Enhancement of Energy Harvesting Efficiency in Mobile Wireless
Sensor Networks
AMIN AL KA’BI
Australian College of Kuwait, KUWAIT
Abstract – Mobile wireless sensor networks suffer from the restricted availability of energy supplies. In this research
work, a proposed method for extending the lifetime of energy-constrained mobile wireless sensor networks (MWSNs)
is presented. This method is based on the fact that RF signal carries both information and energy at the same time.
Hence, by increasing the efficiency of energy harvesting from radio frequency (RF) signals, the lifetime of the wireless
network can be significantly extended. The Simultaneous Wireless Information and Power Transfer (SWIPT) technique
enables harvesting of energy by relay nodes which in turn can be used for wireless data transmission. To enhance the
lifetime of the mobile wireless network, the transmitted RF energy can be recycled at the receiver side. On the other
hand, a balance between energy harvesting and wireless data transmission is required in to maximize the overall
efficiency of the system. Particle Swarm Optimization (PSO) is employed to obtain the optimum resource allocation
policy which maximizes the system energy efficiency. A cost function is framed for this purpose and PSO attains the
maximum energy efficiency by improving the solution of the cost function at each iteration with respect to given
constraints.
Keywords - Simultaneous Wireless Information and Power Transfer (SWIPT); Energy Efficiency; Resource Allocation;
Low Energy Adaptive Clustering Hierarchy (LEACH); Particle Swarm Optimization (PSO)
Received: August 19, 2021. Revised: March 24, 2022. Accepted: April 23, 2022. Published: May 19, 2022.
1. Introduction
The wireless sensors (nodes), which are powered by
cells with limited energy, have restricted the lifetime of a
wireless sensor network. This is an existing basic issue
being faced by sensor networks which are used for long-
haul tasks. Energy conservation techniques can only
reduce the total energy consumption of the system but
cannot compensate for the energy depletion. Deploying
more nodes is undesirable as the deserted nodes may
cause pollution to the surrounding environment.
Replacing the cell or node is only applicable in cases in
which the nodes can be located and physically accessed
by humans or robots [1-4].
Wireless charging technology is a promising solution
for addressing the energy limitations in sensor networks.
The wireless charging technology, along with more cheap
mobile robots, makes the power restoring process
possible and controllable, and hence the power can be
restored to satisfy energy requirements. Close alignment
between the charger and nodes is not required when
compared to the node or cell replacement techniques.
Wireless charging technologies can be classified into
two groups, which are Radio Frequency (RF) based
wireless charging (radiative) and coupling-based wireless
charging (non-radiative). RF waves are used as the
medium for transferring energy in the case of radiative
wireless charging. Here the transfer of energy is on the
basis of the radiative electric field of the RF wave. Non-
radiative wireless charging is commonly utilized in
appliances of daily use due to safety considerations [5-7].
As the RF signal consists of both information and
energy, it is considered as a promising method for
wireless energy transfer where it enables simultaneous
wireless information transfer along with energy
harvesting. To improve the lifetime of the sensor
network, the transmitted RF energy can be recycled at the
receiver side. This technique is referred to as
Simultaneous Wireless Information and Power Transfer
(SWIPT) [8].
In this case, a data transmitting node transfers the
energy together with the data to its cluster head. Based on
Dynamic Power Splitting Scheme, the cluster head
divides the received RF signals into two power streams
with specific power splitting ratios for data forwarding
and energy harvesting, respectively. This method has two
merits: (a) harvesting energy from the RF transmitters,
using the harvested energy in data forwarding, and hence
avoiding the depletion of energy; (b) to improve the
Quality-of-Service (QoS), energy may be harvested from
either interference signals or RF signals of transmitters,
and even antenna noises.
This work focuses on implementing an efficient
resource allocation using particle swarm optimization
with the aim of maximizing energy efficiency.
WSEAS TRANSACTIONS on COMMUNICATIONS
DOI: 10.37394/23204.2022.21.18
Amin Al Kabi
E-ISSN: 2224-2864
129
Volume 21, 2022
2. System Model and Problem
Formulation
2.1 System Model
The system model of the mobile WSN consists of a
mobile collector and N nodes. The sensor network
consists of antennas randomly distributed over the field
under consideration. Periodically, a deployed mobile
collector conducts an information gathering tour
beginning from the sink node. At each tour in the field, it
visits some previously determined anchor nodes, known
as cluster heads for collecting information from the
neighboring sensors through multiple hop transmission
by staying near them for a specific period of time. On the
basis of a clustering protocol known as Low Energy
Adaptive Clustering Hierarchy (LEACH) [9-11], the
sensor nodes are grouped as clusters before starting the
information gathering tour. In this case, each cluster
consists of a cluster head (CH) for collecting the
information sensed by each sensor in its cluster through
relays of other nodes. This collected information is then
uploaded to the mobile collector, as illustrated in Figure
1. The cluster heads also act as anchors for the data
collector. The nodes transmit their sensed information to
the CHs including the energy. The RF energy is also
harvested from the data received by the CH and then the
data is aggregated. A CH consists of a signal processing
unit with a rechargeable cell, an energy harvesting unit,
and a power splitting unit to maintain simultaneous data
forwarding and energy harvesting, as depicted in Figure
2. The circuit in the receiver designed for data forwarding
cannot be used for energy harvesting because of hardware
limitations [1]. Consequently, the energy harvesting unit
and the data processing unit should have separate
circuitries.
Figure 1. SWIPT in a 3 cluster WSN. [1]
At the transmitter side, time-slotted transmission is
employed, and at the receiver side a dynamic power
splitting scheme is employed, which enables the receiver
to process the data and energy harvesting from the
received signal at any instant. The basic principle behind
this technique is illustrated in Fig. 2. The received signal
from transmitter of the node is split dynamically by
the receiver at the CH into two energy streams for
data processing and energy harvesting in ratios
and
respectively.
The transmitting nodes are grouped into clusters such
that the cluster heads lie within the effective coverage
area of the transmitting antennas, then by using an
efficient power management circuit, the received power
is converted to DC using AC/DC converter, and then this
power is transferred to the storage cell to power a sensor.
The energy harvested in the cell helps in lowering the
minimum power transfer requirement , and
hence further limits the power splitting ratio for
harvesting energy, thereby enhancing the rate of data
transmission in the wireless network. The power splitting
unit is assumed as perfect [1], and hence, it will not lead
to any power loss or noise. The power consumed by each
node is fixed as Watts for processing a unit of data
and does not depend on the amount of energy harvested.
And hence, when the data processing rate of a sensor is
R, the total power consumption of the circuit is *R
WSEAS TRANSACTIONS on COMMUNICATIONS
DOI: 10.37394/23204.2022.21.18
Amin Al Kabi
E-ISSN: 2224-2864
130
Volume 21, 2022
Watts. Hence, powering the CHs by more than one
energy source is practically desirable [12].
Figure 2. Model of a receiver with SWIPT [1]
2.2 Communication Model
The communication model of the wireless sensor
network consists of clusters of sensors. Let us assume
that one of these clusters consists of a CH and N−1
sensors, represented as , which
are grouped using LEACH algorithm. The directed graph
of this sensor network is then modelled as X = (M, C); M
= N CH is the group of all nodes, and C corresponds to
the group of all connected links between the sensors and
the CH. The condition for a connected link (i, j) C to
exist is that where indicates the distance
between node and node, denotes the
transmission range of the sensors, which depends on the
transmitted power and gain of the sensor. The channel
between transmitter and the receiver is assumed to be that
of a quasi-static block fading model. The channel gains
are calculated by obtaining the receiver feedback. As
depicted in Figure 2, the corruption of the received signal
occurs due to an Additive White Gaussian Noise
generated from the sensor at the receiver. Then the
received RF signal is then deliverd to a power splitting
unit, at which it is split and then separately fed to the
energy harvesting unit and the information processing
unit.
The capacity of the channel across the transmitter
and receiver can be calculated as
(1)
where W denotes the band-width and denotes the
transmitted power from the transmitter to
receiver, and denotes the channel path loss due to
attenuation, shadowing, and other path losses. The
maximum data rate that can be achieved in the case of
reliable data forwarding from the transmitter to
receiver is always less than channel capacity between
them, i.e.,
(2)
In the case of transfer of energy, according to the rule
of energy conservation, the energy received by the
receiving antenna is always less than the harvested
energy denoted by Joules.
(3)
where represents the coefficient for
harvesting energy from transmitter by receiver
which implies that the entire energy radiated by
transmitter is not harvested by receiver.
shows the efficiency of energy conversion of receiver
in conversion of the received RF signal into electrical
energy for storing in the cell, which is dependent upon
the process of rectification used and the circuit used for
harvesting energy [3]. Maximum values are assumed to
and , i.e., the two sides of the inequalities in (2)
and (3) become equal.
3. Problem Formulation
The resource allocation problem is formulated such
that it maximizes the system energy efficiency (Bit/J).
3.1 End-to-End Data Throughput
The total number of bits conveyed to receivers
successfully per second is known as the end-to-end data
throughput.
(4)
In which { represents the
policy for power allocation, is the policy for power
splitting. To ensure a particular degree of fairness, the
application layer fixes which is a positive weight
accounting for the priorities of different receivers. To
improve the system energy efficiency, the overall system
energy consumption is considered. The weighted energy
consumed by the system needed for reliable
communication is modelled as the total power
dissipation, which is given by
(5)
where, is a constant accounting for the inefficiency
of the transmitter, and represents the data rate.
WSEAS TRANSACTIONS on COMMUNICATIONS
DOI: 10.37394/23204.2022.21.18
Amin Al Kabi
E-ISSN: 2224-2864
131
Volume 21, 2022
3.2 Efficiency of Energy Harvesting
The sum of the weighted number of bits delivered
successfully to the receivers per one Joule of consumed
energy is called as the weighted energy efficiency of the
system and it can be expressed as
(6)
The resource allocation problem (ResAll) is then
formulated into a nonlinear optimization problem:
(7)
which is subjected to
C1:
(8)
C2: (9)
C3:
(10)
C4:
(11)
C5:
(12)
C6:
(13)
C7:
(14)
in which C1 is the minimum power transfer requirement
for power transfer from receiver to
transmitter. It shows that the energy harvested is invalid
in the case when the energy harvested is lesser than the
energy consumed by the circuit for harvesting energy. C2
shows the minimum individual data transfer rate
from transmitter to receiver and it is always less
than the channel capacity. C3 indicates the Quality-of-
Service constraint of the system, which specifies that the
total end-to-end throughput must be greater than the
minimum value of the data rate of the system, . C4
specifies the constraint for a power transmission which
shows that the harvesting energy circuit is capable of
operating in the case when the RF incident power is
greater than the threshold
and less than the
maximum transmitted power
, whose value is
dependent on the hardware limitations of the power
amplifier. A threshold is required for triggering the
charge pump in the circuit for harvesting energy and is
specified in C4. C5 - C7 represent the constraints for
power splitting. C5 indicates that the ratio of power
splitting for energy harvesting is bounded by the lower
limit
and upper limit
. C6 represents the lower
limit and upper limits of the power splitting ratio for data
processing., i.e.,
and
respectively, where
, and
. Power splitting
constraint is specified in C7, which shows the
passiveness of the power splitting unit, and hence no
power gain can be attained by this process of power
splitting.
The theoretical model is practically suitable for
optimization for any node, including the CH node, in the
case when 2 nodes which are inter-connected and are able
to transmit data. Additionally, on the basis of Quality-of-
Service requirements of each node and the system
and are selected in such a way that helps in
attaining a trade-off between system energy efficiency
and the total system capacity. As the value of
increases, the transmit power has to be increased to
satisfy the requirement of greater data rate by reducing
the energy efficiency of the system. Then, based on the
ability of the receiver in dividing the received power, the
values of
and upper limit
are selected.
To maximize the aggregated energy efficiency of all
sensor nodes, the objective function (7) is used. By using
this function, the policy for data rate control , policy
for power splitting , and policy for power allocation
is obtained.
4. Solution of the Optimization
Problem
The resource allocation problem is solved using two
methods, i.e., Resource Allocation Algorithm (ResAll)
[1] and Particle Swarm Optimization (PSO) [2].
4.1 ResAll Algorithm
The ResAll algorithm is based on the iterative
Dinkelback method [13]. Using this algorithm, resource
allocation policies are determined.
Input:
i - index of iteration;
- maximum number of iterations;
n - system energy efficiency;
e - an infinitesimal number;
Output:
- maximum energy efficiency;
- resource allocation policies;
i = 1, n = 0;
for ( )
{
if ( −
return
else
Set
WSEAS TRANSACTIONS on COMMUNICATIONS
DOI: 10.37394/23204.2022.21.18
Amin Al Kabi
E-ISSN: 2224-2864
132
Volume 21, 2022
}
end
This algorithm provides resource allocation policy which
maximizes the energy efficiency This efficiency is further
increased by solving the optimization problem using
Particle swarm optimization (PSO) [2]. Optimum
resource allocation policies are obtained using PSO.
4.2 Particle Swarm Optimization
Particle swarm optimization (PSO) is a computational
method in which optimization is done by trying to
improve a candidate solution problem at each iteration
with respect to a given measure of quality. It is a
population-based method. Here the population of
candidate solutions are known as particles. The objective
of PSO is to find a solution for a constrained
minimization problem based on a particular cost function.
Here the state of the algorithm is represented by a
population, which varies in each iteration until some
criterion is met. Here, the population
is the set of feasible solutions and is referred to as swarm.
These feasible solutions are referred to as
particles, given by . A set
of feasible solutions is considered as the search space in
which these particles move. The number of particles
generally selected is between 10 and 50 in practical for
solving optimization problems.
The population is not changed from generation to
generation in PSOs, instead, the same population is
maintained by updating the particle positions at each. In
PSOs, the particles “interact” or “influence” each other.
… the position vector.
… the ‘historical’ best position.
… the historical best position of the neighboring
particle; it represents the historical best-known position
of the entire swarm in the case of fully connected
topology.
… the velocity; i.e., the step size across and
.
When the algorithm starts, the initial velocities are set
to 0, or to some small random values, and the initial
particle positions are selected in a random manner.
PSO parameters:
In this algorithm, represents the damping factor
known as inertia weight whose value decreases from
around 0.9 to around 0.4 during computation.
represent the acceleration coefficients. In general,
they have values between 0 and 4.
The velocity of the particle is updated as per the equation
(15)
where and represent random variables according to
the uniform distribution U(0,1). The first term of Eq. (15)
is known as the personal component, the middle term
represents the mutual component, and the last one
represents the inertia term.
The particle position is updated based on the equation:
(16)
The termination of this algorithm either occurs once the
fitness value of the particles in the population become
close enough, or when a maximum number of iterations
is reached based on a given cost function. The cost
function Z can be expressed as
(17)
Subjected to constraints C1 to C7.
By using this cost function, the optimization problem can
be solved by using PSO to find the optimum resource
allocation policies ), and hence maximizing the
energy efficiency.
5. Simulation Results
The simulation results using MATLAB v2018b are
discussed here. The simulation settings for ResAll and
PSO are shown in Table I and Table II, respectively.
TABLE I. SIMULATION SETTINGS FOR RESALL
TABLE II. SIMULATION SETTINGS FOR PSO
Nodes are clustered using LEACH (Low Energy
Adaptive Clustering Hierarchy) algorithm, where one
hundred nodes are randomly selected and deployed in an
WSEAS TRANSACTIONS on COMMUNICATIONS
DOI: 10.37394/23204.2022.21.18
Amin Al Kabi
E-ISSN: 2224-2864
133
Volume 21, 2022
100m×100m area. Position of the sink node is fixed at
(50,50). Each node has an initial energy of 0.5J.
Figure 3 shows a comparison of the energy
efficiency vs. the number of iterations of ResAll and PSO
algorithms. It can be seen in Figure 3 that PSO provides
more average energy efficiency compared to ResALL for
a specified number of iterations. This is due to the
improved solution of the cost function Z at each iteration
with respect to the given constraints.
It can be seen that maximum energy efficiency of
12Mb/J is obtained by using PSO while 8M/J is obtained
by using ResAll algorithm.
Figure 3. Energy efficiency vs. number of iterations.
6. Conclusions
The ever-increasing ubiquitous applications of
wireless sensor networks lead to energy scarcity in the
network, which is a serious threat to the lifetime of the
network. To solve this issue, here, Simultaneous Wireless
Information and Power Transfer (SWIPT) technique is
applied to a MWSN. The nodes were clustered using
LEACH algorithm. A resource allocation algorithm is
designed by considering the power splitting capabilities
of relay nodes and cluster heads. Optimal Resource
allocation policies are found out using particle swarm
optimization.
In the proposed method, the received power is split
into two sets of power streams using arbitrary power
splitting ratios. By considering the various power
splitting capabilities of receivers, a Resource Allocation
(ResAll) algorithm is used to find the resource allocation
policies.
In ResAll algorithm, system energy efficiency is
achieved by balancing data rate, energy efficiency, power
splitting ratio, and transmit power. Maximum system
energy efficiency is achieved by balancing transmit
power, data rate, power splitting ratio, and energy
efficiency. This is achieved by framing a cost function
and then improving the solution to the cost function at
each iteration with respect to the constraints. Simulation
result show that the energy efficiency is further increased
by solving the resource allocation problem using particle
swarm optimization.
Acknowledgment
This project was funded “FULLY” by Kuwait Foundation for the
advancement of sciences under project code “PN18-15EE-01”.
References
[1] Songtao Guo, Yawei Shi, Yuanyuan Yang, Bin Xiao, and Fan
Ye, “Energy Efficiency Maximization in Mobile Wireless Energy
Harvesting Sensor Networks,” IEEE Transactions on
Mobile Computing vol. 1, no. 2, pp. 86-91, 2017.
[2] K. Huang and V. K. N. Lau,
“
Enabling Wireless Power
Transfer in Cellular Networks: Architecture, Modeling and
Deployment,” IEEE Trans. Wireless Comm., vol. 13, no. 2,
pp. 902-912, Feb. 2014.
[3] Jun Yang, Hesheng Zhang,
“
Task Allocation for Wireless
Sensor Network Using Modified Binary Particle Swarm
Optimization
,”
IEEE Sensors Journal, vol. 14 no. 3, pp.
882-892.
[4] M. Zhao, J. Li, and Y. Yang,
“
A framework of joint mobile
energy replenishment and data gathering in wireless
rechargeable sensor networks,” IEEE Trans. Mobile
Computing, vol. 13, no. 12, pp. 2689-2705, 2014.
[5] B. Tong, Z. Li, G. Wang and W. Zhang,
“
How wireless
power charging technology affects sensor network
deployment and routing,” IEEE ICDCS, 2010.
[6] R. Zhang and C. K. Ho,
“
MIMO Broadcasting for
Simultaneous Wireless Information and Power Transfer,”
IEEE Trans. Wireless Comm., vol. 12, no.5, pp. 1989-2001,
May 2013.
[7] C. Wang, J. Li, F. Ye and Y. Yang,
“
NETWRAP: An
NDN based real-time wireless recharging framework for
wireless sensor networks,” IEEE Trans. Mobile Computing,
vol. 13, no. 6, pp. 1283-1297, 2014.
[8] Z. Li, Y. Peng, W. Zhang, and D. Qiao,
“
Leach Protocol in
Wireless Sensor Network,”
Proceedings of IEEE 19th Int.
Conf. Network Protocols, 2011, pp. 373–382.
[9] Amin Al Ka’bi, “The effects of mutual coupling and
polarization on the performance of steered-beam adaptive
arrays”, Journal of Computational Electronics, ISSN:
1569-8025, 20(1), 458-466, DOI: 10.1007/s10825-020-
01635-x, Jan. 2021.
[10] W. Dinkelbach,
“
On Nonlinear Fractional Programming,”
Manag. Science, vol. 13, pp. 492-498, Mar. 1967.
[11] Al Ka’bi A. H,” Effect of polarization on the performance
of adaptive antenna arrays,” Journal of Communications,
ISSN: 1796-2021, Volume 15, Issue 9, Pages 661-668,
EID: 2-s2.0-85090368845, September 2020.
[12] K. Anusha and G.I. Shamini,
“
Power and Resource
Allocation Using Modified Binary Particle Swarm
Optimization in Neural Network,” IEEE WiSPNET 2016
conference.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the Creative
Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en_US
WSEAS TRANSACTIONS on COMMUNICATIONS
DOI: 10.37394/23204.2022.21.18
Amin Al Kabi
E-ISSN: 2224-2864
134
Volume 21, 2022
