You are on page 1of 13

A Meta-heuristic Global Path Selection Approach to Prolong the

Lifetime of Heterogeneous Wireless Sensor Network using


Multi-Objective Particle Swarm Optimization.
April 24, 2014
Abstract
Wireless Sensor Networks are collection of compact-
size, relatively inexpensive computational nodes that
measure local environmental conditions or other pa-
rameters and forward such information to a central
point for appropriate processing. Wireless Sensor
Network supports a wide range of useful applications
such as Air trac control, Battleeld management,
Defence system, Habitat monitoring etc. Harnessing
the potential benets of WSNs requires a high-level
of self-organization and coordination among the sen-
sors, to perform the tasks required to support the
underlying application. At the heart of this collab-
orative eort to achieve communications is the need
for the wireless sensor nodes to self-organize into a
multihop wireless network. Consequently, the design
of ecient communications and network protocols for
WSNs becomes crucial for wireless sensor nodes, to
carry out successfully the mission for which they are
deployed. This paper describes an meta-heuristic ap-
proach for prolonging the lifetime of WSNs. Here
the entire region space of network is partitioned into
a number of clusters using Fuzzy C-means algorithm.
The sensor node of one cluster transmits the collected
information to the Base Station either by single hop
or multihop path. Therefore path selection is essen-
tial to provide communication for a longer duration.
This paper propose a multiobjective solution to select
the optimal path for transmitting gathered informa-
tion to the Base Station using Multiobjective Particle
Swarm Optimization. The main objective functions
are 1. Minimize the distance between two participat-
ing sensor nodes along the path. 2. Maximize the
residual energy of this participating nodes. 3. Select
the participating node which has not been selected in
the previous turn.
keyword-Wireless Sensor Network, Multiobjective
Particle Swarm Optimization(MOPSO), Particle
Swarm Optimization(PSO), multihop.
1 Introduction
A WSN is composed of a number of wireless sensor
nodes which form a sensor eld and a sink. These
large number of nodes with low-cost, low-power, and
capable of communication at a short distance perform
limited computation in wireless platform[1]. Wireless
Sensor Networks are extremely versatile and can be
deployed to support a wide variety of applications
in many dierent situations. The sensor nodes can
be stationary or mobile. The deployment of sensor
nodes depends on the nature of the application [2].
In environmental monitoring and surveillance appli-
cations, for example, sensor nodes are deployed in
an ad hoc fashion so as to cover the specic area to
be monitored [3]. In health care applications, smart
wearable wireless devices and biologically compatible
sensors can be implanted strategically within the hu-
man body to monitor vital signs of the patient under
surveillance. After deployment, Sensor nodes are self-
organized into an autonomous wireless ad hoc net-
work, which requires a very little maintenance [4].
1
The main task of wireless sensor nodes is to sense
and collect data from a target domain, process the
data and send the processed information to the base
station either directly or through some intermediate
point. Achieving this work eciently and eectively
requires the development of an energy-ecient rout-
ing protocol to set up paths between sensor nodes
and Base station. The path selection strategy must
be on the benet of the maximization of the lifetime
of the sensor network. The environmental character-
istics and energy limitation of the sensor nodes make
the routing problem very challenging [5,6].
The optimization refers to determining one or more
solutions of a given problem which correspond opti-
mum values of one or more objectives. It becomes one
active area of research as have many real world prob-
lems and is signicantly growing for last few years [7].
Therefore a suitable optimization technique is always
required. Most of the real world problems consists of
several objectives that are need to be optimized at the
same time. To solve multi-objective problems with
conventional mathematical logistic technique, a sin-
gle optimum solution is generated from a set of solu-
tions in one run. Evolutionary computing paradigm
is most suitable to solve multi-objective problems be-
cause Evolutionary algorithms are population based
and generate a set of solutions in one run [8,19].
The main objective of this paper is to propose an
optimized solution to the problem by using multi-
objective particle swarm optimization(MOPSO) al-
gorithm. The main objective of the multi objec-
tive(MO) problem is nding the set of accepted so-
lution is called Pareto front. These acceptable trade
o solutions give more ability to the user to make an
informed decision by seeing a wide range of near opti-
mal solutions that are near optimum from an overall
standpoint. Multiobjective Particle swarm Optimiza-
tion can easily handle constraints of discrete nature.
The main advantages of MOPSO methods are
1. It does not require a prior knowledge of the rela-
tive importance of the objectives
2. There is a set of acceptable trade-o near optimal
solution. Theis set is called Pareto front [35-37] or
optimality trade-o surfaces.
To maximize the lifetime of wireless sensor network
the optimization factors are the followings.
1. Minimize the distance between two participating
sensor nodes along the path.
2. Maximize the residual energy of this participating
nodes.
3. Select the participating node which has not been
selected in the previous turn.
Here, the optimization of these multiple objectives
be solved by adopting a multiobjective optimization
framework. This paper concentrates on multiobjec-
tive strategies, their assumptions, optimization prob-
lem formulations and results.
The rest of this paper is organized as follows: A brief
overview of PSO and MOPSO are outlined in section
II, section III discuss the literature review. Section
IV describe the problem formulation and the pro-
posed algorithmic approach, while section v discuss
the simulation results. Finally a concluding remarks
are given in section VI.
2 PSO and MOPSO - A brief
overview
The Particle Swarm Optimization is a new popula-
tion based optimization technique ,which was rst
proposed be Kennedy and Eberhart in 1995[9]. It was
rst intended for simulating social behaviour , as a
stylized representation of the movement of organisms
in a bird ock or sh school. Kennedy and Eberhart
describes many philosophical aspects of PSO and
Swarm intelligence.The PSO begins with a random
population and searches for tness. PSO is similar
to a genetic algorithm (GA) in that the system is ini-
tialized with a population of random solutions. In the
PSO , particles will evolve by cooperation and com-
petition among the individuals themselves through
generation instead of using genetic operators. Many
variations of PSO have been developed including con-
strained ,multiobjective and combinatorial versions
[10,11,12].
Each particle ies in n-dimensional search space. The
velocity and position of a particle is adjusted dynam-
ically according to its own ying experience and its
companions ying experience. Each particle keeps
track of its coordinate position in the problem space
2
which are associated with the best solution , it has
achieved so far. This value is called pbest. Another
best value that is tracked by the global version of the
particle swarm optimizer is the overall best value and
its location obtained so far by any particle in the pop-
ulation. This location is called gbest [14]. In an iter-
ation , velocity and position of each particle changes
toward its pbest and gbest locations as shown in the
following equations. For an n dimensional search
space, the ith particle of the swarm is represented by
an n-dimensional vector, Xi = (xi1 ,xi2 .....xin ).The
velocity of this particle is represented by another n
dimensional vector Vi = (vi1, vi2...... vin).The pre-
viously best visited position of the ith particle is de-
noted as P
i
= (p
i1
,p
i2
, . . . . . . , . . . ,p
in
). g

is the in-
dex of the best particle in the swarm. The velocity of
the i
t
h particle is updated using the velocity update
equation given by
v
k+1
i
= wv
k
i
+c
1
r
1
(pbest
i
x
k
i
) +c
2
r
2
(gbest
i
x
k
i
) (1)
and the position is updated using
x
k+1
i
= x
k
i
+ v
k+1
i
(2)
Here, x
k
i
is current position of particle i at iteration
k, which has velocity v
k
i
and v
min
v
k
i
v
max
and the
pbest
i
is the historical best position of x
k
i
and gbest
i
is the global best position in the populations history.
The parameter w is the inertia weight factor, c
1
and
c
2
are acceleration constants and r
1
and r
2
are uni-
form random numbers between 0 and 1. The parame-
ter v
max
determines the resolution with which region
between the present position and target position is
searched. If vmax is too high, particles may y past
the good solutions. If v
max
is too small, particles may
not explore suciently beyond local solutions. The
constants c
1
and c
2
represent the weighting of the
stochastic acceleration terms that pull each particle
toward pbest and gbest positions. Low values makes
particles to roam far from target regions before being
tugged back. On the other hand, high values result
in abrupt movements toward, or past, the target re-
gions. It can be proved that for convergence (c
1
+c
2
)
must be less than or equal to 4. The inertia weight
w provides a balance between global and local explo-
ration and exploitation, and on average results in less
iterations required to nd a suciently good solution.
It is typically set at 0.4. Eberhart and Shi (2000) in-
dicate that the optimal strategy is to initially set w
to 0.9 and reduce it linearly to 0.4, allowing initial
exploration followed by acceleration toward an im-
proved global optimum [12,13].
When tackling multiobjective problems however, a
few modications must be made. First, the objective
is to nd not one global best solution but a set of
solutions has come which is compared with Pareto
Front. For this, an archive of non dominated solu-
tions is kept, when all non dominated solutions that
one found at each iteration are stored. The steps fol-
lowed by MOPSO are
1. Initialize the swarm and archive.
2. For each particle in the swarm.
a. Select Leader from the archive.
b. Update Velocity.
c. Update Position.
3. Update the archive of non dominated solutions.
4. Repeat.
In the multiobjective optimization problem, each par-
ticle might have a set of dierent leaders from which
just one can be selected in order to update its posi-
tion. Leader is a particle that is used to guide another
particle towards better regions of the search spacer.
The particle is the member of the swarm. Each par-
ticle represent the potential solution to the problem
being solved. The leaders are usually stored in a dif-
ferent place from the swarm that will call external
archive. The solutions contained in external archive
are used as leaders when the position of the particle
of the swarm has to be updated [19,20].
3 Literature Survey
Many research projects in the last few years have ex-
plored lifetime maximization of WSN from dierent
perspective. Reducing power consumption of sensor
nodes is one of the major issue to maximize the life-
time of sensor node as well as sensor network also.
Clustering of sensor network is one of the solution
to prolong the lifetime of sensor network. There are
so many clustering algorithms in WSN [24]. LEACH
is the rst clustering algorithm that was proposed
3
for reducing power consumption. In LEACH, the
clustering task is rotated among the nodes, based
on duration. The operation of LEACH is divided
into rounds, where each round begins with a setup
phase, when data transfer to the sink node occurs.
In LEACH, a sensor node is elected as the clus-
ter head according to a distributed probabilistic ap-
proach. Non cluster nodes decide which cluster to
join based on the signal strength. LEACH assumes
that all nodes have the same amount of energy ca-
pacity in each election round which is based on the
assumption that being a cluster head results in same
energy consumption for every node. Such an as-
sumption is impractical in most application scenar-
ios. Head extends the basic scheme of Leach by using
residual energy and node degree or density as a met-
ric for cluster selection to achieve power balancing.
It operates in multihop networks, using an adaptive
transmission power in the inter-clustering communi-
cation. Both schemes are fully distributed, terminate
in constant number of iterations and incur low mes-
sage overhead. These methods are suitable for pro-
longing the network lifetime rather than for entire
needs of WSN. In LCA, a node becomes the cluster
head if it has the highest identity among all nodes
within one hop of itself or among all nodes within
one hop of one of its neighbours. LCA [27,28] is im-
proved by LCA2. The LCA2 [30] algorithm as the
cluster head the node with the lowest identication
number among all nodes that are neither a cluster
head nor one within one-hop of the already chosen
cluster heads. In CEFL(Cluster head Election us-
ing Fuzzy logic) [36] algorithm ,fuzzy logic method is
adopted to select the cluster head. Most of the clus-
tering methods rely on synchronization.
Traditional optimization methods include linear,
nonlinear and quadratic programming and interior
point methods [14,15]. The computational complex-
ity grow exponentially with problem size. Resource
requirements and cost of mathematical programming
engines used for linear ,nonlinear programming make
them unattractive for resource constrained nodes.
This is the motivation for metaheuristic algorithm.
Particle swarm optimizer(PSO) is one of the most
accepted evolutionary computation algorithms that
has attracted an incremental interest of scientists and
engineers. While analytical methods suer from slow
convergence and the curse of dimensionality, heuris-
tic based swarm intelligence can be an ecient al-
ternative. PSO, part of the swarm intelligence fam-
ily, is known to eectively solve large-scale nonlin-
ear optimization problems [16,17]. Computational
intelligence-based techniques, such as genetic algo-
rithm (GA) and particle swarm optimization (PSO)
can be solutions to these problems. GA [31] rep-
resents a particular class of evolutionary algorithms
that uses techniques inspired by evolutionary biology
such as inheritance, mutation, natural selection, and
recombination (or crossover). GA can rapidly locate
good solutions, even for dicult search spaces but
having some disadvantages associated with it. In GA
to dene the tness function properly is dicult to
achieve.GA may have a tendency to converge towards
local optima rather than the global optimum of the
problem. The operation on dynamic data sets is dif-
cult to implement.
PSO has its roots in articial life and social psy-
chology, as well as in engineering and computer sci-
ence. It utilizes a population of particles that y
through the problem hyperspace with given veloci-
ties. At each iteration, the velocities of the individ-
ual particles are stochastically adjusted according to
the historical best position for the particle itself and
the neighbourhood best position. Both the particle
best and the neighbourhood best are derived accord-
ing to a user dened tness function. The move-
ment of each particle naturally evolves to an optimal
or near-optimal solution. PSO is a computational
intelligence-based technique that is not largely af-
fected by the size and nonlinearity of the problem,
and can converge to the optimal solution in many
problems where most analytical methods fail to con-
verge. As any routing problem in Wireless environ-
ment is NP-hard problem, so it need to be optimized
[18].
In order to solve the routing problem in Wireless Sen-
sor Network (WSN), which is NP-hard in nature; an
ecient solution needs to be developed. To solve this
optimization problem PSO has been adopted under
the current research work. WSN is a type of wireless
network, where the major concern is the energy con-
straints of the sensor nodes.
4
The main task of sensor nodes is to sense and col-
lect data from a target domain, process the data and
transmit the information back to specic sites where
the underlying application resides. Achieving this
task eciently requites the development of an energy-
ecient routing protocol to set up path between and
the data sink. The path selection must be such that
the lifetime of the network is maximized. The char-
acteristics of the environment within which sensor
nodes typically operate, coupled with severe resource
and energy limitation, make the routing problem very
challenging. In order to increase the lifetime of such
network, the computations at the sensor nodes need
to be minimized. In PSO, the amount of computa-
tion need to be performed is very less in compare
to other existing evolutionary techniques. Hence a
routing algorithm based on PSO is a good choice to
solve the routing problem in WSN. Again, the conver-
gence rate of PSO is also high in compare to other
algorithms. Therefore the number of iterations re-
quired to nd the optimal path in WSN is also less
in case of the proposed algorithm. This in turn will
minimize the space requirement of the system, which
is another critical issue to solve the routing problem
in WSN [20,21].
A novel cluster-based approach is introduced us-
ing PSO [32], which was found to function bet-
ter than LEACH-C(Low Energy Adaptive Cluster-
ing Hierarchy-Centralized)[27, ] protocol in terms of
ecient energy consumption technique. In LEACH-
C, the cluster formation is done at the beginning of
each round using centralized algorithm by the Base
station. The Base station uses the information re-
ceived from each node during the setup phase to nd
a predetermined number of cluster heads and cong-
ures the network into clusters. The cluster group-
ing are then chosen to minimize the non-cluster head
nodes to transmit their data to their respective clus-
ter heads [22,8].
In [33] ,the author proposed a PSO-based cluster
head selection scheme to nd the tness position for
the head nodes, which makes the algorithm semi-
distributed and comparative more energy ecient.
The selection criteria of the objective function are
based on the residual energy, intra-cluster distance,
node degree and head count of the portable cluster
heads. The expected number of packet transmissions
along the estimated path through the cluster head is
reected in the proposed energy consumption model.
The PSO-SD reduces the intra-cluster distance from
the cluster members to the cluster head and retrans-
mission computation for collided packets also assists
in the derivation of overall energy consumption in the
network.
ICWAQ [34] uses ecient and fast searching features
of the ABC algorithm to optimize clustering of the
nodes in the selection process of cluster-heads den-
ing routing gateways. Articial Bee Colony algo-
rithm, simulating the intelligent foraging behaviour
of honey bee swarms is used in this clustering tech-
niques. ICWAQ protocol prolongs the network life-
time as well as minimize signal transfer delays com-
paring other techniques. ICWAQ maintains distance
values by using signal strengths to achieve routing op-
erations without GPS which must incur extra cost.
A meta-heuristic Global path selection approach to
prolong the lifetime of Heterogeneous wireless sensor
network using Multiobjective Particle Swarm Opti-
mization is proposed here. During the process of op-
timization, sensors are formed in clustered architec-
ture. The three objectives for nding one best path
from one source node to the Base station has taken
into consideration so that energy consumption will
be less for data transmission. The objectives of this
proposed scheme are
1) Select the path having minimum Euclidian dis-
tance from source node of one cluster to the base
station 2) Select the next CH node on the path hav-
ing maximum residual energy . 3) If one CH node
has already been selected by i
th
iteration then that
node never be selected in (i + 1)
th
iteration.
The simulation results also prove the improvement of
the performance with increase in number of iterations
in the algorithm.
4 Proposed work
The sensor nodes are highly resource constrained.
Energy is one of the major issues of the sensor nodes.
In wireless sensor network most of the energy is con-
sumed during transmission and it is further increased
5
with the distance, as energy consumption is directly
proportional to the square of the distance among the
nodes. In order to minimize energy consumption and
increase lifetime of the network, we must keep the dis-
tance under consideration and it is possible by the ar-
chitecture design of the network and ecient routing
schemes. In cluster based architectures, cluster head
are over loaded with long range transmissions to the
base station and with additional processing responsi-
bility of data aggregation. Due to these responsibili-
ties, cluster heads nodes are drained of their energy
quickly. It is unsuitable in the case of homogenous
wireless sensor networks that cluster heads are regu-
lar nodes but they communicate for longer distance
with the base station because of inter cluster com-
munication. This leads to energy imbalance among
the clusters. Energy ecient communication tech-
niques mainly focus on minimizing the transmission
energy, while in long range applications the transmis-
sion energy is dominant in the total energy consump-
tion. The transmission energy generally depends on
the transmission distance. There are so many algo-
rithmic approach on Wireless Sensor Network to pro-
long the lifetime of Wireless Sensor Network has been
proposed. This paper describes an meta-heuristic ap-
proach for prolonging the lifetime of WSNs. This
meta-heuristic approach is an iterative process which
guides a subordinate heuristic by combining intelli-
gently dierent concepts for exploring and exploiting
the search space. Some learning strategies are used
to structure information in order to nd eciently
near-optimal solutions. Here the entire region space
of network is partitioned into a number of clusters
using Fuzzy C-means algorithm. The sensor node of
one cluster transmits the collected information to the
Base Station either by single hop or multihop path.
Therefore path selection is essential to provide com-
munication for a longer duration. This paper pro-
poses a multiobjective solution to select the path for
transmitting gathered information to the Base Sta-
tion using Multi-objective Particle Swarm Optimiza-
tion. This algorithm is developed based on certain
considerations on the system such as
1. All sensor nodes are static.
2. All nodes must be heterogeneous in nature with
respect to energy.
3. The Base Station is located outside of the WSNs.
In order to increase the lifetime of the static sen-
sor nodes in WSN, the algorithm is divided into four
stages. The rst stage is the Set up phase where the
clusters are to be formed with n number of randomly
distributed sensor nodes using Fuzzy C-means clus-
tering algorithm [25]. Then the node having maxi-
mum residual energy have been selected as Cluster
Head. The Second stage dene the Fitness function
of particle of the search space to nd the route from
one source node to Base Station .The Fitness func-
tion F dened with three parameter f
1
,f
2
and f
3
.The
functional parameter f
1
leads to minimum Euclidian
distance from one source node of one cluster to the
Base station. The functional parameter f
2
leads to
a node having maximum residual energy along the
route and f
3
denes the selection of a node at time
instant t where the node has not been selected at just
the previous instant of time (t 1) in the path to the
Base station. At the third stage the entire route(from
any source node to Base station) to be selected using
Multiobjective Particle Swarm optimization. At the
last stage the data transmission is allowed along the
optimally selected path and then read the energy sta-
tus of the participating CH nodes. If the energy level
of cluster head is below the threshold level then se-
lection of the next higher energy node has been taken
place. During transmission of data it may be the case
that instead of choosing the best path we have chose
the path where residual energy is maximum, as it is
optimal therefore for every transmission it will select
this path as a result other nodes will be left unused
and energy depletion rate will be faster for this rea-
son step 3.3 has been taken into consideration. The
threshold is the 10 of the corresponding initial energy
of each node.
The algorithm is as follows:
Stage 1: Set up phase (Formation of clusters
and selection of cluster head)
Step1: Randomly deploy n number of sensor
nodes over an application area. Initialize the energy
of each node and xed the coordinate position of
each.
Step2: Apply Fuzzy c-means clustering algorithm to
partition the region space into k number of Clusters
6
using steps 2.1 and 2.2
Step 2.1: choose the nodes having maximum energy
level as the head node in any cluster.
Step 2.2: Construct one rank table based on energy
dierentiation of the nodes.
Stage 2: Determine the Fitness function of
particle to nd out the route (from one source
node to the Base Station)
Step3: Determine the Fitness function to nd out
the routes from one source node of a cluster to the
Base station using steps 3.1, 3.2 and 3.3
Step 3.1: select the path having minimum Euclidian
distance from one source node of a cluster to the
base station using following function:
f1 = min(d) (3)
where d is the Euclidian distance between x
i
and x
j
of two sensor node.
Step 3.2: select the path having maximum residual
energy using the following function:
To compute energy consumption for each trans-
mission sending and receiving, the Radio energy
dissipation model is used .The energy used to
transmit q bit of data at a distance d for each sensor
node is given by:
E
Tx
(q, d) = q
Elec
+ qf
s
d
2
(4)
The energy used to receive data for each node is
given by
E
Rx
(q) = qE
Elec
(5)
Where, E
Elec
is electronic energy, f
s
is power loss
of free space.
The depreciation of energy ED of a node due to
packet transmission is given by
E
D
= E
f
+ E
r
+ E
s
(6)
where E
f
is the energy depreciation of forwarding a
packet and E
r
is the energy depreciation of receiving
a packet and is the rate of depreciation of packet
forwarding and is the rate of packet receiving
and is the rate of depreciation of packet sending.
Therefore, Energy requirement of a node is given by
E = ( + + )ET
x
(q, d) E
D
(7)
Where, (+ + )ET
x
(q, d) is the total energy con-
sumption for d distance based on the usage in the
network and ED is depreciation of energy. The node
having residual energy is
E
resi
= E
i
E (8)
, where E
i
is the initial energy of a node.
f2 = max(E
resi
) (9)
Step 3.3: Keep a node status table to maintain
information of a node that has been used very
recently (i.e at time instant t ). The status table
contains value either 0 or 1 with respect to a node.
If the selected CH node has already been selected
at t
th
round then the node goto the sleep state at
t + 1
th
round else the CH node goto the active state
for participating in the routing . When the node is
in sleeping state its status will be 0 otherwise 1.
f3 = select
t
(i) (10)
,
where select
t(i)
will return 1 when i
th
is selected at
t
th
round otherwise return 0 . In order to evaluate a
solution a set of criteria are combined and the tness
function is obtained that guides the optimization
process. Fitness function F is calculated in the
following way.
F = 0.4 f1 + 0.4 f2 + 0.2 f3 (11)
where,
f1 = min(d) (12)
f2 = max(E
resi
) (13)
f3 = select
t
(i) (14)
7
Stage3: Select the optimal path using MOPSO
Step 4: For all paths (swarmorparticle) from
any source node to base station do steps 4.1 and 4.2.
Step 4.1: Dene a search space region in which we
consider paths, they are guided by their best known
position i.e the source node position and base station
position.
Step 4.2: Initialize position and velocity of particle
.The source x
i
is the initial
position and v
i
is the initial velocity of the particle.
For each particle do steps 4.2.1 and 4.2.2. Step 4.2.1:
Calculate the tness F of the particle using Fitness
function.
Step 4.2.2: Velocity and position of the particle
changes towards its pbest and gbest locations as
shown in the following equations
v
i
(t) = wv
i
(t1)+c1r1(x
pbset
x
i
(t))+c2r2(x
gbest
x
i
(t))
(15)
Where, w is the inertia weight which is set at 0.4
and r1, r2 any random number belonging in the
range (01) . The constants c1 and c2 represent the
weighting of the stochastic acceleration terms that
pull each particle towards pbest and gbest positions.
The inertia weight w provides a balance between
global and local exploration. Update position of the
particle using position
x
i
(t) = x
i
(t 1) + v
i
(t) (16)
Step 4.3: If the destination base station has not
reached along the path then goto Step 4.2
Stage 4: Data transmission phase along
with checking the energy status of the partici-
pating nodes.
Step 5: Allow the data transmission along the op-
timally selected path.
Step 6: If all CH having energy threshold energy
then go to step 4
else if there are any more alternate node in each clus-
ter
then choose the next higher energy node from rank
table
else if 60% of node in network having energy below
the threshold level
then the network is dened as Dark zone
Step 7: Stop.
5 Simulation and Result
The implementation is carried out in C language to
evaluate the performance of our algorithm. There are
200 sensor nodes distributed across the simulation
area of 300mx300m. The base station is positioned
at the outside of the simulation area. The sensor
nodes are heterogeneous in nature i.e the initial
energy of the nodes are dierent. The transmission
radius of each sensor node is set as 10m. For a node
in the sensing state, packets are generated at a con-
stant rate of 1packer/sec. It is also considered that,
once the sensors are deployed in the system they
are xed and mobility is restrained. We calculate
the Euclidian distance between the sensors within
a cluster using the concept of Euclidian distance.
Again we have also considered that the system is
contention and error free, Hence the sensors need not
retransmit data. To compute energy consumption
for each transmission sending and receiving, we used
radio energy dissipation model. The energy used
to transmit q bits of data at a distance d for each
sensor node is
E
Tx
(q, d) = q
Elec
+ qf
s
d
2
(17)
The energy used to receive data for each node is given
by
E
Rx
(q) = qE
Elec
(18)
Where E
Elec
is electronic energy, f
s
is power loss
of free space. For our experiment we have taken
50nj/bits for E
Elec
and 10nj/bits/m
2
for f
s
. In
these experiments, each node begins with an initial
energy rages from 0.4 joule to 0.9 joule and unlimited
amount of data can be sent to the base station via
cluster head. In PSO implementation c1 and c2 are
constants and in our experiments the value of c1 and
c2 is set as 2. This is standard values provided in
8
PSO based implementation as suggested by dierent
literatures [21,22,23]. In order to set the value of w,
it is very critical issue to determine the value of w in
a particular PSO application. If the value of w is set
very small, the convergence rate of PSO will be very
slow. On the Other hand is w value is chosen very
large, the problem may not converge. In that case
it may cross the actual solution point and still the
iteration does not stop. In literatures[22,23] it has
been suggested that the value of w need to be set
between 0.4 to 0.9 for better convergence of PSO.
But the exact value of w need to be determined
experimentally and that depends on specic problem
domain. In our case it has been found experimentally
that w should be set as 0.5 to get best convergence
rate of the problem. In step 3. We are generating
all the possible paths from a source node to the
destination. Each of such paths constitutes a particle
and using PSO (by calculating tness function) with
number of iterations the best particle (or path) will
be selected. From gure 1 it can be observed that
the number of clusters has an important inuence
on the performance of the sensor network. We
have carried out our experiment varying d i.e. the
distance between the cluster head and the base
station from 80, 90 and 100. It is observed that
the energy consumption of the network in minimum
when number of clusters is 5. On the basis of that
we have kept the number of clusters as 5 and the
distance between the sensor nodes and the base
station between 80 to 90 meters. We have carried
out experiments in order to nd out suitable values
for the constants c1, c2 and w in the implementation
of the PSO algorithm. In Figure 2, we it can be
observed that the most suitable value for c1 in our
case is 2.0 that can provide the tness value of
12. Moreover, this graph demonstrates the optimal
value for c1 is 2.0 in our PSO implementation. From
Figure 3, we have determined the optimal value for
c2 is 2.0 in a similar fashion. Hence, we have xed
the values for both c1 and c2 to 2.0 in our proposed
PSO implementation. Then we have tried to nd
a suitable value for w in our PSO implementation.
From Figure 4 it is clear that the optimal value for w
in our case is .5 and the value of the tness function
in that case is 12. Therefore, we have xed the value
of was 0.5 in our experiment. Figure 5 explains how
the values of the tness function vary with number
of generations. Initially it started with a value of
12 and with successive generations the tness value
decreases and nally an optimal solution is obtained.
Hence, our algorithm is capable of nding out the
ttest particle and hence the most promising path
for data transmission from a source to the base
station after number of generations. Selection of
such an optimal path is guided by the appropriate
tness function at each step. Figure 6 demonstrates
the performance of the proposed algorithm in terms
of number of alive nodes with simulation time. From
the result it can be observed that the proposed algo-
rithm outperforms the existing Leach and Leach-C
algorithm in terms of number of alive nodes with
simulation time. This is because in the proposed
algorithm the optimal path between a sensor node
and the base station is based on the appropriate
tness function. But, as the initial set up cost in the
case of PSO is relatively high, the ICWAQ algorithm
[34] works better than the proposed algorithm for
rst 400 rounds. The tness function in ICWAQ
function is chosen in such a way that it results in fast
and ecient searching of roots. But, in the proposed
algorithm, once the initial solution is obtained by
the PSO, it can converge to the solution at a faster
rate. Thus in our algorithm using ecient tness
function, the most promising path gets selected after
successive iterations. Hence the overall network
lifetime is increased in our algorithm with increase
simulation time. Further, an experiment has been
carried out to study the total number of cluster
head selection rounds supported by the alive nodes
in the network. Figure 7 shows the result of the
experiment. Here the performance of the proposed
algorithm is compared with Leach and another
version of existing PSO based cluster head selection
procedure[33]. From the result it is observed that
the lifetime of the network degrades at a faster
rate in case of Leach algorithm, as the cluster head
rotation is performed in each round. In PSO-SD the
gradual decrement in the plot is reported. This is
because the cluster members remain busy most of
the times to forward the data to the base station via
the cluster head. In the proposed algorithm as the
9
cluster head is rotated on demand and depending
on the residual energy of the nodes present in the
cluster, the lifetime of the network is also increased.
Fig 1.Energy consumption rate of CH with varying distance
Fig 2. Suitable value of c1
Fig 3. Suitable value of c2
Fig 4. Suitable value of w
Fig 5. Fitness value of the objective function with number of generations
Fig 6. Number of alive nodes with number of rounds
Fig 7. Number of alive nodes with cluster head selection rounds
10
6 Conclusion
Despite the disparity in the objectives of sensor
applications, the main task of wireless sensor nodes
is to sense and collect data from a target domain,
process the data and transmit the information
back to specic sites where the application resides.
Achieving this task eciently requires the develop-
ment of an energy-ecient routing protocol to set
up paths between sensor nodes and the data sink.
The characteristics of the environment within which
sensor nodes typically operate , coupled with severe
resource and energy limitation, make the routing
problem very challenging. In order to increase the
lifetime of such network, the computations at the
sensor nodes need to be minimised. In PSO, the
amount of computation need to be performed is
very less in compare to other existing evolutionary
techniques. Hence a routing algorithm based on
PSO is a good choice to solve the routing problem
in WSN. In this paper, we have presented clustered
heterogeneous wireless sensor networks where more
powerful sensor nodes act as cluster heads for more
number of rounds. Based upon the simulation
results, the proposed protocol has conrmed that it
provides a longer network lifetime as compared to
LEACH , PSO-SD and ICWAQ. It is observer that
the lifetime of the network degrades at a faster rate
in case of LEACH algorithm ,as the cluster head
rotation is performed in each round. In PSO-SD
the cluster member remain busy most of the time to
forward the data to the Base station via cluster head
and as a result the lifetime of sensor nodes degrades
gradually. ICWAQ algorithm works better than the
proposed algorithm initially as the initial set up
time with PSO based approach is relatively high.
The proposed algorithm selects the most promising
path after successive iteration which exhibits a long
network lifetime. In the proposed algorithm as the
cluster head is rotated on demand and as the nodes
are heterogeneous with respect to initial energy, the
lifetime of the network is also increased. Our future
work includes with the implementation of the mobile
sensor nodes as well as comparing the performance of
the algorithm against other most ecient optimizing
routing algorithm with fault tolerant mechanism.
References
[1] I.F.Akyildiz ,W. Su, Y. Sankara subramniam,
and E. Cayirei ,A survey on sensor networks,
IEEE Commun Mag. vol.40, no.8, pp.102-
114,Aug.2002.
[2] C-Y. Chong, S.P. Kumar.,Sensor Net-
works: evolution,opportunities and chal-
lenges,Proceeding of the IEEE 91(8) (2003)
1247-1256.
[3] H. Wang et al.,Target classication and lo-
calization in habitat monitoring,Proceedings of
IEEE International Conference on Acoustics,
Speech and Signal Processing Hong Kong china
Vol. 4, pp. 844-847,April 2003.
[4] A.B. McDonald, T. Znati, A mobility based
framework for adaptive clustering in wireless
adhoc networks,IEEE Journal on Selected Ar-
eas in Communications Vol. 17,No. 8 ,pp. 1466-
1487,1999.
[5] S. Bandyopadhyay , E. Coyle,An energy e-
cient hierarchical clustering algorithm for wire-
less sensor networks, Proceedings of the 22nd
Annual Joint Conference of the IEEE Computer
and Communications Societies, San Francisco,
California Vol. 3, pp. 1713-1723,2003.
[6] Li Qing, Qingxin Zhu, Mingwen Wang, De-
sign of a distributed energy-ecient clustering
algorithm for heterogeneous wireless sensor net-
works, Computer Communications, Vol. 29, pp.
2230-2237, 2006.
[7] Yang E, Erdogan AT, Arslan T, Barton N,An
Improved Particle Swarm Optimization Al-
gorithm for Power-Ecient Wireless Sensor
Networks,ECSIS Symposium on Bio-inspired,
Learning, and Intelligent Systems for Security
(BLISS) pp.76-82,2007.
11
[8] Rui Wu ,Kewen XIA,yanjun zha,Guodong
LI,Optimal Design on clustering Routing pro-
tocol for Wireless Sensor Network.,Journal
of Computational Information , Vol.
9,No.14(2013),pp 5521-5528.
[9] Kulkarni RV, Venayagamoorthy GKarticle
swarm optimization in wireless-sensor networks:
a brief survey,IEEE Transactions on Systems,
Man, and Cybernatics-Part C. Appl Rev, Vol.
41o.2,pp.262267,2011.
[10] Hu B, Liu S, Li H,Mobile Beacon Node Path
Scheme Based on Particle Swarm Optimiza-
tion in Wireless Sensor Networks, ICEMI
Vol.3,pp.306-309.
[11] K. S. Low, H.A. Nguyen and H. Guo, A Particle
Swarm Optimization Approach for the localiza-
tion of a WSN, IEEE International Symposium
on Industrial Electronics, pp. 1820-1825,2008.
[12] Valeria Loscri,Enrico Natalizio and Francesca
Guerriero , Particle Swarm Optimization
Schemes based on Consensus for wireless sensor
network ,MSWiM,pp. 21-25, Oct 2012.
[13] Sudarmani R. And K.R. Shankar kumar ,Par-
ticle swarm optimization based routing proto-
col for clustered Heterogeneous sensor networks
with mobile sink.,American Journal of Applied
Sciences Vol.10, No. 3,pp. 259-269,2013.
[14] Suchal Sarangi and Biju Thankchan,Particle
swarm optimization based A novel rout-
ing Algorithm for Wireless Sensor Network,
IJREISS,vol.2 Isssue ii,pp 2177-286,Nov.2012.
[15] X. Cao, H. Zhang, J. Shi and G. Cui, Clus-
ter Heads Election Analysis for Multihop Wire-
less Sensor Networks based on Weighted graph
and Particle Swarm Optimization, International
Conference on Natural Computation,pp 599-
603,ieeexplore18-20 oct2008.
[16] Samaneh Hojjatoleslami, Vehe Aghazarian,
Mehdi Dehghan and Nima Ghazanfari Mo-
tiangh, PSO Based Node Placement Optimiza-
tion for Wireless Sensor Networks, IEEE, 2011.
[17] W.Z.W. Ismail and S.A.Manaf, Study on Cov-
erage in WSN using Grid Based Strategy and
Particle Swarm Optimization,IEEE Asia Pa-
cic Conference on Circuits and Systems,pp.
1175-1178 ,2010.
[18] Zhaoxia Wu,Yonjin Nie,Shuqiang Chen,Huifang
Zhang and Lifu Wang ,Double layers cluster-
ing algorithm based on CPSO for wireless sen-
sor networks,Information Technology journal,
Vol.11,No. 12, pp 1737-1743,2012.
[19] Hamid Ali, Waseem Shahzad and Farrukh
Aslam Khan, Using Multi-Objective Particle
Swarm Optimization for Energy-Ecient Clus-
tering in Wireless Sensor Networks,Wireless
Sensor Networks and Energy Eciency: pro-
tocols,Routing and management. Book Chapter
,pp. 291-304.
[20] N. M. Abdul Lati, C. C. Tsimenidis, B. S.
Sharif, C. Ladha, Dynamic Clustering using Bi-
nary Multi-Objective Particle Swarm Optimiza-
tion for Wireless Sensor Networks,IEEE 19th In-
ternational symposium on PIMRC,pp.1-5,2008.
[21] Jamshid Shanbehzadeh, Saced Mehrjoo and Ab-
dol hossein SArrafzadeh, An Intelligent Energy
Ecient Clustering in Wireless Sensor Networks,
Proceedings of the International Multiconfer-
ence of Engineers and Computer Scientists 2001,
Hong kong, vol. 1, March 16-18, 2011.
[22] I. Bhattacharya and S. Samanta, Parameter
selection and Performance Study in Particle
Swarm Optimization, AIP Conference Proceed-
ings Vol. 1298,No.564(2010).
[23] Yuhui Shi and Russel C. Eberhart, Parame-
ter Selection in Particle Swarm Optimization,
Lecture notes in Computer Science, Volume
1447,pp. 591-600,1998.
[24] W. B Heinzelman, A. P Chandrakasan, and
H. Balakrishnan, An application-specic pro-
tocol architecture for wireless microsensor net-
works,IEEE Transactions on Wireless Commu-
nications, vol. 1, no. 4, pp. 660-670, October
2002.
12
[25] S.Dasgupta, P.Dutta, An Improved Leach ap-
proach for Head selection Strategy in a Fuzzy-
C Means induced Clustering of a Wireless Sen-
sor Network,IEMCON 2011 organized by IEM,
Kolkata, in collaboration with IEEE on 5th 6th
of Jan 2011, pp. 203-208,2011.
[26] Indranil Gupta, Riordan, D .Srinivas Sampalli,
Cluster head election using Fuzzy logic for
Wireless Sensor Network,Communication Net-
works and Service Research Conference publica-
tions, pp. 255-260,May 2005.
[27] S. Bandyopadhyay and E. J. Coyle, An Energy
Ecient Hierarchical Clustering Algorithm for
Wireless Sensor Networks,IEEE INFOCOM ,
Vol. 3, pp. 1713-1723 April 2003.
[28] D. J. Baker and A. Epheremides, The Archi-
tectural Organization of a Moblie Radio Net-
work via a Distributed Algorithm, IEEE Trans-
actions on Communications, vol. Com-29, no.
11, November 1981.
[29] P. Tsigas, Project on Moblie Ad Hoc Network-
ing and Clustering for the Course EDA390 Com-
puter Communcation and Distributed Systems,
Manual for University Course.
[30] A. Amis, R. Prakash, T. Vuong, and D. Huynh,
Max-Min D-Cluster Formation in Wireless Ad
Hoc Networks, IEEE INFOCOM, March 2000.
[31] D.E. Goldberg, 1989. Genetic Algorithms in
Search, optimization and Machine Learning,
Addison Wesley, New York, .
[32] Lati N.M.A, Tsemenidis CC, Sheri
BS.Energy-Aware Clustering for Wireless
Sensor Networks using Particle Swarm Opti-
mization The 18th Annual IEEE International
Symposium on Personal, Indoor and Mobile Ra-
dio Communications (PIMRC07), pp.15,2007.
[33] Buddha Singh and Daya Krishan Lobiyal,A
novel energy-aware cluster head selection based
on particle swarm optimization for wireless sen-
sor networks ,in Human centric computing and
Information Sciences, a springer open acess
journal,pp 1-18,2012.
[34] Dervis Karaboga, Selcuk Okdem,Celal OZturk,
Cluster based wireless sensor network routing
using articial bee colony algorithm ,Wireless
Network,Vol. , pp. 847-860.
[35] Ngatchou, P.; Zarei, A.; El-Sharkawi, A.;
Pareto Multi Objective Optimization, Intel-
ligent Systems Application to Power Systems,
Proceedings of the 13th International Confer-
ence, Page(s):84 - 91, 6-10 Nov. 2005.
[36] Berizzi, A., M. Innorta, and P. Marannino.
Multiobjective optimization techniques applied
to modern power systems,In 2001 IEEE Power
Engineering Society Winter Meeting, Jan 28-Feb
1 2001. 2001.
[37] C. A. Coello and M. S. Lechuga, MOPSO: A
proposal for multiple objective particle swarm
optimization, In IEEE Proceedings World
Congress on Computational Intelligence,pages
10511056, 2003.
13

You might also like