informa-
tion on the channel blockage duration is also sent to CBP. For SPCBs,
the blockage duration can be almost exactly predicted thanks to well es-
timated journey time through blockage by the navigation system based
on the average vehicle velocity, GIS-based road map, GPS information,
map-matching algorithms, and vehicular traffic information [107–109].
However, unexpected situation such as car accident and vehicular traffic
jam in tunnels can cause inaccurate estimation of this journey time. In
this case, the CBP recognizes the channel blockage as VPCB based on
real-time vehicular traffic information. Upon receiving the information
on the predicted channel blockage, the CBP of the SOTM node sends
a pause message to the ground station to pause the packet transmission
during the channel blockage as shown in Fig. 5.4. When it receives the
pause message, the ground station pauses the packet transmission af-
ter the time remaining for the beginning of the channel blockage, TBS.
105
When the predicted channel blockage is SPCB, the ground station pauses
the packet transmission during the predicted channel blockage duration,
TCB. On the other hand, if the predicted channel blockage is VPCB, the
ground station pauses the packet transmission until it receives the restart
message. TCB and TBS can be calculated as
TCB =DPB + 2DeGPS
v, (5.2)
TBS =DNB
v− (Tcurr − TBD)− DeGPS
v− TP, (5.3)
where DPB is the length of the obstacle that causes the channel blockage
in meter; DeGPSis the average positioning error in the GPS system in
meter; DNB is the distance between the SOTM node and the obstacle in
meter when the CBP of the SOTM node sends the pause message; Tcurr is
the time when the ground station receives the pause message; TBD is the
time when the CBP of the SOTM node sends the pause message; and v is
the predicted average velocity of the SOTM nodes; TP is the propagation
delay of the satellite link. In the proposed framework, the prediction
would be nearly correct because the pause message is sent within few
seconds before SOTM node meets the channel blockages. However, we
should consider the positioning error in the GPS system [100]. Thus, in
the calculation of TCB and TBS, we consider the guard time for the GPS
positioning error to enhance the accuracy of the prediction for channel
blockages. The detailed pseudocodes for the transmission control of the
proposed framework in the SOTM node and the ground station are shown
in Algorithms 5.1 and 5.2.
106
Algorithm 5.1 Algorithm for the transmission control in the SOTM
node
Require: CBP receives information on the position and PCS.
Ensure: CBP sends a message of the transmission control.
1: if Node in the channel open then
2: if PCS = SPCBs or VPCBs then
3: CBP sends a pause message with the blockage information to
the ground station.
4: end if
5: else
6: if PCS = channel open then
7: CBP sends a restart message to the ground station after
DeGPS/v.
8: end if
9: end if
107
Algorithm 5.2 Algorithm for the transmission control in the ground
station
Require: Ground station receives the message from CBP of the SOTM
node.
Ensure: Ground station pauses or resumes the transmission.
1: if Pause message then
2: if PCS = SPCBs then
3: Ground station pauses the transmission during the predicted
channel blockage duration.
4: end if
5: if PCS = VPCBs then
6: Ground station pauses the transmission until receiving the
restart message.
7: end if
8: else
9: Ground station resumes the transmission.
10: end if
108
In VPCBs, it is hard to predict the end of channel blockage because
of traffic signal and variation of the vehicular traffic load. Therefore, the
CBP of the SOTM node sends the restart message to the ground station
for resuming the transmission when the PCS is channel open.
109
AL
-FE
C S
yste
m
GP
S S
yste
m
Nav
igat
ion
Sys
tem
AL
-FE
C S
yste
m
Cha
nnel
Blo
ckag
eP
redi
ctor
Con
trol
ler
for
pkt.
TX
/RX
Con
trol
ler
for
pkt.
TX
/RX
Cha
nnel
Blo
ckag
e In
fo.
SO
TM
Nod
e
Gro
und
Sta
tion
Fig. 5.3. System architecture with navigation systems.
110
Dis
tanc
e (m
)
Tim
e (s
ec.)
Tra
nsm
issi
on p
ause
Gro
und
stat
ion
rece
ives
the
paus
e m
essa
ge.PBD
v
GP
SeD
GP
SeD
NB
D
CB
TB
ST
PT
SOT
M n
ode
send
s th
e pa
use
mes
sage
Cha
nnel
blo
ckag
e
Fig. 5.4. Example of proposed framework with navigation systems
(SPCB).
111
Predicted Unpredicted
Transmission pause
Channel Blockage (DB)
GPS positioning Error ( )
Channel Open(DO)
Channel Open(DO)
Channel Blockage (DB)
GPS positioning Error ( )GPSeD
GPSeD
Fig. 5.5. Example of the overhead for the proposed protocol.
5.2.4 Benefit and Overhead of Proposed Protocol
In the proposed framework with the navigation system, the resource effi-
ciency of the proposed framework can be enhanced because the transmis-
sion of packets is paused during the channel blockage. For example, in
an environment where there are two channel blockages and one of them
is predicted as shown in in Fig. 5.5, the resource efficiencies of the AL-
FEC with and without the navigation system are 2DO/ (2DO +DB) and
2DO/ (2DO + 2DB), respectively. DO and DB are the average distances
of the channel open and the channel blockage in meter, respectively.
When DO=30, DB=20, the resource efficiencies of the AL-FEC with and
without the navigation system are 0.75 and 0.6, respectively. It is shown
that the transmission pause by the navigation system can enhance the
resource efficiency.
However, the transmission of packets in the proposed framework is
paused during the guard time for the GPS positioning error, in addi-
tion to the duration of the channel blockage. Therefore, the goodput
and the resource efficiency of the proposed framework can be reduced
as compared with the idle goodput and the resource efficiency. The idle
112
goodput and the resource efficiency can be achieved in the case of no the
GPS positioning error. the performance without the GPS positioning
error. In the environment where DO ≤ DeGPS, the proposed framework
is hardly applied. The overhead of the guard time has a significant effect
on the reduction of the goodput and the resource efficiency when the
difference between DO and DeGPSis small. However, when DO � DeGPS
,
the goodput and the resource efficiency of the proposed framework are
almost the same as the idle value. For example, in Fig. 5.5, the effect of
the difference between DO and DeGPSon the goodput and the resource
efficiency is as follows. When the difference is small (DO=6, DB=4, and
DeGPS=2), the goodput and the resource efficiency are 0.4RTX and 0.6667,
respectively. In Fig. 5.5, the goodput and the resource efficiency are
(2DO − 2DeGPS) / (2 (DO +DB))×RTX and (2DO − 2DeGPS
)/(2DO +DB
− 2DeGPS). The calculation of the goodput and the resource efficiency is
explained in Section 5.3 in detail. When the difference is large (DO=30,
DB=20 and DeGPS=2), the goodput and resource efficiency are 0.56RTX
and 0.7368, respectively. The idle goodput and the resource efficiency
are 0.6RTX and 0.75, respectively.
In the case of VPCBs, an additional delay of RTT is needed to resume
the transmission because the ground station should receive the restart
message from the SOTM node. Therefore, the goodput can be reduced
as compared with the idle goodput. The effect of additional delay for
VPCBs on the goodput is explained in Section 5.4 in detail.
113
5.3 Theoretical Analysis
In this section, we theoretically derive the average file transfer time, the
goodput and the resource efficiency for transmitting a file by the Markov
chain model for the proposed framework. A theoretical analysis of the
proposed framework aided by navigation systems is also performed. In
this analysis, we assume that the statistical channel model is configured
by given conditions such as the average distances of the channel open
and the channel blockage, the velocity of the SOTM node, and slot time
[87, 92, 93, 110]. We consider the ratio of predicted channel blockage in
whole channel blockages, the ratio of VPCB in predicted channel block-
ages and the GPS positioning error. The ratio of the predicted channel
blockage has an effect on the enhancement of the resource efficiency. The
GPS positioning error that is the overhead in the proposed framework
increases the file transfer time because the data transmission duration
in the channel open interval is reduced by the guard time for the GPS
positioning error. As a results, the goodput is reduced. The ratio of
VPCB also has an effect on the goodput.
5.3.1 Transfer Time and Goodput
To calculate the average file transfer time, TFT, we model the bidimen-
sional process {s(t), n(t)} with the discrete-time Markov chain depicted
in Fig. 5.6. In this Markov chain, s (t) ∈ {b, o} is the channel state,
n (t) ∈ {0, 1, ..., N} is the number of packets received successfully, and t
is the time measured in slots. N is the required number of packets re-
114
o, 0poo
pbb
pob pbo
b, 0
o, 1
b, 1
o, N-1
b, N-1
o, N
pbb pbb
pob pbo
poo
pob pbo
poo …
…
…
pbo
poo
1
Fig. 5.6. Markov chain model for average file transfer time of the pro-
posed framework.
ceived at the receiver for decoding a file. Thus, N is k + 2 as mentioned
in Section 5.2.1. A slot is equal to the time TS used to transmit one
packet of size LS. TS is LS
RTX. Because the channel model is a two state
Markov chain as shown in Fig. 5.2, the state transition probabilities are
P {o, i|b, i} = pob, i ∈ (0, N − 1)
P {o, i|o, i− 1} = poo, i ∈ (0, N − 1)
P {b, i|o, i− 1} = pbo, i ∈ (0, N − 1)
P {b, i|b, i} = pbb, i ∈ (0, N − 1)
P {o, N |o, N} = 1,
(5.4)
where pxy is the state transition probability from the channel state x ∈
{b, o} to the channel state y ∈ {b, o} in the channel model. Similar to
115
[87], these probabilities can be calculated as
pob = TSTO,
poo = 1− pob,
pbo = TSTB,
pbb = 1− pbo,
(5.5)
where TO and TB are the average time of the channel open and blockage
that can be calculated as TO = DO
v,
TB = DB
v,
(5.6)
In state {o, N} of our model, file transfer is completed. Therefore, similar
to [87], TFT can be calculated from the expected number of times that
the process is in state {o, N} if it is started in state {s(t), n(t)}. The
expected number of times ψ (s (t) , n (t)) from state {s(t), n(t)} to state
{o, N} can be calculated asψ (o, N) = 0,
ψ (o, i) = 1 + pobψ (b, i) + pooψ (b, i+ 1) ,
ψ (o, i) = 1 + pboψ (o, i+ 1) + pbbψ (b, i) ,
(5.7)
where i ∈ (0, N − 1) [110]. From (5.5) and (5.7), the closed form can be
derived as ψ (o, i) =(
1 + pob1−pbb
)(N − i) ,
ψ (b, i) = 11−pbb
{1 + (N − i− 1) (pob + pbo)} .(5.8)
The possible initial states are {o, 0} and {b, 0} in Fig. 5.6. Consequently,
TFT can be calculated as
TFT = (πoψ (o, 0) + πbψ (b, 0))× TS + TP, (5.9)
116
where πo and πb are the steady state probabilities in the channel model.
They can be calculated from the transition probabilities of the channel
model [87].
The average goodput (G) can be defined as
G =LF
TFT
. (5.10)
5.3.2 Resource Efficiency
To derive the resource efficiency of the proposed framework, we consider
the additional resource consumption caused by the ACK-based scheme.
Because the sender can terminate the file transfer upon receiving the
ACK message, the resource is basically wasted in 2TP. Furthermore, if
the R-ACK message is lost, the resource is additionally wasted during
retransmission of ACK messages. The resource usage time due to re-
transmission of ACK messages is 2TP(1 − πb)(πb + 2πb2 + 3πb
3 + ...).
This is approximately 2TPπb
(1−πb). On the other hand, if the S-ACK mes-
sage is lost, the additional resource is only needed to retransmit ACK
messages. However, it is insignificant because the size of ACK messages
is very small. Thus, we do not consider the loss of the S-ACK message
in the resource efficiency. The average resource efficiency is defined as
η =LF
TRURTX
, (5.11)
where TRU is the total resource usage time for complete file transfer with
the proposed framework. For the average file transfer time, the resource
usage time is TFT. Thus, taking into consideration the resource usage
117
time for the average file transfer time and the wasted resource by ACK
exchange, TRU is
TFT + TP
(1 +
2πb
1− πb
)+ ε, (5.12)
where ε is the processing time of the ACK message. However, this time
may be negligible.
5.3.3 Performance Analysis with Navigation Sys-
tems
The predicted channel blockage by the navigation system reduces the
resource usage time to pause the data transmission. However, it can in-
crease the file transfer time. Thus, to calculate the file transfer time with
navigation systems, TFT NAVI, the predicted channel blockage, the ratio of
VPCB and the GPS positioning error should be considered. Therefore,
we derive TFT NAVI based on TFT derived from the Markov chain model
in Fig. 5.6 as shown in Algorithm 5.3.
ω is the average number of channel blockages during the file trans-
fer time TFT except the propagation delay TP and σ is the number of
packets not received in the SOTM node because of the guard time for
the GPS positioning error and the delay due to the transmission of the
restart message. In the case of VPCBs, 2TP is needed to resume the data
reception in the SOTM node by the restart message. ωp and ωc are the
previous and current ω values. γ is the error bound. ω and σ can be
118
Algorithm 5.3 Algorithm to calculate TFT NAVI
1: Calculate TFT, ω, σ.
2: Set N ← k + 2 + σ.
3: Set ωp ← 0.
4: Set ωc ← ω.
5: while |ωc − ωp| > γ do
6: Set ωp ← ωc.
7: Calculate TFT, ω, σ.
8: Set N ← k + 2 + σ.
9: Set ωc ← ω.
10: end while
11: Set TFT NAVI ← TFT.
calculated as
ω =πb (TFT − TP)
TB
, (5.13)
σ =αω ×
(2DeGPS
v+ 2TPβ
)TS
, (5.14)
where α and β are the ratio of predicted channel blockage in whole
channel blockages and the ratio of VPCB in predicted channel block-
ages for 0 ≤ α ≤ 1 and 0 ≤ β ≤ 1, respectively. In the calculation
of σ,2DeGPS
v+ 2TPβ is overhead time per a predicted channel blockage
for pausing data transmission. During the predicted channel blockage
duration and the guard time for the GPS positioning error, the ground
station pauses the data transmission. In the case of VPCBs, the delivery
119
time TP of the restart message is included in the pause duration of the
data transmission. This reduces the resource usage time. Therefore, the
resource usage time with navigation systems is derived as
TRU NAVI = (TFT NAVI − TP) (1− ρ) + 2TP
(1 +
πb
1− πb
)+ ε, (5.15)
ρ = απb
(TB +
2DeGPS
v+ βTP
TB
). (5.16)
In the calculation of ρ, TB +2DeGPS
v+ βTP is the reduced time in the
resource usage time per a predicted channel blockage. Consequently,
the goodput and the resource efficiency with navigation systems can be
calculated as
GNAVI =LF
TFT NAVI
, (5.17)
ηNAVI =LF
TRU NAVIRTX
. (5.18)
5.4 Performance Evaluation
In the performance evaluation, we compare the performance of the pro-
posed framework with that of UDP, TCP Reno, and PEPsal for the pa-
rameters listed in Table 5.1 [111]. PEPsal is conventional PEP solution
in the satellite communication [82]. PEPsal uses the PEP with the TCP
splitting connection and the optimized TCP is applied in the satellite
link [82]. We have implemented an event-driven simulator in MATLAB.
In the proposed framework, LS is varied according to LF because of the
120
Table 5.1. Parameters in the performance analysis.
Parameter Value
Blockage ratio 5%–40%
DB 25–150 m
v 100 km/h
TP 0.25 s
LF 0.5–900 Mbits
Coding scheme(Prop. framework) RaptorQ code
Coding rate(Prop. framework) rateless
LS (Prop. framework) 500–5000 bytes
RTX (Prop. framework) 1 Mbps
Maximum window size (TCP) 65,535 bytes
Segment size (TCP) 1500 bytes
Retransmission timeout (TCP) 1.5 s
size limitation of k. To show results in various environments, we set
up simulation environments by varying each factor such as the channel
blockage distance, the blockage ratio, GPS positioning error, file size, α
(the ratio of predicted channel blockage in whole channel blockages), and
β (the ratio of VPCB in predicted channel blockages) in Table 5.2. The
blockage ratio is
DB
DO +DB
. (5.19)
In SOTM nodes to decode a file, the processing delay exists in the
decoding of Raptor code. However, it is negligible value. In MBMS of
3GPP, the processing delay is from 20 to 200ms. If the optimal decoding
121
algorithm is used, the processing delay is only from 2 to 20ms in the
environment with Intel(R) Xeon(R) CPU @1.60GHz [112].
Initially, we evaluate the reliability of the proposed framework. Fig.
5.7 shows the packet delivery ratio in the application layer for UDP with
and without the proposed framework. It is shown that the proposed
framework offers the fully reliable file transfer in the application layer
thanks to the transmission of repair packets. We also evaluate the ba-
sic performance of the proposed framework as compared with TCP and
PEPsal. Figs. 5.8 and 5.9 shows the file transfer time and goodput when
blockage rate is 33% in the city environment. This channel blockage
statistics are based on the measurements of the field test in Boston, USA
[93]. It is shown that the file transfer time of the proposed framework
is less than that of TCP and PEPsal because the retransmission process
is eliminated in the proposed framework. In the retransmission of TCP
in satellite communications, a long time is consumed by the timeout
and congestion-avoidance scheme owing to the long propagation delay.
Fig. 5.10 shows the resource efficiency in the city environment. For
the transfer of a small file using the proposed framework, the resource
efficiency is greatly reduced because of the wasted resource during the
ACK transmission, and the continuous data transmission regardless of
whether the channel is open or blocked. On the other hand, in the trans-
fer of a large file, the resource efficiency of the proposed framework is
enhanced as compared with that of the transfer of a small file. Because
the wasted resource during the ACK transmission in the proposed frame-
work is static, the overhead due to the wasted resource during the ACK
122
transmission is reduced with increasing file size. In the case of PEPsal,
the resource efficiency can be lower than that of the proposed mechanism
because the optimized TCP is used. The optimized TCP increases the
TCP window size aggressively to enhance the TCP throughput in the
satellite communication with the long propagation delay [82].
123
Table
5.2.
Scen
ariofor
simu
lation.
Scen
.1Scen
.2Scen
.3Scen
.4Scen
.5
Blo
ckage50
m50
m25–150
m50
m25–150
mdistan
ce
Blo
ckage20%
20%20%
,40%
20%5%
–40%ratio
GP
S
10m
0–20m
10m
10m
10m
position
ing
error
File
size0.5–900
Mbits
0.5–900M
bits
0.5–900M
bits
0.5–900M
bits
900M
bits
α0–0.6
0.60.6
0.60–0.6
β0.5
0.50.5
0.1–0.90.5
124
5 10 15 20 25 30 35 400
0.2
0.4
0.6
0.8
1
1.2
Blockage ratio (%)
Pac
ket d
eliv
ery
ratio
UDP w/o AL−FECUDP with AL−FEC
Fig. 5.7. Packet delivery ratio in UDP with and without prop. frame-
work.
125
106
107
108
109
0
2000
4000
6000
8000
File size
Tim
e (s
)
TCP RenoPEPsalProp. (Analysis)Prop. (Simulation)
Fig. 5.8. Average file transfer time in the city environment.
126
106
107
108
109
0
0.2
0.4
0.6
0.8
1
File size
Goo
dput
(M
bps)
TCP RenoPEPsalProp. (Analysis)Prop. (Simulation)
Fig. 5.9. Average goodput in the city environment.
127
106
107
108
109
0
0.2
0.4
0.6
0.8
1
File size
Res
ourc
e ef
fici
ency
, η
TCP RenoPEPsalProp. (Analysis)Prop. (Simulation)
Fig. 5.10. Average resource efficiency in the city environment.
128
Scenario 1 : In this scenario, we demonstrate the high level of ac-
curacy of the derived closed-form equations in Section 5.3 and evaluate
the performance of the proposed framework according to the α and file
size. Figs. 5.11, 5.12 and 5.13 show the average file transfer time, the
average goodput and the average resource efficiency, respectively in the
environment of Scenario 1. It is observed that the theoretical results
(shown by the solid lines) closely match the simulation results (shown
by the markers). In Fig. 5.11, the file transfer time of the proposed
framework is less than that of TCP because the retransmission process
is eliminated in the proposed framework. In the retransmission of TCP
in satellite communications, a long time is consumed by the timeout and
congestion-avoidance scheme owing to the long propagation delay. The
file transfer time increases with increasing α because the overhead of the
guard time for the GPS positioning error increases with increasing α. In
Fig. 5.12, it is shown that the proposed framework outperforms TCP in
terms of the goodput because of its lower file transfer time. The goodput
is improved by about 60%. The goodput decreases with increasing α ow-
ing to the overhead of the guard time. In Fig. 5.13, it is shown that the
resource efficiency is lower in the proposed framework than in TCP. The
resource efficiency increases with increasing α because the transmission
of packets in the proposed framework is paused for the duration of the
predicted channel blockage. When α = 0 in results, the navigation sys-
tem is not used in the proposed framework. Thus, the file transfer time
for α = 0 is the highest value of all. However, the resource efficiency for
α = 0 is the lowest value of all.
129
106
107
108
109
0
500
1000
1500
2000
File size
Tim
e (s
)
TCPProp. (Analy., α = 0.0)
Prop. (Simul., α = 0.0)Prop. (Analy., α = 0.2)
Prop. (Simul., α = 0.2)
Prop. (Analy., α = 0.4)Prop. (Simul., α = 0.4)
Prop. (Analy., α = 0.6)
Prop. (Simul., α = 0.6)
Fig. 5.11. Average file transfer time in the SOTM environment (Scenario
1).
130
106
107
108
109
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
File size
Goo
dput
(M
bps)
TCPProp. (Analy., α = 0.0)
Prop. (Simul., α = 0.0)
Prop. (Analy., α = 0.2)Prop. (Simul., α = 0.2)
Prop. (Analy., α = 0.4)
Prop. (Simul., α = 0.4)
Prop. (Analy., α = 0.6)Prop. (Simul., α = 0.6)
Fig. 5.12. Average goodput in the SOTM environment (Scenario 1).
131
106
107
108
109
0
0.2
0.4
0.6
0.8
1
File size
Res
ourc
e ef
fici
ency
, η
TCPProp. (Analy., α = 0.0)Prop. (Simul., α = 0.0)Prop. (Analy., α = 0.2)Prop. (Simul., α = 0.2)Prop. (Analy., α = 0.4)Prop. (Simul., α = 0.4)Prop. (Analy., α = 0.6)Prop. (Simul., α = 0.6)
Fig. 5.13. Average resource efficiency in the SOTM environment (Sce-
nario 1).
132
Scenario 2 : In this scenario, we show the performance of the pro-
posed framework according to the GPS positioning error. Fig. 5.14 shows
the average goodput in the environment of Scenario 2. It is shown that
the goodput decreases as the GPS positioning error increases because of
its guard time in the proposed framework. Fig. 5.15 shows the average
resource efficiency in the environment of Scenario 2. It is observed that
the resource efficiency with the GPS positioning error is almost equal to
the idle value without the GPS positioning error because of the environ-
ment DO � DeGPSof this scenario.
Scenario 3 : In this scenario, we evaluate the performance of the
proposed framework according to the channel blockage distance. Figs.
5.16 and 5.17 show the average goodput and resource efficiency in the
environment of Scenario 3 with a blockage ratio of 20%. This environ-
ment is characterized by DO � DeGPS. In Fig. 5.16, it is shown that the
goodput is reduced with a short blockage distance. In spite of the same
blockage ratio, the number of channel blockages in a file transfer increases
in the environment of the short blockage distance as compared with the
environment with a long blockage distance. However, in the transfer of
a small file, the goodput with a short blockage distance is larger than
that with a long blockage distance. Because of the short file transfer
time of the small file, the long blockage distance has significant effect on
the increase in the average file transfer time. In Fig. 5.17, it is observed
that the resource efficiencies for various blockage distances are almost
same for the transfer of a large file. Table 5.3 shows the goodput and
the resource efficiency in the environment of Scenario 3 with a blockage
133
Table 5.3. Goodput and resource efficiency(Scenario 3, Blockage rate =
40%, α = 60%, LF = 900Mbits).
DO DB
Goodput (Mbps) Resource Efficiency
TCP Prop. TCP Prop.
37.5m 25m 0.0510 0.3412 0.5373 0.6378
75m 50m 0.1136 0.4703 0.6856 0.7262
112.5m 75m 0.1691 0.5133 0.7473 0.7491
150m 100m 0.2129 0.5347 0.7762 0.7595
ratio of 40% and the file size of 900 Mbits. The goodput and the resource
efficiency are reduced as the difference between DO and DeGPSdecreases
as mentioned in Section 5.2.4. However, the performance of TCP is also
degraded under this condition because of the frequent channel blockage
[113].
134
106
107
108
109
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
File size
Goo
dput
(M
bps)
TCPProp. (Positioning error = 0m)Prop. (Positioning error = 5m)Prop. (Positioning error = 10m)Prop. (Positioning error = 15m)Prop. (Positioning error = 20m)
Fig. 5.14. Average goodput in the SOTM environment (Scenario 2).
135
106
107
108
109
0
0.2
0.4
0.6
0.8
1
File size
Res
ourc
e ef
fici
ency
, η
TCPProp. (Positioning error = 0m)Prop. (Positioning error = 5m)Prop. (Positioning error = 10m)Prop. (Positioning error = 15m)Prop. (Positioning error = 20m)
Fig. 5.15. Average resource efficiency in the SOTM environment (Sce-
nario 2).
136
106
107
108
109
0
0.2
0.4
0.6
0.8
1
File size
Goo
dput
(M
bit/s
ec.)
Blockage distance = 50mBlockage distance = 75mBlockage distance = 100mBlockage distance = 125mBlockage distance = 150m
Fig. 5.16. Average goodput in the SOTM environment (Scenario 3,
blockage rate = 20%).
137
106
107
108
109
0
0.2
0.4
0.6
0.8
1
File size
Res
ourc
e ef
fici
ency
, η
Blockage distance = 50mBlockage distance = 75mBlockage distance = 100mBlockage distance = 125mBlockage distance = 150m
Fig. 5.17. Average resource efficiency in the SOTM environment (Sce-
nario 3, blockage rate = 20%).
138
Scenario 4 : In this scenario, we show the performance of the pro-
posed framework according to β. Fig. 5.18 shows the average goodput
in the environment of Scenario 4 with a blockage ratio of 20%. It is
shown that the goodput is reduced with increasing β. This is because
the additional delay in VPCBs occurs as mentioned above Section 5.2.4.
It cause the reduction of goodput in the proposed framework.
Scenario 5 : In this scenario, we show the performance of the pro-
posed framework for an environment with various channel blockages for
varying the blockage ratio and the blockage distance. Figs. 5.19 and 5.20
indicate the goodput and the resource efficiency in the environment of
Scenario 5. It is clear that the proposed framework outperforms TCP in
terms of the goodput in an environment characterized by various block-
age ratios, blockage distance and α values. The goodput is improved
by about 100%–560% when the blockage ratio and α are 40% and 0.6
,respectively. Furthermore, the resource efficiency can be improved by
the proposed transmission control scheme as compared with the proposed
framework without the aid of navigation systems. The resource efficiency
is improved by about 7%–30% when the blockage ratio and α are 40%
and 0.6, respectively. Furthermore, as shown in the results of Table 5.3
and Fig. 5.20, the goodput and the resource efficiency of TCP are sig-
nificantly degraded in an environment with frequent channel blockage.
Therefore, when the difference between DO and DeGPSis small, the pro-
posed framework can outperform TCP in terms of the goodput as well
as the resource efficiency.
139
106
107
108
109
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
File size
Goo
dput
(M
bps)
TCPProp. (β = 0.1)
Prop. (β = 0.3)Prop. (β = 0.5)
Prop. (β = 0.7)
Prop. (β = 0.9)
Fig. 5.18. Average goodput in the SOTM environment (Scenario 4).
140
010
2030
40 050
1001500
0.5
1
Blockage distance (m)Blockage ratio (%)
Goo
dput
(M
bps)
Prop. (α = 0.0)
Prop. (α = 0.2)
Prop. (α = 0.4)
Prop. (α = 0.6)TCP
Fig. 5.19. Average goodput in the SOTM environment (Scenario 5).
141
010
2030
40 050
1001500.4
0.6
0.8
1
Blockage distance (m)Blockage ratio (%)
Res
ourc
e ef
fici
ency
, η
Prop. (α = 0.0)
Prop. (α = 0.2)
Prop. (α = 0.4)
Prop. (α = 0.6)TCP
Fig. 5.20. Average resource efficiency in the SOTM environment (Sce-
nario 5).
142
5.5 Summary
In this chapter, a fully reliable file transfer framework with AL-FEC is
considered for SOTM systems to enhance the network throughput. To
provide reliable file transfer, we proposed an ACK exchange protocol.
However, because the data is continuously transmitted during the chan-
nel blockage, the resource efficiency is reduced in the proposed frame-
work. Therefore, we also proposed a transmission control scheme to
improve the resource efficiency by utilizing the navigation systems which
enables the data transmission in the proposed framework to be paused
during the channel blockage. We have also theoretically derived the file
transfer time, the goodput, and the resource efficiency to demonstrate
the effectiveness of the proposed framework. Results of the performance
analysis have shown that the proposed framework outperforms legacy
TCP in terms of the goodput in spite of the GPS positioning error. Fur-
thermore, it is shown that the resource efficiency of the proposed frame-
work is improved with the information on the predicted channel blockage
provided by the navigation systems. Therefore, it is expected that the
proposed framework can be helpfully exploited in urgent message trans-
mission and the file transfer services in SOTM systems. Furthermore,
in the case of real-time services such as the video streaming and video
conference, the proposed framework using the static coding rate without
the ACK exchange protocol can be applied.
143
144
6
Conclusion
The dissertation addressed the performance enhancement scheme with
packet level coding to resolve the technical problems in wireless networks.
First, to reduce the power consumption in WSNs, the power saving mech-
anism using NC and duty cycling in the bottleneck zone is proposed. The
first proposed scheme uses RLNC in the packet forwarding to enhance
the energy efficiency and the reliability. Furthermore, the role switching
is exploited to prolong the lifetime of WSNs by means of evenly power
consumption among the nodes in the bottleneck zone. The proposed
scheme has enhanced the energy efficiency and the reliability in the bot-
tleneck zone of WSNs. Second, to use the resource efficiently and provide
the reliable transmission in MBSNs, the MRS with NC is proposed. The
second proposed scheme uses the multicasting routing information and
number of video frame packets to generate coded packets. In addition,
the reliable transmission is provided by the redundancy packet transmis-
sion based on the decoding error rate. In the dissertation, the achievable
145
rate and NC gain of the proposed scheme have been derived in MRS
with NC, theoretically. The proposed scheme has reduced the resource
usage and improved the reliability in MBSNs. Third, to enhance the
network throughput with reliable data transmission for SOTM systems,
the reliable file transfer framework with AL-FEC is proposed. In the
third proposed scheme, the ACK exchange protocol is used to ensure
the reliability of the end-to-end data transfer. In addition, the transmis-
sion control scheme aided by navigation systems is proposed to enhance
the resource efficiency. In the dissertation, the file transfer time of the
proposed file transfer framework has been derived in SOTM systems,
theoretically. The proposed scheme has enhance the network throughput
with the efficient resource usage.
Presently, the wireless communication system is widely used in many
field. In the future, the importance of the wireless communication will
have been increased. Therefore, it is expected that the performance
enhancement schemes with packet level coding of this dissertation can be
helpfully exploited in various wireless networks for the efficient network
design.
146
References
[1] A. Goldsmith, Wireless communications. Cambridge university press,
2005. 1
[2] S. Lee and H. Lee, “Analysis of network lifetime in cluster-based sensor
networks,” IEEE Commun. Lett., vol. 14, no. 10, pp. 900–902, 2010. 2,
20, 21
[3] C. Han, T. Harrold, S. Armour, I. Krikidis, S. Videv, P. M. Grant,
H. Haas, J. Thompson, I. Ku, C.-X. Wang, T. A. Le, M. Nakhai,
J. Zhang, and L. Hanzo, “Green radio: radio techniques to enable
energy-efficient wireless networks,” IEEE Commun. Mag., vol. 49, no. 6,
pp. 46–54, 2011. 2
[4] M. van der Schaar and P. A. Chou, Multimedia over IP and wireless net-
works: compression, networking, and systems. Academic Press, 2011.
2
[5] V. Weerackody and E. G. Cuevas, “Technical challenges and perfor-
mance of satellite communications on-the-move systems,” JOHNS HOP-
KINS APL TECHNICAL DIGEST, vol. 30, no. 2, pp. 113–121, 2011.
2, 65, 68, 96, 100
147
[6] S. Katti, H. Rahul, H. Wenjun, D. Katabi, M. Medard, and J. Crowcroft,
“Xors in the air: Practical wireless network coding,” IEEE/ACM Trans.
Networking, vol. 16, no. 3, pp. 497–510, 2008. 2, 5, 76
[7] P. A. Chou and W. Yunnan, “Network coding for the internet and wire-
less networks,” IEEE Signal Processing Mag., vol. 24, no. 5, pp. 77–85,
2007. 5, 67
[8] T. Ho, M. Medard, R. Koetter, D. R. Karger, M. Effros, S. Jun, and
B. Leong, “A random linear network coding approach to multicast,”
IEEE Trans. Inf. Theory, vol. 52, no. 10, pp. 4413–4430, 2006. 6, 64, 76
[9] N. Cai and R. Yeung, “Secure network coding on a wiretap network,”
IEEE Transactions on Information Theory, vol. 57, no. 1, pp. 424–435,
Jan 2011.
[10] M. Luby, “Best practices for mobile broadcast delivery and playback of
multimedia content,” in Proc. IEEE BMSB, 2012, pp. 1–7. 6, 7, 96
[11] G. Liva, E. Paolini, and M. Chiani, “Performance versus overhead for
fountain codes over Fq,” IEEE Commun. Lett., vol. 14, no. 2, pp. 178–
180, 2010. 6, 96
[12] M. Luby, A. Shokrollahi, M. Watson, T. Stockhammer, and L. Minder,
“RaptorQ forward error correction scheme for object delivery,” RFC
6330, 2011. 7, 96, 97, 98, 100
[13] M. Luby, “Raptor codes: Algorithms and applications,” in Proc. IEEE
ICNC (Qualcomm Distinguished Lectures), 2012. 2, 6, 7, 96
148
[14] “Digital video broadcasting (DVB);framing structure, channel coding
and modulation for satellite services to handheld devices (SH) below 3
ghz,” ETSI EN 302 583, pp. 1–80, 2011. 7
[15] “Digital video broadcasting (DVB);next generation broadcasting system
to handheld, physical layer specification (DVB-NGH),” DVB Document
A160, pp. 1–295, 2012. 7
[16] D. Platz, D. H. Woldegebreal, and H. Karl, “Random network coding
in wireless sensor networks: Energy efficiency via cross-layer approach,”
in Proc. IEEE ISSSTA, 2008, pp. 654–660. 11, 16
[17] S. Hnin Yu and F. Adachi, “Power efficient adaptive network coding in
wireless sensor networks,” in Proc. IEEE ICC, 2011. 11, 16
[18] H. Yongfa, X. Juan, and J. Changjun, “Lifetime maximization in wire-
less sensor networks with network coding,” in Proc. WiCOM, 2007, pp.
2527–2530. 11, 16
[19] L. Junlin and G. AlRegib, “Network lifetime maximization for estima-
tion in multihop wireless sensor networks,” IEEE Trans. Signal Process.,
vol. 57, no. 7, pp. 2456–2466, 2009. 11, 16
[20] L. Wang, Y. Yang, and W. Zhao, “Network coding-based multipath
routing for energy efficiency in wireless sensor networks,” EURASIP J.
Wireless Comm. and Networking, vol. 2012, no. 1, pp. 1–15, 2012. 11,
16
[21] M. Wei, X. Huang, C. Wu, L. Huang, F. Wang, and Y. Zhu, “Network
coding based energy-efficient multi-path routing for wireless sensor net-
work,” in Proc. ACM MoMM, 2012, pp. 240–244. 11, 16
149
[22] L. Wang, Y. Yang, W. Zhao, L. Xu, and S. Lan, “Network-coding-based
energy-efficient data fusion and transmission for wireless sensor networks
with heterogeneous receivers,” International Journal of Distributed Sen-
sor Networks, vol. 2014, pp. 1–13, 2014. 11, 16
[23] P. Glatz, L. Hormann, C. Steger, and R. Weiss, “Opportunistic network
coding for energy conservation in wireless sensor networks,” in Proc.
IEEE CNSR, 2011, pp. 239–246. 11, 16
[24] J. Zou, H. Xiong, C. Li, R. Zhang, and Z. He, “Lifetime and distor-
tion optimization with joint source/channel rate adaptation and net-
work coding-based error control in wireless video sensor networks,” IEEE
Trans. Vehicular Tech., vol. 60, no. 3, pp. 1182–1194, 2011. 11, 16
[25] R. Chandanala, W. Zhang, R. Stoleru, and M. Won, “On combining net-
work coding with duty-cycling in flood-based wireless sensor networks,”
Ad Hoc Networks, vol. 11, no. 1, pp. 490–507, 2013. 12, 16
[26] R. R. Rout and S. K. Ghosh, “Enhancement of lifetime using duty cycle
and network coding in wireless sensor networks,” IEEE Trans. Wireless
Commun., vol. 12, no. 2, pp. 656–667, 2013. 12, 30, 40, 48, 56
[27] ——, “Adaptive data aggregation and energy efficiency using network
coding in a clustered wireless sensor network: An analytical approach,”
Comput. Commun., vol. 40, pp. 65–75, 2014. 11, 12
[28] T. Cui, T. Ho, and J. Kliewer, “Space-time communication protocols for
n-way relay networks,” in Proc. IEEE GLOBECOM, 2008, pp. 1–5. 12,
17
150
[29] D. Gunduz, A. Yener, A. Goldsmith, and H. Poor, “The multi-way relay
channel,” in Proc. IEEE ISIT, 2009, pp. 339–343. 12, 13, 17, 62, 74, 75,
76
[30] L. Ong, S. Johnson, and C. Kellett, “The capacity region of multiway
relay channels over finite fields with full data exchange,” IEEE Trans.
Inf. Theory, vol. 57, no. 5, pp. 3016–3031, 2011. 13, 17
[31] O. Iscan, I. Latif, and C. Hausl, “Network coded multi-way relaying with
iterative decoding,” in Proc. IEEE PIMRC, 2010, pp. 482–487. 13, 17
[32] C. Hausl, O. Iscan, and F. Rossetto, “Resource allocation for asymmetric
multi-way relay communication over orthogonal channels,” EURASIP J.
Wireless Comm. and Networking, vol. 2012, no. 1, p. 20, 2012. 13
[33] A. Amah and A. Klein, “A transceive strategy for regenerative multi-
antenna multi-way relaying,” in Proc. IEEE CAMSA, 2009, pp. 352–355.
13, 17
[34] ——, “Regenerative multi-group multi-way relaying,” IEEE Trans. Veh.
Technol., vol. 60, no. 7, pp. 3017–3029, 2011. 12, 13, 17
[35] F. Vieira and J. Barros, “Network coding multicast in satellite net-
works,” in Proc. IEEE NGI, 2009, pp. 1–6. 13, 17
[36] M. Muhammad, M. Berioli, G. Liva, and G. Giambene, “Instantly de-
codable network coding protocols with unequal error protection,” in
Proc. IEEE ICC, 2013, pp. 5120–5125. 13
[37] S. Sorour and S. Valaee, “A network coded arq protocol for broadcast
151
streaming over hybrid satellite systems,” in Proc. IEEE PIMRC, 2009,
pp. 1098–1102. 13, 17
[38] F. Vieira, D. Lucani, and N. Alagha, “Codes and balances: Multibeam
satellite load balancing with coded packets,” in Proc. IEEE ICC, 2012,
pp. 3316–3321. 13, 17
[39] ——, “Load-aware soft-handovers for multibeam satellites: A network
coding perspective,” in Proc. IEEE ASMS, 2012, pp. 189–196. 13, 17
[40] I. de Fez, F. Fraile, R. Belda, and J. Guerri, “Analysis and evaluation
of adaptive LDPC AL-FEC codes for content download services,” IEEE
Trans. Multimedia, vol. 14, no. 3, pp. 641–650, 2012. 14, 18
[41] M. Luby, M. Watson, T. Gasiba, and T. Stockhammer, “Mobile data
broadcasting over MBMS tradeoffs in forward error correction,” in Proc.
ACM MUM, 2006. 14, 18
[42] I. de Fez, F. Fraile, and J. C. Guerri, “Effect of the FDT transmission
frequency for an optimum content delivery using the FLUTE protocol,”
Computer Communications, vol. 36, no. 12, pp. 1298 – 1309, 2013.
[43] J. Calabuig, J. Monserrat, D. Gozalvez, and D. Gomez-Barquero, “AL-
FEC for streaming services in LTE E-MBMS,” EURASIP J. Wireless
Comm. and Networking, vol. 2013, no. 1, p. 73, 2013.
[44] C. Bouras, N. Kanakis, V. Kokkinos, and A. Papazois, “Embracing Rap-
torQ FEC in 3GPP multicast services,” Wireless Networks, vol. 19, no. 5,
pp. 1023–1035, 2013.
152
[45] ——, “AL-FEC for streaming services over LTE systems,” in Proc. IEEE
WPMC, 2011, pp. 1–5. 14, 18
[46] A. James, A. Madhukumar, E. Kurniawan, and F. Adachi, “Performance
analysis of fountain codes in multihop relay networks,” IEEE Trans.
Veh. Technol., vol. 62, no. 9, pp. 4379–4391, 2013. 14, 18
[47] S. Yousefi, T. Chahed, S. M. M. Langari, and K. Zayer, “Comfort ap-
plications in vehicular ad hoc networks based on fountain coding,” in
Proc. IEEE VTC Spring, 2010, pp. 1–5. 14, 18
[48] H. Chiao, S. Chang, K. Li, Y. Kuo, and M. Tseng, “WiFi multicast
streaming using AL-FEC inside the trains of high-speed rails,” in Proc.
IEEE BMSB, 2012, pp. 1–6. 14, 18
[49] P. Eittenberger, T. Mladenov, and U. Krieger, “Raptor codes for P2P
streaming,” in Proc. 20th Euromicro International Conference on Par-
allel, Distributed and Network-Based Processing (PDP), 2012, pp. 327–
332. 15, 18
[50] S. Chang and H. Chiao, “CRT-based AL-FEC and its application on
streaming over the wimax networks for high-speed rail reception,” in
Proc. IEEE VTC Spring, 2012, pp. 1–6. 14, 15, 18
[51] D. Gozalvez, D. Gomez-Barquero, and T. Stockhammer, “Mobile recep-
tion of DVB-T services by means of AL-FEC protection,” in Proc. IEEE
BMSB, 2009, pp. 1–5. 15, 18
[52] D. Gomez-Barquero, D. Gozalvez, and N. Cardona, “Application layer
FEC for mobile TV delivery in IP datacast over DVB-H systems,” IEEE
Trans. Broadcasting, vol. 55, no. 2, pp. 396–406, 2009. 15, 18
153
[53] V. Pullano, C. Hellge, M. Hensel, G. Corazza, and T. Schicrl, “Appli-
cation layer FEC with long time interleaver and fast tune-in for mobile
satellite TV services,” in Proc. IEEE ICNC, 2013, pp. 108–112. 15, 18
[54] M. Li, S. Chang, H. Chiao, and Y. Kuo, “Implementation of inter-layer
protection scheme for SVC video delivery over WiFi networks,” in Proc.
IEEE ISCE, 2013, pp. 83–84. 15, 18
[55] I. Akyildiz, T. Melodia, and K. Chowdury, “Wireless multimedia sensor
networks: A survey,” IEEE Wireless Commun., vol. 14, no. 6, pp. 32–39,
2007. 19, 21
[56] I. Akyildiz, T. Melodia, and K. Chowdhury, “Wireless multimedia sensor
networks: Applications and testbeds,” Proceedings of the IEEE, vol. 96,
no. 10, pp. 1588–1605, 2008.
[57] I. T. Almalkawi, M. Guerrero Zapata, J. N. Al-Karaki, and J. Morillo-
Pozo, “Wireless multimedia sensor networks: Current trends and future
directions,” Sensors, vol. 10, no. 7, pp. 6662–6717, 2010. 19, 21
[58] D. Guinard, V. Trifa, S. Karnouskos, P. Spiess, and D. Savio, “Interact-
ing with the soa-based internet of things: Discovery, query, selection, and
on-demand provisioning of web services,” IEEE Trans. Services Com-
puting, vol. 3, no. 3, pp. 223–235, 2010. 19
[59] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami, “Internet of things
(iot): A vision, architectural elements, and future directions,” Future
Gener. Comput. Syst., vol. 29, no. 7, pp. 1645–1660, 2013. 19
[60] E. Felemban, C. Lee, and E. Ekici, “Mmspeed: multipath multi-speed
protocol for qos guarantee of reliability and. timeliness in wireless sensor
154
networks,” IEEE Trans. Mobile Comput., vol. 5, no. 6, pp. 738–754,
2006. 21, 30
[61] M. Hamdi, N. Essaddi, and N. Boudriga, “Energy-efficient routing in
wireless sensor networks using probabilistic strategies,” in Proc. IEEE
WCNC, 2008, pp. 2567–2572. 21, 30
[62] W. Qinghua and T. Zhang, “Bottleneck zone analysis in energy-
constrained wireless sensor networks,” IEEE Commun. Lett., vol. 13,
no. 6, pp. 423–425, 2009. 21
[63] “IEEE standard for information technology-telecommunications and
information exchange between systems-local and metropolitan area
networks-specific requirements - part 11: Wireless lan medium access
control (mac) and physical layer (phy) specifications,” IEEE Std 802.11-
2007, pp. C1–1184, 2007. 21, 40
[64] M. Ghaderi, D. Towsley, and J. Kurose, “Network coding performance
for reliable multicast,” in Proc. IEEE MILCOM, 2007, pp. 1–7. 33, 37,
49
[65] K. Han, Y. Liu, and J. Luo, “Duty-cycle-aware minimum-energy multi-
casting in wireless sensor networks,” IEEE/ACM Trans. Netw., vol. 21,
no. 3, pp. 910–923, 2013. 36
[66] (2015) Riverbed modeler. [Online]. Available: http://www.riverbed.
com/ 40
[67] D. Halperin, B. Greenstein, A. Sheth, and D. Wetherall, “Demystifying
802.11n power consumption,” in Proc. ACM HotPower, 2010. 40
155
[68] G. Angelopoulos, M. Medard, and A. Chandrakasan, Energy-aware hard-
ware implementation of network coding, ser. Lecture Notes in Computer
Science. Springer Berlin Heidelberg, 2011, vol. 6827, ch. 14, pp. 137–
144. 41, 49, 82
[69] V. Erceg, “TGn channel model,” IEEE 802.11 document 11-03/0940r4,
2004. 41
[70] V. Rodoplu and T. Meng, “Bits-per-joule capacity of energy-limited
wireless networks,” IEEE Trans. Wireless Commun., vol. 6, no. 3, pp.
857–865, 2007. 41
[71] M. P. Anastasopoulos, T. Taleb, P. G. Cottis, and M. S. Obaidat, “Feed-
back suppression in multicast satellite networks using game theory,”
IEEE Systems Journal, vol. 6, no. 4, pp. 657–666, 2012. 61
[72] Cisco. (2013) Cisco epc satran satellite radio access network. [Online].
Available: http://www.cisco.com 61, 62
[73] DLR. (2012) Network coding satellite experiment (next). [Online].
Available: http://www.dlr.de/ 62
[74] D. Roetter, F. Vieira, and M.-J. Montpetit, “Opportunities for network
coding in satellite networks,” Satellite and Space Communications (SSC)
Newsletter, vol. 21, no. 2, pp. 7–11, 2011. 62
[75] A. Destounis and A. Panagopoulos, “Dynamic power allocation for
broadband multi-beam satellite communication networks,” IEEE Com-
mun. Lett., vol. 15, no. 4, pp. 380–382, 2011. 62
156
[76] J. Jena and P. K. Sahu, “Rain fade and ka-band spot beam satellite
communication in india,” in IEEE RSTSCC 2010, 2010, pp. 304–306.
62
[77] H. Schwarz, D. Marpe, and T. Wiegand, “Overview of the scalable video
coding extension of the h.264/avc standard,” IEEE Trans. Circuits Syst.
Video Technol., vol. 17, no. 9, pp. 1103–1120, 2007. 65, 87
[78] “Digital video broadcasting (dvb);second generation dvb interactive
satellite system (dvb-rcs2);part 2: Lower layers for satellite standard,”
ETSI EN 301 545-2, pp. 1–195, 2012. 65, 69
[79] E. Del Re, G. Gorni, L. Ronga, and M. Castro, “A game theory approach
for dvb-rcs resource allocation,” in IEEE VTC Spring 2008, 2008, pp.
2937–2941.
[80] D. K. Petraki, M. P. Anastasopoulos, and P. G. Cottis, “Dynamic re-
source allocation for dvb-rcs networks,” International Journal of Satel-
lite Communications and Networking, vol. 26, no. 3, pp. 189–210, 2008.
65, 69
[81] M. Ghaderi, D. Towsley, and J. Kurose, “Reliability gain of network
coding in lossy wireless networks,” in IEEE INFOCOM 2008, 2008, pp.
2171–2179. 68
[82] C. Caini, R. Firrincieli, and D. Lacamera, “Pepsal: A performance en-
hancing proxy for tcp satellite connections [internetworking and resource
management in satellite systems series],” IEEE Aerosp. Electron. Syst.
Mag., vol. 22, no. 8, pp. B–9–B–16, 2007. 69, 70, 98, 103, 120, 123
157
[83] P. Davern, N. Nashid, C. J. Sreenan, and A. Zahran, “Httpep: a http
performance enhancing proxy for satellite systems,” International Jour-
nal of Next Generation Computing (IJNGC), vol. 2, 2011. 69, 70, 98,
103
[84] M. Schneider, C. Hartwanger, and H. Wolf, “Antennas for multiple spot
beam satellites,” CEAS Space Journal, vol. 2, no. 1-4, pp. 59–66, 2011.
76
[85] K. Lee, S. Cho, and J. Kim, “Performance evaluation of network coding
in IEEE 802.11 wireless ad hoc networks,” Ad Hoc Networks, vol. 16,
pp. 131–141, 2014. 76
[86] (2012) Yuv video sequences. [Online]. Available: http://trace.eas.asu.
edu/yuv/ 77
[87] H. Rutagemwa, S. Xuemin, and J. W. Mark, “Transfer delay analysis
of WAP 2.0 for short-lived flows,” IEEE Trans. Veh. Technol., vol. 56,
no. 3, pp. 1418–1426, 2007. 78, 114, 116, 117
[88] D. Silva, “Minimum-overhead network coding in the short packet
regime,” in IEEE NetCod 2012, 2012, pp. 173–178. 79
[89] J. Klaue, B. Rathke, and A. Wolisz, “Evalvid-a framework for video
transmission and quality evaluation,” in Computer Performance Evalu-
ation. Modelling Techniques and Tools, ser. Lecture Notes in Computer
Science, 2003, vol. 2794, pp. 255–272. 81
[90] T. Taleb, Y. Hadjadj-Aoul, and T. Ahmed, “Challenges, opportunities,
and solutions for converged satellite and terrestrial networks,” IEEE
Wireless Commun., vol. 18, no. 1, pp. 46–52, 2011. 95
158
[91] J. B. Schodorf, “EHF satellite communications on the move: Exper-
imental results,” MIT Lincoln Laboratory Technical Report 1098, pp.
1–108, 2003. 96, 100, 104
[92] H. Yao, “EHF satellite communications-on-the-move blockage channel
modeling,” MIT Lincoln Laboratory Technical Report 1098, pp. 1–57,
2004. 114
[93] W. M. Smith, “Channel characterization and modeling for satellite com-
munications on the move,” in Proc. IEEE MILCOM, 2005, pp. 821–827.
100, 104, 114, 122
[94] B. Smida, G. Efthymoglou, S. Ghassemzadeh, and V. Tarokh, “On ef-
fects of antenna pointing accuracy for on-the-move satellite networks,”
IEEE Trans. Veh. Technol., vol. 60, no. 4, pp. 1677–1686, 2011. 96
[95] A. Shokrollahi, “Raptor codes,” IEEE Trans. Inf. Theory, vol. 52, no. 6,
pp. 2551–2567, 2006. 96
[96] “Digital video broadcasting (DVB);upper layer FEC for DVB systems,”
ETSI TR 102 993, pp. 1–89, 2011. 96
[97] “Digital video broadcasting (DVB);IP datacast: Content delivery pro-
tocols (CDP)implementation guidelines;part 2: Ip datacast over DVB-
SH,” ETSI TS 102 591-2, pp. 1–64, 2010. 96
[98] C. Fernandez-Prades, L. Lo Presti, and E. Falletti, “Satellite radiolocal-
ization from GPS to GNSS and beyond: Novel technologies and appli-
cations for civil mass market,” Proceedings of the IEEE, vol. 99, no. 11,
pp. 1882–1904, 2011. 97
159
[99] N. Velaga, M. Quddus, A. Bristow, and Y. Zheng, “Map-aided integrity
monitoring of a land vehicle navigation system,” IEEE Trans. Intell.
Transp. Syst., vol. 13, no. 2, pp. 848–858, 2012. 97
[100] N. Drawil, H. Amar, and O. Basir, “GPS localization accuracy classifi-
cation: A context-based approach,” IEEE Trans. Intell. Transp. Syst.,
vol. 14, no. 1, pp. 262–273, 2013. 106
[101] S. Kim, J. Bazin, H. Lee, K. Choi, and S. Park, “Ground vehicle naviga-
tion in harsh urban conditions by integrating inertial navigation system,
global positioning system, odometer and vision data,” IET Radar, Sonar
Navigation, vol. 5, no. 8, pp. 814–823, 2011. 97
[102] Y. Chon and H. Cha, “LifeMap: A smartphone-based context provider
for location-based services,” IEEE Perv. Comput., vol. 10, no. 2, pp.
58–67, 2011. 97
[103] S. Barbeau, R. Perez, M. Labrador, A. Perez, P. Winters, and
N. Georggi, “A location-aware framework for intelligent real-time mobile
applications,” IEEE Perv. Comput., vol. 10, no. 3, pp. 58–67, 2011.
[104] G. Li, N. Zhang, R. Zhong, S. Liu, W. Huang, J. Fan, K.-L. Tan, L. Zhou,
and J. Feng, “TsingNUS: A location-based service system towards live
city,” in Proc. ACM SIGMOD, 2013, pp. 957–960.
[105] T. Ei and F. Wang, “A trajectory-aware handoff algorithm based on
GPS information,” Annals of telecommunications - annales des telecom-
munications, vol. 65, no. 7-8, pp. 411–417, 2010. 97
[106] Opensignal. [Online]. Available: http://opensignal.com 104
160
[107] T. Walters, D. Beesley, and S. Robinson, “Calculation of estimated time
of arrival (eta) based on thoroughfare classification and driving history,”
Nov. 7 2000, US Patent 6,144,917. 105
[108] S. Aben and J. Van Helmert, “Navigation device and method,” Jun. 23
2011, US Patent App. 12/736,799.
[109] (2015) TomTom car navigation. [Online]. Available: http://www.
tomtom.com/ 105
[110] J. Norris, Markov Chains, ser. Cambridge series in statistical and prob-
abilistic mathematics. Cambridge University Press, 1999, no. No. 2008.
114, 116
[111] M. Allman, V. Paxson, and W. Stevens, “TCP congestion control,” RFC
2581, 1999. 120
[112] S. Kim, S. Lee, and S. Chung, “An efficient algorithm for ML decoding
of raptor codes over the binary erasure channel,” IEEE Communications
Letters, vol. 12, no. 8, pp. 578–580, Aug 2008. 122
[113] J. Padhye, V. Firoiu, D. F. Towsley, and J. F. Kurose, “Modeling
TCP reno performance: A simple model and its empirical validation,”
IEEE/ACM Trans. Net., vol. 8, no. 2, pp. 133–145, 2000. 134
161