energy efficient architecture and protocols for cognitive...
TRANSCRIPT
Energy Efficient Architecture and Protocols for Cognitive Radio
based Sensor Networks
Thesis submitted in partial fulfilment
of the requirements for the degree of
Master of Science (by Research)
in
Electronics and Communication Engineering
by
SandhyaSree Thaskani
200832003
Communication Research Center
International Institute of Information Technology
Hyderabad - 500 032, INDIA
January 2011
©Copyright SandhyaSree Thaskani, 2011
All Rights Reserved
International Institute of Information Technology
Hyderabad, India
CERTIFICATE
It is certified that the work contained in this thesis, titled “Energy Efficient
Architecture and Protocols for Cognitive Radio based Sensor Networks” by
SandhyaSree Thaskani, has been carried out under my supervision and is not
submitted elsewhere for a degree.
_______________
Date
________________________
Adviser: Prof. G Rama Murthy
iv
Acknowledgment
I would like to acknowledge and extend my heartfelt gratitude to my advisor, Dr. G. Rama
Murthy for his valuable support and guidance. He has always given me the opportunity to
explore different areas of research. His enthusiasm for research and open mindedness has
been a great source of help for me.
I am thankful to my parents and friends for their moral support.
v
Abstract
Wireless Sensor Networks (WSNs) has been a research topic for more than a decade, and the
range of potential applications has spanned beyond the military domain into commercial
domains such as industrial/building/home automation, lighting control, energy management,
to name a few. Emerging wireless technologies such as Zigbee and IEEE 802.15.4 have
enabled the development of interoperable commercial products, which is important to meet
the scalability and low cost requirements. A key feature for current WSN solutions is
operation in unlicensed frequency bands, for instance, the worldwide available 2.4 GHz band.
However, the same band is shared by other very successful wireless applications, such as Wi-
Fi and Bluetooth, as well as other proprietary technologies. There is evidence that unlicensed
spectrum is becoming overcrowded. As a result, coexistence issues in unlicensed bands have
been subject of extensive research, and in particular, it has been shown that IEEE 802.11
networks can significantly degrade the performance of Zigbee/802.15.4 networks when
operating in overlapping frequency bands. Till now WSNs are working in fixed spectrum
allocation strategy. Using this strategy makes WSNs to interfere with other technologies in
the same band.
Studies sponsored by the Federal Communications Commission (FCC) pointed out that the
current static spectrum allocation has led to overall low spectrum utilization where up to 70%
of the allocated spectrum remains unused (called white space) at any one time even in a
crowded area. The white space is defined by time, frequency and maximum transmission
power at a particular location. Consequently, Dynamic Spectrum Access (DSA) has been
proposed so that unlicensed spectrum users or Secondary Users (SU)s are allowed to use the
white space of licensed users or Primary Users (PU)s spectrum, conditional on the
interference to the PU being below an acceptable level. This function is realized using
Cognitive Radio (CR) technology that enables an SU to change its transmission and reception
parameters including operating frequencies. There are two prominent features of CR. Firstly,
sensing is performed across a wide range of spectrum to identify the white space. Secondly,
data packets are allocated opportunistically to the white space at different channels. This
means that whenever a PU accesses a channel which has been regarded as white space by the
SUs, the SUs must vacate the spectrum as soon as possible.
vi
Cognitive Radio based Sensor Networks (CRSN) are introduced to use DSA. In this networks
sensor node is aided with cognitive radio. In CRSN SUs are WSN. WSN operates on battery
power. Once they drain out of batter it is very difficult to replace them. Hence energy
efficiency is more crucial in CRSN.
This thesis addresses the problem of reducing energy consumption in Cognitive Radio based
Sensor Networks (CRSN). For energy efficiency in CRSN, a spectrum aware architecture is
proposed. In proposed architecture spectrum sensing network is decoupled from data
aggregating network.
Data gathering network is Wireless Sensor Networks (WSN). For data gathering Network
energy efficient routing protocol and TDMA based MAC (Medium Access Control) protocol
is proposed. In addition to this, a Fault Repair Algorithm using Localization and Controlled
mobility to fill Network Holes is proposed. A cross layer design for routing and MAC layer,
by using a light weight token-packet passing is proposed. A routing protocol by placing
uniform virtual grid for topology aware WSN is also proposed. All these protocols are shown
energy efficient when compared to existing protocols. On top of this, energy efficient
network architecture for WSN using a novel Non-Uniform Sampling technique is proposed.
vii
Contents
Acknowledgement iv
Abstract v
1. Introduction 1
1.1 Motivation 2
1.2 Our Approach 2
1.3 Thesis Organization 3
2. Background 4
2.1 Wireless communications systems 4
2.2 Cognitive Radio based Networks 4
2.3 Cognitive Radio Based Sensor Networks 7
2.4 Wireless Sensor Networks 8
2.4.1 Design Factors 9
3. Energy Efficient Architecture for CRSN and its applications 12
3.1 Introduction 12
3.2 Background for CRSN Architecture 13
3.2.1 CRSN Node Structure 13
3.2.2 CRSN Topology 13
3.3 Proposed Energy Efficient Architecture For CRSN 14
3.4 Applications of CRSN 17
3.4.1 CRSN in Telemedicine 17
3.4.2 CRSN in Under water acoustic sensor networks 19
3.4.3 CRSN in Precision agriculture 21
4. Energy efficient virtual grid placement based Routing algorithm 23
4.1 Introduction 23
4.2 Issues Addressed and Underlined Assumptions 24
4.3 Proposed algorithm 24
4.3.1 Setup phase 25
4.3.2 Routing phase 26
4.3.3 Cluster Rotation 28
4.4 Performance Evaluation 29
4.4.1 Network Model 29
4.5 Conclusions 30
5. Mobility Tolerant TDMA Based MAC Protocol for WSN 31
5.1 Introduction 31
5.2 Background 32
viii
5.3 Problem Statement 34
5.4 Assumptions 35
5.5 Proposed Algorithm 36
5.6 Simulation Results 39
5.7 Conclusions 42
6. Energy Efficient cross-layer protocol design by using token passing mechanism 43
6.1 Introduction 43
6.2 Background 44
6.3 Assumptions & Definitions 45
6.4 Proposed Algorithm 46
6.4.1 Sectroid (Group) Formation 46
6.4.2 Token Passing 46
6.4.3 Routing 48
6.5 Simulation Evaluation 49
6.6 Conclusions 51
7. Fault Repair Algorithm using Localization and Controlled mobility 52
7.1 Introduction 52
7.2 Background 53
7.3 Problem formulation 55
7.4 Assumptions and definitions 57
7.5 Proposed Algorithm 58
7.6 Simulation Results 60
7.7 Conclusions 61
8. Sensor Node deployment using an Innovative Non-Uniform Sampling phenomenon 62
8.1 Motivation 62
8.2 Introduction 62
8.3 Motivation for Non-Uniform Sampling 63
8.4 An Approach to Non-Uniform Sampling 63
8.4.1 First engineering approach 66
8.4.2 Second engineering approach 70
8.5 Non-Uniform Sampling: Information -Theoretic Approach 70
8.6 Conclusion 71
9. Conclusions 72
9.1 Future Work 73
Bibliography 74
Related Publications 79
ix
List of Figures
2.1 Spectrum utilization 5
2.2 Spectrum holes 6
2.3 Sensor Nodes scattered in a sensor field 9
2.4 Sensor Node Hardware 11
3.1a Cognitive radio sensor network (CRSN) architecture 14
3.1b possible network topologies or clustered CRSN 14
3.1c a heterogeneous hierarchical CRSN 14
3.2 Hardware structure of a cognitive radio sensor node 15
3.3 network architecture of cognitive Radio Based Sensor Network 16
3.4 Three types of services provided by Sensor Telemedicine Networks 18
3.5 Three types of services provided by Sensor Telemedicine Networks 19
3.6 Cognitive Radio based Sensor Networks in UW-ASN 21
4.1 sensor field at the end of setup phase 27
4.2 Routing of packet from event node to Base Station 28
4.3 Plot of Nodes VS Network lifetime 30
5.1 Schematic of the proposed algorithm 35
5.2 Graph for Delay 40
5.3 Graph for Network Lifetime 41
5.4 Graph for Channel Utilization (I) 41
5.5 Graph for Channel Utilization (II) 42
6.1 Frame structure in our protocol 45
6.2 Part of sensor field divided into sectroids 47
6.3 Schematic of Token Passing Procedure 48
6.4 Schematic of the Proposed Algorithm 49
6.5 Network Life Time 50
6.6 Average number of Redundant Messages 51
7.1 Illustration of control mobility 54
7.2 Model of localized sensor field 58
7.3 Sensor Field with both static and mobile nodes 59
7.4 Plot of Proposed Algorithm VS Levelling 61
7.5 Plot of Proposed Algorithm VS Flooding 61
8.1 Uniform Sampling, grid with bin distance „b‟ 66
8.2 Coarse partition of f(x) 69
1
Chapter 1
Introduction
The research area of Wireless sensor networks (WSN) is an advancing technology which has
opened new research issues. The wireless sensor network comprises of number of sensor nodes
which are equipped with a radio transceiver. The wireless sensor nodes are constrained in energy
supply, computational memory space, speed and bandwidth. In spite of limitations sensor nodes
can reliably and accurately report the surrounding environmental changes.
WSN uses wireless technologies like ZigBee [3] and 802.15.4[4]. These wireless technologies
use unlicensed frequency spectrum such as 433 MHz, 868 MHz (Europe), 915 MHz (North
America), and the 2.45-GHz Industrial-Scientific-Medical (ISM) band. These unlicensed bands
are crowed with other wireless applications, such as Wi-Fi, Bluetooth etc. As a result, coexistence
issues in unlicensed bands have been subject of extensive research [5] [6], and in particular, it has been
shown that IEEE 802.11 networks [7] can significantly degrade the performance of Zigbee/802.15.4
networks when operating in overlapping frequency bands [6]. Hence it is well evident that this
spectrum is over crowed. Till now WSNs are working in fixed spectrum allocation strategy.
Using this strategy makes WSNs to interfere with other technologies in the same frequency band.
According to Federal Communications Commission (FCC) [2], the current static spectrum
allocation has led to overall low spectrum utilization where up to 70% of the allocated spectrum
remains unused (called white space) at any one time even in a crowded area. Dynamic Spectrum
Allocation (DSA) has been proposed so that unlicensed spectrum users or Secondary Users (SU)
s are allowed to use the white spaces of licensed users or Primary Users (PU) s spectrum, with
low interference with PUs. This function can be realized by implementing Cognitive Radio (CR)
in SUs.
In our scenario WSN are SUs. To realize DSA function in WSN, Cognitive Radio based Sensor
Networks (CRSN) has been proposed [1]. CRSN is similar to WSN, but with one difference. In
CRSN radio is CR where as in WSN radio is a normal radio. This CR enables CRSN to sense
spectrum holes and to dynamically switch its parameters to available white space.
2
1.1 Motivation:
One of the most critical issues on the sensor network is efficient energy consumption. Since
sensor nodes play the dual role as both event detectors and data routers, a sensor node draining of
its battery may cause the partition of network in some network topology or make some sensing
area uncovered. In most application scenarios, the replacement of power resources might be
impossible due to a large number of sensor nodes and the difficulty of accessing the sensing area,
so the lifetime of the sensor node shows the strong dependence on the battery lifetime. Sensor
nodes consume mostly their batteries for transmitting and receiving packets. To make the
lifetime of the sensor network longer, in this thesis we proposed energy efficient protocols and
architecture.
1.2 Our Approach:
In CRSN quickly drains out of battery, as sensor node has to sense the spectrum all the time.
Hence, to have better energy efficiency in CRSN we proposed an energy efficient architecture
for CRSN. In this approach, we decouple CRSN in spectrum sensing network and data gathering
network. Spectrum sensing network has the capability to sense spectrum. Data gathering network
has the capability to monitor the field depending on its application (e.g., temperature monitoring)
and data transmission.
Data gathering network is Wireless Sensor Networks (WSN). For data gathering Network energy
efficient routing protocol and TDMA based MAC (Medium Access Control) protocol is
proposed. In addition to this, a Fault Repair Algorithm using Localization and Controlled
mobility to fill Network Holes is proposed. A cross layer design for routing and MAC layer, by
using a light weight token-packet passing is proposed. All these protocols are shown energy
efficient when compared to existing protocols. On top of this, energy efficient network
architecture for WSN using a novel Non-Uniform Sampling technique is proposed.
3
1.3 Thesis Organization:
Thesis has been organized as follows. Chapter2 gives the required background for this thesis. In
chapter3, we proposed an energy efficient architecture for CRSN and applications of CRSN.
In chapter 4, we have proposed a novel routing protocol for topology aware WSN. The key idea
of this routing protocol is to place virtual grid and make clusters. This routing protocol has better
energy efficiency when compared to existing protocols. This has been shown through
simulations.
In chapter 5, we have proposed an energy efficient Medium Access Control (MAC) protocol.
This MAC protocol is a TDMA based protocol. This protocol has shown better energy efficiency
when compared to existing protocols. This has been shown through simulations.
In chapter 6, we have proposed a cross-layer protocol. We have proposed combined routing and
MAC protocol. This protocol uses a light weight token packet for channel utilization and
forwarding a data packet. Through simulations we have shown that this protocol has better
energy efficiency when compared to existing protocols.
In chapter 7, we have proposed a Fault Repair Algorithm. This algorithm uses controlled
mobility of some nodes to fill in the holes formed in the network using localization techniques.
This ensures energy efficiency; it has been shown through simulations.
In chapter 8, we have proposed an innovative Non-Uniform Sampling technique in node
deployment. This thesis is concluded with chapter 9.
4
Chapter 2
Background
2.1 Wireless communications systems:
Wireless communications technology has become a key element in modern society. In our daily
life, devices such as garage door openers, TV remote controllers, cellular phones, personal
digital assistants (PDAs), and satellite TV receivers are based on wireless communications
technology. Today the total number of users subscribing to cellular wireless services has
surpassed the number of users subscribing to the wired telephone services. Besides cellular
wireless technology, cordless phones, wireless local area networks (WLANs), and satellites are
being extensively used for voice- as well as data-oriented communications applications and
entertainment services.
The increasing demand for wireless communications in consumer electronics applications and
personal high-data-rate networks indicate a promising commercial potential. The number of
devices based on multiple wireless standards and technologies are substantially growing but new
problems will arise due to limited availability of radio spectrum. We are, however now at a stage
where the identified problems have to be addressed to enable further growth of these promising
markets and to find a substantial basis for our future information society.
This chapter intends to provide introduction to Cognitive Radio based Networks; Cognitive
Radio based Sensor Networks and wireless sensor networks.
2.2 Cognitive Radio based Networks:
Cognitive radio is a new paradigm of designing wireless communications systems which aims to
enhance the utilization of the radio frequency (RF) spectrum. The term, cognitive radio, can
formally be defined as follows [12]:
A „„Cognitive Radio‟‟ is a radio that can change its transmitter parameters based on interaction
with the environment in which it operates
5
The motivation behind cognitive radio is the scarcity of the available frequency spectrum,
increasing demand, caused by the emerging wireless applications for mobile users. Most of the
available radio spectrum has already been allocated to existing wireless systems, however, and
only small parts of it can be licensed to new wireless applications. Nonetheless, a study by the
Spectrum Policy Task Force (SPTF) of the Federal Communications Commission (FCC) has
showed that some frequency bands are heavily used by licensed systems in particular locations
and at particular times, but that there are also many frequency bands which are only partly
occupied or largely unoccupied [8]. Fig. 2.1 illustrates the spectrum usage. For example,
spectrum bands allocated to cellular networks in the USA [9] reach the highest utilization during
working hours, but remain largely unoccupied from midnight until early morning.
Fig. 2.1 Spectrum utilization
In traditional spectrum allocation based on the command-and control model, where the radio
spectrum allocated to licensed users is not used, it cannot be utilized by unlicensed users and
applications [10]. Due to this static and inflexible allocation, legacy wireless systems have to
operate only on a dedicated spectrum band, and cannot adapt the transmission band according to
the changing environment. For example, if one spectrum band is heavily used, the wireless
6
system cannot change to operate on another more lightly used band. Licensing spectrum leads to
inefficient use of the radio spectrum.
Due to the current static spectrum licensing scheme, spectrum holes or spectrum opportunities
(Fig 2.2) arise. Spectrum holes are defined as frequency bands which are allocated to, but in
some locations and at sometimes not utilized by, licensed users (primary users), and, therefore,
could be accessed by unlicensed users (secondary users) [11].
Fig. 2.2 Spectrum holes
The cognitive radio concept was first introduced in [13, 14], where the main focus was on the
radio knowledge representation language (RKRL) and how the cognitive radio can enhance the
flexibility of personal wireless services. The cognitive radio is regarded as a small part of the
physical world to use and provide information from environment.
The ultimate objective of the cognitive radio is to obtain the best available spectrum through
cognitive capability and re-configurability. Since most of the spectrum is already assigned, the
most important challenge is to share the licensed spectrum without interfering with the
transmission of other licensed users as illustrated in Fig. 2.2. The cognitive radio enables the
usage of temporally unused spectrum, which is referred to as spectrum hole or white space [11].
If this band is further used by a licensed user, the cognitive radio moves to another spectrum hole
7
or stays in the same band, altering its transmission power level or modulation scheme to avoid
interference as shown in Fig. 2.2.
2.3 Cognitive Radio Based Sensor Networks:
The capabilities of cognitive radio may provide many of the current wireless systems with
adaptability to existing spectrum allocation in the deployment field, and hence improve overall
spectrum utilization. Among many others, these features can also be used to meet many of the
unique requirements and challenges of wireless sensor networks (WSN), which are, traditionally,
assumed to employ fixed spectrum allocation and characterized by resource constraints in terms
of communication and processing capabilities of low-end sensor nodes. In fact, a WSN
comprised of sensor nodes equipped with cognitive radio may benefit from the potential
advantages of the salient features of dynamic spectrum access such as:
a. Opportunistic channel usage for bursty traffic: Upon the detection of an event in WSN,
sensor nodes generate traffic of packet bursts. At the same time, in densely deployed
sensor networks, a large number of nodes within the event area try to acquire the channel.
This increases probability of collisions, and hence, decreases the overall communication
reliability due to packet losses leading to excessive power consumption and packet delay.
Here, sensor nodes with cognitive radio capability may opportunistically access to
multiple alternative channels to alleviate these potential challenges.
b. Dynamic spectrum access: In general, the existing WSN deployments assume fixed
spectrum allocation. However, WSN must either be operated in unlicensed bands, or a
spectrum lease for a licensed band must be obtained. Generally, high costs are associated
with a spectrum lease, which would, in turn, amplify the overall cost of deployment. This
is also contradictory with the main design principles of WSN [15]. On the other hand,
unlicensed bands are also used by other devices such as IEEE802.11 wireless local area
network (WLAN) hotspots, PDAs and Bluetooth devices. Therefore, sensor networks
experience crowded spectrum problem [16]. Hence, in order to maximize the network
performance and be able to co-operate efficiently with other types of users, opportunistic
spectrum access schemes must be utilized in WSN as well.
c. Using adaptability to reduce power consumption: Time varying nature of wireless
channel causes energy consumption due to packet losses and retransmissions. Cognitive
8
radio capable sensor nodes may be able to change their operating parameters to adapt to
channel conditions. This capability can be used to increase transmission efficiency, and
hence, help reduce power used for transmission and reception.
d. Overlaid deployment of multiple concurrent WSN: With the increased usage of sensor
networks, one specific area may host several sensor networks deployed to operate
towards fulfilling specific requirements of different applications. In this case, dynamic
spectrum management may significantly contribute to the efficient co-existence of
spatially overlapping sensor networks in terms of communication performance and
resource utilization.
e. Access to multiple channels to conform to different spectrum regulations: Each country
has its own spectrum regulation rules. A certain band available in one country may not be
available in another. Traditional WSN with a preset working frequency may not be
deployed in cases where manufactured nodes are to be deployed in different regions.
However, if nodes were to be equipped with cognitive radio capability, they would
overcome the spectrum availability problem by changing their communication frequency.
Therefore, it is conceivable to provide wireless sensor networks with the capabilities of
cognitive radio and dynamic spectrum management. This defines a new sensor network
paradigm, i.e., Cognitive Radio Sensor Networks (CRSN). In general, a CRSN can be
defined as a distributed network of wireless cognitive radio sensor nodes, which sense an
event signal and collaboratively communicate their readings dynamically over available
spectrum bands in a multi-hop manner ultimately to satisfy the application-specific
requirements.
2.4 Wireless Sensor Networks:
A sensor network is composed of a large number of sensor nodes that are densely deployed as
shown in Fig 2.3. Each sensor node has the capabilities to collect surrounding information and
redirect the information to the sink. Data are routed back to the sink by a multi-hop
infrastructureless architecture through the sink. There are many factors that influence the design
of the sensor network such as fault tolerance, scalability, production costs, sensor network
topology, operating environment, hardware constraints, transmission media, and power
consumption [15].
9
Fig 2.3 Sensor Nodes scattered in a sensor field
2.4.1 Design Factors:
There are many researches made on the design factors, few are addressed in this project.
However, it is very difficult to list all the driving factors which affect the design of sensor
network. These factors are to be considered as they serve as a guideline while designing an
algorithm or a protocol for wireless sensor networks.
Following are the main influencing factors considered while comparing different schemes.
A. Fault Tolerance
The sensor nodes consist of very limited battery and are bound to failure or blockage due to lack
of power, or manufacturing defects or environmental interference. Hence special care should be
taken to overcome these limitations and precautions should be taken in such a way that these
factors do not affect the overall functioning of the sensor network. The new algorithms proposed
should make place for these expected issues. Fault tolerance can be defined as an ability to
10
sustain sensor network functionalities without any interruption due to sensor node failures [17,
18].
B. Scalability
To test the scalability of a sensor network the number of sensor nodes deployed in the area can
be increased from hundreds to thousands depending on the applications. The number of sensors
can reach an extreme value of millions depending on the requirement. At the same time the
density of the sensor networks must be utilised efficiently.
C. Production Costs
A sensor network consists of a large number of sensor nodes deployed. These sensor nodes are
cost efficient as it is very important to limit the network cost. The cost of the network should not
exceed the cost of deploying sensors in traditional sensor network in order to justify the cost. As
a result sensor node cost should be kept low. The state-of-the-art technology allows a Bluetooth
radio system to be less than US$10 [19]. Also, the price of a piconode is targeted to be less than
US$1. The cost of a sensor node should be much less than US$1 in order for the sensor network
to be feasible. The cost of a Bluetooth radio, which is known to be a low-cost device, is even 10
times more expensive than the targeted price for a sensor node.
D. Hardware Constraints
A sensor node consists of four basic components as shown in Fig. 2.4: a sensing unit, a
processing unit, a transceiver unit, and a power unit. There may be additional application-
dependent components for providing properties like mobility. Sensing unit has two subunits:
sensors and analog-to-digital converters (ADCs). The sensors produce analog signals which are
converted into digital signals by the ADC, and then given to the processing unit. The processing
unit has a small storage unit which helps the sensor to coordinate with the neighbouring sensor
nodes in order to the assigned sensing tasks. A transceiver unit connects the node to the network.
An important component of a sensor node is the power unit. There are also application-
dependent subunits. The mobilizer enables the sensor nodes to move when it is required in
completing a specific task.
11
Fig. 2.4 Sensor Node Hardware.
12
Chapter 3
Energy Efficient Architecture for Cognitive Radio based Sensor
Network and its applications
Tremendous growth in wireless communications as well as in Wireless Sensor Networks made
spectrum overcrowded. With recent advances in Cognitive Radio (CR) technology, it is possible
to apply Dynamic Spectrum Access (DSA) model in WSN to get access to less congested
spectrum, possibly with better propagation characteristics. In this chapter, we attempt to improve
the radio spectrum utilization by proposing cognitive wireless sensor networks. Network
architecture of a sensor network plays a key factor in improving the efficiency in cognitive radio
environment. Specifically we propose new spectrum aware network architecture.
3.1 Introduction:
The research area of wireless sensor networks is an advancing technology which has opened new
research issues. Its number of applications in various sectors of military and civilian applications
has created demand for the design of more efficient and effective wireless sensor networks. The
sensor nodes are constrained in energy supply, computational memory space, speed and
bandwidth. In spite of limitations sensor nodes can reliably and accurately report the surrounding
environmental changes.
WSNs are working at 2.4GHz frequency band (unlicensed band). This unlicensed band is
crowded with many other wireless applications such as Bluetooth, Wi-Fi etc. Currently WSNs
are operating in fixed spectrum assignment. This fixed spectrum allocation makes WSN to
interfere with other technologies working at this same band. To overcome this cognitive radio
based sensor networks are introduced.
In general, a cognitive radio sensor network (CRSN) can be defined as a distributed network of
wireless cognitive radio sensor nodes, which sense event signals and collaboratively
communicate their readings dynamically over available spectrum bands in a multihop manner to
ultimately satisfy the application-specific requirements [20].
13
3.2 Background for CRSN Architecture:
Typical communication architecture of a CRSN is illustrated in Fig. 3.1a. Depending on
spectrum availability, sensor nodes transmit their readings in an opportunistic manner to the next
hops and ultimately to the sink. The sink may also be equipped with cognitive radio capability
i.e., a cognitive radio sink. In addition to the event readings, sensors may exchange additional
information with the sink including control data for group formation, spectrum allocation, and
spectrum- handoff-aware route determination depending on the specific topology.
3.2.1 CRSN Node Structure:
The main difference between the hardware structure of a classical sensor and a CRSN node is the
cognitive radio transceiver of a CRSN node (Fig. 3.2). A cognitive radio unit enables the sensor
nodes to dynamically adapt their communication parameters such as carrier frequency,
transmission power, and modulation. CRSN nodes also inherit the limitations of conventional
sensor nodes in terms of power, communication, processing, and memory resources, which also
restrict the features of cognitive radio.
3.2.2 CRSN Topology:
According to the application requirements, CRSNs may exhibit different network topologies, as
explored below.
a. Ad Hoc CRSN: In an ad hoc CRSN such as that in Fig. 3.1a, nodes send their readings to
the sink in multiple hops in an ad hoc manner. This topology imposes less
communication overhead in terms of control data. However, due to the hidden terminal
problem, spectrum sensing results may be inaccurate.
b. Clustered CRSN: It is essential to designate a common channel to exchange various
control data, such as spectrum sensing results, spectrum allocation data, neighbor
discovery, and maintenance information. While it may not be possible to find a network-
wide common channel, it is highly possible in a certain restricted locality [21]. Therefore,
cluster-based network architecture such as the one in Fig. 3.1b is an appropriate choice
for effective dynamic spectrum management with a local common control channel
approach.
14
c. Heterogeneous and Hierarchical CRSN: CRSN architecture may incorporate special
nodes equipped with more or renewable power sources such as actor nodes, which act on
the sensed event in wireless sensor and actor networks (WSANs). Here, actor nodes may
perform additional tasks like local spectrum bargaining. These nodes may have longer
transmission ranges and hence be used as relay nodes. This forms a heterogeneous and
hierarchical topology with ordinary CRSN nodes, high power relay nodes (e.g., cognitive
radio actor nodes), and the sink (Fig. 3.1c).
Fig 3.1. a) Cognitive radio sensor network (CRSN) architecture; b) possible network topologies
or clustered CRSN; c) a heterogeneous hierarchical CRSN.
15
Fig. 3.2. Hardware structure of a cognitive radio sensor node.
3.3 Proposed Energy Efficient Architecture For CRSN:
CRSN is also power constrained as WSN as these cognitive radio based sensor nodes also work
on battery power. In this chapter we propose an energy efficient architecture for CRSN, which is
as follows below.
In CRSN architecture each sensor node senses the spectrum as well as event. From sensed
spectrum information cognitive sensor node dynamically changes its frequency and sends its data
to the sink in multi hop manner. In this architecture sensing nodes have to sense the entire
spectrum all the time, which makes sensor nodes to drain out of battery very quickly.
To increase network life time of CRSN we propose a new architecture in which spectrum
sensing network is decoupled from data gathering network.
The initial sensing network is known as Sensing sensor network. It consists of sensor nodes
arranged in a particular manner to increase the life time of the network. The backbone network is
16
known as Data Transmitting/Data Gathering network. This network is responsible for data
distribution to the operational networks.
By this distinction we are minimizing the load on the sensors in the secondary network i.e.
processing, analyzing and distributing the sensed data on the same front. This multitasking
results in battery drain and decreasing the life time of the network. We have divided the tasks as
a result we have improved network spectrum efficiency. And by distinguishing the secondary
network we have achieved increase in the life time of the network. Therefore we have created
sub networks based on these functionalities as shown below in the Fig. 3.3.
Fig. 3.3. Proposed network architecture of cognitive Radio Based Sensor Network
The data transmitting network is associated with the operational network of unlicensed users.
The sensor nodes are designed in such a manner that the data sensing is made effective and
efficient. The sensors are placed in a grid paradigm for exploiting the characteristics of a sensor
grid. The grid structure provides a seamless access to the network resources in a pervasive
manner.
The sensed spectrum data is collected by one or more base stations and forwarded to the central
head called spectrum allocator. In spectrum allocator the spectrum vacancy chart is dynamically
formed based on the collected data. This corresponding data is distributed to the operational
17
networks in the secondary network. This forwarded information shows the availability of the
spectrum. It is used by the unlicensed users and disturbances to the licensed users are avoided.
3.4 Applications of CRSN:
The applications for WSNs are varied, typically involving some kind of monitoring, tracking, or
controlling. In a typical application, a WSN is scattered in a region where it is meant to collect
data through its sensor nodes. The need for cognitive sensor networks is much more in
applications where the existing communication spectrum is highly assigned, but it is being
underutilized. Some of the applications are mentioned below.
3.4.1. CRSN in Telemedicine:
Telemedicine can be broken into three main categories: store-and-forward, remote
monitoring and interactive services.
Store-and-forward telemedicine involves acquiring medical data (like medical images,
biosignals etc) and then transmitting this data to a doctor or medical specialist at a
convenient time for assessment offline.
Remote monitoring, also known as self-monitoring/testing, enables medical professionals
to monitor a patient remotely using various technological devices.
Interactive telemedicine services provide real-time interactions between patient and
provider, to include phone conversations, online communication and home visits.
The integration of telemedicine with medical micro sensor technology provides a promising
approach to improve the quality of healthcare for people. In hospitals of future there are various
wireless networks that are proposed to be operational. For instance WLAN and may be
Bluetooth networks are proposed to be deployed. Clearly there is a need to share the spectrum
among primary users (WLAN, Bluetooth) and nodes of WSN.
Typical scenarios of implementation of Cognitive Sensor Networks:
In [22] authors proposed a mobile sensor network infrastructure to support the third-generation
telemedicine applications. The main features of the proposed ―3G Mobile Telemedicine based on
Sensor networks‖ are as follows:
18
Sensor Telemedicine Networks are seamlessly integrated with the aforementioned 3G mobile
networks. In the future telemedicine systems, patients will carry medical sensors that sense the
health parameters, such as body temperature, blood pressure, pulse oxymetry, ECG, breathing
activity, and so on. In addition, serious patients can also carry other sensors that help the medical
center carry out remote monitoring. Typical examples are location sensor, motion or activity
sensor, microphone sensor, and camera sensor. Typically a patient will carry a wrist-device
(called as super-sensor) with a stronger battery and higher memory compared to the other
medical sensors, to perform multiple-hop ad hoc communication. Super-sensors will be used to
collect sensing data from body sensors and communicate with other super-sensors. However,
these tiny wrist-devices will not have as much power as today‘s cell phone, to perform two-way
communication with the base stations. They therefore transmit data through multihop routing
algorithm. In Sensor aided Telemedicine Networks there are three types of services. They are as
1. Real-time calls from Ambulance Patients – handoff guaranteed.
2. Handoff-prioritized calls from serious patients or elder people.
3. Non-real-time calls from Cluster-heads who collect medical data from wrist-worn super
sensors of average patients or normal people.
Fig. 3.4. Three types of services provided by Sensor Telemedicine Networks.
Application of CRSN approach in “3G Mobile Telemedicine based on Sensor networks”:
19
In this ―3G Mobile Telemedicine based on Sensor networks‖ WSN is working along with
licensed cellular networks. Robust use of cellular communication, leads to spectrum scarcity for
WSN. CRSN can overcome this spectrum scarcity problem. Fig. 3.5 shows the implementation
of cognitive sensor networks in ―3G Mobile Telemedicine bases on Sensor Networks‖. In this
CRSN scenario in each cell we will have a spectrum coordinator. When any of the three medical
service events occurs they will send spectrum allocation request to coordinator. This coordinator
senses the spectrum and will allocate free spectrum frequency depending on the priority of the
service.
Fig. 3.5. Cognitive Sensor Networks in Telemedicine
3.4.2. CRSN in Under water acoustic sensor networks:
Research has been active for over a decade on designing the methods for wireless information
transmission underwater. Ocean bottom sensor nodes are deemed to enable applications for
oceanographic data collection, pollution monitoring, offshore exploration, disaster prevention,
assisted navigation and tactical surveillance applications. Multiple Unmanned or Autonomous
Underwater Vehicles (UUVs, AUVs), equipped with underwater sensors, will also find
application in exploration of natural undersea resources and gathering of scientific data in
collaborative monitoring missions. To make these applications viable, there is a need to enable
underwater communications among underwater devices. Underwater sensor nodes and vehicles
must possess self-configuration capabilities, i.e., they must be able to coordinate their operation
20
by exchanging configuration, location and movement information, and to relay monitored data to
an onshore station.
Wireless underwater acoustic networking is the enabling technology for these applications.
Under Water Acoustic Sensor Networks (UW-ASN) consist of a variable number of sensors and
vehicles that are deployed to perform collaborative monitoring tasks over a given area. To
achieve this objective, sensors and vehicles self-organize in an autonomous network which can
adapt to the characteristics of the ocean environment.
Main challenges in implementation of UW-ASN:
Unlike normal WSN, UW-ASN communication is done using sound waves.
The main issue is that the acoustic transceiver is bigger than the classical radio one. But
most of these transceivers are for far communications.
Inside the sea water, there are some dynamic effects due to the waves at the sea surface as
well as the movement of fishes or other marine animals, as well as the marine vehicles.
There are also effects of the salinity variations with the water depth on the propagation
parameters, as well the effects of the water pressure with the depth.
Usually the frequencies used in acoustic undersea communications are high (about 100
kHz –1 MHz).But the propagation at that frequencies are limited by small obstacles (at
c=1450 m/s sound velocity, a 1 mm bubble represents an integer wavelength obstacle
1.45 MHz).
Low frequencies are not suitable for communications due the mechanical vibrations.
It is a small room in acoustic spectrum for a secure communication and especially in
transient conditions the available bands are moving quickly.
Cognitive approach:
All the animals inside the sea (under water) communicate using sound waves with different
frequencies. When our UW-ASN operates in these frequencies, our communication can
encounter certain disturbances. On another source for disturbance are bubbles or turbulence. This
bubbles or turbulence operate in certain fixed frequency. So using Cognitive Radio based Sensor
networks in UW_ASN will be yield fruitful results. Cognitive Radio based Sensor Networks;
21
dynamically choose the available and optimal non-disturbance spectrum. Fig. 3.6 shows
architecture of Cognitive sensor networks in UW-ASN.
Fig. 3.6 Cognitive Radio based Sensor Networks in UW-ASN
3.4.3. CRSN in Precision agriculture:
Precision Agriculture refers to a set of technologies that introduce the concept of local variation
into the large-scale mechanization, which is essential to large fields [23]. With the determination
of soil conditions and plant development, these technologies can lower the production cost by
fine-tuning seeding, fertilizer, chemical and water use, and potentially increasing production and
lowering costs. These can be achieved through the approach of agricultural control and
management based on direct chemical, biological and environmental sensing. The Precision
farming system has the following parts:
22
a) Sensing agricultural parameters.
b) Identification of sensing location and data gathering.
c) Transferring data from crop field to control station for decision making.
d) Actuation and Control decision based on sensed data
Agricultural Sensors, positioning systems for detecting location of sensors, actuators like
sprinklers, foggers, valve-controlled irrigation system, etc. are already available in market.
Wireless Sensor Networks (WSN) is well suited for this application and it is very cost effective.
In the forms of future it is very clear that the terrestrial cellular networks, WMAX networks,
Mesh networks and other have to coexist with WSN. Thus approaches to efficient spectrum
utilization are very much needed. Cognitive networking paradigm is an important step in this
direction.
23
Chapter 4
Energy efficient virtual grid placement based Routing algorithm
In this chapter, we propose a novel routing technique for topology aware wireless Sensor
Networks (WSN) and that guarantees energy efficiency of the network. The key idea is to place a
virtual uniform grid. Each grid is considered as a Cluster. After virtual grid placement, a total
network is hierarchically partitioned into rectangular levels. The logical reasoning is that it is
sufficient to know particular ‗level ID and grid ID‘ where event has occurred, instead of knowing
the position of each and every sensor node of the network. Virtual grid placement, benefits in
terms of data aggregation and data filtering and reduces traffic throughout the network. Hence
our protocol deals with localization and fusion of data in addition to routing. Our approach when
applied in the network increases the life time of the network and it also helps to improve the
successful transmission of data packets. We show that compared to existing solutions such as
circular leveling, gossiping, flooding and level controlled gossip, our approach is energy efficient
and it even outperforms other existing algorithms.
4.1 Introduction:
Energy dissipation at sensor node is a major concern, as in many applications sensor have to be
deployed in inaccessible environments. Sensing alone is not an energy consuming activity, but
networking and programming certainly are. So, the major problem lies in activities like routing,
addressing. In this chapter our focus will be on routing problems.
The rapid development of sensor networks have posed many challenges in routing protocols and
trade-offs between reliability and efficient use of energy resources. Epidemic Algorithms [24]
have proven to be unreliable as they reduce the deterministic nature of the route discovery.
Current sensor networks employ query distribution and data collection based on data diffusion
[8]. This model assumes that the base station performs most of the tasks and is permanently
connected with the network. In order to integrate the data base and sensor network technologies
the concept of data-centric protocols was introduced but these protocols requires robust and
efficient methods that support data diffusion [25].
24
4.2 Issues Addressed and Underlined Assumptions:
The following are the issues addressed in our proposed protocol:
Reduce the numbers of nodes involved in the data transfer through virtual grid placement
and leveling (rectangular levels).
Avoids unnecessary flooding (i.e. discard data packets) throughout the network.
Creates an energy efficient path towards the base station and hence is power aware.
Increases the reliability of the gateways (routers) and reduces the complexity of the
overall network.
The assumptions made on the nature of the sensor network are as follows:
The nodes in the network are assumed to be stationary.
The sensor network is densely deployed.
Topology is under control of the user.
All nodes are considered to have similar capabilities in the network.
Base station (BS) is in the centre of the sensor field.
4.3 Proposed algorithm:
To decrease the consumption of energy of the battery, in our proposed algorithm instead of
flooding the data packet into network, initially we are maintaining the data base of the adjacent
cluster heads(CHs) as CH-table and only to selected CHs data packet is being forwarded.
Our proposed algorithm has two phases, setup phase and routing phase. In first phase total
sensor field is divided into permanent grid structure using global location information
irrespective of the number of events. After the sensing field is divided into the grid structure,
sensor nodes decide their grids (clusters) based on the location information and the CH is
randomly selected which is responsible for aggregating the data in the cluster and forward it to
25
sink. Each CH maintains a CH-table. From CH-table next-CH is selected and this process is
explained in section 4.3.2.
When an event occurs, an event node sends a data packet to CH.CH already has the CH-table. It
selects the CH with level number one lesser than current level number and column number one
higher than the current column and row number of grid.
4.3.1 Setup phase:
Setup phase has mainly three steps. They are as follows
a. Placing virtual grid and grid numbering:
The sensor field is divided into a virtual grid using global location information which is provided
by localization system, such as GPS (Global Positioning System) or through techniques such as
[26]. When sensor nodes are deployed in the field, they decide their grids with location
information. The size of the grid is α × α and represented as where represents row number
and represents column number
Now total sensing field is divided into I rows and J columns. Corner left side top most grid (in
reader‘s point of view) is numbered as 00. Column number is incremented on moving left from
grid 00 till the column with Base Station. Column next to Base Station is numbered column
number as J-1and next onwards the column number is decremented on moving right. Row
number is incremented on moving from top (00) to bottom till the row with BS. Row next to BS
is numbered row as I-1and next onward the row number is decremented on moving down. This is
illustrated in Fig. 4.1.
b. Clustering:
Each grid is considered as a cluster. In order to select the head of each cluster, if sensor nodes
begin with equal battery power, all sensor nodes locally flood a packet (head–packet) during the
random period. The announcement of the head-packet is limited within the single cluster by
simply dropping the packet from neighbor clusters. The sensor node which sends the head-packet
first plays the role of CH.
26
c. Rectangular leveling:
Rectangular leveling is done by using ―hop count based method‖. In this method hop count is
used to determine the levels. Initially hop count of CHs is set to infinity (or arbitrary large
number). First BS broadcasts packets with hop count field set as zero. CHs which receive this
packet set their hop count field as zero and level as one and the hop count field in the packet is
incremented by one. These updated packets are broadcasted again. CHs that receive these
packets update their level to „hop count + 1‟, if and only if their current level is higher than ‗hop
count +1‘ .If CHs that are having their level equal to or less than the ‗hop count + 1‘ value of the
received packet, then their do not update their current level value.
This way the whole network is assigned as levels based on their hop count from the sink. Fig. 4.1
depicts the sensing field at the end of setup phase.
4.3.2 Routing phase:
After the end of first phase CHs need to maintain adjacent CHs information in CH-table.
So after the first phase each CH broadcasts information-packet. Information-packet consists of
three fields, level number, grid number and CH ID. CHs which receive this packet update their
CH-table with the information-packet. Being grid like structure each CH has only eight adjacent
CHs.
Node at which an event occurs is known as event node. A packet from event node to Base
Station is routed as follows:
1. Event node sends the packet to CH.
2. CH sends the packet to next-CH. Next-CH is chosen from CH-table if and only if it follows
following conditions
i. Level of next-CH must be one lesser than current CH.
ii. Column or row number must be one greater than current grid column and row number
respectively but not both column and rows.
27
iii. If there is no such CH which satisfies condition (ii) then select next-CH as the one with
level lesser by one of current CH and column and row numbers greater by one of
current grid column and row numbers respectively.
3. Step 2 is followed until packet is send to BS.
This algorithm is illustrated in Fig. 4.2.
Fig. 4.1 sensor field at the end of setup phase, in each grid number in top right represents grid
number and bottom left level number.
28
Fig. 4.2. Routing of packet from event node to Base Station.
4.3.3 Cluster Rotation:
Limited battery capacity of sensor nodes make CHs to drain out of power after a certain amount
of time, which causes failure of transmission. To overcome this issue cluster head rotation
algorithm is implemented.
When CH remains with 5% of its total power it broadcasts a CH-rotate packet. This broadcast is
limited to its cluster (grid) only. Nodes which receive this packet send head-packet, which tells
the remaining power in the particular node to CH. Node with highest power is selected as CH.
29
4.4 Performance Evaluation:
In this section, we have made comparative performance evaluation of our proposed algorithm
with Flooding, Gossip [27] and traditional circular leveling [28]. The results have proven that the
approach proposed by us increase the lifetime of the network and is thus the energy efficient.
These results were plotted by several runs of the experiment.
4.4.1 Network Model:
For the purpose of evaluating the algorithms, we simulated them by varying the number of nodes
in the network. For each algorithm we started with a 100 node network and thereby generating
the number of events that the network could handle. Similarly, the number of nodes has been
varied up to 800 nodes and the corresponding number of events that these networks could handle
was plotted in Fig 4.3. We consider nodes are randomly deployed in a flat sensor field.
The two metrics of interest provided by the simulator are
Number of Nodes: The number of nodes that are present in the network.
Lifetime: The time to failure of say 5% or 10% of nodes closer to the base station is considered
to be the lifetime. The closeness to Base Station is measured based on the level number.
All the events have to pass through nodes near Base Station. When these nodes are dead, total
network is useless. Node parameters are tabulated in Table 4.1.
30
Table 4.1
Node Parameters
Parameters Value
Simulator area 100 X 100 m2
Number of nodes 100 to 800
Grid size 10 X 10 m2
Node initial energy 1000 J
Node transmission range 10 m
Cluster head transmission range 30 m
Transmission power 10 J
Reception power 2 J
Fig. 4.3. Plot of Nodes VS Network lifetime
4.5 Conclusions:
In this chapter we have presented a virtual grid based routing protocol for topology aware WSN,
that maximizes lifetime of the system by involving few nodes i.e., cluster heads in routing from
source to destination. Extension simulations are conducted to evaluate the performance of the
network and to show that our approach is superior to flooding, gossip and traditional leveling for
lifetime maximization.
31
Chapter 5
Mobility Tolerant TDMA Based MAC Protocol for WSN
Recent advancements in wireless communications and electronics have enabled the development
of low cost sensor networks. Among the protocols of wireless Sensor Networks (WSN), Medium
Access Control (MAC) protocols are given more priority as traditional MAC protocols are not
suitable for Wireless communication. As many of the MAC protocols consider the sensor nodes
to be stationary, and when these protocols are used in mobile environment the network
performance decreases. In this chapter, we introduce a new TDMA based MAC protocol which
can be used in mobile wireless sensor network. This protocol uses TDMA based MAC scheme
where the time will be divided into frames and then time slots. These slots are further divided
into sections as channel request (CR), channel allocation (CA) and data section. This MAC
protocol has shown better network lifetime when compared to existing protocols. This has been
shown through simulations.
5.1 Introduction:
The research area of Wireless sensor networks is an advancing technology which has opened
new research issues. A sensor network is composed of a large number of sensor nodes that are
densely deployed either inside the phenomenon or close to it. Some of the application areas are
health, military and home. In military, for example, the rapid deployment, self organisation and
fault tolerance characteristics of sensor networks make them a promising technology. In health,
sensor nodes can also be deployed to monitor patients and assist disabled patients. Some other
commercial applications include managing inventory, monitor product quality and monitor
disaster areas.
Wireless Sensor Networks usually employ performance metrics that are different from those of
more conventional data networks, emphasizing low power consumption and low cost rather than
data throughput or channel efficiency. Since power is consumed every time a network device
accesses the channel, the method by which the device can have a large effect on its power
consumption and therefore the choice of a MAC protocol is crucial for a WSN.
32
The MAC protocol in a wireless multi hop self organising sensor network must achieve two
goals. The first is the criterion of the network infrastructure. Since thousands of sensor nodes are
densely deployed in a sensor field, the MAC scheme must establish communication links for
data transfer. The second objective is to fairly and effectively share the resources between all the
sensor nodes.
5.2 Background:
The Medium Access Control protocols can be broadly classified as following categories. They
are as following:
(1) Scheduling based
(2) Contention based.
In this section some of the scheduling based and contention based MAC protocols are discussed.
A. SMAC:
SMAC [31] is a contention based MAC protocol. SMAC uses three novel techniques to reduce
energy consumption and support self-configuration. SMAC introduced a periodic sleep and wake
up scheduling, which reduces energy consumption in listening to an idle channel. Neighbouring
nodes form virtual clusters to auto-synchronize on sleep schedules. SMAC also sets the radio to
sleep during transmissions of other nodes. And it uses in-band signalling. Finally, SMAC applies
message passing to reduce contention latency for sensor-network applications that require store-
and-forward processing as data move through the network.
B. EMAC:
In EMAC [30], time is divided into so called frames just like in TDMA but each frame is divided
into timeslots and each slot contains three sections: communication request (CR), traffic control
(TC), and data section. Each timeslot can be owned by only one network node. This network
node decides what communication should take place in its timeslot. Other nodes can ask for data
or notify the availability of data for the owner of the timeslot in the CR section. The owner of the
slot transmits its schedule for its data section and broadcasts a table in the TC section, which tells
to which other TC sections the node is listening. After the TC section, the transmission of the
33
actual data packet follows. The Node goes into a standby sleep mode when at a certain time no
transmissions are expected. It releases its slot and starts periodically listening to a TC section of
a frame. When the node has to transmit some data (event driven sensor node), it can just fill up a
CR section of another network node and agree on the data transmission, complete it and go back
to sleep.
C. TRAMA
In this MAC protocol time is divided into number of slots as we do in TDMA [36]. The network
nodes switch in to idle mode or low power mode when they are not transmitting or receiving. A
distribution election scheme is employed to determine which node to use a particular time slot.
Time slots are not allotted to those nodes that do not have any data to transmit. This improves
energy efficiency of the network significantly over TDMA.
D. Mobile TDMA
In [29] authors proposed TDMA-based MAC protocol for mobile sensor networks. This protocol
works by first splitting a given frame into control part and a data parts. The control part is used to
manage mobility, where as nodes transmit messages in data part. In data part, some slots are
reserved for mobile node.
Mobility Models
The most widely used mobility model is random walk mobility model [32] (Brownian
movement). A slight enhancement to this model is Random Waypoint model [33]. In this model
pauses are introduced between changes of direction and speed. A recent work in mobility models
[34], argued that these models do not faithfully reflect reality, and developed a model based on
social theory. The authors argue that, this model is well suited for MANETs. Mobility models of
MANETs may not suitable for sensor networks as applications for sensor networks are tracking,
monitoring, which is not the case of MANETs. However in [29] authors have introduced a novel
mobility models which suit for sensor networks, viz., Intra-cluster mobility and Inter-cluster
mobility. In our protocol we are considering these models.
34
a. Intra-cluster mobility: When nodes are mobile, if they remain within their assigned
cluster, the set of nodes within that cluster remains unchanged, though the topology may
change. This is stated as intra-cluster mobility [29].
b. Inter-cluster mobility: When a node leaves its cluster, it will join in a new cluster due to
mobility, causing the set of nodes in its original cluster to change and in new cluster also.
This is stated as inter-cluster mobility [29].
In [29] authors have proven collision-freedom in intra- and inter-cluster mobility, if each node is
assigned a distinct slot.
5.3 Problem Statement:
In this section we discuss the problems with the present medium access protocols.
The major problem with TDMA is the excessive delay. Each node is given a particular
time slot in which it has to send its data. If the node senses some data after its slot time is
over, then it has to wait till it gets a chance to transmit in its own slot in the next time
frame. This results in poor channel utilization and delay. Some times because of this delay,
in some energy constrained networks like WSN there is probability that the node may die
before it gets its slot in the next time frame and the data is lost with the node itself. Also
lots of energy is being wasted by idle listening, which brings down the network lifetime
significantly.
TRAMA solves the problem of energy efficiency by switching in to low power mode when
the node is not transmitting or receiving. But the problem of delay and poor channel
utilization are not resolved.
EMACS employs the idea of dividing the time slot into sections thereby decreasing the
delay and improving the channel utilization to a certain extent. But this algorithm doesn‘t
consider the case of two nodes compete for a particular time slot whose owner doesn‘t
have any data to transmit. There is no discussion on what priority basis the nodes compete
and get the time slot.
Also none of the above said protocols consider mobility of the network nodes. They
assume that the nodes are stationary or considerably low mobility. In [29] authors proposed
an algorithm for mobile nodes based on TDMA. But this has problems of delay and poor
channel utilization.
35
5.4 Assumptions:
Total network is static in the setup phase.
Cluster head have relatively more battery power and less mobility when compared to
other nodes in the cluster.
Time synchronization for all nodes is done automatically.
Except Cluster Head all other nodes have same capabilities.
Fig. 5.1. Schematic of the proposed algorithm
36
5.5 Proposed Algorithm:
We consider the whole network to be divided in to clusters using FLOC algorithm [35]. Each
cluster has its own Cluster Head. Now as in Fig.5.1 time is divided in to frames (N) and in turn
divided in to time slots. Before every time frame, the Cluster Head allocates a time slot to every
node (n) in the cluster and x slots are left free for the nodes that join the cluster in later part of the
time frame. This x is considered as 10% of number of nodes in a cluster. The node which holds
the current slots is known as owner node.
Whenever a new node comes and joins the cluster, the Cluster Head allocates half of the free
time slots (x/2 slots) to this node. If another comes and joins the cluster, this time it gets half of
the current free slots (x/4). Therefore a cluster will have some empty slot at anytime in the time
frame to allocate to a new node that joins the cluster. Slot allotment to new nodes joined in a
cluster is explained in section 5.5.A. This deals with the problem of mobility of the network
nodes efficiently.
But if some of the mobile nodes leave the cluster or the particular owner node doesn‘t have any
data to send and so the channel resources are wasted. To improve the channel utilization
capability we split each timeslot in to three sections. They are:
1. Communication Request (CR)
2. Channel Allocation (CA)
3. Data Section (DS).
In communication request (CR) section, owner node collects the information about the
surrounding nodes. All those nodes that have data to transmit will put requests to the owner
node for the grant of slot for data transmission. If the owner node doesn‘t have any data to
transmit then it will calculate the Priority Index (PI) of each node that has put a request for grant
of a time slot. Among the slot requested nodes, the node with less PI is given more priority. PI
calculation is explained in section 5.5.B.
In Channel Allocation (CA) section, the node decides whether to send the data and keep the
channel to it or allots the channel to a cluster node which has requested for a slot and has the
highest priority.
37
Packets are transmitted or received in this Data Section. After the communication process is over
the node goes into sleep as often as possible in order to save the battery power.
5.5.A Slot allotment:
Inter-cluster mobility causes the nodes to move from one cluster and join in new cluster. New
joined node needs a slot for its further communication. Cluster Head keeps track of nodes with
the cluster, nodes leaving the cluster. When a new node enters into a cluster it requests Cluster
Head for a slot.
Let r be the number of nodes left the cluster. These r slots are left unused as their owners left the
cluster. x be the number of slots reserved for the nodes which joins cluster because of mobility.
Total number of slots available for this instant is r+x. For the first r nodes which joins the
cluster, Cluster Head allocates r free slots which are left unused as the owner nodes left the
cluster. For next the nodes which joins the cluster, the Cluster Head provides half of the free time
slots. i.e., if Cluster Head has x slots, it will provide x/2 and so on for the next incoming nodes.
5.5.B Calculation of PI:
Priority Index is the measure that decides whether a node gets the time slot of the owner node.
Higher the PI, lower the priority. It is calculated based on the following factors.
Priority Index is inversely proportional to mobility of node. If node has higher velocity
the probability that it completely transmits the packet is less, so priority of packet should
be more and so the priority index should be less.
VPI
1
(5.1)
If battery power (BPc) of the request node is less than threshold (BPth) level, then the
node cannot complete transmission of its packet, so the node should not transmit this
packet. The priority of the packet should be more and hence the priority Index leads to
inverse relationship.
The PI of the packet should be inversely proportional to battery power if the battery
power is less than threshold. If battery power is above the threshold level, then priority
38
index of the packet should be more giving lower priority index values to the packets in
nodes whose battery power is close to threshold value.
BPcPI if BPc ≥ BPth
BPcPI
1
if BPc < BPth (5.2)
)( THC BPBPSgnB (5.3)
Depending on the difference of slot numbers between the owner nodes (so) and requested
(sr) node priority will be assigned.
When the difference is positive, more difference causes more delay for transmission a
packet. To reduce transmission delay, requested node (sr) with more difference should be
given more priority.
When the difference is negative, smaller differences cause more delay for transmission a
packet. To reduce delay, requested node with lesser difference should be given priority.
ro SSPI
1
if ro SS ≥ 0
ro SSPI if ro SS
< 0 (5.4)
)sgn( ro SSD (5.5)
Where sgn stands for signum function which is defined as
Sgn(y) = 1 if y > 0
= 0 y = 0
= -1 y <0
Considering all above equations PI is given as following
PI = B+D-v (5.6)
39
5.6 Simulation Results:
For the comparative evaluation, we simulate the protocols extensively in MATLAB. We
consider a (100x100) m2 sensor field and the nodes (both mobile and static) are randomly
deployed all over the field. Each time frame is assumed to be 1ms long and is divided into 10
equal time slots.
The parameters of the nodes used in our simulation are as shown in the TABLE 5.1.
TABLE 5.1
NODE PARAMETERS
Parameters Values
Simulation area 100 X 100 m2
Communication radius 30m
Sensing radius 3m
Total Initial Energy 1000J
Idle cost 0.0165 J/s
Transmission Power 10 J
Reception power 2 J
For static network:
Events are generated randomly at different times in a time frame. The average delay for a
message generated in a particular time frame is calculated. This average delay is plotted against
the number of events occurred in a time frame. It is evident from the Fig. 5.2 that our proposed
algorithm gives lesser delay when compared to traditional TDMA.
We assume Network Lifetime to be the time taken for 10% of the total nodes in the network to
fail. We then plot the network lifetime obtained by simulating the traditional TDMA and the
proposed protocol. The Fig. 5.3 clearly shows that our algorithm outperforms the TDMA by 30%
approximately.
40
For mobile network:
In this case, 10 mobile nodes are considered to be present initially in the simulation environment.
These mobile nodes (x) have some amount of mobility (ML). A random number of mobile nodes
may join the cluster, which have mobility (MC). Channel Utilization is assumed to be the
percentage of channel that is utilized in a time frame.
When we maintain the mobility (ML) of the nodes previously existing in the cluster to be
constant and calculate the spectrum utility by varying the mobility of the incoming nodes and
vice versa.
From Fig. 5.4 and Fig. 5.5 it is evident that our proposed protocol has good channel utilization
capabilities when compared to mobile TDMA.
Fig. 5.2. Graph for Delay
41
Fig. 5.3. Graph for Network Lifetime
Fig. 5.4. Graph for Channel Utilization (I)
42
Fig. 5.5. Graph for Channel Utilization (II)
5.7 Conclusions:
In this chapter we have proposed energy efficient TDMA based MAC protocol. It has very good
energy conserving properties comparing with traditional TDMA. Another interesting property of
the protocol is that it has relatively very less delay when compared to traditional TDMA.
43
Chapter 6
Energy Efficient cross-layer protocol design by using token passing
mechanism
Most of the Wireless Sensor Networks have limitation of battery power and in most of the
applications it is not possible to replace the battery of a node. Considering this scarce energy and
processing resources of WSN, we try to establish a joint optimization and design of networking
layers i.e., cross layer design which could be a promising alternative to the inefficient traditional
layered protocol design. In this chapter we propose energy efficient cross layer design of the
MAC and Routing protocol namely Energy Efficient cross-layer protocol design by using token
passing mechanism for WSN. This proposed protocol proves to be better than some of the
existing protocols and it is shown with the help of simulations.
6.1 Introduction:
Recent advancements in electronics and wireless communications enabled the manufacturing of
cheap and small sensor nodes. A Wireless Sensor Network (WSN) consists of numerous sensor
motes which sense the data, communicate with each other hop by hop and eventually report the
data to the Base Station. Though WSN started for military applications, gradually it found to
have very useful applications in wide range of areas like Health, Building Monitoring, and
Factory Automation etc. The main constraint in protocol design for WSN is Energy Efficiency as
the motes have limitations due to battery operation. In many of the applications the sensor nodes
are unattended and so battery replacement is not at all possible. Researchers have designed
numerous protocols for each layer in the traditional layered protocol design. But by joint
optimization of multiple layers by cross layer design gives more satisfactory results than our
traditional layered design.
Medium Access Control (MAC) Protocols in Wireless Communications are broadly classified
into three categories namely schedule based, demand based and contention based MAC
Protocols. TDMA, FDMA and CDMA are schedule based MAC Protocols. Traditional methods
like Trunking, Polling comes under demand based protocols. IEEE 802.11 standards define
contention based protocols CSMA/CA. Token Ring is a combination of both contention based
and schedule based protocols.
44
Routing is a very important function carried out by network layer in the protocol stack. The
routing layer protocols mainly emphasize on energy efficiency and throughput. Routing
algorithms are broadly divided as data centric, hierarchical, fault tolerant, localization based [42].
In the proposed algorithm in this chapter we make use of an innovative localization scheme
proposed by [37] to reduce the redundancy. In [37] authors have localized by dividing the sensor
field into levels and sectors. Using this localization technique combined with Token passing
approach we propose a new approach, Energy Efficient cross-layer protocol design by using
token passing mechanism for WSN. This protocol proves to be better than many existing
protocols for static WSN.
6.2 Background:
The previous scientific and experimental research reveals that the interactions between layers of
the network stack play a very important role in improving the performance of the network. The
inter dependency between the local contention and end to end congestion is an important factor
to be considered in the protocol design. The interdependency between these network layers need
cross layer mechanisms for efficient data delivery in WSNs. The design challenges and the
importance of the cross layer design for meeting the application requirements in the energy
constrained networks is discussed in [38].
Receiver based routing has been discussed in [39]. In this authors dealt with energy efficiency,
time delay, multi hop and single hop performance. In [40] a TDMA based MAC scheduling
routing algorithm is proposed where authors made use of sleep schedules which minimize the
wastage of idle listening. Also the nodes will transmit in a particular time slot which is chosen
based on the topology information. In this algorithm, in addition to MAC scheduling, the route
establishment is also dealt with. The physical layer results in [41] are extended in [42] to
compute the TDMA slot lengths to minimize the total energy consumption in a network with
many transmitters and only one receiver.
Some protocols are designed based on the type of applications and other factors. For example, in
[43] for periodic traffic in WSN, the nodes form distributed on-off schedules for each flow in
network. This forms a route and max efficiency could be achieved. The authors also showed a
trade off between on-off schedules and the connectivity in the network.
45
6.3 Assumptions & Definitions:
1. Our first assumption is that all the sensors in the network are Topology – Aware.
2. The whole sensor network is densely deployed.
3. All nodes in network have equal capabilities.
4. Time synchronization for all nodes is done automatically.
5. Token Frame (TC): This is a control packet that circulates within group of nodes. It has
number of fields which are as follows, Direction Flag (DF), Group Number (GN), Source
Address (SA), Destination Address (DA), and Frame Control (FC). Fig. 6.1 shows the
frame structure of token frame.
Token
Frame FC GN SA DA DF
Packet
Frame FC LID SID SA DATA
Acknowledgement
frame FC GN DA SA
Panic
Message FC GN SA
Fig. 6.1 Frame structure in our protocol
6. Acknowledgement (ACK): It is a control packet sent by a node to the adjacent node as a
confirmation for the token packet reception. Fig. 6.1 shows the frame structure of
acknowledgement.
7. Panic Message: It is a control message that is passed to the neighbours by a dying node.
Fig. 6.1 shows the frame structure of panic message.
8. Packet Frame: This frame contains the data. This frame gives Level Id (LID), Sector Id
(SID), source address (SA) and data (DATA). Fig. 6.1 shows the frame structure of
packet frame.
46
9. Network Lifetime: The time taken for 10% of the total number of nodes in the network
to fail is considered to be the Network Lifetime.
10. Direction Flag: This flag determines the direction of flow of the token. When the DF is
set to be 0, the token is given to the left neighbour. If it is set to 1 then it is passed to the
right neighbour, else if it is set to value 2 currently the node is not holding the token.
6.4 Proposed Algorithm:
The Proposed Algorithm is implemented in three main phases. They are as follows.
1. Sectroid (Group) Formation
2. Token Passing
3. Routing
6.4.1. Sectroid (Group) Formation:
Our basic assumption is that BS has capability of transmitting signals at various power levels.
During the initial deployment the base station sends with minimum power level. All the sensor
nodes which receive this broadcast message set their level id as 1. Then the Base Station steps up
the power to the next higher value and broadcasts the signals. Now the nodes which receive the
message for the first time will set their level id to 2. Like this the Base Station keeps on
increasing the power of transmitting signal till all the nodes in the network are covered.
After levelling is completed the base station calculates the signal strength required for the
broadcast to be heard by the farthest node in the network. Using this signal strength and a
directional antenna which has a steerable beam, the sensor network is divided into equiangular
sectors in clockwise or anti-clockwise direction. When the WSN is divided the base station sends
some control packets using which all the nodes assign their sector ids.
6.4.2. Token Passing:
Every sectroid in the sensor network is assumed to be a group. In a group each node has two
neighbours, left neighbour and right neighbour. Each group has two edge nodes, left edge node
and right edge node. Left edge node‘s left neighbour is set to zero and right edge node‘s right
neighbour is set to zero.A token is passed among all the nodes in a group and a node can transmit
its data only when it holds the token. This actually avoids lots of collisions as at any point of
time only a single node (from the whole group) is allowed to transmit the data.
47
Fig. 6.2 Part of sensor field divided into sectroids (groups)
Initially a token is initialised by the left most nodes of all the groups in the whole network. If the
Direction Flag (DF) is set to 0 then the token must be passed on to the left neighbour in the group
else to the right neighbour in the group. Since the token is initialized in the right most node, the
DF is set be 0. The first node holds the token for a fixed amount of time Hold Time (TH) which is
lesser than the transmission time (TX). So that we can make sure that a node holding the token
can complete the transmission completely. After this Hold Time the token is passed on to the
next neighbouring node (left or right depending on the status of DF). When the token reaches the
edge nodes i.e., the right most or the left most nodes, they invert the status of the DF and start
passing the token again. Since there is always a chance that the token frame may be lost due to
several reasons, we also give an acknowledgement packet for every token pass. If any node
doesn‘t get back a token acknowledgement then it reinitializes the token with the same attributes
which it stores temporarily in the cache memory.
48
Fig. 6.3 Schematic of Token Passing Procedure
In case of node failure, that is when the energy of a node reaches certain threshold value then it
sends out a panic message to both its neighbours. The neighbours will exclude the node from the
chain of token passing and they make each other as their right and left neighbours. Typically
when the edge node is about to die, there after the neighbouring node to the edge node itself
becomes an edge node.
6.4.3. Routing:
Whenever a sensor node gets a packet, it will see that it is forwarded only if it is from a higher
level. This makes the flow of data only towards the base station not away from it. This is also
considered as Directional Flooding. The packet is forwarded only when it is coming from the
adjacent sectors. This make the flow of data converge as we move towards the base station.
Effectively the leveling and clustering decreases redundancy to a greater extent. The schematic
flow of the algorithm is shown in Fig. 6.4.
1. When a node receives a data packet, it checks the level and sector id of the packet.
2. If the level id from the source is lesser than node‘s level then the packet is dropped.
3. If the level id from the source is larger than the node, then it checks whether the sector id
is from the neighbouring sectors i.e., sectors with one hop distance. If not the packet is
dropped.
49
Fig. 6.4 Schematic of the Proposed Algorithm
6.5 Simulation Evaluation:
In this section, we compare our proposed algorithm with traditional levelling [28] and PASCAL [37]
algorithms. We calculate the network life time and average number of redundant messages received by
base station using each algorithm. All implementation of protocols are done in MATLAB. Table 6.1
shows node parameters considered for this simulation study. All simulation results in this chapter are
average values after repeating 10 experimentations.
A. Experimental Methodology
In this chapter we considered all nodes are distributed randomly in a flat ground. In case of MAC
for Levelling and PASCAL, we considered simple MAC protocol CSMA/CA. For simplicity
error free links are considered.
50
TABLE 6.1
NODE PARAMETERS
Parameters Values
Simulation Area 200 X 200 m2
Number of Nodes 200 to 700
Initial Energy 500 J
Node Transmission Range 60m
Transmission Power 10 J
Reception Power 2 J
Fig. 6.5 Network Life Time
Fig. 6.5 shows the plot of Network Lifetime and Number of nodes comparing our proposed
algorithm with traditional leveling and PASCAL. The graphs show that the proposed algorithm
gives better network lifetime than leveling protocol and flooding. The network life time is
defined as the number of events taken for the failure of 10% of nodes in the whole sensor
network.
51
Fig. 6.6 Average number of Redundant Messages
Fig. 6.6 shows the plot of average redundant messages and number of nodes comparing our
algorithm with traditional levelling and PASCAL. The graph shows that the proposed algorithms
have less number of redundant messages when compared to levelling and PASCAL protocols.
Experimental results shows that our proposed algorithm has better energy efficiency as the
network life time is better when compared to levelling and PASCAL and average number of
redundant message are lesser than levelling and PASCAL protocols.
6.6 Conclusions:
In this chapter, we have proposed a token passing based cross-layer mechanism to serve the
purposes of both MAC and routing layers. Total network is hierarchically formed into groups.
Each group have a token. A node with token can only transmit data. After every TH time token is
passed to its next neighbour. All the nodes in a group are equally given chance to transmit its
data. By passing a light weight packet scheduling is done. By doing directional flooding routing
is accomplished.
Experimental results have proven that our proposed cross-layer mechanism have better energy
efficiency when compared to traditional levelling and PASCAL algorithms. Average number of
redundant messages in our proposed algorithm is lesser than traditional levelling and PASCAL
algorithms.
52
Chapter 7
Fault Repair Algorithm using Localization and Controlled mobility
The Wireless sensor networks can be used in various applications such as military, health, home
[46]. Depending on the application researchers are resolving different technical issues. In some
applications, a certain segment of the network becomes energy constrained before the remaining
network. The performance gets limited to these constrained sections. In this situation, instead of
rendering the complete network useless, the remaining energy sources should be reorganised to
form a new functional topology in the network [44]. Such energy constrained sections where
many nodes have failed are called Network Holes. In this chapter we propose a new algorithm
which uses controlled mobility [45] of some nodes to fill in the holes formed in the network
using localization techniques.
7.1 Introduction:
The applications envisioned for sensor networks vary from monitoring inhospitable habitats and
disaster areas to operating indoors for intrusion detection and equipment monitoring. In most
cases the network designer would have little control over the exact deployment of the network.
Thus an irregular deployment of sensor nodes is the assumed to be norm for these networks.
Even if the nodes are deployed uniformly at the onset of the network, as time progresses, nodes
will die randomly due to varying traffic characteristics, resulting in a non-uniform network
topology. In some applications, a certain segment of the network becomes energy constrained
before the remaining network. The performance gets limited in these constrained sections. In this
situation, instead of rendering the complete network useless, the remaining energy sources
should be reorganised to form a new functional topology in the network. Such energy
constrained sections where many nodes have failed are called Network Holes.
53
7.2 Background:
In this section, we first discuss a localization technique namely Levelling [37]. Later, we see
about controlled mobility [45].
A. Leveling
Our basic assumption is that the base station has the capability of transmitting signals at various
power levels. During the initial deployment, the base station sends a signal for level 1 with
minimum power level, all the nodes that receive this signal set their level as 1. Next the base
station increases its signal power to reach the next level and sends a level-2 signal. All the nodes
that receive this signal and those nodes that have not set their level id previously will set their
level to 2. This process continues until the base station sends signals corresponding to all the
levels. The number of levels into which the network is divided is equal to the number of different
power levels at which the base station has transmitted the signal. Apart from this level
information, the nodes need not store any location information. Levelling is done internally
without the help of any external facilities such as GPS and in this manner it differs from other
protocols that assume location information.
To counter the effects of fading in wireless channels, an alternative approach i.e., hop-count
based levelling can also be done.
If D is the distance between two levels and R is the transmission radius of any node then
R > 2 * D
The number of levels into which the network is divided depends on the number of transmission
levels at which the base station can transmit.
B. Controlled mobility
The use of mobility has been explored recently in wireless networks. Mobility can be classified
in to three categories. They are Random mobility, Predictable mobility and Controlled mobility.
[51] Authors have discussed various random mobility models available in literature. Random
Walk, Random Waypoint, Random Direction, Gauss-Markov, and Column mobility model are
few random mobility models [51]. In case of Random mobility, nodes move in random fashion
54
and the direction of motion of nodes is also unpredictable. Random mobility has been studied
earlier for improving data capacity [52] and network performance [53]. However in such cases
the latency of data transfer cannot be bounded deterministically. In the next case of predictable
mobility the motion of sensor nodes can be predicted and the use of such predictable motion was
considered. When we consider the case of controlled mobility, we deploy nodes whose mobility
is controlled (may be by pre-programming techniques). Controlled mobility is a key contributing
factor in complete exploitation of the design dimension.
Consider the situation in the Fig. 7.1, which depicts lines of constant volume such as isotherm.
Clearly the most critical region for sampling is in the top left quadrant (from reader point) where
there is a steep gradient. When we deploy a static sensor field, the sensors in the top left region
will be strained more and die sooner than rest of the nodes in the network. Using controlled
mobility of the sensors we redistribute the sensors in such a way that the energy constrained
areas will have the aid of mobile sensor nodes. This is clearly shown in the Fig 7.1.
Fig. 7.1. Illustration of control mobility
55
Table 7.1
Energy Costs for Mobile Nodes
Mobility
Primitive
Energy Cost
(J/m)
System
under test
Indoor
Robot
2.5 Kephra [49]
All Terrain
Robot
60.0 Packbot [50]
Infrastructure Assisted
Robot
10.0 NIMS
7.3 Problem formulation:
Motivation:
In some applications only a part of sensor field is strained and run out of energy. This usually
happens in applications like area monitoring where a part of network lies where the events are
most likely to occur and that part of the network is generally strained more when compared to
the other parts of the network. When this part of the network fails, eventually the whole network
fails as this is a vital section of the whole network. Though the remaining nodes are perfectly
functional, the whole network has to go waste.
Problem:
For example, consider a simple uniformly distributed sensor field as shown in Fig. 7.2. We
assume that all the sensors in the network are uniformly deployed. Therefore, area of the field is
proportional to the number of sensors in that area.
Area of covered by Level 1 = πD2
Area of covered by Level 2 = 3πD2
Area of covered by Level 3 = 5πD2
Area of covered by Level 4 = 7πD2
We generalise this equation as follows
Area covered by level r = (2r-1) πD2 (7.1)
56
Let the probability that a node in level r participate in the transmission of data to the Base Station
be PT(r).
PT(r) =
= + +
= +
PT(r) = (7.2)
By the above probability calculations we can say that nodes in the lower levels of the network
are strained more when compared to nodes in the higher levels. That is a node in level 1 has got
100% probability that it has to participate in every transmission to the base station and eventually
the probability decreases as we move to higher levels. So, the nodes in the lower levels are prone
to quick failure. If the nodes in the lower levels fail and that implies the whole network has
failed, though there are nodes in higher levels are still functioning perfectly.
Discussion
The values in Table 7.2 clearly show that the probability that a transmission occurs in 1st level is
100% and decreases to 5% depending upon the number of levels in the network. As the size of
the network becomes larger (i.e., the number of levels increases), the stress on the inner levels
becomes even more and there is a possibility that they run out of battery.
57
Table 7.2
Probability Chart
Number of
Levels
25 Levels 30 Levels 35 Levels
PT(1) 100% 100% 100%
PT(10) 84% 88.8% 91.84%
PT(15) 64% 75% 81.63%
PT(20) 36% 55.56% 67.35%
PT(25) 7.84% 30.56% 48.9%
PT(30) 6.56% 26.5%
PT(35) 5.63%
7.4 Assumptions and definitions:
1. Base Station has capacity to transmit signals at various power levels.
2. α % of the extra nodes are assumed to have mobility.
3. A mobile node has capability to rotate its direction of motion through 90⁰, 180⁰ and 270⁰
in clockwise direction.
4. Mobile nodes initially don‘t participate in the routing activity till they replace a dying
node.
5. Network Lifetime: The time to failure of 50% of nodes closer to the base station is
considered to be the lifetime. The closeness to BS is measured based on the level number.
6. Network Hole: The segment of network that is energy strained and most of the nodes
becomes non functional in that segment.
7. Panic Message (PM): The message containing the level id of the dying node.
8. Acknowledge Message (AKM): This is a reply to the PM given by a mobile node.
9. Address Query Message (AQM): The message that contains query for the level id of
nodes in that location.
10. Replacement Request Message (RRM): The message sent by dying node in reply to the
AKM.
58
7.5 Proposed Algorithm:
Considering the problem of network holes, we come up with a Fault Repair Algorithm which
uses Localization and controlled mobility to repair the damage that has occurred. Generally,
network lifetime is defined as the time taken for which 10% of the nodes in the network die. But
for the above problem statement which considers energy strain in certain sections of the network,
we redefine the network lifetime to be the time till which 10% of total nodes in the network or
50% of nodes in any level become non-functional.
We consider some percentage (α) of the sensor nodes to have mobility. As per our assumption
BS with its capacity to transmit signals at different power levels, will finish leveling of the
sensor network. Fig. 7.2 shows the model of sensor field after leveling and with mobile nodes.
Fig. 7.2 Model of localized sensor field
When a node‘s energy drops below a critical level it floods a PANIC MESSAGE (PM) in the
network. A static node broadcasts this message on the other hand a mobile node sends an
Acknowledgement Message (AKM) as a reply in the same path using routing tables. The dying
node will respond only to the first AKM, sends a RRM and ignores all further AKM messages.
59
The mobile nodes don‘t participate in sensing or routing activity until and unless they replace a
dying node. After replacing a dying node they don‘t respond to any PMs, they just take part in
the routing and sensing activities.
We define ‗L‘ as | LD – LM|, where ‗LD‘ is the level in which dying node is present and ‗LM‘ is the
current level of the mobile node. As and when the mobile node receives a RRM from dying node
it calculates the value of ‗L‘. If ‗L‘=0, the mobile node is said to be in the destination level. Else
If ‗L‘≠ 0, it moves a distance of ‗D/2‘ in a random direction. Then the mobile node broadcasts an
AQM to know its current level, it then updates the value of ‗L‘. Then checks for the below cases.
This process continues till the destination level is reached.
Fig. 7.3 Sensor Field with both static and mobile nodes
The mobile node follows an algorithmic procedure to reach the level of dying node. The
procedure is as follows:
Case I: If ‗L‘ = 0
Then the mobile node has reached the destination level.
Case II: If ‗L‘ changes
If ‗L‘ increases then take an 180o turn and travel a D/2 distance.
If ‗L‘ decreases then travel a D/2 distance in same direction.
Case III: If ‗L‘ remains same
Then take a 90o turn and travel a D/2 distance.
60
In Fig. 7.3 there is a dying node (indicated by red color) in the level ‗L‘. It floods a panic
message to all the neighbors, which in turn forwards it. Mobile nodes ‗M1‘ and ‗M2‘ receive the
‗PM‘ (The ‗PM‘ route is indicated by red color lines).Both ‗M1‘ and ‗M2‘ respond with an
acknowledgement message which is sent back to the dying node in the same path using routing
tables. (The ‗AM‘ route is indicated by a brown color line). Since ‗AM‘ sent by mobile node
‗M1‘ reaches the dying node first, the dying node sends a replacement request to ‗M1‘ and reject
the ‗AM‘ sent by ‗M2‘. Then ‗M1‘ uses the proposed algorithm to replace the dying node by
coming down to level ‗L‘ from ‗L+1‘.
7.6 Simulation Results:
In this section, we have made comparative performance evaluation of our proposed algorithm
with Flooding and traditional leveling. We consider a (100x100) m2 sensor field and the nodes
(both mobile and static) are randomly deployed all over the field.
The parameters of the nodes used in our simulation are as shown in the Table 7.3
Table 7.3
Node Parameters
Parameters Values
Communication radius 15m
Sensing radius 1.5m
Mobility cost 2.5 J/m
Total Initial Energy 1000J
Idle cost 0.0165 J/s
Transmission Power 10 J
Reception power 2 J
For the purpose of evaluating the algorithms, we carried out simulations in MATLAB by varying
the number of nodes in the network. Initially we consider 100 nodes in the whole network in
which 30% of the nodes are mobile. Then we increase the number of nodes in consecutive steps
of 100 up to 500 nodes.
61
Fig. 7.4, 7.5 show the plots of Network Lifetime and Number of nodes comparing our proposed
algorithm with traditional levelling and flooding. The graphs show that the proposed algorithm
gives better network lifetime than levelling protocol and flooding.
Fig. 7.4. Plot of Proposed Algorithm VS Levelling
Fig. 7.5. Plot of Proposed Algorithm VS Flooding
7.7 Conclusions:
In this chapter, we show that the proposed algorithm allows a sensor network to adaptively
reconfigure and repair itself when few sections of network becomes resource constrained,
instead of rendering the complete network useless, mobile nodes fill the network holes. We see
that introduction of controlled mobility has brought significant network sustainability.
62
Chapter 8
Sensor Node deployment using an Innovative Non-Uniform
Sampling phenomenon
In Wireless Sensor Networks (WSN), the placement of sensor nodes plays a vital role. Rapid
deployment of nodes will not give accurate location of event occurrence. To increase accuracy in
event location we are proposing a novel approach for deciding topology of Wireless Sensor
Networks in this chapter. The key idea is to use an innovative approach, namely Non-Uniform
sampling to deploy sensor nodes at initial network setup.
8.1 Motivation:
Consider sensor network which monitors temperature in a field or a forest. Simplest way to place
nodes is placing them uniformly (using phenomenon of periodic sampling). But it is very clear
that such a sampling scheme doesn‘t in cooperate any information about the phenomenon being
monitored. It is well known that non-uniform sampling schemes are preferable to the periodic
sampling in such situations. In this chapter we propose an innovative approach for non-uniform
sampling. The phenomenon being monitored corresponds to a 1 or 2 or 3 dimensional continuous
time signal. The results of this chapter enable an optimal sampling of a phenomenon being
monitored by wireless sensor networks.
8.2 Introduction:
Many natural and man-made signals are analog (continuous) in nature. Most of the present day‘s
communications deals with digital signals. Two basic operations need to be performed to make a
signal analog to digital, namely sampling and quantization. The process of discretizing an analog
signal, called sampling is an important operation and large body of research literature is
generated. The instants at which the samples are obtained form a stream of uniform events,
which can be depicted graphically as a sampling point process. Characteristic features of the
sampled signals to a large extent depend on the patterns of the point process generated and
used for sampling [53].
In the context of digital signal processing (DSP), usually sampling is considered as deterministic
63
or periodic. In such a model, signal samples are considered at time intervals with a constant and
known duration. Researchers were interested in approximation (of analog signals by discrete
time signals) that is as close as possible using sampling. These efforts culminated in Whittaker-
Kotel‘nikov-Nyquist—Shannon (WKNS) sampling theorem [54]. This theorem ensures that the
original band-limited signal can be recovered from the samples (periodic) provided the sampling
rate is at least equal to (or greater than) twice the highest frequency to which the original analog
signal is band limited. If the sampling rate is below the Nyquist rate, reconstruction error, called
―aliasing‖ occurs. Since a signal is represented by finitely many quantization levels, there is
a reconstruction error (of original signal by the samples) due to quantization noise.
8.3 Motivation for Non-Uniform Sampling:
As was established relatively long ago, the application of sampling alone is not sufficient. The
sampling model is not applicable when fluctuations in sampling instants cannot be ignored or
when signal samples can be obtained only at irregular or even random time intervals. Studies
have indicated that randomness in sampling is not always harmful, sometimes random
irregularities in sampling can even be beneficial. These irregularities, if properly introduced,
provide for instance, such a useful effect as the suppression of aliasing. And such sampling itself,
usually, is considered as non-uniform. Existing non-uniform sampling might be realized either as
randomized or pseudo-randomized sampling. The ultimate goal of various sampling schemes is
to decrease the data rate (number of bits to represent an analog signal) while at the same time
providing sufficient amount of accuracy. Our goal in this chapter is to approach ―optimal
sampling‖ using an innovative approach. Also, in literature, there are few results devoted to non-
uniform sampling of non-band-limited one/two/three/multi dimensional functions. In the
following section, we discuss one possible approach to reconstruction of non-band limited
signals from non-uniform sampling.
8.4 An Approach to Non-Uniform Sampling:
To describe the approach presented in this section, we need the following information
from real analysis.
Lp Space: A function f(x) is said to belong to L
p Space if
64
(8.1)
The class of functions in L2 is said to belong to Hilbert space. Normally, many functions which
arise in practical applications belong to Hilbert space.
Consider a one dimensional function f(x). As in the mathematical discipline, Real Analysis,
decompose f(x) in the following manner:
f(x) = g(x) – h(x), ( g(x) = f(x) for all ―x‖ such that f(x) > 0 ) (8.2)
where g(x), h(x) are non-negative functions corresponding to the positive part and negative
part of the function f(x).
In the following, we discuss the non-uniform sampling of g(x). The same approach holds true for
the non-uniform sampling of h(x). It is very clear that once g(x) and h(x) are reconstructed from
the samples, the function f(x) can easily be reconstructed.
Consider the function g(x). Let it belong to Lp space for some . To be specific, let the
function belongs to L1 (clearly all bounded amplitude signals of finite duration are integrable). In
the following, we discuss the non-uniform sampling of g(x). Normalize the function in the
following manner i.e. define a new function r(x) such that
(8.3)
It is clear that r(x) is a probability density function. We now sample this probability density
function. Our goal is to design a sampling scheme such that r(x) can be reconstructed as
accurately as possible from the samples. The following approximation approaches are
considered.
The probability density function r(x) has an associated probability distribution function
m(x). The problem of sampling boils down to finding a piecewise linear distribution
function that approximates (as closely as possible) the distribution function
corresponding to r(x). The approximation procedure can be iterative (using a sequence of
approximating rectangles). In statistics literature, there are well developed
procedures for approximating a density function (or equivalently the
65
distribution function) with respect to some useful/meaningful metric (between
the original density and the approximating densities). Those results are invoked
in the context of non-uniform sampling.
The probability density function r(.) is associated with a random variable R. Our goal
is to approximate R by a random variable M as closely as possible. From
probability literature, Least Mean Square estimation approach is very clear. The
minimum mean square estimate of R using M is the conditional expectation of R
with respect to M. The formal details of this approach follow from [55].
We now propose a new non-uniform sampling approach.
Essential Idea:
Consider the probability density function obtained through the above procedure as in (8.3). For
the purposes of OPTIMAL SAMPLING (resulting in minimum possible reconstruction
error), it is logical to include more samples in the region where there is large
probability mass and small number of samples in the region where there is small
probability mass.
Note: The problem of approximating a probability density function by a set of rectangles
is well studied in statistics literature. They are effectively transferred to arrive at the
notion of optimal sampling.
Two engineering approaches to the problem are summarized in the following. For the sake of
simplicity, let the support (domain) of the probability density be finite (bounded). It
should be noted that both the engineering approaches are HYBRID SAMPLING
approaches i.e. partly uniform sampling and partly non-uniform sampling approaches.
8.4.1 First engineering approach:
A. Coarse partition determination:
Consider the RANGE of the probability density function r(x). Based on the minimum as
well as maximum possible values, divide the range into finitely many intervals. Using
these intervals, find the corresponding COARSE partitions on the domain. i.e., Divide the
domain (support) of the probability density function , r(.) into finitely many COURSE
partitions. Using one of the various numerical integration methods, compute the
66
probability mass in each of the regions of the course partition.
B. Fine partition determination:
Decide the ―smallest probability mass ‖ in any region of the ―fine partition‖. Using this
mass, divide each COARSE partition into FINE PARTITIONS.
8.4.2 Second engineering approach:
Consider the case where the domain of the function is bounded.
1. Based on the dynamic range of the domain of the function, divide it into equally
spaced samples. This constitutes the COARSE PARTITION. It corresponds to uniform
sampling.
2. Using one of the numerical integration techniques, compute the area in each of the
intervals of the COARSE PARTITION. Since the function is normalized, these areas
correspond to probability values. It is most logical to assign, large number of samples to
the interval where there is large probability mass. Thus the FINE PARTITION leads
to non-uniform sampling of the intervals in the COARSE PARTITION.
First let us consider one of the intervals obtained after COARSE sampling. Consider
the following diagram illustrating non-uniform sampling of the function restricted to
this interval.
Fig. 8.1 Uniform Sampling, grid with bin distance ‗b‘
67
Locally the function is approximated by a monotone increasing/decreasing function. Suppose
the total area of the function restricted to this interval be P0. Suppose, in this interval, the
function is approximated by rectangles whose base is bj and height is hj. Thus our goal
is to find, a finite set of rectangles approximating the function restricted to this interval.
Thus, we readily have
0
1
pbhM
j
jj (8.4)
Let bj=b for j=1, 2, 3… M
Locally, on the interval, the function is approximated utilizing a tangent
approximation to the function. Let the slope of the tangent (gradient) line be m. Thus, we
readily have that
h1+mb=h2
h2+mb=h2
.
.
.
hi+mb=hi+1 for 1≤i≤ (M-1)
Furthermore, we readily have
.0
1
phbM
i
i
Utilizing the above sequence of equations, we readily have
012
)1(pbm
MMhMb
68
In the above expression, the known quantities are {h1, m, p0}. We also know the length of the
interval, say . This length is divided into M smaller units. Thus we have
Mb
Substituting for ‗b‘ (from above equation), in previous equation
012
)1(pm
MhM
M
(8.5)
The above equation can be explicitly solved for M and the original non-negative function can
be sampled.
3. Optimal sampling of coarse partition:
Let us consider the probability density function f(x) with the following statistics:
Conditional mean
Conditional variance
69
Fig. 8.2. Coarse partition of f(x)
Initially f(x) is coarse partitioned (have k uniform intervals) as shown in above figure. Our main
goal in this chapter is to allocate samples in such a way that the reconstruction error is minimum,
given that
N1 + N2 . . . . . . . . +Nk = M
Where Ni is the number of samples in ith
coarse partition (ith
interval)
In statistics literature for optimal allocation of the samples in ith
interval, we have [58], [59]
(8.6)
Selection of Ni is the optimal way to allocate samples in ith
interval (coarse partition).
70
Notes: The above novel non-uniform sampling techniques can easily be generalized for two or
three or multi-dimensional signals.
8.5 Non-Uniform Sampling: Information -Theoretic Approach:
In the following, we briefly summarize the ideas from source coding in communication theory.
Suppose we have a Discrete Memory less Source i.e., the source output random variables are
Independent, Identically distributed (iid).
Consider the case where the source output random variables are DISCRETE i.e., each random
variable 1X assumes, say k values with probabilities }.,....,,{ 21 kppp It is well known from
source coding theorem that the average code length is greater than or equal to entropy.
Huffman coding achieves the lower bound on average code length.
Now we return to the problem of sampling. Consider the COARSE PARTITION of the function.
It is clear from the second engineering approach that the intervals in the COARSE
partition are all of the same length. Let the number of intervals be k and let the corresponding
probability masses be }.,....,,{ 21 kqqq
We have a random variable X that is non-negative and continuous. The support of such a
random variable is divided into, say L intervals i.e.;
Probability {0 < X < 1x } = 1q ,
Probability { 1x X < 2x } = 2q , and so on till
Probability { kk xXx 1 } = Lq .
It is clear that },....,,{ 21 Lqqq are known.
Also, unlike Huffman coding, we want to assign large number of bits the interval with larger
probability mass. The total number of samples is fixed (M).
Problem Formulation:
Let ni be the number of samples allocated to ith
interval.
71
Unlike Huffman coding, instead of keeping a lower bound on expected number of samples, we
propose to minimize variance of a number of samples.
i.e., minimize and maximize subjected to and ni are integers.
This problem is being attempted for solution in [57].
8.6 Conclusion:
In this chapter, we have presented an innovative approach of non-uniform sampling. By placing
a sensor node at each sample point, nodes can be placed non-uniformly in the sensor field. Non-
uniform placement of nodes increases accuracy in calculation of event location.
72
Chapter 9
Conclusions
In this thesis, we have proposed an energy efficient architecture for Cognitive Radio based
Sensor Networks. In this architecture we have decoupled spectrum sensing network from data
gathering network. Spectrum sensing network is enabled with Cognitive Radio and has
capabilities of Dynamic Spectrum Allocation. Data gathering network is Wireless Sensor
Network; this network monitors the field and allows data transmissions.
We have presented energy efficient protocols for data gathering network. In chapter 4, we have
presented an energy efficient routing protocol. In which initially we have placed a virtual grid
and then each grid is considered as a cluster. Using cluster head information data packets are
routed to sink (Base Station) in an energy efficient way. This protocol has shown better energy
efficiency when compared to existing routing protocols. It has been shown through simulations.
In chapter 5, we have proposed a TDMA based Medium Access Control (MAC). This protocol
uses TDMA based MAC scheme where the time has been divided into frames and then time
slots. These slots are further divided into sections as channel request (CR), channel allocation
(CA) and data section. This protocol has very good energy conserving properties comparing with
traditional TDMA. Another interesting property of the protocol is that it has relatively very less
delay when compared to traditional TDMA.
In chapter 6, we have proposed an energy efficient cross layer design of the MAC and Routing
protocol namely Energy Efficient cross-layer protocol design by using token passing mechanism
for WSN. Total network is hierarchically formed into groups. Each group have a token. A node
with token can only transmit data. After every TH time token is passed to its next neighbour. All
the nodes in a group are equally given chance to transmit its data. By passing a light weight
packet scheduling is done. By doing directional flooding routing is accomplished. Experimental
results have proven that our proposed cross-layer mechanism have better energy efficiency when
compared to traditional levelling and PASCAL algorithms. Average number of redundant
messages in our proposed algorithm is lesser than traditional levelling and PASCAL algorithms.
73
In chapter 7, we have proposed fault repair algorithm. This proposed algorithm allows a sensor
network to adaptively reconfigure and repair itself when few sections of network becomes
resource constrained, instead of rendering the complete network useless, mobile nodes fill the
network holes. We see that introduction of controlled mobility has brought significant network
sustainability.
In chapter 8, we have presented an innovative approach of non-uniform sampling. By placing a
sensor node at each sample point, nodes can be placed non-uniformly in the sensor field. Non-
uniform placement of nodes increases accuracy in calculation of event location.
9.1 Future Work:
In our thesis we have decoupled spectrum sensing network and data gathering network. Then we
have proposed energy efficient protocols for data gathering network. One can investigate energy
efficient spectrum sensing and spectrum allocation protocols for spectrum sensing network.
74
Bibliography
[1] Cavalcanti D., Das, S., Jianfeng Wang,Challapali, K. ,― Cognitive Radio Based Wireless Sensor
Networks‖, ICCCN '08, Aug. 2008.
[2] FCC Spectrum Policy Task Force, ―Report of the spectrum efficiency working group,‖
FCC, Technical Report 02-155, Nov 2002.
[3] ZigBee Alliance, Zigbee Specification, Document 053474r14, November 2006.
[4] IEEE 802.15.4 Standard, ―Wireless medium access control (MAC) and physical layer (PHY)
specifications for low-rate Wireless Personal Area Networks (LR-WPANs),‖ 2003 Edition.
[5] I. Howitt, J. Gutierrez, ―IEEE 802.15.4 Low Rate – Wireless Personal Area Network Coexistence
Issues,‖ in Proceedings of IEEE WNCN 2003.
[6] D. Cavalcanti, R. Schmitt, and A. Soomro, ―Achieving Energy Efficiency and QoS for Low-Rate
Applications with 802.11e,‖ in the IEEE WCNC 2007, Hong Kong, March 2007.
[7] IEEE 802.11 Standard, ―Wireless LAN medium access control (MAC) and physical layer
(PHY) specifications,‖ 1999 (Reaff 2003) Edition.
[8] FCC, ET Docket No 03-237 Notice of inquiry and notice of proposed Rulemaking,
November 2003. ET Docket No. 03- 237.
[9] M. A. McHenry, P. A. Tenhula, D. McCloskey, D. A. Roberson, and C. S. Hood,
―Chicago spectrum occupancy measurements & analysis and a long-term studies
proposal,‖ in Proceedings of First InternationalWorkshop on Technology and Policy for
Accessing Spectrum (TAPAS). ACM, 2006, p. 1.
[10] M. Buddhikot, ―Understanding dynamic spectrum access: models, taxonomy and
challenges,‖ in Proceedings of IEEE International Symposium on New Frontiers in
Dynamic Spectrum Access Networks (DySPAN), April 2007, pp. 649–663.
[11] S. Haykin, ―Cognitive radio: brain-empowered wireless communications,‖ IEEE
Journal on Selected Areas in Communications, vol. 23, no. 2, pp. 201–220, 2005.
[12] FCC, ET Docket No 03-222 Notice of proposed rulemaking and order, December
2003.
[13] J. Mitola III, Cognitive radio for flexible mobile multimedia communication, in:
Proc. IEEE International Workshop on Mobile Multimedia Communications
(MoMuC) 1999, November 1999, pp. 3–10.
[14] J. Mitola III, Cognitive radio: an integrated agent architecture for software defined
75
radio, Ph.D Thesis, KTH Royal Institute of Technology, 2000.
[15] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, ―A Survey on Sensor
Networks‖, IEEE Communications Magazine, Vol. 40, No. 2, pp. 102-114, Aug.
2002.
[16] G. Zhou, J. A. Stankovic, S. H. Son, ―Crowded Spectrum in Wireless Sensor
Networks‖, in Proc. Third Workshop on Embedded Networked Sensors (EmNets
2006), 2006.
[17] C. Shen, C. Srisathapornphat, and C. Jaikaeo, ―Sensor Information Networking
Architecture and Applications,‖ IEEE Pers. Commun., Aug. 2001, pp. 52–59.
[18] G. Hoblos, M. Staroswiecki, and A. Aitouche, ―Optimal Design of Fault Tolerant
Sensor Networks,‖ IEEE Int‟l. Conf. Cont. Apps., Anchorage, AK, Sept. 2000, pp.
467–72.
[19] J. M. Rabaey et al., ―PicoRadio Supports Ad Hoc Ultra- Low Power Wireless
Networking,‖ IEEE Comp. Mag., 2000, pp. 42–48.
[20] Ozgur B. Akan, Osman B. Karli, and Ozgur Ergul ,―Cognitive Radio Sensor
Networks‖, The Magazine of Global Internetworking - Special issue title on
networking over multi-hop cognitive networks, Volume 23 Issue 4, July/August
2009.
[21] J. Zhao, H. Zheng, and G. Yang, ―Distributed Coordination in Dynamic Spectrum
allocation Networks,‖ Proc. IEEE DySPAN, Nov. 2005, pp. 259–68.
[22] Fei Hu, Sunil
Kumar, ―QoS Considerations in Wireless Sensor Networks for
Telemedicine‖, proceeding of SPIE ITCOM Conference, September 2003, Orlando,
FL.
[23] N. Zhang M. Wang, and N. Wang;‖ Precision Agriculture – a Worldwide Overview‖
Computers and Electronics in Agriculture, v. 36, pp. 113-132, 2002.
[24] Deepak G., B. Krishnamachari, Alec Woo, Davi Culler, Deborah Estrin, Stephen
Wicker, ―An Empirical study of Epidemic Algorithms in Large Scale Multihop
Wireless Networks‖, Intel Research, IRB-TR-02-003, Mar. 14, 2002.
[25] K.Kalpakis, K. Dasgupta and P. Namjoshi, ―Maximum lifetime data gathering and
aggregation in wireless sensor networks‖, in preceesing of IEEE International
Conference on Networking (NETWORKS‘02), Atlanta, GA, Aug. 2002.
76
[26] Joe Albowicz, Alvin Chen, and Lixia Zhang, ―Recursive Position Estimation in
Sensor Networks, IEEE International Conference on Network Protocols (ICNP),
2001.
[27] Zyymunt J.Has.Joseph, Y.halpe, and Li(Erran), ― Gossip-based Ad hoc routing‖,
IEEE/ACM Transactions on Networking, Vol.15, No.3, June 2006.
[28]Santhosh Bhima, Anil Gogada, and G. Rama Murthy, ―Level Controlled Gossip
based Tsunami Warning Wireless Sensor Networks‖, Sensor and Transducers
Journal (ISSN 1726-5476), PP.27-34, vol. 106, No.7, July 2009.
[29] Arshad Jhumka, Sandeep Kulkarini ―On the design of mobility-tolerant TDMA-
based media access control (MAC) protocol for mobile sensor networks‖, 4th
international conference on Distributed computing and internet technology,
Bangalore, India, December 2007.
[30] S. Dulman, L. Hoesel, T. Nieberg, P. Havinga, ―Collaborative Communication
Protocols for Wireless Sensor Networks‖, European Research on Middleware and
Architectures for Complex and Embedded Systems Workshop, Pisa, Italy, April 2003.
[31] W. Ye, J. Heidemann, and D. Estrin, ―An energy-efficient MAC protocol for
wireless sensor networks,‖ in In Proceedings of the 21st International Annual Joint
Conference of the IEEE Computer and Communications Societies (INFOCOM
'02), New York, June 2002.
[32] Einstein, A.,―Investigations on the Theory of the Brownian Movement‖. Dover
Publications,Mineola, NY (1956).
[33] Johnson, D.B., Maltz, D.A.,―Dynamic Source Routing in ad hoc Wireless
Network‖,ch. 5, pp. 153–181. Kluwer Academic Publishers, Dordrecht (1996).
[34] Musolesi, M., Mascolo, C., ―Designing mobility models based on social networks
theory‖, Mobile Computing and Communications Review 1(2) (2007) .
[35] Demirbas, M., Arora, A., Mittal, V., Kulathumani, V.: ―A fault-local self-stabilizing
clustering service for wireless ad hoc networks‖. IEEE Trans. Parallel Distrib.
Systems 17(9), 912–922 (2006).
[36] V. Rajendran, K. Obraczka, and J.J. Garcia-Luna-Aceves, ―Energy- Efficient,
Collision-Free Medium Access Control for Wireless Sensor Networks‖, The First
ACM Conference on Embedde Networked Sensor Systems (SenSys 2003), Los
77
Angeles CA, Nov. 2003.
[37] Md.Aquil Mirza, Ramamurthy Garimella, “PASCAL: Power Aware Sectoring
Based Clustering Algorithm for Wireless Sensor Networks”, The International
Conference on Information Networking (ICOIN), Chiang Mai, Thailand. January 20-
24th,2009.
[38] A. J. Goldsmith and S. W. Wicker, ―Design challenges for energy-constrained ad
hoc wireless networks,‖ IEEE Wireless Communications Magazine, 2002.
[39] P. Skraba, H. Aghajan, A. Bahai, ―Cross-layer optimization for high density sensor
networks: Distributed passive routing Decisions,‖ in Proc. Ad-Hoc Now‟04,
Vancouver, July 2004.
[40] L. van Hoesel, T. Nieberg, J. Wu, and P. J. M. Havinga, ―Prolonging the lifetime of
wireless sensor networks by cross-layer interaction,‖ IEEE Wireless Communications,
vol. 11, no. 6, pp. 78 - 86, Dec. 2004.
[41] S. Cui, A. J. Goldsmith, and A. Bahai, ―Energy-constrained modulation
optimization,‖ IEEE Transactions on Wireless Communications, vol. 4, no. 5, pp.
2349–2360.
[42] S. Cui, A. J. Goldsmith, and A. Bahai, ―Joint modulation and multiple access
optimization under energy constraints,‖ GLOBECOM, 2004.
[43] M. L. Sichitiu, ―Cross-layer scheduling for power efficiency in wireless sensor
networks‖, INFOCOM, 2004.
[44] Saurabh Ganeriwal, Aman Kansal and Mani B. Srivastava “Self Aware Actuation
for Fault Repair in Sensor networks” Proceedings of the 2004 IEEE, international
Conference on Robotics & Automation New Orleans, LA Aprll2004.
[45] Aman Kansal, Mohammad Rahimi, Deborah Estrin, William J Kaiser, Gregory J
Pottie, and Mani B Srivastava ―Controlled Mobility for Sustainable Wireless Sensor
Networks” in Proceedings of the 1st Annual IEEE Communications Society
Conference on Sensor and Ad Hoc Communications and Networks (SECON '04), pp.
1–6, Santa Clara, Calif, USA, October 2004.
[46] M Grossglauser and DNC Tse., ―Mobility increases the capacity of adhoc wireless
networks‖, IEEE/ACM Transactions on Networking (TON), 10(4):477–486, 2002.
[47] D. Estrin, R. Govindan, J. Heidemam, S. Kumar, ‗Next Century Challenges:
78
Scalable Coordination in Sensor Networks‖, ACM Mobicom, Seattle, WA, August 1999
[48] G. T. Sibley, M. H. Rahimi, G. S. Sdchatme, ―Robornote: A Tiny Mobile Robot
Platform for Large-Scale Sensor networks―, EEE ICM, 2002.
[49] KephraRobot .http://www.k-team.com/robots/kephra/.
[50] Packbot. http://www.packbot.com/.
[51] Tracy Camp, Jeff Boleng, Vanessa Davis. ―A Survey of Mobility Models for Ad-
hoc Network Research‖, Wireless Communication & Mobile Computing (WCMC),
vol 2, no 5, pp. 483-502, 2002.
[52] Aman Kansal, Mohammad Rahimi, Deborah Estrin, William J Kaiser. ―Controlled
mobility for sustainable wireless sensor networks‖, IST Annual IEEE Comm‘s Soc
Conference on Sensor and Ad-hoc Comm, 2004.
[53] http://www.edi.lv/dasp-web/sec-5.htm
[54] A.V. Oppenheim, A. Willsky and H. Nawab, ―DiscreteTime Signal Processing‖,
Prentice Hall of India.
[55] A. Papoulis, ―Probability, Random Variables and Stochastic Processes‖, Mc Graw
Hill.
[56] F. Marvasti, ―Nonuniform Sampling: Theory and Practice‖, Kluwer
Academic/Plenum Publishers, NY 2001
[57] G. Rama Murthy, ―Innovative Approaches to Non-Uniform Sampling‖, IIIT
Technical Report in Preparation.
[58] Cochran, W.G. ―Sampling Techniques‖, Asia Publishing House.
[59] B.L. Agarwal, ―Basic Satistics‖, Paperback Publishers.
79
Related Publications
1. SandhyaSree Thaskani, K. Vinod Kumar, G. Rama Murthy, ―Mobility Tolerant TDMA Based
MAC Protocol for WSN”, accepted to IEEE- ISCI-2011.
2. SandhyaSree Thaskani, K. Vinod Kumar, G. Rama Murthy, ― Energy Efficient Cross-
Layer Design Protocol by Using Token Passing Mechanism for WSN”, accepted to
IEEE- ISCI-2011.
3. K. Vinod Kumar, G. Lakshmi Phani, K. Venkat Sayeesh, SandhyaSree Thaskani, G. Rama
Murthy, ―Fault Repair Algorithm using Localization and Controlled mobility in WSN‖ ,
SaCoNaS Globecom 2010 proceddings, December 2010.
4. SandhyaSree Thaskani, G.Rama Murthy, ―Application of Topology Under Control Wireless
Sensor Networks in Precision Agriculture‖, IETE 41st Mid-Term Symposium, April 2010.
5. SandhyaSree Thaskani, G.Rama Murthy, ―A Novel Routing / Fusion algorithm for Topology
aware Wireless Sensor Networks‖, Global Journal of Computer Science and Technology
(GJCST), Vol. 10 Issue 4, June 2010.
6. G.Rama Murthy , SandhyaSree Thaskani and Narendra Ahuja, ―Innovative approach to Non
uniform sampling‖, International Journal of Recent Trends in Engineering, Vol 2, No. 3,
November 2009.