Threshold Sensitive Routing Protocol For Wireless Sensor Networks
Threshold Sensitive Routing Protocol For Wireless Sensor Networks
Threshold Sensitive Routing Protocol For Wireless Sensor Networks
Introduction
1
Introduction
2
Chapter 1 Introduction
consideration whether after being the CH it will sustain for next rounds or not.
Biological systems exhibit scalability, robustness, resistance to individual failures
thereby following the principle of “improvement” which considers every single
individuals capacity. Hence, recently evolutionary algorithms (EAs) have attained
researcher’s attention so as to develop hierarchical cluster-based routing protocols for
WSNs. Primary goal of these genetic algorithm based cluster routing protocols is to
form clusters dynamically so that energy consumption in the network is diminished
thus, extending network lifetime. In genetic algorithm based routing like GA, HCR
and ERP, the location of nodes exploited and cluster formation done as selection
through set of computed solutions or chromosomes [4], [5], [6]. They have already
proved their efficiency when compared to LEACH. However, they are still liable to
remarkable modifications which can make them more stable. Stability is said to
be attained for an interval by the network when there is no dead node present in
the network. After this period network nodes tend to die in a frequent manner.
The work done in this paper tries to attain more stability and network lifetime by
introducing transmission distance and residual energy as the main factors for the
fitness function used in EAs. Further, final path is also formulated which tends to
minimize the burden of direct or multi-hop transmission by the cluster heads directly
to BS.
This thesis work demonstrates the influence of factors stated in formulating set
of solutions through EAs selection, crossover, mutation and fitness evaluation. And
proves the liability of proposed methodology through comparison of network life and
energy dissipation with protocols namely, LEACH, TEEN and ERP.
3
Chapter 1 Introduction
(b) Transceiver: Transmitter and receiver are the communication devices used
in a sensor node. Transmitter and receiver integrated together in a single
device is known as a transceiver. The transmission channel used are radio
frequency (RF), infrared and optical communication. As the lasers (optical
communication) need line-of-sight to communicate and infrared’s broadcasting
capacity is limited, WSN applications make use of radio frequency based
communication. Its operational state comprise of transmit, receive, sleep and
idle.
(c) External Memory: In general, on-chip flash memory is used due to their low
cost and storage capacity. However, requirements of memory is application
dependent. The memory comprise of two categories of memory namely, user
memory which contains application related or personal data and program
4
Chapter 1 Introduction
memory. Program memory is used for programming the device and may also
contain devices identification data.
(d) Power Source: Sensor nodes consume energy while sensing, communicating
and processing data. But, more energy is required while communication of
data. Power storage is done in batteries or capacitors. A wide variety of
sensor nodes try to scavenge energy from the nearby environment.
(e) Sensors: A sensor node may comprise of one or more small size, low energy,
adaptive sensors. Sensor is a hardware device which produces a quantifiable,
measurable response when a physical phenomenon like temperature, pressure
being sensed changes. Thus, they measure the physical data of the parameter
being monitored. The signal produced is analog which is digitized by the ADC
(analog-to-digital converter) and then propagated to the micro controller for
further processing. They are classified into three categories namely, passive
omni-directional sensors, narrow-beam sensors and active sensors.
(a) Area observing: Area observing is a basic use of wireless sensor network. In
area observing, the WSN is deployed over an area where some phenomenon
like temperature, pressure etc is to be observed. Such as in military sensors
are utilized to detect any adverse interruption.
(b) Monitoring health care: The two kinds of medical applications are wearable
and embedded. In wearable, devices are utilized on the body surface of a
human or positioned closely to the client. Whereas, the implantable devices
5
Chapter 1 Introduction
(e) Data logging: Wireless sensor networks are used for gathering information
while observance of environment such as checking the temperature, pressure,
heat etc. The facts thus collected can then be used to show how the functioning
frameworks. WSNs provide with ”live” information contrary to the customary
logging systems.
6
Chapter 1 Introduction
7
Chapter 1 Introduction
8
Chapter 1 Introduction
1.3.6 Scalability
If one wants to add or remove sensor nodes from the field and have the operation
going on it should be possible in the WSN. For this to be possible routing protocols
should also be scalable. It can be achieved by number of nodes not transmitting
every while
9
Chapter 1 Introduction
1.4 Motivation
A sensor network is said to be accurate and reliable only when sensed data forwarded
comes from large number of member nodes aggregated data. This can happen only
when nodes do not die abruptly making the network unreliable. And the battery
power decides how long network will be alive. Thus, death of first node decides
stability of the network. Presently, protocols stay alive for long time but after
10
Chapter 1 Introduction
certain point of time degrade steeply. The degradation starts after the death of first
node in the network. This fact led to work upon factors which can reduce early
death of nodes over time.
1.6 Summary
This chapter introduced about the trending wireless sensor networks with a
description regarding sensor node architecture, applications of WSN and the design
challenges of routing in WSN. The next chapter overviews previous work done int
he field of WSN routing protocols.
11
Chapter 2
Literature Review
This chapter gives a brief description of various categories of sensor network routing
protocols.
12
Chapter 2 Literature Review
forwarded.
2.2.3 Cougar
The authors have stated that the existing sensor networks assume sensors to be
preprogrammed for a particular function and transmit data to a centralized front-end
13
Chapter 2 Literature Review
where further data aggregation, analysis and storage is done [9]. This approach lacks
changing the system behavior as when needed and also, trades costly communication
for local computation which are cheap. Thus, they have introduced the Cougar
approach to tasking sensor networks through declarative queries which does not
resemble preprogrammed sensors. And thus, have also introduced sensor network as
a processing platform. Its approach is driven by two main factors i.e., declarative
queries and in-network data aggregation and processing. In declarative queries users
and application programs issued queries are processed remains abstracted from the
network layer functions. This is attained by a query proxy which lies between the
network layer and application layer and interacts with both the layers present at
every nodes query layer. Reduced energy consumption acquired by in-network data
aggregation and processing by the query proxy at each node.
14
Chapter 2 Literature Review
source information all along until the edge of network. From that point extreme
node communicates to BS. Thus, grid formed and BS forwards query to nearest
dissemination point. And through grid query finally reaches source which sends
back according to query received. On change of network topology there can occur
abundant overhead with respect to the recalculation of grid [11].
15
Chapter 2 Literature Review
16
Chapter 2 Literature Review
2.2.8 Pegasis
The authors have proposed an improvement over Leach protocol named, Pegasis.
Comprises of three phases namely, chain formation, leader selection and data
transmission [14]. A single node has the responsibility to send the fused data to
BS every round (along chain) which is chosen randomly every round. On select
From all ends of the network data received in a chain. Each node between the chain
or chains receive and send data. Thus, nodes communicate or disseminate only to
neighbor after fusing with ones own data. Chain is made among sensors through
greedy approach and not like traveling salesman problem because it is intractable.
This is done keeping in mind the radio communication energy parameters. First
order radio model as in Leach used here. For construction of chain all nodes have
global knowledge about the network. Pegasis considers that every round any single
node will transmit to BS but practically, multi-hop communication comes in far
away node scenario. Also, there can be an excessive delay when chain head is at the
far end of network.
17
Chapter 2 Literature Review
18
Chapter 2 Literature Review
19
Chapter 2 Literature Review
Multi-hop communication
S. Lindsey et al, 2002
[14]
Communication chain created (Greedy Dynamic topology
Algorithm) needed can be an
overhead
Sensor node sends its data to its nearest Method to obtain node
neighbor location is uspecified
Chain head selection random and transmit to
sink
A. Manjeshwar et al,
2001 [15]
Reactive Multiple level cluster
formation adds to
Transmits only when HT and ST satisfied complexity
A. Manjeshwar et al,
2002 [10]
Advancement of [15] Method to implement
threshold based
function is an overhead
Support periodic data collection and
time-critical situations
20
Chapter 2 Literature Review
of an optimizer which selects the best solution based on fitness function. Fitness
function is based upon parameters such as energy consumption, size of cluster,
distance to sink, distance among cluster members.
Evolutionary based routing protocol is also a GA based routing protocol in
which fitness function comprise of clusters compaction and separation together with
number of clusters [5].
2.3 Summary
This chapter described about the various paradigms used to classify sensor network
routing protocols and based on them the routing protocols designed till date.
Chapter 3 carries on with the proposed methodology.
21
Chapter 3
Proposed Methodology
3.1 Background
In genetic algorithm based energy efficient clustering each node in the network is
represented in a chromosome by a bit. 1s represent head nodes and 0s represent
normal nodes. -1 denote dead nodes. The basic framework of GA is used as given
in Algorithm 1. Population consists of a number of individuals called chromosomes
[4]. This set of population define the solution to the problem, here clustering.
Initialization of population done randomly. After that, new generations evolve
through selection, crossover, mutation and test based on fitness for a predefined
number of times. The fitness function quantitatively defines the ”enhancement”.
The selection process decides the chromosomes which will crossover and create new
chromosomes. Crossover is a binary genetic operator which acts upon two parents [4].
These include single-point, two-point, multi-point, arithmetic operator. It is done
after the selection process depending upon the rate of crossover which in general,
is around 80 to 95 percent. Chromosomes with better fitness value are selected
to crossover. Offspring thus produced join the population already present either
replacing the chromosomes which crossover or adding to the population. Through
crossover the generation evolved will only have the characteristics of the parents
22
Chapter 3 Proposed Methodology
23
Chapter 3 Proposed Methodology
transmit packet to the BS. This leads to a network with less stability period. Stress
is put on stability period because it is observed in various routing methods that
node death graph changes steeply sometimes after the stability period is over i.e.,
after the first node dies. Also, after this point reliability of data cannot be sured
since, nodes die abruptly until all nodes are dead over a period of rounds.
Hence, we have redefined the fitness function to be used in the GA based
hierarchical clustering. It is a protocol based on the genetic algorithm and the
threshold sensitiveness. Threshold sensitive means the network nodes transmit data
only when there is a change based upon given threshold. To efficiently design
clustering solutions through GA, we need to consider factors namely transmission
distance, average minimum distance, energy left per node, cluster head transmission
energy and cluster head count. These factors constitute the fitness function.
(a) Transmission distance (T D): It is the total distance a packet covers from
the point of formation to the point where it gets aggregated i.e., CH. This
factor needs to be minimized. Equation (3.1) gives the transmission distance
of normal member nodes.
CHs
XX
TD = dis(mn, CHi ) (3.1)
∀n∈Cli
i=1
(b) Average minimum distance (AM D): It is the average of distance among
all the cluster heads in a network. Distance evaluated as the Euclidean distance
between cluster heads. This needs to be maximized conforming that the
clusters formed are not very close to each other. Equation (3.2) gives the
average minimum distance of cluster nodes. nc denotes the number of cluster
head.
X
AM D = dis(CHi , CHj )/nc (3.2)
∀Cli ,Clj ,Cli 6=Clj
24
Chapter 3 Proposed Methodology
(c) Energy left per node (EL): It is the total residual energy of every node
present in the network. This factor needs to be maximized conforming that the
energy per node is dissipated uniformly over the network so as to increase the
stability period. Equation (3.3) gives the summation of energy left per node.
Eo i denotes the initial energy of node i and EDi denote the energy dissipated
by node i.
n
X
EL = (Eoi − EDi ) (3.3)
i=1
mn
X nc
X
T ED = ET Xi + (ERXj + EDAj ) (3.4)
i=1 j=1
Initially a random set of individuals or chromosomes are taken. For these the
fitness function value is calculated and the general GA steps are imposed on the
chromosomes as given in Algorithm 1. Fitness function is calculated for each
chromosome as in Equation (3.5). f f denotes the fitness factor and wt denotes
the weight of corresponding fitness factor initially taken to be equal for all the three
factors considered.
X
F itness Chromo(i) = (wti ∗ f fi ) (3.5)
i
f f1 = T D ÷ AM D (3.6)
25
Chapter 3 Proposed Methodology
f f2 = T ED ÷ EL (3.7)
f f3 = nc (3.8)
3.2.2 Assumptions
Base station has the knowledge regarding each sensor node in the field. This
includes ID-location of the nodes and distance of node from BS.
Base Station is located far away from the sensor node field.
Each sensor node has the knowledge of its own energy and location of the BS.
26
Chapter 3 Proposed Methodology
3.2.3 Phases
For every number of transmission of data packets WSN undergoes three phases.
(a) Cluster formation phase: Base station implements the genetic algorithm.
It has the information regarding every node in the sensor network. And each
network node has an identity (ID) known to BS and the node itself. On
computation of clusters and regarding cluster heads through GA, BS transmit
the information to respective cluster heads. On that event each cluster head
transmits a signal to the corresponding member nodes so that they can send the
sensed data for further number of rounds to this cluster head for aggregation.
CH percentage is predefined and so initialization done through randomness as
given in Equation 3.11.
(b) Path formation phase: In this phase, chain of cluster-heads of all the
clusters are formed. Every cluster node does not need to propagate the data
packet directly to the BS. A path is formed among cluster heads such that
data aggregation could be done along the path and final aggregated data is
send to the BS. For this to happen, BS transmits the information regarding
the path formation to the extreme cluster head of chain formed. That node
sends aggregated data along the chain to next node. And finally data sent
through cluster head with utmost capacity. This path evaluation is done at
BS based on the Euclidean distance between cluster heads, between BS and
27
Chapter 3 Proposed Methodology
every cluster head and the energy left with them. Figure 3.1 depicts the path
among the cluster heads towards the BS.
(c) Data transmission phase: Data transmitted only when the two conditions
are satisfied: Firstly, the sensed value should be greater than the user
adjustable, predefined value termed as hard threshold (HT ). And secondly,
when difference between latter sensed value and former sensed value is greater
than another user adjustable, predefined value termed as soft threshold
(ST ). In case of data with utmost importance i.e., abrupt increase in sensed
parameter, node directly sends the data to the cluster head and then gets
propagated to the BS.
28
Chapter 3 Proposed Methodology
3.3 Summary
This chapter gave an in-depth explanation of GA based method used for cluster
formation and the fitness function factors used. It gave formulations regarding the
methodology and step-wise phases involved in the protocol. The next chapter gives
a quantified evaluation of method used in this chapter through simulation results.
29
Chapter 4
Simulation Result
30
Chapter 4 Simulation Result
q
do = f s /mp (4.1)
Graph plotted in Figure 4.1 shows the nodes alive vs number of rounds for
GA-TSRP, ERP, TEEN and LEACH. It is clear from the figure that nodes are
alive for greater extent of rounds in case of GA-TSRP than the others. The steep
downfall of the number of nodes starts from around round number 900, 800, 550,
450 for GA-TSRP, ERP, TEEN and LEACH respectively.
Graph in Figure 4.2 displays total energy left in the sensor network with
increasing round numbers. LEACH tend to show steep decrease in the residual
energy. Whereas, TEEN and ERP has better record of energy left. GA-TSRP
shows very little steepness for considerable number of rounds proving that uniform
usage of node energy done for a longer period of rounds.
Graph plotted in Figure 4.3 shows the total energy dissipation (cumulative) over
the rounds for all the four approaches. This shows that in GA-TSRP nodes dissipate
much uniformly than others for a longer period of time (rounds). Reasons for this
increase in network life-time could be stated as below:
31
Chapter 4 Simulation Result
Figure 4.1: Comparison based on nodes alive in network over the rounds
Figure 4.2: Comparison based on energy left in network over the rounds
32
Chapter 4 Simulation Result
Figure 4.3: Comparison based on cumulative energy dissipation over the rounds
Fitness function takes into consideration the residual energy present with nodes
which will be selected as CHs in any iteration further. It also, takes into
consideration both the transmission distance factor relating to cluster heads
and member nodes.
All the cluster heads and especially ones at the extremes of the network does
not have to send the aggregated data packets directly to the BS. Instead,
a chain is formed among the cluster heads, every time there needs to be a
transmission such that only the cluster-heads qualifying on both distance and
residual energy basis becomes the end conveyor to BS. In LEACH, TEEN and
ERP all the cluster heads send the sensed and aggregated data directly to the
BS.
Cluster head does not depend totally on the threshold value and chances, as in
LEACH and TEEN but also the energy left of the node after the transmission
in previous rounds.
33
Chapter 4 Simulation Result
4.2 Summary
Tis chapter gave an analytical evaluation of proposed method thus, proving the
efficiency achieved with respect to network lifetime and energy associated. The next
chapter gives an overview of possible future works related to the method analyzed.
34
Chapter 5
Simulations show that GA-TSRP is efficient than the former protocols with respect
to stability period and substantial energy dissipation. These are the outcome
of addition of transmission distance and residual energy factors into the fitness
function of GA for cluster formation. And also, the node capability based route
formation among the cluster heads to BS. Effect of variant weight(wt) on fitness
function(f f ) can be further scrutinized to give more accurate fitness function. And
thus, determining which factor plays a role to which extent so as to determine energy
efficient solution. Clear picture can be drawn when worked upon wireless simulation
environment by considering other sensor node activities and shortcomings too. Thus,
future work can be done regarding testing the proposed method in a network based
simulator and deduce weight’s importance in fitness function.
35
Bibliography
[1] Mohamed Guerroumi, Al-Sakib Khan Pathan, Nadjib Badache, and Samira Moussaoui.
Strengths and weaknesses of prominent data dissemination techniques in wireless sensor
networks. International Journal of Communication Networks and Information Security
(IJCNIS), 5(3), 2013.
[2] Jamal N Al-Karaki and Ahmed E Kamal. Routing techniques in wireless sensor networks: a
survey. Wireless communications, IEEE, 11(6):6–28, 2004.
[4] Sajid Hussain, Abdul Wasey Matin, and Obidul Islam. Genetic algorithm for hierarchical
wireless sensor networks. Journal of Networks, 2(5):87–97, 2007.
[5] A Attea Baraa and Enan A Khalil. A new evolutionary based routing protocol for clustered
heterogeneous wireless sensor networks. Applied Soft Computing, 12(7):1950–1957, 2012.
[6] Sajid Hussain and Abdul Wasey Matin. Base station assisted hierarchical cluster-based
routing. In Wireless and Mobile Communications, 2006. ICWMC’06. International
Conference on, pages 9–9. IEEE, 2006.
[7] Wendi Rabiner Heinzelman, Joanna Kulik, and Hari Balakrishnan. Adaptive protocols for
information dissemination in wireless sensor networks. In Proceedings of the 5th annual
ACM/IEEE international conference on Mobile computing and networking, pages 174–185.
ACM, 1999.
[8] Chalermek Intanagonwiwat, Ramesh Govindan, and Deborah Estrin. Directed diffusion: a
scalable and robust communication paradigm for sensor networks. In Proceedings of the 6th
annual international conference on Mobile computing and networking, pages 56–67. ACM,
2000.
[9] MJ Handy, Marc Haase, and Dirk Timmermann. Low energy adaptive clustering hierarchy
with deterministic cluster-head selection. In Mobile and Wireless Communications Network,
2002. 4th International Workshop on, pages 368–372. IEEE, 2002.
36
Bibliography
[10] David Braginsky and Deborah Estrin. Rumor routing algorthim for sensor networks.
In Proceedings of the 1st ACM international workshop on Wireless sensor networks and
applications, pages 22–31. ACM, 2002.
[11] Fan Ye, Haiyun Luo, Jerry Cheng, Songwu Lu, and Lixia Zhang. A two-tier data dissemination
model for large-scale wireless sensor networks. In Proceedings of the 8th annual international
conference on Mobile computing and networking, pages 148–159. ACM, 2002.
[12] Ming Liu, Jiannong Cao, Guihai Chen, and Xiaomin Wang. An energy-aware routing protocol
in wireless sensor networks. Sensors, 9(1):445–462, 2009.
[13] Yong Yao and Johannes Gehrke. The cougar approach to in-network query processing in
sensor networks. ACM Sigmod Record, 31(3):9–18, 2002.
[14] Stephanie Lindsey and Cauligi S Raghavendra. Pegasis: Power-efficient gathering in sensor
information systems. In Aerospace conference proceedings, 2002. IEEE, volume 3, pages
3–1125. IEEE, 2002.
[15] Arati Manjeshwar and Dharma P Agrawal. Teen: a routing protocol for enhanced efficiency
in wireless sensor networks. In Parallel and Distributed Processing Symposium, International,
volume 3, pages 30189a–30189a. IEEE Computer Society, 2001.
[16] Arati Manjeshwar and Dharma P Agrawal. Apteen: A hybrid protocol for efficient routing and
comprehensive information retrieval in wireless sensor networks. In Parallel and Distributed
Processing Symposium, International, volume 2, pages 0195b–0195b. IEEE Computer Society,
2002.
[17] Georgios Smaragdakis, Ibrahim Matta, and Azer Bestavros. Sep: A stable election protocol
for clustered heterogeneous wireless sensor networks. Technical report, Boston University
Computer Science Department, 2004.
[18] Femi A Aderohunmu, Jeremiah D Deng, et al. An enhanced stable election protocol (sep) for
clustered heterogeneous wsn. Department of Information Science, University of Otago, New
Zealand, 2010.
[19] S Faisal, Nadeem Javaid, Akmal Javaid, MA Khan, Safdar Hussain Bouk, and ZA Khan. Z-sep:
Zonal-stable election protocol for wireless sensor networks. arXiv preprint arXiv:1303.5364,
2013.
[20] Sajid Hussain, Abdul Wasey Matin, and Obidul Islam. Genetic algorithm for energy efficient
clusters in wireless sensor networks. In ITNG, pages 147–154, 2007.
37
Bibliography
[21] Hari Rangarajan and JJ Garcia-Luna-Aceves. Reliable data delivery in event-driven wireless
sensor networks. In Computers and Communications, 2004. Proceedings. ISCC 2004. Ninth
International Symposium on, volume 1, pages 232–237. IEEE, 2004.
[22] V Loscri, G Morabito, and S Marano. A two-levels hierarchy for low-energy adaptive clustering
hierarchy (tl-leach). In IEEE Vehicular Technology Conference, volume 62, page 1809. IEEE;
1999, 2005.
[23] Deepak Ganesan, Ramesh Govindan, Scott Shenker, and Deborah Estrin. Highly-resilient,
energy-efficient multipath routing in wireless sensor networks. ACM SIGMOBILE Mobile
Computing and Communications Review, 5(4):11–25, 2001.
[24] Ming Liu, Jiannong Cao, Guihai Chen, and Xiaomin Wang. An energy-aware routing protocol
in wireless sensor networks. Sensors, 9(1):445–462, 2009.
38