Performance Analysis of Hybridization of [PIO-GSO] Algorithms in
Wireless Sensor Networks
K.THAMIZHMARAN, K.PRABU
Dept. of ECE, GCE, Bodinayakkanur, Theni, Tamilnadu, INDIA
PG & Research Dept. of CS, Sudharsan College of Arts & Science, INDIA.
Abstract: - In wireless sensor networks (WSN), clustering is treated as an energy efficient technique employed to
achieve augmenting network lifetime. But, the process of cluster head (CH) selection for stabilized network
operation and prolonged network lifetime remains a challenging issue in WSN. In this research, presents a novel
Hybridization of Pigeon Inspired with Glowworm Swarm Optimization (HPIGSO) algorithm based clustering
innovation in WSN. This innovative HPIGSO algorithm integrates the good characteristics of Pigeon Inspired
Optimization (PIO) algorithm and Glowworm Swarm Optimization (GSO) algorithm. The proposed algorithm
operates on three major stages namely initialization, cluster head selection and cluster construction. Once the nodes
are deployed, the initialization process takes place. Followed by, Base Station (BS) executes the HPIGSO
algorithm and selects the cluster heads effectively. Subsequently, nearby nodes joins the cluster head and becomes
cluster members, thereby cluster construction takes place. Finally, the cluster members send the data to cluster
heads which is then forwarded to the base station via inter-cluster communication. The performance of the
proposed HPIGSO method has been evaluated and compared with QOGSO, PIOA-DS, ALO, GOA and FFOA.
Finally the proposed HPIGSO algorithm provides prolonged the lifetime of WSN over the existing clustering
techniques
Keywords: Clustering, Augmenting Network lifetime, PIO, GSO, optimization algorithm.
Received: December 27, 2021. Revised: October 24, 2022. Accepted: December 2, 2022. Published: December 29, 2022.
1. Introduction
Recently, WSN has become a predominant one
which is highly efficient in real-time applications.
WSN observes the atmosphere and predicts the
modifications happening in target regions. Some of
the physical changes in the environment were
vibration, sound, pressure, humidity, intensity,
temperature, and so forth. The domain of WSN is
applied in diverse areas such as armed forces, habitat
monitoring [1], bio-medical sector, health
observation, smart home tracking as well as
inventory management system [2]. As an inclusion,
clustering [3] is developed which helps in dividing
the geographical region into tiny sectors. The main
purpose of applying clustering is to divide the load
equally to all nodes as head of the cluster, called as
CH. The election of CH is one of the major tasks
which helps in better data transmission. Practically,
the cluster can contain a CH with maximum number
of CM. The key objective of CH is to modify the
nodes within a cluster [4]. However, the proper CH
selection [5] with best potential is essential to
manage the network’s power-efficiency. Thus, the
meta-heuristic approaches were Computational
intelligence (CI) methods like Artificial Bee Colony
(ABC), Artificial Immune Systems (AIS),
Reinforcement Learning (RL), and Evolutionary
Algorithms (EA) have been applied to proceed
clustering task and to resolve NP-hard optimization
problem. Transmitting the data to a BS or sink from
the sensor node via optimal CH [6] is a complicated
operation. The optimal CH selection process results
in minimum power consumption, latency, distance
etc. When compared with all other methods, the
optimal CH election process in WSN remains a
challenging issue. Various studies have been
developed to determine the optimal CH selection
process in WSN. Mehra et al. [7] presented a Fuzzy-
Based balanced cost CH Selection method (FBECS)
which has been constrained with residual energy
(RE), distance and node density are considered to be
the input for Fuzzy Inference System (FIS). For the
selection of optimal CH, the Eligibility index has
been determined for each node. Priyadarshini and
Sivakumar [8] applied load balancing by triggering
the Adelson-Velskii and Landis (AVL) tree rotation
clustering approaches. The developers have divided
the unique area network into massive clusters by
novel and improved K-means clustering methods.
Mann and Singh [9] projected an improved ABC
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2022.21.40
K. Thamizhmaran, K. Prabu
E-ISSN: 2224-2872
336
Volume 21, 2022
with optimal solution search function for improvising
system efficiency. Furthermore, a population
sampling algorithm has been employed for a
student’s distribution. It is mainly used for enhancing
the global convergence of deployed meta-heuristic
approaches. Elhabyan et al. [10] established a Pareto
optimization-relied method for handling the problems
involved in finding best network configuration. In
order to estimate the efficiency, the proposed
technique has assumed a few metrics such as the
number of CHs, the number of the clustered nodes,
and link supremacy over CMs. A fractional ABC
dependent based multi-objective CH selection
(FABCMOCHS) approach has been presented as an
energy effective clustering model to expand the
sensor nodes’ duration with improved network power
[11]. FABC-MOCHS is mainly utilized for managing
the convergence present in ABC by adding fitness
function (FF) along with latency, travelling distance
and power application to reduce the problem. A
combination of ACO and ABC model-based
clustering scheme (ACO-ABCA-CS) was presented
for effective CH election under the mutual prevention
of limitations [12]. The problem of stagnation in
ACO and delayed convergence of ABC can be solved
by mutual modification in exploitation and
exploration phases. A dynamic scour bee-based CS
(DSB-CS) has been projected for increasing the scout
bee and maintaining the count of active nodes as well
as CH power in a system [13]. It is highly applicable
due to the advantages of ABC and FABC for
increasing the duration of a network and power by
using the best CH election approach. The
concatenated Simulated Annealing as well as
differential evolution-based CH selection (SADE-
CHS) model has been established to enhance the
power effectiveness by using clustering [14]. The
SADE-CHS is mainly used for eliminating the
overload of sensor nodes which is related with CH, as
it is a major reason for immediate death of sensor
nodes that leads to improper CH election process. It
highly focuses on the network extension by removing
the possibility of premature death of CH. A
combined PSO as well as HSA-based CH selection
approach has been applied to retain the energy
balance as well as network duration [15]. PSO-HSA-
CHS method was presented by integrating the
dynamic ability of PSO and the higher exploring
potential of HSA meta-heuristic approach for
selecting the best CH in a system. To maximize the
network lifetime, this paper presents a new HPIGSO
algorithm based clustering innovative in WSN.
2. The Proposed HPIGSO Algorithm
Fig. 1 depicts the workflow of the presented HPIGSO
algorithm. Once the nodes are deployed, the
initialization process takes place. Followed by, BS
executes the HPIGSO algorithm and selects the CHs
effectively. Subsequently, nearby nodes join the CH
and become CMs, thereby cluster construction takes
place. Finally, the CMs send the data to CHs which is
then forwarded to BS via inter-cluster
communication.
Fig. 1. Block diagram of HPIGSO algorithm
2.1. PIO Algorithm
The PIO algorithm is stimulated from the homing
characteristics of pigeons. A pigeon is a familiar bird
commonly employed for message passing by
Egyptians, also by military forces. For idealizing the
homing features of pigeons, 2 operators were
developed by utilizing a few principles: Map and
compass operator: the pigeon’s predict the earth
using magneto reception to design the map mentally.
It considers the altitude of the sun as a range to alter
the way. Landmark operator: If the pigeons fly near
to a target, it relies on landmarks. When it is well-
known with the landmarks, then it flies to the target.
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2022.21.40
K. Thamizhmaran, K. Prabu
E-ISSN: 2224-2872
337
Volume 21, 2022
When it is distant from the target and unknown to the
landmarks, it is following the pigeons that are well-
known through the landmarks.
2.1.1. Map and compass operator
It determines the position and velocity of
pigeon in a ‐dimension search space which are
updated for all iterations. A novel position and
velocity of pigeon at the ‐th iteration is
computed with the subsequent equations:
where implies map and extent factor, rand is an
arbitrary number, and is the present global
optimal position, and that is attained by relating
every the positions between each pigeon.
While an optimal position of each pigeon is assured
by utilizing map and compass. With relating each
domain position, it can be apparent which the right
centred pigeon’s location is optimized. All the
pigeons alter the flying direction by subsequent and
particular pigeons based on Eq. (1) that are illustrated
by dark arrows. A thin arrow refers the former flying
way relative to in Eq. (2). A vector
value of these 2 arrows is its subsequently flying
way.
2.1.2. Landmark operator
Here, maximum pigeons are reduced by in all
generations. But, the pigeons are distant from the
target, and it is different through the landmarks.
Assume be the center of any pigeon’s position
at the th iteration, and assume all pigeons are flying
directly to the target. A position updating rule to
pigeon at the th iteration is provided by:
where is the efficiency of a
pigeon. Followed by, it selects
. To maximize optimization problems, we
have to select . To all
individuals pigeon, a better position of the Nc‐th
iteration is indicated with , and
. The center of each pigeon
is their aim in all iterations. The half of each pigeon
which are distant from their target is following the
pigeons which are near to their target which also
implies that 2 pigeons can be at the similar position.
A pigeon that is near to their target (a pigeon in
encircle) would fly to the required place with
sufficient speed.
2.2. Conventional GSO algorithm
Here, a swarm of glowworms are organized in a
random manner on the solution space. A brighter
individual implies an optimal position. Utilizing a
probabilistic method, all agents are inspired by a
neighbour with better luciferin intensity in local
decision fields. A density of a glow worm's neighbors
influences its decision radius and defines the size of
its local decision field: if the neighbor density is
minimum, a local decision field is extended; else, it
may be limited to enable the swarms to divide into
lesser groups. The above procedure is followed till
reaching the termination criteria. Currently,
maximum individuals collect brighter glowworms.
Followed by, the GSO contains 5 important stages:
luciferin-update stage, neighborhood select stage,
moving probability computer stage, movement stage,
and decision radius added stage.
2.2.1. Luciferin Update stage
A luciferin update is based on the FF of preceding
luciferin value, and regulation is provided by
where, indicates the luciferin measure of
glowworm at time represents the luciferin
decompose constant, implies the luciferin
improvement constant; is the
position of glowworm at time , and
signifies the value of the fitness
at glowworm ’s position location at time
2.2.2. Neighborhood Select Phase.
The neighbors of glowworm at time have
the brighter ones and are expressed as
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2022.21.40
K. Thamizhmaran, K. Prabu
E-ISSN: 2224-2872
338
Volume 21, 2022
where, signifies the using Euclidean distance
among glowworms and at time , and
signifies a decision radius of glowworms at time
2.2.3. Moving Probability Computer Phase
The glowworm utilizes a possibility rule in order to
send other glowworms containing superior luciferin
level. A possibility of glowworm move
towards the neighbor is expressed by:
2.2.4. Movement Phase
Assume glowworm chooses glowworm
with ; the discrete‐time method of
glowworm is offered by (9)
where, signifies the Euclidean norm operator,
and is the step‐size.
2.2.5. Decision Radius Update Phase
During all updates, the decision radius of glowworm
is provided as following:
where, shows a constant, indicates the sensory
radius of glowworm , as well as refers to an
attribute for balancing the adjacent value.
2.3. Hybridization of PIO and GSO
algorithms
This section discusses the HPIGSO algorithm, which
integrates PIO and GSO algorithms. Actually, the
GSO algorithm has the ability to deal with non-
linear, multimodal issues. But it gets stuck to solve
high dimensional problems and fails to convergence
faster. At the same time, the PIO algorithm has the
ability of faster convergence.
Fig. 2. Flowchart of HPIGSO algorithm
At this point, the validation of the objective function
is carried out initially and the evaluated fitness gets
sorted. Then, find the optimal five fitness values and
choose an index. If it is higher than five, execute the
PIO algorithm update, otherwise execute the GSO
update. Through the hybridization, the multi-
objective CHs election leads to minimum delay, and
maximum energy saving. Besides, the negative
searching ability is discarded by the HPIGSO
technique, whereas the enhanced searching capability
can be used for faster convergence. Therefore, the
HPIGSO algorithm has achieved a better CH
selection process. The processes involved in the
HPIGSO method are illustrated in Fig. 2.
3. Performance Validation
This section investigates the experimental outcome of
the HPIGSO technique under different dimensions.
The presented HPIGSO technique has been simulated
using MATLAB. Moreover, a set of measures
applied to investigate the results of the network
lifetime, network stability, count of active nodes as
well as number of inactive nodes. The parameter
settings involved in the experimentation are given in
Table 1.
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2022.21.40
K. Thamizhmaran, K. Prabu
E-ISSN: 2224-2872
339
Volume 21, 2022
Table 1 Parameter Settings
Parameters
Values
Network Size
100 x 100 m2,
500 x 500 m2
No. of Nodes (N)
100, 300, 500
No. of BS
1
Initial energy (E0)
0.5
Energy fraction for intermediate nodes
() and advanced nodes ()
1, 2
No. of gateway nodes (m) and advanced
nodes fraction (m0)
m=0.1,
m0=0.2
Energy need to transmit and receive Eelec
50nJ/bit
Threshold distance (d0)
80m
Amplifying power required for smaller
distance dd0 (Eefs)
10pJ/bit/m2
Amplifying power required for smaller
distance dd0 (Emp)
0.0013
pJ/bit/m4
Energy utilization incurred when data
aggregation (Eda)
5
nJ/bit/signal
Data packet Size
2000 bits
Population size (P)
100
Selection Method
Rank
Selection
Method
No. of generation
30
No. of runs
20
3.1. Alive Nodes Analysis of HPIGSO
technique on varying Node Count
Fig. 3 portrays the results analysis of HPIGSO
technique interms of count of alive nodes within the
node count of 100. The figure exhibited that the
FFOA model has resulted in a least number of alive
nodes over the compared methods. In line with that,
the GOA and ALO algorithms have led to a higher
and closer number of alive nodes. Continuing with
this, the PIOA-DS algorithm has started to modify
network lifetime with a somewhat supreme number
in alive nodes. Simultaneously, the QOGSO
algorithm has tried to exhibit near optimal results
with the higher count of alive nodes. At last, the
proposed HPIGSO technique has shown better
outcome by attaining high alive nodes. For example,
under the execution round of 1000, the HPIGSO
technique has attained higher count of 51 alive nodes
whereas the lowest of 42, 32, 16, 12 and 4 alive
nodes are attained by QOGSO, PIOA-DS, ALO,
GOA and FFOA algorithms. Likewise, under the
execution round of 2000,the HPIGSO model has
displayed a massive number of 18 alive nodes
whereas the minimum of 14, 11, 6, 4 and 0 alive
nodes are attained by QOGSO, PIOA-DS, ALO,
GOA and FFOA algorithms.
Fig. 3.Alive node analysis of HPIGSO technique
under 100 nodes
3.2. Dead Nodes Analysis of HPIGSO
technique on varying Node Count
Fig. 4 demonstrates the results analysis of the
HPIGSO approach by means of number of inactive
nodes within node value of 100. The diagram
illustrates that the FFOA technique has provided a
reasonable number of dead nodes than the previous
modules. On continuing this, the GOA and ALO
models have resulted in fewer and fewer dead nodes.
Similarly, the PIOA-DS algorithm has been
initialized to showcase better network lifetime with a
lower normal number of inactive nodes. Meanwhile,
the QOGSO approach has attempted to show near
best results with a minimum number of dead nodes.
Finally, the projected HPIGSO approach has
exhibited moderate outcome by accomplishing lower
count of dead nodes. For sample, under execution
round of 1000, the HPIGSO model has achieved
minimum number of 49 dead nodes while the
maximum of 58, 68, 84, 88 and 96 dead nodes are
reached by QOGSO, PIOA-DS, ALO, GOA and
FFOA methodologies. In line with this, under the
execution round of 2000,the HPIGSO scheme has
depicted a lower number of 82 dead nodes and the
maximum of 86, 89, 94, 96 and 100 dead nodes are
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2022.21.40
K. Thamizhmaran, K. Prabu
E-ISSN: 2224-2872
340
Volume 21, 2022
achieved by QOGSO, PIOA-DS, ALO, GOA and
FFOA algorithms.
Fig. 4. Dead node analysis of HPIGSO technique
under 100 nodes
3.3. Network Lifetime (Alive) Analysis of
HPIGSO Algorithm on varying Node Count
Fig. 5 showcased the competing analysis of the
HPIGSO method with respect to stability duration,
HND and network lifespan. The figure illustrates that
the HPIGSO model has achieved better network
stability than the related technologies. Followed by,
the HPIGSO technology has delayed the HND to a
higher extent than previous models. Here, the
HPIGSO scheme has demonstrated a maximum
network lifetime. By seeking into the reached results,
it is assured that the HPIGSO model has
accomplished supreme function than alternate
models.
Fig.5. Network lifetime (alive) analysis of the
HPIGSO method
The above mentioned figures indicated that the
HPIGSO algorithm has achieved maximum network
lifetime denoting the network stability, and stability
period.
4. Conclusion
This paper has innovated a new HPIGSO algorithm
based clustering technique in WSN, which integrates
the characteristics of PIO and GSO algorithms. The
proposed algorithm operates on three major stages
namely initialization, CH selection and cluster
construction. Once the nodes are deployed, the
initialization process takes place. Followed by, BS
executes the HPIGSO algorithm and selects the CHs
effectively. Subsequently, nearby nodes joins the CH
and becomes CMs, thereby cluster construction takes
place. Finally, the CMs send the data to CHs which is
then forwarded to BS via inter-cluster
communication. The proposed HPIGSO algorithm
involves an objective function using residual energy,
distance and energy. The proposed method has the
ability to select the CHs in an optimal way; thereby
network (alive) lifetime can be maximized. An
elaborate experimental validation takes place and the
results of HPIGSO algorithm has attained maximum
network lifetime compared to QOGSO, PIOA-DS,
ALO, GOA and FFOA. In future, the network
lifetime can be further increased by the use of data
aggregation mechanisms.
References
[1] Madhuri, B., Aniruddha, G., Rahul, R., 2013.
Identification and classification of flood prone
areas using AHP, GIS and GPS. J. Disaster
Adv. 6, 120–131.
[2] Sherifi, I., Baholli, I., 2015. Information
Systems and Online Media in Albania:
Challenges and Expectations, The 8th
International Conference on Economic
Sciences, Vienna, Austria, 8–15
[3] Zhu, E., Ma, R., 2018. An effective partitioned
clustering algorithm based on new clustering
validity index. Appl. Soft Comput. 71, 608–
621.
[4] K.Thamizhmaran, M.Anitha and
Alamelunachippan (2018) “Reduced End-To-
End Delay for MANETS using SHSP-
EA3ACK Algorithm”, Journal on
Communication Engineering and System, Vol.
7, No. 3, pp. 8-15.
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2022.21.40
K. Thamizhmaran, K. Prabu
E-ISSN: 2224-2872
341
Volume 21, 2022
[5] K.Thamizhmaran (2018) "Performance
Analysis of On-Demand Routing Protocols in
MANETs Using (SHSP-EA3ACK)
Algorithms", Ciit, International Journal of
Digital Signal Processing, Vol. 10, No. 3, pp.
50-54.
[6] K.Thamizhmaran (2020), “EE-ATPSP –
Evaluation Node Life Time for WSNs”, i-
manager’s Journal on Wireless
Communication Network Vol. 8, No. 4, pp.
27-35.
[7] Mehra, P.S., Alam, M.N.D., 2017. Zonal based
approach for clustering in heterogeneous
WSN. Int. J. Inf. Technol., 1–9
[8] Priyadarshini, R.R., Sivakumar, N., 2018.
Cluster head selection based on minimum
connected dominating set and bi-partite
inspired methodology for energy conservation
in WSNs. J. King Saud Univ. Comp. Inf. Sci.
Available online 19.
[9] Kumar, R., & Kumar, D. (2015). Multi-
objective fractional artificial bee colony
algorithm to energy aware routing protocol in
wireless sensor network. Wireless Networks,
22(5), 1461–1474.
[10] Sengathir, J. (2018). A hybrid ant colony and
artificial bee colony optimization algorithm-
based cluster head selection for IoT. Procedia
Computer Science, 143(1), 360–366.
[11] K. Thamizhmaran, R. Santosh Kumar Mahto,
and V. Sanjesh Kumar Tripathi (2012)
“Performance Analysis of Secure Routing
Protocols in MANET”, International Journal of
Advanced Research in Computer and
Communication Engineering, Vol. 1, No. 9,
pp. 651-654.
[12] Akshayadevi Arivazhagan, K.Thamizhmaran
and N.Thamilselvi (2015) “Performance
Comparison of on Demand Routing Protocols
under Back whole For MANET”, Advance
Research in Computer science and software
Engineering, Vol. 5, No. 3, pp. 407 – 411.
[13] K.Prabu and K.Thamizhmaran (2016) “Cluster
Head Selection Techniques and Algorithm for
Mobile Ad-hoc Networks (MANETS)”,
Advance Research in Computer science and
software Engg, Vol.6, No. 7, pp. 169-173.
[14] K.Thamizhmaran, M.Anitha and
Alamelunachippan (2017) “Comparison and
Parameter Adjustment of Topology Based (S-
EA3ACK) for MANETs”, International
Journal of Control Theory and Application,
Vol. 10, No. 30, pp. 423-436.
[15] K.Thamizhmaran, M.Anitha and
Alamelunachippan (2017) “Performance
Analysis of On-demand Routing Protocol for
MANET Using EA3ACK Algorithm”,
International Journal of Mobile Network
Design and Innovation (Inderscience), Vol. 7,
No. 2, pp. 88-100.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the Creative
Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en_US
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2022.21.40
K. Thamizhmaran, K. Prabu
E-ISSN: 2224-2872
342
Volume 21, 2022
