Content uploaded by Rajeev Paulus
Author content
All content in this area was uploaded by Rajeev Paulus on Dec 15, 2022
Content may be subject to copyright.
Received: 21 February 2022 Revised: 18 October 2022 Accepted: 20 October 2022
DOI: 10.1002/cpe.7535
RESEARCH ARTICLE
Coverage hole aware optimal cluster based routing for wireless
sensor network assisted IoT using hybrid deep recurrent
neural network
Abhishek Srivastava Rajeev Paulus
Department of Electronics and Communication
Engineering, SHUATS, Prayagraj, India
Correspondence
Abhishek Srivastava, Department of
Electronics and Communication Engineering,
SHUATS, Prayagraj, India.
Email: abhishek200959@gmail.com
Summary
Nowadays, the wireless sensor network (WSN) with IoT is intended to monitor
real-worldphysicalorenvironmentalphenomena in a number of applications, including
foreign areas such as health and habitat monitoring. The WSN-IoT network gener-
ates huge volume of data, which has to be processed and accessed by the remote
users. Due to this large volume of data generation and resource constraint ability make
achieving optimal cluster based routing in WSN-IoT. The location of the sensor nodes
significantly affects the accuracy of the information collected, which determines the
quality of service provided by the application system. WSN can have multiple conflicts,
which can create different coverage holes. These holes will break the existing overlap
or connection and affect the required operation of the networks. Therefore, it is essen-
tial to find and repair the coverage holes to ensure the full functioning of the WSN
as a motivation of this study. In this paper, we suggest a novel Coverage Hole aware
Optimal Cluster based Routing (CHOCR) scheme for WSN-IoT. First, we propose Mod-
ified Lichtenberg optimization (MLO) algorithm for balanced clustering which improve
the performance of coverage hole. Second, we develop a linear equilibrium optimiza-
tion based decision making (LEO-DM) technique to subtract trust value of each IoT
node using multiple restraints in cluster and consider the highest trusted node is act
as cluster head (CH). After that, a hybrid deep recurrent neural network (HD-RNN) is
developed for intermediate node selection to frame the routing between two nodes.
Finally, we simulate our proposed CHOCR scheme on the NS3.26 simulator. Accord-
ing to energy consumption, network longevity, number of nodes that are still alive
and packet delivery ratio, packet loss ratio and throughput, end-to-end latency and
delay of our proposed CHOCR routing system, we compare it to other current routing
schemes.
KEYWORDS
cluster head, coverage hole, neighboring node selection, optimalcluster routing, WSN-IoT
1INTRODUCTION
Detecting and monitoring life events using a wireless sensor network (WSN) is a novel technology. Sensor nodes in WSNs are often deployed in
open, unprotected areas1,2 and are powered by a huge number of batteries. Sensor terminals may communicate at a wide variety of power levels,
Concurrency Computat Pract Exper. 2022;e7535. wileyonlinelibrary.com/journal/cpe © 2022 John Wiley & Sons, Ltd. 1of21
https://doi.org/10.1002/cpe.7535
2of21 SRI VA S TAVA AND PAULUS
both from the sensor and from the battery. Modern routing protocols and quick technical advancements have made touch nozzles commonplace in
the microelectronics sector. WSNs may be used for target tracking, environmental monitoring, and war tracking.3In order to monitor and record
environmental changes, the WSN makes use of thousands of sensor nodes powered by tiny batteries that are placed in the open.4As a result, it
is difficult to provide complete coverage when using a random deployment method in which the placements of the sensor nodes are not precisely
regulated.5
Major network standardization projects, including IETF 6 LoWPAN, RPL, and COAP, are at the forefront of WSN IoT applications.6There are
severalways in which the Internet of Things (IoT) may be used as a kind of philanthropy, from healthcare to transportation to industrial automation.7
IoT sensor nodes are built on the IoT network with restricted power, memory, and computing capabilities.8Physical media is managed such that
losses are minimized.9
Wireless Sensor Networks (WSN) nodes are able to process, store, and communicate with other nodes on the network.10 To get the most
out of RPL, you’ll need to take a closer look at each currency’s performance metrics.11 The IETF established the low power network rout-
ing protocol (LPNRP) as the low power network standard that is specifically tailored for wireless networks.12 Additionally, it’s vital to evaluate
the network’s strength and dependability while evaluating these technologies for their ability to perform and be efficient. WSN is a vast net-
work with a large number of self-contained, low-power, low-cost, and compact touch nodes.13 Data may be collected and synchronized by
sensors of this kind. The coordination between nodes14 is an important function of this kind of network. Sensor nodes are able to do cal-
culations locally and relay only partly processed data because of this. These factors may have a negative impact on WSN dependability and
predictability.
Live monitoring of mobile targets and emergency scenarios in surveillance systems in combat contexts are two examples of crucial situations
whendependabledataroutingisimportant. To accomplish a common goal, the Internet ofThingsdependsona huge number of sensors and actuators
working together.15 Thus, WSNs are essential for getting context-specific information and enabling decision-making in the real world. In a WSN, a
source node is a sensor using wireless technology; this sensor may collect data from the surrounding environment and transmit it to a central place
known as the control station.16 Alternative topologies maybe employed since the WSN does not have any physicalinfrastructure. The most common
topology for large network design is a tree-based or hierarchical one.
IntheInternetofThings (IoT) paradigm, physicalitemsare being transformed into intelligentdevicesthat can be networkedoverIP and affecting
current views of the Internet’s future.17 A concept known as Internet of Things (IoT) will be critical to future Internet development. Application
areas for the WSN include tracking of trafficand military activities, weather forecasts, landslide detection, and fire detection. IoT and cyber-physical
systems rely on WSNs as their foundation.18 The Internet of Things (IoT) is a relatively new notion that refers to everyday objects that have Internet
Protocol (IP) connections and are therefore fully interconnected with the Internet.
In IoT, a WSN is a monitoring system composed of unmanned nodes that are aware of their environment and synchronous nodes that gather
data and act as Internet gateways.19 Sensor nodes and synchronization are not necessarily instantaneous in isolated WSNs, when sync nodes are not
always present. The WSN’s lifespan may be extended if the programme does not need direct data collection via the useof data unit storage and data
transmission.20 WSN-IoT faces a number of challenges, including network performance, connectivity, security, integration, and energy efficiency
which mainly cause of coverage hole. The coverage hole is a location in the WSN-IoT where no data identification or transmission is available, so
identifying such a site can be difficult. Once the holes are found, another important task is to determine the edges of the border that reflect the
cover holes. WSN-IoT has problems with limited battery capacity, intermittent disconnection due to multi-jump connection, and short transmission
range. To overcome such problems, we proposed a novel technique for detecting coverage gaps in WSN-IoT networks is called “coveragehole aware
optimum cluster based routing” (CHOCR).
1.1 Contributions
The novel contributions of our proposed CHOCR method are:
1. We use a Modified Lichtenberg optimization (MLO) algorithm for balanced clustering which improve the performance of coverage hole.
2. We propose a Linear Equilibrium Optimization based Decision Making (LEO-DM) technique to subtract trust value of each IoT node spending
multiple constraints in cluster and consider the highest trusted node is act as Cluster Head (CH).
3. We propose a hybrid deep recurrent neural network (HD-RNN) for intermediate node selection to frame the routing between
two nodes.
Finally, the presentation of our proposed CHOCR scheme is related with several current state-of-the-art routing schemes such as Repair, HWSN
and MO-EPO schemes in terms of liveliness consumption, network lifespan, the number of alive nodes, packet delivery ratio, packet loss ratio,
throughput, end-to-end latency and delay.
SRI VA S TAVA AND PAULUS 3of21
1.2 Organization of paper
The paper’s remaining sections are follows: Coverage-hole finding and retrieval methodology for WSN-IoT are discussed in Section 2. CHOCR’s
problem statement and network model are detailed in Section 3. In Section 4, a mathematical model depicts the workings of the proposed CHOCR
system. Discussion of the findings and results comparisons may be found in Section 5. Section 6concludes the paper.
2RELATED WORKS
In this section, we study and analyze the recent works related to the coverage-hole detection and efficient cluster created beating protocols in
WSN-IoT.
Vaiyapuri et al.21 developed an IoT enabled cluster based routing (CBR) protocol for information centric wireless sensor networks (ICWSN),
named CBR-ICWSN. The developed model employed a black widow optimization (BWO) based clustering technique to select the optimal set of
cluster heads (CHs) effectively. Besides, the CBR-ICWSN technique involved an oppositional artificial bee colony (OABC) based routing process for
optimal selection of paths. A series of simulations validated the performance of the CBR-ICWSN technique and the results are examined under
several aspects. The experimental results of the CBR-ICWSN technique outperformed other methods in terms of network lifetime and energy
efficiency.
An optimized Orphan-LEACH (O-LEACH) was developed by Senthil et al.22 to facilitate the formation of a novel process of clustering, which
resulted in minimized usage of energy as well as enhanced network longevity. A hybrid optimization utilizing simulated annealing with Lightning
Search Algorithm (LSA) (SA-LSA), and particle swarm optimization with LSA (PSO-LSA) Algorithm was introduced. Those developed techniques
effectively managed the CH selection achieving optimal path routing and minimization in energy usage, resulting in the enhanced lifespan of the
WSN. The developed technique’s performance surpassed other techniques.
Sujanthi et al.23 designed a novel Secure Deep Learning (SecDL) approach for dynamic cluster-based WSN-IoT networks. A new
One Time-PRESENT (OT-PRESENT) cryptography algorithm was designed to achieve high-level security for aggregated data. Then,
the ciphertext was transmitted to mobile sink through optimal route to ensure high-level QoS. For optimal route selection, a novel
Crossover based Fitted Deep Neural Network (Co-FitDNN) was developed. They employed the concept of data mining to authenti-
cate the IoT users. All IoT users were authenticated by Apriori based Robust Multi-factor Validation algorithm which maps the optimal
authentication feature set for each user. The developed SecDL approach achieved security, QoS and energy efficiency. Finally, the net-
work was modeled in NS-3.26 and the results showed improvement in network lifetime, throughput, packet delivery ratio, delay and
encryption time.
Pan et al.24 have planned a frivolous and distributed geographic multicast beating technology to overcome the aforementioned chal-
lenge. There are three stages to our plan. To reach the multicast destination, the first step is to choose an intermediate node. The sec-
ond step closes the loop and shortens the root that was formed in the first. Finally, assess whether or not the selected connections can
be combined. The suggested method effectively reduces the transmission connections of the created multicast path and shortens the way,
according to simulation findings. They also put a scheme created on a ZigBee-related platform to the test to see if it can be employed in
IoT applications.
Devi et al.25 have recommended submitting information based on the content. Interconnected data on the same route reduces route
traffic. Thus, barriers are reduced, traffic is reduced, and battery energy is saved. Stability also ensures that the proposed method is supe-
rior to modern technology. IoT employed in a variety of industries. Collecting data from such networks, on the other hand, will boost
traffic.
Thangaramya et al.26 have proposed the IoT routing that is intelligent WSN is a significant issue that must be addressed in order to improve
network service quality (QoS). Furthermore, preventing massive packet loss and packetloss, quick power loss, network inequality, and node failure
is difficult due to the high power required for IoT-based touch network connection. It slows down packet delivery and reduces packet delivery. As
a result, it’s critical to track power consumption at the nodes in order to recover the network’s overall presentation and develop effective routing
solutions employing intelligent mechanical technology.
SLP difficulties may be solved and power consumption reduced by using a random routing system by He et al.27 In SRR, data packages are
delivered to Phantom sources in various regions and routed in all instructions to reach the sync node. Limits on route strategycontrol and dynamism
feasting were also established by IMD. Theoretical investigation and test results confirm that certain protocols can effectively reduce retaliation
and targeted attacks while maintaining security and network life.
In an Internet of Things (IoT) scenario, Sumathi et al.28 created artificial intelligence tree routing based on RNN and ZigBee protocol. Accord-
ing to simulation results, NEWTR outperforms AODV routing protocol in terms of network lifetime by 5.549% and energy consumption (EC)
by 5.817%.
4of21 SRI VA S TAVA AND PAULUS
A thorough examination of MAC protocols in WBANs was created by Javadpour et al.29 using a technical security analysis technique.
In their study, the MQoS and security flaws of time-based, contention-based, and hybrid protocols are compared. In WBANs, which could
be compromised by a cyberattack, they have taken into consideration delay, packet loss, and energy usage as performance evaluation
criteria.
Green dependability algorithm for the Transmission Control Protocol/Internet Protocol (TCP/IP protocol) in fog computing was created by
Mahmoodi Khaniabadi, et al.30 The created algorithm does not call for significant modifications to the TCP/IP protocol or necessary hardware. It
was built on sending fewer packets over the network by taking use of TCP and User Datagram Protocol (UDP) differences. User Datagram Pro-
tocol (UDP) and Transmission Control Protocol (TCP) vary in that UDP sends and receives half as many packets as TCP does overall. As a result,
UDP had half as many full packets as TCP did. Their approach was developed so that minor packet losses in applications like voice and online
video did not significantly impair the final product. Therefore, in certain circumstances, the UDP protocol might take the place of TCP. The mini-
mum acceptable Quality of Service (QoS) of the entire network serves as the basis for choosing between the two. So long as the QoS standards
are met, the UDP protocol will be used. The dynamic switching between UDP and TCP was made more efficient by gauging the current level of
network noise.
Gopika et al.31 have described energy efficient protocols and mechanisms, low power and loss networks (RPL), power generation, bio-inspired
routing, ambiguous logic approaches, and IPv6 routing protocols for sustainable approaches. The research community is constantly striving to
understand and address the challenges of achieving the highest version of this technology.
Recent studies have focused mainly on the use of energy from sensor nodes to increase sensor life. Seyyedabbasi et al.32 have introduced
a new routing procedure founded on ant colony optimization for multiple managers that properly manage live network possessions. Opera-
tors use special methods to determine the ants’ next target and to control pheromone renewal and evaporation. This technique receipts into
explanation several important limitations when selecting a future destination, such as future destination power, buffer size, travel speed, and
distance in different situations. The method finds the best route with low power ingesting and prolongs the life of the network. Kaur et al.33
have examined the latest steering protocols in sensor networks and develops exploit strategies for different approaches. The design of rout-
ing protocols is one of the most promising of these systems, showing the best features needed to send information. This paper instigates
with a thorough explanation of the work connected with the Foundation. In adding, this study presents new routing protocols to upsurge
the effectiveness of touch plans connected to the Internet. The comparative analysis with drawbacks of state-of-art schemes are summarized
in Table 1.
3PROBLEM STATEMENT AND NETWORK MODEL
3.1 Problem statement
WSN-IoT faces a number of challenges, including network performance, connectivity, security, integration, and energy efficiency which mainly
cause of coverage hole. The coverage hole is a location in the WSN-IoT where no data identification or transmission is available, so identi-
fying such a site can be difficult. Once the holes are found, another important task is to determine the edges of the border that reflect the
cover holes. WSN-IoT has problems with limited battery capacity, intermittent disconnection due to multi-jump connection, and short transmis-
sion range. IoT has been used to build mechanization, smart cities, smart farms, smart homes, smart marketing organization, and shrewd grid
schemes. Grouping and routing are done separately using different solutions, so efficient solutions cannot be provided in footings of energy inges-
tion and network generation. Now, gathering large amounts of information from a network of limited resources is complex process with many
challenges. Traditional network protocols offer a number of network algorithms that used without changing the Internet of Things operating
system.
Resource-limited WSN-IoT networks can further improve the data operatingsystem, including single-event reliability challenges and maximum
latency for multicast.34 In addition, the sensor terminals are battery operated and can be used in adverse conditions, and the sensor terminals can-
not be charged or replaced after use. In this case, the sensitivity energy of the sensor block is an important factor in the sensitivity field. Strong
routing protocols are required to increase network-operating hours and balance power consumption. Recent activities have focused on several
routing criteria that parents have used to determine the most effective route. This will take into account the dimensions of the work source, resid-
ual power, battery reduction code, serial usage, expected gear ratio, road maintenance, congestion, and barriers to network expansion in a variety
of situations. However, standard routing standards apply to specific applications that support routing. In addition, the detachment of the terminals
from the improper position and the remaining range of power contacts are important factors in selecting a terminal as a relay terminal. This causes
hotspots on the network when the relay cluster head fails to transmit during transmission. The proposed routing scheme is not only increase the
network’s lifespan but also establish a high packet delivery rate, so as to raise the count of live nodes in the region and to manage optimized data
transmission.
SRI VA S TAVA AND PAULUS 5of21
TA B L E 1 Summary of research gap
Ref Protocols Methodology Parameters Remarks
21 Cluster based routing
(CBR) protocol
Black widow
optimization (BWO)
Network lifetime and energy
efficiency
Performance of the CBR-ICWSN models need
to be enhanced using machine learning
techniques
22 Orphan-LEACH
(O-LEACH)
protocol
SA-LSA, PSO-LSA
Algorithm
Lifespan Need enhancement of protocol
23 OT-PRESENT
protocol
Co-FitDNN Network lifetime,
throughput, packet
delivery ratio, delayand
encryption time
Need to extend work with more than one sink
nodes to further reduce the energy
consumption
24 Multicast routing
protocol
Centralized multicast
routing
Number of Nodes To make changes to their scheme to allow for
node mobility. The following two factors must
be considered in order to support mobility.
25 ETERNAL protocol Cost-effective Accuracy Because homogeneous data is collected,
integrated, and processed into a single group,
a dynamic route allows a good load balance
and controls the energy required for these
nodes.
26 Neuro-Fuzzy based
protocol
New member joining Energy consumption Their work allowed to improve routing
protocols by using new trust mechanisms to
ensure efficient and secure routing.
27 CASER protocol Energy-efficient Network size They are seeking for more energy efficient
approaches to tackle SLP problems founded
on the various basis nodes or mobile
synchronization nodes.
28 ZigBee protocol Artificial intelligence tree
routing based on RNN
Accuracy Need extension of the work
29 MAC protocols,
hybrid protocols
Technical security
analysis approach
Delay, packet loss, and
energy consumption
The criteria in WBANs, which may be degraded
under a cyber attack
30 Transmission control
protocol/Internet
protocol (TCP/IP),
user datagram
protocol (UDP)
Green reliability
algorithm
Accuracy Need extension of work
31 Cluster based
protocol
Mobility aspect Adaptive technique Their method has been evaluated in multi-hop
networks to compare several matrices with
hybrid energy efficient distributed cluster
methods (HEED) such as network latency and
throughput.
32 CDXY protocol Single-based Accuracy It could be utilized in a variety of real-world
situations and applications that generate and
analyze enormous amounts of data.
33 SAWQE protocol Encryption and isolation Efficiency Increase in lifespan of the network.
3.2 Objectives
The main objectives of proposed coverage hole aware optimal cluster based routing (CHOCR) are:
1. To introduce a novel optimization technique for efficient balanced cluster framework this enhances the coverage hole aware data transfer.
2. To develop a new decision making technique to subtracttrust value of every IoT node using multiple restraints in cluster and consider the highest
trusted node is act as CH.
6of21 SRI VA S TAVA AND PAULUS
3. To prolong the quantity of alive sensor nodules over time with the help of multi-objective optimal network monitoring technique.
4. To propose new machine learning technique for intermediate node selection to frame the routing between source to destination.
3.3 Network model
Figure 1shows the network model of proposed CHOCR scheme which consist of three fold process are cluster formation, CH selection and neigh-
boringnodeselectionfor route computation. The network modelillustrateshow the cluster formation isdone,CHis select, and optimal pathisselect.
Our proposed CHOCR scheme should maximize throughput and minimalize energy feasting and recovers the coverage-hole detection through bal-
anced clustering and constructs multiple paths based on hybrid deep learning technique. Source nodes maynow pick a data transmission route from
a variety of options thanks to the routing protocol’s selection-based approach.
4PROPOSED METHODOLOGY
In this segment, we describe the working function of proposed CHOCR scheme which consists balanced clustering, CH selection and route
computation. Figure 2shows the workflow of proposed model.
4.1 Load balanced clustering using modified Lichtenberg optimization algorithm
In WSN-IoT, cluster formation used to solve many barriers to accessing the network by modifying the structure of sensor nodes. The main goals of
clustering methods are high-energy efficiency, bandwidth reuse, goal tracking, data collection, and network life. Finally, clustering is an important
energy saving routing technology. There are several reasons why a cover hole may form. The coverhole is created by the irregular light of the nodes,
incorrect network topography, node movement, friendly atmosphere, and energy drainage. Therefore, it is very important to find such cover holes,
FIGURE 1 Network model of proposed scheme
SRI VA S TAVA AND PAULUS 7of21
FIGURE 2 Workflow of proposed model
as the presence of holes can reduce the performance of WSN-IoT in the event of poor contact, excessive power consumption, and loss of valuable
data. The coverage hole plays a key role in detecting WSNs and other applications due to the potential understanding of coverage and errors in the
communication range. Holes in WSNs are rarely avoided because of the different geographical environment of the monitoring area, such as pools,
obstructions,orphysicaldamage to nodes. Ignoring whole detection affects geo-rootperformance,dataoverload, andexcessive power consumption
of hole boundary nodes. In addition, the flow of information may affect the overall capacity of the diaphragm network. These transformational
algorithms are inspired by natural events and used to solve real-world engineering issues. They are neither good nor terrible, however they are only
more efficient for a certain task. This means that the Lichtenberg optimization technique may be used to tackle problems that havemany objectives.
We further modify Lichtenberg optimization algorithm35 to make efficient prediction algorithm that is, modified Lichtenberg optimization (MLO)
which is very suitable for coverage-hole detection in our case.
Modified Lichtenberg optimization (MLO) was used to determine and characterize the crack distribution in thin plates of composite materials.
With just four and eight complex variables to work with, the system successfully recognizes the existence, direction, and intensity of marginal and
central fracture amplitudes. The diffusion limited aggregation (DLA) theory serves as the numerical foundation for the algorithm’s development.
Binary matrix (0 and 1) is laid up as a graph with a central integer denoting a particle constant. The cluster is arranged using matrix values with zero
values for single and empty spaces for the team. Initially, each value in the range is specifiedas a cluster particle and their number (Np) in the cluster.
A bitmap picture may be used to depict this matrix form (black and white). Picture production begins with two-dimensional matrices whose row and
column numbers are all identical to the image creation radius (Rc) (diameter). Welded beam design, labor necessary for the four variables, and the
non-linear control of the chosen test issue the width and length of the welded area, and the depth and thickness of the beam. The goal is to reduce
the total output cost under appropriate control of shear stress (τ), bending stress (σ), buckling load (q), and maximum deflection (δ). According to
Rao (2009), mathematical modeling Equations (3)and(4) represent a problem and provide a corresponding objective function.
where
y1=g,y2=l,y3=sand y4=a
min F(y)=1.1047y2
1y2+0.04811y3y4(14 +y2)(1)
Subject to
h1(Y)=𝜏(Y)−𝜏max ≤0
h2(Y)=𝜎(Y)−𝜎max ≤0
8of21 SRI VA S TAVA AND PAULUS
h3(Y)=y1−y4≤0
h4(Y)=1.1047y2
1+0.04811y3y4(14 +y2)−0.5 ≤0
h5(Y)=0.125 −y1≤0
h6(Y)=𝛿(Y)−𝛿max ≤0
h7(Y)=q−qd(Y)≤0
0.1 ≤yj≤2.0,j=1,4
0.1 ≤yj≤10.0,j=2,3
𝜏(Y)=(𝜏′)2+2𝜏𝜏′y2
2r+𝜏N2(2)
As long as the MLO is coded this way, it will always have a gap between points, resulting in errors in the optimizer and reducing its overall accuracy.
Using the modified Lichtenberg optimization (MLO) technique, Algorithm 1 explains how cluster formation works.
𝜏′(Y)= q
y1y22
𝜏N(Y)=Mr
Im
Then compute optimal solution as follows:
m=ql+y2
2(3)
r=y2
2
4+y1+y3
22(4)
i=2y1y2
2y2
2
12 +y1+y3
22 (5)
𝛿(Y)= 4ql3
ey3
3y4
(6)
𝜎(Y)= 6ql3
y2
3y4
(7)
qd=4.013
l21−y3
2le
4H
eH y2
3y6
4
36 (8)
If the MLO is programmed exactly like this, there will always be a gap between one point and the other, which will cause gaps when used in the
optimizer and compromise the optimizer accuracy. Algorithm 1 describes the working function of the cluster formation using modified Lichtenberg
optimization (MLO) algorithm.
Algorithm 1. Clustering using MLO algorithm
Input: Node location, position, and movement direction
Output: Optimal solution of location, position, and movement direction
1 Set objective function and search space
min F(y)=1.1047y2
1y2+0.04811y3y4(14 +y2)
2if
M=2, weight LF, end if
3if
M=0, Generate a LF, end if
SRI VA S TAVA AND PAULUS 9of21
4Ifref=0
h1(Y)=𝜏(Y)−𝜏max ≤0uptoh7(Y)=q−qd(Y)≤0
5 Apply random scale and rotation
h1(Y)=𝜏(Y)−𝜏max ≤0uptoh7(Y)=q−qd(Y)≤0
6 Calculate the fitness y(j)
7End
4.2 Trust computation and CH selection using LEO-DM technique
After cluster formation, we compute the cluster head (CH) using the following design parameters. The energy consumption of IoT sensor node is
resulting from basic liveliness model, which deliberate both the spreader and headset part energy necessities. The energy ingesting of a node is
comparativeto square of distance D2when the propagation distance (D)less than the threshold distance (D0), otherwise it is proportional to D2.
The total energy ingesting of each node in the network for transmits and receives nbit data packet.
Etotal =T(n,D)+R(n)(9)
where T(n,D)and R(n)are energy ingesting of conveying and getting node.
T(n,D)=
n×Eelec +n×𝜀fs ×D2;ifD<D0
n×Eelec +n×𝜀mp ×D4;ifD≥D0
(10)
R(n)=n×Eelec (11)
where Eelec the energy is degenerate per bit to run the spreader or headset circuit, intensification liveliness for free space model 𝜀fsand for
multi-path model 𝜀mpdepends on the spreader speaker model and D0is the beginning broadcast aloofness. The received signal strength (RSS) is
resolute by the detachment and programmeenergy, if the IoT sensor node conducts packet with energy T(n,D), the nodes customary signal strength
RSS, with the detachment of D, can be articulated as follows:
RSS =T(n,D)
4𝜋D2
i
+Ta,a1∕a2(12)
Movement, distance, and relative velocity are determined by the signal strength of the current sample, and select sample points that exceed the
limit, Δt1=Δt2=Δtbut such points do not exist in the sample domain. Where Di1,Di2and Di3denotes a changed detachment is subtracted from
the cosines laws as follows:
D2
i1=D2
i2+a1a2
2−2Di2⋅a1a2⋅cos(𝛼)(13)
D2
i3=D2
i3+a1a2
2−2Di2⋅a1a2⋅cos(𝛽)(14)
The present place of node is a,and can move to a1and a2in two position points individually. Contemplate cos(𝛼)=−cos(𝛽)and simply above
reckoning to calculate velocity (v)as follows:
2a1a2
2=D2
i1+D2
i2−2D2
i3(15)
v=
2D2
i1+D2
i2−2D2
i3
2Δt(16)
The undertaking period for IoT sensor node from present location ato the stimulated position a1or a2is uttered as the aloofness Ta,a1∕a2alienated
the node’s velocity and it can find by sign law as follows:
Ta,a1∕a2=R⋅sin 𝜗
sin 𝛽⋅v(17)
10 of 21 SRI VA S TAVA AND PAULUS
Finally, we frame the mobility (M) model for IoT sensor node as follows.
M=Ta,a1∕a2=Δt⋅R⋅sin 𝜗
sin 𝛽⋅D2
i1+D2
i2−2D2
i3
2
(18)
The number of nodes in the network with the neighbors set to the center is used to determine the traffic rate (T) between two points in time (t). For
base station time (t), this complaint was used to deduct from a node’s total traffic rate (BS).
T(i,t)=dis tan ce xi(t),BS(t)(19)
Objective and control are integrated into a set of equations or inequalities in the mathematical optimization paradigm. Engineering design and
financial portfolio selection are two domains where update models are often employed.
4.3 CH selection using LEO-DM technique
Linear equilibrium optimization (LEO) is a meta-heuristic optimization approach that aims to maintain a fair balance between research and opera-
tional phases by using LEO.As a result, local optimism is being depleted and search quality is being improved by using local space in pursuit of a good
answer. The dynamic mass balance of a control module system served as an inspiration for the LEO algorithm. As a preliminary step, the first order
is the normal difference, expressing the general mass equation, where the change in time-mass is equal to a sum of the mass entering and departing
the system.
Udq
ds =PQeq −PQ +H(20)
Linear equilibrium optimization (LEO) begins with an initial population depending on the number of particles and the size of the characteristic
dimension, as is the case with all meta-heuristic algorithms. The initial random population is shown in the equation below.
qinitial
j=qmin +randj(qmax −qmin)(21)
where qinitial
jcharacterizes the initial attentiveness vector of the jth particle and qmin and qmin are the minimum and the maximum concentration of
particles respectively, and randjgoes to [0, 1] and n is the number of the subdivisions in the populace. The optimization process comes to a conclusion
when the global equilibrium level is improved, and no knowledge of optimization is gained during the earliest stages of optimization. Four particles
are assumed to be ideal throughout the optimization process. The average of these four candidates should be included as a fifth candidate. When
compared to other optimization procedures, the number of particles that are picked has no fixed value. Here are five items that maybe used to build
avector.
−−−−→
qeq.pool =−
−−−→
qeq.(1),−−−→
qeq.(2),−−−→
qeq.(3),−−−→
qeq.(4),−−−−−→
qeq.(avg)(22)
It is then that exact vocabulary seeks balance between study and exploitation, and finally strives for greater optimization by using the exponential
term (e). Because the crowd density changes over time, the next illustration shows a random vector with a range of [0, 1].
e=E→
∈(s−s())(23)
In this case, where, s diverges with the difference of the iteration I which is defined as:
s=1−i
max jK2⋅i
max j(24)
In the above equation, jrepresents the current frequency and the maximum frequency K2. It is a variable that controls the ability to develop the
ability to exploit. It is easy to increase search speed by enhancing search and exploitation capabilities, as shown by the following equation.
s() =1
∈ln −K1sign(M−0.5)1−e−∈.s+s(25)
SRI VA S TAVA AND PAULUS 11 of 21
Where, K1indicates the ability to learn. Its value increases further K2exploitation and decreases searches. To ensure a successful working phase,
the generation factor plays a crucial role in providing the right solution to the optimization task. The well-known 1-d space model is one of several
models for estimating production rates.
GH=
G() ⋅E→
∈(s−s())(26)
where, G() is the original value and ∈is the decay continuous. To achieve a more symmetric search decoration and measured outcome are modified
as follows:
GH=
G() ⋅
E(27)
E−0=HDQ(Qeq−∈Q)(28)
HDQ =
0.5 ⋅MifM>HQ
() else
(29)
As a group control parameter, HDQ refers to a potential generation time during an upgrading process. Last but not least, the LEO upgrading
legislation must be taken into account.
Q=qeq +Q−qeqe+f
∈U(1−e)(30)
The Algorithm 2 describes the working function of CH selection using linear equilibrium optimization based decision making (LEO-DM) technique.
Algorithm 2. CH selection using LEO-DM technique
Input: Widespread feature space, populace size, max iteration
Output: Trust degree of IoT nodes
1 Initialize the particle’s populace
k1=2, k1=1, GP =.05
2 while j<max jdo
3for
l=1,…number of particles (n) do
4 Calculate fitness particle
5endif
6endfor
7qinitial
j=qmin +randj(qmax −qmin)
8 Equilibrium pool, −−−−→
qeq.pool =−
−−−→
qeq.(1),−−−→
qeq.(2),−−−→
qeq.(3),−−−→
qeq.(4),−−−−−→
qeq.(avg)
9 Assign s=1−i
max jK2⋅i
max j
10 Choose a applicantrandomly from the symmetry pool
11 Produce accidental number ∈and M
e=K⋅sign(M−0.5)e−∈.s−1
12 Concept HDQ =0.5 ⋅MifM>HQ
() else
13 Concept
GH=
G() ⋅
E
14 Update concentration Q=qeq +Q−qeqe+f
∈U(1−e)
15 End
4.4 Intermediate node selection for routing using hybrid deep recurrent neural network
4.4.1 Hybrid deep recurrent neural network
LSTM layers and conventional neural network layers are used to extract information from a wide variety of sensory inputs in a hybrid deep neural
network. In addition, the specific approach achieves sensor integration at the data level by combining data from multiple sensors. Both types of
12 of 21 SRI VA S TAVA AND PAULUS
networks exhibit transient dynamic behavior. A hybrid deep recurrent neural network (HD-RNN) is an activated acyclic map that has been modified
and replaced with a rigid leading neural network. A typical continuous neural network consists of an input layer, a hidden layer, and an output layer.
Enter the time range [y=y1,y1,…y1]. At time s, the output gsof the hidden layer as well as the xsof the output layer of HD-RNN are designed as
follows:
gs=fZgggs−1+Syg ys+ag
xs=𝜎Zgygs+ax
where gsis the output of the unseen coating at time s, xsis the output of the output layer, F() is the beginning meaning of the unseen coating, σ
is the beginning meaning of the output layer, xsis the input at time s,gs−1is the output of the preceding hidden layer, and Zgg ,Syg are the weight
matrix conforming to the output of the preceding hidden layer and the weight matrix agreeing to the input at time s, correspondingly. Why is the
weight matrix conforming to the output layer, and ag,axare the aberrations conforming to the hidden layer and the output layer, correspondingly.
An HD-RNN repeater module has just one part that has to be covered during assembly. The HD-RNN suffers from a gradient vanishing issue in the
back broadcasting process. So the LSTM network is designed to address HD-current RNN’s issues. The information that has to be eliminated from
the cellular state is computed in the hidden stage. For storage, materials are pre-calculated at this step. The “output gate” determines the cell state
we famine to output, which is the basis for starting the output. Input and the hidden layer are shown in the following manner:
Fs=𝜎ZgFgs−1+ZyF ys+aF(31)
js=𝜎Zgjgs−1+Zyj ys+aj(32)
Os=𝜎Zgogs−1+Zyo ys+ao(33)
Ds=tan gZgdgs−1+Zyd ys+ad(34)
where
Ds=F∗
sDs−1+j∗
s
Ds
gs=O∗
stan g(Ds)
xs=𝜎(Zxgs+ax)
where ysis the input arrangement at time s; gs−1is the output of the LSTM system cells at time s−1; ZyF,Zyj,ZyO ,ZgF,Zgj ,ZgO,Zgd ,Zyd and Zxrepresent
the corresponding weight matrixes; aF,ajand aOare the mistakes of the overlooking gate, input gate and output gate, correspondingly; by is the
calculation deviation; gs, it, and Osare the states of the overlookinggate, input gate and output gate correspondingly; and 𝜎() is the sigmoid beginning
function.
Dsis the impermanent state of the input at time s, Dsis the present cell state, tan g() is the tan gactivation function, gsis the current cell
output, xsis the fore seen charge at time s,and“∗” denotes the Hadamard product. There are only “update gates” and “reset gates” in the GRU
network, which is a very prevalent irregular of LSTM at current. WsSignifies the “update gate”, and Wsdetermines the amount of state information
gs−1from the previous moment that is transferred to the current state gs. The superior Wsis, the more material from the preceding second that is
approved into the present state. Rssignifies the “reset gate”, which panels the grade of inspiration of the preceding state gs−1on the contender state
gs. The advancing spread function of HD-RNN is updated as follows:
Rs=𝜎ZgRgs−1+ZyR ys+aR(35)
Ws=𝜎WgWgs−1+WyW ys+aW(36)
gs=tan gZyggs+Zhh gs−1+ad(37)
gs=(1−ws)∗gs−1+w∗
s
gs)(38)
where xs=𝜎(Zxgs+ax)and “∗” represents the Hadamard product, and ZyR,Zyw ,andZycharacterize the weight mediums of “input state”. ZgR,Zgw,
and Zgg characterize the weight matrixes of “state”. Zxrepresents the heaviness medium from the hidden layer to the output layer. aR,aw,
agand ax
SRI VA S TAVA AND PAULUS 13 of 21
are the balances. 𝜎() and tan g() characterize the sigmoid and tan gbeginning meanings, respectively. Algorithm 3 describes the working function of
intermediate node selection for routing using HD-RNN.
Algorithm 3. Intermediate node selection for routing using HD-RNN
Input: Time s, ZyR,Zyw ,andZy
Output: gs,xsZx(hidden layer)
1 Initialize the values for the input parameters
2 Set a sample training as
n=y(j),x(j),j=1,2,. …,n
3 Determine the LSTM network using
Fs=𝜎ZgFgs−1+ZyF ys+aF
4 Determine the GRU network using
Rs=𝜎ZgRgs−1+ZyR ys+aR
5 Modify the weight using
gs=tan gZyggs+Zhh gs−1+ad
6 Modify the offset using
xs=𝜎(Zxgs+ax)
7End
5RESULTS AND DISCUSSION
In this result, we validate our proposed Coverage Hole aware Optimum Cluster based Routing (CHOCR) scheme through two different simulation
scenarios. The simulations are conducted using NS3.26 simulator with two cases are impact of sensor nodes and impact of rounds.
5.1 Simulation setup
The summary of proposed routing scheme is summarized as follows: MLOalgorithm is used for balanced clustering which improve the performance
ofcoveragehole.TheLEO-DM technique is used to subtracttrust value of everyIoTnode using manifold restraintsinclusterand consider the highest
trusted node is act as CH. After that, HD-RNN is used for intermediate node selection to frame the routing between two nodes. The reproduction
strictures are described in Table 2. We examined IoT nodes and distributed them at random inside an 800×600 m2network space as part of our
TABLE 2 Simulation setup
Parameters Val ues
Simulation time 100 s
Node initial energy 100 J
Node transfer power 1.4 W
Node receiving power 1.0 W
Transmission range 250 m
MAC protocol IEEE 802.15.4
Traffic type CBR
Node density 200–1000
Data size 88 Mbps
Node mobility 20 m/s
Mobility model Random waypoint
Network size 800 ×600 m2
Number of simulation rounds 10, 20, 30, 40 and 50
14 of 21 SRI VA S TAVA AND PAULUS
conceptual design. The Euclidean distance between nodes is what separates them. IEEE 802.15.4 MACprotocol and CBR traffic type were selected.
Depending on your needs, we have IoT nodes that range from 50 to 250 and can send data at a rate of up to 88Mbps per node. Sensor nodes have
an initial transmit and receive power of 1.4 and 1.0 W, respectively. Each node in the network consumes 100 joules of energy during startup. Each
sensor node has 512 bytes of data storage. The proposed CHOCR scheme’s simulation results are compared to those of the state-of-the-art Repair,
HWSN, and MO-EPO systems34 in terms of energy consumption, network generation, the number of active nodes, packet delivery ratio,packet loss
ratio, throughput, end-to-end latency35 and delay.
5.2 Varying node density
In this case, the number of nodes ranges from 20 to 100, with increments of 20, 40, 60, 80, and 100, and simulation rounds set at 50. The proposed
CHOCR scheme is associated with the existing state-of-art Repair, HWSN, and MO-EPO schemes. Figure 3demonstrates the residual energy of
proposed and existing state-of-art schemes. The plot portrays the residual energy of proposed CHOCR scheme is 18.934%, 12.501% and 6.320%
higher than the existing Repair, HWSN and MO-EPO schemes respectively.
Figure 4displays the network generation of proposed and prevailing state-of-art schemes. The plot clearly represents the network life-
time of proposed CHOCR scheme is 13.858%, 8.044% and 3.850% higher than the existing Repair, HWSN, and MO-EPO schemes individu-
ally. Figure 5shows the quantity of alive nodes of proposed and existing state-of-art schemes. Compared to the current Repair, HWSN, and
MO-EPO systems, the graphic clearly shows that the number of active nodes in the projected CHOCR scheme will be 30.385%, 18.712%,
and 7.560% greater. Proposed vs present state-of-the-art delivery ratios are shown in Figure 6. The plot obviously represents the packet
delivery ratio of proposed CHOCR scheme is 18.699%, 10.605% and 4.979% higher than the existing Repair, HWSN, and MO-EPO schemes
respectively.
FIGURE 3 Residual energy with varying nodes
FIGURE 4 Network lifetime with varying nodes
SRI VA S TAVA AND PAULUS 15 of 21
FIGURE 5 Number of alive nodes with varying nodes
FIGURE 6 Percentage of packets delivered with varying nodes
FIGURE 7 Ratio of packet loss when nodes change
Figure 7displays the packet loss ratio of proposed and existing state-of-art schemes. The plot evidently represents the packet loss
ratio of proposed CHOCR scheme is 66.200%, 61.481% and 53.236% lower than the existing Repair, HWSN, and MO-EPO schemes
respectively.
Figure 8demonstrates the end-to-end delays of proposed and prevailing state-of-art schemes. The plot obviously depicts the end-to-end
delay of proposed CHOCR scheme is 42.632%, 33.995% and 21.963% lower than the existing Repair, HWSN, and MO-EPO schemes
respectively.
The throughput achieved by the suggested CHOCR in comparison to earlier research works is shown in Figure 9A,B. The amount of data
that is successfully transferred over a given period of time to a particular destination is known as throughput. In WSN-IoT, it successfully
16 of 21 SRI VA S TAVA AND PAULUS
FIGURE 8 End-to-end delay with variable nodes
FIGURE 9 (A) Analysis on throughput based on network size. (B) Analysis on throughput based on packet size
SRI VA S TAVA AND PAULUS 17 of 21
estimates the volume of data that the sink node has received. According to the analysis, throughput increases as network size increases while
decreasing as packet size increases. However, CHOCR outperforms with other works in both situations. The suggested study addresses QoS,
energy efficiency, and security as the three main WSN-IoT objectives, which is the primary cause of this significant variance. Therefore, the pro-
posed work produces a higher throughput. With regard to both network size and packetsize, the proposed CHOCR achieves throughput greater
than 90% during the specified time. CHOCR achieves throughput when n=100, whereas the HWSN have 65% throughput. This is 34% less
than the CHOCR. That is, the earlier works cannot consistently perform well. However, the CHOCR can outperform any given situation of a
time frame.
The earlier works had substantially lower throughput when the packet size is high. The HWSN approach specifically offers a throughput of
55%. The primary cause is the previous approaches could not handle huge packet sizes. For instance, HWSN takes a less-than-ideal channel for
data transmission, CNN uses more energy, QoS routing uses more bandwidth, and Secure WSN-IoT uses both time and energy. As a result, the past
research studies are inappropriate for a WSN-IoT context with limited resources. The proposed CHOCR increases throughput while also enhancing
QoS, security, and energy efficiency. The suggested CHOCR is hence appropriate for the WSN-IoT scenario.
5.3 Varying simulation rounds
With a fixed IoT sensor node of 100, the number of simulation rounds maybe varied between 10, 20, 30, 40, and 50. Existing state-of the art Repair,
HWSN and MOEPO systems are all linked to the proposed CHOCR scheme.
Figure 10 shows the residual energy of proposed and prevailing state-of-art schemes. The plot evidentlydepicts the residual energy of proposed
CHOCR scheme is 16.871%, 9.763% and 4.438% higher than the existing Repair, HWSN, and MO-EPO schemes individually. Figure 11 shows the
network generation of proposed and existing state-of-art schemes. The plot evidently depicts the network generation of proposed CHOCR scheme
is 11.135%, 6.895% and 3.988% higher than the existing Repair, HWSN, and MO-EPO schemes respectively.
FIGURE 10 Residual energy with varying simulation circles
FIGURE 11 Network generation with varying simulation circles
18 of 21 SRI VA S TAVA AND PAULUS
Figure 12 displays the amount of alive nodes of proposed and present state-of-art schemes. The plot obviously depicts the quantity of alive
nodes of proposed CHOCR scheme is 35.181%, 24.255% and 11.700% higher than the existing Repair, HWSN, and MO-EPO schemes correspond-
ingly. Figure 13 displays the packet delivery ratio of proposed and present state-of-art schemes. The plot obviously portrays the packet distribution
ratio of proposed CHOCR scheme is 24.991%, 15.559% and 7.884% higher than the existing Repair, HWSN, and MO-EPO schemes correspond-
ingly. Figure 14 depicts the packet loss ratio of proposed and current state-of-the-art designs. The CHOCR system is 56.083 percent less likely to
lose a packet than the current Repair, HWSN, and MO-EPO schemes respectively. The end-to-end delay of proposed and current state-of-the-art
FIGURE 12 Number of alive nodes with variable simulation rounds
FIGURE 13 Packet delivery ratio with varying simulation rounds
FIGURE 14 Packet loss ratio with varying imitation rounds
SRI VA S TAVA AND PAULUS 19 of 21
FIGURE 15 End-to-end delay with varying reproduction rounds
FIGURE 16 Throughput with variable simulation rounds
TABLE 3 Time and space complexity
Proposed algorithms Time complexity Space complexity
MLO O (n log (n)) O (log (n))
LEO-DM O(n) O(n)
HD-RNN O(nt*(ij +jk)) 2 n +1
CHOCR O (n log (n)) +O(n) +O(nt*(ij +jk)) O (log (n)) +O(n) +2n+1
methodsisshowninFigure 15. The proposed CHOCRsystemis39.040percent, 29.529%, and 18.073 percent fasterthantheexisting Repair,HWSN,
and MO-EPO schemes correspondingly in terms of end-to-end delays.
Figure 16 shows the throughput of proposed and existing state-of-art schemes. The plot evidentlydepicts the throughput of proposed CHOCR
scheme is 11.135%, 6.895% and 3.988% higher than the existing Repair, HWSN, and MO-EPO schemes respectively.
5.4 Time and space complexity
In MLO, the node of the packet is log (n) and for each node, the linear time taken is n. So the total time complexity of MLO is O (n log (n)). And the
space complexity of MLO is O (log (n)) because the node of the packet is log (n).
LEO-DM step takes time O(n), so the overall algorithm takes expected time O(n) and uses space O(n).
Fortraining a HD-RNN that has 4 layers with respectively i,j,knodes with ttraining examples and nepochs which has resultant Time complexity
O(nt*(ij +jk)). HD-RNN has space complexity 2 n+1. Table 3shows the Time and space complexities.
20 of 21 SRI VA S TAVA AND PAULUS
6CONCLUSION
In this paper, we have proposed a coverage hole aware optimal cluster based routing (CHOCR) scheme for WSN-IoT. The following research con-
tributions are achieved in this work. Modified Lichtenberg optimization (MLO) algorithm is proposed for balanced clustering which improved the
performance of coverage hole. Then, a linear equilibrium optimization based decision-making (LEO-DM) technique is proposed to multiply trust
value of each IoT node using numerous restrictions in cluster and consider the highest trusted node is act as CH. After that, a hybrid deep recur-
rent neural network (HD-RNN) is developed for intermediate node selection to frame the routing between two nodes. Finally, we simulated our
proposed CHOCR scheme using NS3.26 simulator and the performance analyzed through two simulation scenarios such as varying node density
and varying number of rounds. From the simulation results, we conclude that the presentation of proposed CHOCR scheme is very efficient than
the existing state-of-art schemes. It is found that CHOCR is 10.357% more efficient than current systems in terms of energy consumption and net-
work longevity and the number of nodes that are still alive in the network, as well as packet delivery ratio, packet loss ratio, throughput, end-to-end
latency and delay for simulated cycles. The proposed CHOCR scheme achieves 98% efficiency in residual energy, 95% efficiency in Network lifetime
with varying nodes. The end-to-end delay of proposed CHOCR scheme is 42.632%, 33.995%, and 21.963% lower than the existing Repair, HWSN,
and MO-EPO schemes respectively. At given period of time, the proposed CHOCR achieves throughput is higher than 90% with respect to both
network size and packetsize.
CONFLICT OF INTEREST
The authors declare no conflicts of interest.
DATA AVAILABILITY STATEMENT
Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.
ORCID
Abhishek Srivastava https://orcid.org/0000-0001- 7020-7654
REFERENCES
1. Watt AJ, Phillips MR, Campbell CE-A, Wells I, Hole S. Wireless sensor networks for monitoring underwater sediment transport. Sci Total Environ.
2019;667:160-165.
2. Koriem SM, Bayoumi MA. Detecting and measuring holes in wireless sensor network. J King Saud Univ Comput Inf Sci. 2020;32(8):909-916.
3. Qu Z, Huihui X, Zhao X, Tang H, Wang J, Li B. A fault-tolerant sensor scheduling approach for target tracking in wireless sensor networks. Alex Eng J.
2022;61(12):13001-13010.
4. YadavAS, Khushboo K, Singh VK, Kushwaha DS. Increasing efficiency of sensor nodes by clustering in section based hybrid routing protocol with artificial
bee colony. Procedia Comput Sci. 2020;171:887-896.
5. Farsi M, Elhosseini MA, Badawy M, Ali HA, Eldin HZ. Deployment techniques in wireless sensor networks, coverage and connectivity: a survey. IEEE
Access. 2019;7:28940-28954.
6. Morabito R, Jiménez J. IETF protocol suite for the internet of things: overviewand recent advancements. IEEE Commun Standards Mag. 2020;4(2):41-49.
7. Iqbal R, Doctor F, More B, Mahmud S, Yousuf U. Big data analytics and computational intelligence for cyber–physical systems: recent trends and stateof
the art applications. FutureGener Comput Syst. 2020;105:766-778.
8. Al-Qurabat AKM, Mohammed ZA, Hussein ZJ. Data traffic management based on compression and MDL techniques for smart agriculture in IoT. Wirel
Pers Commun. 2021;120(3):2227-2258.
9. Oliveira L, Rodrigues JJPC, Kozlov SA, Rabêlo RAL, Victor HC, de Albuquerque. MAC layer protocols for internet of things: a survey. Future Internet.
2019;11(1):16.
10. Jondhale SR, Maheswar R, Lloret J. Fundamentalsof wireless sensor networks. Received Signal Strength Based Target Localization and Tracking Using Wireless
Sensor Networks. Springer; 2022:1-19.
11. Najmi KY, AlZain MA, Masud M, Jhanjhi NZ, Al-Amri J, Baz M. A survey on security threats and countermeasures in IoT to achieve users confidentiality
and reliability. Mater Today: Proc. 2021;20(14):3814.
12. Sales O, Frederico YM, VieiraAB, Silva EF. Energy consumption evaluation of a routing protocol for low-power and lossy networks in mesh scenarios for
precision agriculture. Sensors. 2020;20(14):3814.
13. Tao K, Chang H, Jin W, Tang L, Miao J. MEMS/NEMS-enabled energy harvesters as self-powered sensors. Self-Powered and Soft Polymer MEMS/NEMS
Devices. Springer; 2019:1-30.
14. Zhuo X, Liu M, Wei Y, Guanding Y, FengzhongQ, Sun R. AUV-aided energy-efficient data collection in underwater acoustic sensor networks. IEEE Internet
Things J. 2020;7(10):10010-10022.
15. Ray PP, Dash D, De D. Edge computing for internet of things: a survey, e-healthcare case study and future direction. J Netw Comput Appl. 2019;140:1-22.
16. Gite P, Shrivastava A, Krishna KM, Kusumadevi GH, Dilip R, PotdarRM. Under water motion tracking and monitoring using wireless sensor network and
machine learning. Mater Today: Proc. 2021.
17. Omolara AE, Alabdulatif A, Abiodun OI, Alawida M, Alabdulatif A, Arshad H. The internet of things security: a survey encompassing unexplored areas
and new insights. Comput Secur. 2022;112:102494.
18. Bhatia V, Sunita K, Vivek J. Overview of the role of the internet of things and cyber-physical Systems in Various Applications. Handbook of Research of
Internet of Things and Cyber-Physical Systems. Apple Academic Press; 2022:3-30.
SRI VA S TAVA AND PAULUS 21 of 21
19. Poongodi T, AnuRathee RI, Suresh P. IoT sensing capabilities: sensor deployment and node discovery, wearable sensors, wireless body area network
(WBAN), data acquisition. Principles of Internet of Things (IoT) Ecosystem: Insight Paradigm. Springer; 2020:127-151.
20. Kumar S, Subhasri D, Rangan VP, Maneesha VR. Reliable network connectivity in wireless sensor networks for remote monitoring of landslides. Wirel
Netw. 2020;26(3):2137-2152.
21. Vaiyapuri T, Velmurugan Subbiah Parvathy V, Manikandan NK, Gupta D, Shankar K. A novel hybrid optimization for cluster-based routing protocol in
information-centric wireless sensor networks for IoT based mobile edge computing. Wirel Pers Commun. 2021;1-24.
22. Senthil GA, Raaza A, Kumar N. Internet of things energy efficient cluster-based routing using hybrid particle swarm optimization for wireless sensor
network. Wirel Pers Commun. 2022;122(3):2603-2619.
23. Sujanthi S, NithyaKalyani S. SecDL:QoS-aware secure deep learning approach for dynamic cluster-based routing in WSN assisted IoT. Wirel PersCommun.
2020;114(3):2135-2169.
24. Pan MS, Yang SW. A lightweight and distributed geographic multicast routing protocol for IoT applications. Comput Netw. 2017;112:95-107.
25. Devi MD, Geetha K, Saranyadevi K. Content based routing using information centric network for IoT. Procedia Comput Sci. 2017;115:707-714.
26. Thangaramya K, Kulothungan K, Logambigai R, Selvi M, Ganapathy S, Kannan A. Energy aware cluster and neuro-fuzzy based routing algorithm for
wireless sensor networks in IoT. Comput Netw. 2019;151:211-223.
27. He Y, Han G, Wang H, Ansere JA, Zhang W. A sector-based random routing scheme for protecting the source location privacy in WSNs for the internet
of things. Future Gener Comput Syst. 2019;96:438-448.
28. Sumathi AC, Javadpour A, Pinto P, Sangaiah AK, Zhang W, Khaniabadi SM. NEWTR: a multipath routing for next hop destination in internet of things
with artificial recurrent neural network (RNN). Int J Mach Learn Cybern. 2022;1-21.
29. Javadpour A, Sangaiah AK, Ja’fari F, et al. Toward a secure industrial wireless body area network focusing MAC layerprotocols: an analytical review. IEEE
Trans Ind Inform. 2022.
30. Khaniabadi M, Shadi AJ, Gheisari M, Zhang W, Liu Y, Sangaiah AK. An intelligent sustainable efficient transmission internet protocol to switch between
user datagram protocol and transmission control protocol in IoT computing. Expert Syst. 2022;e13129.
31. Gopika D, Panjanathan R. Energy efficient routing protocols for WSN based IoT applications: a review. Mater Today: Proc. 2020.
32. Seyyedabbasi A, Kiani F. MAP-ACO: an efficient protocol for multi-agent pathfinding in real-time WSN and decentralized IoT systems. Microprocess
Microsyst. 2020;79:103325.
33. Kaur L, Kaur R. A survey on energy efficient routing techniques in WSNs focusing IoT applications and enhancing fog computing paradigm. Glob Transit
Proc. 2021;2(2):520-529.
34. Jain JK. A coherent approach for dynamic cluster-based routing and coverage hole detection and recovery in bi-layered WSN-IoT. Wirel Pers Commun.
2020;114(1):519-543.
35. Pereira JLJ, Francisco MB, Diniz CA, Oliver GA, Cunha SS Jr, Gomes GF. Lichtenberg algorithm: a novel hybrid physics-based meta-heuristic for global
optimization. Expert Syst Appl. 2021;170:114522.
How to cite this article: Srivastava A, Paulus R. Coverage hole aware optimal cluster based routing for wireless sensor network assisted IoT
using hybrid deep recurrent neural network. Concurrency Computat Pract Exper. 2022;e7535. doi: 10.1002/cpe.7535