Extending Lifetime of Wireless Sensor Networks by Distributing the
Overload on Cluster Heads
Mohammed Kayed
1
, Ahmed Anter
1
and Ahmed Hassan
2
1
Faculty of Computers and Information, Beni-Suef University, Beni-suef, Egypt, 62511
2
Faculty of Science, Beni-Suef University, Beni-suef, Egypt, 62511
Keywords: Wireless Sensor Networks, Cluster Head, Wireless Power Transfer, Particle Swarm Optimization
Abstract: Clustering sensor nodes in power-constrained Wireless Sensor Network (WSN) is an efficient step to enhance
energy efficiency and extend the network lifetime. Clustering gives many advantages (e.g., data aggregation
and less number of transmissions) that greatly reduces the energy consumption of the WSN. A Cluster Head
(CH) node is selected for each cluster to receive data from the cluster’s nodes, aggregate them, and finally
transmit these data to a Base Station (BS). However, the overhead on the CH nodes is still a problem for the
network lifetime, which causes premature death for those overloaded nodes. Energy harvesting is one of the
most energy optimization techniques that make the WSN rechargeable and so extends the lifetime of the
network. Traditional techniques such as solar and wind harvesting are not reliable because they are neither
constant nor always available. Another type of energy harvesting is the Wireless Power Transfer (WPT)
approach in which a node enables to transfer energy to other nodes. According to the energy consumption
theory in WSN, about 90% of energy is left unused after the premature death of overloaded nodes. If this big
surplus of energy is used to recharge the overloaded nodes, it will greatly extend the lifetime of the WSN. So,
in this paper, we use the WPT technology and the multi-objective Particle Swarm Optimization (PSO) to
extend the network lifetime. The target of our proposed approach is to minimize the amount of this surplus of
energy. All nodes in the WSN could transfer energy to the CH nodes to distribute the overall load. At the
same time, the optimal amount of energy, transferred by each node, must also be convenient to its residual
energy. Therefore, this paper tries to eliminate the overhead on the CH nodes and therefore extend the lifetime
of the clustered WSN. Our simulation results show an encouraging result to extend the lifetime of the WSNs
as compared with the common Leach algorithm.
1 INTRODUCTION
WSN is one of the most effective factors that have
achieved recent smarting and technological progress
in many vital fields such as healthcare, military, and
smart cities (Yang et al., 2015; Wu et al., 2018). WSN
consists of a set of small, cheap, and smart devices
called sensor nodes. Each sensor node is responsible
for three main tasks: sensing an interesting area,
processing the sensed data, and finally transmitting
the interesting events to the BS in order to take the
convenient actions remotely. Each sensor node is
powered by a limited capacity battery, which makes
the WSN power-constrained and with a limited
lifetime. This will, in turn, has a negative effect on the
availability and the performance of this efficient
technology. Many energy optimization techniques
have been proposed to extend the lifetime of the
WSN. One of the most energy-efficient approaches is
a clustering step in which sensor nodes are grouped
into clusters, where each cluster has a cluster head
(CH) node (Arghavani et al., 2017). In each data
transmission round, a sensor node transmits the
sensed data to its CH node, which aggregates the
sensed data of all its cluster members in one data
packets (data aggregation) and finally transmits the
aggregated data to a BS. A clustering approach has
achieved many features that not only greatly extend
the lifetime of the WSN, such as data aggregation and
less number of transmission which greatly reduce the
communication overhead, but also improve the
performance and throughput of the WSN. This
improvement will be done by ensuring the
connectivity between the nodes and the BS,
facilitating the security for the whole WSN by
securing only the CH nodes rather than all nodes in
Kayed, M., Anter, A. and Hassan, A.
Extending Lifetime of Wireless Sensor Networks by Distributing the Overload on Cluster Heads.
DOI: 10.5220/0009424201370143
In Proceedings of the 1st International Conference on Industrial Technology (ICONIT 2019), pages 137-143
ISBN: 978-989-758-434-3
Copyright
c
2020 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
137
the network (the CH nodes are considered as
gateways for the network) and finally provide
scalability for the WSN (Afsar et al., 2014). However,
all clustering approaches have a common dangerous
problem which is the huge overhead on the cluster
head (CH) nodes (Kuila et al., 2013) as CH nodes are
responsible for receiving data from all their cluster
members, aggregating them and finally transmitting
the aggregated data to the base station (BS).
Therefore, the CH nodes are exposed to huge energy
consumption, which causes premature death for the
nodes and harms the lifetime of the WSN.
Energy harvesting is one of the hot approaches to
recharge the batteries of the sensor nodes from other
energy sources. Traditional energy harvesting
approaches convert the energy from natural sources,
such as solar or wind harvesting, into electric energy
to replenish the energy of the sensor nodes. However,
traditional energy harvesting is not reliable because it
depends on the current environmental conditions that
are not constant or always available. Another energy
harvesting approach is WPT, by which the nodes can
exchange energy with each other to recharge their
batteries. Based on the energy consumption theory, at
the same time, when the first node depletes its energy,
90% of the initial energy in the network still exists
(Yu et al., 2018). So in this paper, we propose to use
this big surplus of energy as a source to replenish the
battery of the overloaded nodes, using WPT, and so
extend the lifetime of the WSN. Practically, each
node intends to keep its residual energy large as much
as possible to be able to perform its assigned tasks,
but at the same time, it should transfer an amount of
energy to the overloaded CH nodes, to increase the
lifetime of the WSN, which is considered as a multi-
objective optimization problem. Therefore, we use a
particle swarm optimization algorithm to determine
the optimal amount of energy to be transferred from
each node to reduce the overhead on the CH nodes.
At the same time, the amount of energy transferred
does not harm the lifetime of the source nodes.
Software-defined network (open flow) is a recent
approach in networking where the control is
decentralized from the data plane and embedded in a
device called controller (Kobo et al., 2017). Open
flow is used here to choose the optimal route for the
power transfer.
The main contributions of this paper could be
summarized as follows:
- We eliminate the centralized overhead on the
CH nodes, which exists in the traditional
clustering algorithm, by distributing the load
over all nodes in the network. The target of our
proposed approach is to minimize the number
of energy surpluses at the time of premature
network death (i.e., extend the lifetime of the
network).
- To the best of our knowledge, no prior work
tries to use the WPT technology and the multi-
objectives particle swarm optimization to
extend the network lifetime.
- At the same time, we conserve the topology of
the clustered sensor nodes which achieve the
pre-discussed energy-efficient features.
- Our approach does not require any external
energy source, which is costly and not suitable
for all WSN applications such as healthcare to
recharge the nodes.
The rest of this paper is organized as follows.
Section 2 covers the related works. The system model
is described in Section 3. We explain the proposed
algorithm in Section 4. The performance and
evaluation of the proposed algorithm are analyzed in
section 5. Finally, Section 6 concludes our work.
2 RELATED WORKS
Many clustering algorithms have been proposed to
extend the lifetime of the WSN. Leach algorithm is
one of the most common clustering algorithms in
which the CH nodes are selected randomly
(Heinzelman et al., 2000). Whereas each node firstly
generates a random value between 0 and 1, and then
if the generated value is less than a predefined value,
this node is selected as CH node. The other nodes
assign to the nearest CH node. Leach has some
drawbacks because the nodes with less residual
energy may be selected as CH nodes, and this causes
premature death for them. Also, the CH node
communicates directly with the BS, and this causes a
huge communication overhead. The unequal clusters
that often formed by the traditional clustering
algorithms make the load of the CH nodes
unbalanced. The smaller the number of nodes in a
cluster, the less load of the CH for this cluster.
According to (Zhao et al., 2018), each node
initially generates a cluster that includes itself. After
that, the generated cluster iteratively merges with the
nearest neighboring cluster until a predefined number
of clusters is reached. The distance between two
clusters can be calculated by the longest distance
between any two nodes or the shortest distance
between any two nodes in the two clusters. The CH
nodes must have higher residual energy, closer to the
center of the cluster, and closer to the BS. The
algorithm differentiates between large and small
cluster sizes to balance the load between all CH
ICONIT 2019 - International Conference on Industrial Technology
138
nodes. It considers the cluster whose energy
consumption more than 1.5 of the average energy
consumption for all clusters as a large cluster.
Therefore, the cluster head role in large clusters is
performed by two CH nodes: secondary CH node,
which is responsible for receiving the sensed data
from all cluster members, aggregating them and
transmitting them to the primary CH node, which is
responsible for transmitting the aggregated data to the
BS. So, the load is reduced and balanced between the
CH nodes in either large or small clusters. However,
the algorithm requires computation and time
overhead in the merge process to change the number
of clusters from the number of nodes (initially) to the
predefined number of clusters. Also, using two CH
nodes in many clusters increase the number of
overloaded nodes, and this harms the lifetime of the
WSN. Generally, reducing the load on the CH nodes
by rotating the CH role or balancing the load between
CH nodes is efficient but not optimal solution because
in the usual case the CH node is overloaded because
of the nature of its role and this cause energy
consumption for the CH nodes and this cause
premature death for them (Yadav et al., 2017).
3 SYSTEM MODEL
We consider a network of sensor nodes that are
distributed randomly in an interesting area and a BS,
which is located in the middle of this area. Each
sensor node is powered by a rechargeable battery.
Also, let each sensor node supports the open flow
protocol, and the controller is embedded in the BS.
Each node is assumed to have a unique ID value, and
it supports the wireless power transfer capability. In
order to achieve the most efficient power transfer, we
use an approach called “strong resonant coupling” to
transfer power between the source and the destination
nodes. Strong resonant coupling is considered as one
of the best WPT approaches regarding the
transmission range and the transmission efficiency
(Menon et al., 2013). We use the same radio model
for energy as in (Heinzelman and W.B, 2000). The
energy consumption for transmitting bits a distance
is calculated using equation 1.
,
ε
for
ε
for
(1)
Where
is the consumed energy for
transmitting bits for distance ,
is the energy
required by the electronics circuit, ε
, and ε
are
the energy required by the amplifier in the free space
and the multipath, respectively. The energy
consumed for receiving bits can be calculated by
equation 2.
(2)
4 THE PROPOSED APPROACH
The proposed approach to extending the lifetime of a
WSN has two main steps: clustering and
optimization. Initially, in the first step, clusters are
formed according to the Leach algorithm
(Heinzelman et al., 2000). Each node will generate a
random number between 0 and 1. If the generated
value is less than a predefined threshold, the node is
considered as a CH. After that, each node transmits
its status (residual energy, location, and whether it is
a CH or not) to the controller. Therefore, the
controller has a global view of the whole network
(network map), and so it can implement the proposed
optimization algorithm. Our proposed optimization
algorithm, which shall be discussed in the next
subsection, determines the optimal amount of energy
that will be transferred from each node to compensate
for the overloaded CH nodes. The controller then
submits to each node the ID of its corresponding CH,
the amount of energy which the node can transmit to
the overloaded CH nodes through the network, and
the best routes for the energy transfer.
4.1 PSO Algorithm
PSO is a heuristic-based computational algorithm
inspired by birds flock which searches for food
location. The candidate solution, in PSO, is called a
particle. Each particle has two parameters: location
and velocity. By them, the particle moves through the
search space towards an optimal solution. PSO is an
iterative algorithm. In each iteration, it evaluates the
particle position using a fitness function and records
the optimal local value of the particle and the optimal
global value for all particles. Using the local and
global optimal values, it calculates the updated
velocity, which is used to know a new location of the
particle (new candidate solution). The velocity and
location of the particle are updated using equations 3
and 4, respectively.
1
1
1
1
1
1
1
3
1
4
Where
is the velocity of particle at round
,
is the position of particle at round ,
and
Extending Lifetime of Wireless Sensor Networks by Distributing the Overload on Cluster Heads
139
are two positive constants,
is the local best of
the particle , is the global best of all particles,
and
are two random number between [0,1]. The
PSO algorithm has different advantages: ease of
implementation, efficient solution, and computational
and memory efficiency usage.
4.2 PSO-Based Power Transfer
The global view of the controller to the whole
network allows it to determine the CH for each node
easily and to calculate the load assigned to all nodes
(CH nodes/member nodes). Using the PSO algorithm,
the controller will determine the optimal amount of
energy needed to be transferred to the CH nodes to
compensate for their overload. For distributing the
CH overload over the nodes, there are two possible
approaches. First, the overload of the CH node in a
cluster is locally distributed over the member nodes
of this cluster only. Second, the overall CHs
overloads in the network are globally distributed over
all nodes in the WSN. In most clustering algorithms,
we can observe that CHs close to the BS bear a huge
load than the CHs away from the BS. Therefore
according to the first approach for the clusters near
the BS, member nodes in such overloaded clusters are
likely forced to transfer a big amount of energy even
if their residual energy is not sufficient, which is not
efficient for their lifetime. Although, at the same time,
there may be other nodes (almost have higher energy)
that transfer a small amount of energy because of the
less overload on their CHs. So, the second approach
is more efficient because it balances this overload on
all nodes in the network. Therefore, the dimension of
the particle/solution here is equal to the number of
nodes in the network. Each value corresponds to the
possible amount of energy that could be transferred
from each node in the network to eliminate the overall
CHs Loads.
4.2.1 PSO Fitness function
Our optimization problem has two basic
objective/fitness functions
and
that aimed to
characterize the optimal amount of energy. The first
function allows each node to transfer an amount of
energy, which is suitable to its residual energy (i.e.,
the more residual energy of a node the larger amount
of energy to be transferred). The function
can be
calculated as in equation 5.
∑
∑
∑
(5)
Where is the number of nodes,
is the residual
energy of the node ,
is the amount of the
transferred energy from the node . When the value of
is small, this means that the ratio of the energy of
the node to the energy of all other nodes (the first
modulus term) is approximately equal to the ratio
between the amount of transferred energy from the
node to transferred energy from all other nodes (the
second modulus term). This means the node will
transfer an amount of energy which is suitable to its
residual energy. In other words, when residual energy
is high as compared to the energy of other nodes, the
node is supposed to transfer a larger amount of energy
than the transferred energy by these other nodes and
vice versa.
The second objective function considers other
loads for a node, such as data transmissions and
sensing. For example, there may be a case in which
two nodes have an equal or close amount of energy,
but the load on one of them is large as compared to
the second node. So, it will not be fair to transfer the
same amount of energy from the two nodes. This
function tries to maximize the minimal residual
energy over the whole nodes after the process of
energy transfer. The function
is calculated using
equation 6.
min
–
(6)
Where
is the residual energy of the node,
is the amount of the transferred energy from the node,
and
is the load assigned to the node .
Therefore, the algorithm considers both the energy of
the node and the load of the other tasks assigned to
the node in order to extend the lifetime of the source
node. We can observe that the two objective functions
are in conflict. Our optimization problem has a
constraint that the sum of the transferred energy from
all nodes must be equal to the overall load on all CH
nodes in the network. That is:
∑
∑
_
Where _ is a load of a CH node, and is the
number of CH nodes in the network. Most bio-
inspired optimization algorithms are applied for
unconstrained-optimization problems. So, we will
use an efficient and common constraint handling
technique, which is a penalty function. Penalty
function converts the constrained optimization
problem into an unconstrained optimization problem
which can be solved by a bio-inspired optimization
algorithm easily. This is done by adding a term, called
"quadratic loss function," to the objective function
and converting the constraint to an objective in the
objective function. So the adjusted objective function
becomes:
ICONIT 2019 - International Conference on Industrial Technology
140
1
1
_
Quadratic loss function becomes squared to make
the constraint more severe to be applied. is a
constant, and its value is ranged from 10 to 100, and
is a weight value.
Therefore, the overload of the CH nodes in the
network is completely distributed over all nodes. So,
this huge load is divided into smaller loads (because
of the big number of nodes that bear this load
together) and be assigned to the whole nodes in the
network based on the node energy and the other task
loads.
4.2.2 Energy Transfer and Delay Reduction
After the algorithm determines the optimal amount of
energy that needs to be transferred from each node,
the controller submits an MSG to each node, which
contains its CH id, the amount of energy required, and
the optimal route for the energy transferring. There
are two main approaches in power transfer: a single-
hop approach in which the energy is transmitted
directly from the source to the destination, where
there are no intermediate nodes, and a multi-hops
approach in which the energy is transmitted hop by
hop (passing through multiple nodes) from the source
to the destination. Generally, the energy transfer
process is affected by the distance between the nodes
(Han et al., 2018). The longer distance between the
source and destination, the less energy transfer
efficiency is achieved. Also, there exists a limited
distance value, after which the receiver cannot
receive any amount of energy (exceeds the power
transmission range). We can observe that the multi-
hops approach is more efficient than the single-hop
because it can extend the transmission range where a
source can transmit energy to another node, which is
not in its transmission range. Also, the multi-hops
approach improves the transfer efficiency because the
distance between the source and any intermediate
node is less than the distance between the source and
the destination. The energy is transferred with the
shortest path, determined by the global view feature
of the controller, in the multi-hops approach.
Numerous studies have been conducted to
demonstrate that using a resonant repeater (inside
each sensor node) to achieve energy transmission is
the most effective method (Han et al., 2018). Also,
the repeater improves the strength of the wave and
compensates for any loss during transmission. So, the
repeater with the shortest path generated by the global
view is used to guarantee the best energy transmission
efficiency. Each CH harvests an amount of energy
equals the load of its CH role.
Our proposed algorithm uses the “simultaneous
energy transmissions” mechanism to reduce the delay
in the case where the transferring node is far from the
target CH node (not in its cluster). Whereas, if the
CH, after harvesting energy from its cluster members,
still needs an extra amount of energy to eliminate its
assigned CH load completely, the controller chooses
the nearest set of nodes whose residual energy allows
to transfer the extra amount of energy to the CH node.
In this way, this set of nodes will transfer an amount
of energy larger than its optimal amount, which is
determined by the algorithm. The controller then
chooses another set of nodes that is nearest to the first
set to compensate for the extra load of the first set
(second level). Iteratively, this power transfer process
is repeated level by level simultaneously until
reaching the nodes that are supposed to transfer
energy to the far CH node from them. So, the far node
logically transfers energy for the CH node at the same
time the closer node to the CH node, and so delay is
reduced.
5 PERFORMANCE
EVALUATION
We simulate our proposed algorithm under Matlab.
The performance is compared with the Leach
algorithm in terms of network lifetime and the
elimination of the CH load. We apply our algorithm
in a network consists of 100 nodes and distributed
randomly in 100 100 meters, as shown in Figure 1.
Figure 1: A simulated network distribution.
The BS is located at the center of the network (50,
50) to achieve the best performance. The simulation
parameters and values are listed in Table 1.
Extending Lifetime of Wireless Sensor Networks by Distributing the Overload on Cluster Heads
141
Table 1: This caption has one line, so it is centered.
Parameter Value
Network size
100 100
Nodes number 100
BS position (50,50)
Data packet size (bits) 4000
Control packet size (bits) 100
Initial node energy (J) 0.1
Transmission range (m) 20
C1
1
.5
C2 2.0
W 1
C 10-100
Generally, the optimal solution of the PSO is
based on the number of particles and the number of
iterations. So, we ran the algorithm many times with
different values for the two parameters and found that
the best value has occurred when the number of
particles equals 200, and the number of iterations
equals 2000, as shown in Figure 2.
Compared to the Leach algorithm, our algorithm
shows an efficient extension for the lifetime of the
wireless sensor networks. The first node in Leach is
died at around 284, but in our algorithm, the first node
is died at around 350. Also, in Leach, the last node
has died at around 315, but in our proposed algorithm,
the last node died at around 400, as shown in Figure
3. This extension of life occurs because the huge
overhead on the cluster head nodes is efficiently
eliminated and distributed over the whole nodes. By
dividing this overhead over a big number of nodes,
there are no overloaded nodes, and also the average
energy consumption of the CH nodes is
approximately equal to the average energy
consumption for ordinary member nodes in around as
shown in Figure 4. This will enhance the energy
efficiency of the clustering algorithms, whereas the
centralized overhead on the CH nodes is distributed
over the whole nodes.
Figure 2: The best cost based on a different number of
particles and iterations.
Figure 3: Lifetime extension for the proposed algorithm is
compared to Leach algorithm
Figure 4: Elimination of CH nodes overhead in the
proposed algorithm is compared to the Leach algorithm.
6 CONCLUSIONS
In this paper, we eliminate the centralized overhead
on the CH nodes by integrating the WPT and PSO
algorithm, where each node can transfer an amount of
energy, determined by the PSO model, to the
overloaded CH nodes. Therefore, this huge load is
divided into small values and distributed over the
whole nodes based on their residual energy and load.
Although our algorithm eliminates the centralized CH
load, it keeps on the clustered topology, which
provides energy-efficient features for extending the
lifetime of the WSN. We plan to apply other
optimization algorithms and compare them with the
PSO algorithm on other simulated networks.
ICONIT 2019 - International Conference on Industrial Technology
142
REFERENCES
Arghavani, M., Esmaeili, M., Esmaeili, M., Mohseni, F., &
Arghavani, A. (2017). Optimal energy aware clustering
in circular wireless sensor networks. Ad Hoc
Networks, 65, 91-98.
Afsar, M. M., & Tayarani-N, M. H. (2014). Clustering in
sensor networks: A literature survey. Journal of
Network and Computer Applications, 46, 198-226.
Kuila, P., Gupta, S. K., & Jana, P. K. (2013). A novel
evolutionary approach for load balanced clustering
problem for wireless sensor networks. Swarm and
Evolutionary Computation, 12, 48-56.
Yu, S., Liu, X., Liu, A., Xiong, N., Cai, Z., & Wang, T.
(2018). An adaption broadcast radius-based code
dissemination scheme for low energy wireless sensor
networks. Sensors, 18(5), 1509.
Yang, J., Zhou, J., Lv, Z., Wei, W., & Song, H. (2015). A
real-time monitoring system of industry carbon
monoxide based on wireless sensor
networks. Sensors, 15(11), 29535-29546.
Wu, F., Redouté, J. M., & Yuce, M. R. (2018, October). A
Self-Powered Wearable Body Sensor Network System
for Safety Applications. In 2018 IEEE SENSORS (pp.
1-4). IEEE.
Heinzelman, W. R., Chandrakasan, A., & Balakrishnan, H.
(2000, January). Energy-efficient communication
protocol for wireless microsensor networks.
In Proceedings of the 33rd annual Hawaii international
conference on system sciences(pp. 10-pp). IEEE.
Zhao, Z., Xu, K., Hui, G., & Hu, L. (2018). An Energy-
Efficient Clustering Routing Protocol for Wireless
Sensor Networks Based on AGNES with Balanced
Energy Consumption Optimization. Sensors, 18(11),
3938.
Yadav, R. K., Gupta, D., & Lobiyal, D. K. (2017). Energy
Efficient Probabilistic Clustering Technique for Data
Aggregation in Wireless Sensor Network. Wireless
Personal Communications, 96(3), 4099-4113.
Han, G., Guan, H., Wu, J., Chan, S., Shu, L., & Zhang, W.
(2018). An uneven cluster-based mobile charging
algorithm for wireless rechargeable sensor
networks. IEEE Systems Journal.
Heinzelman, W. B. (2000). Application-specific protocol
architectures for wireless networks (Doctoral
dissertation, Massachusetts Institute of Technology).
Menon, K. U., Vikas, V., & Hariharan, B. (2013, July).
Wireless power transfer to underground sensors using
resonant magnetic induction. In 2013 Tenth
International Conference on Wireless and Optical
Communications Networks (WOCN) (pp. 1-5). IEEE.
Kobo, H. I., Abu-Mahfouz, A. M., & Hancke, G. P. (2017).
A survey on software-defined wireless sensor
networks: Challenges and design requirements. IEEE
access, 5, 1872-1899.
Extending Lifetime of Wireless Sensor Networks by Distributing the Overload on Cluster Heads
143
