8/25/22, 12:05 PM View Letter

Date: Aug 23, 2022

To: "Dushmanta Kumar Das" dushmantakumardas29@gmail.com
From: "Computer Communications" support@elsevier.com
Subject: Decision on submission to Computer Communications

Manuscript Number: COMCOM-D-21-01580R2

Multiple Objective Optimization-based DV-hop Localization for Spiral Deployed Wireless Sensor Networks using Non-
inertial Opposition-based Class Topper Optimization (NOCTO)

Dear Dr Das,

Thank you for submitting your manuscript to Computer Communications.

I am pleased to inform you that your manuscript has been accepted for publication.

My comments, and any reviewer comments, are below.

Your accepted manuscript will now be transferred to our production department. We will create a proof which you will be
asked to check, and you will also be asked to complete a number of online forms required for publication. If we need
additional information from you during the production process, we will contact you directly.

We appreciate you submitting your manuscript to Computer Communications and hope you will consider us again for
future submissions.

Kind regards,
Carla Fabiana Chiasserini
Editor-in-Chief

Computer Communications

Editor and Reviewer comments:

Reviewer #1: The requested corrections have been made. The article is acceptable in this form.

Reviewer #3: After carefully checking, I confirm that the authors have completely addressed all my concerns, I therefore
recommend accepting this paper for publication

More information and support

FAQ: When and how will I receive the proofs of my article?

https://service.elsevier.com/app/answers/detail/a_id/6007/p/10592/supporthub/publishing/related/

You will find information relevant for you as an author on Elsevier’s Author Hub: https://www.elsevier.com/authors

FAQ: How can I reset a forgotten password?

https://service.elsevier.com/app/answers/detail/a_id/28452/supporthub/publishing/kw/editorial+manager/

For further assistance, please visit our customer service site:

https://service.elsevier.com/app/home/supporthub/publishing/. Here you can search for solutions on a range of topics,
find answers to frequently asked questions, and learn more about Editorial Manager via interactive tutorials. You can also
talk 24/7 to our customer support team by phone and 24/7 by live chat and email.

#AU_COMCOM#

To ensure this email reaches the intended recipient, please do not delete the above code

__________________________________________________
In compliance with data protection regulations, you may request that we remove your personal registration details at any
time. (Use the following URL: https://www.editorialmanager.com/comcom/login.asp?a=r). Please contact the publication
office if you have any questions.

https://www.editorialmanager.com/comcom/ViewLetter.aspx?id=383335&lsid={8D4BE095-F3AF-4895-868A-2EA7A619E09C} 1/1
REVISED Manuscript (text UNmarked) Click here to view linked References

Multiple Objective Optimization-based DV-hop

Localization for Spiral Deployed Wireless Sensor
Networks using Non-inertial Opposition-based Class
Topper Optimization (NOCTO)
Tapan Kumar Mohanta1 and Dushmanta Kumar Das2
1
Department of ECE, ICFAI University Kamalghat, Tripura,India.2 Department of
Electrical and Electronics Engineering, National Institute of Technology Nagaland,
India.(tapank838@gmail.com, dushmantakumardas29@gmail.com)

Abstract
The problem of localization is one of most important issues in wireless sen-
sor networks. Furthermore, it is critical to monitor and evaluate the data
gathered. For a variety of factors, such as upkeep, lifespan, and breakdown,
the fixed density of these beacons may be increased or decreased. Because
of its robustness, flexibility, and economic viability, a well-known technique
for locating wireless sensor network nodes is the Distance Vector-Hop (DV-
Hop) algorithm. As a result, researchers continue to look for ways to de-
velop it. A new Non-inertial Opposition based Class Topper Optimization
(NOCTO) based enhanced DV-Hop localization algorithm is proposed. It
also focuses through an optimized formulation to compute the average hop-
size with weight of beacon nodes in order to reduce the localization error with
estimated distance between the beacon and the dumb node, due to improved
localization accuracy. For spiral deployed 2D wireless sensor networks, this
paper proposes a multi-objective NOCTO-based DV-Hop localization. The
simulation results indicate that our suggested multi-objective function out-
performs some existing techniques.
Keywords: DV-Hop, Non-inertial Opposition based Class Topper
Optimization, Wireless Sensor Networks, Multiple objective

Preprint submitted to Computer Communication June 23, 2022

1. Introduction
Wireless sensor networks (WSNs) are becoming a common technology
that is attracting research attention. WSNs are made up of a large number
of small sensor devices called sensor nodes that collect and communicate data
from their sensing environment [1]. WSN localization is a critical problem
that holds the collected data more usable for a variety of applications. Fur-
thermore, without knowing the actual location within each sensor node, the
obtained data is meaningless [2]. Furthermore, the location of a sensor is just
as critical as the data itself. As a result, the data and position information
must be communicated jointly such that the precise location of the collected
information by the sensor node can be determined. Several applications
are designed to work in extreme environments, such as battlefields, border
surveillance, rain forests [3, 4], remote control of hazardous areas, routing
[5, 6, 7, 8], and so on. Although the Global Positioning System (GPS) is
commonly used for sensor localization, it is not a viable choice due to WSN’s
cost constraints. As a result, researchers have proposed a variety of WSN
localization algorithms [9].

There are numerous localization algorithms available today, that can be

classified into two parts: range-based and range-free algorithms [10]. To cal-
culate a node’s location, range-free localization algorithms estimate the dis-
tance within nodes and do not need outside hardware. Under these types of
algorithms, the closest nodes that are aware of their positions are used to lo-
calize dumb nodes. The centroid algorithm [11], DV-Hop [12, 13], amorphous
[14], APIT [15], and MDS- MAP [16, 17] are some of the most well-known
range-free localization algorithms in the literature.

Range-based localization algorithms rely on precise measurements meth-

ods that necessitate expensive equipment to determine the distance or angle
data between a node and also its counterpart, allowing for high-accuracy
location determination [18]. Received Signal Strength Indicator (RSSI) [19],
Time of Arrival (ToA) [20], Angle of Arrival (AoA) of signals packets [21],
and Time Difference of Arrival (TDoA) [22] are some examples of range-
based localization algorithms found in the literature. We’re interested in
range-free algorithms in our paper because they don’t require any additional
devices. Many existing research work focused on using single objective op-
timization to solve localization problems [23]. However, numerous studies

2
have demonstrated the importance of multiobjective optimization in solving
multiobjective complex issues [24]. Various biological techniques have been
presented in the past to reduce the localization error [25].

Being able to withstand dynamic changes conventional optimization ap-

proaches are not resistant to flexible changes in the environment, and pro-
viding a solution necessitates a complete restart. On the other hand, the
proposed algorithm can be used to adjust solutions for changing conditions.
It has numerous applications. The proposed algorithms can be applied to
any problem that can be expressed as an optimization problem. More con-
ventional optimization techniques can be combined with the proposed algo-
rithms. One of the advantages of proposed algorithm is the ability to solve
problems for which no human knowledge exists. Human knowledge should
be used when it is required and available. But, it is frequently insufficient
for automated problem-solving habits. Power system operations and control,
chemical processes, vehicle routing problems, mobile networking, batch pro-
cess scheduling, multi-objective optimization problems, model optimization
problems, image processing, and pattern recognition problems are all exam-
ples of applications where the proposed algorithm can be used.

The main motivation of the proposed method is to reduce energy con-

sumption, increase network lifetime, and improve network reliability. It also
significantly reduces computational time, network complexity, and node en-
ergy consumption. Other localization strategies produce accurate estimates,
but they are quickly influenced by the complexity of indoor environments’
dynamic changes. To accomplish this motivation, the proposed method or-
ganises sensor nodes in a spiral deployed in a wireless sensor network. In this
paper, we propose a multiobjective NOCTO based on a DV-hop localiza-
tion algorithm is proposed to improve localization accuracy. We considered
six single objective functions in this study, and the single-objective functions
were turned into multiobjective functions for greater accuracy. The proposed
algorithm reduces communication between dumb nodes and beacon nodes
by calculating the hop size of all beacons at dumb nodes. It significantly
reduces computational time while also lowering node energy consumption.
The simulation results show that our proposed multiobjective NOCTO al-
gorithm outperforms existing single- and multiobjective NOCTO algorithms.

The proposed method includes the following innovations:

3
. An optimization algorithm is proposed for making node energy consump-
tion more uniform, lowering energy consumption and increasing net-
work lifetime in spiral deployed 2D wireless sensor networks.

. Six single objective functions are used in the proposed method, which were
then transformed into multiobjective functions for better accuracy.

. The proposed technique reduces communication among dumb nodes and

beacon nodes by calculating the hop size of all beacons at dumb nodes.
It significantly reduces computational time and node energy consump-
tion.

. We adopt a new non-inertial opposition based class topper optimization

algorithm to correct the approximate locations of the dumb nodes,
thereby minimising the localization error.

. Experiment results show that our method not only significantly reduces
localization errors and increases network coverage, but it also signifi-
cantly improves accuracy.

The paper is organized as follows: Section 2 presents related works;

Conventional DV-Hop algorithm uses multiobjective NASG-II in Section 3.
Section 4 includes a Non-inertial Opposition-based Class Topper Optimiza-
tion (NOCTO) Algorithm and Multiobjective Functions. Section 5 explains
Performance Metrics. Results from the simulation are shown in Section 6.
Finally, in Section 7, conclusions are drawn.

2. Associated works

4
 Table 1: The disadvantages and shortcomings of the previous methods

Techniques Disadvantage shortcoming of algorithm

multiple objective It did not enhance the stability This study does not look into the impact of
NASG-II DV-Hop [26] period and increased the delay. radio irregularities.
Teaching Learning The effects of noise, signal fading, Although collinearity improves localization
based Optimization and mobility on the localization precision, Because it is measured for all bea-
(TLBO) [27] process are discussed in this pa- con nodes, the average hop distance error is
per. relatively large.
Grey Wolf Optimiza- High computational overload A different evolution algorithm is used to
tion (GWO) [28] solve the problem. The network design is a
little more complicated.
Runner-Root Algo- Data success rates are low when This paper does not demonstrate the effect
rithms (RRA) [29] reliability is high. of network irregularities on the localization

5
algorithm.
firefly algorithm and It had no effect on the stability All nodes are outfitted with GPS, which is
fuzzy logic [30] period and increased the delay. very expensive.
An energy-aware clus- High computational overload Sensor node energy resources are limited
tering and two-level
routing [31]
Dragonfly algorithm High overhead for updating the The impact of network abnormalities on the
[32] sink table and MS movement, se- localization algorithm is not demonstrated
lecting cluster heads close to the
sink and incorrectly distributing
them in the network.
 Table 1: continued

Techniques Disadvantage shortcoming of algorithm

A tree-based data ag- Node failure is not supported. The weighted coefficient is inaccurate due to
gregation [33] the variety of network topologies in networks,
contributing to the calculated distance’s in-
accuracy.
Secure Hybrid Struc- Scalability and mobility are not This paper does not show mobile sinks with
ture Data Aggregation supported. appropriate mobility models and a dynamic
(SHSDA) [34] structure to manage node additions and re-
movals.
Energy Efficient Lay- High hardware overhead There is no mention of proper mobility mod-
ered Routing Protocol els or using a dynamic structure to manage
(EELRP) [35] the added or removed nodes

6
a routing and data ag- High computational overhead In this paper, it is not clear how to update
gregation [36] the sink table and MS movement, as well as
how to select cluster heads near the sink and
distribute them improperly in the network.
Fractional order Class High computational complexity The implementation of beacon nodes at the
Topper Optimization sensor network’s boundary is insufficient for
(FCTO) [37] isolated areas in this paper
 Because of its simplicity, feasibility, and low hardware requirements, DV-
Hop is a common choice among WSN localization algorithms. However, the
precision with DV-Hop is insufficient to meet more stringent requirements.
Therefore, researchers have suggested a variety of improved DV-Hop-based
schemes. In this section, some enhanced algorithms based on multi-objective
DV-Hop are briefly mentioned.

In [26], a multi-objective DV-Hop localization algorithm based on NSGA-

II is suggested with an improved constraint strategy applied to all beacon
nodes to improve DV-Hop location estimation precision. It provides greater
accuracy at a small increase in computational costs. To use the Grey Wolf
algorithm to get a more accurate estimate of the average hop distance as
determined by each beacon node, Kaur et al. [28] suggested of grey wolf
optimization. The algorithm has been shown to provide greater precision
with a small increase in the cost of computing. The use of multi-objective
functions to solve difficult issues involving several objects and to keep track
of them in order to keep the model’s size down. However, the proportion of
selection is used to determine the global optimum solution. The influence of
radio irregularity is not examined in this work.

Kanwar and Kumar [29] have suggested a range-free localization using

runner-root algorithms. They change the average hop size of the anchor
nodes by refining a correction factor, and the changed hop size is further
optimised by a line search algorithm. The precision of localization is fur-
ther improved utilizing a runner-root method. The Teaching Learning based
Optimization (TLBO) method was used in [27] to improve range free local-
ization. The authors introduce a correction factor to alter the hop size of
anchor nodes, which is then further improved using a line search algorithm.
However, holes, non-uniform node distribution, network sparsity, and irreg-
ular radio patterns may reduce the suggested algorithm’s performance.

A multipath routing scheme for homogeneous WSNs is presented in [30].

The proposed routing method is intended to reduce energy consumption as
well as balance load, resulting in increased network lifetime. We also want
to lower the packet loss rate. Clustering network nodes, exploring paths
between CHs, and maintaining paths are the three phases of the proposed
routing method. A 2-level routing algorithm with two phases, clustering and
routing, is proposed in [31]. CHs are chosen during the clustering phase

7
based on residual energy, length from the BS, centrality, and the number
of neighbours. BCHs, as well as CHs, are chosen during this phase. Each
clustering-setup phase has two clustering-steady phases, as determined by
BCH selection. As a result, clustering’s overhead is minimised. Routing is
the second phase, which has 2 levels: intra-cluster routing and inter-cluster
routing. Each cluster is divided into 4 sections, each with a CH centrality,
and nodes are located in these 4 sections in intra-cluster routing. In inter-
cluster routing, CHs are layered according to their distance from the BS.

A safe data aggregation scheme is described in [32]. The goal of the data
patterns is to reduce redundant data in the network, decrease sensor node
energy consumption, and maintain data accuracy and precision. Intra-cluster
data aggregation, inter-cluster data aggregation, and data transfer are the
three phases of a secure data aggregation method. In [33], a binary tree is
used to organise the sensor nodes. After that, data aggregation is conducted
on the tree’s middle nodes, and an analysis of the data is sent to the root
node. They must first be approved in order to prevent unauthorised data ag-
gregation requests. The aggregation process then begins, with an enhanced
cyclic redundancy code (CRC) being used hop by hop to ensure data aggre-
gation accuracy and reliability. Meanwhile, the data packets obtained from
their children are subjected to cumulative functions imposed by intermediate
nodes. As a result, the workload of the sensors in the network is significantly
reduced. This pattern continues, and aggregated data is eventually sent to
the base station.

In [34], a secure combination data aggregation method called SHSDA is

proposed, which is based on a mixture of star as well as tree structures. The
network is divided into four sections, each of which is geographically split
into 4 equal parts. The network is divided into four sections, each of which
is divided geographically into four equal parts. A star structure is formed
among the nodes. The best node in terms of residual energy and centrality
is chosen as the root of a star structure in each section. An energy-efficient
layer routing protocol has been proposed by Hajipour and Barati [35]. This
technique divides the network into circles that are concentric. The circles
are then divided into eight equal sectors at a 450 angle. Then, for each sec-
tion, an agent is chosen. Each section’s agent is in charge of gathering and
aggregating the data sensed by the nodes in that section. When the agent
receives the data, it adds error detection and correction codes before sending

8
it to the lower layer of the same sector’s agent. The process is repeated until
all of the data has arrived at the base station.

In [36], a two-level energy-aware routing and a clustering method is pro-

posed. The network is clustered, and the most suitable CHs are chosen based
on the main factors related to energy consumption. A rendezvous region is
then built to create a communication substructure between CHs, and the
nodes in this region are referred as backbone nodes. Then, in this area, a
tree is created to send the aggregated data to the sink via CHs. This tree
provides a proper substructure for sending data to the sink, allowing CHs
to send data to the sink with least amount of overhead and energy, and in
a short amount of time. Because of the use of this microstructure, CHs do
not need to use route discovery to send data to the sink. Mohanta and Das
[37] developed a better DV-Hop algorithm based on the Dynamic Beacon
Node Set based on experiments (DBNS). The DBNS IDV-Hop steps 1 and
2 are identical to the DV-Hop steps. The DANS IDV-Hop construction’s
third step, a Dynamic Beacon Node Set (DBNS), uses the Binary Fractional
Order Class Topper Optimization (BFCTO) algorithm to optimise the dumb
nodes. Fractional Order Class Topper Optimization (FOCTO) is used in the
fourth step of the DBNS IDV-Hop to correct the localization effect.

The novelty of the paper is that a multi-objective non-inertial opposition

mechanism is introduced into CTO, resulting in a significantly different CTO
variant known as Non-inertial Opposition CTO (NOCTO). NOCTO, unlike
most CTO variants, is executed on a sorted student. Instead of learning from
the best positions in history, each student can then correct their behaviour
by learning from better students in the class (demonstrators, i.e., (st ) and
(ct )). To reduce the burden of parameter setting, a two-dimensional param-
eter control strategy has been proposed in the proposed NOCTO to improve
its robustness to the search dimension (the number of objective functions)
of the problem to be optimised. To the best of our knowledge, there is no
optimization algorithm that employs both an optimization model and an ad-
justment to the minimum hop as an enhancement for spiral deployed 2D wire-
less sensor networks for the DV-Hop algorithm at the heart of the research.
For improved estimation accuracy and coverage, we suggest a multi-objective
Non-inertial Opposition based Class Topper Optimization(NOCTO) DV-Hop
localization method for spiral deployed 2D wireless sensor networks with min-
imum hop optimization in this paper. Table 1 shows the disadvantages and

9
shortcomings of the previous methods.

3. Algorithm of conventional DV-Hop using multiobjective NASG-

II
The localization technique is essentially a free localization technique that
is based on a protocol for distance routing. For number of hops of beacon
nodes as well as the minimum hop distance besides WSNs, calculating the
distance between dumb nodes or unknown nodes and beacon nodes is used.
Different paths form in a network topology among dumb nodes and beacon
nodes that are not linear due to non-uniform connectivity with wireless sen-
sor nodes. Therefore, Some errors have been identified at the period of the
algorithm in the node position method [29].

Step 1 The minimum hop amount is defined for unknown nodes and bea-
con nodes of step 1. By transmitting signals through beacon nodes by
vector protocol system, the neighbouring nodes can be shown their lo-
cation. Information exists in the form of Hi , aj and bj in which id will
be the identity. aj and bj have been coordinates, and Hi has been the
hop count for the i beacon node. First, 0 is set of the value for Hi [38].
The nodes obtain data from the broadcast and record the hop amount
and localization of the vector’s beacon nodes. The value of Hi must
be increased by 1 through this process [27]. In this update process,
if any node receives the same id group, the new received data will be
compared with the original value of Hi . The nodes obtain broadcast
data and keep track of the hop amount and localization for the beacon
nodes of the vector.

Step 2 Minimum hop count and average hop distance are determined to
find the distance between unknown nodes and beacon nodes. In this
method, the average hop distance for the entire network can be deter-
mined by obtaining the position and hop amount for nodes of beacons,
as described in the previous stage. Then this data is transmitted to the
entire network. Furthermore, most nodes require average hop distance
information from a beacon node that is nearer to them [26]. The fol-
lowing equation provides the typical distance of the hop (jpi ) and the
hop range (pi ) between the i(ai , bi ) beacon node and the other (beacon)

10
 node j(aj , bj ) can just be computed as:
P q
i6=j (ai − aj )2 + (bi − bj )2
jpi = P , (1)
i6=j h(ij)

where jpi is the minimum hop distance and h(ij) is the number of hops
from beacon nodes i and j. The distance between beacon nodes and
dumb nodes is expressed in the following formula.

piU = jpiU × Hopmin , (2)

where jpiU is the step distance, Hopmin is the step count between the
i beacon nodes and piU denotes the estimated distance between the
dumb node U and the beacon node i.

Step 3 Let the U dumb node coordinates be (a, b), and the ith beacon node
coordinates be (ai , bi )(1 ≤ i ≤ n) analyzed in stage 3. Correspondingly,
the distance between the beacon node ith to the unknown or dumb
node U is pi (1 < i < n). The coordinates of the dumb node have been
determined as follows:

(a − a1 )2 + (b − b1 )2 = p21 ;
(a − a2 )2 + (b − b2 )2 = p22 ;
....................................... (3)
2 2
(a − an ) + (b − bn ) = p2n ;

a21 − a2n − 2(a1 − an )a + b21 − b2n − 2(b1 − bn )b

= p21 − p2n ;
..................................................................... (4)
a2n−1 − a2n − 2(an−1 − an )a + b2n−1 − b2n − 2(bn−1
− bn )b = p2n−1 − p2n ;

The matrix resulting from (4) is shown below:

SA = T, (5)

11
  
2(a1 − an ) 2(b1 − bn )
 2(a2 − an ) 2(b1 − bn ) 
S=
 .................................  ,
 (6)
2(an−1 − an ) 2(b1 − bn )
 
a21 − a2n + b21 − b2n + p2n − p21
 a22 − a2n + b22 − b2n + p2n − p22 
T = , (7)
 .......................................... 
2 2 2 2 2 2
an−1 − an − bn−1 − bn + pn − pn−1
 
a
A= . (8)
b

Finally, the least square approximation is used to calculate the un-

known node coordinates as given below:
−1
A = (S T S) S T T. (9)

Step 4 To correct the estimated positions of the unknown nodes, the multi-
objective NASG-II algorithm is used. The mathematical formulation of
NASG-objective II’s functions using DV-hop localization is as follows:
X q
f 1 = min( (a − ai )2 + (b − bi )2 − piu ), (10)
i=1,2..n

2×R
X q
f 2 = min( (a − ai )2 + (b − bi )2 − ). (11)
i=1,2..n
3

Fig. 1 depicts the flowchart of DV-hop localization.

4. A Non-inertial Opposition-based Class Topper Optimization (NOCTO)

Algorithm and Multiobjective Functions
Because of their versatility, simplicity, and robustness, metaheuristic op-
timization algorithms are widely used for optimization in both experimental
and industrial problems. One of the most recent and widely used algorithms
in this field is Class Topper Optimization (CTO) [39, 40].

12
 State

Beacon and dumb nodes deployment

The Beacon node transmits its network coordinates

Calculation of the distance between beacon nodes

The lowest number of hops between all nodes with

the fastest path algorithm

Each beacon node calculates the average size of the

hop and transmits the network packet

Dumb nodes calculate the distance to beacon nodes

Method for trilateral of the position of dumb nodes

Optimization of NOCTO to find the correct position

estimate

End

Figure 1: DV-hop localization shows the flowchart.

In accordance with their respective section topper (st ), each student (si )
in each section learns by section topper (st ):

lt+1 = w × lt + c × φ1 × (st − si ), (12)

st+1
i = sti + lt+1 . (13)
In order for the section topper (st ) to be the best student in the section, he
or she must be pleased to compete among the class topper (ct ) that section
topper (st ) learns for the class topper (ct ):

lt+1 = w × lt + c × φ2 × (ct − st,i ), (14)

st+1
t,i = st,i
t + l1 t+1 . (15)

13
w is the inertial weight factor, which maintains the balance between global
and local exploration. w can be written as follows [37]:
 
E wmax − wmin
w = wmax − ×E , (16)
Emax

where wmin is the lowest weight factor; wmax is the highest weight factor.
Opposition-based learning (OBL) [41] is coupled with CTO in this study
to improve exploratory behavior while keeping the convergence rate high.
It successfully accelerates as well as increases the likelihood of reaching the
global optimum by employing an OBL strategy in CTO. After each updated
version, a fresh set of student solutions consisting of Class Topper (Ct ), Sec-
tion Topper (St ) and least fit student S. The difference between the given
vector z and the vector zct is determined using each aspect (i) with a given
vector z as well as the vector zct , the distinction among them is determined
with the following formula:

dif f (i) = |z(i) − zct (i)| . (17)

If the value of dif f (i) exceeds the threshold, the number of parameters with
an quoted jno is increased, i.e. jno = jno +1. Let n be the complete dimension.
If (n − jno ) < jno , the dimensions with a discrepancy between z and zct is
greater than the threshold are in opposition used in following equation:

z(j) = ua(j) + la(j) − z(j). (18)

where j ∈ {i : dif f (i) ≤ threshold}, z is the subsequent vector, ua(j) is the

maximum bound of j th dimension with in the existing population, and la(j) is
the minimum bound of j th dimension in the existing population. In general,
OBL is applied by using the function’s maximum and minimum bounds. Let
z be the ideal dimension value and z + δ be the current value, where δ is a
very small amount. On opposition, the value becomes bj + aj − z − ∂, which
may or may not be close to the best solution unless z = (bj + aj )/2. As a
result, applying opposition learning when the method converges can cause
needless exploration, which can be detrimental. The bounds from the class
are used for opposition of a student to achieve a balance between discovery
and exploitation. In addition, an adaptive elite mutation is used in this paper
to effectively prevent convergence. AEM considers the best student in the
class (the class topper) to be an elite student. As shown below, the test

14
position (Ct ) (class topper) is generated [42]:
arctan(xm)
Ct∗ = Ct + sign.(C + Q ), (19)

where sign is a symbolic variable with a random value of −1 or 1. Parameter

C represents the undetermined constant [43]. From Eq. (20) shows its value.

0.5, other;
C= 1, 10 ≤ st d < 10−1 ;
−2
(20)
1.5, ≤ st d < 10−2 ,


and xm is a mutation degree [43].It can be represented as follows:

rdi iter
xm(i) = (1 − ) · exp(−λ · ), (21)
rdmax itermax
When confronted with a large number of multiobjective problems, the charac-
teristic motivates it to search for and find the final dominant solution as long
as the computation overhead is sufficient. As a result, increasing the con-
vergent speed of the CTO algorithm is both worthwhile and necessary. This
section introduces a new Non-inertial velocity formula (NIV) that replaces
the traditional CTO velocity update formula. The Non-inertial velocity for-
mula (NIV) is divided into three types. The precise expression is as follows:

Type-I: (NIV-D is motivated by the exchange of information with each

other.)

Each student (s) in each section learns from section topper (st ):

l1t+1
i = w.(str1 − str2 ) + c1 φ1 (stt,i − sti ), (22)

Section Topper (st ) is taught by Class Topper (ct ):

l1t+1
i = w.(str1 − str2 ) + c2 φ2 (ctt,i − stt,i ). (23)

where i ∈ {1, 2, ..., M ), w is a differential coefficient used to regulate the

scope of the search.

Type-II: (NIV-U was inspired by group learning and knowledge ex-

change.)

15
 Each student (s) in each section learns from Section topper st :

l2t+1
i = w.(v t − v t−1 ) + c1 .φ1 (stt,i − sti ), (24)

Section Topper (st ) is taught by Class Topper (ct ).

l2t+1
i = w.(v t − v t−1 ) + c2 .φ2 (ctt,i − stt,i ). (25)

where v t and v t−1 are the average student positions of the tth and (t − 1)th
generations, respectively.

Type-III: (NIV-R was inspired by the acquisition of random information

in human society.)

Each student (S) in each section learns from Section topper ((St )):

l3t+1
i = w.φw + c1 .φ1 (stt,i − sti ), (26)
Section Topper (St ) is taught by Class Topper (Ct ).

l3t+1
i = w.φw + c2 .φ2 (ctt,i − stt,i ), (27)
where φw is a random variable with a uniform distribution.

4.1. Mathematical analysis of learing form

In subsequent empirical experiments, NIV-U was shown to be the best
NIV formula. As an example, consider NIV-U to analyse convergence and
provide a stability region for a class.

A model of student learning: Substitute Eqs (24) and (25) into Eqs. (13)
and (15), respectively, and the trajectory equation of students’ learning
is given as follows in the (t + 1)th iteration:

st+1
i = sti + ω(v t − v t−1 ) + β(stt,i − sti )

=(1-β)sti + ω(v t − v t−1 ) + β(stt,i ) (28)

16
 Assuming β = c1 φ1 .
Similarly,
t
st+1 t
t,i =(1-β1 )st,i + ω(v − v
t−1
) + β1 (ctt,i ), (29)
Assuming β1 = c2 φ1 .
As a result, (28) and (29) are easily deduced in the (t + 2)th iteration:
st+2
i =(1-β)st+1
i + ω(v t+1 − v t ) + β(st+1
t,i ), (30)
Assume
Pi1 = ω(v t+1 − v t ) + β(st+1
t,i ), (31)
Similarly,
t+1
st+1
t,i =(1-β1 )st,i + ω(v
t+1
− v t ) + β(ct+1
t,i ), (32)
Assume
Pi2 = ω(v t+1 − v t ) + β(ct+1
t,i ), (33)
P is referred as the ”stability point,” and the order-2 difference recur-
rence equation of student learning is denoted as follows:
st+2
i =(1-β)st+1
i + Pi1 , (34)

Analysis of trajectory and convergence: The corresponding matrix for

Formula Eq (34) is as follows:

st+2
   t+1 
i si
st+1
i
 = X .  sti  (35)
1 1 3×1
where  
1−β 0 0
X = 1 0 0 , (36)
0 0 1
Solve the characteristic equation associated with coefficient matrix X:

|X − Iγ| = γ(1 − β − γ)(γ − 1) = 0

⇒ γ1 = 0, γ2 = 1 − β, γ3 = 1 (37)

17
As a result, the following student trajectories can be obtained:

s(t) = q1 + q2 γ2t , (38)

where q1 and q2 are determined by the kinetic system’s initial condi-

tions. So, substituting Eq. (42) into Eq. (34) yields the following:
1
q1 = .P, (39)
β
1
q2 = s0 + .P, (40)
β
The initial learing stage of the students is denoted by s0 . If a student’s
trajectory is convergent, formula (42) must be satisfied:

β > 1 and kγ2 k < 1 ⇒ |1 − β| 1 ⇒ 0 < β < 2, (41)

As a result, the order-2 stable region is obtained as follows:

SVD = {β : 1 < β < 2}, (42)

In SVD , the trajectory sequence of students will be stably convergent

to the stability point P .

lim si (t) = lim (q1 + q2 γ2t );

t→∞ t→∞

ω(v t − v t−1 ) + β st+1

t,i
= , (43)
β

where v t is trending to v t−1 as evolution goes on. Further derive:

ω (v t − v t−1 )
Pi1 = + st+1
t,i ;
c1

ω (vt − v t−1 )
= + pi1 . (44)
c1

18
 Similarly, in the case of the class topper
ω (v t − v t−1 )
Pi2 = + ct+1
t,i ;
c2

ω (vt − v t−1 )
= + pi2 , (45)
c2

where pi1 and pi2 denote the ”stability point” of the traditional CTO system
[40]. The non-inertial velocity formula (NIV) is used to replace the classic
CTO examination improvement velocity update formula, and a non-inertial
opposition-based class topper optimization, or NOCTO for short, is con-
structed using a combination of the OBL and AEM strategies. Fig. 2 shows
the NOCTO algorithm in detail.

The following steps are included in the proposed algorithm of multi-

objective NOCTO:

Step 1. Dumb node locations are determined using the minimum hop dis-
tance calculated by each beacon (i.e., one-hop-size) from another bea-
con in the network. The greater the precision of this estimated distance,
the effective will be the approximate positions.

Step 2. Instead of using the conventional approach, we define to measure the

average hop distance among beacons. As a result, we use the polyno-
mial approximation to reduce estimated localization error and increase
localization accuracy. The following polynomial is used to calculate the
approximate hop distance from beacons i and other anchors m.
rim = κ0 + κ1 pim + κ2 p2im , (46)

where κ0 , κ1 and κ2 are the coefficients.

   
p2i1 pi1 1   ri1
 p2 pi2 1 κ2
    ri2 
 
 i2
 .. .. ..  κ1 =  ..  , (47)
 . . . κ  . 
2 0
pij pij 1 rij

19
 State

Set no. of Sections, no. of students in sections, each student having

subjects, C= 2.5 to 0.5, W=0.8

1) Opposition Student Generation (OS) in class (C), 2) Record Class

Topper (CT) and Section Topper (ST) with a Performance Index (PI).

Learns from No If Student Yes Learns from Class

Section Topper = ST Topper

No Update Section Topper (ST) and Class Topper (CT) by equation

(8,9,10,11) and NIV

Recalculate Performance Index (PI) of students

Update Section Topper (ST) and Class Topper (CT)

If max
Iteration
Yes
End

Figure 2: The proposed Multiobjective NOCTO flow chart

where p is the number of hops among beacon nodes. We solve the fol-
lowing equation to find its polynomial function to estimate the distance
among nodes:

−1
κ = (P T P ) P T R, (48)
Then, for each dumb node, the distance between itself and the beacon
node is calculated as follows:

rij = κ0 + κ1 pij + κ2 p2ij , (49)

where rij and pij represent the distance as well as minimum number
of hops between the i0 th dumb node and the j 0 th beacon node, respec-

20
 tively. As a result, we get the following matrix form:
rest = hop × κ, (50)
The actual distance between beacon i and j shall be calculated by
q
rtrue = (ai − aj )2 + (bi − bj )2 . (51)
Error between nodes i and j of the beacon is given as:
rerror = rest − rtrue . (52)
Now, we have added a rectification factor, and this is defined as:
rerror
τ=
, (53)
s
where s is the number of beacon nodes. The τ rectification factor is
used by adding it to the previous hop size to change the hop size of
the beacon node. The adjusted distance between i beacon nodes and
k dumb node are determined as [37]:

M od
rik = (HopSize + τ ) × Hik . (54)
.

Step 3. We use a multiobjective non-inertial opposition-based topper class

optimization to increase the position accuracy of the sensor nodes. We
used six single-objective functions and three multiobjective functions
to develop our proposed technique in this paper. As a result, the follow-
ing is a proposed algorithm that employs single-objective optimization
functions:

X q
f1 (a, b) = min( (a − ai )2 + (b − bi )2 − rik
M od
), (55)
i=1,2....M

where f1 (a, b) is the objective function that represents the distance

error; in this case, function f1 (a, b) is considered in the proposed opti-
mization (NOCTO).
X q
f2 (a, b) = min( (a − ai )2 + (b − bi )2 − rik
M od 2
), (56)
i=1,2....M

21
 where f2 (a, b) is the 2nd order objective function of f1 (a, b) that repre-
sents the distance error.
X q 2×R
f3 (a, b) = min( (a − ai )2 + (b − bi )2 − ), (57)
i=1,2..M
3

where f3 (a, b) is the objective function that represents the distance

error; in this case, function f3 (a, b) is obtained from [23].
X q 2×R 2
f4 (a, b) = min( (a − ai )2 + (b − bi )2 − ), (58)
i=1,2..M
3

where f4 (a, b) is the 2nd order objective function of f3 (a, b) that repre-
sents the distance error.
X q
f5 (a, b) = min( (a − ai )2 + (b − bi )2 − ritM od ), (59)
i=1,2....M

where f5 (a, b) is the objective function that represents the distance

error; in this case, function f5 (a, b) is considered as [40]. where ritM od is
the distance between beacon and dumb node and is given by

ritM od = (hopsi + θ) × HopSizeavg , (60)

where θ is the factor used to change the beacon node hop size by ap-
plying it to the previous hop size.

rappro − rtrue
θ= , (61)
s
X q
f6 (a, b) = min( (a − ai )2 + (b − bi )2 − ritM od )2 . (62)
i=1,2....M

where f6 (a, b) is the 2nd order objective function of f5 (a, b) that repre-
sents the distance error.

The proposed algorithm considers the following multiobjective functions,

which are combination of single objective functions:
m1 = f1 and f2

22
 X q
m1 (a, b) = min( (a − ai )2 + (b − bi )2 − rik
M od
)
i=1,2....M

and
X q
min( (a − ai )2 + (b − bi )2 − rik
M od 2
), (63)
i=1,2....M

m2 = f3 and f4

2×R
X q
m2 (a, b) = min( (a − ai )2 + (b − bi )2 − )
i=1,2..M
3
and
2×R 2
X q
min( (a − ai )2 + (b − bi )2 − ), (64)
i=1,2..M
3

m3 = f5 and f6

X q
m3 (a, b, ) = min( (a − ai )2 + (b − bi )2 − ritM od );
i=1,2....M

and
X q
min( (a − ai )2 + (b − bi )2 − ritM od )2 . (65)
i=1,2....M

Fig. 2 depicts the flowchart of NOCTO and Table 2 lists the notations
used in the proposed method.

5. Metrics of Performance
The location performance of various objective functions has been analyzed
using the following metrics:

23
 Table 2: Notations

Notation Description
lt+1 Learn from
lt Previous learn from
w Minimum weight factor
c Stochastic acceleration term
φ1 , φ2 and φw Random variables
s Student
st Section Topper
ct Class Topper
κ0 polynomial function
rest Estimation distance
rtrue True distance
τ Rectification factor
t number of dumb nodes
(ptr tr
i , qi ) Actual positions of dumb node i
(pest est
i , qi ) Estimated positions
R Communication range of sensor node
st d Standard deviation of fitness
λ Undetermined constant
iter Number of iteration
itermax Maximum number of iteration
rdi i0 th dimension of students
rdmax maximum value of rdi
Emax Number of examination

24
5.1. Localization error
We investigated the effect of the total number of nodes, beacon nodes,
and range on simulation results. The following is the localization error:
t q
tr 2 tr 2
(aest est
P
i − ai ) + (bi − bi )
LE = i=1 , (66)
t×R

where LE stands for localization error.

5.2. Localization error variation

Variance measures how far each number in a set deviates from the mean.

v
u t q 2
u ( (aest
uP tr 2 est tr 2
t i=1 i − ai ) + (bi − bi ) − LE × R)
LEV = ,
t × R2
(67)

where LEV denotes the variation in localization error.

5.3. Localization accuracy and coverage

Localization parameters assess the coverage and accuracy of localization,
and these performance parameters are derived using the formula:
t q
tr 2 tr 2
(aest est
P
i − ai ) + (bi − bi )
i=1
LA = , (68)
t × R2

where LA denotes localization accuracy.

t
R2 (ai , bi )
S
i=1
Coverage = , (69)
Area

where t denotes the number of unknown or dumb nodes, (atr tr

i , bi ) is true
positions of unknown or dumb node i. The estimated dumb node i positions
are (aest est
i , bi ). R is the sensor node’s communication range.

25
 Table 3: Parameters for simulation.

Parameters Value
Border length 100 × 100m2 to 300 × 300m2
Number of sensor nodes 200 to 450
Beacon nodes (M) 20 to 160
Range (R) 25 to 50
Algorithm of Optimization NOCTO
Maximum iterations 50
Simulation Tool LabVIEW@2015

6. Result and discussions

For our system, we proposed a Non-inertial Opposition-based Class Top-
per Optimization algorithm with the following parameters C = 4, W = 2.5
to 0.05, Itermax = 20, as well as the random variables in between 0 to 1.
For 2D, the minimum transmission range of a node is 25 m2 . The value of
the communication range has been found to be dependent on node density.
The value of the communication range has been found to be dependent on
node density. The greater the communication range, the lower the node den-
sity. The node distribution is viewed in Fig. 3. The red circle represents
the beacon node, while the blue pentagram represents the dumb nodes. The
approximation error with localization is represented by the blue straight line.
Table 3 lists the parameters that were used in the simulation. The following
parameters were used to evaluate the results:

i The total number of beacon nodes.

ii The total number of nodes.

iii The message radius of a nodes.

iv Communication Cost.

6.1. Total number of beacon nodes

It has been observed that as the total number of beacon nodes within
the network area increases, the localization error in the proposed objective
functions decreases. This is because as the number of beacon nodes increases,

26
 250

200

150
Meters

100

50

0
0 50 100 150 200 250
Meters

Figure 3: The distribution of sensor nodes was localized using spiral beacon nodes deploy-
ment

the number of hops between dumb and beacon nodes decreases, and the
network becomes more connected. We kept the number of dumb nodes and
the node’s message radius constant in this case. Fig. 4, 5, 6, and 7 show
how the total number of beacon nodes affects localization error, localization
error variance, localization accuracy, and localization coverage, respectively.
It is noted in Table 4 that the m1 has less localization error and localization
error variation. Table 5 shows that the multi-objective function m1 has the
improved Localization Accuracy and coverage.

6.1.1. On beacon nodes, the effects of localization error

It can be seen in Fig. 4 that the localization error decreases as the total
number of beacon nodes increases. For simulation purposes, a 250 × 250 m2
area is assumed to have fixed 200 dumb nodes and a message radius of 25m.

27
 0.45
S.O. f3 (spiral)
S.O. f4 (spiral)
S.O. f5 (spiral)
0.4 S.O. f6 (spiral)
M.O. m2 (spiral)
M.O. m3 (spiral)
M.O. m [25]
0.35 S.O. f2 (spiral)
S.O. f1 (spiral)
Localization Error

M.O. m1 (spiral)

0.3

0.25

0.2

0.15

0.1
20 40 60 80 100 120 140 160
The amount of beacon nodes

Figure 4: The total number of beacon nodes used for LE is changing

The number of beacon nodes ranges from 20 to 160. For example, the total
number of beacon nodes is set to 20. According to Table 4, the proposed
multi-objective function m1 has less localization error in f 1, f 2, m [26], m3,
m2, f 6, f 5, f 4 and f 3 by approximately 0.01%, 0.08%, 0.09%, 0.15%, 0.16%,
0.25%, 0.31%, 0.45%, and 0.51%, respectively.

6.1.2. On beacon nodes, the effects of localization error variance

According to Fig. 5, the variance of the localization error decreases as
the total number of beacon nodes increases. For simulation purposes, a
250 × 250 m2 area is assumed to have fixed 200 dumb nodes and a message
radius of 25m. The number of beacon nodes ranges from 20 to 160. For
example, the total number of beacon nodes is set to 20. Our proposed multi-
objective function m1 has a 0.001%, 0.10%, 0.13%, 0.17%, 0.20%, 0.30%,

28
 0.035
S.O. f3 (spiral)
S.O. f4 (spiral)
S.O. f5 (spiral)
S.O. f6 (spiral)
0.03 M.O. m2 (spiral)
M.O. m3 (spiral)
Localization Error Variance

S.O. f2 (spiral)
M.O. m [25]
0.025 S.O. f1 (spiral)
M.O. m1 (spiral)

0.02

0.015

0.01

0.005
20 40 60 80 100 120 140 160
The amount of beacon nodes

Figure 5: The total number of beacon nodes used for LEV is changing

0.35%, 0.40%, and 0.60% lower localization error variation in f 1, f 2, m [26],

m3, m2, f 6, f 5, f 4 and f 3, respectively, according to Table 4.

6.1.3. On beacon nodes, the effects of localization accuracy

According to Fig. 6, the localization accuracy increases as the total num-
ber of beacon nodes increases. For simulation purposes, a 250 × 250 m2 area
is assumed to have fixed 200 dumb nodes and a message radius of 25m. The
number of beacon nodes ranges from 20 to 160. For example, the total
number of beacon nodes is set to 20. Table 5 shows that the proposed multi-
objective function m1 improves Localization Accuracy m2, m [26], m3, f 1,
f 2, f 5, f 6, f 4 and f 3 by approximately 0.48%, 1.38%, 1.33%, 1.73%, 1.86%,
6.12%, 6.13%, 6.14%, and 6.16%, respectively.

29
Table 4: A comparison of the effects of the proposed (multiobjective) algorithm on LE
and LEV beacon nodes with single-objective algorithms and multiple objective NASG-II
[26] were conducted.

Number of nodes 20 50 80 100 130 150

[S.O. spiral] LE (%) 0.57 0.7 0.8 0.86 0.82 0.91
f3 LEV (%) 0.65 0.65 0.75 0.85 0.68 1
[S.O. spiral] LE (%) 0.51 0.65 0.75 0.75 0.75 0.8
f4 LEV (%) 0.45 0.325 0.3 0.45 0.38 0.87
[S.O. spiral] LE (%) 0.37 0.5 0.6 0.66 0.62 0.91
f5 LEV (%) 0.40 0.625 0.6 0.75 0.58 1
[S.O. spiral] LE (%) 0.31 0.47 0.58 0.6 0.6 0.74
f6 LEV (%) 0.35 0.225 0.2 0.35 0.28 0.77
[M.O. spiral] LE (%) 0.22 0.27 0.32 0.48 0.4 0.5
m2 LEV (%) 0.25 0.2 0.18 0.24 0.2 0.55
[M.O. spiral] LE (%) 0.21 0.25 0.31 0.46 0.36 0.46
m3 LEV (%) 0.22 0.2 0.15 0.21 0.18 0.51
M.O [26] LE (%) 0.158 0.19 0.25 0.28 0.27 0.35
m LEV (%) 0.18 0.16 0.15 0.17 0.16 0.31
[S.O. spiral] LE (%) 0.148 0.19 0.24 0.27 0.26 0.3
f2 LEV (%) 0.15 0.1 0.12 0.14 0.1 0.33
[S.O. spiral] LE (%) 0.07 0.09 0.12 0.14 0.13 0.16
f1 LEV (%) 0.047 0.04 0.042 0.046 0.16 0.22
[S.O. spiral] LE (%) 0.06 0.041 0.1 0.12 0.11 0.14
m1 LEV (%) 0.046 0.035 0.04 0.043 0.13 0.2

30
Table 5: A comparison of the effects of the proposed (multiobjective) algorithm on LA and
Coverage beacon nodes with single-objective algorithms and multiple objective NASG-II
[26] were conducted.

Number of bea- 20 50 80 100 130 150

con nodes
[M.O. spiral] LA (%) 6.28 5.43 4.2 3.41 3.07 2.75
m1 Coverage 0.96 3.85 0.96 0.23 0.11 0
(%)
[M.O. spiral] LA (%) 5.8 4.73 4.5 3.1 2.91 2.62
m2 Coverage 0.9 3.5 0.9 0.21 0.1 0
(%)
M.O [26] LA (%) 4.9 3.93 3.55 2.1 1.91 1.62
m Coverage(%) 0.8 2.5 0.8 0.11 0.091 0
[M.O. spiral] LA (%) 4.95 3.93 3.55 2.1 1.91 1.62
m3 Coverage(%) 0.75 2.5 0.8 0.11 0.091 0
[S.O. spiral] LA (%) 4.55 0.43 0.41 0.39 0.38 0.37
f1 Coverage 0.65 1.25 0.5 0.1 0.07 0
(%)
[S.O. spiral] LA (%) 4.42 0.38 0.35 0.26 0.21 0.187
f2 Coverage 0.6 2.25 0.4 0.09 0.05 0
(%)
[S.O. spiral] LA (%) 0.16 0.43 0.41 0.39 0.38 0.37
f5 Coverage 0.3 1.25 0.5 0.1 0.07 0
(%)
[S.O. spiral] LA (%) 0.15 0.38 0.35 0.26 0.21 0.187
f6 Coverage 0.25 2.25 0.4 0.09 0.05 0
(%)
[S.O. spiral] LA (%) 0.14 0.12 0.1 0.08 0.07 0.062
f4 Coverage 0.2 1 0.2 0.08 0.03 0
(%)
[S.O. spiral] LA (%) 0.12 0.1 0.08 0.075 0.06 .058
f3 Coverage 0.15 0.5 0.2 0.064 0.02 0
(%)

31
 60

50
M.O. m1 (spiral)
Localization Accuracy(%)

M.O. m2 (spiral)
M.O. m [25]
40 M.O. m3 (spiral)
S.O. f1 (spiral)
S.O. f2 (spiral)
S.O. f5 (spiral)
S.O. f6 (spiral)
30 S.O. f4 (spiral)
S.O. f3 (spiral)

20

10

0
20 40 60 80 100 120 140 160
The amount of beacon nodes

Figure 6: The total number of beacon nodes used for error LA is changing

6.1.4. On beacon nodes, the effects of localization Coverage

Fig. 7 show the effects of the beacon on localization coverage, for simula-
tion purposes, a 250 × 250 m2 area is assumed to have fixed 200 dumb nodes
and a message radius of 25m . The number of beacon nodes ranges from 20
to 160. Table 5 contains the associated analysis data. For example, the total
number of beacon nodes is set to 20. Table 5 shows that, multi-objective
function m1 has increased coverage by approximately 0.06%, 0.16%, 0.21%,
0.31%, 0.36%, 0.66%, 0.71%, 0.76%, and 0.81% in m2, m [26], m3, f 1, f 2,
f 5, f 6, f 4 and f 3, respectively.

6.2. Total number of nodes

It has been discovered that as the number of nodes within the area in-
creases, the localization error in the proposed objective functions decreases.

32
 1

0.9

0.8
M.O. m1 (spiral)
M.O. m2 (spiral)
0.7 M.O. m [25]
M.O. m3 (spiral)
S.O. f1 (spiral)
0.6
Coverage

S.O. f2 (spiral)
S.O. f5 (spiral)
S.O. f6 (spiral)
0.5 S.O. f4 (spiral)
S.O. f3 (spiral)
0.4

0.3

0.2

0.1

0
20 40 60 80 100 120 140 160
The amount of beacon nodes

Figure 7: The total number of beacon nodes used for Coverage is changing

This is due to the fact that as the number of nodes increases, the average
number of neighbours for each node grows, resulting in a well-connected net-
work. In this case, we kept the number of beacon nodes and the message
radius of each node constant. Fig. 8, 9, 10, and 11 depict the effect of the
total number of nodes on localization error and variance, respectively.

6.2.1. Localization errors affect the total number of nodes

In Fig. 8, it has been observed that the localization error decreases with
an increase in the total number of dumb nodes. For simulation, an area of
250 × 250 m2 is considered to have fixed 35 beacon nodes and a message
radius of 25m. Total nodes vary from 100 to 350. For example, the total
number of nodes is set to 200. It is noted in the Table 6 that our proposed
multi-objective function m1 have approximately 0.097%, 0.087%, 0.077%,

33
 0.95
S.O. f3 (spiral)
S.O. f4 (spiral)
0.9 S.O. f5 (spiral)
S.O. f6 (spiral)
M.O. m2 (spiral)
0.85 M.O. m3 (spiral)
M.O. m [25]
Localization Error

S.O. f2 (spiral)
0.8 S.O. f1 (spiral)
M.O. m1 (spiral)

0.75

0.7

0.65

0.6

0.55
200 250 300 350 400 450
The amount of nodes

Figure 8: The total number of beacon nodes used for LE is changing

0.085%, 0.059%, 0.058%, 0.013%, 0.012%, and 0.011% less localization error
in f 3, f 4, f 5, f 6, m2, m3, m [26], f 2 and f 1, respectively.

6.2.2. Localization error variance affect the total number of nodes

As shown in Fig. 9, the localization error variance decreases as the total
number of nodes increases. For simulation, an area of 250 × 250 m2 is con-
sidered to have fixed 35 beacon nodes and a message radius of 25m. Total
nodes vary from 100 to 350. For example, the total number of nodes is set
to 200. It is noted in the Table 6 that our proposed multi-objective function
m1 have approximately 1.78%, 1.43%, 1.38%, 1.33%, 0.71%, 0.70%, 0.01%,
0.009%, and 0.005% less localization error variation in f 3, f 4, f 5, f 6, m2,
m3, m [26], f 2 and f 1, respectively

34
Table 6: A comparison of the effects of the proposed (multiobjective) algorithm on LE and
LEV number of nodes with single-objective algorithms and multiple objective NASG-II
[26] were conducted.

Number of nodes 200 250 300 350 400 450

[S.O. spiral] LE (%) 0.13 0.16 0.17 0.205 0.209 0.24
f3 LEV (%) 1.85 1.97 1.89 1.9 2.3 2.25
[S.O. spiral] LE (%) 0.12 0.15 0.16 0.204 0.208 0.23
f4 LEV (%) 1.5 1.87 1.79 1.8 2.2 2.2
[S.O. spiral] LE (%) 0.11 0.14 0.15 0.202 0.207 0.22
f5 LEV (%) 1.45 1.77 1.69 1.7 2 2.15
[S.O. spiral] LE (%) 0.118 0.124 0.135 0.142 0.149 0.16
f6 LEV (%) 1.4 1.3 1.1 1.2 1.55 1.63
[M.O. spiral] LE (%) 0.092 0.097 0.099 0.01 0.015 0.12
m2 LEV (%) 0.78 0.69 0.65 0.80 0.96 1.47
[M.O. spiral] LE (%) 0.091 0.096 0.098 0.01 0.014 0.12
m3 LEV (%) 0.77 0.68 0.64 0.79 0.95 1.46
M.O [26] LE (%) 0.046 0.050 0.055 0.060 0.08 0.070
m LEV (%) 0.081 0.075 0.073 0.091 0.25 0.22
[S.O. spiral] LE (%) 0.045 0.068 0.07 0.073 0.078 0.08
f2 LEV (%) 0.075 0.38 0.35 0.52 0.9 1.21
[S.O. spiral] LE (%) 0.044 0.048 0.049 0.045 0.05 0.065
f1 LEV (%) 0.071 0.065 0.063 0.091 0.15 0.21
[S.O. spiral] LE (%) 0.033 0.037 0.039 0.042 0.043 0.045
m1 LEV (%) 0.066 0.062 0.06 0.081 0.13 0.17

35
 0.08
S.O. f3 (spiral)
S.O. f4 (spiral)
0.07 S.O. f5 (spiral)
S.O. f6 (spiral)
M.O. m2 (spiral)
Localization Error Variance

M.O. m3 (spiral)
0.06 S.O. f2 (spiral)
M.O. m [25]
S.O. f1 (spiral)
M.O. m1 (spiral)
0.05

0.04

0.03

0.02

0.01
200 250 300 350 400 450
The amount of nodes

Figure 9: The total number of beacon nodes used for error LEV is changing

6.2.3. Localization accuracy affect the total number of nodes

In Fig. 10, it has been observed that the localization accuracy increases
with an increase in the total number of nodes. For simulation, an area of
250×250 m2 is considered to have fixed 35 beacon nodes and a message radius
of 25m. Total nodes vary from 100 to 350. For example, the total number
of nodes is set to 200. It can be seen from Table 7 that the our proposed
multi-objective function m1 has approximately 11.5%, 11%, 10.5%, 10%, 3%,
2%, 1%, 0.7%, and 0.5% improved Localization Accuracy in f 3, f 4, f 6, f 5,
f 2, f 1, m3, m [26], and m2,respectively,

6.2.4. Localization coverage affect the total number of nodes

Fig. 11 show the effect of the total number of nodes on Localization
coverage. For simulation, an area of 250 × 250 m2 is considered to have fixed

36
 40
M.O. m1 (spiral)
M.O. m2 (spiral)
M.O. m [25]
35 M.O. m3 (spiral)
S.O. f1 (spiral)
S.O. f2 (spiral)
30
Localization Accuracy(%)

S.O. f5 (spiral)
S.O. f6 (spiral)
S.O. f4 (spiral)
25 S.O. f3 (spiral)

20

15

10

0
200 250 300 350 400 450
The amount of nodes

Figure 10: The total number of beacon nodes used for LA is changing

35 beacon nodes and a message radius of 25m. Total nodes vary from 100
to 350. For example, the total number of nodes is set to 200. Table 7 shows
that, multi-objective function m1 has increased coverage by approximately
9%, 8%, 7%, 6%, 3.7%, 2.6%, 1.5%, 1%, and 0.4% in f 3, f 4, f 6, f 5, f 2, f 1,
m3, m [26], and m2, respectively,

6.3. Message radius of sensor nodes

It has been observed that as the message radius of the node within the
area increases, the localization error in the proposed objective functions de-
creases. This is because, as message radius increases for a fixed number of
dumb and beacon nodes, the network becomes well-connected. In this case,
we kept the number of dumb nodes and beacon nodes constant. Fig. 12, 13,
14, and 15 depict the effect of the node’s message radius on localization error,

37
Table 7: A comparison of the effects of the proposed (multiobjective) algorithm on LA
and Coverage number of nodes with single-objective algorithms and multiple objective
NASG-II [26] were conducted.

Number of nodes 200 250 300 350 400 450

[M.O. spiral] LA (%) 12 4.3 4.75 2.5 1.8 1.46
m1 Coverage 9 4 1 .25 1 0.5
(%)
[M.O. spiral] LA (%) 11.5 3.35 4.45 2.8 1.5 2.3
m2 Coverage 7.6 4.5 0.8 0.24 0.8 0.43
(%)
M.O. [26] LA (%) 11.3 3.31 4.42 2.5 1.3 2.1
m Coverage(%) 7.5 4.4 0.6 0.23 0.7 0.42
[M.O. spiral] LA (%) 11 3.3 3.45 1.8 1.5 1.3
m3 Coverage 8.6 3.5 0.9 0.23 0.5 0.33
(%)
[S.O. spiral] LA (%) 3 1 1.4 1.7 0.9 0.53
f1 Coverage 4 1.5 0.8 0.5 0.01 0
(%)
[S.O. spiral] LA (%) 2 1 1.2 1.9 1.5 0.4
f2 Coverage 3 0.9 0.3 0.09 0.125 0.142
(%)
[S.O. spiral] LA (%) 1.9 0.9 1.1 1.7 1.2 0.2
f5 Coverage 2 0.7 0.2 0.08 0.122 0.141
(%)
[S.O. spiral] LA (%) 1.9 0.9 1.1 1.7 1.2 0.2
f6 Coverage 2 0.7 0.2 0.08 0.122 0.141
(%)
[S.O. spiral] LA (%) 1 0.9 1.4 1.8 1.11 0.4
f4 Coverage 1 0.2 0.08 0.03 0.01250.11
(%)
[S.O. spiral] LA (%) 0.5 0.3 0.9 1.2 0.8 0.28
f3 Coverage 0 0 0 0 0 0
(%)

38
 1
M.O. m1 (spiral)
0.9 M.O. m2 (spiral)
M.O. m [25]
M.O. m3 (spiral)
0.8 S.O. f1 (spiral)
S.O. f2 (spiral)
0.7 S.O. f5 (spiral)
S.O. f6 (spiral)
S.O. f4 (spiral)
0.6 S.O. f3 (spiral)
Coverage

0.5

0.4

0.3

0.2

0.1

0
200 250 300 350 400 450
The amount of nodes

Figure 11: The total number of beacon nodes used for Coverage is changing

localization error variance, localization accuracy and coverage, respectively.

6.3.1. The message radius is influenced by localization errors

The effect of message radius on localization error is depicted in Fig. 12.
For simulation purposes, a 250 × 250 m2 area with 200 dumb nodes and
35 beacon nodes is assumed. The message radius varies between 25 and
50 metres. It has been discovered that as message radius increases, the
localization error decreases. For example, the message radius is set to 25.
According to Table 8, the our proposed multi-objective function m1 has
approximately 0.27%, 0.26%, 0.25%, 0.2%, 0.14%, 0.13%, 0.12%, 0.09%, and
0.02% less localization error in f 3, f 4, f 5, f 6, m2, m3, m [26], f 2 and f 1,
respectively.

39
Table 8: The proposed (multiobjective) algorithm and multiple objective NASG-II [26]
were compared to single-objective DV-Hop algorithms in terms of range effects on LE and
LEV

Transmission 25 30 35 40 45 50
Range (m)
[S.O. spiral] LE (%) 0.29 0.49 1.15 0.9 0.89 0.75
f3 LEV (%) .627 2.5 3.35 5.27 7.29 0.3
[S.O. spiral] LE (%) 0.28 0.48 1.13 0.85 0.88 0.74
f4 LEV (%) .626 2.4 3.34 5.26 7.28 0.2
[S.O. spiral] LE (%) 0.27 0.47 1.1 0.88 0.87 0.73
f5 LEV (%) 0.625 2.2 3.33 5.25 7.26 0.1
[S.O. spiral] LE (%) 0.22 0.37 1.3 0.86 0.84 0.6
f6 LEV (%) 0.5 2.1 3.1 5.14 7.12 0.7
[M.O. spiral] LE (%) 0.166 0.27 1.1 0.82 0.81 0.26
m2 LEV (%) 0.37 1.9 2.2 4.13 6.2 0.05
[M.O. spiral] LE (%) 0.165 0.25 1.08 0.85 0.8 0.24
m3 LEV (%) 0.35 1.5 2.1 4.11 6.1 0.03
M.O. [26] LE (%) 0.162 0.26 1.08 0.72 0.71 0.2
m LEV (%) 0.33 1.8 2.1 4.11 6.1 0.03
[S.O. spiral] LE (%) 0.11 0.21 1.06 0.62 0.61 0.13
f2 LEV (%) 0.25 1.6 1.8 3.10 5.1 0.02
[S.O. spiral] LE (%) 0.04 0.08 0.2 0.11 0.1 0.21
f1 LEV (%) 0.23 2.62 2.7 4.1 5.1 0.018
[S.O. spiral] LE (%) 0.02 0.06 0.1 0.09 0.08 0.11
m1 LEV (%) 0.2 1.62 1.7 3.1 4.1 0.017

40
 0.5
S.O. f3 (spiral)
S.O. f4 (spiral)
S.O. f5 (spiral)
0.45 S.O. f6 (spiral)
M.O. m2 (spiral)
M.O. m3 (spiral)
0.4 M.O. m [25]
S.O. f2 (spiral)
S.O. f1 (spiral)
Localization Error

0.35 M.O. m1 (spiral)

0.3

0.25

0.2

0.15

0.1
25 30 35 40 45 50
Range

Figure 12: Variation in localization error as range changes

6.3.2. The message radius is influenced by localization error variance

From Fig. 13, it has been observed that the localization error variance
decreases with the increase in message radius. For simulation purposes, a
250 × 250 m2 area with 200 dumb nodes and 35 beacon nodes is assumed.
The message radius varies between 25 and 50 metres. It has been discov-
ered that as message radius increases, the localization error decreases. For
example, the message radius is set to 25. According to Table 8, the pro-
posed multi-objective function m1 has approximately 0.42%, 0.41%, 0.4%,
0.3%,0.17%,0.15%, 0.14%, 0.05%, and 0.03% less localization error in f 3, f 4,
f 5, f 6, m2, m3, m [26], f 2 and f 1, respectively.

41
Table 9: The proposed (multiobjective) algorithm and multiple objective NASG-II [26]
were compared to single-objective DV-Hop algorithms in terms of range effects on LA and
Coverage .

Transmission 25 30 35 40 45 50
Range (m)
[M.O. spiral] LA (%) 3.1 2.53 1.74 1.54 0.90 0.05
m1 Coverage 9.5 5 1.2 0.28 1.2 0.8
(%)
[M.O. spiral] LA (%) 3 2.03 1.64 1.34 0.80 0.03
m2 Coverage 9 4 1 0.25 1 0.5
(%)
M.O. [26] LA (%) 2.9 2.71 2.73 1.64 0.75 0.04
m Coverage 8.9 3.9 0.95 0.24 0.45 0.37
(%)
[M.O. spiral] LA (%) 2.6 2.01 2.53 1.24 0.70 0.05
m3 Coverage 8.6 3.5 0.9 0.23 0.5 0.33
(%)
[S.O. spiral] LA (%) 2.2 0.9 0.85 0.74 0.09 0.025
f1 Coverage 4 1.5 0.8 0.5 0.01 0
(%)
[S.O. spiral] LA (%) 2 0.8 0.76 0.72 0.07 0.012
f2 Coverage 3 0.9 0.3 0.09 0.125 0.142
(%)
[S.O. spiral] LA (%) 1.9 0.7 0.66 0.62 0.06 0.011
f5 Coverage 2 0.8 0.2 0.08 0.124 0.141
(%)
[S.O. spiral] LA (%) 1.8 0.6 0.65 0.61 0.05 0.011
f6 Coverage 1 0.7 0.1 0.07 0.122 0.14
(%)
[S.O. spiral] LA (%) 0.66 0.52 0.46 0.44 0.08 0.025
f4 Coverage 0 0 0 0 0 0
(%)
[S.O. spiral] LA (%) 0.4 0.3 0.25 0.4 0.07 0.024
f3 Coverage 0 0 0 0 0 0
(%)

42
 0.35
S.O. f3 (spiral)
S.O. f4 (spiral)
S.O. f5 (spiral)
0.3 S.O. f6 (spiral)
M.O. m2 (spiral)
M.O. m3 (spiral)
Localization Error Variance

S.O. f2 (spiral)
0.25 M.O. m [25]
S.O. f1 (spiral)
M.O. m1 (spiral)

0.2

0.15

0.1

0.05

0
25 30 35 40 45 50
Range

Figure 13: Variation in localization error variance as range changes

6.3.3. The message radius is influenced by localization accuracy

According to Fig. 14, the localization accuracy increases as the message
radius increases. For simulation purposes, a 250 × 250 m2 area with 200
dumb nodes and 35 beacon nodes is assumed. The message radius varies
between 25 and 50 metres. For example,the communication range is set to
25. According to Table 9, our proposed multi-objective function m1 has
approximately 2.7%, 2.44%, 1.3%, 1.2%,1.1%,0.9%, 0.52%, 0.2%, and 0.1%
improved Localization Accuracy in f 3, f 4, f 6, f 5, f 2, f 1 m3, m [26], and
m2 respectively

6.3.4. The message radius is influenced by localization coverage

Fig. 15 shows the effect of the beacon on localization coverage. For sim-
ulation purposes, a 250 × 250 m2 area with 200 dumb nodes and 35 beacon

43
 40
M.O. m1 (spiral)
M.O. m2 (spiral)
35 M.O. m [25]
M.O. m3 (spiral)
S.O. f1 (spiral)
Localization Accuracy(%)

S.O. f2 (spiral)
30 S.O. f5 (spiral)
S.O. f6 (spiral)
S.O. f4 (spiral)
S.O. f3 (spiral)
25

20

15

10

5
25 30 35 40 45 50
Range

Figure 14: Variation in localization accuracy as range changes

nodes is assumed. The message radius varies between 25 and 50 metres.

For example, the message radius is set to 25. Table 9 shows that, multi-
objective function m1 has increased coverage by approximately 9.5%, 9.5%,
8.5%, 7.5%, 6.5%, 5.5%, 0.9%, 0.6%, and 0.5% in f 3, f 4, f 6, f 5, f 2, f 1 m3,
m [26], and m2, respectively.

6.4. Communication Cost analysis of the algorithm

The computational cost of an algorithm can be used to determine its com-
plexity. The total number of messages sent from sender to receiver by the
sensor nodes during the localization process represents the communication
cost. Step 1 involves beacon nodes communicating with all dumb nodes in or-
der to broadcast their location. At this point, the traditional single-objective,

44
 1
M.O. m1 (spiral)
0.9 M.O. m2 (spiral)
M.O. m [25]
M.O. m3 (spiral)
0.8 S.O. f1 (spiral)
S.O. f2 (spiral)
S.O. f5 (spiral)
0.7 S.O. f6 (spiral)
S.O. f4 (spiral)
0.6 S.O. f3 (spiral)
Coverage

0.5

0.4

0.3

0.2

0.1

0
25 30 35 40 45 50
Range

Figure 15: Variation in coverage as range changes

traditional multiobjective, and proposed multiobjective NOCTO algorithms

all have the same communication cost. Traditional single-objective and mul-
tiobjective algorithms only communicate between nodes in step 2, whereas
the proposed multiobjective NOCTO algorithm eliminates all communica-
tion between beacon nodes and dumb nodes by calculating the hop size of
beacons at dumb nodes. As shown in Fig. 16, the elapsed time of all al-
gorithms is calculated to determine the computational cost of all objective
functions. The tic-toc function measures elapsed time, which is the total
time taken from the start of the localization process to the end. To measure
elapsed time, tic collaborates with the toc function. The tic function saves
the current time, and the toc function calculates the elapsed time using that
value. Fig. 16 depicts the total number of dumb nodes vs. the elapsed time
of all objective functions. The results show that our proposed multiobjective

45
 900 M.O. m [25]
M.O. m2 (spiral)
M.O. m3 (spiral)
800 M.O. m1 (spiral)
S.O. f1 (spiral)
S.O. f2 (spiral)
700 S.O. f3 (spiral)
S.O. f4 (spiral)
S.O. f5 (spiral)
Elapsed Time(s)

S.O. f6 (spiral)
600

500

400

300

200

100
100 150 200 250 300 350
Total number of dumb nodes

Figure 16: Changes in elapsed time as node density changes

functions take less time to compute than the traditional method. As a result,
our proposed multiobjective localization scheme has a lower communication
cost.

7. Conclusion
In this paper, we present a range-free DV-Hop localization method in 2D
spiral space using Non-inertial Opposition-based Class Topper Optimization.
The suggested approach converts six single objective functions based on the
DV-Hop localization algorithm into three multiobjective functions in order
to reduce localization errors. The goal of multiobjective localization is to
reduce localization errors while also increasing positional precision. The re-
sults show that among all proposed objective functions, the multi-objective
function m1 has the least localization error and localization error variance

46
and also improves LA(Localization Accuracy) and coverage. The simula-
tion results reveal that among all single and multiobjective functions and
NASG-II [26], the multi-objective function m1 has greater placement accu-
racy. Development of the circular and spiral beacon nodes deployment of
the multi-hop-based approach in a 3D system and on the physical staging
ground is also one of the potential research topics to be worked out.

References
[1] J. V. Capella, J. C. Campelo, A. Bonastre, R. Ors, A reference model
for monitoring iot wsn-based applications, Sensors 16 (11) (2016) 1816.

[2] T. Najeh, H. Sassi, N. Liouane, A novel range free localization algorithm

in wireless sensor networks based on connectivity and genetic algorithms,
International Journal of Wireless Information Networks 25 (1) (2018)
88–97.

[3] Y. K. Joshi, M. Younis, Restoring connectivity in a resource constrained

wsn, Journal of Network and Computer Applications 66 (2016) 151–165.

[4] J.-H. Huh, K. Seo, An indoor location-based control system using blue-
tooth beacons for iot systems, Sensors 17 (12) (2017) 2917.

[5] W. Balid, H. Tafish, H. H. Refai, Intelligent vehicle counting and clas-

sification sensor for real-time traffic surveillance, IEEE Transactions on
Intelligent Transportation Systems 19 (6) (2017) 1784–1794.

[6] H. Mostafaei, M. U. Chowdhury, M. S. Obaidat, Border surveillance

with wsn systems in a distributed manner, IEEE Systems Journal 12 (4)
(2018) 3703–3712.

[7] A. Boubrima, W. Bechkit, H. Rivano, Optimal wsn deployment models

for air pollution monitoring, IEEE Transactions on Wireless Communi-
cations 16 (5) (2017) 2723–2735.

[8] V. Angayarkanni, V. Akshaya, S. Radha, Design of a compressive sens-

ing based fall detection system for elderly using wsn, Wireless Personal
Communications 98 (1) (2018) 421–437.

47
 [9] G. Song, D. Tam, Two novel dv-hop localization algorithms for randomly
deployed wireless sensor networks, International Journal of Distributed
Sensor Networks 11 (7) (2015) 187670.

[10] E. Q. Shahra, T. R. Sheltami, E. M. Shakshuki, A comparative study of

range-free and range-based localization protocols for wireless sensor net-
work: Using cooja simulator, in: Sensor Technology: Concepts, Method-
ologies, Tools, and Applications, IGI Global, 2020, pp. 1522–1537.

[11] A. Kumar, S. C. Gupta, A new initial centroid finding method

based on dissimilarity tree for k-means algorithm, arXiv preprint
arXiv:1509.03200.

[12] L. Gui, X. Zhang, Q. Ding, F. Shu, A. Wei, Reference anchor selection

and global optimized solution for dv-hop localization in wireless sensor
networks, Wireless Personal Communications 96 (4) (2017) 5995–6005.

[13] A. Kaur, P. Kumar, G. P. Gupta, A novel dv-hop algorithm based on

gauss-newton method, in: 2016 Fourth International Conference on Par-
allel, Distributed and Grid Computing (PDGC), IEEE, 2016, pp. 625–
629.

[14] L.-z. Zhao, X.-b. Wen, D. Li, Amorphous localization algorithm based on
bp artificial neural network, International Journal of Distributed Sensor
Networks 11 (7) (2015) 657241.

[15] S. Li, X. Ding, T. Yang, et al., Analysis of five typical localization

algorithms for wireless sensor networks, Wireless Sensor Network 7 (04)
(2015) 27.

[16] E. Tian, A non-ranging fusion location algorithm for concave regions,

Wireless Personal Communications (2022) 1–15.

[17] B. Risteska Stojkoska, V. Kirandziska, Improved mds-based algorithm

for nodes localization in wireless sensor networks, arXiv e-prints (2016)
arXiv–1606.

[18] A. Kaur, P. Kumar, G. P. Gupta, A weighted centroid localization algo-

rithm for randomly deployed wireless sensor networks, Journal of King
Saud University-Computer and Information Sciences 31 (1) (2019) 82–
91.

48
[19] J. Du, J.-F. Diouris, Y. Wang, A rssi-based parameter tracking strategy
for constrained position localization, EURASIP Journal on Advances in
Signal Processing 2017 (1) (2017) 1–10.

[20] X. Liang, H. Zhang, T. Lyu, H. Xiao, T. A. Gulliver, A novel time of

arrival estimation algorithm using an energy detector receiver in mmw
systems, EURASIP Journal on Advances in Signal Processing 2017 (1)
(2017) 1–13.

[21] M. Kamal, Comparison of localization algorithms for unmanned aerial

vehicles, in: Intelligent Technologies and Applications: Second Interna-
tional Conference, INTAP 2019, Bahawalpur, Pakistan, November 6–8,
2019, Revised Selected Papers, Vol. 1198, Springer Nature, 2020, p. 258.

[22] S. He, X. Dong, W.-S. Lu, Localization algorithms for asynchronous time
difference of arrival positioning systems, EURASIP Journal on Wireless
Communications and Networking 2017 (1) (2017) 1–14.

[23] P. Wang, F. Xue, H. Li, Z. Cui, L. Xie, J. Chen, A multi-objective dv-hop

localization algorithm based on nsga-ii in internet of things, Mathemat-
ics 7 (2) (2019) 184.

[24] Z. Sun, L. Tao, X. Wang, Z. Zhou, Localization algorithm in wireless sen-

sor networks based on multiobjective particle swarm optimization, Inter-
national Journal of Distributed Sensor Networks 11 (8) (2015) 716291.

[25] B. Peng, L. Li, An improved localization algorithm based on genetic

algorithm in wireless sensor networks, Cognitive Neurodynamics 9 (2)
(2015) 249–256.

[26] V. Kanwar, A. Kumar, Multiobjective optimization-based dv-hop local-

ization using nsga-ii algorithm for wireless sensor networks, International
Journal of Communication Systems 33 (11) (2020) e4431.

[27] G. Sharma, A. Kumar, Improved dv-hop localization algorithm us-

ing teaching learning based optimization for wireless sensor networks,
Telecommunication Systems 67 (2) (2018) 163–178.

[28] A. Kaur, P. Kumar, G. P. Gupta, Nature inspired algorithm-based im-

proved variants of dv-hop algorithm for randomly deployed 2d and 3d

49
 wireless sensor networks, Wireless Personal Communications 101 (1)
(2018) 567–582.

[29] V. Kanwar, A. Kumar, Dv-hop-based range-free localization algorithm

for wireless sensor network using runner-root optimization, The Journal
of Supercomputing (2020) 1–18.

[30] A. N. Shahbaz, H. Barati, A. Barati, Multipath routing through the

firefly algorithm and fuzzy logic in wireless sensor networks, Peer-to-
Peer Networking and Applications 14 (2) (2021) 541–558.

[31] A. Mosavifard, H. Barati, An energy-aware clustering and two-level rout-

ing method in wireless sensor networks, Computing 102 (7) (2020) 1653–
1671.

[32] E. Yousefpoor, H. Barati, A. Barati, A hierarchical secure data aggrega-

tion method using the dragonfly algorithm in wireless sensor networks,
Peer-to-Peer Networking and Applications 14 (4) (2021) 1917–1942.

[33] E. Hasheminejad, H. Barati, A reliable tree-based data aggregation

method in wireless sensor networks, Peer-to-Peer Networking and Ap-
plications 14 (2) (2021) 873–887.

[34] M. Naghibi, H. Barati, Shsda: secure hybrid structure data aggregation

method in wireless sensor networks, Journal of Ambient Intelligence and
Humanized Computing 12 (12) (2021) 10769–10788.

[35] Z. Hajipour, H. Barati, Eelrp: energy efficient layered routing protocol

in wireless sensor networks, Computing 103 (12) (2021) 2789–2809.

[36] S. S. Sharifi, H. Barati, A method for routing and data aggregating in

cluster-based wireless sensor networks, International Journal of Com-
munication Systems 34 (7) (2021) e4754.

[37] T. K. Mohanta, D. K. Das, Advanced localization algorithm for wireless

sensor networks using fractional order class topper optimization, The
Journal of Supercomputing (2022) 1–29.

[38] S. Kumar, D. Lobiyal, An advanced dv-hop localization algorithm

for wireless sensor networks, Wireless personal communications 71 (2)
(2013) 1365–1385.

50
[39] P. Das, D. K. Das, S. Dey, A new class topper optimization algorithm
with an application to data clustering, IEEE Transactions on Emerging
Topics in Computing.

[40] T. K. Mohanta, D. K. Das, Class topper optimization based improved lo-

calization algorithm in wireless sensor network, Wireless Personal Com-
munications (2021) 1–20.

[41] S. Dhargupta, M. Ghosh, S. Mirjalili, R. Sarkar, Selective opposition

based grey wolf optimization, Expert Systems with Applications 151
(2020) 113389.

[42] L. Kang, R.-S. Chen, N. Xiong, Y.-C. Chen, Y.-X. Hu, C.-M. Chen,
Selecting hyper-parameters of gaussian process regression based on non-
inertial particle swarm optimization in internet of things, IEEE Access
7 (2019) 59504–59513.

[43] L. Kang, R.-S. Chen, W. Cao, Y.-C. Chen, Non-inertial opposition-

based particle swarm optimization and its theoretical analysis for deep
learning applications, Applied Soft Computing 88 (2020) 106038.

51