Improving Network Lifetime in IoT Sensor Network Based on Particle Swarm Optimization, Clustering, and Mobile Sink

Full text

Improving Network Lifetime in IoT Sensor Network
Based on Particle Swarm Optimization, Clustering, and
Mobile Sink
Khalid A. Darabkh*1, Asma’a B. Amareen1, Muna Al-Akhras1, and Wafa’a K. Kassab2
1Department of Computer Engineering, The University of Jordan,
Amman, 11942, Jordan
2Department of Infrastructure and Information Security, Ministry of Finance
Amman, 11118, Jordan
*Corresponding Author Email: k.darabkeh@ju.edu.jo
Abstract— Recently, extending the network lifetime and
reducing energy consumption have become pressing necessities
in the Internet of Things (IoT). Meanwhile, the energy-
constrained sensor nodes pose a challenging problem that
diminishes the nodes’ lifetime, and so as the entire network
system due to energy depletion. Therefore, in this paper, we
propose an innovative energy-aware cluster-based protocol
considering an adaptive Mobile Sink (MS) movement and
utilizing Particle Swarm Optimization (PSO). Accordingly, the
circular network area is divided into clusters, where each cluster
has its own elected cluster head, depending on the PSO method.
The MS, in the proposed work, intends to balance the energy
among nodes, thereby avoiding the hotspot problem. In
particular, it moves in a circular pattern with constant angular
velocity to ensure better coverage. This movement is initiated
from the center of the network area and then goes along the
radius of the network on a forward and backward basis. The
simulation results obtained are so promising and particularly
show that our proposed protocol outperforms its counterparts in
terms of the network lifetime.
Keywords— IoT; Clustering; PSO; Power-aware; Mobile sink
I. INTRODUCTION
Over the last decade, the exponential growth of the Internet
of Things (IoTs) has led to a revolution in both industrial and
academic fields, which has a significant impact on improving
the quality of human being’s life [1] [2]. The main idea behind
IoT technology is to provide a global communication among
all the objects or things that surround us, which will allow
them to accomplish complex tasks [3]. Generally, IoT
environment consists of physical and virtual objects that are
smoothly integrated into the communication network [4][5].
Likewise, Wireless Sensor Networks (WSNs) consist of a
group of spatially distributed sensors that sense and store the
events of surroundings to finally process the gathered
information in order to get a proper decision or action. Since
WSNs covers a big portion of IoT applications, it plays an
important role in evolving and growing IoT technology [6].
Sensors in these networks run on batteries, whose energy is
restricted and cannot be replaced or recharged. Thus,
optimizing the resources in such complicated environments,
plays vital importance in preserving the energy of these nodes
to maximize the network lifetime as much as possible [7][8].
To this end, scholars and researchers have focused on
enhancing and solving networks energy efficiency issues [9].
One of the best solutions, that is adopted by a large number
of researchers, is employing clustering techniques [10]. As
mentioned, sensors are in charge of sending the data they
collect from the surrounding to the Base-Station (BS) node
either by a single hop or a multi-hope way, which will keep
data transmission in free space channel propagation model.
Employing clustering methods succeeded in meeting
researchers’ expectations in designing energy efficient
routing algorithms for such energy-constraint networks. In
this mechanism, sensors are distributed into many sets that are
named clusters, where Cluster Head (CH) node is responsible
for gathering sensed data from its cluster members, and if
required from other nearby CHs [11] [12] [13]. Moreover,
clustering method can conserve communication bandwidth,
support scalability and reduce routing problem among all
nodes through transmitting the data by CHs members toward
BS. CH nodes have a vital role in reducing amount of traffic
that is transmitted through the network area, by compressing
the collected data before sending it to the BS, which will
diminish both delay, packet drop and energy depletion issues.
This will lead to a heavy burden on the CHs, particularly those
that are nearby the BS, triggering energy hole issue and hence
accelerating the first node to die problem [14]. It is worth
mentioning that BS node has a significant impact on
preserving the network energy [15]. As a result, introducing
Mobile Sink (MS) node into network area has been shown to
be an effective method to mitigate the energy hole issue and
to balance the energy consumption [16]. Additionally, it
brings new challenges and opportunities by improving the
positioning accuracy of sensors, while dynamically changing
the network topology to allow direct communication between
CHs and MS node [17]. In light of this, the network traffic is
no longer concentrated on the nodes near to the sink node,
which will alleviate the hot spot issue by avoiding the
bottleneck rests upon these nodes [18].
The authors in [19] suggested an Energy-Efficient Cluster-
based Dynamic Routes Adjustment (ECDRA) approach for
WSNs with MSs, where wireless sensors are distributed
uniformly over circular network area that is partitioned into
mutual clusters. Two mobile sinks are moving at the outer
border of this network in a counter-clockwise way, which will
distribute the network traffic evenly between them. In the first
time, CH will be selected from the center of each sector,
whereas the threshold energy will be the main factor that
manages heads’ selection on consecutive rounds. Nodes will
transmit sensed data directly to their cluster heads when the
distance between them is less than a threshold value,
otherwise multi-hop method will take place in the
transmitting process. It is worth stating that CHs are
responsible of identifying the location of sink node in each
round, which will reduce the amount of control packets and
prolong the life time of the network.

Surekha et al, proposed new clustering algorithm named
Cluster-Chain Mobile Agent Routing (CCMAR), where a
Mobile Agent (MA) sensor is in charge of collecting and
compressing all the data gathered by other nodes [20]. In this
work, the network area is segmented into a number of
clusters, where each cluster has a head called a cluster chain.
Each round consists of two stages, wherein the first stage the
nodes belong to the same cluster will compose a chain to
aggregate the sensed data. A mobile agent will be sent from
sink node, in the second stage to collect the compressed data
from clusters chain nodes. The MA node path will be
determined based on several criteria such as path loss,
residual energy and signal strength. However, CCMAR
suffers from frequent data forwarding, causing more energy
depletion. To overcome this problem, an Energy Efficient
Routing Algorithm with MS support for WSNs (EEMAR)
was developed by Wang et al [21].
The circular area in EEMAR is partitioned into an equal
number of sectors. Cluster members' weight is considered in
CHs selection, which takes into account many factors such as
the distance between the node and sink and node’s residual
energy. Additionally, transmitting data directly or through
multi-hop method is employed based on specific criteria. The
CHs depend on Greedy algorithm to transmit the aggregated
data to MS node that moves in a static circular radius. Wang
et al, developed new protocol that selects CH for each cluster
based on Particle Swarm Optimization (PSO) algorithm [22].
In this work, PSO considers many parameters in selection
process, which are node’s residual energy and the distance
between sensor nodes and MS. The movement of the sink
node is adaptive based on specific conditions to enlarge the
network life time and diminish the amount of the consumed
energy. Interestingly, circular movement of sink nodes, with
a predetermined radius and a constant angular velocity,
strengthens the connectivity among the network sensors,
thereby reducing communication overhead and delay when
gathering the data from CH nodes.
In our proposed protocol, sensors are randomly deployed over
a circular area that is split into identical sectors. On the other
hand, the network lifetime is divided into equal rounds
intervals. The first round only is consisting of both setup and
steady state phases, while the others involve just the steady
stage. Simulation results indicate that our proposed protocol
outperforms other protocols in term of the network life time
as will be discussed shortly.
The remainder of this paper is arranged as follows: Section II
presents the proposed protocol, section III demonstrates and
discusses the simulation results. Finally, paper is concluded
in Section IV.
II. THE PROPOSED PROTOCOL
This section illustrates the proposed protocol thoroughly,
which aims to minimize energy consumption by proposing an
efficient clustering algorithm in which a combination of both
node's residual power and its distance to the MS to elect the
CH are considered bearing in mind that, the CH’s energy is
depleted faster compared with other nodes. Referring to this,
our proposed protocol takes into an account the power
consumption balancing between nodes and preventing the
earlier death for the CH as much as possible. In particular, the
PSO, which is considered one of the best optimization
methods, is used for CH selection. Also, the sink employed,
in this work, is mobile and its movement is proposed in a way
to achieve a full network coverage. The following subsections
discuss the proposed protocol extensively.
A. Protocol’s Assumptions
In the proposed protocol of this study, several assumptions
are taken into consideration as follows:
• Sensors are deployed randomly and are static.
• All nodes have the same initial energy and non-
rechargeable batteries and they are able to predict the
location of the other nodes through data exchanges.
• The sensing region is partitioned into equal-sized sectors
(clusters), with each sector having a CH selected
depending on PSO.
• Sensor nodes are connected to clusters according to their
positions where each one has its own ID and they
communicate with the CH via either a single or multi-hop
communication.
• MS has unlimited energy, communication range, as well
as constant angular velocity and it has a well-designed
circular trajectory.
B. Network Model
In this model, the sensor nodes are randomly deployed in
circular sensing area with radius R, and the total number of
sensors is denoted as N. Fig. 1 shows the network model
considered for this work. The sensing area is divided into
equal angular sectors or clusters. Each sector is called 
that is known by all sensors where m refers to the number of
clusters formed. Every sensor has a predefined time division
multiple access time slot to transmit its packet, and when
transmission is done, the sensor's transceiver goes into sleep
mode to conserve more energy. At the beginning of a set-up
phase, every sensor sends an information message to the sink
includes its ID, position and residual power. Accordingly, the
sink generates a table that includes all of this information.
This can be used to elect a sensor to be a CHj

1<j<m

for each
cluster. The routing path is determined based on finding the
minimum energy consumption path between a Sensor Node
(SNi) and CHj. As a result, SNi may send its information to
CHj directly or use another sensor node

SNk as a relay which
in turn sends a packet includes its own information and SNi's
information directly or to another relay node and so on until
transmitting to CHj directly. A sensor node is considered
dead when it has little or insufficient power to send a packet
to the sink. At this case, this sensor is going to be abandoned
from the sink's table and if this sensor is a relay node for
another sensor SNi, then the routing path should be reformed.
Intriguingly, if a CH is dead, then the MS will broadcast a
packet indicating the time of re-clustering all over the
network and this is to warrantee having all sensors
synchronized. The network lifetime gets portioned into
rounds. The first round has two phases, specifically, the set-
up phase and the steady phase, while subsequent rounds may
only include the steady phase or sometimes both. During the
set-up phase, the MS will typically begin dividing the area
into sectors, selecting CHs for each sector using the PSO
algorithm, which will be explained in more detail shortly.
C. Network Area Division
The sensing field is divided into equal angular clusters, as
conducted by [21] as shown in Fig. 1. It’s worth mentioning

that the angular cluster helps in minimizing the reclustering
and balancing the network load. However, each cluster angle
(
)
A is defined as:
,
2
n
C
π
A= (1)
where n
Cis the number of clusters in the network.
D. Sink Movement
In this work, the MS is proposed to move in a circular pattern
with radius Rms and a constant angular velocity, noting that
Rms is changed every number of rounds i depending on a
quarter addition of the network radius to its prior value. After
reaching three quarters of the network radius,

Rms will
reversely follow the same forward trend. It is noteworthy to
mention that the MS firstly broadcasts its initial position and
angular velocity which will make it possible for each sensor
node to predict the sink's instantaneous position at any time.
It is known that the angular velocity (ω, radians/sec) is
identified as the ratio of the change in angular rotation (∆θ,
radians) to the change in time (∆t, sec). To have it clearer, the
new MS location, which is, for example, at position k (
)
is being calculated as follow:
,
01
01
1
tan
0
/
1
1
tan
1
1







−
−
−
−
−
=






−
−
=
−x
k
x
y
k
y
ms
P
k
ms
P
k
θ
(2)
,
1
12
ωθθ
×+
−
=RT
kk (3)
,
2
sin,
2
cos, 
















×=








→kk
ms
R
y
k
ms
P
x
k
ms
P
k
ms
P
θθ
(4)
where, 
refers to the initial position of MS,



denotes for the last (previous) position of MS, and
represents the round time.
E. CH Selection using PSO
In this part, the process of selecting the CHs is described. The
selection of CHs is more demandable in homogeneous
networks due the similarity in power for all nodes [23] [24].
In this work, we use the PSO to find the best CH for each
cluster. It is worth stating that the MS disseminates a HELLO
message for all sensor nodes. This message involves the
related information of the MS, including the initial coordinate
and the speed of the MS. Then, the nodes response with a
HELLO message contains the node’s ID, the node’s
coordinates and the node’s energy. Based on that, the MS will
have information about all nodes and further all nodes will
have information about the MS. After that and as mentioned
earlier, the MS starts the set-up state by dividing the network
area into equal-sized clusters, distributing nodes among them
and the CHs get elected based on PSO algorithm [25]. When
initiating the process of PSO, which is an iterative approach,
each particle is deployed with a random position and velocity.
In each iteration, the particle finds its own personal best that
is named “

best ” and also the global best that is named
“ ”. To achieve the  solution, it uses its

best and
 with the objective of updating the position and velocity.
It is assumed that fitness function () of PSO represents the
residual power of the sensor node along with the distance
between the sensor node and the MS. Technically speaking,
the PSO contains a predetermined number of particles ()
named a swarm, as particle gives a potential solution. Each
particle  has two parameters, including its position and its
velocity, where the fitness function () evaluates the quality
of solution provided by each particle. The primary principle
of PSO is to find the best particle’s positions with the best
result of fitness function (). In more details, assume that 

is a function that expresses the distance between a sensor node
and the MS (

). We need to maximize
 for optimal
CH selection as the distance between the node and the MS
becomes as short as possible. Also, we assume that 
 is a
function which basically refers to the ratio of node’s residual
energy () to the node’s initial energy () bearing in mind
that this ratio should be maximized to achieve the optimum
CH election. The two objective functions ( 
,


) are
normalized between the range of 0 and 1. Noting that, these
two functions will be used to find the fitness function () in
the PSO and the appropriate way is to maximize their linear
combination. Therefore, the optimal linear combination for
CH selection is as follows:
(
)
,
2
1
1ffF ×−+×=
αα
(5)
where,
(
)
.1
1
0 ,/1
1<<
→
−= f R
msn
Df (6)
.1
2
0 ,/
2<<= f
o
I
E
R
Ef (7)
where R refers to the radius of the network and α denotes for
a parameter that normalizes the combination of 
 and 

subject to 0 <α<1.
PSO algorithm is repetitive, where in the initialization stage
of PSO, the nodes in each cluster gets arranged in a matrix
contains the ID, position and the energy of the nodes. Then
the fitness function (F) is computed based on the matrix
values. The node that has the maximum fitness function ()
is considered as the  . Thereafter, the algorithm forms a
fitness matrix of all nodes in the cluster. Noting that, the
particles are initialized by allocating random coordinates to
the nodes. Every particle announces itself as a local or

bestsolution, and then it searches for the  solution. To
find the  solution, during each iteration, every particle
uses its own

best and  solution to update its position
and velocity in order to reach the  solution, according to:





−××+





−××+×=
+)()(
22
)()(
11
)()1( t
x
P
t
x
best
Grc
t
x
P
t
x
best
Prc
t
x
V
t
x
V
ω
(8)
Fig. 1. CH selection








−××+







−××+×=
+)()(
22
)()(
11
)()1( t
y
P
t
y
best
Grc
t
y
P
t
y
best
Prc
t
y
V
t
y
V
ω
(9)
)1()()1( +
+=
+t
x
V
t
x
P
t
x
P (10)
)1()()1( +
+=
+t
y
V
t
y
P
t
y
P (11)
where, (
 
) refer to the current position of particle on x, y
coordinates respectively in an iteration t. Vx, Vy denote the
velocity of particle on x, y coordinates respectively in an
iteration t.ω refers to the inertia weight.   refer to the
acceleration coefficients where 0!  ! ". #
 # refer to
the random values where 0 ! #
 #! " . To reach an
acceptable value of Gbest , which is the particle that has the
maximum fitness, the updating process is repeated. When
having a new updated position, the particle recalculates the
fitness function (F) and accordingly updates the $%&' and
Gbest, as follows:




>
=.,
, ,
Otherwise
i
Pbest
))
i
estfitness(Pb)
i
(Pf (fitness i
i
P
i
Pbest (12)


>
=.,
,,
Otherwise
Gbest
est))fitness(Gb)
i
stitness(Pbe if (f
i
Pbest
Gbest (13)
As a summary, the PSO algorithm works respectively in each
iteration as follows: Initially, every node represents a

bestsolution for itself. Secondly, two random variables (#
,
#) are generated randomly between [0,1]. After that, every
particle updates the velocity and position to find the nearest
point to the solution or a candidate solution. Then, the fitness
value is evaluated for each particle to compare the fitness
values of candidate and old solution then update the $%&'.
Accordingly, if the fitness of the old solution is greater than
(or the same as) the fitness of the candidate solution, then the
old solution will remain with keeping the as before.
However, if the fitness of the old solution is less than the
fitness of the candidate solution, then the gets updated.
After that, the fitness of the candidate and the fitness of
Gbestare compared. In other words, if the fitness of the Gbestis
less than the fitness of the candidate solution, then the
candidate will certainly become the new Gbest . Finally, the
CH will be elected when the iteration reaches its maximum
value or the Gbestis equal to the $%&' and the velocities are
equal to zero, for all entries in the fitness matrix, in the next
iterations.
F. Radio Energy and Channel Propagation Models
To transmit or receive a packet, then () (Nano Joule/bit)
will be the energy consumed by the transmitter or receiver per
a bit to run the electronic circuit. Thus, the L-bit packet
transmission consumes precisely * + () , Furthermore,
there is an energy consumed for amplifying the packet to
reach its destination. In other words, there are two channel
propagation models used, namely, free space (fs) model and
multipath model (mp). As a matter of fact, the signal is
assumed to propagate according to one of these two models
until it reaches the receiver situated at a distance d.
Considering

- and

., which donate for the amplification
coefficients of one bit in free space model and multipath
model correspondingly, then the threshold distance (/0) of
aforementioned channel propagation models can be
represented by:
.
0mp
fs
d
ε
ε
= (14)
Consequently, the total energy consumed for a packet
transmission (12* /3) is as follows:





≥××+×
<××+×
=
.,
4
,,
2
,
o
d if dd
mp
L
elec
EL
o
d if dd
fs
L
elec
EL
d)(L
Tx
E
ε
ε
(15)
While the energy consumed for receiving a packet (2*3)
is represented by:
.
elect
EL(L)
Rx
E×= (16)
III. SIMULATION RESULTS AND DISCUSSION
This section outlines the experimental setup of the network
and analyze the efficiency of the proposed protocol. In this
work, simulations were conducted using MATLAB R2018
software over a circular network area with various radius. The
simulation parameters are shown in Table 1.
Table 1. Simulation parameters
Parameter Name Value
Sensing field radius (R) [100, 200, 300, 400] m
Radius of MS Variant [0.25,0.5,0.75]
×
R
Speed of the MS [π/20, π/10, π/5]
Number of clusters 5
Number of rounds for a new MS
radius
400
Number of nodes (N) 100
Packet length (L) 500 bits
The initial energy for each node 0.5 J
The amount of energy that
required to transmit one bit
50 nJ/bit
The amount of energy that is
dissipated by the transmission
amplifier
fs
ε
10 pJ/bit/
4

mp
ε
0.0013 pJ/bit/
4
5
The threshold distance which
determines the use of either free
space model or two ray model
87
It is worth mentioning that numerous simulations were
performed to evaluate the performance of the proposed
protocol, in term of the network lifetime that is measured by
the number of rounds.
Fig. 2 shows that nodes in CCMAR protocol died earlier and
faster compared with the other protocols. This is due to the
direct communications between CHs and MS and also to the
structure of the chain that leads to a heavy burden of
forwarding data. MS nodes’ movement in the outer side of the
circular area in ECDRA protocol leads to consume more
energy from the sensors that are located in the network
boundaries. Additionally, the way used for CHs selection,
which is based on a random predetermined energy threshold,
has led to minimize and shorten the network life time faster
than both EEMAR and proposed protocols.
It is clearly shown in Fig. 2 that the proposed protocol
outperforms significantly the other protocols in term of the
network lifetime. This is due to the impact of the following
ideas. Firstly, the MS moves in a dynamic circular way, which
will place the sink relatively close to the nodes and covering
the entire network. By this way, the long communication
distances along with the power consumption will be
decreased. In more detail, the MS in the proposed protocol
moves in a circular path along the network with various radius
on a forward and reverse basis. The radius of this path is

changed every number of rounds. It is worth mentioning that
employing different network’s radii for various rounds are
simulated and examined, where it is found that the best
pattern for the MS is to change its movement radius every i
rounds for a quarter of the radius. Accordingly, for a
consecutive of i rounds, the  , in the forward direction, is
changed based on this vector; [0.25,0.5,0.75] ×Rms whereas,
in the reverse direction, it is changed based on this vector;
[0.25,0.5,0.75] ×Rms . Secondly, the proposed protocol
employs PSO technique for CHs selection without
transmitting extra control packets. There are many factors
considered in CH selection, such as being the nearest to the
sink and having the highest residual energy. In this method,
the CH role will be rotated among the nodes, which will help
in distributing the energy consumption evenly among all
sensor nodes. Thirdly, the proposed protocol makes it
possible to locate the MS by CHs without the need for
frequent broadcast messages, which would cause a loose of
energy. Finally, the set-up phase takes place at the first round
only, which will contribute in reducing the consumption of
the network energy in control packets transmission.
Fig. 2. Network lifetime considering different protocols
IV. CONCLUSION
In this work, we proposed a novel IoT protocol focusing on
the clustering technology as well as the sink mobility to
maximize the network lifetime through minimizing the
amount of energy consumption. In other words, there are
novel mechanisms that were employed in this work, such as
clustering that divides the network area into a number of
identical sectors. Each sector selects its CH based on PSO
algorithm considering many parameters as the node’s residual
energy and how far a node is from the MS. Furthermore, the
MS movement is chosen in a method that ensures balancing
the energy dissipation among all sensor nodes by designing a
circular sink path with a dynamic radius and constant angular
velocity. Interestingly, it is clearly proven that the ideas
incorporated in this work contribute significantly in
improving network efficiency in terms of energy
consumption and network lifetime.
REFERENCES
[1]
X. Li and L. D. Xu, "A Review of Internet of Things
—
Resource
Allocation," IEEE Internet of Things Journal, vol. 8, no. 11, pp. 8657-
8666,
2021.
[2]
A. A. Laghari, K. Wu, R. A. Laghari, M. Ali and A. A. Khan , "A Review
and State of Art of Internet of Things (IoT),"
Archives of Computational
Methods in Engineering,
vol. 29, p. 1395
–
1413, 2022.
[3]
W. Kassab and K. A. Darabkh, "A
-
Z Survey of Internet
of Things:
Architectures, Protocols, Applications, Recent Advances, Future Directions
and Recommendations,"
Journal of Network and Computer Applications,
vol. 163, p. 102663, 2020.
[4]
K. A. Darabkh, W. K. Kassab and A. F. Khalifeh, "LiM
-
AHP
-
G
-
C: Life
Time Maximizing Based on Analytical Hierarchal Process and Genetic
Clustering Protocol for the Internet of Things Environment,”,"
Computer
Networks,
vol. 176, p. 107253, 2020.
[5]
K. A. Darabkh and W. K. Kassab, "Maximizing the Life Time of Wireless
Sensor Networks Over IoT Environment," in
2020 Fifth International
Conference on Fog and Mobile Edge Computing (FMEC)
, Paris, France,
2020.
[6]
K. A. Darabkh, S. M. Odetallah, Z. Al
-
qudah, A. F. Khalifeh and M. M.
Shurman, "Energy–Aware and Density-
Based Clustering and Relaying
Protocol (EA-DB-
CRP) for Gathering Data in Wireless Sensor Networks,"
Applied Soft Computing,
vol. 80, pp. 154
-
166, 2019.
[7]
N. Chen, T. Qiu, M. Daneshmand and D. O. Wu, "Robust Networking:
Dynamic Topology Evolution Learning for Internet of Things,"
ACM
Transactions on Sensor Networks,
vol. 17, no. 3, pp. 1
-
23, 2021.
[8]
K. A. Darabkh, M. Z. El
-
Yabroudi and A. H. El
-
Mousa, "BPA
-
CRP: A
Balanced Power-
Aware Clustering and Routing Protocol for Wireless
Sensor Networks,"
Ad hoc Networks,
vol. 82, pp. 155
-
171, 2019.
[9]
L. Krishnasamy, K. R. Dhanaraj, D. G. Gopal , R. T. Gadekallu, M. K.
Aboudaif and E. A. Nasr, "A Heur
istic Angular Clustering Framework for
Secured Statistical Data Aggregation in Sensor Networks," Sensors,
vol.
20, no. 17, p. 4937, 2020.
[10]
K. A. Darabkh, J. N. Zomot, Z. Al
-
qudah and A. F. Khalifeh, "Impairments
-
Aware Time Slot Allocation Model for Energy-constrained Multi-
Hop
Clustered IoT Nodes Considering TDMA and DSSS MAC Protocols,"
Journal of Industrial Information Integration,
vol. 25,
p. 100243, 2022.
[11]
K. A. Darabkh, J. N. Zomot and Z. Al
-
qudah, "EDB
-
CHS
-
BOF: Energy
and Distance Based Cluster Head Selection with Balanced Objective
Function Protocol,"
IET Communications, Special Issue: Future of
Intelligent Wireless LANs,
vol. 13
, no. 19, pp. 3168
-
3180, 2019.
[12]
A. Shahraki, A. Taherkordi, Ø. Haugen and F. Eliassen, "A Survey and
Future Directions on Clustering: From WSNs to IoT and Modern
Networking Paradigms,"
IEEE Transactions on Network and Service
Management,
vol. 18, no
. 2, pp. 2242
-
2274, 2021.
[13]
J. Daniel, S. F. V. Francis and S. Velliangiri, "Cluster head selection in
wireless sensor network using tunicate swarm butterfly optimization
algorithm,"
Wireless Networks,
vol. 27, p. 5245
–
5262, 2021.
[14]
A. A. Jovith, M. Mathapati, M. Sundarrajan, N. Gnanasankaran, S. Kadry,
M. N. Meqdad and S. M. Aslam, "Two-
Tier Clustering with Routing
Protocol for IoT Assisted WSN," Computers, Materials & Continua,
vol.
71, no. 2, p. 3375
–
3392, 2022.
[15]
K. A. Darab
kh, E. N. AL
-
zoubi, F. A. Al
-
naimat and A. F. Khalifeh,
"Mobile sink optimization for enhancing data delivery in wireless sensor
networks," in
2020 IEEE International IOT, Electronics and Mechatronics
Conference (IEMTRONICS)
, Vancouver, BC, Canada, 2020.
[16]
G. Xie and F. Pan, "Cluster
-
Based Routing for the Mobile Sink in Wireless
Sensor Networks With Obstacles," IEEE Access, vol. 4, pp. 2019-
2028,
2016.
[17]
S. Yalçın and E. Erdem, "TEO
-
MCRP: Thermal exchange optimization
-
based clustering routing pr
otocol with a mobile sink for wireless sensor
networks," Journal of King Saud University -
Computer and Information
Sciences,
vol. In press, 2022.
[18]
V. Agarwal, S. Tapaswi and P. Chanak, "A Survey on Path Planning
Techniques for Mobile Sink in IoT-En
abled Wireless Sensor Networks,"
Wireless Personal Communications,
vol. 119, p. 211
–
238, 2021.
[19]
J. Wang, J. Cao, S. Ji and J. Park, "Energy
-
efficient cluster
-
based dynamic
routes adjustment approach for wireless sensor networks with mobile
sinks,"
The Journal of Supercomputing,
vol. 73, no. 7, p. 3277
–
3290, 2017.
[20]
S. Sasirekha and S. Swamynathan, "Cluster
-
chain mobile agent routing
algorithm for efficient data aggregation in wireless sensor network,"
Journal of Communications and Networks, vol. 19, no. 4, pp. 392 -
401,
2017.
[21]
J. Wang , Y. Gao , W. Liu, A. K. Sangaiah and H.
-
J. Kim , "Energy Efficient
Routing Algorithm with Mobile Sink Support for Wireless Sensor
Networks,"
Advances in Sustainable Computing for Wireless Sensor
Network
s,
vol. 19, no. 7, 2019.
[22]
J. Wang , Y. Gao, W. Liu , K. A. Sangaiah and H.
-
J. Kim , "An Improved
Routing Schema with Special Clustering Using PSO Algorithm for
Heterogeneous Wireless Sensor Network," Sensors,
vol. 19, no. 3, p. 671,
2019.
[23]
D.
Pandita and K. R. Malik, "A Survey on Clustered and Energy Efficient
Routing Protocols for Wireless Sensor Networks,"
International Journal of
Trend in Scientific Research and Development (ijtsrd),
vol. 2, no. 6, pp.
1026
-
1030, 2018.
[24]
A. Kardi and
R. Zagrouba, "RaCH: A New Radial Cluster Head Selection
Algorithm for Wireless Sensor Networks,"
Wireless Personal
Communications,
vol. 113, p. 2127
–
2140, 2020.
[25]
A. G. Gad, "Particle Swarm Optimization Algorithm and Its Applications:
A Systematic Review,"
Archives of Computational Methods in
Engineering,
p. 2531
–
2561, 2022.