performance enhancement schemes with packet-level coding in …winner.ajou.ac.kr › publication ›...

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


4.4.1 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


5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 143

6 Conclusion 145

References 147

ix

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



nario 2). . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

5.16 Average goodput in the SOTM environment (Scenario 3,

blockage rate = 20%). . . . . . . . . . . . . . . . . . . . 137


nario 3, blockage rate = 20%). . . . . . . . . . . . . . . . 138




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

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

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.


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


6: else


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.


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.


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


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


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


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


Pow

er c

onsu

mpt

ion

(W)


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


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


Tra

ffic

load

(K

bps)


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


Net

wor

k lif

etim

e (s

)


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


Pac

ket d

eliv

ery

ratio


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


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

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.


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)


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)


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


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)


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. (Analy., α = 0.4)



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)


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)


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


nario 3, blockage rate = 20%).

138


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.


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)


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


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


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

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

http://www.riverbed.com/

http://www.riverbed.com/

[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

http://www.cisco.com

http://www.dlr.de/

[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

http://trace.eas.asu.edu/yuv/

http://trace.eas.asu.edu/yuv/

[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

http://opensignal.com

[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

http://www.tomtom.com/

http://www.tomtom.com/

performance enhancement schemes with packet-level coding in …winner.ajou.ac.kr › publication ›...

Documents