University 

of Huddersfield Repository

Long, Hua

Research routing and MAC based on LEACH and S­MAC for energy efficiency and QoS in 
wireless sensor network

Original Citation

Long, Hua (2013) Research routing and MAC based on LEACH and S­MAC for energy efficiency 
and QoS in wireless sensor network. Doctoral thesis, University of Huddersfield. 

This version is available at http://eprints.hud.ac.uk/23739/

The University Repository is a digital collection of the research output of the
University, available on Open Access. Copyright and Moral Rights for the items
on this site are retained by the individual author and/or other copyright owners.
Users may access full items free of charge; copies of full text items generally
can be reproduced, displayed or performed and given to third parties in any
format or medium for personal research or study, educational or not­for­profit
purposes without prior permission or charge, provided:

• The authors, title and full bibliographic details is credited in any copy;
• A hyperlink and/or URL is included for the original metadata page; and
• The content is not changed in any way.

For more information, including our policy and submission procedure, please
contact the Repository Team at: [email protected].

http://eprints.hud.ac.uk/
Research Routing and MAC based on LEACH and S-MAC for

Energy efficiency and QoS in Wireless Sensor Network

Hua Long

A thesis submitted to the University of Huddersfield

in partial fulfilment of the requirements for

the degree of Doctor of Philosophy

The University of Huddersfield

October 2013

1
 ABSTRACT
The wireless sensor is a micro-embedded device with weak data processing capability and
small storage space. These nodes need to complete complex jobs, including data monitoring,
acquisition and conversion, and data processing. Energy efficiency should be considered as one of
the important aspects of the Wireless Sensor Network (WSN) throughout architecture and protocol
design. At the same time, supporting Quality of Service (QoS) in WSNs is a research field, because
the time-sensitive and important information is expected for the transmitting to to the sink node
immediately. The thesis is supported by the projects entitled “The information and control system
for preventing forest fires”, and “The Erhai information management system”, funded by the
Chinese Government. Energy consumption and QoS are two main objectives of the projects. The
thesis discusses the two aspects in route and Media Access Control (MAC).
For energy efficiency, the research is based on Low Energy Adaptive Clustering Hierarchy
(LEACH) protocol. LEACH is a benchmark clustering routing protocol which imposes upon cluster
heads to complete a lot of aggregation and relay of messages to the base-station. However, there are
limitations in LEACH. LEACH does not suit a wide area in clustering strategy and multi-hop
routing. Moreover, routing protocols only focus on one factor, combining the clustering
strategy and multi-hop routing mechanism were not considered in routing protocol for performance
of network.
QoS is supported by the MAC and routing protocol. Sensor MAC(S-MAC) makes the use of
the periodically monitoring / sleeping mechanism, as well as collision and crosstalk avoidance
mechanism. The mechanism reduces energy costs. Meanwhile, it supports good scalability and
avoids the collision. However, the protocols do not take the differentiated services. For supporting
QoS,A new route protocol needs to be designed and realized on embed platforms, which has WIFI
mode and a Linux operation system to apply on the actual system.
This research project was conducted as following the steps: A new protocol called RBLEACH
is proposed to solve cluster on a widely scale based on LEACH. The area is divided into a few
areas, where LEACH is improved to alter the selecting function in each area. RBLEACH creates
routes selected by using a new algorithm to optimize the performance of the network. A new
clustering method that has been developed to use several factors is PS-ACO-LEACH. The factors
include the residual energy of the cluster head and Euclidean distances between cluster members
and a cluster head. It can optimally solve fitness function and maintain a load balance in between
the cluster head nodes, a cluster head and the base station. Based on the “Ant Colony” algorithm
and transition of probability, a new routing protocol was created by “Pheromone” to find the
optimal path of cluster heads to the base station. This protocol can reduce energy consumption of
cluster heads and unbalanced energy consumption. Simulations prove that the improved protocol
can enhance the performance of the network, including lifetime and energy conservation.
Additionally, Multi Index Adaptive Routing Algorithm (MIA-QR) was designed based on network
delay, packet loss rate and signal strength for QoS. The protocol is achieved by VC on an
embedded Linux system. The MIA-QR is tested and verified by experiment and the protocol is to
support QoS. Finally, an improved protocol (SMAC -SD) for wireless sensor networks is proposed,
2
in order to solve the problem of S-MAC protocol that consider either service differentiation or
ensure quality of service. According to service differentiation, SMAC-SD adopts an access
mechanism based on different priorities including the adjustment of priority mechanisms of channel
access probability, channel multi-request mechanisms and the configuring of waiting queues with
different priorities and RTS backoff for different service, which makes the important service
receive high channel access probability, ensuring the transmission quality of the important service.
The simulation results show that the improved protocol is able to gain amount of important service
and shortens the delay at the same time. Meanwhile, it improves the performance of the network
effectively
Key words: WSN, RBLEACH, PS-ACO-LEACH, MIA-QRP, SMAC-SD

3
 ACKNOWLEDGEMENTS

I would like to express my most sincere appreciation and thanks to those who have helped me.
First of all, I would like to extend my heartfelt thanks to my supervisors, Professor Joan Lu
and Professor Qiang Xu. Professor Joan Lu and Professor Qiang Xu have been a tremendous
mentor for me. They are humble and approachable. Their minds are full of knowledge and
academic rigor. I would like to thank them for encouraging my research in the process of topic
selection and ideas of writing. In addition, their innovative mind and unique perspectives have
profoundly inspired me in terms of my research and career.
A special thanks to my family. I am extremely grateful to my parents and my husband who
have given me endless support.
The thesis is supported by the projects entitled “The information and control system for
preventing forest fires”, and “The erhai information management system”, which come from the
Chinese Government. Special thanks for the Fund Committee of YUNNAN Science and
Technology Agency, Which is belong Chinese Government.
I thank for the Members of our group, especially Yubin Shao, Qin zhi Du, they give good
suggest for experment simulations.
Finally, I would like to thank all the teachers who have given their kind assistance. Thank you
all for your time, advice and hard work.

4
CONTENTS

CHAPTER 1 INTRODUCTION .......................................................................... 10

1.1 MOTIVATIONS OF THIS STUDY ...................................................................... 12

1.1.1 Energy Efficiency.................................................................................. 13

1.1.2 QoS issue .............................................................................................. 13

1.2 CONTRIBUTION OF THIS STUDY .................................................................... 14

1.3 THESIS OUTLINE ........................................................................................... 14

CHAPTER 2 BACKGROUND ............................................................................ 16

2.1 ARCHITECTURE OF WSN .............................................................................. 16

2.2 LIMITATIONS OF WSN .................................................................................. 19

2.3 OVERVIEW OF RESEARCH ISSUES .................................................................. 21

2.4. WSN ROUTING PROTOCOL ........................................................................... 24

2.4.1 WSN routing protocol’ characteristics.................................................. 24

2.4.2 Routing Category .................................................................................. 26

2.4.3 Clustering algorithm requires ............................................................... 27

2.4.4 Target of clustering protocol design ..................................................... 28

2.5 TYPICAL CLUSTERING ALGORITHM ANALYSIS AND COMPARISON ................. 29

2.5.1 LCA....................................................................................................... 29

2.5.2 RCC....................................................................................................... 29

2.5.3 EEHC .................................................................................................... 30

2.5.4 HEED .................................................................................................... 31

2.5.5 DWEHC ................................................................................................ 33

2.5.6 EEUC .................................................................................................... 34

2.5.7 LEACH ................................................................................................. 34

2.6 ANALYSIS ON LEACH ALGORITHM .............................................................. 36

2.6.1 Network model...................................................................................... 36

2.6.2 Forming process of cluster .................................................................... 38

2.7 EXTENDED LEACH ...................................................................................... 42

5
 2.7.1 Extended LEACH ................................................................................. 42

2.7.2 Reasons to select LEACH protocol as the improved object in the

research .................................................................................................................. 44

2.8 SWARM INTELLIGENCE AND CLUSTERING ALGORITHM ................................. 46

2.8.1 PSO's basic principles and mathematical models ................................. 46

2.8.2 ACO’s basic principles and mathematical models ............................... 50

2.9 ADVANTAGES OF APPLICATION OF PSO AND ACO IN WSN ......................... 55

2.9.1 Advantages of application of PSO in WSN .......................................... 55

2.9.2 Advantages of application of ACO in WSN ......................................... 57

2.10 SUPPORT QOS’ ROUTE AND MAC PROTOCOL ............................................. 59

2.10.1 Design requirements for ADHOC network QoS routing protocol ..... 60

2.10.3 Typical QoS routing protocols in ADHOC Network and Performance

Analysis.................................................................................................................. 60

2.10.4 Introduction to Embedded Linux routing modules ............................. 64

2.10.5 MAC Protocol of WSN to support QoS ............................................. 66

2.11 SUMMARY .................................................................................................. 68

CHAPTER 3 RBLEACH PROTOCOL................................................................ 69

3.1 ASSUMING THE NETWORK PARAMETERS ....................................................... 69

3.2 DEFINE DIVIDED AREA .................................................................................. 70

3.3 DEFINE THE NUMBER OF CLUSTER HEADS ASSIGNED..................................... 71

3.3.1 The Energy Consumption Calculation .................................................. 71

3.3.2 Energy Monitoring ................................................................................ 73

3.4 CLUSTER HEAD SELECTION ........................................................................... 73

3.4.1 Energy parameters ................................................................................ 73

3.4.2 Selected the cluster head ....................................................................... 74

3.5 THE FORMATION OF CLUSTERS...................................................................... 74

3.6 MULTI-HOP DATA TRANSMISSION AMONG CLUSTERS .................................... 75

3.7 SUMMARY .................................................................................................... 78

6
CHAPTER 4 PS-ACO-LEACH ........................................................................... 79

4.1 MAIN PROBLEMS........................................................................................... 79

4.2 ASSUMPTIONS OF PS-ACO-LEACH CLUSTERING ROUTING PROTOCOL ....... 79

4.3 DESIGN OF PS-ACO-LEACH CLUSTERING ROUTING PROTOCOL ................. 80

4.3.1 Clustering strategy of PS-ACO-LEACH .............................................. 80

4.3.2 Weight analysis of the judgment factors ............................................... 84

4.3.2 Routing process of PS-ACO-LEACH .................................................. 93

4.3.3 Routing maintenance and strategy for routing failure .......................... 96

4.5 SUMMARY .................................................................................................... 96

CHAPTER 5 MIA-QRP PROTOCOL IMPLEMENTATION............................. 97

5.1 INTRODUCTION TO MIA-QRP PROTOCOL ..................................................... 97

5.2 DESIGN OF PROTOCOL MESSAGE ................................................................... 99

5.2.1 Common head of protocol message .................................................... 100

5.2.2 Format of the “hello” message ............................................................ 101

5.2.3 Format of “response” message............................................................ 102

5.2.4 Format of update message .................................................................. 103

5.2.5 Format of delete message.................................................................... 104

5.2.6 Format of “ack” message .................................................................... 104

5.3 PATH DISCOVERY ........................................................................................ 105

5.3.1 Active path discovery ......................................................................... 105

5.3.2 Process of negative path discovery ..................................................... 112

5.4 PATH MAINTENANCE................................................................................... 115

5.4.1 Regular updating strategy of routing .................................................. 116

5.4.2 Routing deletion strategy .................................................................... 116

5.5 FINITE-STATE MACHINE MODE OF PROTOCOL.............................................. 119

5.6 ROUTING ALGORITHM ................................................................................. 121

5.6.1 Routing algorithm classification and design requirements ................. 121

5.6.2MIA-QRP routing algorithm................................................................ 122

7
 5.7 DESIGN AND IMPLEMENTATION OF MIA-QRP PROTOCOL .......................... 126

5.7.1Embedded development condition ....................................................... 126

5.7.2 Design and transplantation of wireless NIC driver ............................. 127

5.7.3 Drive function and main data structure of network device ................. 128

5.8 ARCHITECTURE OF PROTOCOL SOFTWARE .................................................. 132

5.8.1 Software layers model......................................................................... 133

5.8.2 Threading model ................................................................................. 135

5.8.3 Data stream of protocol ....................................................................... 136

5.8.4 Design of the key module and Process flow ....................................... 137

5.9 SUMMARY .................................................................................................. 146

CHAPTER 6 SMAC-SD PROTOCOL .............................................................. 147

6.1 THE IMPROVED SMAC PROTOCOL .............................................................. 147

6.2 MECHANISMS OF SMAC-SD PROTOCOL .................................................... 147

6.3 THEORETICAL ANALYSIS OF THE SMAC-SD .............................................. 152

6.4 SUMMARY ................................................................................................... 159

CHAPTER 7 SIMULATION AND ANALYSIS ............................................... 160

7.1 SIMULATION ANALYSIS OF RBLEACH PROTOCOL ..................................... 160

7.1.1 The parameters and settings of the simulation .................................... 160

7.7.2 Performance indicator of simulation ................................................... 162

7.7.3 Experiment results of simulation and comparison of relative work ... 163

7.2 SIMULATION ANALYSIS OF PS-ACO-LEACH ............................................ 166

7.2.1 Performance indicator ......................................................................... 166

7.2.2 Simulation results and evaluations...................................................... 166

7.2.3 Evaluations for simulation performance indicator .............................. 167

7.3 EMULATION OF MIA-QRP ......................................................................... 172

7.3.1 Creation of protocol testing environment ........................................... 172

7.3.2 Support setting of nodes to data package for forwarding ................... 173

7.3.3 Network configuration of testing nodes .............................................. 174

8
 7.3.4 Protocol performance testing .............................................................. 175

7.3.5 Compare with QoS protocol ............................................................... 183

7.4 SIMULATION OF SMAC-SD PROTOCOL ON THE NS2 .................................. 184

7.4.1 Realization of the SMAC-SD protocol ............................................... 184

7.4.2 Simulation scene and Parameter settings ............................................ 185

7.4.3. Performance Analysis of Low-load Protocol ..................................... 189

7.4.4 Performance Analysis of high-load Protocol ...................................... 194

7.4.5 General Analysis ................................................................................. 199

7.5 SUMMARY ................................................................................................... 203

CHAPTER 8 CONCLUSIONS AND FUTURE WORK ................................... 204

8.1 CONCLUSIONS............................................................................................. 204

8.2 FUTURE WORK ............................................................................................ 205

REFERENCES ................................................................................................... 207

APPENDIX1 ....................................................................................................... 223

APPENDIX2 ....................................................................................................... 224

9
 Chapter 1 Introduction

Wireless Sensor Network (WSN) is a self-organizational Network constituted of

hundreds of small-volume, low-electric energy and low-cost sensor nodes integrating

sensing, data collection, processing and wireless communication ability. It aims at

perceiving, collecting and processing the information of an object monitored in a

geographic area covered by a network and transmitting the information to the

background for data processing [1]. WSN has great potential for many applications in

scenarios, such as military target tracking and surveillance [2,3], natural disaster relief

[4], biomedical health monitoring [5,6], hazardous environment exploration, seismic

sensing [7], intelligent agriculture, intelligent transportation[8,9], and etc.

A typical WSN includes sensor nodes, sink nodes, infrastructure network and

sensor network management nodes. The nodes have resource constraints. The resource

constraints include limited amount of energy, short communication range, low

bandwidth and limited processing and storage in each node. Nodes transmit the data

monitored to a sink node through other nodes in a multi-hop way. The design of

communication protocol of WSN must be considered as follows: correctness, stability,

fairness, as of the traditional computer network. Except for energy efficiency,

simplicity should conform to the principles of WSN protocol. The design of the WSN

protocol should be a minimum use of power and extend the life cycle. Meanwhile,

Quality of Service (QoS) in WSNs is an on-going research field, due to the services

differentiations. The protocol should be designed to support QoS.

Typical protocols include flat routing, hierarchical and routing based on

geographic position. Optimal consumption of energy for WSN routing protocols is

essential. Hierarchical is a well-known technique with special advantages in terms of

scalability. There are numerous routings that have been developed for WSN data

communication networks. The typical examples of the hierarchical routing protocol

include:
10
 Low Energy Adaptive Clustering Hierarchy(LEACH)[10], Power-Efficient

GAthering in Sensor Information Systems(PEGASIS) [11], Tree based Energy

Efficient protocol(TREEPSI )[12].

The three protocols represent typical WSN clustering routing algorithms of three

different topological structures - cluster, chain and tree. PEGASIS is based on

chain-based protocol. This proposal builds all sensor nodes to form a chain according

to Greedy algorithm that the sum of edges must be minimum in wireless sensor

networks. In PEGASIS, each node communicates only with a close neighbor and takes

turns transmitting to the base station, thus reducing the amount of energy spent per

round. TREEPSI is tree-based protocol in which WSNs select a root node in all the

sensor nodes, the length of path form end leaf node to root/chain node in TREEPSI is

the shortest. The data will not send for a longer path. Therefore, TREEPSI can reduce

power consumption in data transmission less than the others. LEACH routing protocol

is the first hierarchical routing protocol based on data aggregation. The initial idea is to

allocate the energy load of the whole network to each sensor node on average through

equal probability, random and cyclical selection of cluster heads, thus achieving the

purpose of reducing network energy cost and extending the life cycle of the network.

The advantage of LEACH is that each node has the equal probability to be a cluster

head, which makes the energy dissipation of each node be relatively balanced.

However, it requires nodes to have high communication ability and it has low

expansibility. It is not appropriate for widely-scale networks [13, 14, 15, 16].

The recent research work [17,18,19] is focused on clustering strategy, and

multi-hop routing mechanism. The former is responsible for the topology optimization

of a network. The latter will be responsible for choosing the best path to transmit the

data, which it senses in the network, to the base station by multi-hop routing mode.

However, How to use the LEACH for widely area, Whether it is possible to combine

clustering and routing to improve network performance? At the same time, if the

swarm intelligence is introduced into the WSN, is there an advantage? This issues need
11
more research.

Another aspect of research is focused on QoS [20]. Because WSN is limited by

bandwidth delay, jitter, available bandwidth and packet loss. It is required to design

and realize protocol, which supports QoS.

In the access control protocol in WSN, idle listening, collision, overhearing,

control-packet overhead, traffic fluctuation cause energy consumption. IEEE802.11

[21], Time MAC (T-MAC)[22] and S-MAC[23] are proposed to decrease the energy

costs and to support good scalability and avoid collision. S-MAC makes use of the

periodically monitoring / sleeping mechanism and collision and crosstalk avoidance

mechanism. However, the protocols do not include the differentiated services. Because

WSNs are very different from traditional networks, supporting Quality of Service (QoS)

in WSNs is still a largely unexplored research field. Based on S-MAC, an improved

protocol SMAC-SD is proposed. The new protocol added the principle of distinction

traffic, the mechanism of priority channel access based on traffic, and the mechanism

of Multi-channel request mechanism.

The route is also influenced by QoS. How to choose a route based on QoS

parameter is a research issue. The literatures [24,25,26,27] put forward these solutions.

But most of the methods only consider a QoS parameter.

Over all, Energy efficiency and QoS will be studied in the thesis.

1.1 Motivations of this Study

The thesis is supported by the projects entitled “The information and control

system for preventing forest fires”, and “The Erhai information management system”,

which come from The Chinese Government. The projects have two main objectives for

monitoring sensors: energy consumption and QoS. The thesis contains the two issues

based on route and MAC.

12
 1.1.1 Energy Efficiency

LEACH protocol is a clustering protocol, which is simple and easy to

implement. However, the following problems should be taken into consideration:

1. How to use LEACH for a wide area? Due to the fact that the cluster

head selection is randomly distributed, it will lead to a lack of uniformity in the energy

distribution of sensor nodes in that the nodes, which are far away from the base

station, will easily fail in a wide area.

2. Usually protocols only focus on one factor in terms of clustering strategy or

multi-hop routing? For example, in researching multi-hop routing, the clustering

strategy is simple. Equally, the cluster could be slightly neglected for route

optimization. In fact, combining the clustering strategy and multi-hop routing

mechanism will optimize the performance of network. This combination can be

focused on clustering strategy or multi-hop routing.

3. LEACH routing protocol was put forward based on one parameter in WSN,

such as hop counts or residual energy. Essentially, the route can be influenced by many

factors, such as distance, energy and others. What if it could select multi-parameters as

a judgment to select routes? What if combining Particle Swarm Optimization (PSO)

and Ant Colony Optimization (ACO) is used for route?

1.1.2 QoS issue

1. Different technical communities may perceive and interpret QoS in different

ways. In our project, information includes video, image and monitoring data. QoS is

accepted as a measure of the service quality that the network offers to the applications.

The purpose of MAC protocol is to focus on the underlying network, which provides

applications to the QoS.

2. QoS is needed to support the routing protocol. How can we design and embed

the routing protocol and realize these goals?

13
 1.2 Contribution of this Study

1. A new protocol called RBLEACH is proposed to solve cluster on a widely

scale. The widely scale divided into small scale and LEACH is improved to alter

selecting function in each area. RBLEACH creates routes select by using a new

algorithm to optimize the performance of network.

2. A new clustering method has been developed which uses several factors is

PS-ACO-LEACH, the factors include the residual energy of cluster head, Euclidean

distances between cluster members and a cluster head. Fitness function is optimally

solved between cluster head nodes, a cluster head and the base station. It also maintain

a load balance.

3. Based on the “Ant Colony” algorithm and transition of probability, a new

routing protocol was created by “Pheromone” to find the optimal path of cluster heads

to the base station. This protocol can avoid energy consumption of cluster heads and

unbalanced energy consumption.

4. The Multi Index Adaptive Routing Algorithm (MIA-QR) was designed based

on network delay, packet loss rate and signal strength. Protocol data formats are

designed, and MIA-QR is achieved by Visual C++ (VC) on an embedded Linux

system.

5. According to service differentiation, SMAC-SD is proposed based on S-MAC.

It adopts an access mechanism based on different priorities. The protocol adjusts

priority mechanisms of channel access probability, channel multi-request mechanisms,

and configure waiting queues with different priorities. The protocol can make

important service get high channel access probability and ensure the transmission

quality of important service.

1.3 Thesis Outline

The thesis is organized as follows. In Chapter 2, it presents the main research

14
related with the thesis. In Chapter 3, a method for improving LEACH protocol

RBLEACH is proposed. In Chapter 4, PS-ACO-LEACH is proposed combining PSO

and ACO, considering the combination between clustering and routing, and a complete

mathematical analysis to the new protocol. In Chapter 5, based on the embedded

platform that uses multi-QoS parameters, an MIA-QRP adaptive routing algorithm is

put forward and the protocol is designed and implemented. In Chapter 6, improved

SMAC-SD Protocol is proposed based on S-MAC. In Chapter 7, simulation and

analyze of protocols are discussed for improved protocols. Chapter 8 is the summary of

the thesis and overview of future works.

15
 Chapter 2 Background

2.1 Architecture of WSN

WSN [21] nodes consist of managed nodes, aggregation and sensor nodes. A

large number of sensor nodes are distributed in the monitoring region, and are

responsible for the perception and acquisition of specific target temperature, humidity,

density, light intensity, pressure and other physical characteristics, and then transmit

these data to the aggregation node through a single-hop or multi-hop manner. The user

releases monitoring tasks and data query through the management of node WSN

network configuration and management. Typical WSN network structure is as shown

in the Figure 2-1 [22][23]:

Satellite and mobile

aggregation
network

Task management
node
users

Monitoring area Sensor nodes

Figure2-1 Network structure of WSN

1.Wireless sensor nodes: WSN is composed of a large number of wireless sensor

nodes[28], which are of a micro-embedded system, containing perception modules

responsible for the perception of the external information, and are responsible for

calculating the data processing module for processing data. A wireless communication

module of sensor node is responsible for sending and receiving information.The energy

module provides energy, and each of the wireless sensor nodes can be used as a

16
network node terminal. The sensor nodes can also be used as a router for data

perception, collection, storage, integration and forwarding.

2. Aggregation nodes: WSN system has to contain one or more node base stations

(BS), also known as the aggregation node. The Aggregation node [29] is mainly

responsible for transmitting data from sensor nodes for processing (such as

compression, fusion), and then transfers it from the external network to the user. The

data can be directly transmitted to the BS, or the BS in a multi-hop manner by the

nodes. So it can be seen that the base station node is a transmission station for WSN

and external network (users) to communicate with each other and that it can support

two or more different protocol conversions between different protocols. The node base

station typically has more computing power than ordinary sensors, more storage space

and more energy reserves, and thus supports the general interaction of the entire WSN

and external network (users).

3. Management node: Users manage and configure the wireless sensor node

through the management nodes [30], publishing tasks, collecting data and inquiry data.

On the contrary, the wireless sensor node will perceive and collect data passing to the

convergence node in a single-hop or multi-hop manner, transmit the data from the

Internet or a mobile network to a remote data processing center for processing or

directly transmit them to the user.

4. From the perspective of network structure, it has a planar structure or

hierarchical structure. In the planar structure [31], all nodes have the equal status with

no difference in grades and levels. The network is divided into different levels with the

selected node as the cluster head node and a cluster head perform as the

communication backbone network, which is responsible for transmitting information in

the network to the base station.

17
 Sensor Module The wireless communication
Processor
module
Processor Receiver and
Sensor AC/DA Network MAC
Storage transmitter

Power

Figure 2.2 Wireless sensor network nodes module

The sensor nodes consist of four components: the sensor module, the processor

module, wireless communication module and power supply module. Shown in Figure

2.2[32].

The converted information and data are collected and monitored by the sensor

module it can store information.

Information is processed by processor.

Receiving and transmitting data from the own node to other nodes are undone by

the wireless communication module.

The energy used by each node is supplied by energy modules.

WSN communication protocol architecture [33] is shown below, and can be

divided into five layers: respectively; application layer, transmission layer, network

layer, data link layer and physical layer. The horizontal management plane can be

divided into the energy management plane, mobility management plane and task

management plane. The following is the description of each of the specific functions.
m
WSN protocol stack is shown in Figure 2.3[34]: e
g
a
n
a
t
n M
t e k
n
e m y s
m e t a
e g
a i
l T
g
a y n i
n g a
M b
o
Application Lay a r M
M e
n
E
Transmission control layer

Network Lay

Data Link Lay

Physical Lay

Figure 2.3 WSN Protocol

18
 The physical layer addresses the needs of a robust modulation, transmission and

receiving techniques.

Data link layer provides data framing, frame monitoring, media access and error

control.

The network layer takes care of routing the data supplied by the transport layer.

The transport layer helps to maintain the flow of data if the wireless sensor

network application requires it.

Depending on the sensing tasks, different types of application software can be set

up and used on the application layer.

The power management plane manages how a sensor node uses its power and

manages its power consumption among the three operations (sensing, computation, and

wireless communications). The mobility management plane detects and registers the

movement mobility of sensor nodes as a network control primitive. The task

management plane (i.e., cooperative efforts of sensor nodes) balances and schedules

the events’ sensing and detecting tasks from a specific area.

2.2 Limitations of WSN

Limitations of WSN include:

1. Limitation of Power

Power limit sensor nodes are usually tiny with a low price, carrying limited

energy[35][36]. There is no device transmitting external light, heat and tidal energy

into electricity. Besides, it’s impractical to change sensor nodes manually for the

amount, distance and complex environmental reasons. As the precious energy is

carried by sensors, its priority is to make full use of the energy and extend the life

cycle of WSN. Due to the consumption of a vast majority of energy by the sensor in

the wireless communication module design of WSN routing protocols, energy saving

has become a need to focus on issues for consideration.

19
 2. Limitation of Communication ability

It can be seen that there exists a positive correlation between the energy

consumption of communicating nodes with the distance. The longer the distance is, the

greater the energy consumption will be with a sharp increase. Therefore, under the

premise of guaranteeing the communication link, it is required to try to reduce the

distance of the single-hop communication, meanwhile, to design WSN routing one also

needs to optimize the communication mechanisms to meet the communication needs of

nodes in the network.

3. Limitation of computation and storage ability

The wireless sensor is a micro-embedded device, with weak data processing

capability and small storage space. These nodes finish complex work, including the

data monitoring, acquisition and conversion, the data processing [37]. For instance, one

would need to address how to take full advantage of the node itself, the computing

power and storage space and how to use the idea of working together to optimize

computing and storage node challenged in the process of WSN routing design.

To sum up, the following problems need to be solved emphatically in WSN: (1)

energy saving: due to a great number of nodes and wide distribution in the wireless

sensor network, a battery supply is a widely used method. But the energy is limited and

it is difficult to change and recharge the power for the battery supply. In order to make

maximum use of the network and guarantee the working time, the energy should be

saved as much as possible in the network; (2) network scalability: continuous

invalidation would happen in network working due to a limited amount of node energy.

The new nodes need to be added to enlarge the lifetime of the network. Thus, the

topology of network is always changeable. WSN protocol needs to overcome this

network scalability to adapt to this topological change; (3) network efficiency: the

measurements to determine network efficiency include fairness, real-timing and

utilization rate of the broad band etc.

20
 2.3 Overview of research issues

In the case of WSNs, the following aspects must also be considered [38]: network

topology control, network contract, information security, time synchronization,

positioning technology, data fusion and management, developed on embedded

operating system, application-layer technology

1. Network topology control

A good network topology is generated automatically by the topology control [39],

it is possible to improve the efficiency of the MAC protocol for time synchronization,

data fusion and targeting many aspects of the foundation, the sensor node can also save

energy to extend the life cycle of the entire network. Under the premise of

guaranteeing coverage to meet the communication link, the methods for forming

network topology in order to ensure efficient data forwarding is the focus of current

research.

2. Network contract

WSN has a dynamically changing topology; the nodes in the network are

constantly changing, medium access control (MAC) protocol is an important factor to

network contract. Network contract is time division multiple access (TDMA),

frequence division multiple access and Coding division multiple access. Another

method is competition-based MAC protocol, such as CSMA or SMAC [40,41].

3. Information security

As a task-based data-centric network, WSN not only collects data, but also needs

to address, integrate and transmit data. Reliable and secure transmission and efficient

data fusion on WSN is of paramount concern [42]. Information security should be paid

attention to the problem, in order to ensure the message's authenticity, integrity,

robustness, confidentiality, availability.

4. Time synchronization

Time synchronization is a new hot spot in the study of WSN technology. Many

21
wireless sensor network applications require the sensor nodes to maintain clock

synchronization. For WSN’s own characteristics in the synchronization range, the

synchronization accuracy and energy consumption have special requirements, so it is

not applicable to make traditional synchronizations in WSN. In the design of WSN

time synchronization mechanisms [43], it is required to take into account the stability,

convergence, energy perception and scalability and other factors simultaneously. But

because of the diversity of WSN applications, it is difficult to meet all application

requirements for time synchronization mechanism.

5. Positioning technology

The location information is very important for the wireless sensors and the single

collection of data [44], the majority of the data must be included for the location

information in order to make it sense. Determining the data collection node location or

a time position is one of the most basic functions of WSN. Deployed in the monitoring

region of the wireless sensor nodes it must be able to get its own position, and

positioning mechanisms need to consider self-organization, robustness, distributed

computing and energy efficiency.

6. Data fusion

Due to the limited scope of the perception of a WSN, the data collected usually

does not have a high degree of accuracy and redundancy. In order to improve the data

accuracy and reduce the amount of data transmitted to save energy, it is needed to send

and transmit data with WSN, use their computing power and storage capacity and

realize relay data transmission fusion [45] to remove redundant information, so as to

save node energy.

7. Data management

WSN can be seen as a distributed database. By simplifying the logical structure of

data query management and database management [46] in WSN, WSN observers can

achieve a logical view of the data stored in the network and network separation without

hindering the specific implementation process.

22
 8. Developed on embedded operating system

Each wireless sensor is a micro-embedded system containing simple data

acquisition, process control [47], wireless communications and energy supply module

with data processing algorithms input and data transfer protocol, application-specific

support.

The new generation of WSN operating system needs to handle multiple

applications concurrently, for control of hardware resources to respond to the

multi-application, the question of how to efficiently use the wireless sensor nodes

concurrency-intensive and high degree of modularity is a new challenge for WSN

operating systems.

9. Application-layer technology

WSN application layer is the application by the individual-oriented software

components, and publishes a variety of commands to perform a variety of specific

tasks to WSN. WSN application level is focused on application system development

and coordination between multi-tasking. Study of the application layer is only half

completed: application-oriented system services, based on perceptual data

understanding, decision making and actions of theory and technology [48,49].

10. QoS supporting

Different multimedia applications have different QoS requirements expressed in

terms of end to-end QoS parameters in traditional network[50].QoS requirements

generated by the applications of WSNs may be very different and traditional

end-to-end. Some new QoS parameters are desired for the measurement of the delivery

of the sensor data in an efficient and effective way. In WSN, The protocol need QoS

routing, and QoS Medium Access Control (MAC)to support QoS, includes QoS model,

QoS resource reservation signaling. They are current research issues.

The thesis is supported by the projects, which are provided financial aid by the

Chinese Government “The information and control system for preventing forest fires ”,

“Erhai information system” and etc.. The projects have two main problems for
23
monitoring sensors: energy consumption, and QoS. The thesis will research the two

issues based on route and MAC.

In conclusion, the thesis will focus on reduce energy consumption and ensure QoS.

The issues regarding MAC and route protocol will be discussed.

2.4. WSN routing protocol

2.4.1 WSN routing protocol’ characteristics

Routing protocols in WSN differ from traditional routing protocols in several

ways [51]. For one, sensor nodes do not have Internet protocol (IP) addresses, so

IP-based routing protocols cannot be used in a WSN. The design of network protocols

in a WSN needs to be scalable. It should easily manage communication among many

nodes and propagate sensor data to the base station. The protocol should meet network

resource constraints such as limited energy, communication bandwidth, memory, and

computation capabilities. By meeting these constraints, a sensor network’s lifetime can

be prolonged. Lastly, the protocol should address issues of efficiency, fault tolerance,

fairness, and security.

WSN routing protocol has the following characteristics:

1. Energy priority. The Energy of a WSN node is limited, and prolonged survival

of the entire network is an important goal, the energy consumption of nodes and

balance of network energy is an important factor [52,53,54].

2. Based on local topology information. WSN uses multi-hop communication

modes in order to save energy, nodes with limited memory resources and computing

resources, cannot store large amounts of routing information. Because of local node

topology information storage, how to implement a simple and efficient routing

mechanism is a fundamental problem in wireless sensor network [55,56,57,58].

3. Data-centric. WSN has a large number of nodes deployed randomly;

monitoring data is focused on the data forwarding path, which is importance. The
24
monitoring data have redundancy. Data-centric route finishes aggregation of data for

the energy-efficient. [59]

4. Related-applications. WSNs are applied in different environments, data

communication patterns are different, and there is no routing mechanism suitable for

all applications, which is a manifestation of sensor network. Designers need to analyze

the needs of each specific application with specific routing mechanisms and design a

routing mechanism which can adapt to those needs.

For these characteristics of WSN routing mechanisms, depending on applications

when designing routing mechanisms, it should meet the following requirements[55]:

1. Energy efficiency. WSN routing protocol not only needs to select the energy

consumption of transmission paths, but also considers entire network considerations,

chooses balanced energy consumption for the entire network and requires simple and

efficient implementation transmission of information.

2. Scalability. In WSN, area ranges or different node densities are detected,

resulting in different network sizes; if a node fails, a new node is added, node mobility,

it will make the network topology change dynamically, requiring scalable routing

mechanisms which can adapt to changes in the structure of the network.

3. Robustness. Routing mechanisms requires a certain fault tolerance because of

the exhaustion of energy or environmental factors which may cause failure of sensor

nodes, the surrounding environment also can affect the quality of wireless

communication links.

4. Fast convergence. Routing mechanisms are required to quickly to adapt to the

dynamic changes in network topology, limited node energy and communication

bandwidth.

5. Quality of Service (QoS): dynamic configuration of the network resources in

order to enable higher data efficiency issues from networking support and QoS

considerations.

25
 2.4.2 Routing Category

In general, routing in WSNs can be divided into flat-based routing,

hierarchical-based routing, and location-based routing depending on the network

structure [60]. In flat-based routing, all nodes are typically assigned equal roles or

functionality. In hierarchical-based routing; however, nodes will play different roles in

the network. In location-based routing, sensor nodes' positions are exploited to route

data in the network. A routing protocol is considered adaptive if certain system

parameters can be controlled in order to adapt to the current network conditions and

available energy levels.

As figure2.4 shows, the network is divided into a group of clusters, each cluster

selected a node as a cluster head, the job of each cluster head is to transmit information

to the base station after it finishes aggregation of the data, and removes redundant

information. Figure 2.4 shows the structure relationship between the various parts of

WSN in the cluster [61,62,63].

Figure 2.4 Structural relationships between nodes, cluster heads and base stations in

WSN
26
 After the network is clustered, cluster members are flexible, variable levels of

network formation. Its advantages are [64,65]:

1. The storage space is reduced for routing table on a single node;

2. Clustering limits the scope of interaction within the cluster, avoids excessive

exchange of information between the sensor nodes, and reduce the information stream

in the broadband of communication;

3. Clustering can effectively stabilize the network topology, reduce topology

maintenance overhead;

4. The sensors only need to focus on the connection between the cluster heads; the

impact of inter-cluster heads doesn’t change the hierarchy;

5. The cluster head can be determined by optimal management mechanism, and

further enhance network operation, extend battery life and individual nodes throughout

the network life cycle;

6.The cluster head can schedule activities within a cluster, the nodes can wake up

based on a low-power sleep mode to reduce the energy consumption rate, allowing the

sensor nodes to run in web mode and determining the send and receive timing to avoid

retry limit coverage redundancy to prevent an MAC layer of conflict [66].

7.The cluster head in a cluster can collect data, reducing the number [67] of the

relay packets.

2.4.3 Clustering algorithm requires

Regarding applications of WSNs, clustering is proposed for different structures

and design goals. The following are the structural parameters and the impact on

network clustering.

1. Network dynamics: The nodes’ moving and the energy’s limitation cause

topology change [68].

2. Network data processing: As the node may generate a lot of redundant data,

similar data on multiple node aggregation packets reduces the amount of information
27
sent [10]; one must assign a cluster head for each cluster to achieve data forwarding.

3. Node Deployment: sensor nodes have the same energy, computing and

communication capabilities when the nodes are randomly distributed; they are called

isomorphic sensor network. Cluster heads communicate without sensors; the cluster

head may not have the ability to reach the base station. The connectivity of a cluster

head is an important part of the clustering mechanism [69].

2.4.4 Target of clustering protocol design

In order to meet different application requirements, various clustering algorithms

goals are distinguished.Route hops in order to satisfy the delay-sensitive date, and the

routing performances will be disscussed.

1. The maximum lifetime

A node starts working until its energy is depleted, combining clustering routing

and adaptive clustering so as to maximize the lifetime of the network and streamline

routing design goals[70][71].

2. Load balancing

Balance of node load to extend the lifetime of a network is the goal of clustering

[72]. Load balancing is especially useful in energy constrained sensor network because

the relative energy level of the nodes does affect the network lifetime more than their

absolute energy level. When all sensors have equal initial energy and equal chances to

become sources, network could maximize its lifetime if all sensor sensors dissipate

energy at the same rate, since no loss of connectivity would result from node failure.

3. Fault tolerance

To avoid losing information, the cluster head should be fault tolerant [73].

Intuitive re-clustering is the role of the cluster nodes cluster head, rotation is also a

tolerant approach and also has the advantage of balancing the load.

4. Enhanced connectivity to reduce delay

The purpose of communication is to ensure the existence of available paths from

28
each cluster head to the point of convergence; connectivity becomes a design goal and

constraint between clusters.

2.5 Typical clustering algorithm analysis and comparison

2.5.1 LCA

Mustapha Khiati, Djamel Djenouri proposed Linked cluster algorithm (LCA) [74].

The algorithm is to build a network topology that can effectively solve the problem of

the movement of nodes. Every node broadcasts its ID and monitors the transmission of

the other nodes. In the next round, the nodes broadcast the monitored the set of

neighbors, and let every node know its 1-hop and 2-hop neighbors. Node x is defined

as the cluster head if it has the highest ID amongst the neighbors, or if it does not have

the highest in the near 1-hop neighbor but there is at least one node y in the neighbors

making node x become the highest ID node in the 1-hop neighbor of node y. When

nodes move, the cluster head can form a backbone for the network by means clustering,

which allows the cluster members to connect. LCA algorithm is prepared to maximize

the connectivity of the network.

2.5.2 RCC

K. Xu, M. Gerla proposed random competition based clustering( RCC) [75].The

protocol is designed for a mobile WSN network, RCC is the clustering based on

random competition. RCC algorithm focuses on the stability of the cluster and follows

the rule of “first claim, first win”. After receiving the message of the first node

broadcast, the neighboring nodes will join in the cluster as a member and give up the

right of being a cluster head. In order to maintain the cluster, every cluster head will

regularly broadcast the message of cluster head. The delay between message receiving

and sending which may cause conflicts can force the neighboring nodes to broadcast

the message simultaneously due to the fact that other nodes cannot sense the ongoing
29
message. To avoid this problem, RCC makes judgments by using a random timer and

node ID’s. Every node in the network will reset their time value randomly before

broadcasting the message of cluster head. During a random time, a node will simply

cancel its self-sent cluster head message if it receives the broadcasted message of

cluster head from another node. Since the random timer cannot completely solve this

problem, the nodes with a lower ID will become the cluster head if the conflict is

upheld. Although frequent movements of the nodes directly affect the stability, RCC is

more stable than a traditional clustering mechanism. A cluster head in the self-adaptive

clustering mechanism gives up its own position when sensing a node with a lower ID;

but a cluster head in the RCC gives up its own position only when a cluster head

moves close to it.

B K

B K
J
A

C E
F
D
G

Figure 2.5 Three-level cluster structures

2.5.3 EEHC

S. Bandyopadhyay, E. Coyle proposed Energy Efficient Hierarchical Clustering

30
algorithm (EEHC) [76] is the distributive clustering algorithm with the objective of

maximizing the network survival time. The cluster heads collect the data picked up by

the sensors in the clusters, and send the collected report to a focal point. In the stage of

initialization (also called the single-stage clustering), every node claims itself as the

cluster head with frequency p to the neighbor node in its communication range. These

cluster heads claim themselves as the “volunteer” cluster heads. All cluster heads that

are in the k-hop range of cluster heads receive the claims by direct communication or

repost. Any cluster head will become the nearest cluster member if it is not the cluster

head and receives the claims. The forced cluster head is neither the cluster head nor the

node that belongs to the clusters. If the claim cannot get a node with the set time t, time

t can be calculated by the duration when a message to the k-hop’s node is received. If

the claims are not in the any k-hop of the volunteer cluster head, this node is defined as

the forced cluster head. In the second stage, the process is developed to allow the

multi-level clustering, which is to build the h level cluster layers ,as shown in Figure

2.5. The process of recursive clustering repeats and forms an echoed level. The

algorithm ensures the n-hop connection between cluster head and focal points. If the

highest level is set as the h level, the sensor node will send the collected data to the

higher level cluster head. The data at level-1 can be transferred and aggregated to

level-2; the rest may be deduced by analogy. At the highest level of a clustering

structure, the cluster head will send the data to the focal points.

2.5.4 HEED

O. Younis, S. Fahmy proposed Hybrid Energy-Efficient Distributed Clustering

(HEED) [77]. The protocol can be selected from the deployment of sensor nodes. The

energy and communication cost of HEED must be considered comprehensively in the

selection of cluster head. This algorithm cannot select the cluster head node randomly,

because only the nodes with residual energy can become the cluster head node. There

are three main characteristics for this algorithm:

31
 First, the probability of being a cluster head is low in the transmission range

between two nodes. Second, the energy cost is not balanced for all nodes. Third, the

connectivity between the cluster heads can be maintained by adjusting the selection

probabilities for the given transmission range of nodes.

At HEED, every node can certainly shine upon the cluster, and communicate with

the cluster head directly. The algorithm can be classified into three stages:

C prob
Initialization phases: Firstly, set the initial proportion of all nodes in the

cluster head. This ratio is used to limit the initial cluster heads for other sensors

declared. Each sensor in accordance with the formula (2-1) as a cluster head

CH prob
determines the probability

CH= C prob × Eresidual / Emax

prob
(2.1)

Eresidual is residual energy of the current node Emax is the maximum energy when

CH prob
the battery is fully charged. is not allowed to fall below a certain threshold

Pmin , Pmin and Emax is inversed.

Repeat phase: In this phase, each sensor node goes through several iterations until

the cluster head with minimum transmission power transmission is found, if not the

cluster head is not heard, the cluster head node is allowed to send a message to its

CH prob
neighbors notifying of changes. Finally, each sensor , is multiplied and it goes

on into the next iteration. When it reaches 1, the stage is stopped. Therefore, the sensor

nodes can advertise to their neighbors; there are two types of cluster head states:

CH prob < 1
Temporary Status: If , the sensor node becomes a temporary cluster

head. If a lower power node cluster head is found, it can switch to the next iteration of

general node status.

CH prob
Final state: If is 1, the node becomes permanently cluster head.

Final phase: In this phase, each sensor node according to their status as a final
32
decision, with the selection of the least expense of a cluster head or declared itself as

the cluster head.

2.5.5 DWEHC

P. Ding, J. Holliday, A. Celik put forward Distributed Weight-Based Energy-Efficient

Hierarchical Clustering (DWEHC) [78], as shown in Figure 2.6. Every node calculates

its’ self-weight after locating the neighbor nodes. The weight is the function of the

nodes energy and the proximity of the neighbors. In the neighborhood, the node with

V8
V9
V6
V3
V7
V2
V0
H

CLUSTER
V1
RANGE
V4

V5

Figure 2.6 DWEHC cluster topology

maximum weight becomes the cluster head, and other nodes become the members. In

this stage, since the nodes directly connect with the cluster head, they are defined as

first-class members. In order to get the cluster head with the minimal energy, the nodes

will keep adjusting the relations between members. The nodes will check the minimum

power of non-cluster heads, and will judge them self to be a first level member or a

second level member with a 2-hop to be the cluster head if the distance between the

node to the neighbors is known. If the nodes can determine topology inside the cluster

with the most effective energy use, the processing will stop. In order to limit the

number of levels, every cluster is distributed in a range that each member node should

give.
33
 2.5.6 EEUC

Heinzelman W, Chandarkasan A, Balakrishnan H. proposed Energy Efficient

Uneven Clustering EEUC [79]. The algorithm classifies the routing organization into

two parts: intra-cluster communication and communication between the cluster heads

and focal points. The intra-cluster communication uses single-hop, which is easy to

achieve, while the communication between the cluster heads and focal points uses

multi-hop to prevent energy waste caused by long-distance data transmission. EEUC

algorithm uses non-uniform competitive radius so as to relatively decrease the number

of cluster members that are closed to the focal points. Thus, the cluster head can save

energy used when reposting the data in order to balance the energy cost of cluster

heads. Moreover, not only is it important to consider the position of the candidate

nodes to the focal points, but also the residual energy of candidate nodes when the

cluster head selects the next hop nodes of its routing. The experiment results show that

this routing algorithm can effectively solve the problems of non-uniform cluster head

energy cost under multi-hop communication. This algorithm can also modify the

energy cost of every node in the network and effectively prolong the lifetime of the

network.

2.5.7 LEACH

Heinzelman W, Chandrakasan A, Balakrishnan H proposed Low Energy Adaptive

Clustering Hierarchy [LEACH][10]. Based on the clustering, LEACH protocol is a

low-cost protocol and clustering algorithm with the self-adaption. The cluster head

nodes collect the data of sensor nodes from the clusters, and then the data is processed

and sent to the focal points, which is the cluster head with a layered structure. It is very

easy to select the information transmission path and save routing information without

saving a large amount of information in complex functions. LEACH routing protocol

can be mainly divided into two stages: building of the clusters and table processing.
34
The total time of these two stages is called a round. In order to decrease the cost of the

protocol, the duration of the stable processing stage is longer than the building of

clusters. However, very few nodes in LEACH usually die prematurely because of too

much energy in the network. The sequence of the death is from the distant to the close.

Thus, the communication of the region far away from the base station will be

unresponsive early in the round and the functional region of the whole network

decreases quickly while the life of network is strongly reduced.

This algorithm is one of the most popular clustering algorithms. In this thesis

improvements to the LEACH algorithm are proposed and its detailed processing steps

will be described in the next chapter.

To sum up, the comparison of various algorithms is as shown in Table 2.1[80, 81,

82].

35
 Table 2-1 Comparison of various algorithms

Converg- Cluster
Clustering Node Location Energy Failure Balanced Cluster
ence overlap
approaches mobility awareness efficient recovery clustering stability
time ping

LCA O (n) Possible No Required No Yes OK Moderate

RCC O (n) Yes No Required N/A Yes Good Moderate

EEHC O ( k1 + k2 + ⋅⋅⋅ + No No Required Yes N/A OK N/A

HEED O (1) Stationa No Not Yes N/A Good High

ry required

DWEHC O (1) Stationa No Required Yes N/A Very High

ry good

EEUC O (1) No No Required Yes N/A Good High

LEACH O (1) Fixed No Not No Yes OK Moderate

BS required

2.6 Analysis on LEACH algorithm

2.6.1 Network model

The network model uses the typical settings of the topology control in WSN,

which is the same with LEACH. N nodes are randomly distributed in the region with

the determined directions. The communication abilities of these isomorphic nodes are

weak. The calculation abilities are relatively weak, and their energy is limited. These

nodes will not move after being settled down, and they can automatically connect to

each other to build the network without the manual help. There is only one base station

in the whole network. The network which is built by the connections of nodes has the

following properties [10]:

1) The energy consumption of the nodes in the network is different in every round
36
 2) It is no need to know the detailed position of the nodes in the network and it is

only needed to know the relative distance with the other nodes.

3) There is a network synchronization clock; each sensor node senses the

monitoring parameter of the surrounding environment at a constant rate.

4) All sensor nodes in the network perceive the environment with a constant rate.

Thus, there must be data information sent to the terminal.

5) The life cycle of the network is the duration from the operation of network to

the first node death.

To sum up, it is clear that the design of the algorithm should uniformly distribute

the energy cost of the network system into all nodes. The design also needs to

guarantee scalability of the whole network and guarantee that the nodes can deal with

the relative operations in different positions. Moreover, the design should reduce the

negative effects to the network

The communication of the network uses both a symmetrical communication

channel model and a free space model. In the environment of the network, the node

energy cost of message sending with the constant distance threshold d is in proportion

to the distance between the node sending and receiving. The node energy cost of a

message sending can be calculated based on formula (2-2) and (2-3) by first order

radio model [83].

Kbit data ETx (k , d )

_ elect

Transmitting Power
circuit amplifier

ETx − elect ETx _ amp

E Rx _ elect ( k , d )
Kbit data
Receiving
circuit

kE elect

Figure 2.7 First order radio model

The energy cost produced in the data transmission of the receiving nodes is as

37
 shown in formula (2-2) [10,83]

ETx ( k , d ) ETx −elect ( k ) + ETx − amp ( k , d )

=

k ∗ Eelect + k ∗ ε fs ∗ d 2 , d < d 0

=
k ∗ Eelect + k ∗ ε mp ∗ d , d ≥ d 0 …….(2-2)
4

E Rx (k , d ) = Eelec k …… …(2-3)

Where k is the length of the data frame in the protocol, d is distance between the

two nodes, Eelec is the power consumption of sending and receiving, determined by the

circuit itself, and ε amp is the sending energy cost of an amplifier in the protocol, the

transmission distance is less than threshold, the power amplifier

ε fs of free space

model is adopted. Conversely, when the transmission distance is equal or greater than

the value, the power amplifier ε amp of multi-path attenuation model is adopted.

d0
If the transmission distance is less than threshold , the power amplifier Eelec is
the power consumption of sending and receiving, decided by the circuit itself.

ε fs
d0 =
ε mp (2-4)

2.6.2 Forming process of cluster

1. About “round”

The concept of the “round” has been introduced in LEACH [10]. The working

process of the whole sensor network is divided into several rounds with two stages in

each round: cluster creation stage and stable data communication stage. The cluster

creation stage is mainly to send the control message without sending the actual sensor

data. All nodes organize the clustering by broadcasting what is called the “short

message”, one node in every cluster is set as the cluster head. After cluster formation,

the cluster head is responsible for creating a TDMA program in the nodes. In the stage
38
of clustering, all nodes broadcast the “short message” for communication by CSMA

protocol. In the stage of stable data communication, the nodes send the data collected

continuously by every frame to the cluster head nodes based on the TDMA program.

Then the cluster head focuses the important data and sends them to the aggregation

nodes. This is regarded as a recognizable method for reducing communication traffic.

Figure 2.8 illustrates the above processes.

Stabilization stage Round

Time

Figure 2.8Working process of LEACH protocol

2. Election method of cluster head.

In the creation stage, the sensor nodes create a random number from 0 to 1. If this

number is lower than the threshold T ( n ) , the node will be the cluster head. T ( n ) can

be calculated based on formula (2-4)[12].

 p
1 − p[r mod( 1 )] n∈G
T ( n) =  p

0 other
(2-5)

Where,
p is the percentage of the expected number of the cluster head in all

nodes, which is the possibility of the nodes to be selected as cluster heads, r is the

number of rounds. G is the set of all nodes not selected as the cluster head in the

recent 1 / p rounds. In every round, T (n) is set as 0 if the nodes have been selected

as the cluster head so that there is no node selected as the cluster head. For those sensor

nodes not selected as the cluster head, they will be selected with the probability of

T (n) . The threshold T (n) of the residual nodes that will be selected as the cluster
39
heads increases as more and more nodes are selected to be the cluster head. However,

the random number percentage of the probabilities that the nodes creation is lower than

the value of the T (n) will increase. Thus the percentage of the nodes to be selected as

the cluster head will increase. As there is only one node not selected as the cluster head,

T (n) = 1 , which means this node certainly can be selected.

4. Cluster head number

The cluster head is randomly created in the region of the whole network in

LEACH protocol, including the position and number of cluster head. This kind of

random creation will lead to excessive cluster heads. In every round of LEACH

protocol, the number of the selected cluster heads follows the binomial distribution.

Assume that the percentage of the cluster head is p = k / N , the number of the selected

cluster heads in every round is X. The formula (2-6) can be obtained:

N
P { X= k=
}   p k (1 − p )
N −k
(2-6)
k 

5. Cluster head distribution

LEACH protocol randomly produces the cluster head in the whole region. This

random distribution makes the size and scope of each cluster different, as shown in

Figure 2.8. According to the analysis in Figure 2.9, it is known that this difference is

caused by clustering distribution and will cause the prerequisites of LEACH protocol

algorithm not to be ensured.

40
 100

90

80

70

60

50

40

30

20

10

0 20 40 60 80 100
Head of cluster Nodes Base Station

Figure 2.9 LEACH Cluster Distributions

The condition that LEACH protocol can effectively operate is described as

follows: the first condition is the beginning of cluster formation. The energy of all

nodes is the same at this moment. The second condition is cluster formation. The

energy consumed is generally similar. In the first life cycle of the network, the first

prerequisite is easily satisfied when the cluster head is selected in the first round.

However, the second prerequisite cannot be satisfied if the range of the clustering is

not constant. The network model is assumed as in section 2.61, that all the sensors

perceive the surroundings. Therefore, the data is sent to the users. The nodes transmit

the data at a constant rate, thus the clustering range affects the workload of the cluster

head. As mentioned above, the nodes energy that is consumed by the range of cluster

and the cluster position produced and distributed randomly by LEACH protocol has an

obvious difference in data transmission. Meanwhile, there is a large difference between

41
the contained nodes and the cluster head in terms of distance. Hence, the

communication functions between the nodes have large differences. In the following

round, the algorithm protocol will select a node that consumes more energy than the

previous round as the cluster head, which will conversely accelerate the death of the

nodes.

2.7 Extended LEACH

2.7.1 Extended LEACH

LEACH has been improved for various applications. Zuo Chen,Kai Chen

proposed LEACH-MF, the protocol adopts the methods of multi-layer clustering and

eliminates redundant information [84], B. Manzoora, N. Javaida, O. Rehmana, etl

propose Quadrature-LEACH (Q-LEACH) for homogenous networks which enhances

stability period, network life-time and throughput quiet significantly[85]. Thiemo

Voigt proposed SLEACH [86], it provides a better network lifetime as compared to

LEACH, some nodes are facilitated by solar power and these nodes act as CHs which

mainly depend upon their solar status in LEACH. Farooq, M.O, Dogar, A.B. proposed

Multi-Hop LEACH [87], it is more Energy efficient than LEACH. In Multi-Hop

LEACH cluster-heads a long distance from Base station transmit data to next

cluster-head closer to Base station instead of direct transmission to Base station. Wu

Xinhua, Wang Sheng proposed LEACH-C [88] .To LEACH-C, at the initialization of

every round, each node has to report its residual energy, resulting in copious energy

consumption. Sunkara Vinodh Kumar and Ajit Pal proposed Assisted LEACH

(A-LEACH) [89]; the protocol achieves lessened and uniform distribution of dissipated

energy by separating the tasks of Routing and Data Aggregation. Mustapha Khiati,

Djamel Djenouri proposed Broadcast over Duty-Cycle LEACH (BOD-LEACH)[90],

which adds new common static and dynamic broadcast periods to support and

accelerate broadcasting. This reduces broadcast latency and ensures scalability. JIANG
42
Yang，SUN Liu-lin[91]and others proposed a model to calculate the number of cluster

head according to energy consumption. Hu Junping, Jin Yuhui, Dou Liang proposed

LEACH-F [92], it makes some changes to the basis of LEACH. The formation of

clusters is same as LEACH-C. The base station uses simulated annealing algorithms to

generate clusters. Meanwhile, base stations generate a cluster head list for each cluster

to indicate the order of selection of nodes for cluster heads in turn. Once a cluster is

formed, its structure will no longer change. Compared to LEACH and LEACH-C, the

greatest advantage of LEACH-F is that it is unnecessary to establish clusters in each

cycle and the cost for establishing clusters is reduced. However, LEACH-F is not

appropriate for real network applications, because it cannot process the addition,

failure and motion of nodes dynamically and it increases signal interference among

clusters. The generation of cluster heads in HYENAS[93] is basically the same as

LEACH-C. It is generated by the base station with a centralized algorithm. However, it

is necessary to determine whether to cluster again according to the similarity of the

chosen clusters. The similarity of clusters can be completed according to the process of

adaptive learning. Jing Chen, Hong Shen proposed HYENAS. It avoids the

recombination of clusters and saves the cost for clustering in each cycle.

To conclude, LEACH protocol allows every node to have the chance to be the

cluster head in the whole work cycle because of the expectations, making the load

uniform, to some extent. However, the distribution and the number of cluster heads

have some randomness and there are still some cluster heads that die prematurely,

which cause a decrease in the life of the network. In order to prevent this phenomenon

from happening, the nodes with high ability should be selected as cluster head as much

as possible, and the energy used by the selected cluster head should be reduced as

much as possible.

43
 2.7.2 Reasons to select LEACH protocol as the improved object in the

research

LEACH protocol is a low energy cost, self-adapting clustering routing protocol

designed for WSN according to the layer routing. After LEACH protocol, many

clustering protocols [85,86,87,88,89,90,91,92,93] are proposed and designed based on

LEACH protocol. A lot of simulation experiments prove that this algorithm has a

better stability and longer life cycle compared with the general plane routing

algorithms and static clustering algorithms.

LEACH routing protocol has many characteristics in solving the problems of node

life cycles in WSN. For example, it reduces the number of the node transmissions and

the aggregation messages. The cluster head nodes are responsible for focusing the

message sent from different nodes in the cluster, and send the collected message to the

base station. The conflicts between or in the clusters are reduced by the MAC layer

mechanism based on TDMA/CDMA[94]. The protocol is able to reselect the cluster

head nodes to guarantee balanced use of energy in a WSN. However, there are still

disadvantages of LEACH protocol [95,96]:

1. LEACH protocol cannot guarantee an even distribution of all cluster head

nodes in the space within specified settings. In some cases, the cluster head nodes

selected by the algorithm may concentrate in some narrow region, making some of the

member nodes unable to join any cluster, or cause them to consume too much energy

in the process of data communication, thus resulting in the premature death of those

nodes, and even making it impossible to communicate with the cluster head nodes.

2. Asymmetrical distances between the cluster heads and the base station cause a

non-uniform load to occur.

In LEACH protocol clustering algorithm, the cluster head is selected to produce a

single-hop algorithm. The cluster nodes directly communicate with the base station

after aggregating the data message sent from the other nodes in the cluster. In a small

44
network scope, most communications which applied free space models between the

cluster head and base station can be ensured, which can eliminate the energy

differences consumed by the cluster head communication in different regions.

However, once the covered regions in the network become larger, the differences of

the energy cost between the nodes will become larger if the cluster head nodes select

single-hop communication again. This can cause a situation to occur in which the

nodes far from the base station die earlier than the nodes near the base station, which

will influence the characteristics of the whole network.

Figure 2.10 is a screenshot of the experiment simulation of LEACH protocol. As

shown in Figure 2.10, as the region scope of the whole network becomes larger, the

nodes loads become uneven. The life cycle of the nodes near the base station are

obviously longer, but the nodes far from the base station die prematurely because of

the excessive energy cost caused by the communication between the nodes and the

base station.

Dead nodes

Active nodes

Cluster head

Figure 2.10 Network topology at round 400 in a region set at 500m*500m

However, if the base station is on the edge of the sensor nodes and the sensor

nodes are in a wide range. The question which needs to be researched is “How to
45
improve LEACH to be more suitable for current networks

2.8 Swarm intelligence and clustering algorithm

Swarm Intelligence (Swarm Intelligence, SI), is no intelligence or a simple

intelligent individual without a centralized control case, by the simple act of a single

individual, making the entire group exhibit some intelligent behavior, so as to solve

some specific problems. Mark Millonas proposed to build a SI system which would

meet five basic principles [97,98].

1. Proximity principle (Proximity Principle): the group of individuals can obtain

information on the surrounding environment or an individual has the ability to be able

to evaluate and execute a simple calculation on time or space;

2. Quality Principle: the group of individuals can respond to changes to key

factors of the environment (Including other individuals within the group);

3. Principle of Diverse Response: different individuals in the group have different

changes in response to the behavior exhibited by the diversity within the group;

4. Stability principle: not every change in the environment would lead to changes

in behavior patterns of the entire group;

5. Adaptability Principle: environmental changes that occur in the event of

changes in population worth paying opportunities, groups must be able to change their

behavior patterns;

The ideas presented in SI provided a basic method to complex distributed

problems in which there is not centralized control and does not provide a global model.

2.8.1 PSO's basic principles and mathematical models

Proposed Particle Swarm Optimization (PSO),[99,100] by Dr. Eberhart and Dr.

Kennedy, is based on iterative optimization tools, the system is initialized within a

group of random solutions, through an iterative search it finds the optimal solution.

1. The basic principles of PSO

Assuming a scenario as follows:

46
 (1) A flock of birds in a random search for food.

(2) In this situation there is only one piece of food in the area.

(3) All the birds do not know where the food is.

(4) All the birds know their current location and how far away from the food

they are.

The basic PSO algorithm, each optimization problem is a bird in the search space,

called "particle." All particles have an optimized adaptation value (Fitness Value)

function to determine, each particle has a flying speed and can determine their

direction and distance. The particles follow the optimal particle current search in the

solution space. Reynolds study found that birds flying only use a limited number of

neighbors to track it, but the end result is the whole flock seems under control by a

center, that complex global behavior is caused by the interaction of simple rules.

2. PSO mathematical model

Basic PSO algorithm is based on the parallel optimization to search technology.

The population (Swarm) is constituted by the number of particles and the particles are

known as population size or scale (Swarm Size). Each particle has a position vector

and velocity vector, their number represents the number of problem solution space. If it

is assumed that the search space is a problem in dimensional space, the position vector

and the particle velocity vector as shown in Equation (2-7), (2-8), the position where

the component represents dimensional position, vector represents the velocity vector

the dimension velocity component.

X i = [ xi1 , xi 2 ,..., xid ]

（2-7）

Vi = [ vi1 , vi 2 ,..., vid ]

（2-8）

The basic PSO algorithm initializes a group of random particles (random

solutions), each particle "flies" in the search space through iterations to find the

optimal solution. In the iterative process, the particles, by tracking, adjust their position

to be updated by two "extremes" constantly. One extreme is the optimal solution found

47
by the particle itself, this solution is called "individual extreme" or individual optimal

Pi = [ pi1 , pi 2 ,..., pid ]

solution, the individual particles of the extreme is ; the other

extreme is the entire population which currently has an optimal solution

Pg =  pg1 , pg 2 ,..., pgd  , becomes global minimum or global optimal solution, they then

find the optimal values of these two particles (2-9, 2-10) to update their velocity and

the new position according to the formula:

1) wvij (t ) + c1r1  pij (t ) − xij (t )  + c2 r2  pgj − xij (t ) 

vij (t +=
（2-9）

xij (t + 1)= xij (t ) + vij (t + 1)

（2-10）
1 ≤ i ≤ n,1 ≤ j ≤ d

Where,
c1 、c2 are positive constants which is called the acceleration factor; r1 、r2

are random numbers between

[0,1] ; w is called inertia factor. j (1 ≤ j ≤ d ) changes

   
in the scope of the position  − x j ,max , x j ,max  and velocity range  −v j ,max , v j ,max  ,

respectively, and the dimension (range translational processing by adding symmetry),

xij vij
iteration, if or is beyond their border, and or take the boundary values. The

initial position and velocity of the particle swarm generates a random number, and then

the formula (2-9, 2-10) to iterate until the condition is satisfied.

As shown from equation (2-9), the three factors represent the three parts of the

particle velocity update.

(1) The first part reflects the influence of the next iteration, contacting the

particles current state, combining inertial particle performance, the particles ability to

balance global and local search. Therefore, it is called the inertia factor.

(2) The second part reflects the understanding of the particles “own experience",

namely the impact of the particles memory, this part of the particles has a global search

ability, to avoid falling into local minutia. Thus Venter and Socbeiski also defined this

as "self-awareness” “(Self-Confidence)".

48
 (3) The third part reflects the entire population of the "social experience”, this part

helps to improve the global search ability of the particle, and therefore of the iterative

process before considering the entire populations’ collective "experience", Venter and

Socbeiski also define this as (Swarm-Confidence)".

At present, all particles are divided into several parts in the basic PSO algorithm,

the particles are overlapping with neighbor particles, where the "neighbors" does not

general position of the particle based Euclidean distance exhaustive, but rather refers to

the number of particles adjacent to the neighbor relationship and this iterative process

is typically remain unchanged. So each particle

Pi and the neighboring historic

Pl = [ pl1 , pl 2 ,..., pld ]

optimal subgroups adjust their positions , that formula (2-9, 2-10)

exchanges according.

The basic PSO algorithm processes are shown as follows:

Begin

(1)The initialization of various random group velocities and the position of each

particle;

(2) Calculate the fitness function of particles. The position of each particle and the

fitness function are stored in an individual, the most optimal value of fitness function

and position are stored in the global variable

While （Abort condition is not satisfied）do

According to the formula (2-9) and (2-10), Update the velocity and position of

particles:

If the fitness value of particle is better than the current value ，updated to the

current value of the current position and particle fitness function; If the particle fitness

value than the current global values, with fitness function optimal particle position

update global values;

End while

End

49
 2.8.2 ACO’s basic principles and mathematical models

Ant colony optimization (Ant Colony Optimization, ACO [101, 102]) algorithm is

called the ant colony algorithm. It is proposed by Colorni, Dorigo and Maniezzo. The

basic principle is analyzed based on a mathematical model.

1. Ant behavior description

Ants, although not communicating visually, are able to communicate. They

release special secretions (Pheromones) along their path - to find their way [102].When

they encounter an intersection which they cannot pass, they select a random path, and

release (information) on the path. The longer the path of the ants, the smaller the

amount of information is released. When the ants come again across this intersection, a

larger amount of information on the path is secreted, thus a positive feedback

mechanism is formed. The amount of information on the optimal path always grows,

but the amount of information on other paths wills gradually abate. Ant colonies will

eventually find the optimal path. Meanwhile the colony is also able to adapt to changes

in the environment. For instance, if suddenly there is an obstacle on the path, ants can

quickly navigate and find the optimal path. Although the capacity of a single ant is

limited, the entire ant pheromone behavioral system has a high level of

self-organization, the exchange of information between the paths of the ants, and

optimal paths are found. Figure 2-11 illustrates further the search principle of a colony.

Figure2.11 Ant foraging Schematic

Figure 2.11, the set A point is a nest, D point is the food source, and the gray

area of the triangle is an obstacle which cannot be passed. Because of the presence of
50
obstacles, the Ant colony reaches D through E or F from A. The distance between

the points is equal to that shown in Fig 2-11 (a).

Assuming that 30 ants reach point D from A, the amount of information is left

as 1. For convenience, suppose the information residence time is 1unite time. In the

initial moment, because no information exists on the path AB, BE, EC, CD, EC, AF,

FD, located in A and D an ant can randomly select a path, from a statistical point of

view and can be considered that an equal probability exists for ants to choose AB,

AF, CD, FD, as shown in 2-11 (b), afteran unitnit of time, the amount of information

in path ABECD, AFD path has two times the amount of information. After another

period of time, there will be 20 ants who reached D from F, Fig 2-11 (c) shows, over

time, more and more ants will probably select path AFD, eventually fully choosing

path AFD, in order to find the shortest path from the nest to the food source.

2. The basic principle of the mechanism of ant colony algorithm

Simulated foraging behavior of an ant colony is a smart model. The algorithm is

based on the following basic assumptions:

(1) Information and the environment are considered by communication between

ant pheromones and each ant only responds to the local environment around them, they

can only have an impact on its local surroundings.

(2）An ant’s response to the environment is decided by their internal model.

Because ants are genetically modified organisms, ant behavior is actually an adaptive

gene expression, that ants are reactive and adaptive.

(3) To a single ant, each ant is only making independent selections according to

the environment; for the ant colony, they can form highly ordered behavior through

self-organizational processes.

The above assumptions and analysis, the parade mechanism of a basic ant colony

algorithm has two basic stages: adaptation stage and collaborative stage. In the

adaptation phase, candidate solutions are continuously adjusted based on information

accumulated in its structure, the more ants on the path leave a greater amount of
51
 information, so the paths are selected more; the amount of information that will

decrease over time. In the collaborative stage, candidate solutions are produced

through informational exchange. It is similar to the mechanism of automatic machine

learning.

Self-organization mechanisms of ant colony algorithms are a dynamic process of

entropy increase which is not affected by outside influences, the dynamic evolution is

from disorder to order, its logical structure is shown in Figure 2.12.

Ant activity planning

Individual Pheromone update Individual

ants A management ants B
Incremental Incremental
pheromone pheromone

Pheromone

Decision Point

Expression issues

Optimization
combinatorial problems

Figure 2.12 The logical structure of the basic ant colony algorithm

2. ACO mathematical model

Assuming that
bi (t ) is the number of ants at time t for element i ,the τ ij (t ) is

the amount of information in t time on the path (i, j ) , said scale TSP, the n is

{τ }
n

total number of ant colony, then; m = ∑ b=

i (t ) is
Γ ij (t ) ci , c j ⊂ C
the residual
i =1

lij
amount of information on the connection between two congregate city C at the t

time . In the initial moments, the amount of information h is equal to each path and set

τ ij (0) = const
up , the basic ant colony is implemented by a directed graph

=g (C , L, Γ)
.
52
 Ants k (k = 1, 2,..., m) determine their transfer direction during movement

according to the amount of information on each path. Taboo table

tabuk (k = 1, 2,..., m)

record the ants k with the currently traversed the city, their
tabuk makes adjustments

according to a dynamic evolutionary process. In the search process, the transition

probability of ants is calculated based on the amount of information and the

pijk (t )
information of each path. is transition probability of element at the time from

the elements (city) i to the element (city) j .

[ ]
 τ ij (t ) β g[η ik (t )]β
j ∈ allowed k
 , if
 ∑ [τ is (t )] g[η is (t )]
α β
pij (t ) =  s⊂allowed k
k


0 else
 （2.11）
allowed k = {C -tabu k }
In formula (2.11), which allows the selection of the next

city for ant k , α is a heuristic factor, it indicates the relative importance of the track,
the larger the value is, the stronger the collaboration between ants; The more ants tend

to choose the path that other ants travel; β shows the relative importance of visibility,

reflecting the ants heuristic information in motion the process to select the path in the

ant the degree of attention, the greater its value, the state transition probabilities closer

ηij (t )
greedy rule; as heuristic function, its expression is as follows

1
ηij (t ) =
dij
（2.12）

dij
In the above formula, means the distance between the two adjacent cities.

d η (t ) p k (t )
Ants k , the smaller ij is, ij and ij are greater. Obviously, this heuristic

function, said the desired level which ants moved from i the elements (city) to the

element (city) j the desired level.

In order to avoid submerged remnants heuristic information by too much residual

53
pheromone, or at the completion of each ant has completed step traversal of all cities
n (ie, the end of a cycle), the information should be updated to deal with the residue.

This update strategy mimics the characteristics of the human brain memory, stored in

the information system of the brain, while constantly on the information stored in the

brain gradually fade over time, or even forgotten. Whereby the amount of information

according to the following timing t + n rules on the adjustment path (i, j )

τ ij (t + n)= (1 − ρ )gτ ij (t ) + ∆τ ij (t )
（2.13）
m
∆τ ij (t ) =∑ ∆τ ijk (t )
k =1 （2.14）

Formula (2.13, 2.14),

ρ indicates pheromone evaporation rate, 1 − ρ taking into

account the residual pheromones, in order to prevent the accumulation of information

which has no limit, the

ρ range is ρ ⊂ 0,1) ; ∆τ ij (t ) is pheromone increments on

∆τ ij (0) =
0 ∆τ ijk (t )
the cycle path (i, j ) , is the initial time, which means the amount of

information on the first path left ant k in this cycle (i, j ) .

Depending on the pheromone update strategy, Dorigo M presents three different

basic ant colony algorithm models [98], referred to as Ant-Cycle model, Ant-Quantity

Model and Ant-Density model, the difference lies in the different methods of seeking

∆τ ijk (t )
.

1.In Ant-Cycle model

Q
 the ant k in cycle (i, j)
∆τ ijk (t ) =  Lk

0 otherwise
(2.15)

2. Ant-Quantity model:

54
 Q
 d the ant k in cycle (i, j) form t to t + 1
∆τ ijk (t ) =  ij
0 otherwise

(2.16)

3. Ant-Density model:

Q the ant k pass (i, j) form t to t + 1


∆τ (t ) = 
k
ij
0 otherwise
 （2.17）
Q is pheromone intensity, which affects the degree of convergence of the

algorithm to a certain extent;

Lk is ant k in this cycle as the total length of the path to

go.

The difference between formula (2.15) and (2.16) (2.17), (2.16) and (2.17) are use

of the local information, i.e., after the completion of the ant pheromone update path;

the formula (2.15) is the overall utilization of the information, i.e., after the completion

of a cycle Claim ant pheromone on all paths, better performance when solving TSP, so

the commonly used formula is (2.15) as the basic model of ant colony algorithm.

2.9 Advantages of application of PSO and ACO in WSN

2.9.1 Advantages of application of PSO in WSN

There is growing concern about the application of PSO in the field of combination,

optimization and network routes [103][104]. PSO is a swarm intelligence optimization

algorithm simulating intergroup social behaviors of birds in the natural world and

seeking optimal solutions through mutual cooperation among individuals. Due to the

maturity, simple realization, good self-organization and adaptation of the PSO

algorithm, it shows a strong vitality and potential in some classical combination

optimization problems and routing problems. Compared to traditional optimization

algorithms (such as linear programming and dynamic programming), it has the

55
following characteristics:

1. All particles preserve the relevant knowledge of the optimal solution. In an

environment with uncertainty, individuals constantly improve their own adaptability

through learning and adapt to the environment better through cooperation and have

good adaptation and self-organization.

2. Potential concurrency greatly improves the efficiency of searching and rate of

convergence;

3. The algorithm has a strong applicability and effectiveness and good robustness

in different environments.

Currently, PSO has been used in multicast routing and route optimization with

QoS constraints. In WSN, PSO is mainly used in deployment optimization, network

coverage and clustering routing optimization etc. Xue Wang, Jun-Jie Ma, Sheng Wang,

and Dao-Wei Bi [105] puts forward a distributed energy efficiency coverage algorithm

combining PSO and simulated annealing algorithm (SAA) which uses PSO and SAA

to optimize energy consumption of SWN and obtain optimal deployment. The result

shows that this distributed algorithm saves network energy after optimizing

deployment. This algorithm improves network energy efficiency when applied in WSN

target tracking. Santi P [106] puts forward a wireless sensor network routing algorithm

based on effective avoidance of low energy or no energy areas PSO which converts

routing problems to linear programming problems and uses PSO to solve the routing

problem of low energy area avoidance on this basis. It has a good ability to adapt to

any network deployment and meets the requirement of adapting to its environment. Hu

et al. have proposed PSO-Traffic for topological planning for a real world traffic

surveillance application [107]. Li et al. have proposed a mixture of stationary and

mobile nodes and particle swarm genetic optimization (PSGO) as a remedy to

coverage holes [108]. The PSGO hybrid is employed to determine redeployment

positions of mobile nodes in order to improve average node density. Hong et al. have

PSO Multi-Base for optimal positioning of multiple base stations in a two tier WSN
56
[109]. The two-tier network consists of nodes that can communicate only with the

application nodes they are assigned to.

However, it is not applicable to networks with high requirements for real-time

performance and its expansibility has certain limitations.

2.9.2 Advantages of application of ACO in WSN

Ant colony is a group composed of ant individuals with simple functions under

certain rules. It is characterized by heuristic and distributed calculations etc [98]..

Therefore, it is appropriate to use ant colony algorithms in routing algorithm design of

WSN.

1. Individuals of ant colony algorithms only have simple functions, but can jointly

complete complicated tasks according to simple rules. Therefore, it is feasible to use

ant colony optimization algorithms to solve the design of WSN routing algorithms at

this level.

2. Ant colony algorithm has a good superiority in solving dynamic problems. It

completes a task through the work of each ant and their interactions. In the execution

of ant colony algorithm, some individual factors will not influence optimal solutions.

Other individuals in the group are not disturbed and operate according to predefined

rules. Therefore, ant colony algorithm has a good ability to adapt to changes of

network topology and can seek a new data path in a short amount of time.

3. Local work. Nodes in WSN only record relevant information of neighbor nodes.

It is impossible to store the routing information of the whole network in each node.

Traditional routing protocol is obviously not applicable to WSN. In network routing

algorithms based on ant colony algorithm, each node achieves global routing through

calculations and influence of local information.

4. Parameter association. In the design of network routing algorithms, part

characteristic parameters of WSN such as link quality and residual energy of nodes and

parameters in ant colony algorithm such as path pheromones, path distance and path
57
angle can be associated. WSN route selection can be realized through parameter

matching of ant colony algorithm.

5. Multiple routing. Routing table of each node records the concentration of

pheromones on the path with a neighbor node. Nodes select the next path through the

use of a probability calculation based on the concentration of pheromones. The

possibility of selection among multiple paths can be supported through parameter

control. Routing designed with this idea has a good fault tolerance and stability of data

transmission.

In recent years, the application of ACO in WSN routing has gradually become a

hotspot:

Xin Guan, Guan L., Wang X.G., Ohtsuki introduces routing algorithms[110]

based on ant colony system and uses multi-ant colony competitive mechanisms to

solve the problem of energy balance and congestion control in the routing processes of

WSN.

Ziyadi, M, Yasami, K., Abolhassani, B [111] uses adaptive clustering

characteristics of ant colony to put forward a clustering protocol with energy

awareness, which effectively reduces the energy consumption of data transmission and

aggregation and greatly improves the network life cycle and energy consumption of

data transmissions.

Salehpour A.A, Mirmobin. B, Afzali-Kusha A, Mohammadi S [112] puts forward

an efficient WSN two-stage structural routing algorithm based on ant colony

optimization. Member nodes in the first stage send data to cluster heads. Among

cluster head nodes in the second stage, ant colony optimization algorithm is used to

produce efficient routing. This algorithm has little delay and good performance in

power consumption and load balance.

Ghasem Aghaei R. Rahman A.M. Rahman M.A, puts introduces the many-for-one

data transmission routing protocol based on ant colony which generates the shortest

path by using ant colony, implements congestion control mechanisms during

58
many-for-one perception data transmission and reduces end-to-end time delay and loss

of data package [113].

However, there are the following problems:

1. Single emphasis. Whether for clustering strategy or multi-hop routing,

generally only one part is considered. For example, the research on multi-hop routing

uses a simple clustering strategy, while the research on clustering neglects routing

optimization.

2. Factors considered by clustering strategy are incomplete. Currently, many new

ideas are introduced into the research on clustering strategy, such as solving “hotspot”

problems based on graph theory and using non-uniform clustering thought, based on

geographical location and energy awareness etc. However, these strategies generally

only focus on considering some factors influencing clustering in WSN.

3. Clustering results mostly have randomness. Current clustering algorithms

generally use the method of producing random numbers and setting threshold values to

increase or decrease the weight of one or more factors determining cluster heads. Such

cluster head selection is random and is not the optimal value.

4. The weight of each decision factor is not analyzed in detail when PSO is used

to obtain the optimal cluster head. For example, in research using PSO to improve

clustering algorithms such as literatures [113], they generally take the same value for

the weight of each decisive factor or give a subject value.

2.10 Support QoS’ route and MAC protocol

Quality of Service (QoS, Quality of Service) is a comprehensive evaluation of the

communication system; QoS can reflect a reasonable degree of distribution of network

resources and information services strength for a communication system to carry. The

definition of QoS in RFC2386 [114] (Request for Comments) is: QoS refers to a series

of service requirements when transmitting the data stream in the network. Typical QoS

parameters include delay, bandwidth, hops, packet loss rate and delay jitter.
59
 2.10.1 Design requirements for ADHOC network QoS routing protocol

Adhoc is a self organization network, the following are requirements [115]:

1. Quick route discovery, because the network nodes have mobility characteristics,

when moved to the network nodes, routing protocols should be able to protect the node

and quickly establish communication with the network routing;

2. QoS parameter reasonable selection, the parameters should select the

appropriate evaluation to optimize allocation of limited resources for different network

environments

3. Algorithm is feasible for multi-index constraint routing strategy, processing

algorithms should be able to draw meaningful optimal executable programs.

4. The routing protocol can perceive any dynamic changes in the network

topology.

5. Low Bandwidth, ADHOC network bandwidth resources are limited, protocols

should be compressed as much as possible to eliminate unnecessary overhead, using

lower bandwidth resources;

6 .To reduce the overhead of host resources, ADHOC network nodes should be

able to reduce the routing protocol transmission time and data transfer times while the

use of efficient routing algorithms, reduce CPU overhead, memory and other

resources;

2.10.3 Typical QoS routing protocols in ADHOC Network and Performance

Analysis

Typical QoS routing protocols in ADHOC Network and Performance Analysis

are researched by IETF MANET working group. Typical QoS routing protocols are

described as follows:

1. P.Gupta, P.R.Kumar proposed STARA（system and traffic dependent adaptive

routing algorithm)[116], it is a protocol with a minimum average path delay time as the

QoS parameter index, is a table-driven proactive routing protocol. Each node needs to
60
periodically update the routing tables of the nodes and to have higher storage capacity

requirements,

2. Anelise Munaretto, Hakim Badis proposed LS-QoS (Link State Base QoS

Routing) [117]. The protocol selects routes by using the parameter of average survival

time of the route and packets error rate. Network nodes periodically broadcast link

state data to calculate the survival time, communication link and packet error rate, the

routing table is in a dynamic state of adjustment. The protocol requires resources to

periodically maintain link data, increasing the routing overhead and energy

consumption of nodes and the need to take up storage space to save the link packet

error rate and link lifetime.

3. Shigang Chen,Klara N proposed TBP（Ticket-Based Probing）[118].This

protocol selects routes by the delay, bandwidth cost, it belongs to on-demand QoS

routing protocol. The protocol dynamically changes the routing table by sending a

number of packets (exploring packages) from the source node to the destination node.

Each packet arrives at an intermediate node, according to the link status information

and the neighbor status, the protocol determines the path to explore and discover the

packet routing. If the exploring package finally reaches the destination node, the

routing meets the QoS requirements of the path, the packet is sent by the destination

node along the path in the reverse direction of resource reservation. If the source node

does not receive the resource package, it indicates the establishment of QoS routing

failure.

4.SHI JIAN, ZOULING proposed LBRM（Local Broadcast Routing Message）

[119]. The protocol is a protocol which selects routing by delay, bandwidth and QoS

parameters as indicators of cost, it is designed based on the TBP protocol, and belongs

to on-demand QoS routing protocol. LBRM dynamically changes due to changes in

network topology. The protocol applies changes slowly in the network topology.

5.Chenxi Zhu, M. ScottCorson proposed QAODV [120], the routing protocol

selects the bandwidth of the QoS parameters as indicators, through improvements to

61
AODV protocol; it is also a flat-demand routing protocol. The protocol effectively

reduces network latency and increases network throughput, mainly used in small-scale

low-speed mobile networks.

6.Huangc, DA IF, WUJ proposed LTBR (Location-Aided Ticket-Based QoS

Routing) [121], the protocol selects delay, bandwidth and QoS parameters as indicators

of cost; it is a flat-demand routing protocols. The protocol is similar to the TBP

protocol. Only the path information is a binding agreement in the choice of location

information of nodes, effectively reducing the length of QoS routing.

7.Sivakumar, R.; Sinha, P.; Bharghavan, V. proposed CEDAR (Core Extraction

Distributed ADHOC Routing)[122],the protocol is an on-demand hierarchical routing

protocol with bandwidth. The protocol introduces the concept of core nodes to build a

network through a core node, mainly used for small and medium ADHOC networks.

8. Prasun Sinha , Raghupathy Sivakumar , Vaduvur Bharghavan proposed

MCEDAR (Multicast Core Extraction Distributed ADHOC Routing)[123] protocol is a

protocol extension which comes from CEDAR, it still uses bandwidth as the QoS

parameter index, and it is a hierarchical routing protocol. Through the expansion of the

multicast protocol it establishes a hybrid multicast structure.

9. Wang jianxin, et al proposed QRME (QoS routing protocol based on maximum

expiration time) [124], the protocol selects the maximum connection time as a

parameter indicator. Mobile node location information is located by GPS and mobile

information. It can effectively reduce the time and cost of connection establishment.

10、Li Layuan, Li Chunlin proposed HQMRP (Hierarchical QoS Multicast

Routing Protocol) [125], the protocol selects delay and bandwidth as the QoS

parameter indexes. It is a clustering, structure-based QoS multicast routing protocol

which exchanges routing information between cluster head nodes by each cluster, the

cluster node only needs to maintain routing information within the cluster. The

protocol needs to consume some energy and take up some storage space.

Characteristics of each QoS routing protocol are shown in Table 2-2. According
62
to Table 2-2 QoS analysis and comparison of various routing protocols, it shows that

demand distributed QoS routing protocols direct its development, with the size of the

structure, according to ADHOC network it can be used separately in plane distribution

and hierarchical structure and relatively large network for general use clustering

network topology. Referring to the different network application services, you can

select different QoS parameter index routing protocols.

There are many routing protocols to support QoS, but most of the existing

protocols usually use a single criterion based on node energy, bandwidth and delay

QoS indicators. In this thesis, the average delay link, the average packet loss rate and

signal strength indicators have been chosen as the node QoS parameters, to select the

optimal route to meet specific requirements.

63
 Table 2-2 Comparison of routing protocols

Protocol Support Active / QoS Distributed Network

Multicast demand Operation Topology

STARA NO Active Delay Yes Flat

LS-QoS NO Demand Packet rate, Yes Flat

Survival

time

TBP NO Demand Delay, Yes Flat

Bandwidth

LBRM Yes Demand Delay, Yes Flat

bandwidth

QAODV NO Demand Bandwidth Yes Flat

LTBR NO Demand Delay, Yes Flat

bandwidth

CEDAR NO Demand Bandwidth Yes Hierarchical

MCEDAR Yes Demand Bandwidth Yes Hierarchical

QRME Yes Demand Maximum Yes Flat-based

connection routing

time

HQMRP Yes Demand Bandwidth, Yes Hierarchical

Delay

2.10.4 Introduction to Embedded Linux routing modules

Linux operating system has powerful network functions, including network

routing modules. Routing Linux kernel supports rule-based strategy [126,127], the

rules include two selectors and an action: a selector chooses the format of the packet,

including the source address, destination, Type of Service(TOS), etc.; while the action

64
is defined by a specific treatment according to selector conditions. The action of core is

divided into 5 classes: UNICAST, BLACKHOLE, UNREACHABLE, PROHIBIT and

NAT. UNICAST which represents a routing table entry from the routing table;

BLACKHOLE Drops packets; UNREACHABLE reports that the destination network

is unreachable; PROHIBIT prohibits access; NAT expresses address translation. In

addition, the network can match the kernel and creates three rules for all packets:

local_rule (local rule), main_rule (main rule), and default rule (the default rule). The

rule-based routing mechanism kernel shown in Figure 2-13, is set in accordance with

local rules, custom rules, the main rule, the default rule followed by the

implementation of the order.

Local Rule
Local rules route
table Selector
Source Address
Destination
address
TOS
Entrance
Custom Equipment
Rule
Custom routing
ACTION： table
UNICAST
BLACKHOLE Routing table
Main UNREACHABLE entries:
Main rules routing PROHIBIT Next hop
table NAT Export

The default
The default rule Routing
table

Figure 2-13 Rule-based routing mechanism

Rules for routing table management and routing table entries, Linux kernel

provides a network interface to handle Net link rules in Net link mechanisms, and the

routing information is related:: RTM_NEWRULE, adds routing rule entries;

RTM_DELRULE, deletes the entries of a routing rule; RTM_NEWROUTE, adds a

routing table entry; RTM_DELROUTE, deletes a routing table entry.

This technology provides a basis for achieving self-developed protocols on

Linux-based embedded systems.

65
 2.10.5 MAC Protocol of WSN to support QoS

The medium access control (MAC) is responsible for the resource distribution of

the channel. The MAC can be classified into the time division multiple access,

frequency division multiple access, and coding division multiple access. Another

method is competition access control and the other types [128].

For WSN, MAC protocol has three reasons to cause energy consumption. The

first reason is the collision. When a transferred data packet is abandoned because of the

channel competition or transmission damage, the following re-transmission will

increase the loss of energy. The second reason is the listening. A node listens if other

nodes are sending the data. Listening will cause energy loss. The third reason is the

energy cost of the control message.

The Time Division Multiple Access (TDMA) [129] distributes the individual data

sending and receiving by time slot for the wireless sensor node. The data transmission

does not need too much control information; the node can fall asleep in the free time

slot. These characteristics make the WSN with TDMA mechanism avoid energy waste.

Based on the applications of the wireless sensor, the ideal TDMA MAC protocol can

be classified into the clustering networks based on TDMA type MAC protocol,

DEANA protocol [130], TRAMA protocol [131] and DMAC [132], etc. However

there are still some disadvantages in the TDMA mechanism, such as the strict time

synchronization and the poor network scalability.

The competition-based MAC protocol uses the channel according to its needs. A

node needs to use the channel by the competition to send the data. The sent data must

be resent if a collision happens until the data is sent successfully or the data is

abandoned directly. The typical competition-based MAC protocols are IEEE802.11

MAC protocol [133], S-MAC [23], T-MAC protocols [134] and so on.

Aiming at the shortfall of S-MAC protocol on the fixed duty cycle, the T-MAC

protocol makes improvements. It adjusts activity time dynamically according to traffic

66
on the basis of remaining the cycle’s length and sends messages with a burst mode,

thus reducing idle listening time [135].

Sift protocol [136] is a MAC protocol based on the competition and oriented

sensor networks aiming at even-driving. The main idea of this protocol is to use a fixed

backoff window, and select the appropriate transmission probability distribution for

nodes in different time slots, so that detecting a plurality of nodes that in a same event

can send messages without collision in the time slots before the contention window.

But it suffers from increased idle listening and system-wide time synchronization,

which reduces its energy efficiency.

With the wide use of wireless sensor networks and the diversification of user

needs, more and more services need adaption and satisfaction. In the case of an

accident, fire, destruction of public facilities and other unexpected incidents, users

expect the time-sensitive important information to be able to be transmitted to the sink

node immediately.

S-MAC protocol makes use of the periodically monitoring / sleep mechanism and

collision and crosstalk avoidance mechanism to reduce energy consumption. The

protocol is to reduce energy consumption and to support a good scalability and

collision avoidance. However, the S-MAC protocol uses the solid frame length and

duty cycle. The frame length is limited by the delay requirement and the size of the

cache. The active time of the nodes relies on the rate of the message. When the

network load is very small, the long time of free monitoring will lead to serious

transmission delay. Thus, the S-MAC is not suitable for the applications of the high

requirement of real-time. It cannot identify the priorities of the transaction and it

cannot supply the different access mechanism for the different transactions. Thus,

under the conditions of limited channel, the S-MAC cannot guarantee that the data

with high priority has the superiority in the channel competition, and it cannot ensure

the timely and effective transmission of the important transactions.

67
 2.11 Summary

1. When the sensor nodes are deployed in a widely range, LEACH protocol needs

to be improved to ensure network performance, this is an issue to be studied;

2. Particle swarm and ant colony algorithm in line with the characteristics of

WSN. Taking into account the combination of clustering and routing, PSO applied to

the ant colony clustering and applies to routing and clustering considered together with

route optimization to achieve improved routing algorithms for WSN research content.

3. There are a variety of supported QoS routing protocols. The actual

development of the route requires practicality. The special focus of this study is how to

design protocols for resource-constrained embedded systems.

4. S-MAC protocol widely used in the wireless sensor networks is unable to meet

service differentiation. Due to the lack of S-MAC protocol, there is a need for

improvement to satisfy distinct transmission services based on service differentiation

in wireless sensor networks.

68
 Chapter 3 RBLEACH PROTOCOL

As discussed in Chapter 2, analysis for LEACH protocol, cluster heads are

selected according to random numbers, but communication cost and residual energy of

cluster nodes are not considered, the nodes whose remaining energy the lowest can still

be selected as the cluster heads, which means their lifetime is short; the network nodes

are farther away from the base station which means the communication cost is more,

resulting in unbalanced energy consumption; in the data transfer phase, due to data

integration and data forwarding，nodes who consume more energy have an early death.

LEACH is not suitable for wide areas. This thesis proposes RBLEACH protocol for

widely area.

3.1 Assuming the network parameters

In this project, there is a scene, where a lot of sensors are distributed on a wide

area and the Base Station is in the corner of the area, as Lake monitoring network.

The network model used is the same as in LEACH. It only limits region, which

the nodes are randomly distributed in the square, BS is located at one end of the

rectangle, according to the assumption §2.61. The network parameters are show as

follows:

1. N nodes randomly distributed inside the area Q, Q is square (b*b).

2. The nodes are homogeneous, meaning nodes have the same initial energy,

independent of the choice the communication distance.

3. The cluster heads can merge data according to data fusion rate δ .

4. The nodes are fixed and they can automatically constitute networks without

artificial help.

5. There is only one Base Station (BS) in the whole network; BS is located at one

end of the rectangle.

69
 6. A node knows the relative distance between itself and the other nodes

(including the base station).

7. There is a network synchronization clock; each sensor node senses the

monitoring parameter of the surrounding environment at a constant rate.

8. Electromagnetic waves are transmitted by a first-order radio model (2.2)and

(2.3), assume the distance is d < d 0

ETx (k , d ) = Eelec k + ε amp kd 2

…….(2-2)

E Rx (k , d ) = Eelec k (2-3)

3.2 Define divided area

Taking the location of the Base Station as the center of coordinate, sensors are

distributed a rectangle on the first quadrant. The quadrant will be divided into arc areas

by the radius
Ri = Ani , as shown in Figure 3-1, which A is the unit length split from

side b.

Area=Si
es

b
al
Sc
1
C+

Qi

BS
90
A C
2A

Figure 3.1 Divided areas

70
 ni = 1,2,3.....C 0 ≤ Ri ≤ b (3-1)

Shown in Figure 3.1, the entire network area is divided into C + 1 region:

C = b / A (3-2)

Qi is ith region which is divided.

Qi ∈ Q

C +1
Q = ∑ Qi ,
i =1

Si = π ( Ri2+1 − Ri2 ) / 4 (i ≠ C )
(3-3)

S i = b 2 − πRc2 / 4 (i = C )

ri Qi ,
The distance between a node to the Base Station in region i is in

Ri ≤ ri ≤ Ri +1

Definitions：

R i+1 + R i
Ri =
2 (3-4)

is the number of the The distance between a node to the Base Station in
Ni

region nodes in
Qi

3.3 Define the number of cluster heads assigned

3.3.1 The Energy Consumption Calculation

As shown in figure 3.1,

N _ CH j
is the number of the cluster heads in Qi

( j = 1,2,3.... )

In regional i = 1 , the nodes are directly connected to the base station;

71
 N _ CH j = 0
'

2.
i ≥ 2 ，cluster heads has N _ CH j

Consider the energy consumption of the nodes when data is transmitting, the

energy consumed by the power amplifier and receiving data according to (2.2)
Ni
The average number of nodes is − 1 except cluster head in a cluster,
N _ CH j

Each cluster members to send 1bit data in one round, a k data fusion energy

consumption is
EDA , Data fusion rate is η , When the data need to transmit, The

amount of data after fusion is:

Ni
K =η *k * (3-5)
N _ CH j

Then the energy consumption of the cluster head in a round is

Ni Ni
ECH = k * Eelect * ( − 1) + k * E DA * + K * Eelect + K * ε amp * RtoBS
N i _ CH j N i _ CH j

(3-6)

Then the energy consumption of the cluster number in a round is

EN NO _ CH = KEelect + K * ε fs * RtoBS
(3-7)

The total energy consumption of cluster nodes and the cluster head is

E total = ∑ N i _ CH i * {ECH + (
Ni
− 1) * E NO _ CH }
N i _ CH j
j (3-8)
S 1
The average size for each cluster node is N CH , Node density ρ = S i
i

i_ i
N i _ CH i

(3-9)

The Mathematical expectation of the square of the distance between and BS is

RtoBS = ∫∫ r 2 ρ (r ,θ )dθdr

0 ≤θ ≤π , Ani ≤ r ≤ Ani+1
2 (3-10)
The expectation square distance from node to the cluster head is :

72
 Si
E ( RtoCH ) = ∫∫ ρr 2 drdθ =
8πN i _ CH j
(3-11)

Combining (2.2), (3-11) and (3-8),

∂Etotal
=0
∂N i _ CH j
(3-12)

The optimal number of cluster heads:

Ni ε fs S i Ni ε fs S i
N i _ CH jbest = ≈
4 2π ε amp Rtobs 4 2π
2
ε amp Ri2 (3-13)

3.3.2 Energy Monitoring

After the first cluster head is selected based on region, for the region Qi , each

round, the sink node will calculate the remaining average energy of each cluster

Assuming there are numbers of cluster heads

N _ CH j= l

Assuming the residual energy of each cluster is:

Definition：
Eavgi1 + Eavgi 2 ..... + Eavgim
Eavgi =
l
(3-14)

Eavgi
Represents the remaining energy in the region Qi

3.4 Cluster head selection

3.4.1 Energy parameters

In order to make the nodes with the highest energy be elected as the cluster head,

thus introducing energy parameters:

Definition:

Eij Q
Is the remaining energy of j node in i , its normalized value of the

73
remaining energy.

Eij (3-15)
β =j
E avgi Definition:

The relative distance from the node to the base station

Ri
R (i ) = 1 −
2b (3-16)

3.4.2 Selected the cluster head

Cluster head Selection is based on two things: First, the optimal number of nodes

in the network, Second the number of nodes in the front wheel in the diagram n was

selected as cluster head.

That is, each node according to a formula (3-17). The first round of random r

calculated values associated with the energy and the distance of the value between 0-1.

Definition: The cluster head selection

(3-17)

Is the corresponding weight value α 1 and α 2

(α1 + α 2 ) = 1

TQi is calculated and arranged α 1 > 0, α 2 > 0, by order of size, G is a node

set which has not been selected in the cluster head which determines the number of

clusters by format.

3.5 The formation of clusters

After the selection of the cluster head, the cluster N _ CH j head

broadcasts “SMS” to all nodes using MAC protocol of CSMA.

74
 N j is the nodes in Qi

If

E Rx j ( L, d j ) = E elect L ≤ E RThreshod
，

Node
N i belongs to the i cluster

If
N i is at the boundary of Qi and Q j ,

If

E Rx j ( L j , d ) > E Rxi ( Li , d ) N i Belongs to the i cluster

E Rx j ( L j , d ) < E Rxi ( Li , d ) N i Belongs to the j cluster

Else if

E Rx j ( L j , d ) = E Rxi ( Li , d ) Nodes can randomly select the i cluster or the j

cluster.

3.6 Multi-hop data transmission among clusters

In the stage of data transmission of LEACH protocol, the communication between

the head of cluster and base station is the single-hop, which is data transmission from

each head of cluster to the base station. As shown in Figure 3.2, all the heads of

clusters focus the data, which come from all member nodes in the clusters and transfer

the data to the base station. However, it is obvious that this single-hop method does not

work. The head of cluster far away from the base station needs to consume much

energy to send the data to the base station, thus causing early death of nodes.

75
 Head of cluster Base station

Figure 3.2 Single-hop of the head of cluster and base station

The single-hop does not work based on the analysis in Figure 3.2, so the

multi-hops should be considered in RBLEACH. With the multi-hops, each head of

cluster does not need to transfer the data with a long distance, and the energy cost of

the sensor nodes can be reduced to elongate the life cycle of the networks. Thus, in this

thesis, the multi-hops data transmission between the head of clusters is employed. The

transmission between the sensor nodes is as shown in Figure 3.3.

Head of cluster Base station

Figure 3.3 Transmission between multi-hops in the head of clusters

Compare Figure 3.2 and Figure 3.2, it is known that the data is not transferred to

the base station directly from each head of cluster. The selected head of clusters in

each round connects to each other to form a link with several branches. The data
76
collected from each head of cluster is transferred and aggregated in this link. Finally,

all the data is fused and aggregated, ready to be sent to the base station.

This thesis contributes the relative analysis based on following analysis:

In the process of data transmission, the table of all sensor nodes contains

T (T= M + K )
time slots. Each node member (except the cluster head) transmits data

i (i ≤ M )
collected to cluster head nodes at time slot . Cluster head nodes aggregate data

received and then integrate them. At the former M time slots, the transmission of data

in all clusters is completed. At subsequent K time slots, cluster head data is transmitted.

Cluster head nodes integrate data received from other cluster head nodes and data

saved by itself and send data obtained after integration to the next hop of cluster head

nodes at time slot M + j until all data in base station nodes is fully integrated.

The current cluster head maintains the energy list

{E ( CH ) , E ( CH ) , ⋅⋅⋅, E ( CH )}
1 2 m
of a neighbor cluster head, where refers to the

residual energy of cluster head of the ith cluster at the current moment. Assume that

CH1 , CH 2 , ⋅⋅⋅, CH m are neighbor cluster heads around CH k (When cluster head CH k

has data information to be transmitted to the base station, first, it is required to

calculate whether the base station is in the communication scope (radius) of its cluster

head; if it is, it will directly transmit data to the base station; otherwise, it will give the

process of transmitting data to the next hop, i.e. the process of cluster head jump. The

residual energy of neighbor cluster head nodes and the distance between this cluster

CH j Pj
head node and base station should first be considered. To cluster head ， is

called the probability of cluster head jump. Its calculation is shown in formula (3-11).

E (CH j ) α
Pj = [ ] [exp(− Ri )]β （3-20）
E max

In formula (3-11),
Emax refers to highest residual energy among all neighbor

77
cluster heads of cluster head
CH k ; α and β are two factors used to control the

energy and position of cluster heads. After the calculation of relative probability of

different neighbor cluster heads being selected, the one with a higher probability value

will become the jump cluster head of this node. Therefore, it is obvious that, the more

the residual energy there is and the closer it is to the base station, the higher the

probability of being selected as jump cluster head will be. However, the cluster head

node selected as a routing hole does not have the chance of being selected as a jump

cluster head. Therefore, such path selection mode obtained based on the residual

energy of neighbor cluster head nodes and the distance between this cluster head node

and base station.

3.7 Summary

This chapter proposed the performance of the RBLEACH algorithm. The regional

division, selection of cluster head and path depend on residual energy and distance,

these changes can improve the performance of the protocol, such as prolongs the

lifetime of each sensor node and death of the nodes, the simulation results are showed

in chapter 7.

RBLEACH is used in wide area,but it only focus on cluster select, it is need to

conside

(energy and distance as criteria for selecting head

78
 Chapter 4 PS-ACO-LEACH

4.1 Main problems

Clustering strategy or multi hop routing are research hot issues. However,

protocols only focus on one factor. For example, in researching multi-hop routing, the

clustering strategy is simple; conversely, the cluster could be slightly neglected for

route optimization. RBLEACH is used in the wide area,but it only focuses on the

selection of the cluster head, whether is the performance of network combining

the clustering strategy and multi hop routing mechanism to be optimized? Based on

these problems, in this chapter, a WSN clustering routing protocol is proposed, which

is based on the particle swarm optimization and ant colony optimization

(PS-ACO-LEACH).

4.2 Assumptions of PS-ACO-LEACH clustering routing protocol

The default scene of this protocol is a process, which is to randomly distribute the

wireless sensor nodes into a limited square area and then send the data collected from

the nodes to the base station.

The following assumptions are made of the networks considering the networks

model except §2.61

1. All sensor nodes send the solid bits of data to the head of the cluster every

time. The bits will not increase or decrease because of data amount. The information

(position, residual energy and whether being the head of cluster or not) of the sensor

nodes are all contained in the data packets with a solid length. The sensor nodes will

not send the extra data.

2. The base station uses the external power supply and the energy will not be

exhausted.

3. The setting information sent from the base station is very simple and short.
79
There is no energy cost when the sensor receives this information.

4. The radio communication models used by WSN is the same with that in section

2.6

4.3 Design of PS-ACO-LEACH clustering routing protocol

4.3.1 Clustering strategy of PS-ACO-LEACH

The clustering part of the PS-ACO-LEACH uses the centralized control strategy,

which executes the algorithm at the base station which has inexhaustible energy. The

protocol applies to the same ideas with the literature [8,45]. There are two rounds in

each round: creation of the cluster and steady-state stage. The creation of the cluster

can be implemented by the following three steps:

(1) The base station collects the nodes’ information

(2) The basic PSO is applied to find out the optimized solution of the fitness

function adjudged on the multi-factors, so as to find out the optimized clustering.

(3) The base station publishes the clustering information to the head of the cluster,

and the head of the cluster publishes the clustering information to the nodes in the

clusters.

LEACH considers the impact of the node in the cluster. Based on these two

judgment factors, in this thesis, the Euclidean distance to the head of the cluster and the

Euclidean distance from the head of the cluster to the base station are also considered

in order to improve the distribution of the cluster.

According to the energy information of all nodes, the base station calculates the

average energy of the node. The node whose node energy is more than or equal to the

average energy, is selected as the head of the cluster node. The candidate for the head

of the cluster node is set to narrow the selection range of the head of the cluster to

guarantee that the head of the cluster node finally selected has sufficiently large

amounts of energy and has the ability to play the role of head of the cluster.
80
 Assume that there are N nodes in the network, Kth node is selected as the head of

the cluster, and there are M candidates for the head of the cluster (M>>K). Thus, in

total there are Q = CM probabilities for the head of the cluster combinations.
K

Modify the problem and describe it as: decide a group as the optimized solutions

(K solutions) in M candidate head of cluster nodes to make the value of the optimized

solutions minimal.

cos t=α f1 + β f 2 + λ f 3 + η f 4 (4-1)

Where f1 、f 2 、f3 and f 4 are the evaluation factors considered in the selection of the

head of the clusters. α , β , λ and η represent the weights of each judgment factor in

the fitness function respectively. Formulas (4-2) to (4-5) are the mathematical formula

for the judgment factors

 
f1
k =1,2,..., K
∀ni ∈C p ,k
(
max  ∑ d (ni , CH p ,k ) / C p ,k ∗ d ( BS , NC ) )
 (4-2)
N' K

2 f = ∑ E (n ) / ∑ E (CH p ,k )
i
=i 1 = k 1 (4-3)

{
f3 = max d ( BS , CH p ,k ) / d ( BS , NC )
k =1,2,..., K
} (4-4)
 
f4 max  min {d (CH p ,k , CH p ,e )} − l / d ( BS , NC ) 
k =1,2,..., K e =1,2,..., K
 e ≠ k 
(4-5)
Table 4-1 shows the parameters explanations from Formulas (4-2) to (4-5). It

d ( BS , NC )
should be noted that plays the role of dimensionless elimination.

d (parameter1, parameter2)
is the Euclidean distance between two parameters.

81
 Table 4-1 The meaning of parameters in fitness function

k Head of cluster ID K Head of cluster Amount

N Amount of nodes N' Amount of current active nodes
i Nodes ID ni Node i
C p ,k Cluster k in the particle p Amount of the nodes contained
C p ,k

by cluster k in the particle p

CH p , k Head of cluster of cluster k BS Base station

in the particle p

NC Networks region center CH p ,e E head of cluster in the particle p

l Idea value of the distance

between the nearest clusters

Table 4-2 The meaning of fitness function factors

f1 Compact cluster judgment factor: calculate the maximal average

Euclidean distance from the nodes in the each cluster to the corresponding

cluster heads. The longer distance, the worse compact.

f2 Head of cluster energy judgment factor: it equals to the summation of

the current residual energy of all sensor nodes divided by the summation of

the current residual energy of head of cluster nodes.

f3 “Hot zone” Related judgment factor: equals the maximal Euclidean

distance from the head of cluster to the base station divided by the distance

from Base station to the region center. The less value of the distance, the

larger probabilities that the head of cluster concentrate in to the hot zone.

f4 Clustering uniformity judgment factor: calculate the Euclidean

distance from each cluster head to the nearest cluster head, and calculate

the difference between the value of this Euclidean distance and the ideal

distance. The less difference, the more uniform distribution of the cluster.

Based on the definition of the fitness function, the clustering method

82
corresponded with the minimal fitness value should be satisfied by the following:

(1) The average Euclidean distance from the nodes to the corresponding cluster

head should be short, which is shown by the compact distribution of the nodes in the

cluster that is quantified by

f1

(2) The summation of the cluster head energy is large, and is quantified by
f2

(3) The cluster head is close to the base station, and is quantified by
f3

(4) The distribution between the cluster heads is reasonable, and is quantified by

f4

This type of network clustering can minimize the energy cost in the clusters,

balance the energy cost and reduce the probabilities that the nodes far from the base

station become the cluster heads in order to maximally elongate the life of the nodes

and the networks.

Flow chart of the PS clustering algorithm is as shown in Figure 4-1

83
 Selecting the candidate
cluster head sets

PSO optimized
clustering

Whether particle P
Yes No
represents optimal ?

Cluster head Waiting optimal

broadcasts cluster head to
broadcast

Waiting for Non-cluster head node

other nodes sends position, energy,
information and information to the
cluster head

TDMA slot Waiting

allocation allocated slot

Transmission
of data, and
stable
operation

End

Figure 4-1 Flow chart of the PS-based clustering algorithm

4.3.2 Weight analysis of the judgment factors

The judgment factors are analyzed based on the idea: the weight analysis of each

judgment factor needs to compare the difference on a relative fair “Platform”. The “fair”

means that each judgment factor has a similar influence to the selection of the cluster

head in some ideal networks environment and in the initial state.

Assume this scene: all nodes and all cluster heads in the networks are distributed

84
uniformly, as shown in Figure 4-2. This kind of network topology and clustering are

considered temporarily to be ideal.

Calculate the initial value of each judgment factor in the above-mentioned

network environment and adjust the weight parameter of each factor accordingly to

make the final calculating value of each factor the same. Adjust slightly the value of

each weight parameter and observe the life cycle for different value cases.

Analyze the results according to the rationality and influence of the simulation

results for the judgment factor.

Figure 4-2 Network topology of the nodes uniform distribution

The following is the analysis of the initial value of four judgment factors under an

ideal network environment.

1. For f1 , each cluster head is responsible for a rectangular area under the

assumptions shown in Figure 4-2. The area of the region should follow formula (4-6).

The number of the cluster members in the region should follow formula (4-7)

S= ( X ∗Y ) ∗ p (4-6)

A = N ∗ p −1 (4-7)
85
 Where S is the area that each cluster head is responsible for, A is the node number

of a cluster, X is the length of the network region, Y is the width of network region,

and P is the percentage of cluster heads. Assume that no matter what the value of X or

Y, the region that the cluster head is responsible for is close to the square which means

the length along the side is l = S . The nodes in the cluster are distributed uniform.

If the nodes are moved around the center diagonally with some degrees, as shown in

Figure 4-3, all nodes in the cluster are distributed uniformly along the diagonal. The

2
d (ni , CH p=
,k ) ∗l
distance from the nodes to the head of cluster is 4 . Assume that the
2 2
X  Y 
coordinate of the base station is
(x y
bs , bs ) , d ( BS , NC )=  − xbs  +  − ybs 
2  2  can
be obtained. Thus, formula (4-8) can be concluded:

f1 = d (ni , CH p ,k ) / d ( BS , NC )

 2  X 2
 Y 
2

=  ∗ ( X ∗ Y ) ∗ p  /  − xbs  +  − ybs 
 4   2  2 
(4-8)

Figure 4-3 Equivalent position of cluster members in evenly distributed nodes

2. For f 2 , the energy of all nodes is equal to the initial condition. Based on formula

86
(4-3), the value of f 2 is same with the reciprocal of cluster head probabilities.
1
f2 = (4-9)
p

3. For f3 , the distribution of cluster heads is assumed to be uniform. There are

N∗p
cluster heads in the region, in which there are A cluster heads on the long side

and B cluster heads on the short side. The value of d ( BS , CH p ,k ) and (

d BS , NC )
in

formula (4-4) is analyzed. The base selection is selected in different locations because

of different application environments and purposes. The location might be in the center

of the region, at the edge of the region or someplace far away from the region. The

location of the base station is assumed as ( xbs , ybs ) , then the solutions for formula (4-11)

to solve values of A and B can be obtained:

A∗ B = N ∗ p

A X (4-10)
 B = Y

  X
=A  ( N ∗ p)∗ 
  Y 
 (4-11)
  Y 
B=  ( N ∗ p)∗ 
  X

The location of the base station is set as (0,0) , and it can be concluded that the

coordinate of x of the farthest cluster head to the base station in the current network is

 1 X  1 Y
 X − ∗  , the coordinate of y is  Y − 2 ∗ B  . According to formula (4-4), the
 2 A  

value of f3 can be obtained as:

2 2
 1 X   1 Y 
=
f3   X − 2 ∗ A  − xbs  +   Y − 2 ∗ B  − xbs  / d ( BS , NC )
     

87
 2 2
     
     
 1 X  − x  +  Y − 1 ∗ Y − x 
 X − 2 ∗  bs   bs 
X  2  Y  

  ( N ∗ p)∗  
  
   ( N ∗ p)∗   

  Y     X  
=
2 2
X  Y 
 − xbs  +  − ybs 
 2  2  (4-12)

Figure 4-4 Cluster head node distribution in uniformly distributed nodes

X
4. For f 4 , the optimal distance between the clusters is l = A , as shown in Figure
4-4. However, because the distance between the clusters in the ideal environment are

all optimized value, f 4 = 0 .

The network environment is assumed and shown in Table 4-3. (It is named as the

first category environment). Based on the formulas 4-11, 4-12 and 4-15, the value of f1 ,

f 2 , f3 and f 4 can be obtained as: f1 = 0.15 , f 2 = 10 , f3 = 2 , f 4 = 0 . In order to make

d (CH p ,k , CH p ,e )
the proportions closer, the difference between the late of the WSN

life cycle and l is assumed as 20. This means that the value of f 4 in the late

d (CH p ,k , CH p ,e )
is approximately 0.5. It is easy to set the least common multiple of

f i (i = 1, 2,3, 4 ) to be 30. The result of fi multiplying weight is 30, and the values in

Table 4-5 can be concluded.

88
 Table 4-3 Parameter settings of network environment.
X region length 100 (m) Y region width 100 (m)

Location of base (0,0) N amount of the 100

station sensor nodes

ε fs 10 ∗10−12 0.0013 ∗10−12

energy cost ε fs of
ε mp
energy cost
(J/bit/m2) (J/bit/m4)
the free space channel ε mp
of the free space
model
channel model

Initial energy of the 0.1(J) nodes uniform

nodes distribution type

Table 4-4 Weight Value of the judgment factors

α 200 β 3
λ 15 η 60

The value in Table 4-5 can be obtained by normalizing the values in Table 4-4. In

Table 4-5, ρk is the weight of the corresponding factor f k , and the value of k can be

ranged as k = 1, 2,..., m where m is the number of the factors in the self-set fitness

function. The fitness function probably selects more different factors that will

influence the clusters because of the different applications and considering

development in the future.

ρ
=ρk =
m
k
, ( k 1, 2,..., m )
∑ρ
i =1
i
(4-13)
Table 4-5 Normalizing the weight of judgment factors
α 0.71 β 0.02
λ 0.05 η 0.22

As shown in Table 4-5, it is unreasonable to set the value of weights as the same

or set the weights to be subjective, which probably would cause the completely failure

of one or more judgment factors

89
 After the creation of a fair “platform”, the value of each weight should be adjusted

to be 5 times or 20 times more. Then the difference of the simulation results between

the adjusted weight combination and the initial weight combination is observed to

determine which judgment factor is more important. It is noted that the simulation

mentioned is to peel off the multi-hops routing mechanism that the cluster head

transfer the data to the base station with single-hop like LEACH, and the networks

nodes are distributed randomly. The purpose is to determine if the judgment factors

joining the PS-ACO-LEACH are reasonable, and to analyze the importance of each

judgment factor in the fitness function.

In Figure 4-5, the solid line indicates the rising curve of the dead nodes when the

values of the weight parameter α , β , λ , η of the judgment factors are averaged. The

dotted line indicates the rising curve of the dead nodes when the values of the weight

parameters are taken in the following Table 4-5. The region indicated by the arrow is

enlarged for the comparison of two simulation curves. It is observed that the life cycle

of weight parameters calculated by formulas (4-8), (4-9) and (4-12) is improved. It is

noted that the simulation curves straight up when the number of dead nodes is close to

90, which is caused by the added settings in the program. When the proportion of the

dead nodes is close to 90%, WSN is invalid. This will result in the rapid death of

residual nodes. The following simulation figure is set with the same assumptions.

90
 Figure4-5 Comparison of weight parameters

Figures 4-5 to 4-8 show the simulation results that increase the values of α , β , λ

and η by 5 times and 20 times. Compared with the parameters in Table 4-5, the

following can be concluded:

(1). The simulation curves have a high degree of coincidence, but their

separations are not obvious, they need to be enlarged to see the tiny differences

(2) The value of the solid line is decided based on Table 4-4. Its life cycle is the

longest, demonstrating that considering the four judgment standards in this thesis

comprehensively and evenly is better than considering some judgment factor solely.

(3) The simulation curve in Figure 4-5 and Figure 4-8 has a high degree of

coincidence in the former and latter parts of the WSN statement, demonstrating that the

cluster compact judgment factors and “hot zone” relative judgment has a large

influence on the selection of the cluster head.

(4) The solid lines in Figure 4-5 and Figure 4-7 are always on the right side,

demonstrating that the cluster head energy judgment factors and the clustering

uniformity judgment factors have few influences on the cluster head. The separation in

Figure 4-9 is a little larger, demonstrating that the cluster head energy judgment factors
91
have a minimal influence in this network environment.

Figure 4-6 Comparison of weight parameters

Figure4-7 Comparison of weight parameters

92
 Figure 4-8 Comparison of weight parameters

Conclusions for the above analysis:

(1) Two judgment standards proposed in this thesis are feasible, effective and

have a positive influence on the elongation of the life cycle of proposed networks.

(2) In this thesis, the methods of analysis for the weight of judgment factors and

the weight parameters obtained based on the analysis methods are optimized in the

network environment. The analysis methods are universal and can be used to compare

the weight parameter of each judgment in the different network environments.

(3) In this network environment, the influence of each judgment is ranked from

high to low: f3 > f1 > f 4 > f 2

4.3.2 Routing process of PS-ACO-LEACH

PS-ACO-LEACH uses the basic ant colony algorithm [46-48] to select the

shortest path from the cluster head to the base station. According to ACO, The

transition probability is computed by

93
  [τ ij (t )]α + [η ik (t )}β
 , j ∈ allowed k
pijk (t ) =  ∑ [τ ij (t )] + [η ik (t )
α β

 s∈allowed k
0 else
(4-14)

α and β have to be discussed.

allowed k = {C -tabu k }
is all the states allowing ant k th to select for the next

step.

1
ηij (t ) =
dij
(4-15)

η ij dij
is the heuristic information, is the distance between the two adjacent

cities.

The information system is similar to a human’s brain, because the

information will fade with time, and even be forgotten. The amount of information

on path (i, j ) will be expressed in the moment of t + n :

τ ij (t + n) = (1 − ρ )τ ij (t ) + ∆τ ij (t )
(4-16)

∆τ ij (t )
is carried over the amount of information for k th ant in this cycle path.
ρ is a pheromone evaporation coefficient.

The flow chart is shown in Figure 4-9.

94
 Start

Parameter initialization, the K

ants are placed on the cluster
head

1 plus the
number of
iterations

Ant K = 1

Ant K = K +1
No

In accordance with the state transition No

probability formula (4-14) ,selecting the
next element

Modify taboo
table

k> = total number of ant k?

Yes

According to formula (4-15,4-16)

，updating the amount of
information

Yes

Satisfied with the end condition？

Storage calculate routes for

the cluster head

End

Figure 4-9 Flow chart of using basic AC to find optimized path

95
 4.3.3 Routing maintenance and strategy for routing failure

Affected by the environment and energy, the topology of WSN is changing

frequently. Both the death of the old nodes due to energy depletion and addition of the

new ones will contribute to changes in topology of the networks so that the routing

process fails. The PS-ACO-LEACH protocol uses the artificial ants to construct the

path, which can deal with the addition of new nodes and death of the old ones

conveniently. As a new node enters the network, the ants can rapidly construct new

communication paths according to the developed probability transmission regulation.

As an old node dies because of energy depletion or for other reasons, its previous

hop-node will find the death caused by the data transmission time-out. Then the

information consistence of this path will decrease to zero to prevent continuous visits

to the dead node. All the characteristic information of the dead node in its near nodes

list will be deleted. Finally, the next hop-node will be calculated according to the

probability transmission regulation.

4.5 Summary

According to the characteristics of WSN and in combination with two kinds of

swarm intelligence algorithms, PS-ACO-LEACH is proposed. Firstly, it improved

particle swarm fitness function in the clustering stage, considering the residual energy

of cluster head, Euclidean distance of a cluster member to a cluster head, Euclidean

distance between clusters, cluster head nodes to base station of the Euclidean distance

to optimize the cluster head configuration; Secondly, in the routing phase of ant colony

algorithm to find the best route from nodes to the base station of cluster head, cluster

head along this path transmission data have been collected.

96
 Chapter 5 MIA-QRP protocol implementation

5.1 Introduction to MIA-QRP protocol

The network is Adhoc, which can origanize the network itself without relying on

a pre-existing infrastructure. This network is one kind of WSN. RBLEACH and

PS-ACO LEACH are used for ADHOC, but the route protocls don’t consider QoS.

The routing protocols of Adhoc support the ones of the QoS [111-123], but most of the

existing protocols are criticized by one factor among the node energy, the bandwidth,

the delay and the other single QoS indexes. In this thesis, the average link delay, the

average package loss rate and the node signal strength are selected as the parameters of

QoS for the special requirements. Multiple Indicator Adaptive QoS Routing protocol

(MIA-QRP) is suggested.

Classified by routing strategy, MIA-QRP belongs to table-driven routing protocol

[117], which can support at most two-hops. During the operation of the protocol, each

node maintains a network topology table. The nodes obtain the link relation by

exchanging the topology routing information. When the topology is updated, it will use

the routing algorithm to select the optimized path. The stored information of node

topology table is as shown in Table 5-1.

Table 5-1 Node list information structure

Mac IP ParentMac ParentIP Lost Delay Quality TTL Link_state

The definition of each field is as follows:

MAC means the MAC address of nodes.

IP means the IP address of nodes.

Parent Mac means the parent MAC of nodes

Parent IP means the parent IP of nodes

Loss means the package loss rate parameters of the link formed by the nodes and

parent nodes.
97
 Delay means the delay parameters of the link formed by the nodes and parent

nodes.

Quality means the nodes’ signal strength of receiving the parent node signal

TTL means the live time of nodes.

Link_statie means the link condition between the nodes and parent nodes.

The relative definitions in this protocol are described as follows:

In order to manage the nodes information normally and conveniently, the first

stored information of the topology table is set with the relative information of each

node. The MAC and IP in the table fill the MAC address and the IP address of the each

node respectively. The parent nodes MAC and the parent nodes IP fill the

00-00-00-00-00-00 and the Loop address 127.0.0.1. The value of the lost is 0, value of

the delay is 0, value of quality is 1, value of TTL is 0 and that of the link state is 2.

Relay nodes: in a communication link, if the data communication between node A

and node C should be finished by the data transmission of node B, then node B is the

relay node of node A and node C.

Parent nodes: In a communication link, if node B is the relay node of the link

from node A to node C, then node B is the parent node of node C, and node A is the

parent node of node B.

Direct neighbor nodes: other nodes which are in the direct communication

distance range of node A are the direct neighbor nodes of node A. The topology

information of the direct neighbor nodes sets the parent MAC and the parent IP to be

the MAC address and the IP address of node A respectively.

Line of sight: the communication link built with the direct neighbor nodes is the

line of sight.

Multi-hop path: the communication link between the indirect neighbor nodes is

multi-hop path.

Two-way link: If node A and node B can receive a signal from each other, then

the communication link between node A and node B is two-way link.

98
 One-way link: If only one node can receive the signal from the other node, the

communication link between the two nodes is one-way link.

All nodes maintain their own network topology table. The nodes obtain each path

to communication with other nodes, and finish the selection of their optimized path and

the division of the alternate path by combining the path selection algorithm programs.

Finally, the settings of the nodes routing table are finished. All node information in this

protocol is operated and maintained by this network topology table.

The operations of this protocol are finished by the cooperation of seven types of

messages, namely, “hello”, “response”, “update”, “rsreq”, “reqack”, “Ack” and

“delete”. The hello message is a broadcast message, in which the nodes will first send

this message to the surrounding nodes as they start up. During the operations of the

protocol, the nodes will regularly broadcast the hello message to the surrounding nodes

to maintain their own network topology information. The surrounding nodes will

response to the response message when they receive the hello message to determine the

direct neighbor nodes to form the line of sight. When one node has a new path, it will

send the updated message to the neighbor nodes directly so that the neighbor nodes can

update the network topology table in time. In order to guarantee the efficiency of the

node information, this protocol sets a detection timer to regularly maintain each node.

Once an expired node is found, the “delete” message will be sent to the neighbor nodes.

The nodes detect their own network topology table after receiving the delete message

and delete the information of the expired nodes. In order to prevent routing loops, an

“rsreq” path detection message and a “reqack” path confirmation message are designed.

The “ack” message is to finish the verification of the two-way link to ensure that the

protocol can construct all two-way linked networks.

5.2 Design of protocol message

As described in 5.1, this routing protocol is finished by the cooperation of five

types of messages, “hello”, “response”, “update”, “ack” and “delete”. The functions
99
of each message are exhibited in Table 5-2

Table 5-2 Function of protocol message

Hello message Claim self-existence by broadcasting to the surrounding nodes

Response Send the node network topology information to the nodes of

sending “hello” message

Update Send the updated information of the network topology to the direct

neighbor nodes

Delete Send the routing deletion information to the direct neighbor nodes

Ack Detect the message in the adjacent nodes to remove the one-way

link

5.2.1 Common head of protocol message

Seven types of message in this protocol have the similar format, which is called

the common head of the protocol message

0 7 15 31
Command（6、7） Version1 F Length（16）

Ipv4 address

MAC

MAC(cont’)

Source IPv4 Address（ip_src）

Destination IPv4 Address（ip_dest）

Figure 5-1 Public radical format of protocol messaging

Value definitions of each field command: This field indicates the type of message,

which occupies eight bits. The value of this field is 1, which means it is a “hello”

message. The rest may be deduced by analogy, when the values of the fields are

2,3,4,5,6 and 7, the fields are response message, update message, delete message, ack

message, rsreq message and reqack message.

Version: This field shows the protocol version of this field, which occupies 4 bits
100
and is designed for the upgrading of the protocol. In this thesis, the value of the first

version in this protocol is 1.

F: This field indicates the mark field. When the value of the command field is 5,

the value of this field is significant. If the value of the F field is 1, it means it needs the

other side to send the “ack” message for the confirmation. The value of the F field is 2,

which means it does not need confirmation.

Length: This field indicates the total length of the protocol message, including the

length of the protocol head, which occupies 2 bits.

Ipv4 address: This field indicates the ipv4 address of this node, which occupies 4

bits.

MAC: This field indicates the physical address of the node, which is actually the

MAC address of the NIC and occupies 6 bits.

5.2.2 Format of the “hello” message

The format of the “hello” message is the same as the public radical of the message.

It just needs to give the relative value to each field. The value of the command is 1, the

value of version is 1, the value of F is 1, the value of length is 16, and the IPv4 and

MAC fields fill the IP address and the MAC address of this node respectively, as

shown in Figure 5-2:

0 7 15 31

Command Version.1 F Version.（16）

Ipv4 address

MAC

MAC(cont’)

Figure 5-2 Format of “hello” message

101
 5.2.3 Format of “response” message

The “response” message is the response to the “hello” message. The node

receiving the “hello” message supplies the network topology information to the node

sending the “hello” message so that the node that sends the “hello” message can create

its own network topology table as soon as possible. The message format is shown in

Figure 5-3
0 7 15

Command（2） Version1 F Length（16）

Ipv4 address

MAC

MAC(Cont’)

n routing entries

5-3 Format of the “response” message

In the “response” message, the radical of the message command field value is 2.

There are “n” routing entries after the public radical. Each routing entry indicates the

path information from one node to one of its direct neighbor nodes. The format of each

routing entry is shown in Figure 5-4

0 7 15 31

Ipv4 address

Link_state Delay

Lost Quality

Figure 5-4 Routing entry format

The length of each routing entry is 12 bits, including IPv4 address, link state,

delay, package loss rate and signal strength. The definition of each field is described as

follows:

IPv4 address (IP) occupies 4 bits and indicates the IP address of the direct

neighbor nodes.

Link state occupies 2 bits. It is used to express the link state between direct
102
neighbor nodes and the original node. This field can supply criteria for routing

selection in the routing algorithm. If the value is 0, it means that the link between the

nodes is broken. If the value is 1, it means that the link between the nodes is one-way

link. If the value is 2, it means that the link between the nodes is two-way link.

Delay occupies 2 bits. It shows the statistical average value of the communication

delay between the original node and the direct neighbor nodes, including processing

delay and transmission delay.

Package Loss Rate occupies 2 bits. It shows the average value of the

communication package loss rate between the original node and direct neighbor nodes.

Signal strength (quality) occupies 2 bits. It shows the signal strength of the direct

neighbor nodes detected by the original node at its location.

It is known from the definitions of the message format that the total length of the

message minus the length of the message public radical is equal to the length of the

entry part of the routing. Each routing entry occupies 12 bits. Thus, the number N of

the routing entry in each response message can be calculated with the following

formula:
len − 16
N=
12 (5-1)
Where len is the length of the whole message, 16 is the length of the public

radical of the protocol message and 12 is the length of one routing entry.

5.2.4 Format of update message

The “update” message is used to report the finding of a new path or the updating

of a path to the direct neighbor nodes. It is an important part of the path updating and

maintenance. Each “update” message contains a path condition, and each condition is

expressed by one routing entry. The “update” message has a length of 28 bits and the

value of command field is 3. The message format is shown in Figure 5-5.

103
0 7 15 31

Command(3) Version(1) F

IPV4(Address)

MAC (Address)
MAC Address

IPV4(Address)

Link_state C Delay C

Lost C Quality C

Figure 5-5 Formate of update message

5.2.5 Format of delete message

The “delete” message format is similar to the “response” message format. Both of

them contain public message radicals and “n” routing entries. Its command value is 4.

The detailed message format is shown in Figure 5-6

0 7 15

Command（2） Version1 F Length（16+12n）

Ipv4 address

MAC

MAC(Cont’)

n routing entries

Figure 5-6 Format of the “delete” message

The routing entry in the “delete” message is the routing entries that need to be

deleted. Most of them are expired node paths.

5.2.6 Format of “ack” message

The “ack” message is mainly used for protocol message confirmation between the

neighbor nodes. The neighbor nodes confirm a two-way link based on the “ack”
104
message. The length and the format of the message are similar to the “request”

message. But the value of the command field is 5. The format of the “ack” message is

shown in Figure 5-7

0 7 15 31

Command Version.1 F Length.（16）

Ipv4 address

MAC

MAC(cont’)

Figure 5-7 Format the of “ack” message

5.3 Path discovery

In this protocol, the path discovery mechanism includes positive path discovery

and negative path discovery. The positive path discovery means that the nodes send the

“hello” message to the neighbor nodes to require positive routing information. The

negative path discovery means nodes receive the “update” message from the neighbor

nodes, in which the path is constructed based on the “update” message. The positive

path discovery is the base of the negative path discovery.

5.3.1 Active path discovery

The active path discovery of nodes includes three steps: perception of new nodes

to the path, confirmation of link state, and perception of the neighbor nodes to the new

nodes. The network topology is shown in Figure 5-9. The distance between the

neighbor nodes is in the one-hop, but the distance that is between the non-neighbor

nodes is beyond the direct communication range. It is assumed that nodes B, C and D

can communicate with each other directly, and the network topology shown in Figure

5-8 has been formed. Setting node A as an instance, the path active discovery of the

protocol is described as follows:

105
 B
D
C
A

Figure 5-8 Nodes network topology

After the switch of node A’s hardware, node A first initializes self-hardware, then

it operates the content of this protocol. Node A first broadcasts a “hello” message to

the neighborhood the content of this message is shown in Figure 5-9. After sending the

“hello” message, node A is in the condition of loop waiting. After receiving the

protocol message from the other nodes, node A will operate the relative messages.
0 7 15 31

Command（1） Version1 F Length（16）

Ip A

Mac A

Mac A

Figure 5-9 Content of the “hello” message

Based on the node network topology in Figure 6-10, node B and node C will

receive the “hello” message sent from node A. According to the construction

regulation of the direct path, node B and node C will create the corresponding network

topology information based on the received “hello” message. The received “hello”

message can only confirm the one-way communication between the Link A-B and

Link A-C. The link state field is set as 1, the detailed generated network topology

information is shown in Figure 5-10.

Node MAC Node IP Parent node Parent node Packet Delay Signal Living Link
MAC IP loss rate strength time state
Mac A IP A Mac B IP B - - - 0 1

Mac A IP A Mac B IP B - - - 0 1

Figure 5-10 Topology information content of node B and node C

106
 The rules of one-way direct path constructed by the “hello” message are described

as follows:

The nodes MAC address in the radical of the MAC and of the “hello” message

The nodes IP address in the radical of the IP and “hello” message

Parent node MAC, filling in the MAC address of this node

Parent node IP, filling in the IP address of this node

Package loss rate, because there is no two-way link confirmed. The measured data

is not generated.

Delay, because there is no two-way link confirmed, the measured data is not

generated.

Signal strength, because there is no confirmed two-way link, the measured data is

not generated.

Life time, because the path is just constructed, 0 is filled in.

Link state, because the “hello” message is just received, 1 is filled in to express

the one-way link.

After node B and node C generate their own one-way topology information table,

node B and node C will check self-network topology information tables, and extract

and package the corresponding node information to the response message, which will

be sent to node A. The routing entry of the response message must be a two-way direct

path. The response message sent from the node B and node C to node A are shown in

Figure 5-11 and figure 5-12.

107
 0 7 15 31

Command(0x02) 0x01 F Lenght（28）

Ip B

Mac B

Mac B

Ip C

link_state C Delay C

Lost C Quality C

Figure 5-11 Response message of node B

0 7 15 31

Command ( 0x 02 ) 0 x 01 F Length （ 16）

Ip C

Mac C

Mac C

Ip B

link_state= =2 Delay B

Lost B Quality B

Ip D

link_ state
= =2 Delay D

Lost D Quality D

Figure 5-12 “Response” message of node C

After node A receives the “response” message sent from node B and node C. The

network paths are constructed based on the message content and two direct paths and

three multi-hop paths are obtained, as shown in Figure 5-13

108
Node MAC Node IP Parent node Parent node Packet Delay Signal Living
MAC IP loss rate strength time Link state
Mac B IP B Mac A IP A Lost B Delay B Quality B 0 2

Mac C IP C Mac A IP A Lost C Delay C Quality C 0 2

Figure 5-13 Direct path information of node A

Rules of construction of the direct path:

Node MAC, the MAC address in the “response” message radical

Node IP, the IP address in the “response” message radical

Parent MAC, the MAC address filled in node A

Parent IP, the IP address filled in node A

Package loss rate, node A measures the statistical average value of package loss

rate of the communication between node A and the node related to the “response”

message.

Delay, node A measures the statistical average value of the delay of

communication between node A and the node related to the “response” message

Signal strength, node A measures the signal strength value of the node related to

the “response” message.

Life time, as the path is just constructed, 0 is filled in.

Link state, because the “response” message is just received, 2 is filled in to

express the two-way link.

Meanwhile, according to the “response” message entry message, node A can

construct three multi-hop paths. The content of network topology information content

is shown in Figure 5-14.

109
 parent node parent node packet Signal
NodeMAC NodeIP loss rate Delay Living time link state
MAC IP strength

Mac B IP B Mac A IP A Lost B Delay B Quality B 0 2

p
Mac C IP C Mac A IP A Lost C Delay C Quality C 0 2 o
h
it
lu
Mac C IP C Mac B IP B Lost C Delay C Quality C 0 2 M

Mac B IP B Mac C IP C Lost B Delay B Quality B 0 2

Mac D IP D Mac C IP C Lost B Delay B Quality B 0 2

Figure5-14 Network topology information of node A

The construction rules of the multi-hop paths are described as follows:

Node mac, mac is filled in 00-00-00-00-00-00-00 because it is not the direct path.

Node IP, extracted from the routing entry information of the response message.

Parent mac, the mac address of the response message radical

Parent IP, the IP address of the response message radical

Package loss rate, the nodes related to the response message radical measure the

statistical average value of the package loss rate of the communication between the

nodes and the relative routing entry nodes delay, the nodes related to the response

message radical measure the statistical average value of the delay of communication

between the nodes and relative routing entry nodes

Signal strength, the nodes related to the response message radical measure the

signal strength of relative routing entry nodes.

Living time, because the path is just constructed, 0 is filled in

Link state, the response message is promised to merely send the two-way path

information, thus 2 is filled in.

B
D
C
A

Figure 5-15 Network topology structure of node A

110
 Node A receives the “response” message by sending a “hello” message to the

neighbor nodes. Based on the detailed content of the “response” message, node A can

obtain the network topology structure of the neighbor nodes, and modify the network

topology structure information table of node A to create the communication path with

the neighbor nodes. The network topology structure of node A is shown in the dotted

line in Figure 5-15.

After node B and the node C receive the “hello” message from node A, they can

only confirm the link A-B and link A-C as one-way paths, but not as the complete

communication path between nodes B, C and node A. In order to modify the network

topology information table of node B and node C, after node A finishes the operation

of the “response” message, it needs to feedback the “ack” path confirmation message

to node B and node C so that the direct neighbor nodes can confirm the creation of

two-way path. As node A has already confirmed the communication link as a two-way

link, it just needs to send the “ack” link confirmation message to the direct neighbor

nodes. The value of the field F of the message radical is 2, which means that the direct

neighbor nodes do not need to reply to the “ack” message for link confirmation. The

detailed “ack” message content sent from node A is shown in Figure 5-16.
0 7 15 31

Command（5） Version(1) 2 Length（16）

Ip A

Mac A

Mac A

Figure 5-16 Content of “ack” message

After receiving the “ack” link confirmation message, node B and node C will

correct the link state field of the information related to network information topology

table of the nodes, and correct 1 to 2 expressing the two-way link. They will also call

the package loss rate, delay and signal strength of the relative link obtained from the

111
link measurement program. The content of the corrected network topology information

table is shown in Figure 5-17.

parent node parent node packet Signal

NodeMAC NodeIP loss rate Delay Living time link state
MAC IP strength

Mac A IP A Mac B IP B Lost A Delay A Quality A 0 2

Mac A IP A Mac C IP C Lost A Delay A Quality A 0 2

Figure 5-17 Network topology information of the node B and the node C

After the “ack” link confirms the completion of message operations, the network

topology information tables of node A, node B and node C are modified, and the

communicated network topology between nodes is created. However, node D does not

update its own network topology information table, thus it does not have the multi-hop

path information to node A. This routing protocol also designs an updated routing

renewing message, by which node D modifies the topology information table of the

nodes. The update process of the network topology information table of node D is

introduced as follows:

5.3.2 Process of negative path discovery

Based on the network topology shown in Figure 5-8, after node C finishes the

“ack” link confirmation message and modifies its own network topology information,

it will send the updated routing message to the neighbor nodes. As the constructed C-D

link cannot be confirmed as a two-way link, the value of link state of the network

topology information will be corrected to 1 when node C sends the “update” message

to node D in order to confirm the C-D link state again. The content of the “update”

routing message sent from node C to node D is shown in Figure 5-18.

112
 0 7 15 31

Command（3） Version1 F Length（28）

Ip C

Mac C

Mac C

Ip A

link_state=2 Delay A

Lost A Quality A

Figure 5-18 Content of the “update” message

After node D receives the update routing message, it will create a D-C-A

multi-hop path from node D to node A according to the message content and to the

multi-hop path construction rules. Because the link state of path C-D needs to be

updated, the value of the link state of the created network topology information table

will be set as 1. The content of network topology information table created by node D

is shown in Figure 5-19

parent node parent node packet Signal

NodeMAC NodeIP loss rate Delay Living time link state
MAC IP strength
Mac C IP C Mac D IP D Lost C Delay C Quality C 0 1

Mac A IP A Mac C IP C Lost A Delay A Quality A 0 1

Figure 5-19 Network topology information of node D

After node D updates the path topology information based on the “update”

message, it needs to send the “ack” link state request confirmation message to confirm

the D-C path as a two-way link. Because node D needs node C to feed back the “ack”

link confirmation message, the value of F field is set as 1. The content of the “ack”

message sent is shown in Figure 5-20.

113
 0 7 15 31

Command（5） Version1 1 Length（16）

Ip D

Mac D

Mac D

Figure 5-20 Content of “ack” request message sent by node D

After node C receives the ack link state request message from node D, it modifies

its network topology information table first, and corrects the value of the link state of

the relative routing information from 1 to 2. It also marks the relative path as a

two-way available path. After modification of the information, node C will feed back

the ack link state confirmation message to node D. The value of the F field in the

message is set as 2. The detailed content is shown in Figure 5-21.

0 7 15 31

Command（5） Version1 2 Length（16）

Ip C

Mac C

Mac C

Figure5-21Ack confirmation message sent from node C

After node D receives the ack link state confirmation message, it will correct the

value of the path state of the network topology information based on the message

content, and set the value of link state of relative D-C link and D-C-A link as 2 to

express the relative path as the two-way path. The corrected topology information is

shown in Figure 5-22.

114
Node MAC Node IP Parent node Parent node Packet Delay Signal Living
MAC IP loss rate strength time Link state
Mac C IP C Mac D IP D Lost C Delay C Quality 0 2

Mac A IP A Mac C IP C Lost A Delay A Quality A 0 2

Figure5-22 Topology information content of node D

After the update path message is completed and the network topology information

table of node D is modified by the ack link confirmation message, the node topology

shown in Figure 3-10 obtains updated communicated topology, in which there is

accessible topology path information between each node and the path discovery is

completed. The final network topology is shown in Figure 5-23.

B
D
C
A

Figure 5-23 Network topology of nodes

5.4 Path maintenance

This routing protocol is based on the table-driven routing protocol. Each node

needs to regularly maintain self-routing table information. This routing protocol

implements the work of nodes routing table maintenance, mainly by setting the node

timing, updating the timer and the node inquiring timer. The nodes update the timer

mainly through the real-time effectiveness of self-node maintenance, and by regularly

claiming the existence of the self-nodes. The nodes inquiring timer is used to regularly

inquire the content of the network topology information table, and timely delete the

expired invalid node information.

115
 5.4.1 Regular updating strategy of routing

In this ADHOC network, the nodes can randomly move. The network topology is

characterized by dynamic change. The network topology information created between

the nodes can be changed at any time. The routing protocol needs to, on a timely basis,

determine the effectiveness of each link to guarantee the feasibility of network

communication.

This routing protocol implements the update of routing information by designing

the regular updating timing of nodes to guarantee the path effectiveness between the

nodes. Each communication node has a regular updating timer. Once the prescribed

amount of time elapses, the node will regularly broadcast the “hello” message to the

surrounding nodes to claim the existence of the nodes. Due to different application

conditions of the ADHOC networks, the external influences on the nodes are different,

and the moving conditions of the nodes are different. The detailed interval is set

according to the specific circumstances.

When the time of the timer comes, the nodes will not only send the “hello”

message, but also update the value of the lifetime field of each path in the network

topology information table of the nodes. The nodes will also set regular intervals to

achieve the objective of recording the lifetime of a path. When receiving the “hello”

message, the surrounding nodes will timely correct the relative path lifetime of the path

information, which is reset to 0, based on the content of the “hello” message, and mark

the effectiveness of the link.

5.4.2 Routing deletion strategy

In order to timely update the invalid routing information of the nodes in the node

network topology information table, an inquiry timer is designed in this protocol. The

nodes regularly travel through the network topology information table to check the

recorded value of the path lifetime field, and to judge if the lifetime of each path
116
exceeds the set maximum lifetime. If the lifetime of each path exceeds the set

maximum lifetime, the path is judged as an expired invalid path, and the corresponding

node is the invalid node. This node will delete the corresponding routing information

in the network topology information, and package the invalid nodes into the “delete”

message sent to the surrounding direct neighbor nodes. The surrounding direct

neighbor nodes will timely update the network topology information table according to

the “delete” message to remove the invalid routing information contained by the

expired nodes.

Node D in Figure 5-24 is assumed to be an invalid node, and the deletion process

of the path is described. If node D does not have failing node, the set request timer will

detect the maximum value that the C-D path lifetime exceeds the set value in the

network topology information table after a while. Meanwhile, node C judges node D as

an invalid one and deletes the routing information related to node D in the network

topology information table of self-node. The network topology information of node C

is shown in Figure 5-24.

Node MAC Node IP Parent node Parent node Packet Delay Signal Living time
MAC IP loss rate strength Link state
Mac A IP A Mac C IP C Lost B Delay B Quality B TTL 2

Mac B IP B Mac C IP C Lost A Delay A Quality A TTL 2

Mac B IP B Mac A IP A Lost B Delay B Quality A TTL 2

Mac A IP A Mac B IP B Lost A Delay A Quality A TTL 2

Figure 5-24 Network topology information of node C

After deleting the information of the invalid node D, node C will package node D

into the “delete” message and send it to the direct neighbor nodes. The state value of

the link in the routing entry contained in the “delete” message is set as 0, which

indicates an invalid link. The content of the “delete” message is shown in Figure 5-25.

117
0 7 15 31

Command(4) Version(1) F Length(28)

IP C

MAC C
MAC C

IP D

Link_state =2 Delay D

Lost D Quality D

Figure5-25 Deleted message sent by the nodes

After node A and node B receive the “delete” message, the lifetime of path A-C

and path B-C will be set to 0 according to the message content. The information of the

invalid nodes in the nodes network topology information table, which is the recorded

information of node D, will be deleted. After the deletion, the topology information of

node A and node B are shown in Figure 5-26 and Figure 5-27.
parent node parent node packet Signal
NodeMAC NodeIP loss rate Delay Living time link state
MAC IP strength
Mac B IP B Mac A IP A Lost B Delay B Quality B TTL 2

Mac C IP C Mac A IP A Lost C Delay C Quality C TTL 2

Mac C IP C Mac B IP B Lost C Delay C Quality C TTL 2

Mac B IP B Mac C IP C Lost B Delay B Quality B TTL 2

Figure 5-26Network topology information of node A

Node MAC Node IP Parent node Parent node Packet Delay Signal Living time
MAC IP loss rate strength Link state
Mac A IP A Mac B IP B Lost A Delay A Quality A TTL 2

Mac C IP C Mac B IP B Lost C Delay C Quality C TTL 2

Mac C IP C Mac A IP A Lost C Delay C Quality C TTL 2

Mac A IP A Mac C IP C Lost A Delay A Quality A TTL 2

Figure 5-27 Network topology information of node B

118
 After node A, node B and node C complete the update of the network topology

information table, they will delete the expired path information of node D. The

network topology structure is shown in Figure 5-28.

B
D
C
A

Figure 5-28 Network topology structure

5.5 Finite-state machine mode of protocol

Many protocol modes always involve the Finite-state machine (FSM) [137],

which makes an accessible analysis by the state transmission and detects all kinds of

mistakes in the protocols and the standards. The FSM of the protocol is shown in

Figure 5-30. From the view of the form, the FSM mode of this protocol can be

regarded as a triad (S,M,T).

In this triad,

S indicates all possible states set in the protocol;

M indicates the setting initializing the transmission of the protocol state;

T indicates the setting of the transmission between protocol states.

The event trigger mode is used in the design of this protocol. As a protocol event

comes, the protocol transfers from the free state to the events processing state. After

the processing is done, the protocol will transfer to a free state.

The events that initiate the state transmission (Figure 5-29).

Protocol initiation

Protocol exit

Regular sending of “hello” message

The path lifetime exceeds the maximal living time

Regular deletion of the link QoS parameter index

119
 Receiving the hello message

Receiving the response message

Receiving the update message

Receiving the delete message

Receiving the ack message

Figure 5-29 FSM mode of protocol

Protocol state set:

Over all, there are two states in this protocol: free waiting state and protocol event

processing state. In the free waiting state, the protocol machine does nothing and waits

for the coming events that initiate the protocol state transmission. The protocol event

processing state is a brief busy state of the protocol machine to process all kinds of the

protocol events. The detailed protocol events processing states are described as

follows:

Protocol initiation

Protocol message packaging

Protocol message sending

 Protocol message transferring

120
 New path construction

Path information updating

 deleting the path

Two-way link confirmation

Protocol resource releasing

The core of this protocol is the processing of the protocol message. The protocol

machine makes the relative processing based on the received message. It returns to the

waiting state after processing. The two timers use the interrupt mechanism. After the

interruption mechanism, the interruption processing is done, and then the protocol

retunes to the waiting state. From the FSM mode mentioned above, the protocol always

repeats wait for the mission-processing when the mission returns back to the waiting

state, and there is no deadlock in the protocol mode. Thus, the protocol has no mistakes

in flow-processing design.

5.6 Routing algorithm

5.6.1 Routing algorithm classification and design requirements

The routing algorithm can be classified into two types: self-adaptive routing

algorithm and non-self-adaptive routing algorithm [138]. The self-adaptive routing

algorithm can feel the change of the networks and change the routing strategy

according to the measurements and estimates. In the non-self-adaptive routing

algorithm, the routing information has already been downloaded to the local when the

networks are launched. When the network topology or other networks statistics

changes, the nodes won’t change the routing strategy according to the network change.

These two routing algorithms have their own advantages and disadvantages. The

self-adaptive routing algorithm can timely change the routing strategy according to

network change, but it is very complex to calculate with this algorithm and has a high

121
consumption. Although the flexibility of non-self-adaptive routing algorithm is low the

calculation is simple and the consumption will be low [139].

ADHOC networks are characterized by randomized moving of the nodes and a

dynamic change of network topology. The radio channel will constantly change, which

is influenced by the environment. Qos routing should timely select an optimized

network topology path based on the link condition in order to supply the reliable QoS

assurance. In order to implement the dynamic aesthesia to the change of the networks,

self-adaptive routing algorithm is used in this protocol, which can improve an effective

QoS routing assurance routing to ADHOC networks.

5.6.2MIA-QRP routing algorithm

ADHOC networks are mainly used for emergency communication and military

communications. Now, there are many routing protocols supporting the QoS, but most

of them use the node energy, with bandwidth, delay and some single QoS indicator as

the judgments. In this protocol the average delay of link, average package loss rate and

node signal strength are selected as the judgment standards to select the optimized

routing that can satisfy the special requirements.

1． QoS

(1) Delay

The delay is the time needed to transfer a message or a grouping from a node to

another in the network. It includes sending delay, broadcast delay, processing delay

and queuing delay (delay = sending delay + broadcast delay + processing delay +

queuing delay). The average delay index is regarded as one of the link judgment

indices [140] [141]. For the multi-hop ADHOC networks, the total average delay of the

link equals to the accumulation of the average delay of each single-hop link. The

average delay reflects the effectiveness of network transmission. The less the delay is,

the more effective it is. Especially for voice packages, too much delay will badly

influence the quality of communication.

122
 n
delay = ∑ delayi
i =1
(5-2)

(2) Package loss rate

Package loss rate (loss) is the proportion of lost data packages in the total sent

data packages during the link transmission. For the multi-hop ADHOC networks, the

total average package loss rate of the link is the product of average package loss rate of

each single-hop link. The package loss rate is related to the length of the data package

and the sending frequency it also reflects network reliability. The less the package loss

rate is, the more reliable it will be.

n
lose = ∏ losei (5-3)
i =1

(3) Signal strength index

The signal strength [142] indicates the signal strength of the nodes in the relative

surrounding points, which can reflect the communication quality of nodes and the

condition of residual energy, and is regarded as one of the QoS judgment indexes in

order to increase the stability of the routing.

1 n
strength = ∑ strengthi
n i =1 (5-4)

2. MIA-QRP judgment mode

In this routing protocol a Multiple Indicator Adaptive QoS Routing (MIA-QRP) is

proposed to improve the network service quality [143,144,145] by comprehensively

balancing delay, package loss rate and signal strength of the nodes, the MIA-QRP can

overcome the defects of a single consideration of the QoS index in the traditional

algorithm, and select the optimized path timely by a self-adaptive algorithm

determining the weight of each index with the entropy weight coefficient method.

For the original node S and the destination node D, the delay, package loss rate

and the signal strength of each link can be calculated by the formula (5-2), the formula

(5-3) and the formula (5-4) respectively. The set of the judgment indexes of the S and
123
 Rj R j 2 R j3
D, which is ={ R j1 , , | j =1,2, ……, n }, can be obtained by combining each
R j1 R j2
R j3
link. is average delay index, is link loss packets index, is the link signal
strength index. Due to magnitude and dimension problem of each index, the judgment

Rj
indexes matrix is standardized as follows:

 R jk − min R jk
 max R − min R For signal strength index
 jk jk
B jk = 
 maxR jk − R jk For average delay index and ink loss packets index
 maxR jk − min R jk


(5-5)

 R11 R12 R13   B11 B12 B13 

R =  R21 R22 R23  ⇒ B =  B21
 B22 B23  (5-6)
 R31 R32 R33   B31 B32 B33 

After the standardization, the value of the non-dimension and the magnitude is

from 0 to 1. Assume that the relative weight of each index is ω1 ≥ 0、ω 2 ≥ 0、ω 3 ≥ 0 ,

and ω1 + ω 2 + ω 3 = 1 , then the weighted assessment mode of each path is shown as

follows:

u j = ω1 H j1 + ω 2 H j 2 + ω 3 H j 3
(5-6)

H jk = − B jk log B jk , k = 1，
2，3
where
After the weight of each index is determined, the weighted assessment value of

each path can be calculated. Thus the path with the highest assessment value can be

selected.

3. Decision of weight

Based on the extreme property of entropy, when each system condition has the
1
same probabilities, which is Pj = , j = 1,2 ……, n , the maximum value of the entropy
n
is:

124
 1 1 1
H max = lg n = H  , ,…… 
n n n (5-7)

When the entropy of one index gets larger, the difference of the relative index

value between each link is less. Then the effects of the judgment index are less, and its

weight value is less. The weight value of the judgment index mentioned above can be

determined based on the characteristics of the entropy.

According to the standardized matrix B produced by judgment index matrix R, the

close degree mode of each judgment index can be determined:

D jk
d jk = n 3 (5-8)
∑∑ D jk
j =1 k =1

 B jk
 max{B } For signal strength index
 jk
D jk = 
 min{B jk } For average delay index and ink loss packets index
 B jk

(5-9)

According to the definition of entropy, the importance of each judgment index can

be measured by the following condition entropy:

n d jk  d jk 
H k = −∑ lg 
(5-10)
j =1 dk  dk 

n
d k = ∑ d jk , k = 1,2,3
j =1

The H k is normalized by the H max based on the extreme property of the

entropy to get the assessment importance entropy that characterizes the judgment

index:

Hk 1 n d jk  d jk 
hk =
H max
=− ∑ lg ，k = 1,2,3
lg n j =1 d k  d k  (5-11)

In order to comprehensively evaluate each index, the comprehensive weights of

the judgment indexes can be determined by hk :

125
 1 − hk 3
ωk = 3
0 ≤ ωk ≤ 1 ∑ω k =1
∑ (1 − h )
k =1
k
k =1

(5-12)

Thus, the judgment weights of the delay, package loss rate and the signal strength

ω1、ω 2、ω3 can be determined. Related to the path weighted assessment mode, the

nodes can select the path with the highest judgment to be the optimized path from the

original node S to the destination node D by calculating the judgment value of each

path.

5.7 Design and implementation of MIA-QRP protocol

5.7.1Embedded development condition

1. Hardware resources

Development mainframe: CPU Frequency 3.0GHz, Memory 1GB, Hard disk

512G

Target board: The friendly Arm mini development developed board, arm 9

Processor, 256M and flash

Wireless transceiver: 802.11xUSB Wireless network card, which uses the link

rt73 USB Wireless LAN chip and supports 802.11a, b, g protocol.

2. Software resource

Fedora 10.0 Linux Operating system, Linux 2.6.25-14 Kernel version

Vmware7.0; Vmware7.0 Virtual machine software

The target board operates the embedded Operating system mirror, Linux 2.6.32.2

Kernel version.

Rt73 Wireless network card chip driver source code

126
 5.7.2 Design and transplantation of wireless NIC driver

As shown in Figure 5-30, based on the TCP/IP network layer mode, the system

structure of the Linux networks driver can be divided into 4 layers [146], namely

network protocol interface layer, network device interface layer, device driven function

layer and network device and media layer.

Figure 5-30 Architecture of network device driver

The network device and media layer is in the lowest layer of the network device

driver structure. It supplies directly the driver to access the physical network device.

The function of device driven function layer is to improve the operator interface to

access the outside device for the upper program, and to implement the driver of the

device. It starts the sending operation by the function of hard_start_xmit(), and

implements the operations by the interrupt trigger in the network device. The network

device interface layer improves an unified structure net_device, which is used to

describe the characteristics and operations of detailed network device to the network

protocol interface layer. This structure is the container of each function in the networks

protocol interface layer. The network device interface layer is responsible for macro

plan of device driven function layer of the hardware. The network protocol interface

layer improves an unified sending and receiving interface to the networks layer

protocol without considering the upper protocol. The dev_queue_xmit() function is

used to send the data, the netif_rx() function is used to receive the data. The design of
127
this layer makes it is unnecessary to consider the detailed implementation of the

protocol for the last layer of the protocol.

5.7.3 Drive function and main data structure of network device

1. Drive function of network device

The main functions achieved by the network device driver include system

initialization, data package sending and receiving etc. The system initialization is

implemented by defining the initialization function init() in the data structure

net_device. The initialization function init() will detect the hardware information of the

device to determine whether to arrange the resource spaces to this device or not. After

deletion, the initialization function will fill the information of this device into the data

structure net_device and add this new device to the new device list.

The data package sending and receiving is the most important content in the

design of the network device driver. Whether the programming of this driver is good or

not will directly influence the overall quality of the driver. The sending of network

device needs to use the dev_queue_xmit() function in the protocol interface layer, by

which the pointer of the hard_header function in the data structure net_device is

adjusted, the data in the socket buffer saved in the sk_buff will be read, and the data

will be sent to the physical device. The receiving of the data package is always a

netif_rx() function. When driver programs of lower layer hardware receive the data

package, the hardware will create an interrupt signal. Then, the data in the socket

buffer sk_buff will be queued in the upper protocol queue by the netif_rx_schedule

function for the convenience of the following processing.

2. Main data structure of the networks driver

All network devices are abstracted as an unified interface in the Linux, which is

the network device interface. It expresses the operation condition of the network device

in the kernel by the structure variables of the structure net_device type, including

128
loopback devices and hardware network device interface. The kernel administrates the

entire network device from the device list in which the dev_base is the head pointer.

sk_buff structure is used to administrate the back T structure and plays an

important role in the sending and receiving of data package. The data of each network

layer in the Linux is sent by the sk_buff. In the network protocol interface layer, both

kernel functions the dev_queue_xmit() and the netif_rx() use the sk_buff as the

parameters.

3. Design and analysis of rt73 wireless network card driver

In the Linux, most of the drivers are based on the development approach of the

module, and they can be loaded and unloaded freely. The design of this network card

driver is developed according to the module, which sets the module_init as the driver

entrance and sets the module_exit as the driver exit.

The main tasks to be completed by the design of the network device driver

include network device registration, network device initialization, processing of data

sent and received etc. It is also required to make the immediate processing in case of

timeout and breakage of transmission.

(1) Initialization of the network device driver

The work of module_init is to register the rtusb driver to the kernel USB Drive

chain. The corresponding program is return usb_register(&rtusb_driver), where the

definition of the rtusb_driver is described as follows:

struct usb_driver rtusb_driver = {

#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,15)

.owner = THIS_MODULE,

#endif

.name="rt73",

.probe=usb_rtusb_probe,

.disconnect=usb_rtusb_disconnect,

.id_table=rtusb_usb_id,
129
 #ifdef CONFIG_PM

.suspend = rt73_suspend,

.resume = rt73_resume,

#endif

};

The usb_rtusb_probe function is the most important, and is mainly used for device

detection and initialization of the needed kernel resource. The rt73_suspend and the

rt73_resume are related to the power administration, such as how to work when it

receives the suspension, and how to recover at the resumption. The

usb_rtusb_disconnect will call the interface after the usb device is removed from the

system. The rtusb_usb_id is the usb devices ID lists driven by the rtusb_usb_id.

After the module initialization and registration to the system, the driver will scan

the system bus to check if the relative devices of this driver are on the bus. If there are

devices on the bus, the probe function will be called, or the device is inserted to the

system after the initialization of the driver.

According to the design of network device driver, most members in the

net_device structure need to be instantiated. The relative content of this network card

driver is defined as follows:

netdev->priv_flags = INT_MAIN;

netdev->open = usb_rtusb_open;

netdev->hard_start_xmit = usb_rtusb_sendpkt;

netdev->stop = usb_rtusb_close;

netdev->do_ioctl = rt73_ioctl;

(2).Sending of network data package

According to the instantiated code content of the net_device structure, the data

sending of the networks card is finished by calling the usb_rtusb_sendpkt, which will

call the RTMPSendPackages function to finish the sending of the data. The

dev->hard_start_xmit is a callback function, and its value is set as dsr_dev_start_xmit.

130
This function is not called by the drive layer, but it is called by the upper protocol. For

example, when the IP or ARP layer needs to message, it can decide use which network

device (net_dev) to transfer the data according to the destination IP address, and it will

call its hard_start_xmit callback function. The call and control of the TCP/IP protocol

stack will have been finished in the kernel, and they will use the callback function until

the data comes to the device. This is the principle of the driver implementation. The

network drive is the same as other drivers.

For the network card, the true data sending is finished by the USB. The

RTUSBKickBulkOut function will be called in the RTMP Send Package function to

finish the data sending.

(3) Receiving of the networks data package.

In the process of data package receiving, an interrupt will be generated when the

data packages arrive at the network device, and inform the driver to receive the data

packages. The interrupt processing program will apply for a period of addresses to the

system as the sk_buff buffer to save the received data. After the received data is saved

to the buffer, it is still needed to fill in the other parts information, such as the structure

of specified devices, data frame types and link layer data types. Finally, by using the

protocol interface layer function netif_rx() calling the RTUSBRxPackage(data)

function sending the data packages to the upper layer for processing.

There is 802.11 protocol stack processing in the RTUSBRxPackage function.

After the USB part receives the data, it will call the receiving function of network card

driver for simple protocol stack processing and copy the data to the upper data layer.

4. Compilation and transplantation of the wireless LAN driver

After the design of network card driver, the network card needs to be recompiled

in order to use the network card driver on the embedded Linux platform. This driver

uses the arm-linux-gcc-4.3.2 cross-compiler based on the mini2440 development board

for the compilation. The processing of compilation to the driver is described as

follows:
131
 Correct the compilation files

Enter “2010_0817_RT73_Linux_STA_v1.1.0.4/Moude” catalog and open the

“Makefile” file under the catalog. Make the following corrections:

Correct the platform settings, use the embedded platform, PLATFORM= CMPC

(2) Correct the compiler settings, use cross-compiler and link tools

CC :=arm-linux-gcc , LD :=arm-linux-ld

(3) Correct the kernel compiled head file catalog, LINUX_SRC =

/root/linux-2.6.32.2, this catalog is the saving path of the kernel head file.

2. Compilation

(1) Execute the command “make old config && make prepare && make

modules_prepare” to make preparation for the compilation of the kernel module.

(2) Enter the catalog “2010_0817_RT73_Linux_STA_v1.1.0.4/Moude”, and

execute “make -C /root/linux-2.6.32.2 M=`pwd` modules” to compile the networks

card driver. After the compilation, the driver kernel module files are generated finally:

rt73.bin, rt73.koand rt73sta.dat. The network card can be used on the destination board

if these files are downloaded and installed on the destination board.

5.8 Architecture of protocol software

There are two modes of design of embedded Linux based software, namely kernel

module design and user application software design. The kernel module development

program usually begins to run into the kernel state with the start of the operation

system. The development requirements of kernel program are high, which needs to

completely consider the calling mechanism of the kernel program. Designed for the

safety and convenience, this protocol applies the user application program so that the

users can design and administrate the program easily.

132
 5.8.1 Software layers model

According to the OSI design mode, layer modular design is also used in this

routing protocol to divide the protocol into four layers, each of which implements the

relative function. The layers call each other by the interface, and complete the function

of the routing protocol with help of each other. As shown in Figure 5-31, starting from

the top, this routing protocol software model can be divided into routing algorithm

execution layer, protocol event processing layer, protocol message analysis layer and

original network data package capture layer.

Figure 5-31 Model of protocol layers

1. Original data package capture layer.

This is the bottom layer in the protocol layer model, mainly responsible for

supplying the original data related with this protocol to the upper layer. The function of

this layer is implemented according to the function of the OSI based data link layer.
133
The captured original data is the MAC frame data of the link layer, which has the

transparent and reliable data characteristics. The data captured by this layer will be

saved to the buffer of the original data for the use of the upper layer. Meanwhile, in

order to obtain the QoS parameter index of the link, this layer implements the statistics

function for the link average package loss rate and the average delay and the detection

function of the nodes signal strength to integrate the captured data and the detected

data, which is regarded as the basic data processed by the upper layer routing.

2. Protocol message analysis layer.

This layer is mainly responsible for extracting the data related to this protocol

from the original data buffer saved from the original data package capture layer, and

saves the data to the protocol data buffer for the calling of the upper layer. As the data

captured from the lower layer are all Ethernet data frames captured from the data link

layer, in addition to the data message types of this protocol, there are still other

networks protocol data frame types, such as IP, ARP data, PPP data. Thus, the type

mark field data of the message in this protocol needs to be designed.

Based on the Ethernet format, the types of Ethernet are marked and distinguished

according to the value of Length/EtherType field. The different value indicates the

different types of the data frame. When the value of the Length/EtherType field is

more than or equal to the hexadecimal 0x0600, it indicates the MAC client computer.

For example, the value of 0x0800 means IP datagram; the value of 0x0806 means the

ARP datagram, and the value of 0x8863 indicates PPPoE protocol datagram. In order

to mark the datagram type of this protocol, the value of the Length/EtherType field in

the Ethernet is set as 0x6c6c, this layer just needs to analyze the data frame with the

value of the type marked field, which equals to 0x6c6c, in the original networks data

buffer. The other types of data frame will be abandoned directly.

3. Protocol event processing layer.

This layer is the core of this routing protocol. It creates and maintains the network

topology path table, generates and updates the path information, and supplies the
134
foundation for the routing algorithm. This layer consists of two modules: protocol

message processing and timing clock event processing. The protocol

message-processing module is responsible for confirming the communication link,

path findings, path maintenance, and inquiry and confirming path state. The clock

provides basic timing for the protocol, and assists the nodes to accomplish the regular

message request, position update service and maintenance of life cycle. This layer

consists of four parts: update timer module, deletion timer module, path findings

module and path maintenance module, which work together to accomplish the

processing of the protocol message and calculate the optimized path to the destination

nodes.

4. Routing algorithm execution layer.

The assignment of this layer is to calculate the optimized path to the destination

node according to the content calling routing algorithm in the network topology

information table. Meanwhile, it is responsible for maintaining the system kernel

routing table and adding the optimized path to the system kernel routing table and

deleting the expired path.

5.8.2 Threading model

Based on the protocol layers model, this protocol applies the technologies of the

multi-threading which designs three threading and three timers to accomplish the basic

work of the protocol. The three basic threading models are the original network data

package capture threading, the protocol data package analysis threading and the

protocol message processing threading. These three threads keep working with the start

of the protocol and will be destroyed until this protocol quits.

In this protocol, the method of creating the temporary threading is used to

implement the function of three basic timers. When the time of the timer comes, a

temporary threading will be created to implement the corresponding assignment. This

threading is destroyed after the assignment is accomplished. The creation of the timer
135
starts with the beginning of the protocol, and the timer will stop until the protocol

quits.

The routing selection algorithm of this protocol is designed as a call function. The

part processing assignments of the protocol message processing threading is called in

need to save the cost and improve the performance. The model of the protocol

threading is shown in Figure 5-32.

Figure 5-32 The model of the protocol threading

5.8.3 Data stream of protocol

According to the protocol layers design model, the data stream of this protocol is

shown in Figure 5-33.

Figure 5-33 Data stream of protocol

136
 The data of this protocol is captured and obtained according to the networks data

stream of the data link layer. The obtained data frame is saved to the protocol original

data buffer by the processing of the networks original data package capture layer. The

protocol data analysis layer will extract the data of this protocol from the original data

buffer to save the data to the protocol message buffer queue. The protocol

message-processing layer will make the further process to the protocol message buffer

queue to generate the network topology path information table. Finally, the protocol

message processing threading selects the optimized path of the network topology by

calling the routing algorithm, and the selected optimized path is added to the system

kernel routing table.

5.8.4 Design of the key module and Process flow

1. Synchronization of multi-threading

In the process of protocol data stream, the networks original data capture

threading will timely extract the networks data frame and save it to the networks

original data buffer. Meanwhile, the protocol message analysis threading will timely

extract the data in the networks original data buffer to extract and filter this protocol

message. The protocol threading makes a shared access to the data in the networks

original data buffer. If there is no relative threading synchronization, the unforeseen

data accessing errors will be generated. There is a shared access to the protocol

message buffer queue between the protocol message analysis threading and the

protocol message processing threading.

The Linux provides several process methods for the synchronization of

multi-threading, but the usual methods are mutual exclusion, condition variable and

semaphore. This protocol implements the data synchronization of the multi-threading

by the method that combines the mutual exclusion and the condition variable, and

implements the gradual visit of the multi-threading to the shared resource by using the

mutual exclusion. This protocol also combines a condition variable to control the
137
access and the dormancy of threading in case of multi-threading access blocked caused

by only using the mutual exclusion, and increases the resource utilization and the code

execution efficiency.

The mutual exclusion and the condition variable used in this routing protocol are

defined as follows:

pthread_mutex_t raw_lock;

pthread_mutex_t pro_lock;

pthread_mutex_t topo_lock;

prthread_cond_t cond_raw;

prthread_cond_t cond_pro;

Where the mutual exclusion raw_lock and the condition variable cond_raw are

used for synchronization shared access of the network original capture threading and

the protocol message analysis threading to the networks original data buffer. The

mutual exclusion pro_lock and the condition variable cond_pro are used for the

synchronization shared access of the protocol message analysis threading and the

protocol message processing threading to the protocol message buffer queue.

Where the mutual exclusion and the condition variable are used, mutual exclusion

locked and unlocked function, condition dormancy and the wake-up function are

described as follows:

pthread_mutex_lock(pthread_mutex_t *mutex), the mutual exclusion locked

function

pthread_mutex_lock(pthread_mutex_t *mutex), the the mutual exclusion

unlocked function

pthread_cond_wait(pthread_cond_t *cond,pthread_mutex_t *mutexz, Consition

waiting, which leads the threading to the freedom waiting state

pthread_cond_broadcast(pthread_cond_t *cond), condition activation, which

wakes up the free waiting threading

138
 The threading synchronization flow used by the mutual exclusion and the

condition variable is shown in Figure 5-34.

Figure 5-34 Synchronization flow chart of multi-threading

2. Data buffer queue

Because this protocol needs to timely receive the networks data package and

processes the protocol message, it is needed to design a data buffer for this protocol in

case of loss of data received and processed. The design can effectively process the

saved data in case of the data lose.

The data buffer of this protocol, which applies to the loop queue, is implemented

by the structure. The saving unit of the loop queue consists of saving data and data

length. The detailed data structures are defined as follows:

typedef struct raw_package_unit{

unsigned char data[1518];

unsigned short len;

}raw_unit,*praw_unit;

3. The sending and receiving of the data package

All original data of this protocol is captured by the data link layer. In the

condition of Linux, the libpcap data package and the creation of the socket are used to

implement the capture of the networks data frame, where the libpacap is the open
139
source implementation code of the network’s programming on the Linux stage [147].

Based on the practical experience of the author to the Linux networks programming,

this protocol is implemented by applying the creation of socket under the Linux. The

direct access of the data in the networks data link layer is implemented by creating the

networks original socket to capture the networks original data package.

The function prototype of the Linux networks socket interface is the int socket (int

domain, int type, int protocol. In order to directly access the data in the data link layer,

the parameters are set as follows:

Domain is socket field, and is set as PF_PACKAGE

“Type” is the type of the socket. The usual types are the stream socket

(SOCK_STREAM), the data gram socket (SOCK_DGRAM) and the original socket

(SOCK_PACKAGE. The socket is set as SOCK_PACKAGE in this protocol.

The protocol is the socket protocol, and is set as htons (0x0003) in this protocol.

The data in this protocol capturing the threading needs to timely catch the

networks data, and set the networks socket as the non-blocked state. Then the function

fcntl (s, F_SETFL, O_NONBLOCK) is called to finish the settings.

After the creation and settings of original socket, the function send to() and

recvfrom() is used to implement the access the data in the data link layer. The

prototype of the function interface is described as follows:

sendto(int sock, const void *buf, size_t len, int flags, const struct sockaddr *to,

socklen_t tolen);

int recvfrom(int sockfd,void* buf,int len,unsigned int flagsstruct sockaddr*

from,int* fromlen).

140
 Figure 5-35 Flow chart of protocol message processing

4. Processing flow of protocol message

The path findings and maintenance in this protocol are implemented by the

effective processing of the protocol event process layer to this protocol message. The

processing protocol message is the key to process the data in this protocol. The

message processing flow in this protocol is shown in Figure 5-36.

According to the protocol message processing flow chart, this routing protocol

extracts the protocol message one by one from the protocol message buffer queue first.

Then the protocol runs the switch branching processing based on the type of the

message, and calls different message processing programs to implement the message

processing.

The protocol message processing designed in the protocol has a pretreatment.

Each type of the message processing branches firstly constructs a piece of network

topology information by the pretreatment of the message. Then the network topology

table is inquired to detect if there is the same network topology information. If no, a
141
new piece of topology information will be added in the network topology information,

otherwise, the content of the network topology information table is updated according

to the message content. The protocol message pretreatment procedures are shown in

Figure 5-36.

Figure 5-36 Message pretreatment procedures

In this protocol, an optimized path selection algorithm is designed. According to

the protocol threading model, the routing algorithm will call in the message processing

as needed

5. Protocol timer

In this protocol, three modules are designed which accomplish the periodical

sending of node broadcast message, expired path information detection and link

parameter indexes detection respectively. The function of each timer has been

described in the last section. Now the operations of the timers are described as follows:

Linux arranges three internal timers for each assignment [148]. When the time

comes, the ITIMER_REAL real-time timer will send the SIGALRM signal to the

process, ITIMER_VIRTUAL timer will send the SIGVTALRM signal to the process

and the ITIMER_PROF timer will send the SIGPROF signal to the process. An initial

142
value is given to the timer, which decreases with time and sends the signal to the

process until it decreases to 0, and then it will return to its initial value. The

ITIMER_REAL real-time timer is used for the implementation of this protocol, which

can timely send the SIGALRM signal to the process and run the relative functions.

Linux uses timers, which need the initialization works and relative signal

processing mechanism. In the design of this protocol, the initialization and signal

processing mechanism of the timers are accomplished by calling the corresponding

function. The prototype of each function is:

int setitimer(int which, const struct itimerval *value, struct itimerval *ovalue)

The set timer function is used to accomplish the timer initialization. Each

parameter is set as follows:

Which is the type of the timer, which is set as ITIMER_REAL

Interval value of the timer. The detailed interval of the timer can be set by this

structure, which can be accurate to a microsecond.

ovalue is used to save the previous value. It is not used usually, and is set as the

NULL here.

int sigaction(int signum,const struct sigaction *act ,struct sigaction *oldact)

The sigaction function is used to create the signal processing mechanism of the

timers. Each parameter is set as follows:

signum is the signal type, which is set as the SIGALRM signal type sent from the

ITIMER_REAL timer act is the structure type. The call function of the SIGALRM

signal processing can be set by the member sa_handler of the structure.

The object pointed by the old act is used to save the processing to the old

corresponding signal, which is set as NULL.

In the design of the timer, the Linux system calls the timer signal to execute the

function by the method of creating the temporary threading, and accomplishes the

operation assignments of the timer. After completion of assignment, the threading

resource is released.
143
 6. Design of routing algorithm program.

According to the protocol-threading model designed in this protocol, the routing

algorithm of this protocol runs the calling the execution by the protocol message

processing the threading to decrease the recourse cost largely, and improve the

executive efficiency of the program. The core idea of this protocol algorithm design is

to calculate the optimized path from the original node to the destination node according

to the path information in the network topology information table and the MIA-QRP

algorithm theory, and to improve a reliable QoS security service to the AD HOC

networks.

In the program of protocol message processing, after execution of the “response”

message and “ack” message, the function executed by the routing algorithm will be

called. The execution procedures of routing algorithm is shown in Figure5-37.

Figure 5-37 Execution procedures of routing algorithm program

When the routing algorithm program is called, it will first read the path

information between the original nodes to the destination node from the network

topology information table in order to detect if there is only one accessible path. If

there is only one path, then it will be directly added to the system routing table,
144
otherwise all the accessible information will be collected from the topology

information table to calculate the optimized path in all accessible paths by calling

MIA-QRP path selection algorithm, and to update the system routing table and ARP

table. When all the path information of the nodes is processed, the assignment is over

and will quit.

7. System routing table and ARP table operation

The network communication routing selection of Linux system is implemented by

the inquiry operation of system routing table. The path information obtained from this

protocol is applied by the other networks program, and is added to the system routing

table. Thus, the system() function is further called to update and maintain the system

routing path.

The operation of using system() function to the system routing table[148] is

described as follows:

For example, add it to 192.168.1.170 direct routing

system (’route add –host 192.168.1.170’)

add it to 192.168.1.171 routing and forwarded by the 192.168.1.170

system (’route add –host 192.168.1.171 gw 192.168.1.171’)

delete 192.168.1.170 direct routing to the nodes

system (’route del –host 192.168.1.170’)

delete the 192.168.1.171 routing and forward it by 192.168.1.170

system (’route del –host 192.168.1.171 gw 192.168.1.171’)

In the Linux system, the kernel routing table accomplishes the routing function

with the connection of the ARP table. The corresponding relation between the IP

address and the hardware address can be inquired by the ARP table. The ARP table is

also used in the networks address analysis. The ARP table information should be

updated when the routing table is added. The operation of the system ARP table is

described as follows:

Assume that the IP address of the direct routing node is 192.168.1.170, its MAC
145
address is D8:5D:4C:7C:13:DF

system (’arp –s 192.168.1.170 D8:5D:4C:7C:13:DF’)

system(’arp –d 192.168.1.170 D8:5D:4C:7C:13:DF’)

If the IP address of the node is 192.168.1.171, the next hop from this node to the

next destination node is 192.168.1.170

add the corresponding ARP table: system(’arp –s 192.168.1.171

D8:5D:4C:7C:13:DF)

Delete the corresponding ARP table: system(’arp –d 192.168.1.171

D8:5D:4C:7C:13:DF’)

5.9 Summary

According to the multiple QoS parameters including delay, loss of package and

quality, a new routing algorithm is proposed. This routing protocol is implemented by

five kinds of protocol messages, including network route discovery and maintenance,

and through the programming, it is to realize link parameters acquisition, selection of

optimal path and update of system routing table. In the development of protocol

implementation, wireless network card driver is analyzed. According to the design

principle of the protocol, the software architecture of the protocol is given in this

chapter. The protocol is designed in detail.

146
 Chapter 6 SMAC-SD Protocol

6.1 The improved SMAC protocol

Route protocol need to be supported by MAC,S-MAC protocol can guarantee the

nodes to be periodical monitoring/sleeping mechanism to reduce the energy cost.

However, in order to satisfy differentiated Services in WSN, S-MAC protocol should

be improved .The thesis proposed an improved protocol: SMAC-SD.

The main principles of the SMAC-SD protocol are: it follows the periodical

monitoring/sleeping mechanism of S-MAC protocol, the message transmission

mechanism and the selection and maintenance scheduling, and the time

synchronizations and NAV mechanism. Its improvements include principle of the

differentiated services; priority mechanism based on service queue; channel

multi-requirements mechanism.

6.2 Mechanisms of SMAC-SD Protocol

1. Secure transport mechanism of the service differentiation.

(1). Principle of the differentiated Services

The standard of the differentiated services is related to the traffic of application

lay.

From the point of view of network QoS, how the data is delivered to the sink and

corresponding requirements are concerned. Generally, there are three basic data

delivery models, i. e. , event-driven, query-driven, and continuous delivery models

[149]. Before presenting the application requirements, some factors that characterize

are provided as follows:

Real-time voice, image, or video data: Real-time data is delay-constrained and has

a certain bandwidth requirement. Packet losses can be tolerated to a certain extent.

Non-real-time data: The sink may want to collect periodic data from the sensor
147
field.

In this project, information of time-sensitive will be taken in consideration.

This protocol defines two transaction classes: important transaction and general

transaction.

Important Transaction: low tolerance to the delay and high priority in the

networks transmission. It has a higher priority access channel queue mechanism.

General Transaction: high tolerance to the delay, and low priority in the networks

transmission.

For the different WSN, the users can define which transaction is the important

transaction depending on the actual demand of using.

(2) Probability priority mechanism based on service queue

SMAC-SD defines two queues of access category (AC), and each queue has

different priorities of the data transmission.

Assume:

The minimal backoff windows are

CWmin ; maximal backoff window is CWmax

and the Arbitration Interframe Space is (AIFS).

CWmin [1], CWmax [1], and AIFS [1] are for the high priority access class;

CWmin [0], CWmax [0] and AIFS [0] are for the low priority access class

where CWmax [0] ≥ CWmax [ 1]，CWmax [0] ≥ CWmax [ 1] ，AIFS[0] ≥AIFS[ 1].

Two transmission queues are arrayed in the base station. Every sequence supports

an access classification, and its behavior is similar to DCF mechanism [145]. Every

queue runs a DCF function. The bigger backoff windows is, the lower the priorities of

access control is.

The access classifications are divided into two types. It is assumed that there are n

active base stations in the WSN and every base station has the complete two queues,

there are 2n queue object, which means there are 2n base stations, their channel

148
compete by IEEE802.11DCF, as figure 6-1.

SMAC SDSMAC

Access Category

Only one AC is AC is
kind queue high low

DIFS AIFS[AC_1] AIFS[AC_0]

CWmin=31 CWmin[AC_1] CWmin[AC_0]
Cwmax=1-23 Cwmax[AC_1] Cwmax[AC_0]

transmission transmission

Figure 6-1 Flow Chart of the Probability priority mechanism of the Channel

Access

3) Channel multi-requirements mechanism

SMAC-SD protocol uses the probability priority mechanism of the channel access,

which can guarantee the higher efficiency of accessed data to the channel compare

with the general transaction. However, the probability of accessing the channel is not

completely equal to the successful probability of accessing the channel. The successful

probability of accessing the channel for the important transaction is related to the data

stream of the general transaction. The successful probability of accessing the channel

for a priority transaction in the nodes is [150,151,152]:

Psi = nτ i (1 − ∑τ i ) n −1 ∏ (1 − τ i ) (6-1)
i∈m

Psi is the successful probability of accessing the channel of the transaction data

with the priority

i , n is the number of the nodes, τ i is the access probability of the

transaction with the priority

i , and m is the number the access transaction.

With the analysis, the general transaction will influence the important transaction
149
by the following aspects: the data packet of the general transaction collides with the

important transaction; generally, the channel occupied time after the successful sending

in the general transaction makes the delay to the channel of the data access in the

important transaction. This will lead the channel competition advantages of the data

stream of the important transaction to be reduced by the large existed data stream of

the general transaction under the intense condition in the networks.

According to the probability priority mechanism of channel access, the channel

multi-requirements mechanism improves the handshake mechanism of the queues with

high priority.

The access mechanism of all transactions in the SMAC protocol is

RTS-CTS-DATA-ACK. RTS-CTS means Request To Send/Clear To Send. Based on

the principle of the differentiated services. Access of channel supplies one more

passing RTS frame requirement. The mechanism is showed:

1. Important transaction: the interFrame space of the waiting queues with high

priority in the important transaction is adjusted as follows:

Contention
window
Important UIFS
access service RTS RTS DATA
SIFS DIFS Contention
window
Receive
point SIFS CTS SIFS ACK

DIFS

Collision RTS DIFS

Figure 6-2 Access and RTS/CTS mechanisms for Important Transaction

If a node requires sending RTS request frame is successful, the receivers will

immediately responds to agree to send the CTS frame, and the senders begin to send

the data after receiving the CTS.

If the first requirement is failed, then it will resend the RTS requirement with the

UIFS interval, UIFS=SIFS+2*slot time. If the requirement is successful at this time,

the receivers will responds CTS frame, and the senders begin to send the data after
150
receiving. If the requirement is failed again, it turns into the backoff process.

2. General transaction: the timing chart of the waiting queues with low priority

in the important transaction is adjusted as follows:

Contention
window
General access
RTS DATA
service
SIFS DIFS Contention
window

Receive point SIFS CTS SIFS ACK

DIFS

Collision RTS
DIFS

Figure 6-3 Access and RTS/CTS mechanisms for General Transaction

For the data stream of the general transaction with the low priority, Access and

RTS/CTS mechanisms is showed by figure 6-3.

If a node requires sending RTS request frame is successful, the receivers will

immediately responds to agree to send the CTS frame, and the senders begin to send

the data after receiving the CTS.

If the first requirement is failed, then it turns into the backoff process.

The flow chart of the multi-appointments mechanism based on the service

differentiation is shown as follows:

151
 Begin

Type of traffic

General
Important
traffic

NO NO
RTS request RTS request

YES YES
Received Reserving channel
Received
CTS again after UIFS Backoff

Send data NO
RTS request Send data

YES

Received
Backoff

Send data

Figure 6-4 Multi-channel request mechanism flowchart

6.3 Theoretical analysis of the SMAC-SD

Assume that there are n nodes in the WSN sharing a radio channel without

considering the problems of the hidden terminal and channel fading. For the channel

access process of a node, set s (t ) as the back off level of the node at the moment, the

range of the value is

[0, m] , where m is the value of maximal back off. The b(t )
152
stands for the value of back off counter at t moment, the range of the value is [0, Wi −1 ]

where
Wi = 2iW , the range of i is [0, m] ; W = CWmin ， CWmax = 2mW . The two

random processes described by s (t ) and b(t ) are obviously not the Markov process.

Regard these two non- Markov random process as an entirety and use two-dimensional

random process {s (t ), b(t )} to demonstrate the channel competition process of this

node. The status transmission is shown in Figure 6.5.

Figure 6-5 Flow chart of the two-dimensional Markov process

Assume the probabilities of data frame collision are not related to the former

collision numbers, and the collision probability

p is constant and independent. At

the status {s (t + 1), b(t + 1)} of t +1

moment, two-dimensional random process is

only related to the former status {s(t)，b(t)} at t moment. It has no relationship with

former status such as {s(t-1)，b(t-1)}, and the time and the status are discrete. Above all,

the random process {s (t ), b(t )} is a two-dimensional Markov chain. To simplify the

153
analysis, set the transmission probability as:
P{i1 , j1 | i0 , j0 } = P{s(t + 1) = i1 , b(t + 1) = j1 | s(t ) = i0 , b(t ) = j0 }
(6-2)
P{i, j | i, j + 1} = 1 0 ≤ i ≤ m, 0 ≤ j ≤ Wi − 2
(6-3)

P{i + 1, j | i,0} = p 0 ≤ i ≤ m − 1,0 ≤ j ≤ Wi +1 − 1

Wi + 1 (6-4)

P{m, j | m,0} = p 0 ≤ j ≤ Wm − 1 (6-5)

Wm
P{0, j | i,0} = (1 − p)W0 1 ≤ i ≤ m,0 ≤ j ≤ W0 − 1
(6-6)
Formula (6.2) shows that the backoff counter will subtract 1 immediately when

the backoff begins. Formula (6.3) shows that the backoff level will add 1 if the backoff

level does not achieve the maximal value m and if the sending of the data frame is

failed. The probabilities of the value of the counter is obtained randomly from the

range of [0, Wi +1 − 1] ; Formula (6.4) shows that when the backoff level get to the

maximal value m, the value of the backoff level keeps constant and the probabilities of

the value of the counter is obtained randomly from the range of [0, Wi − 1] if the

sending of the data frame is failed. Formula (6.5) shows that the backoff level value

will migrate to the minimal value 0 if the sending is successful at any backoff level.

It is defined:

bi , j = lim P{s (t ) = t , b(t ) = j} 0 ≤ i ≤ m,0 ≤ j ≤ Wi − 1 (6-7)

Based on the Figure 6-5, the following status transmission formula can be

obtained:

bi −1, 0 p = bi , 0 0<i<m


 bm −1, 0 p + bm , 0 p = bm , 0 i=m (6-8)

bi , 0 = p ib0, 0 0<i<m


bm, 0 = p 1 − pb0, 0 i=m
m (6-9)

From Figure 6-5, it is concluded:

For any j ∈ [0, Wi − 1]

154
  m

 (1 − p ) ∑ bk , 0 i=0
w − j k =0
bi , j = i  pbi −1, 0 0<i<m (6-10)
wi 
p(bm −1, 0 + bm, 0 ) i=m



All bi , j can be shown by b0, 0 and p. From Figure 6-5, it can be known that
m WI −1

∑ ∑b
i = 0 WI
i, j =1

Because

Wi − j
bi , j = bi , 0
Wi
Thus,
2(1 − 2 p )(1 − p )
b0, 0 =
(1 − 2 p )(W + 1) + pW [1 − (2 p ) m ] (6-11)

The probability of sending data frame at any moment is:

m m
1 − p m +1
τ = ∑ bi ,0 = ∑ p ib0,0 = b0, 0
i =0 i =0 1− p (6-12)

Formula (6.11) can be used for all priority transaction data.

For the transaction with different priorities, the probability of sending the data

frame at any moment from the nodes of the real-time transaction with high priorities is:
m =1
2(1 − p1 )(1 − p1 )
τ1 = m +1 m +1 (6-13)
W1[1 − (2 p1 ) ](1 − p1 ) + (1 − p1 )(1 − p1 )

The probability of sending the data frame at any moment from the nodes of the

general transaction with low priority is:

m +1
2(1 − p2 )(1 − p2 )
τ2 = m +1 (6-14)
W2 [1 − (2 p2 ) m +1 ](1 − p2 ) + (1 − p2 )(1 − p2 )

Where
W1 and W2 are the minimal competition window of the real-time

transaction and the general transaction, respectively.

P1 and P2 are the collision

probabilities of sending frame of the transaction stream of the important transaction

155
and the general transaction, respectively.

From formula (6.11), the probability of sending the data frame of the nodes at any

moment is τ , which is decided by the collision probabilities

p . The channel status

of sending the collision event of a node is: in the transmission networks with the node

numbers n , the collision happens in the data frame of a node. This collision happens

when there is at least one node sending the data in the n − 1 nodes. Set the

probabilities of sending the data frame of every node as ι , then,

p = 1 − (1 − τ ) n −1 (6-15)

For the transactions with different priorities, the probabilities that there is a

collision taking place on the nodes n1 of the real-time transaction with high priority is:

p = 1 − (1 − τ 1 ) n1 −1 (6-16)

The probabilities that there is a collision taking place on the nodes n2 of the

real-time transaction with low priority is:

p = 1 − (1 − τ 1 ) n1 (1 − τ 2 ) n2 −1 (6-17)

Analysis for the Throughput

In this protocol, S means the throughout of the system, which is defined as

following:

E[Load of one slot ]

S=
E[ Time of slot] (6-18)

Before the analysis the throughput, the following parameters regionalized

Set Ptr as the probabilities when the channel is busy. The busy of channel means

that there is at least one node sending the data during the channel competition process

of all nodes with the priority transaction. Thus:

Ptr = 1 − (1 − τ 1 ) n1 (1 − τ 2 ) n2
(6-19)
Ps1 and Ps 2 are the data frame successful sending probabilities of the
156
 important transaction and the general transaction, respectively;
Ps11 and Ps12 are

the data frame successful sending probabilities of the important transaction and the

general transaction without the collision, respectively; Pc1 is the collision probability

between the real-time transaction data frames,

Pc 2 is the collision probability between

the general transaction data frames when there is no transaction; Pc11 is the collision

successful probability between the real-time transaction data frames,

Pc12 is the

collision failed probability between the real-time transaction data frames.

Ps1 = Ps11 + Ps12 + Pc11
(6-20)
Ps11 = n1τ 1 (1 − τ 1 ) n1 −1 (1 − τ 2 ) n2
(6-21)
Ps12 = n1τ 1 (1 − τ 1 ) n1 −1
[1 − (1 − τ 2 ) ]
n2

(6-22)
Ps 2 = n2τ 2 (1 − τ 1 ) (1 − τ 2 )
n1 n 2 −1

(6-23)
Pc1 = 1 − (1 − τ 1 ) n1 − n1τ 1 (1 − τ 1 ) n1 −1
(6-24)
1
Pc11 = ( ) n1 Pc1 (6-25)
2
1
Pc12 = [1 − ( ) n1 ]Pc1 (6-26)
2
Pc 2 = (1 − τ 1 ) n1 [1 − (1 − τ 2 ) n2 − n2τ 2 (1 − τ 2 ) n2 −1 ]
(6-27)
Ts11 ， Ts12 ， Ts 2 ， Tc11 ， Tc12 ， Tc 2 response to the channel time of successful

sending data and the conflict.

Ts11 = RTS + 3SIFS + CTS + DATA + ACK + DIFS
(6-28)
Ts12 = 2 RTS + 4SIFS + 3slottime + CTS + DATA + ACK + DIFS
(6-29)
Ts 2 = RTS + 3SIFS + CTS + DATA + ACK + DIFS
(6-30)
Tc11 = 2 RTS + 4SIFS + 3slottime + CTS + DATA + ACK + DIFS
(6-31)

157
 Tc12 = 2 RTS + SIFS + slottime + DIFS
(6-32)
Tc 2 = RTS + DIFS
(6-33)
Based on the above analysis, the saturation throughput of the important

transaction and the general transaction are:

Ps1E[data ]
S1 =
(1 − Ptr )θ + Ps11Ts11 + Ps12Ts12 + Ps 2Ts 2 + Pc11Tc11 + Pc12Tc12 + Pc 2Tc 2 (6-34)

Ps 2 E[data ]
S2 =
(1 − Ptr )θ + Ps11Ts11 + Ps12Ts12 + Ps 2Ts 2 + Pc11Tc11 + Pc12Tc12 + Pc 2Tc 2 (6-35)

where θ is the duration of an empty slot, E[data ] is the average payload time of
transmission. It is known from analysis formulas (6-15) and (6-16), that the failure

probabilities caused by collision will be influenced by the transaction data stream with

higher priorities. From the formulas (6-19) and (6-22), when the probabilities of

sending frame and the number of nodes of the real-time transaction with high priorities

and the general transaction with the low priorities are same, the successful probabilities

of the real-time transaction is higher than the one of the general transaction. Thus the

similar conclusion can be obtained for the saturation throughput

Under the collision between the data stream of the important transaction and the

general transaction, the delay of data stream of the important transaction with the

successful sending (TSs12)

TDs12 = RTS + 2 * SIFS + floor (2.999 * rand ) * slottime + ACK

(6-36)

Under the collision between the data streams of the important transaction, the

delay of data stream of the important transaction with the successful sending (TDc12)

TDc11 = Tc11
(6-37)

Under the collision between the data streams of the important transaction, the

delay of data stream of the important transaction with the failed sending (TDc12)

TDc12 = Tc12
(6-38)

158
 Under the collision between the data streams of the general transaction, the delay

of data stream of the general transaction with the successful sending (TSs12)

TDc 2 = Tc 2
(6-39)

6.4 summary

A new protocol SMAC-SD is proposed to support QoS. It includes service

distinction, priority mechanism based on service queue, and channel requests service

differentiation mechanisms constituting secure transport mechanism. The SMAC-SD

protocol ensured the network's context of the high-quality transmission on diverse

business, at the same time to establishing the analytical model by formula deduction to

verify the correctness of the theory.

159
 Chapter 7 Simulation and analysis

7.1 Simulation analysis of RBLEACH protocol

7.1.1 The parameters and settings of the simulation

Simulation process:

1. The simulation starts and the simulation programs begin. The simulation scene

is set and each energy parameter is added into the settings.

2.Initialization settings of the whole system.

3. Preset function is set based on the relative theories and the cycle setting is done.

4. Implement the results and the simulation is ended.

The settings of the simulation scene are described as follows:

The 400 sensor nodes are randomly distributed into the network region with the

area of 200m × 200m .The preventing position of the base station is ( 0,0 )

160
Table 7.1 Used parameters in the simulation

Parameter Value

Network size ( 200m × 200m )

Base station position ( 0,0 )

Amount of nodes 400

Expected percentage of the head of cluster P 0.1

Fusion probability 0.6

E0 initial 0.05J
energy of single node

ETX 50 ∗ 0.000000001J / bit

transmission energy cost

ERX receiving 50 ∗ 0.000000001J / bit

energy cost

E fs 10 ∗10−12 J / bit / m 2
channel model of the free space

Emp 0.0013 ∗10−12 J / bit / m 4

channel model of the Multipath fading

EDA energy 5 ∗10−12 J / bit

cost of the data fusion

d0 border Threshold 84

(α1 ,α 2 ) (0.6,0.4)

(α , β ) (0.6,0.4)

161
 Figure7.1 shows a network of 400 nodes randomly produced according to the

above parameters

Figure 7.1 Random networks of 400 nodes

7.7.2 Performance indicator of simulation

The network simulation needs to set many performance indicators based on

practical situations, and to set a realistic parameter to these performance indicators. All

indicators and parameters are implemented by the implementation of the simulation

objectives initially set that need to be solved by the new protocols. In this experiment,

the results of the simulation are to prolong the life cycle of the networks. The energy

cost of nodes decreases in each round, and the energy cost of sensor nodes of data

collection and transmission can be balanced in the whole network. In order to achieve

this objective, the simulation implements LEACH algorithm and RBLEACH algorithm;

then compare the performance of the two algorithms. The standards to evaluate the

RBLEACH algorithm are described as follows:

1. Network life: duration from the first dead node to the last one.

2. Proportion of the amount of cluster heads.

3. Comparison of residual energy in each round.

4. Comparison of energy cost of cluster heads in each round.

162
 7.7.3 Experiment results of simulation and comparison of relative work

1. Comparison of network life

Figure 3.8 demonstrates the relation between the amount of dead nodes and the

network working time (rounds). Compared with LEACH, RBLEACH proposed in this

thesis can greatly prolong the life cycle of a network because every sensor node in the

cluster shares the energy cost evenly. Meanwhile, the node with more energy is first

selected as the cluster head, and the proportion of cluster head distribution is balanced

in the cluster head selection. It synchronizes the energy cost of each sensor node and

delays the moment of the first death of a node. The RBLEACH algorithm prolongs the

life of nodes 30% more than when using LEACH algorithm.

Figure 7.2 Comparison of amount of dead nodes

Figure 7.2 shows the death rate of sensor nodes. It is known from the first death of

a node that the death of all nodes will be concentrated in the last 40 rounds when using

RBLEACH algorithm, which is better than the death rate when using LEACH

algorithm, illustrating that the algorithm can balance the energy used by the nodes and

that the networks can supply high quality service.

2. Comparison of cluster head amounts

163
 Figure 7.3 shows the number of cluster heads selected with different rounds in the

LEACH protocol and RBLEACH protocol. It is obvious that the amount of the live

sensor nodes in RBLEACH protocol is more than that of LEACH protocol. The

amount of the cluster heads selected with proportion of RBLEACH protocol is less

than that of LEACH protocol. Thus, the energy cost in the process of transmission in

RBLEACH protocol is reduced, achieving the objectives of reducing energy cost.

Figure7.3 Comparison of amount of cluster heads

3. Comparison of residual energy for each round

Figure 7.4 shows the comparison of total residual energy when using LEACH

protocol and when using RBLEACH protocol for each round. It can be known from

Figure 3.10 that the total residual energy when using RBLEACH protocol is much

more than that of LEACH protocol, illustrating that the life cycle of the whole network

when using RBLEACH protocol is much longer than that when using LEACH

protocol.

164
 Figure7.4 Comparison of residual energy for each round

4. Comparison of the energy cost of the cluster heads for each round.

Figure 7.5 shows the simulation results of cluster head energy cost of LEACH

protocol and RBLEACH protocol for each round. It’s shown in Figure 7.5that the

maximum value of the cluster head energy cost of RBLEACH protocol is less than the

minimum value of the cluster head energy cost of LEACH protocol for all rounds.

Thus, under the condition where the same initial energy is supplied, the cluster head

energy cost of RBLEACH protocol is less than that of LEACH protocol. Therefore, the

life cycle of RBLEACH protocol can be prolonged.

Figure 7.5 comparisons of the residual energy for each round

In conclusion, whether examining the lifetime or total residual energy of the

sensor nodes in the networks, RBLEACH protocol is better than LEACH protocol. In
165
RBLEACH protocol, the whole network is divided into several regions and the cluster

heads are selected by proportion, which obviously contributes to a more uniform

distribution of the cluster heads in the network. The node energy cost of the data

transmission from the member nodes to the cluster head and from the cluster head to

the base station is lower, resulting in a longer life for each node in the network.

Compared with LEACH protocol, the multi-hops selected by the cluster head in the

clusters are used for data transmission to save energy between the sensor nodes. This

relatively improves the lifetime of each sensor node in the network.

7.2 Simulation analysis of PS-ACO-LEACH

7.2.1 Performance indicator

Throughout this thesis, the performance of the PS-ACO-LEACH is analyzed from

the following aspects:

1. Load balance: the routing protocol of WSN should distribute the network load

to each node evenly as much as possible and prevent few nodes or some region nodes

dying too early, which can affect the networks performance. This chapter is to analyze

the protocol based on the distribution of dead nodes.

2. Life cycle: the duration from the operation of the networks to the death of the

nodes is shown as “round”, which can be expressed by two durations: duration from

the first death and duration of the death of half the nodes, if there are half of the

sensors exhausting their energy, the WSN is considered to be dead.

3. Energy cost: the routing protocol of WSN should be good to save the energy of

the nodes in the networks and prevent the unnecessary consumption. This chapter

contributes to the analysis from the total residual energy of the nodes in each round.

7.2.2 Simulation results and evaluations

1. Design of simulation environment.

166
 In order to compare the performance difference between LEACH and

PS-ACO-LEACH in different ranges, the simulations are operated in two different

scenes respectively. The area of scene 1 is 100m*100m, and the coordinate of the base

station is (0, 0). The nodes of scene 2 are distributed into the region with an area of

200m*200m, and the coordinate of the base station is (0, 0). The relative parameters

are shown in Table 7-2, and the detailed physical definitions are described in 4.2.2.

Table 7-2 The detailed parameters of the network environment

Experimental parameters Scene 1 Scene 2

Region area 100m*100m 200m*200m

nodes amount 110 440

Location of the base station (0, 0) (0, 0)

initial energy 0.1J 0.1J

p 0.1 0.1

length of the data packet 4000 bit 4000 bit

Eelect 50nJ / bit 50nJ / bit

ε fs 10 pJ / bit / m 2 10 pJ / bit / m 2

ε mp
0.0013 pJ / bit / m 4 0.0013 pJ / bit / m 4

EDA 5nJ / bit 5nJ / bit

d0 87m 87m

7.2.3 Evaluations for simulation performance indicator

1. Load balance

Assume that the area of the region is 200m*200m, there are 400 nodes and the

base station is located at (0,0). Figure 7-6 and Figure7-7 show the distribution of dead

nodes of the WSN operation in the LEACH protocol and PS-ACO-LEACH in

early-term. Figure 7-8 and Figure 7-9 show the distribution of the dead nodes to these

167
two protocols in medium-term. The nodes far away from the base station begin to die

first due to single-hop routing used in LEACH protocol. Moreover, the distribution of

the cluster heads are more reasonable and the energy cost is less because of improved

clustering strategy and the multi-hop mechanisms used in PS-ACO-LEACH protocol,

thus the dead nodes are distributed evenly. LEACH protocol has a considerably large

part of its region that cannot be covered in the medium-term and late-term of the

network operation which means the network will fail. However, PS-ACO-LEACH

protocol can cover a large part of the region in the medium-term and late-term of the

networks operation. Thus, the PS-ACO-LEACH protocol is better than LEACH

protocol when considering the load balance.

Figure 7-6 Distribution of dead nodes in early term of LEACH

168
 Figure 7-7 Distribution of dead nodes distribution in early term of

PS-ACO-LEACH

Figure 7-8 Distribution of dead nodes in the medium term of LEACH

Figure7-9 Distribution of dead nodes distribution in the medium term of

169
PS-ACO-LEACH

2. Life cycle

Figure 7-10 shows the network life cycle in the 100m*100m scenario. The dotted

line represents the rising curve of dead nodes in the CRP-PSAC protocol, and the solid

line represents the rising curve of dead nodes in the LEACH protocol. In scenario 1,

the moment of the first nodes death (FND) in the CRP-PSAC is the 200th round and

that of the LEACH is the 166th round. For the half node death (HND) moment, the

CRP-PSAC is the 200th round and the LEACH is the 225th round, illustrating that that

the CRP-PSAC can distribute the energy cost more evenly into all the nodes to prevent

early death of a single node due to energy loss.

Figure7-11 shows the network life cycle in the 200m*200m scenario. In scenario

2, the advantages of the CRP-PSAC are further expressed. The moment of its FND and

HND are the 186th round and the 267th rounds respectively. The moment of LEACH’s

FND and HND are the 29th round and the 131st round respectively. This is because the

communication distance between the cluster head and the base station increases in the

LEACH protocol under the large scenario environment, and the single communication

energy loss of the cluster heads increases. However, in the CRP-PSAC protocol, the

communication between cluster heads and the base station uses a multi-hop routing

method, thus the single communication energy loss will not increase with the increase

of the area. Moreover, the PS-ACO-LEACH protocol considers the judgment standard

for multiple cluster heads to distribute network energy amongst all the sensors as much

as possible.

It is noted that the all curves are drawn by repeating the simulation 100 times to

calculate the average value in order to make the simulation results convincing. The

number of network nodes is set to the 110th and the 440th rounds respectively because

of programming needs. However, the curves are cut at the 100th and the 400th round in

the figures.

170
 Figure 7-10 Relation between dead nodes and time in the 100m*100m scene

Figure 7-11 Relation between dead nodes and time in the 200m*200m scene

3. Energy cost

Figure 7-12 and Figure 7-13 show the relation between network energy cost and

time in the LEACH and the PS-ACO-LEACH protocol under the 100m*100m and the

200m*200m environments respectively. The figures intuitively show the relation

between the networks residual energy and time. As shown in the figures,

PS-ACO-LEACH can save the network residual energy costs compared with LEACH.

The rate of the energy cost decreases, and the total residual energy of the

PS-ACO-LEACH is lower than that of LEACH at any moment. This is because most

of the nodes in LEACH have died over time, but there are still many live nodes in the

PS-ACO-LEACH. Thus, in LEACH, the covered area of the cluster head is larger, the

collected information is more, and the energy cost is relatively more.

When the area of the network region becomes larger, the energy-saving effects of

PS-ACO-LEACH are more obvious, which can be explained in the same way as was

171
explained in Figure 7-11

Figure 7-12 Relation between network residual energy and time in the

100m*100m scene

Fiure 7-13 Relation between network residual energy and the time in 200m*200m

scene

7.3 Emulation of MIA-QRP

7.3.1 Creation of protocol testing environment

To test the operation of this protocol, verify the correctness of routing protocol

and the good characteristics of protocol performance, it is not only needed to

172
accomplish the software design of the protocol and drive design and compilation of

wireless LAN, but also to set the forward of the data package supported by the

destination machine and to set the node networks card.

7.3.2 Support setting of nodes to data package for forwarding

The ADHOC network is characterized by moving multi-hops, and the nodes can

randomly move in the networks. In the process of the network communication, each

node functions as both the client and router. When one node becomes the forwarding

node of one communication link, it needs to supply the forwarding of data package for

the network data communication between the original node and the destination one.

Thus, to test this protocol needs to modify the routing function of nodes, so that the

node can forward the data package.

There is a complete TCP/IP protocol stack in the Linux, which supports the data

package forwarding. The Linux operation system has a powerful service, and has

functions of the network service of the client computer and the data package

forwarding of the router. However, the Linux is just used as the client computer in the

default condition without using the forwarding of the Linux to the data package. To

meet the need of this routing protocol testing, the destination machine regarded as the

nodes; need to start the forwarding function. The following are the detailed setting

commands:

sysctl -w net.ipv4.ip_forward=1

After start of the Linux system, imputing the above setting commands can

accomplish the forwarding of system to the network data package. As Linux operation

system does not start this function in the default condition, the data forwarding will be

shut down when the system is restarted. To reduce the repetition of data forwarding

configuration of system set in the each testing, the data package forwarding settings

can be written into system files to read the configuration file itself to start the

173
forwarding. The correction methods of the system configuration files are described as

follows:

First, open the configuration files, the command is: # vim /etc/sysctl.confined the

configuration item and correct “the net.ipv4.ip_forward = 0” to “net.ipv4.ip_forward =

1”. After correction, save it and quit. In order to make the set configuration come into

effect timely, the command line needs to be input as “# sysctl –p” Input the command

“# sysctl net.ipv4.ip_forward” to check if the data forwarding starts.

7.3.3 Network configuration of testing nodes

The friendly arm company development board mini2440 is used for the testing

nodes in this protocol use. There are three tested nodes, and the networks card of each

node use rt73 USB wireless LAN. Each nodes target board needs to install the

compiled network card driver. The installation steps of the Rt73 USB wireless LAN

are described as follows:

First, download the compiled files to the mini2440 development board and

execute the following commands:

# mkdir /etc/Wireless

# mkdir /etc/Wireless/RT73STA

# cp rt73.bin /etc/Wireless/RT73STA/rt73.bin

# dos2unix rt73sta.dat

# cp rt73sta.dat /etc/Wireless/RT73STA/rt73sta.dat

# /sbin/insmod rt73.ko

After the above operation, the compiled driver is successfully installed to the

system, and the networks card can work generally.

After load of the networks card driver, it is need to start the networks and

configure network nodes. The basic configuration information of the network nodes is

described as follows:

174
 Table 7-3 Node configuration

Node name IP address MAC address

A 192.168.1.172 D8:5D:4C:7C:15:2F

B 192.168.1.171 D8:5D:4C:7D:95:1F

The node configuration and the network settings are described as follows:

(1)start the networks card: if config rausb0 up, rausb0 is the name of the networks

card interface

(2)set the IP address of the nodes: the IP address corresponding to the if config

rausb0.

(3) set the networks card to ad-hoc networks mode: iwconfig rausb0 mode ad-hoc

(4) set SSID for the node: iwconfig rausb0 essid “dsh”, which improves the

security of the networks

(5) download the protocol sdap to the development board and correct the

competence: chmod +x sdap

(6) execute the protocol modul./sdap

Through the above settings, the neighbor nodes have the function of data

transmission of using USB wireless LAN.

7.3.4 Protocol performance testing

There are many indexes to measure the performance of the routing protocol. This

thesis is mainly to judge the protocol by the primary path finding velocity, path

maintenance effectiveness, data package forwarding and protocol cost etc.. The

performance testing and the results of the analysis are provided.

1. Path findings testing

Due to limited resources, testing networks consist of the three networks nodes.

The distribution of the network node is shown in Figure 7-14

175
 B

A
C

Figure 7-14 Distribution of testing node network

Based on the testing results of nodes, the operation of this protocol is described.

Start node B and node C before testing node A, and build path topology between node

B and node C. According to the design theory of this routing protocol, node A will

build a network topology information table with node B and node C after it starts to

form the A-B and the A-C two direct path and form the A-B-C and the A-C-B two

multi-hop path. Then the routing algorithm is called to select a optimized path between

the nodes to be added to the system routing table.

The network topology information of node A can be printed, as shown in Figure

7-15, after it starts and operates the routing protocol for a while.

Figure 7-15Network topology information of node A

As shown in Figure 7-15, there are 5 pieces of information in the network

topology information table of node A. The first one is the network information of node

A itself, which is designed and produced according to this protocol and has no

corresponding path. The second one and the third one are the direct path topology

information, which is related to the nodes networks configuration information in Table

7-3. It is known that the second one and the third one record the topology information

of path A-B and path A-C respectively. The mac address recorded in the fourth one

and the fifth one is 00-00-00-00-00, which records the multi-hop path information and

176
corresponds to topology information of the path A-B-C and the path A-C-B,

respectively. The network topology information produced in node A is in compliance

with the design theory of this protocol.

In order to inquire the selected final routing information of this protocol by the

MIA-QRP routing selection algorithm, the system routing table and the ARP table

information of the node A can be observed by using the routing information and arp

address analysis table information of the output system of the command route and

command arp respectively, as shown in Figure 7-16and Figure 7-17.

Figure 7-16 System routing table information

Figure 7-17 System ARP table information

According to the recording of system routing table information, path A-B and

path A-C are finally selected in the routing algorithm designed in this routing protocol

finally as the optimized path for communication between node A and nodes B,C.

Meanwhile, the system ARP table records the physical address corresponded to the IP

address of node C and node B, which corresponds to the system routing table

information recording. The network topology structure of node A is as show in the

dotted line in Figure 7-18.

A
C

Figure 7-18 Network topology structure.

177
 2. Path maintenance testing

Based on the distribution of the nodes in the networks shown in Figure 7-14, each

node has produced the network topology path information after the routing protocol

operation. Now shut down the main computer of node C, and further observe the

change of the network topology information of node A. The printed and outputted

network topology information of the A after shutdown of node C and waiting for a

while is shown in Figure 7-19.

Figure 7-19 Network topology information of node A

Based on the topology path information of node A, node A only has the

information record of the direct path that arrives at node B. The network topology

information produced before or related to node C has been deleted, which is in

compliance with the results of this routing protocol design. The system routing table

and the ARP table information of the node A, as shown in the Figure 7-20 and the

Figure7-21 respectively, are examined by the route command and arp command.

Figure 7-20 System routing table information of node A

Figure 7-21 ARP table information of node A

Based on the information between Figure 7-16 and Figure7-20 and by comparing

the ARP table information between Figure 5-44 and Figure 5-46, the node C becomes

invalid after it is shut down. Node A maintains and deletes the network topology

information related to node C after the routing, and updates the system routing table

information and ARP table information. There is only path information record of the

path A-B left. Based on this, node A can effectively run dynamic perception to the
178
network topology, timely update the network topology information, delete the expired

nodes and maintain the path.

3. Multi-hop path selection testing

The tested nodes are redistributed by using the T-shaped corridor in the laboratory

building, as show in Figure 7-22. The direct communication link between node A and

node C is influenced by the wall, so the signal strength of node C tested by node A is

largely decreased. Meanwhile, the commutation delay between node A and node C and

the package loss rate are influenced with different degrees. However, node B is set at a

corner, so there is no any blocking between node B and nodes A, C. The design of this

distribution is mainly to test if node A and node C can select the multi-hop path that

regards node B as the relay node by the routing selection algorithm.

Figure 7-22 Distribution structure of nodes

Start the mechanism of node B and node C and build the network topology path of

node B and node C, then start the mechanism of node B. Export the network topology

information of node A after the operation for a while, as shown in Figure 7-23.

Figure 7-23 Network topology information of node A

It is known from Figure 7-24 that node A has captured the network topology

179
information of node B and node C, and created a two-way direct path and two

multi-hop paths with each node. The system routing table and ARP table information

of node A, as shown in Figure 7-24 and Figure 7-25 respectively, are examined by the

route command and arp command.

Figure 7-24 System routing table information of node A

Figure 7-25 ARP table information of node A

They are shown from the system routing table information that the path from node

A to node C (IP address is 192.168.1.170) needs to go through the gateway (IP address

of node B). The physical address corresponding to the networks IP address of node C

is recorded as D8:5D:4C:7D:95:1F (MAC address of node B). This can completely

demonstrate that this routing protocol sets the multi-hop path A-B-C as the

communication path between node A and node C, and node B plays the role of relay

node.

4. Multi-hop path performance testing

For multi-hop path A-B-C formed in Section 5.2.3, each index and performance

of the link can be detected and the data forwarding of the node B can be supported by

using ping command on node A. First, the simple ping command is used to test the link

connectivity.

From the testing results shown in the Figure 7-26, the communication data

package between the node A and the node C is implemented by the forwarding of the

node B. The response data package received from the node A is sent back from

192.168.1.171 (that is the IP address of the node B). Thus this can verify that node B

has the function of data package forwarding and the function to be the gateway router

180
and it can be the relay node of node A and node C.

Figure 7-26 Ping command-testing results

In order to further detect the communication link delay between node A and node

C and the package loss rate performance, the statistic data can be obtained by the

extended instruction of ping command. The experiment tests the link performance by

changing the ping package number and the ping package sizes. The ping commands

used are described as follows:

#ping –q –c package number n –s package size m

The ping package number set in this testing are 50, 100, 300, 500, 700, 1200 and

1500. The ping package sizes are 1KB, 10KB and 50KB. The testing results under the

condition that the ping package number sent by the node A is 50 and the ping package

size sent by the node A is 1KB is shown in the Figure 7-27.

Figure 7-27 Link performance testing results

After the testing, the final testing results can be counted. The statistic results of

the link delay are shown in Figure 7-27.

According to the testing results, for the two-hop A-B-C link, the communication

delay of the link is stabilized at about 11ms when the sizes of the sent data package are

181
1KB and 5KB, and the package loss rate is stabilized less than 5%. The link

performance is tending towards stability and the fluctuation is slight. However, when

the sent data package size increases to 50KB, the communication delay of the link will

increase with the increasing of the data package number, and the package loss rate

increases to higher than 10%. Every aspect of the performance decreases greatly.

Figure7-28 Statistic results of link delay

5. Protocol cost

During operation of the protocol, using the top command can examine the

real-time occupation condition of protocol to the system CPU and the memory. The

results are shown in Figure 7-29.

Figure 7-29 Protocol CPU and memory cost

182
 It is shown in Figure 7-29 that the utilization of the protocol to the system CPU

and the memory are 0.4% and 27.0% respectively. Through the testing of protocol cost,

the occupation of this protocol to the system CPU is small. The occupation of this

protocol to the memory resource is up to 27%, including a buffer space of 15036K. But

this protocol is based on the mini2440 testing platform. The total memory size is 64M.

Thus, the memory cost of this protocol is small so that it can adapt to the using of the

selected development environment.

7.3.5 Compare with QoS protocol

Literature [153] reported the QoS protocol realization on Linux. According to zhu

Lin’s work, the performances of QoS are compared:

Test environment

Three nodes A,B,C, their IP address are set:

A(192.168.0.1),

B(192.168.0.2)

C((192.168.0.3)

Embedded nods A has a WIFI interface, it is used access the public network. IN

WIFI model, the Qos parameters are set according to the

Table 7.4 QoS parameters

Downlink Bandwith 90kbit/s

Maximum download Bandwith 24kbit/s

Maximum Uplink Bandwith 24bit/s

Minimum Bandwith 16kbit/s

Let A,B,C access the FTP server, and it will get average download speed:

183
 Table 7.5 QoS testing result1

Node average download speed average download

For literature speed for MIA-QRP

A 41kbit/s 43kbit/s

B 25kbit/s 32 kbit/s

C 24bit/s 31 kbit/s

Let B,C access the FTP server, A is not work, B,C’ average download speed

Table 7.6 Table QoS 7.6 testing result 2

Node Average download Average download

speed speed

For literature [148] MIA-QRP

B 25kbit/s 32 kbit/s

C 24bit/s 32 kbit/s

The result shows MIA-QRP has more accurate flow control on every IP than

literature [153].

7.4 Simulation of SMAC-SD protocol on the NS2

7.4.1 Realization of the SMAC-SD protocol

The SMAC-SD protocol is used to improve the S-MAC protocol. The realization

of the new protocol in the NS2 needs the finishing of the C++ protocol code and the tcl

debugging code. Compared to the code of the S-MAC protocol, the improvement is

made.

The modified sdsmac.h file and the sdsmac.cc file in the new SMAC-SD protocol

will add the file into NS2.

Simulation scene and parameter setting, tel sript in order to analyze the

performance, 20 nodes and two data streams of the different transaction with different

original node and destination node are set in the experiment. The protocol uses the
184
 AODV routing protocol and control the network traffic by the data transmission rate.

Related parameters are described as follows:

Table 7.7 Simulation parameters

Parameter Value Parameter Value

Routing protocol AODV maximal window 1024

data stream CBR minimal window 32

agency protocol UDP size of the data 512

packet

simulation time 100s Slot duration 20us

7.4.2 Simulation scene and Parameter settings

1.Simulation objective: the SMAC-SD protocol is simulated by the Low-load

network and high-load network, respectively. By analyzing the throughput and the

system delay, the improved SMAC-SD protocol is proved to supply a effective

transmission for the important transaction.

2. Simulation scene: In the using of WSN, the basic communication model

collects the data for the nodes of the sensor, and then sends to the aggregation nodes

which will send the data to the users. Because the nodes will self organize the wireless

networks and the users are in the limited networks, the nodes of sensor cannot

communicate with the users directly. Thus, the combination of the wireless networks

and the limited networks is involved. In order to be closer to the using reality, the basic

simulation scene of the SMAC-SD protocol is described as shown in Figure 30:

185
 Ad1
S1

S2

S3

route
route
BS
1
SN-

SN

Ad1

Figure 7-30 Basic simulation scene of SMAC-SD Protocol

3. Simulation experiment: in the basic simulation scene of SMAC-SD Protocol,

based on the Implementation requirements of the simulation objectives, this simulation

consists of the experiment one and the experiment two. The SMAC-SD protocol

performance is proved to meet the design objective by comparing the experiment one

and the experiment two.

(1) Simulation experiment one: simulation of SMAC-SD Protocol at the low-load

networks.

(2) Experiment one: By the simulation model and the supporting settings of the

parameters, the networks is set in the low-load networks. The SMAC-SD protocol is

simulated.

(3) Simulation scene of the experiment one: set four mobile wireless nodes, one

base station and three wired nodes (one of them is set as the routing). The four wireless

nodes are the original nodes, which can set two data streams of the different

transactions. The wired nodes are objective nodes, and different objective nodes

receive the different transaction data stream. The original nodes 0,1 send the important

transaction data stream, which is received by the objective node 1. The original nodes

2,4 send the general transaction data stream, which is received by the node 2. The

wireless node 0 is responsible for the routing. The topology of the simulation scene of

the experiment one is shown as figure 7-31 :

186
 S1

S2 Ad1

route
route
BS
S3

S4
Ad2
Ad1

Figure 7-31 Topology of Low-load simulation experiment Scene

The related simulation parameters of the important transaction data stream are

shown as table 7.8:

Table 7.8 Simulation parameters

Parameter Value Parameter Value

Routing protocol DSDV maximal window 7

data stream CBR minimal window 15

agency protocol UDP size of the data packet 256

Priority 0(High) Send packet rate 100Kb

AIFS 2 Simulation time 50s

The related simulation parameters of the general transaction data stream are

shown as table 7.9:

187
 Table 7.9 Simulation parameters

Parameter Value Parameter Value

Route AODV maximal window 15

Data Flow CBR minimal widow 31

Protocol agent UDP size of the data 256

packet

Traffic 1（low） Send packet rate 100Kb

priorities

AIFS 2 Simulation time 50s

(4). Simulation experiment two: simulation of SMAC-SD Protocol at the

high-load networks

Experiment two: By the simulation model and the supporting settings of the

parameters, the networks are set in the high-load networks. The SMAC-SD protocol is

simulated.

Simulation scene of the experiment two: set 8 mobile wireless nodes, one base

station and three wired nodes (one of them is set as the routing). The 8 wireless nodes

are the original nodes, which can set two data streams of the different transactions. The

wired nodes are objective nodes, and different objective nodes receive the different

transaction data stream. The original nodes 0,1, 2, 3 send the important transaction data

stream, which is received by the objective node 1. The original nodes 4, 5, 6, 7 send

the general transaction data flow, which is consider receiving by the node 2. The

wireless node 0 is responsible for the routing. The topology of the simulation scene of

the experiment two is shown in Figure 30and Figure 31:

188
 S0
Ad1
S1

S2

S3

route
route
S4

BS
S5

S6

Ad1
S7

Figure 7-32 Topology of High-load simulation experiment Scene

The related simulation parameters of the important transaction data stream and the

general transaction data stream are same with table 7.8 and table 7.9.

7.4.3. Performance Analysis of Low-load Protocol

In this simulation experiment, the wired link is set as 10 M, the wireless link is set

as 1m, the sending rate of every wireless nodes are 100kb, and there are 4 wireless

nodes. The network is in the condition of relax facing to the load. The trace file is

obtained after the simulation. The following results are obtained by the analysis of the

AWK tool:

1. Throughput

Due to the large amount of the obtained data, the data analyzed by awk needs to

be treated to be shown with the graphic. Take the low-sdsmac-through as an example.

There are 7626 rows data in the original file of sdmac-put-important. In order to show

the change trend of the throughput with the time, and value of the throughput is set

equidistantly with the probability of 0.01. One group is taken out every 100 rows.

There are totally 76 groups of data.

Under two transactions, the chart of the throughput as a function of time in the

SMAC-SD protocol at the low-load and the chart of the throughput as a function of

time in the S-MAC protocol at the low-load are shown respectively as follows figure:
189
 Figure 7-33 Throughput of the Low-load SMAC-SD protocol in the Different

Trnasactions

Figure 7-34 Throughput of the Low-load S-MAC procotol in the Different

Transactions
190
 In the simulation, set the bandwidth wireless link as 1M, there are 4 wireless

nodes, the rate of sending is 100 kb/s, the percentage of load is less than 40% of the

maximal load of the channel. The channel is in the low-load condition with the

reasonable competition. It can be concluded from analyzing the above two figures at

the low-load:

In the SMAC-SD protocol, the important transaction is set as the high priority 0,

the general transaction is set as the low priority 1. The effective transmission of the

important transaction is guaranteed by different size of the competition window and by

channel request mechanism of multi-sending RTS frames in the access slot. As shown

in the Figure 7-33, the important transaction almost uses all channel resources from 0

to 15s. However, the general transaction begins to receive the data packet from the 15s,

and there is no obvious difference throughput between the important transaction and

the general transaction after 15s. Thus, the ability to seize the channel of the important

transaction data is better than the general transaction data. Under the premise of good

transmission of the data for the two transactions, the SMAC-SD protocol can firstly

ensure the transmission of the important transaction.

S-MAC protocol does not divide the priority of the transaction, all types of the

transactions compete fairly for the channel. As illustrated in the Figure 7-33 and 7-34,

the data Throughputs of different types of transactions are similar, and the slight error

can be omitted.

3. System delay

The time interval of the system delay data that is analyzed by awk is small, the

jitter amplitude is large, and the drawn pictures are not clear. In order to show the

performance of the system delay, low-sdsmac-delay is set as an example. SMAC-SD

delay-important original data is valued equidistantly with 0.01 probabilities, one group

is taken out every 10 rows.

Under two transactions, the chart of the throughput as a function of time in the

SMAC-SD protocol at the low-load and the chart of the throughput as a function of
191
time in the S-MAC protocol at the low-load are shown respectively as follows

Figure 7-35 System Delay of the Low-load SMAC-SD protocol in the Different

Transactions

Figure 7-36 System Delay of the Low-load S-MAC protocol in the Different

Transactions

192
 Under the condition of low-load, it is known from the above the two figures7-35

and 7-36:

In the SMAC-SD protocol, from 0s to 15s, the system delay of the important

transaction data stream is concentrated with 0.0072s and is waved constantly with the

margin of 0.0003s. After 15s, the entire of the delay increases to be concentrated with

0.008s and is waved with the margin of 0.004s. In the general transaction, there is no

transmission before 15s, and the system delay is empty. After 15s, the system delay is

concentrated with 0.010s and is waved inconstantly with the margin of 0.005s. Thus,

the wave range of the important transaction is lower than the one of the general

transaction in the SMAC-SD. The delay in the important transaction is more stable,

which can ensure the effective transmission of the important transaction.

In the S-MAC protocol, there is no big difference of the system delay between the

important transaction and the general transaction. Both of them is concentrated with

0.009s and is waved with the margin of 0.005s.

4. Packet loss rate

Table 7-10 Packet loss rate of the different transactions at low-load in the

SMAC-SD protocol and S-MAC protocol

Important transaction General transaction

Packet loss Packet loss Packet loss Packet loss

Protocol type
number rate number rate

SMAC-SD 1285 0.144418 2672 0.299820

S-MAC 1314 0.147442 1146 0.128591

From the data in the Table 7-10, it shows:

Under the condition of the same load of the important transaction and general

transaction in the SMAC-SD protocol, the packet loss number of the important

transaction is 1285, the packet loss number of the general transaction is 2672; the

193
packet loss rate of the important transaction is 14.45%, the packet loss rate of the

general transaction is 29.98%. The packet loss rate of the important transaction is about

50% of the one of the general transaction, but the packet loss probabilities of two

transactions are small. Thus, under the condition of the low load, the SMAC-SD

protocol can effectively ensure the transmission of the different transactions, but it will

firstly ensure the transmission of the important transaction. The performance of the

important transaction is better.

Under the condition of the same load of the important transaction and general

transaction in the S-MAC protocol, the packet loss number of the important transaction

is 1314, the packet loss number of the general transaction is 1146; the packet loss rate

of the important transaction is 14.74%, the packet loss rate of the general transaction is

12.86%. Thus, the packet loss probabilities of the different data are similar in the

S-MAC protocol. The slight error can be omitted.

7.4.4 Performance Analysis of high-load Protocol

In this simulation experiment, the wired link is set as 10 M, the wireless link is set

as 1m, the sending rate of every wireless nodes are 125kb, and there are 8 wireless

nodes. The network is in the condition of full-load. The trace file is obtained after the

simulation. The following results are obtained by the analysis of the AWK tool:

1.Throughput

Due to the large amount of the obtained data, the data analyzed by awk needs to

be treated to be shown with the graphic. Take the high-sdsmac-through as an example.

There are 10180 rows data in the original file of sdmac-put-important. In order to show

the change trend of the throughput with the time, and value of the throughput is set

equidistantly with the probability of 0.01. One group is taken out every 100 rows.

There are totally 101 groups of data.

Throughout two transactions, the chart of the throughput as a function of time in

the SMAC-SD protocol at the high-load and the chart of the throughput as a function
194
of time in the S-MAC protocol at the high-load are shown respectively as figure 7-37

and figure 7-38

Figure 7-37 Throughput of the high-load SMAC-SD protocol in the Different

Transactions

Figure 7-38 Throughput of the high-load S-MAC protocol in the Different

Transactions

It can be concluded from analyzing the above two figures at the low-load:
195
 In the SMAC-SD protocol, the important transaction is set as the high priority 0,

the general transaction is set as the low priority 1. The effective transmission of the

important transaction is guaranteed by different size of the competition window and by

channel request mechanism of multi-sending RTS frames in the access slot. As shown

in the Figure 5-9, the important transaction almost uses all channel resources from 0 to

15s. After 15s, the throughput of the important transaction remains as much as five

times than the one of the general transaction. However, the general transaction begins

to receive the data packet from the 15s, and the throughput increases slowly after 15s.

Thus, the ability to seize the channel of the important transaction data is better than the

general transaction data considering the full-load and the channel competition. There is

large difference throughput between the important transaction and the general

transaction. The excellent performance of the important transaction is ensured by

sacrificing the transmission performance of the general transaction. At the same time,

under the extreme conditions of the full-load system, the general transaction can seize

the channel to ensure that there is no transmission hunger in the general transaction and

ensure the basic requirement of the transmission.

S-MAC protocol does not divide the priority of the transaction, all types of the

transactions competite fairly for the channel. As illustrated in the Figure 5-10, the data

Throughputs of different types of transactions are similar, and the slight error can be

omitted.

2. System delay

The time interval of the system delay data that is analyzed by awk is small, the

jitter amplitude is large, and the drawn pictures are not clear. In order to show the

performance of the system delay, high-sdsmac-delay is set as an example. There is

2712 rows of data in the sdsmac-delay-important original data. It is valued

equidistantly with 0.02 probabilities. One group is taken out every 50 rows and there

54 rows of data being obtained.

Under two transactions, the chart of the throughput as a function of time in the
196
SMAC-SD protocol at the full-load and the chart of the throughput as a function of

time in the S-MAC protocol at the full-load are shown respectively as figure 7-39 and

figure 7-40.

Figure 7-39 System Delay of the high-load SMAC-SD protocol in the Different

Transactions

197
 Figure 7-40 System Delay of the high-load S-MAC protocol in the Different

Transactions

Comparing the figure 7-39 and figure 7-40, it shows: in the SMAC-SD protocol

with high-load networks, from 0s to 15s, the system delay of the important transaction

data stream is concentrated with 0.0075s and is waved with low margin, which is

shown in Figure 5-11. After 15s, the entire of the delay begins to increase. After 20s, it

is concentrated with 0.01s and waved. The margin of waving is 6 times more than the

one from 0s to 15s. In the general transaction, there is no transmission before 15s, and

the system delay is empty. After 15s, the system delay is concentrated with 0.070s and

is waved inconstantly with the margin of 0.007s. The reason for the empty delay of the

first 15s in the general transaction is shown in Figure 5-9. The channel is used for the

transmission of the important transaction in the first 15s, and there is no data

transmission for the general transaction during this time. Thus, the concentration of the

delay and the wave range of the important transaction are lower than the ones of the

general transaction in the SMAC-SD. The delay in the important transaction is more

stable, which can ensure the effective transmission of the important transaction in the
198
 SMAC-SD protocol.

In the S-MAC protocol, there is no big difference of the system delay between the

important transaction and the general transaction. Both of them is concentrated with

0.0085s and is waved with the margin of 0.0005s.

3.Packet loss rate

Table 7-11 Packet loss rate of the different transactions at high-load in the

SMAC-SD protocol and S-MAC protocol

Type of Important transaction General transaction

protocol packet Loss packet packet Loss packet

SMAC-SD 4671 0.314503 3583 0.482494

S-MAC 3349 0.225491 3385 0.488738

In the full-load SMAC-SD protocol, the packet loss number of the important

transaction is 4671, the packet loss number of the general transaction is 3583; the

packet loss rate of the important transaction is 31.45%, the packet loss rate of the

general transaction is 48.25%. The packet loss rate of the important transaction is about

60% of the one of the general transaction. The packet loss number of the important

transaction is lower. Thus, the full-load SMAC-SD protocol can effectively ensure the

transmission of the important transaction and basically ensure the transmission of the

general transaction.

Under the condition of the same load of the important transaction and general

transaction in the S-MAC protocol, the packet loss number of the important transaction

is 2849, the packet loss number of the general transaction is 3885; the packet loss rate

of the important transaction is 22.55%, the packet loss rate of the general transaction is

48.87%. Thus, there are no differed priorities in the transactions of the S-MAC. The

packet loss rate is random.

7.4.5 General Analysis

It is concluded by analyzing the general performance of the SMAC-SD protocol

199
and the S-MAC protocol without dividing the types of the transactions:

1.Total throughput of the system

Figure 7-41 Change Trend of the Throughput as a function of the load in the

SMAC-SD protocol and the S-MAC protocol

From Figure 7-41, the total throughput of the SMAC-SD protocol is lower

than the one of the S-MAC protocol, and both of them slightly decrease with the

increasing of the load. There are two reasons leading to this change: (1) SMAC-SD

protocol focuses on the assurance for the performance of the important transaction. The

performance of the important transaction is higher than all other transactions in the

SMAC-SD protocol. However, the performance of the general transaction decreases.

Thus, the system performance of the SMAC-SD decreases slightly. (2) the access slot

and the competition window of the SMAC-SD protocol increases, which lead to the

increasing of the free timing. To sum up, the total throughput of the SMAC-SD

protocol system decreases slightly.

2.System packet loss rate

200
 Figure 7-42. The change of the packet loss rate as a function of the load in the

SMAC-SD protocol and the S-MAC protocol

Figure 7-42 shows the result of simulation, because the throughput of the

SMAC-SD protocol is slightly lower than the one of the S-MAC protocol, the packet

loss rate of the SMAC-SD is slightly higher than the one of the S-MAC protocol.

3.Average system delay

201
 Figure 7-43 the change of the packet loss rate as a function of the load in the

SMAC-SD protocol and the S-MAC protocol

From the Figure 7-43, the average system delay of the SMAC-SD protocol is

lower than the one of the S-MAC protocol. For the data that is sensitive to the delay,

SMAC-SD protocol is better

To sum up, although the total throughput of the SMAC-SD protocol slightly

decreases compared with the one of the S-MAC protocol, the throughput of the

important transaction is higher than all transactions in the S-MAC protocol. At the

same time, the average delay of the SMAC-SD protocol increases badly compared with

the one of the S-MAC protocol. Moreover, the delay of the important transaction in the

SMAC-SD protocol is much better than the average value.

Under the condition of the intense competition in the channel, the SMAC-SD

protocol can effectively protect the transmission of the important transaction, which is

in accordance with the objectives.

202
 7.5 summary

1. To RBLEACH, it can be concluded from the results of the experimental

simulation that RBLEACH has largely prolongs the lifetime of each sensor node. It

also solves the problem of the premature death of the nodes in LEACH protocol caused

by the far-to-close death and by the premature communication paralysis in the region

far away from the base station.

2. To PS-ACO-LEACH, Simulation results show that this protocol can avoid

network node energy consumption and energy consumption imbalance, which is

caused by unsatisfactory cluster head selection or poor routing paths. Compared with

LEACH, PS-ACO-LEACH improves the robustness of the network more effectively,

promotes balance of network energy consumption and enhances the network life cycle

effectively.

3. To MIA-QRP , it select route based on the multiple QoS parameters, including

delay, loss of package and quality. The test results show that the protocol can find a

path quickly and maintain the path effectively.

4.To SMAC-SD, the new protocol was verified by the NS2 simulation platform.

The simulation results show that the improved protocol can gain the throughput of vital

service and shorten the delay at the same time, and improve the performance of

network effectively.

203
 Chapter 8 Conclusions and Future Work

8.1 Conclusions

1. Comparing the RBLEACH algorithm and the LEAHC algorithm, the energy

cost of communication between each member node, head of cluster and base station is

less, the life cycle of the networks is elongated, and the multi-hop data transmission

between the clusters largely save the energy of each sensor node compared with the

LEACH protocol.

2. PS-ACO-LEACH protocol prevents too much energy cost or non-balanced

energy cost caused by the unsatisfactory selection of head of cluster or poor routing

path. Compared with LEACH, PS-ACO-LEACH protocol effectively improves the

network scalability, contributes to the balanced energy cost of network, and elongates

the life cycle of networks effectively.

3. MIA-QRP self-adaptive routing algorithm can rapidly find the path and make

effective maintenance to the path. The optimized path to the destination node can be

calculated according to the MIA-QRP self-adaptive routing algorithm. This protocol

supports multi-hop forwarding and occupies few system resources.

4. According to service differentiation, SMAC-SD adopts an access mechanism

based on different priorities. It can adjust priority mechanism based on service queue

and channel multi-request mechanisms, and configure waiting queues with different

priorities and RTS retries for different service with different priorities, which can make

vital service get high channel access probability and ensure the transmission quality of

vital service. The simulation results show that the improved protocol can gain the

throughput of vital service and shorten the delay at the same time, and improve the

performance of network effectively.

204
 8.2 Future work

1. In this thesis, the clusters are designed with the proportioned selections and

communication path between clusters in a special divided area. However, in actual

condition, there is possible difference density in the distribution of sensor nodes in the

network. The distributions do not show the uniform and random assumed in this thesis.

Thus, the division for the network should be judged by the density, which makes the

problem more complex. In the future, a clustering algorithm with optimized function

should be considered in order to adapt to all the changing factors.

2. For PS and ACO, in this thesis, the PS optimization clustering just uses the

basic PS with a slow convergence rate, high-energy consumption of nodes, and a

massive consumption of time. This basic PS also possibly results in the part

optimization, not the overall optimization. In the future, a more detailed research for

the PS should be developed to apply to the improved PS, so as to optimize the

clustering. The basic ACO is also used in the stage of multi-jumping routing. Simply

seeking for the shortest routing in the ESN routing protocol is not the best program. In

the future, an improved AC should be developed in this part and should be connected

with energy, location and other information to optimize the routing selection in order

to decrease the energy consumption of routing and elongate the lifetime of the network.

Meanwhile, because of large amount of weight parameters and complex relation

between each parameter in the PS fitness function, it is a very complex problem to

decide the best parameters to optimize the settings of clustering groups. In this thesis, a

calculation method for the optimized parameters is concluded for the random

distribution of the nodes. However, this method is not universal, and cannot be applied

for the non-uniform distribution of the nodes (e.g. random Gaussian distribution). The

author will further research the optimized weight parameters combination in the fitness

functions.

3. In this thesis, the MIA-QRP self-adaptive QoS routing protocol just considers

205
the delay of link, package loss rate and strength of the node signal, which can be only

adaptive for the network application of the part QoS protection and needs. According

to the application prospects of the WSN, The networks should dynamically perceive

the networks application and the QoS service demands, and comprehensively evaluate

the QoS service guarantee. Moreover, because the limited resource can only build a

simple testing environment, the protocol function cannot be evaluated currently in the

conditions with a large amount of nodes.

4. For QoS, since WSNs have to interact with the environment, their

characteristics can be expected to be very different from other conventional data

networks. Thus, while WSNs inherit most of the QoS challenges from general wireless

networks, including severe resource constraints, unbalanced traffic, data redundancy,

multiple sinks, packet criticality and so on, it will is studied how to design MAC

protocol to satisfy QoS. .

206
 References

[1] Anna Hac. Wireless Sensor Network Designs [M].Hoboken, NJ: John Wiley

& Sons Ltd,2003.

[2] Jennifer Yick, Biswanath Mukherjee, Dipak Ghosal. Wireless sensor

network survey [J].Computer Networks, 2008: 2292–2330.

[3] J. Yick, B. Mukherjee, D. Ghosal, Analysis of a Prediction-based Mobility

Adaptive Tracking Algorithm, in: Proceedings of the IEEE Second International

Conference on Broadband Networks (BROADNETS), Boston, 2005.

[4] M. Castillo-Effen, D.H. Quintela, R. Jordan, W. Westhoff, W. Moreno,

Wireless sensor networks for flash-flood alerting, in: Proceedings of the Fifth IEEE

International Caracas Conference on Devices, Circuits, and Systems, Dominican

Republic, 2004.

[5] T. Gao, D. Greenspan, M. Welsh, R.R. Juang, A. Alm, Vital signs

monitoring and patienttracking over a wireless network, in: Proceedings of the 27th

IEEE EMBS Annual International Conference, 2005.

[6] G. Wener-Allen, K. Lorincz, M. Ruiz, O. Marcillo, J. Johnson, J. Lees, M.

Walsh, Deploying a wireless sensor network on an active volcano, Data-Driven

Applications inSensor Networks (Special Issue), IEEE Internet Computing,

March/April 2006.

[7] F. Ehlers, D. Sofge, M. Chitre and J. Potter, “Distributed Mobile Sensor

Networks for hazardous Applications”, International Journal of Distributed Sensor

networks, vol. 2012, (2012), pp. 1-3.

[8] Vivek Katiyar,Prashant Kumar,Narottam Chand,An Intelligent Transportation

Systems Architecture using Wireless Sensor Networks,International Journal of

Computer Applications (0975 – 8887) Volume 14– No.2, January 2011

[9]KSHITIJ SHINGHAL,WIRELESS SENSOR NETWORKS IN

AGRICULTURE: FOR POTATO FARMING,International Journal of Engineering

207
Science and Technology,Vol. 2(8), 2010, 3955-3963

[10] Heinzelman W, Chandrakasan A, Balakrishnan H. Energy-efficient

Communication Protocol for Wireless Microsensor Networks[C]. In Proceeding of the

33rd,Annual Hawaii Int’l Conf.on System Sciences. Maui：IEEE Comoputer Society,

2000: 3005-3014.

[11] Lindsey S, Raghavendra C S. Pegasis: Power-efficient Gathering in Sensor

Information Systems[C]//Proc. of IEEE Int’l Conf. on Computer Systems. San

Francisco, USA: [s. n.], 2002.

[12] S.S. Satapathy and N. Sarma, TREEPSI: tree based energy efficient protocol

for sensor information ,Wireless and Optical Communications Networks 2006, IFIP

International Conference 11-13,April 2006 .

[13] Rajesh Patel,Sunil Pariyani,Vijay Ukani,Energy and Throughput Analysis of

Hierarchical Routing Protocol (LEACH) for Wireless Sensor Network,International

Journal of Computer Applications (0975 – 8887),Volume 20– No.4, April 2011

[14] Renold, A.P. ; Poongothai, R. ; Parthasarathy, R. , Performance analysis of LEACH with

gray hole attack in Wireless Sensor Networks, Computer Communication and Informatics (ICCCI),

2012 International Conference on, Date 10-12 Jan. 2012

[15] Ravneet Kaur, Deepika Sharma2 and Navdeep Kaur, Comparative Analysis

Of Leach And Its Descendant Protocols In Wireless Sensor Network, International

Journal of P2P Network Trends and Technology- Volume3Issue1- 2013,P51-54

[16] Qian Liao, Hao Zhu, An Energy Balanced Clustering Algorithm Based on

LEACH Protocol, Proceedings of the 2nd International Conference On Systems

Engineering and Modeling (ICSEM-13)

[17] Nguyen Duy Tan, Longzhe Han, Nguyen Dinh Viet, and Minho Jo,An

Improved LEACH Routing Protocol for Energy-Efficiency of Wireless Sensor

Networks,Smart Computing Review, vol. 2, no. 5, October 2012,P360-390

[18] Rajashree.V.Biradar , Dr. S. R. Sawant , Dr. R. R. Mudholkar, Dr. V.C .Patil,

Multihop Routing In Self-Organizing Wireless Sensor Networks IJCSI International

208
Journal of Computer Science Issues, Vol. 8, Issue 1, January 2011 155 ISSN (Online):

1694-0814,P155-164

[19]Akyildiz IF, Su W, Sankarasubramaniam Y, Cayirci E. A survey on sensor

networks. IEEE Communications Magazine, 2002,40(8): 102−114

[20] R.Sumathi and M.G.Srinivas,A Survey of QoS Based Routing Protocols for

Wireless Sensor Networks,J Inf Process Syst, Vol.8, No.4, December

2012,http://dx.doi.org/10.3745/JIPS.2012.8.4.589

[21]LAN MAN Standards Committee of the IEEE Computer Society, Wireless

LAN medium access control (MAC) and physical layer (PHY) specification,IEEE,

New York, NY, USA, IEEE Std 802.11-1997 edition, 1997.

[22] van Dam T, Langendoen K. An adaptive energyefficient mac protocol for

wireless sensor networks. In: Proc. of the 1st ACM Conf. on Embedded Networked

Sensor Systems (SenSys). New York: ACM Press, 2003. 171−180.

[23] Ye W, Heidemann J, Estrin D. An energy-efficient MAC protocol for

wireless sensor network. In: Proc. of the INFOCOM 2002. San Francisco: IEEE

Computer Society, 2002.

[24]Vithya, G. ; St. Joseph''s Coll. Of Eng., Chennai, India ; Vinayagasundaram,

B.Actuation sensor with adaptive routing and QOS aware checkpoint arrangement on

Wireless Multimedia Sensor Network, Recent Trends in Information Technology

(ICRTIT), 2011 International Conference on 3-5 June 2011 Page 444 - 449.

[25]Jafarzadeh, S.Z. ; Dept. of Comput. Eng., Ferdowsi Univ., Mashhad, Iran ;

Moghaddam, M.H.Y. Design of energy-aware QoS routing protocol in wireless sensor

networks using reinforcement learning,Electrical and Computer Engineering (CCECE),

2014 IEEE 27th Canadian Conference on 4-7 May 2014 page1 - 5

[26]Rui Dai ; Broadband Wireless Networking Lab., Georgia Inst. of Technol.,

Atlanta, GA, USA ; Pu Wang ; Akyildiz, I.F. Correlation-Aware QoS Routing

With Differential Coding for Wireless Video Sensor Networks,Multimedia, IEEE

Transactions on (Volume:14 , Issue: 5 )page 1469 - 1479.

209
 [27]Jing Feng ; Inst. of Meteorol., PLA Univ. of Sci. & Technol., Nanjing, China ;

Xiaoxing Yu ; Zijun Liu ; Cuihan Wang,A QoS Enhanced Routing Protocol in

Wireless Sensor Networks,Computer Communications and Networks (ICCCN), 2011

Proceedings of 20th International Conference on July 31 2011-Aug. 4 2011,page 1 - 6

[28] Akyildiz I，Su W, Cayirci E. Wireless sensor network [J]. Computer

Networks. 2002，vol.38: 393-422.

[29] Jiming Chen, Linrui Zhong, SUN You-Xian, wireless sensor networks

Communications System [J]. Sensing Technology,2006,19(4):1290-1295.

[30] Edgar H. Callaway. Wireless sensor networks: Architectures and

protocols[M]. BocaRaton: CRC Press, Jan.2004.

[31] Yongxin Liu,Zhaoxia Wang,Junqi Zhao,Zhenda Sun,A wireless sensor

network node design based on ZigBee protocol,Automatic Control and Artificial

Intelligence (ACAI 2012), International Conference on 3-5 March 2012,P:325 – 328

[32] Considine, J.; Li, F.; Kollios, G. & Byers, J. (2004). Approximate

aggregation techniques for sensor databases, In Proceedings of ICDE, pp. 449-460,

April, 2004.

[33] T.H. Kim and S. Hong, “Sensor Network Management Protocol for

State-Driven Execution Environment,” in Proc. ICUC Conf., Oct. 2003

[34] I.F. Akyildiz, J.N. and A. E. Kamal , “Routing techniques in wireless sensor

networks : A survey”, IEEE Wireless Communications , vol.11, no.6, pp.6-28, 2004.

[35] Jiming Chen, Linrui Zhong, SUN You-Xian, wireless sensor networks

Communications System [J]. Sensing Technology,2006,19(4):1290-1295.

[36] Edgar H. Callaway. Wireless sensor networks: Architectures and

protocols[M]. BocaRaton: CRC Press, Jan.2004.

[37] Nets-noss: Creating an architecture for wireless sensor networks.

[38] C-Y. Chong, S.P. Kumar, Sensor networks: evolution, Opportunities, and

challenges, Proceedings of the IEEE 91 (8) (2003) 1247–1256.

[39] Eliana Stavrou, Wireless Sensor Networks, part 2: Limitations,Web Hosting Articles

210
 [40] W. Ye, J. Heidemann, and D. Estrin, “Medium Access Control with

Coordinated Adaptive Sleeping for Wireless Sensor Networks,” IEEE/ACM Trans.

Net., vol. 12,no. 3, June 2004, pp. 493–506.

[41] A. El-Hoiydi, “Spatial TDMA and CSMA with Preamble Sampling for

Low Power Ad Hoc Wireless Sensor Networks,” Proc. ISCC 2002, July 2002, pp.

685–92.

[42] Al-Sakib Khan Pathan,Hyung-Woo Lee,Choong Seon Hong,Security in

Wireless Sensor Networks: Issues and Challenges,Feb. 20-22, 2006 ICACT2006

[43] Jeremy Elson and Deborah Estrin, Time Synchronization forWireless Sensor

Networks, In Proceedings of the First Workshop on Hot Topics in Networks

(HotNetsI),2829,October 2002, Princeton, New Jersey,

USA.http://www.acm.org/sigcomm/HotNetsI

[44] Mohammad Reza Gholami, Positioning Algorithms for Wireless Sensor

Networks, Communication Systems Group Department of Signals and Systems

chalmers university of technology,Gothenburg, Sweden 2011

[45] Sasikala V and C. Chandrasekar, Energy Efficient Multipath Data Fusion

Technique for Wireless Sensor Networks, ACEEE Int. J. on Network Security , Vol.

03, No. 02, April 2012

[46] Haiying Shen; Fayetteville, AR ;Ting Li,Data Management of Wireless

Sensor Networks,Consumer Communications and Networking Conference, 2009.

CCNC 2009. 6th IEEE

[47] David E. Culler , Arch Rock Corp. TinyOS: Operating System Design for

Wireless Sensor Networks, Networking & Communication

[48] KAZEM SOHRABY,DANIEL MINOLI,TAIEB ZNATI,WIRELESS

SENSOR NETWORKSTechnology, Protocols, and Applications,A JOHN

WILEY&SONS,INC. PUBLICATION

[49] Gyanendra Prasad Joshi, Seung Yeob Nam and Sung Won Kim, Cognitive

Radio Wireless Sensor Networks: Applications, Challenges and Research

211
Trends,Sensors 13, no. 9: 11196-11228.

[50] Dazhi Chen and Pramod K. Varshney,QoS Support in Wireless Sensor

Networks:A Survey, International Conference on Wireless Networks - ICWN , pp.

227-233, 2004

[51] Shamsad Parvin,Muhammad Sajjadur Rahim,Routing Protocols for Wireless

Sensor Networks: A Comparative Study,International Conference on Electronics,

Computer and Communication (ICECC 2008),PP891-894

[52] G. Gupta, M. Younis. Load-balanced clustering in wireless sensor

networks[C]. Proceedings of the International Conference on Communication (ICC

2003), Anchorage, Alaska, 2003.

[53] D. J. Baker and A. Epheremides, “The Architectural Organization of a

Moblie Radio Network via a Distributed Algorithm,” IEEE Transactions on

Communications, vol. Com-29, no. 11, November 1981.

[54] P. Tsigas, “Project on Moblie Ad Hoc Networking and Clustering for the

Course EDA390 Computer Communcation and Distributed Systems,” Manual for

University Course.

[55] A. Amis, R. Prakash, T. Vuong, and D. Huynh, “Max-Min D-Cluster

Formation in Wireless Ad Hoc Networks,” IEEE INFOCOM, March,2000

[56] K. Akkaya, M. Younis. A survey on routing protocols for wire less sensor

networks[J]. Journal of Ad Hoc Networks, 2005, 3(3):325-349.

[57] M. Younis, M. Youssef, K, Arisha. Energy-aware management in

cluster-based sensor networks[J]. Computer Networks, 2003, 43(5):649-668.

[58] Y. T. Hou, Y. Shi, H. D. Sherali. On energy provisioning and relay node

placement for wireless sensor networks[J]. IEEE Transactions on Wireless

Communications, 2005,4(5):2579-2590.

[59] G. Jolly, M. Younis. An energy efficient, scalable and collision less MAC

layer protocol for wireless sensor networks[J]. Wireless Communications and Mobile

Computing, 2005, 5(3):285-304.

212
 [60] Qianfeng Jiang,Manivannan D,"Routing.Protocol for Sensor Network"[A].

In:Consumer Communciations and Networking Conference[C].2004.93-98

[61] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy

Efficient Communication Protocol for Wireless Micro Sensor Networks,”

Proceedings of IEEE HICSS, Jan 2000.

[62] C. F. Chiasserini, I. Chlamtac, P. Monti, and A. Nucci, “Energy Efficient

Design of Wireless Ad Hoc Networks,” Proceedings of European,Wireless, Feb 2002.

[63] S. Bandyopadhyay and E. J. Coyle, “An Energy Efficient Hierarchical

Clustering Algorithm for Wireless Sensor Networks,” IEEE INFOCOM,April 2003.

[64] K. Dasgupta, K. Kalpakis, P. Namjoshi. An efficient clustering-based

heuristic for data gathering and aggregation in sensor networks[C]. Proceedings of the

IEEE Wireless Communications and Networking Conference(WCNC 2003), New

Orleans, LA, 2003.

[65] Jianping Pany Y. Thomas Houz Lin Caiy Yi Shiz Sherman X.

Sheny,Topology Control for Wireless Sensor Networks,MobiCom’03, September

14–19, 2003, San Diego, California, USA.p286-299

[66] Yao Liang Rui Liu,Routing Topology Inference for Wireless Sensor

Networks,ACM SIGCOMM Computer Communication Review,Volume 43, Number 2,

April 2013

[67] T. Zhao, W. Cai, and Y. Li. MPIDA: A sensor network topology inference

algorithm. In Proc. Int.Conf. on Computational Intelligence and Security,PP 451-455,

2007.

[68] Y. Yang, Y. Xu, and X. Li. Topology tomography in wireless sensor

networks based on data aggregation. In Proc. Int. Conf. on Communications and

Mobile Computing, pages 37{41, 2009.

[69] Bhaskar Krishnamachari, Deborah Estrin, Stephen Wicker, Modelling

Data-Centric Routing in Wireless Sensor Nwetwork,

[70] Kemal Akkaya, Mohamed Younis,A survey on routing protocols for wireless
213
sensor ,Ad Hoc Networks 3 (2005) 325–349
[71] Jae-Hwan Chang, Maximum lifetime routing in wireless sensor networks, Browse

Journals & Magazines > Networking, IEEE/ACM Transaction Volume:12 Issue:4

[72] Jianmin Zhang, Load-balanced Route Discovery for Wireless Sensor

Networks, JOURNAL OF NETWORKS, VOL. 5, NO. 9, SEPTEMBER

2010,p1060-1067

[73] Hai Liu, Amiya Nayak, and Ivan Stojmenovi´c, Fault-Tolerant

Algorithms/Protocols in Wireless Sensor Networks, S. Misra et al. (eds.), Guide to

Wireless Sensor Networks, Computer Communications and Networks, DOI:

10.1007/978-1-84882-218-4 10,p261-291

[74] D. J. Baker, A. Ephremides. The architectural organization of a mobile radio

network via a distributed algorithm[J]. IEEE Transactions on Communications, 1981,

29(11):1694-1701.

[75] K. Xu, M. Gerla, A heterogeneous routing protocol based on a new stable

clustering scheme[C]. Proceeding of IEEE Military Communications Conference

(MILCOM 2002), Anaheim, CA, October 2002.

[76] S. Bandyopadhyay, E. Coyle. An energy efficient hierarchical clustering

algorithm for wireless sensor networks[C]. Proceedings of the 22nd Annual Joint

Conference of the IEEE Computer and Communications Societies (INFOCOM 2003),

San Francisco, California, 2003.

[77]O. Younis, S. Fahmy.: A Hybrid, Energy-Efficient, Distributed clustering

approach for Ad Hoc sensor networks[J]. IEEE Transactions on Mobile Computing,

2004, 3(4):325-349.

[78]P. Ding, J. Holliday, A. Celik, Distributed energy efficient hierarchical

clustering for wireless sensor networks, in: Proceedings of the IEEE International

Conference on Distributed Computing in Sensor Systems(DCOSS’05), Marina Del

Rey, CA, June 2005.

[79]Heinzelman W,Chandarkasan A,Balakrishnan H.Energy-efficient

214
Communication Protocol for Wireless Microsensor Networks[C].In Proceeding of the

33rd Annual Hawaii Int’l conf. on System Science. Maui: IEEE Computer

Society,2000:3005-30

[80]M. Ettus. System Capacity, Latency, and Power Consumption in

Multihop-routed SS-CDMA Wireless Networks. In Radio and Wireless Conference

(RAWCON ’98), pages 55–58, Aug. 1998

[81]X. Lin and I. Stojmenovic. Power-Aware Routing in Ad Hoc Wireless

Networks. In SITE, University of Ottawa, TR-98-11, Dec. 1998

[82]T. Meng and R. Volkan. Distributed Network Protocols for Wireless

Communication. In Proc. IEEEE ISCAS, May 1998.

[83]Wendy Beth Heinzelman. Application-Specific Protocol Architecture for

Wireless Networks. PhD thesis, Department of Electrical Engineering and Computer

Science, Massachusetts Institute of Technology, 2000.

[84] Zuo Chen,Kai Chen,An improved Multi-hop Routing Protocol for

Large-Scale Wireless Sensor Network Based on Merging Adjacent

Clusters,JOURNAL OF SOFTWARE, VOL. 8, NO. 8, AUGUST 2013,pp2080-2085.

[85] B. Manzoora, N. Javaida, O. Rehmana, etl,Q-LEACH: A New Routing

Protocol for WSNs,The 4th International Conference on Ambient Systems, Networks

and Technologies (ANT 2013), the 3rd International Conference on Sustainable

Energy Information Technology (SEIT-2013),Volume 19, 2013, Pages 926–931

[86] Thiemo Voigt, Hartmut Ritter, Jochen Schiller,Adam Dunkels, and Juan

Alonso. Solar-aware Clustering in Wireless Sensor Networks. In Proceedings of the

Ninth IEEE Symposium on Computers and Communications, June 2004.

[87] Farooq, M.O, Dogar, A.B. ; Shah, G. MR-LEACH: Multi-hop Routing with Low Energy

Adaptive Clustering Hierarchy, Sensor Technologies and Applications (SENSORCOMM), 2010

Fourth International Conference on 18-25 July 2010

[88] Wu Xinhua ,Wang Sheng, Performance Comparison of LEACH and LEACH-C Protocols

by NS2, Distributed Computing and Applications to Business Engineering and Science (DCABES),

215
2010 Ninth International Symposium on 18-25 July 2010

[89] Sunkara Vinodh Kumar and Ajit Pal ,Assisted-Leach (A-Leach) Energy

Efficient Routing Protocol for Wireless Sensor Networks, International Journal of

Computer and Communication Engineering, Vol. 2, No. 4, July 2013

[90] Mustapha Khiati, Djamel Djenouri,BOD-LEACH: broadcasting over

duty-cycled radio using LEACH clustering for delay/power efficient dissimilation in

wireless sensor networks, International Journal of Communication Systems,24 SEP

2013

[91] JIANG Yang，SUN Liu-lin，AO Wen-jun，YUAN Min,Research on optimal

cluster-head number of LEACH routing protocol for WSN,Application Research of

Computers,Vol. 27 No. 11Nov． 2010

[92] Hu Junping, Jin Yuhui, Dou Liang. A Time-based Cluster-Head Selection

Algorithm for LEACH[J]. Computers and Communications,2008.ISCC 2008.IEEE

Symposium on, 2008,1172-1176

[93]. Jing Chen, Hong Shen. MELEACH-L: More Energy-efficient LEACH for

Large-scale WSNS[J]. Wireless Communication and Sensor Networks,2008.WCSN

2008.Fourth International Conference on, 2008, 1-6.

[94] Rajendran V，Obraczka K，Garcia-Luna-Aceves J．Energy-efficient，

collision-free medium access control for wireless sensor networks．Embedded

Networked Sensor Systems，November 2003，LosAngeles，CA．181-192

[95] Mortaza Fahimi Khaton Abad, Mohammad Ali Jabraeil Jamali,Modify

LEACH Algorithm for Wireless Sensor Network,IJCSI International Journal of

Computer Science Issues, Vol. 8, Issue 5, No 1, September 2011,ISSN (Online):

1694-0814

[96] Lan Tien Nguyen, Xavier Defago, Razvan Beuran,Yoichi Shinoda,An

Energy Efficient Routing Scheme for Mobile Wireless Sensor Networks,IEEE ISWCS

2008,pp568-572

[97] E. Bonabeau, M. Dorigo, and G. Theraulaz. Swarm Intelligence: From

216
Natural to Arifical Systems. New York：Oxford University Press，1999.

[98] Ajith Abraham, et.al. Swarm Intelligent Systems[M]. Studies in

Computational Intelligence,Volume 26: 3-25

[99] Kenndy J, Eberhart R.C. Paricle Swarm Optimization. In: IEEE International

Conference on Neural Networks, IV. Piscataway, NJ: IEEE Service

Center,1995.1942-1948.

[100] Shi, Y.; Eberhart, R.C. (1998). "A modified particle swarm optimizer".

Proceedings of IEEE International Conference on Evolutionary Computation.

pp. 69–73.

[101] Marco Dorigo, Vittorio Maniezzo, Alberto Colorni. The Antsystem:

Optimization by a Colony of Cooperating Agents [J]. IEEE Transactions on Systems,

Man, and Cybermetrics, 1996, 26 (1):1-13.

[102] Marco Dorigo, Luca Maria Gam bardella. Ant Colony System: A

Cooperative Learning Approach to the Traveling Salesman Problem [J]. IEEE

Transactions on Evolutionary Computation, 1997, 1 (1): 53-66.

[103]Marco Dorigo, Ant System: Optimization by a Colony of Cooperating

Agents,IEEE TRANSACTIONS ON SYSTEMS,MAN, AND CYBERNETICS-PART

B:CYBERNETICS,VOL,26,NO,1,February 1996

[104] Raghavendra V. Kulkarni, and Ganesh Kumar Venayagamoorthy, Particle

Swarm Optimization in Wireless Sensor Networks: A Brief Survey,IEEE

TRANSACTION ON SYSTEM,MAN AND CYBERNETICS-PART

C:APPLICATIONS AND REVIEW

[105] Xue Wang,Jun-Jie Ma, Sheng Wang,Dao-Wei Bi,Distributed Particle

Swarm Optimization and Simulated Annealing for Energy-efficient Coverage in

Wireless Sensor Networks,Sensors (Basel). 2007 May; 7(5): 628–648

[106] Santi P Silence is folden with high probability: maintaining a connected

backbone in wireless sensor networks [A].In; Proc. of 1st European Workshop on

Wireless Sensor Networks[C].Betlin,2004

217
 [107] J. Hu, J. Song, M. Zhang, and X. Kang, “Topology optimization for urban

traffic sensor network,” Tsinghua Science & Technology, vol. 13,Apr 2008.

[108] J. Li, K. Li, and W. Zhu, “Improving sensing coverage of wireless sensor

networks by employing mobile robots,” in Proceedings of the International Conference

on Robotics and Biomimetics (ROBIO), 2007,pp. 899–903.

[109] T. P. Hong and G. N. Shiu, “Allocating multiple base stations undergeneral

power consumption by the particle swarm optimization,” in Proceedings of the IEEE

Swarm Intelligence Symposium (SIS), 2007,pp. 23–28

[110]Xin Guan, Guan L., Wang X.G., Ohtsuki, A Novel Routing Algorithm based

on Ant Colony System for Wireless Sensor Networks[C]. Proceedings of 18th

International Conference on Computer Communications and Networks(ICCCN 2009),

2009: 1-5.

[111] Ziyadi, M, Yasami, K.,Abolhassani, B, Adaptive Clustering for Energy

Efficient Wireless Sensor Networks based on Ant Colony Optimization [C]. Seventh

Annual Communication Networks and Services Research Conference (CNSR '09),

2009: 330-334.

[112] Salehpour A.A, Mirmobin. B, Afzali-Kusha A, Mohammadi S. An Energy

Efficient Routing Protocol for Cluster-Based Wireless Sensor Networks Using Ant

Colony Optimization[C].International Conference on Innovations in Information

Technology (IIT 2008), 2008: 455-459.

[113] Ghasem Aghaei R. Rahman A.M. Rahman M.A, Gueaieb W. Ant

Colony-Based Many-to-One Sensory Data Routing in Wireless Sensor Networks[C].

International Conference on Computer Systems and Applications (AICCSA 2008),

2008: 1005-1010.

[114] http://www.rfc-base.org/rfc-2386.html
[115] Prasun Sinha, QoS Issues in Ad-Hoc Networks, Ad Hoc Networks,2005,

Publisher:Springer US, pp 229-247

[116] P.Gupta, P.R.Kumar. A System and traffic dependent adaptive routing

218
algorithm for ad hoc networks.The 36th Conference on Decision and Control.San

Diego,California,1997:2375-2380.

[117] Anelise Munaretto, Hakim Badis, Khaldoun Al Agha, Guy Pujolle, A

Link-state QoS Routing Protocol for Ad Hoc Networks, Mobile and Wireless

Communications Network, 2002. 4th International Workshop on 2002

[118] Shigang Chen,Klara N.Distributed quality-of-service routing in Ad hoc

networks.IEEE Journal on Selected Areas in Communications,1999,17(8):1488-1505.

[119]SHI JIAN, ZOULING, A QoS based distributed multicast routing

algorithm in Ad Hoc networks, JOURNAL OF CHINA INSTITUTE OF

COMMUNICATIONS, 2003, 24(6)

[120] Chenxi Zhu, M. ScottCorson. QoS routing for mobile ad hoc networks.

INFOCOM,2002,6(2): 958-967

[121] HUANG C, DA IF, WUJ. On-Demand location-aided QoS routing in ad

hoc networks. In：Proc．Of the Int’l Conf．On Parallel

Processing(ICPP)．2004:502—509.
[122] Sivakumar, R.; Sinha, P. ; Bharghavan, V.,CEDAR: a Core-Extraction Distributed Ad

hoc Routing algorithm, Selected Areas in Communications, IEEE Journal

on (Volume:17 , Issue: 8 )

[123] Prasun Sinha , Raghupathy Sivakumar , Vaduvur Bharghavan, Multicast

Core-Extraction Distributed Ad hoc Routing, In Proc. of the Wireless Communications

and Networking Conference, 1999

[124] Wang jianxin,et al,A quality of service routing protocol based on mobility

prediction in mobile ad hoc networks, Journal CSUT Vol 10, No1, 2003: 53—57.

[125] Li Layuan, Li Chunlin, A QoS multicast routing protocol for mobile ad-hoc

networks, Information Technology: Coding and Computing, 2005. ITCC 2005.

International Conference on 4-6 April 2005

[126] Wang Lei. Research and implementation of Linux data packet capture

technology based on Gigabit network. Ji'nan: Shandong University,2007

219
 [127] Chang Feng Meng Zhuanliang ,Based on ARM-Linux network driver

designs. Communication technology.2012.6(45):32-35.

[128]Yigitel M A，Ozlem Durmaz Incel，Cem Ersoy．QoS-aware MAC protocols

for wireless sensor networks：A survey『J]．Computer Networks，2011(SUPPL)：l-23．

[129] http://onlinelibrary.wiley.com/doi/10.1002/0470867787.ch6/summary.

[130] Bao L, Garcia-Luna-Aceves J J. A new approach to channel access

scheduling for ad hoc networks.[C] In: Proc 7th Annual Int’l Conference on Mobile

Computing and Networking,(MobiCOM 2001), Rome, Italy, July 16-21, 2001.

210~221

[131] Rajendran V, Obracazka K, Garcia-Luna-Aceves J J.Energy-efficient,

collision-free medium access control for wireless sensor networks[A]. Proc 1st Int’l

Annual Joint Confon Embedded Networked Sensor Systems (SenSys)[C],

Nov.5-7,2003, Los Angeles, CA. 181~192

[132]Lu G, Krishnamachari B, Raghavendra C. An adaptive energy-efficient and

low-latency MAC for data gathering in wireless sensor networks.[C] In: Proc 18th Int’l

Parallel and Distributed Processing Symp (IPDPS’04), April 26-30, 2004,Santa Fe,

New Mexico. 224~230

[133] IEEE Computer Society LAN MAN Standards Committee. IEEE Std

802.11-1999, Wireless LAN Medium Access Control (MAC) and Physical Layer

(PHY) specifications. 1999

[134]Langendoen K, Van Dam T. An adaptive energy-efficient MAC protocol for

wireless sensor networks.[C] In: Proc 1st Int’l Annual Joint Conf on Embedded

Networked Sensor Systems (SenSys), Nov. 5-7,2003, Los Angeles, CA. 2003

[135] Langendoen K, Van Dam T. An adaptive energy-efficient MAC protocol

for wireless sensor networks.[C] In: Proc 1st Int’l Annual Joint Conf on Embedded

Networked Sensor Systems (SenSys), Nov. 5-7,2003, Los Angeles, CA. 2003

[136]Jamieson K, Balakrishnan H, Tay Y C. Sift: A MAC protocol for

event-driven wireless sensor networks.[J],MIT-LCS-TR-894, 2003.

220
 [137] Lee, David Y. AT&T Bell Labs, Murray Hill, NJ, USA Yannakakis,

Mihalis.Principles and methods of testing finite state machines-a survey.Proceedings of

the IEEE.Aug 1996:1090-1123.

[138] Matthias Grossglauser, David N.C. Tse, Mobility Increases The Capacity of

Wireless Ad Hoc Networks. INFOCOM 2001. Twentieth Annual Joint Conference of

the IEEE Computer and Communications Societies. Proceedings. IEEE, Volume: 3,

2001. Page(s): 1360 -1369 vol.3.

[139]Schult N,Mirhakkak M,LaRocca D, Strater J, Routing in mobile Ad Hoc

networks, Military Communications Conference Proceedings, IEEE, Volume: 1,1999:

10 -14 vol.1.

[140] Wang, Yunbo Dept. of Comput. Sci. & Eng., Univ. of Nebraska - Lincoln,

Lincoln, NE, USA Vuran, Mehmet C.; Goddard, Steve.Cross-Layer Analysis of the

End-to-End Delay Distribution in Wireless Sensor Networks.Networking, IEEE/ACM

Transactions on. Feb.2012:305-318.

[141] Xing Yang, Li zhicun, Gao qian, A Delay guaranteed QoS routing protocol.

Microcomputer Information. 2012.5(28):127-129.

[142] Wang S Y,Liu J Y,Huang C C.Signal strength-based routing protocol for

mobile ad hoc networks, Proceedings of the 2005 IEEE International Conference on

Advanced Information Networking and Applications, 2005:17-20.

[143]Xiuqin Yi,Xiaotong Zhu.Application of Entropy Weight Coefficient Method

to the Evaluation of the profitability of listed firms.Computer Application and System

Modeling (ICCASM), 2010 International Conference on.Oct.2010:108-110.

[144]Jinbo Song, Yan Liu, Danrong Song.Multi-grade Fuzzy Comprehensive

Evaluation of BOT Projects Service Quality Based on Fuzzy Entropy Weight

Coefficient Method.2012 Fifth International Conference on Business Intelligence and

Financial Engineering (BIFE).Aug.2012:556-560.

[145]FengKong,Zhiguang Zhang,Ying Liu.Research on improvement of grey

relation analysis method based on ideal points.Computing, 2007. WiCom 2007.

221
International Conference on Wireless Communications, Networking and

Mobile.Sept.2007:5717-5720.

[146] CHANG Feng,MENG Chuan-liang,development of Network Driver

Program based on Embedded Linux,Communications Technology,2012, 45(6):32-35.

[147] Wang yang, Zhuchaoxia,Research on Linux Multithread Synchronization

Based on Semaphore, COMPUTER AND DIGITAL ENGINEERING,2004,4:20-22.

[148] Ye Dun-fan, Zhou Fe fan; Min Liang-liang.Design and Implementation of

High-Precision Timer in Linux.Computer Science and Information Engineering.2009

WRI World Congress on Volume:7 Digital Object

Identifier:10.1109/CSIE.2009.277.Publication Year:2009,Page(s): 341- 345.

[149] Sung-Ju Lee and Mario Gerla,A Simulation Study of Table-Driven and

On-Demand Routing Protocols for Mobile Ad Hoc Networks,IEEE Network

July/August 1999,p48-54

[150] G. Bianchi, “IEEE 802.11: saturation throughput analysis,” IEEE

Commun.Lett., vol. 2, pp. 318-320, Dec. 1998.

[151] G. Bianchi, “Performance analysis of the IEEE 802.11 distributed

coordination function,” IEEE J. Select. Areas Commun., vol. 18, pp. 535-547,Mar.

2000.

[152] Yang Xiao．Performance Analysis of IEEE802.1le EDCF Under Saturation

Condition[C] ．Proceedings of 2004 IEEE International Conference on

Communications．IEEE，June2004：170-174

[153] Zhu Lin,Research and Implementation OnWireless Router Based on

Embedded Linux,Chengdu University of Technology in Communications and

information technology, Master’s thesis,2009.

222
 Appendix1

Published papers

1.Hua Long, Qian Lei , Joan Lu ,A New RSSI-based Centroid Localization

Algorithm Using Virtual Reference Tags, INFOCOMP2013

2. Hua Long, Joan Lu, S .Chao, Jim Yip, Handoff algorithm improvement and its

analysis for mobile networks. In: Proceedings of the 2007 International Conference on

Communications. CSREA Press, pp. 66-71. ISBN 1-60132-030-2

3.Hua Long，Joan Lu, , Yip, J.; Hua Long Jiang, An Improved Access Scheme

Based on Multiple-Hop Priority and Cross-Layer Design in 802.11e , Wireless

Communications, Networking and Mobile Computing, 2008. WiCOM '08. 4th

International Conference

4.Hua Long, Joan Lu,Yubin Shao, Guantao and Yip, Yau Jim (2007) A Clustering

Algorithm for Block Mobility in Ad-hoc Network. Communications of SIWN, 5. ISSN

1757-4439

5.Hua Long, Joan Lu, Yubin Shao ,Guantao and Yip, Yau Jim ，A new backoff

scheme based on channel state for IEEE 802.11e WLANs，Proceedings of the fifth

annual meeting of China Institute of Communications,ISBN:9787121059261, PRESS：

Publishing House of electronics industry, 2008-2-1

6.Y Che, Hua Long, Joan Lu, YJ Yip,An investigation in self-similar study and

implication of 3G mobile traffic, 01/2007; In proceeding of: Proceedings of the 2007

International Conference on Communications in Computing, CIC 2007, June 25-28,

2007, Las Vegas, Nevada, USA,Source: DBLP

223
 Appendix2

kernel program

1．RBLEACH algorithm

clear all %

close all %

clc %

%% Simulation parameters

MC = 100; % Monte Carlo simulation times

%xm=100; %x

%ym=100; %y

xm=200; %

ym=xm; %

%sink.xd=0.5*xm;%

%sink.yd=0.5*ym;%

% Base stations are placed in the region boundary, is arranged in the

corner
sink.xd=0;%BS x

sink.yd=0;% BS y

N=400; % NUMBEROF SENSORS

p=0.1; % The percentage of the expected cluster head

Cc=0.6 % The fusion rate;

rmax=200; %Toll rounsa

Rmax=200 % the total number of rounds;

CM=32; % The size of the control information

CM=32;% control message size

224
 DM=4000; % Data size

DM=4000;% data size

E0=0.05; % Single node initial energy

The initial energy of single node E0=0.05;%

ETX=50*0.000000001; % Transport energy consumption（J/bit）

ETX=50*0.000000001;% transmission energy consumption (J/bit)

ERX=50*0.000000001; % Receive energy consumption（J/bit）

ERX=50*0.000000001;% acceptance of energy consumption (J/bit)

Efs=10*10^(-12); %（free space）Channel model（J/bit/m2）

Efs=10*10^ (-12);% free space (free space) channel model (J/bit/ meters square)

Emp=0.0013*10^(-12); %（multipath fading）Channel model（J/bit/m4）

Emp=0.0013*10^ (-12);% multipath attenuation (multipath fading) channel model

(4 J/bit/ m)

EDA=5*10^(-12); % Data fusion energy consumption（J/bit）

EDA=5*10^ (-12); energy consumption data fusion (J/bit)

do=sqrt(Efs/Emp); %Efs and EmpBoundary threshold

Do=sqrt (Efs/Emp); demarcation threshold%Efs and Emp, value is 84

Dead = zeros(1,rmax); % Death of nodes per round

Dead = zeros (1, Rmax);% the death of nodes for per round

CH_Num = zeros(1,rmax); % Cluster heads for per round

CH_Num = zeros (1, Rmax);% every wheel cluster characteristic

R_total = 0; % Survival Round

R_total = 0;% survival number of rounds

R_first = 0;% the first death round number

E_left = zeros (1, Rmax);% total residual energy per round

E_CH = zeros(1,rmax); % Each cluster head energy consumption

E_CH = zeros (1, Rmax);% each cluster head energy consumption

for mc = 1:MC
225
 %% The initialization of scene
figure(1)

hold off

for i=1:N

S(i).xd=rand(1,1)*xm;

S(i).yd=rand(1,1)*ym;

S(i).G=0;

S(i).E=E0;

S(i).type='N';

plot(S(i).xd,S(i).yd,'o');

hold on;

end

plot(sink.xd,sink.yd,'x');

Plot (sink.xd, sink.yd,'x');

network area

Partition%% network area

Partition%% network area

QN = 6; % The numbers of area

QN = 6;% number region division

QN = 6;% number region division

QN = 6;% number region division

QN_num=zeros (1, QN);% of the number of nodes in each region, survival

QN_num=zeros (1, QN);% of the number of nodes in each region, survival

QN_num=zeros (1, QN);% of the number of nodes in each region, survival

CN_num=zeros (1, QN);% of each region the number of cluster heads

CN_num=zeros (1, QN);% of each region the number of cluster heads

Rc=zeros (1, QN);% cluster head ratio of each region

Sqn=zeros (1, QN);% of the area of each region

226
R = xm/ (QN-1);% zone radius

R = xm/ (QN-1);% zone radius

R = xm/ (QN-1 zone radius);%

dd=R/10;

Dd=R/10;

Dd=R/10;

Dd=R/10;

% is calculated for each area

Sqn(1)=1/4*pi*R^2;

Sqn (1) =1/4*pi*R^2;

Sqn(QN)=xm*ym-1/4*pi*(QN*R-R)^2;

Sqn (QN) =xm*ym-1/4*pi* (QN*R-R) ^2;

for qn=2:QN-1

for qn=2:QN-1

Sqn(qn)=1/4*pi*(qn*R)^2-1/4*pi*(qn*R-R)^2;

Sqn (QN) =1/4*pi* (qn*R) ^2-1/4*pi* (qn*R-R) ^2;

end

end

% calculated for each node which belong to the area

for i=1:N

for i=1:N

qn = min(floor(distance(sink,S(i))/R),5);

QN_num(qn+1)=QN_num(qn+1)+1;

S(i).Q=qn+1;

end

end

% of each region the number of cluster heads

CN_num(1)=0;
227
CN_num (1) =0;

CN_num(2)=1;

CN_num (2) =1;

for qn=3:QN

for qn=3:QN

CN_num(qn)=floor(Sqn(qn)/Sqn(1));

CN_num (QN) =floor (Sqn (QN) /Sqn (1));

end

end

% painting area division map

alpha=pi/2:-pi/32:0;

Alpha=pi/2:-pi/32:0;

Xround=zeros(QN-1,length(alpha));

Xround=zeros (QN-1, length (alpha));

Yround=zeros(QN-1,length(alpha));

Yround=zeros (QN-1, length (alpha));

for qn=1:QN-1

for qn=1:QN-1

for i=1:length(alpha)

Xround(qn,i)=R*qn*sin(alpha(i));

Yround(qn,i)=R*qn*cos(alpha(i));

end plot(Xround(qn,:),Yround(qn,:),'r')

Plot (Xround (QN, Yround, (QN:),:),'r')

end

end

he cluster head ratio of each area%

Disp ('each region of the cluster head ratio');

disp ('region of 1: 0');

228
 Disp (0 'region of 2: (fixed 1 cluster head)');

for r=1:rmax

figure(1);

hold off; % Round picture redrawn

plot(sink.xd,sink.yd,'x');% Draw the base station

hold on;

for qn=1:QN-1

plot(Xround(qn,:),Yround(qn,:),'r')

end

dead=0; % The initial death nodes 0

end

%% Death statistics node

QN_num=tttemp;

for i=1:N

if(S(i).E<=0) % If the energy is equal to 0

S(i).type='D';

plot(S(i).xd,S(i).yd,'r.');

QN_num(S(i).Q)=QN_num(S(i).Q)-1; % A surviving node is

calculated for each region

dead=dead+1; % Cumulative mortality nodes

if(dead==1) % If the node is the first death occurred

if(flag_first_dead==0)

first_dead=r; % Record the first round of death

node

flag_first_dead=1;

end

end

else
229
 S(i).type='N';

plot(S(i).xd,S(i).yd,'o');

end

endflag_first_dead=0;%

Flag_first_dead=0;% the first death node flag variables

Cid=zeros(2/p,rmax);

Cid=zeros (2/p, Rmax);

Cnum=zeros(2/p,rmax);

Cnum=zeros (2/p, Rmax);

tttemp=QN_num;

Tttemp=QN_num;

for r=1:rmax

for r=1:rmax

figure(1);

Figure 1.

Round picture redrawnHold off;% every wheel picture re

Plot (sink.xd, sink.yd,'x');% drawing base station

hold on;

Hold on;

for qn=1:QN-1

For qn=1:QN-1

plot(Xround(qn,:),Yround(qn,:),'r')

Plot (Xround (QN, Yround, (QN:),:),'r')

end

End

dead=0; %

Dead=0;% initial death node number is 0

statistical dead node%%

230
 QN_num=tttemp;

QN_num=tttemp;

for i=1:N

for i＝1：N

if(S(i).E<=0) %

If (S (I).E<=0)% if the energy is less than or equal to 0

S(i).type='D';

S (I).Type='D';

plot(S(i).xd,S(i).yd,'r.');

Plot (S (I).Xd, S (I).Yd,'r.');

QN_num(S(i).Q)=QN_num(S(i).Q)-1; %计算每个区域的存活节点

QN_num (S (I) =QN_num (.Q) S (I).Q) -1;% of each region was calculated

survival nodes

Dead=dead+1;% cumulative death of nodes

If (dead==1)% if it is the first death of nodes

if(flag_first_dead==0)

If (flag_first_dead==0)

First_dead=r;% records first appeared dead node sequence

flag_first_dead=1;

Flag_first_dead=1;

end

end

end

end

else

Else

S(i).type='N';

S (I).Type='N';
231
plot(S(i).xd,S(i).yd,'o');

Plot (S (I).Xd, S (I).Yd,'o');

end

end

Dead (R) = Dead (R + dead/MC);% record each round of death of nodes

If (dead > = 0.95*N)% of all death exit the loop node

break;

Break;

end

End

%% each region to select cluster head -Leach protocol

% no cluster head region 1, region 2 only 1 cluster head

cluster=0;

Cluster=0;

CN_num=zeros(1,QN);

CN_num=zeros (1, QN);

CN_num(1)=0;

CN_num (1) =0;

CN_num(2)=1;

CN_num (2) =1;

Ptemp=floor(QN_num(2)*rand)+1;

Ptemp=floor (QN_num (2) *rand) +1;

Temp=0;

C2_id=0;

C2_id=0;

for i=1:N

For i=1:N

if(S(i).type ~= 'D' && S(i).Q == 2)

232
 If (S (I).Type ~ ='D'& & S (I).Q = = 2)

% 2 regional cluster head selection

temp=temp+1;

Temp=temp+1;

if (Ptemp == temp)

If (Ptemp = = temp)

cluster=cluster+1;

Cluster=cluster+1;

S(i).type='C';

S (I).Type='C';

C (cluster).Xd=S (I).Xd;% will this node as cluster head logo

C(cluster).yd=S(i).yd;

C (cluster).Yd=S (I).Yd;

C(cluster).Next=0;

C (cluster).Next=0;

Plot (S (I).Xd, S (I).Yd,'k*');% mapping of this cluster head

C(cluster).distance=distance(sink,S(i));

C (cluster).Distance=distance (sink, S (I));

C (.Id=i cluster);% this cluster head ID

C2_id=i;

C2_id=i;

C(cluster).Q=S(i).Q;

C (cluster).Q=S (I).Q;

Packet_To_Bs (=1 cluster);% and transmitted to the base station data packet

number 1

end

End

end
233
 End

% of other regional cluster head selection

if(S(i).type ~= 'D' && S(i).Q > 2)

If (S (I).Type ~ ='D'& & S (I).Q > 2)

p=Rc(S(i).Q);

P=Rc (S (I).Q);

if(rand<= p)

If (rand<= P)

cluster=cluster+1;

Cluster=cluster+1;

S(i).type='C';

S (I).Type='C';

C (cluster).Xd=S (I).Xd;% will this node as cluster head logo

C(cluster).yd=S(i).yd;

C (cluster).Yd=S (I).Yd;

C(cluster).Next=0;

C (cluster).Next=0;

Plot (S (I).Xd, S (I).Yd,'k*');% mapping of this cluster head

C(cluster).distance=distance(sink,S(i));

C (cluster).Distance=distance (sink, S (I));

C (.Id=i cluster);% this cluster head ID

C(cluster).Q=S(i).Q;

CN_num(S(i).Q)=CN_num(S(i).Q)+1;

packet_To_Bs(cluster)=1;% The number of packets sent to

the base station for 1

end

end

end
234
 CH_Num(r)= CH_Num(r) + cluster/MC; %

for i=1:N

if(S(i).type=='N'&&S(i).E>0)%

if (S(i).Q == 1)

Er1=0;

Et1=ETX*DM+Efs*DM*distance(S(i),sink)^2; % Free

space fading

elseif (S(i).Q == 2)

Er1=ERX*CM;

Et1=ETX*(CM+DM)+Efs*(CM+DM)*distance(S(i),S(C2_id))^2; %By the spatial

fading else

Er1=ERX*CM*CN_num(S(i).Q)*0.5;

min_dis = 999999;

Pcc=0;

for cc=1:cluster

if (C(cc).Q==S(i).Q)

if (distance(C(cc),S(i))<min_dis)

min_dis=distance(C(cc),S(i));

Pcc=cc;

end

end

end

if(min_dis==999999)

Et1=ETX*DM+Efs*DM*distance(S(i),sink); %

elseif (min_dis>do)

packet_To_Bs(Pcc)=packet_To_Bs(Pcc)+1;

235
Et1=ETX*(CM+DM)+Emp*(CM+DM)*min_dis^4; %

else

packet_To_Bs(Pcc)=packet_To_Bs(Pcc)+1;

Et1=ETX*(CM+DM)+Efs*(CM+DM)*min_dis^2; %

end

end

S(i).E=S(i).E-Er1-Et1*0.8;% Residual energy

end

end

%% Computing cluster head routing - greedy algorithm

for c1=1:cluster

if (C(c1).Q>2)

min_dis=9999;

pmin=0;

flag=1;

qq = C(c1).Q;

while (flag)

qq = qq-1;

for c2=1:cluster

if (C(c2).Q==qq)

if (distance(C(c2),C(c1))<min_dis &&

distance(C(c2),C(c1))<0.8*distance(sink,C(c1)))

min_dis=distance(C(c2),C(c1));

pmin=c2;

end

end

end
236
 if (min_dis ~= 9999)

plot([C(pmin).xd C(c1).xd],[C(pmin).yd

C(c1).yd],'g');

C(c1).Next=pmin;

if (qq==2)

packet_To_Bs(pmin)=packet_To_Bs(pmin)+packet_To_Bs(c1)/2;

else

packet_To_Bs(pmin)=packet_To_Bs(pmin)+packet_To_Bs(c1)/10;

end

flag=0;

elseif (qq==2)

flag=0;

end

end

end

end

for cc=1:cluster

if (C(cc).Next == 0)

plot([C(cc).xd sink.xd],[C(cc).yd sink.yd],'g');

end

end

%% Calculate the residual energy of cluster head

for c=1:cluster

CEr1=ERX*CM*(packet_To_Bs(c)-1);% Cluster members to

receive information about energy consumption

CEr2=ERX*DM*(packet_To_Bs(c)-1);% Cluster members received information on

237
energy consumption data CEt1=ETX*CM+Efs*CM*R*R;% cluster

head broadcast energy clusters if (C(c).Next == 0)

CEt2=(ETX+EDA)*DM*packet_To_Bs(c)+Efs*DM*cc*packet_To_Bs(c)*dd^2; %

Decline of free space

else

CEt2=(ETX+EDA)*DM*packet_To_Bs(c)+Efs*DM*cc*packet_To_Bs(c)*dd^2; %

Decline of free space

end

S(C(c).id).E=S(C(c).id).E-CEr1-CEr2-CEt1-CEt2;%此

E_CH(r) = E_CH(r) +(CEr1+CEr2+CEt1+CEt2)/MC;

end

for i=1:N

if (S(i).type=='N' && S(i).E>0 )

E_left(r) = E_left(r) + S(i).E/MC;

end

end

end

R_total = R_total + r/MC;

R_first = R_first + first_dead/MC;

end

save RcLeach Dead CH_Num R_total R_first E_left E_CH

2.PSO-LEACH

%% Cluster head selection algorithm - PSO -Leach

clear all
238
close all

clc

%% simulation parameters

%xm=100;%x axis range

%ym=100;%y axis range

% because of the larger values of do, modify the range

Xm=200;%x axis range

Ym=200;%y axis range

% sink.xd = 0.5 * xm;% base x-axis

% sink.yd = 0.5 * ym;% y-axis base

% Stations are generally on regional boundaries, located on the corner

sink.xd = 0;% base x-axis

sink.yd = 0;% y-axis base

N = 200;% number of sensors

p = 0.1;% expected percentage of cluster head

cc = 0.6;% fusion rate

rmax = 1000;% of total number of rounds

CM = 32;% the size of the control information

DM = 4000;% data size

E0=0.05; % The initial energy of single node

ETX=50*0.000000001; % Transport energy consumption（J/bit）

ETX=50*0.000000001;% transmission energy consumption (J/bit)

ERX=50*0.000000001; % Acceptable energy consumption（J/bit）

ERX=50*0.000000001;% acceptance of energy consumption (J/bit)

Efs=10*10^(-12); % free space’s Channel model（J/bit/M）

Efs=10*10^ (-12);% free space (free space) channel model (J/bit/ meters square)

Emp=0.0013*10^(-12); % multipath fading’s Channel model（J/bit/米的4次方）

Emp=0.0013*10^ (-12);% multipath attenuation (multipath fading) channel model

239
(4 J/bit/ m)

EDA=5*10^(-12); % Data fusion energy consumption（J/bit）

EDA=5*10^ (-12); energy consumption data fusion (J/bit)

do=sqrt(Efs/Emp); %Efs和Emp’ Boundary threshold，值越为84

Do=sqrt (Efs/Emp); demarcation threshold%Efs and Emp, value is 84

%% scene initialization

figure(1)

for i=1:N

S(i).xd=rand(1,1)*xm;

S(i).yd=rand(1,1)*ym;

S(i).G=0;

S(i).E=E0;

S(i).type='N';

plot(S(i).xd,S(i).yd,'o');

hold on;

end

plot(sink.xd,sink.yd,'x');

%% Each round of iteration

flag_first_dead=0;% The first sign of the variable node death

Cid=zeros(2/p,rmax);

Cnum=zeros(2/p,rmax);

for r=1:rmax

hold off; % Round picture redrawn disp(r)

figure(1);

plot(sink.xd,sink.yd,'x');%

hold on;

dead=0; %

%% Death statistics node

240
 for i=1:N

if(S(i).E<=0) %

S(i).type='D';

plot(S(i).xd,S(i).yd,'r.');

dead=dead+1; % Cumulative mortality nodes

if(dead==1) %

if(flag_first_dead==0)

first_dead=r; %

flag_first_dead=1;

end

end

else

S(i).type='N';

plot(S(i).xd,S(i).yd,'o');

end

end

Dead(r)=dead; %

if(dead == N) % All died out of the loop node

break;

end

%%select head of cluster

%f the number of rounds is precisely an integer multiple of a cycle, is set

S1(i).G=0

if(mod(r-1, round(1/p) )==0)

for i=1:N

S(i).G=0;

end
241
 end

[C,S]=GroupHead(S,N,dead,sink,p,r);

num_cluster=length(C); %

CH_Num(r)=length(C); %

%% Assigned to each cluster head node and compute the energy

consumption of cluster members

for i=1:N

if (S(i).type ~= 'D')

if num_cluster >0

min_dis=distance(S(i),sink);

min_dis_cluster=0;

for c=1:num_cluster

temp=distance(S(i),C(c));

if(temp<min_dis)

min_dis=temp;

min_dis_cluster=c;

end

end

if (min_dis_cluster>0)

C(min_dis_cluster).ptb=C(min_dis_cluster).ptb+1; %

This node to join the cluster head plus a number of packets

Er1=ERX*CM*(num_cluster+1); % This cluster head node receives

control information of each energy else

min_dis=distance(S(i),sink);

Er1=0;

end

else

min_dis=distance(S(i),sink);
242
 Er1=0;

end

% This node sends a join message and transmit energy

consumption data

if (min_dis>do)

Et1=ETX*(CM+DM)+Emp*(CM+DM)*min_dis^4; %

else

Et1=ETX*(CM+DM)+Efs*(CM+DM)*min_dis^2; %

Multipath fading

end

S(i).E=S(i).E-Er1-Et1; %

Residual energy

end

end

%% Cluster head routing decisions made for each cluster head to the base

station of the information required for the path

C=HeadRoute(C,sink);

%% Computing cluster heads energy consumption

for c=1:num_cluster

CEr1=ERX*CM*(C(c).ptb-1); %

Cluster members to receive information about energy consumption

CEr2=ERX*DM*(C(c).ptb-1); %

Cluster members received information on energy consumption data

if (N-dead<=num_cluster)

CEt1=0;

else
243
CEt1=ETX*CM+Efs*CM*(sqrt(xm*ym))*(sqrt(xm*ym)); %

Information broadcast cluster head energy clusters

end

% headof cluster sent to the base station after fusion energy

if (min_dis>do)

CEt2=(ETX+EDA)*DM*C(c).ptb+Emp*DM*cc*C(c).ptb*C(c).distance^4; %

Multipath fading

else

CEt2=(ETX+EDA)*DM*C(c).ptb+Efs*DM*cc*C(c).ptb*C(c).distance^2; %

Decline of free space

end

S(C(c).id).E=S(C(c).id).E-CEr1-CEr2-CEt1-CEt2; % After this

round of the node residual energy

Cid(c,r)=C(c).id;

Cnum(c,r)=C(c).ptb;

end

end

244