Chapter 1

INTRODUCTION

1.1 WIRELESS SENSOR NETWORKS

Wireless Sensor Networks (WSNs) have revolutionized the way we monitor and
interact with the environment. These networks comprise spatially distributed sensor
nodes that are designed to monitor physical or environmental conditions such as
temperature, pressure, humidity, or motion and transmit the collected data to a central
location for processing [1]. The versatility and applicability of WSNs have made them
indispensable in various domains, including smart agriculture, healthcare, industrial
automation, military surveillance, and disaster management. However, the efficient
operation of WSNs is hampered by several constraints, the most critical of which is
energy consumption. This report delves into the application of swarm optimization
techniques to address the issue of energy expenditure in WSNs, focusing on
enhancing their operational efficiency and extending their lifespan. Wireless Sensor
Networks (WSNs) represent a critical technological advancement in the field of
wireless communication and distributed computing [2]. They consist of numerous
small, energy-constrained sensor nodes that can sense, process, and communicate
data. These networks find applications in diverse areas, such as environmental
monitoring, healthcare, military surveillance, and smart cities. However, their
effectiveness is often hindered by energy constraints. Sensor nodes, which are
typically powered by batteries, have limited energy resources. As a result, one of the
primary challenges in deploying WSNs is minimizing energy expenditure while
ensuring reliable and efficient data transmission.

A Wireless Sensor Network (WSN) is a network of small, low-power, and

inexpensive sensor nodes deployed over a geographical area. Each node in a WSN is
equipped with sensors, a microcontroller, a communication module, and a power
source [3]. The primary objective of a WSN is to collect data from the environment,
process it, and transmit it to a central location for further analysis. WSNs are widely
used in applications such as agricultural monitoring, disaster management, military
surveillance, and healthcare systems. The architecture of a WSN typically consists of
three main components: sensor nodes, sink nodes (or base stations), and a

1
communication network. Sensor nodes are responsible for sensing, processing, and
transmitting data, while sink nodes aggregate data from multiple sensors for further
analysis. The communication network, often based on wireless protocols like Zigbee
or Bluetooth, connects these components [4]. However, due to their small size and
low cost, sensor nodes are constrained in terms of processing power, memory, and
battery life, making energy optimization a key concern.

1.1.1 CLASSIFICATIONS OF WIRELESS SENSOR NETWORKS:

The cognitive radio networks can be classified into different types, based on different
parameters used, such as topology, mobility, application, and communication
methods, as shown below:

1.1.1.1 Classification of WSNs based on the Node Deployment (Topology):

a) Flat WSNs: All nodes have equal roles and responsibilities. Every sensor node
collects data and sends it to the base station (BS) or sink.
b) Hierarchical (Clustered) WSNs: Nodes are organized in clusters. Each cluster has
a Cluster Head (CH) responsible for aggregating and forwarding data to the BS [5].
This structure enhances scalability and energy efficiency.

1.1.1.2 Classification of WSNs based on Mobility:

a) Static WSNs: Sensor nodes remain stationary after deployment. These are the
most common due to simpler communication strategies.
b) Mobile WSNs: Sensor nodes or the base station can move. Mobility helps in
dynamic environments like wildlife monitoring or battlefield surveillance.

1.1.1.3 Classification of WSNs based on Power Source:

a) Battery-Powered WSNs: Most common, where nodes operate until the battery
drains.
b) Energy-Harvesting WSNs: Nodes harvest energy from solar, wind, or vibrations,
enabling longer or even perpetual operation [6].

2
 Flat WSNs Hierarchical
(Clustered)
WSNs

Static WSNs
Based on Node Deployment

Based on
WSNs
Mobility

Mobile
Based on Power Source WSNs

Battery- Energy-
powered harvesting

Fig 1.1: Classification of Wireless Sensor Networks

1.1.1.4 Classification of WSNs based on communication type utilized during
transmission:
a) Single-hop: Figure 1.2 shows that in single-hop communication, each sensor node
sends its data directly to the base station, which is simple and results in lower latency
but consumes high energy for distant nodes and limits scalability [7]. The
communication protocols are straightforward due to the absence of routing through
intermediate nodes.

b) Multi-hop:
Figure 1.3 shows that multi-hop communication allows nodes to transmit data through
intermediate nodes until it reaches the base station, significantly improving energy
efficiency and scalability in larger networks [8]. However, this method introduces
more complexity in routing and may increase data transmission delays.

The choice between single-hop and multi-hop communication in WSNs depends on

various factors, including network size, node distribution, energy constraints, and
application requirements [9]. Single-hop communication is suitable for small-scale

3
networks with nodes in close proximity to the base station, offering low latency and
simplicity. In contrast, multi-hop communication is more appropriate for large-scale
deployments, providing better energy efficiency and scalability at the cost of
increased complexity [10].

1.1.2 CHARACTERISTICS OF WIRELESS SENSOR

NETWORKS
a) Scalability:
WSNs are highly scalable, meaning they can support a large number of sensor nodes
distributed over vast areas [11]. This allows them to adapt to different application
sizes, from small-scale indoor monitoring to wide-scale environmental sensing.

b) Energy Efficiency:
Energy conservation is a critical feature of WSNs since sensor nodes typically operate
on limited battery power. Efficient communication protocols, sleep scheduling, and
energy-aware routing are used to prolong network lifetime.

c) Self-Organization:
Sensor nodes in WSNs can self-organize and configure themselves without
centralized control. This adaptability enables the network to recover from node
failures and adjust to changes in the environment or topology [12].

d) Data-Centric Communication:
WSNs focus on the data being collected rather than the nodes themselves. Sensor
nodes often use data aggregation and compression to minimize redundancy and
reduce the total amount of data transmitted, thereby saving energy.

1.1.3 FUNCTIONS OF WIRELESS SENSOR NETWORKS

1.1.3.1 Environmental Sensing: WSNs gather data from physical environments
using sensors [13]. This includes temperature, humidity, pressure, motion, light, etc.
These sensors are deployed in diverse terrains (forests, oceans, cities) to monitor
changes and phenomena in real-time.

4
 Environmental
Sensing
Data Data
Processin Aggregatio
g and n and
WSNs
Computati Fusion
on

Wireless Actuation and

Communication Decision-Making

Fig 1.2: WSNs

1.1.3.2 Data Processing and Computation: Once the data is sensed, nodes locally
process the data using embedded microcontrollers. This reduces redundancy and
saves energy. Processing may include filtering, compression, and even preliminary
analytics to extract meaningful insights before transmission [14].

1.1.3.3 Wireless Communication: Nodes transmit data wirelessly either directly to the
base station (single-hop) or through intermediate nodes (multi-hop) [15].
Communication protocols are optimized for low power, reliability, and efficiency.
This step is vital for ensuring scalability and coverage in large-scale WSN
deployments.

1.1.3.4 Data Aggregation and Fusion: To reduce the amount of data sent and conserve
bandwidth, aggregation techniques are used. Nodes combine similar data from
different sources to eliminate duplicates and summarize it meaningfully [16]. Data
fusion can further integrate different types of sensor data for complex decision-
making.

5
 Fig 1.3: Functions of WSNs

1.1.4 CHALLENGES OF WIRELESS SENSOR NETWORKS:

Wireless Sensor Networks (WSNs) face several critical challenges that hinder their
efficient operation and widespread deployment. One of the foremost issues is energy
constraint, as sensor nodes are battery-operated and often deployed in remote or
inaccessible areas, making battery replacement impractical. This necessitates the
development of energy-efficient protocols to prolong network lifetime [17]. Another
major challenge is scalability and network management, as WSNs may consist of
hundreds or thousands of nodes, requiring efficient routing, load balancing, and
synchronization strategies [18]. Additionally, data aggregation is essential to reduce
redundancy and communication overhead, but it must be done carefully to avoid data
loss or duplication. Security and privacy pose significant concerns, given the wireless
medium’s vulnerability to attacks such as eavesdropping, data tampering, and denial-
of-service. Lightweight but effective security mechanisms are essential in such
resource-constrained environments. Lastly, communication reliability is a persistent
issue due to limited bandwidth, packet loss, interference, and dynamic environmental
conditions. Ensuring consistent and accurate data transmission in such a setting
remains a complex task, requiring robust and adaptive communication protocols [19].
These challenges must be addressed comprehensively to unlock the full potential of
WSNs in real-world applications.

The advantages of Wireless Sensor Networks (WSNs) are as follows:

a) Remote Monitoring and Automation: WSNs enable real-time monitoring of

6
remote or hazardous environments such as volcanic zones, military borders, and
industrial plants [20]. They allow automated data collection without human
intervention, enhancing safety and reducing operational costs.

b) Scalability and Flexibility: WSNs are highly scalable and can be deployed in
diverse sizes and topologies (star, mesh, tree). Their self-configuring nature allows
them to adapt to changes in the network, such as node failure or addition, without
manual reconfiguration.

c) Energy-Efficient Communication: Advanced protocols in WSNs focus on

minimizing energy usage by optimizing routing paths, cluster-head selection, and
sleep scheduling [21]. This energy-awareness prolongs the network's operational life,
especially in battery-dependent scenarios.

d) Integration with IoT and Smart Systems: WSNs can seamlessly integrate with
Internet of Things (IoT) platforms, enabling smart applications in healthcare (e.g.,
patient monitoring), agriculture (e.g., soil moisture sensing), and smart cities (e.g.,
traffic and pollution monitoring), boosting efficiency and responsiveness [22].

1.2 SWARM OPTIMIZATION TECHNIQUE:

1.2.1 INTRODUCTION:
Swarm Optimization is a nature-inspired computational technique that mimics the
collective behavior of decentralized, self-organized systems found in nature, such as
bird flocks, fish schools, or insect colonies. These algorithms use a population of
simple agents, often called particles or individuals, that interact locally with one
another and with their environment to find optimal solutions in complex search spaces
[23]. Each agent updates its position based on its own experience and the experience
of its neighbors, promoting both exploration and exploitation of the solution space
[24] . Swarm optimization techniques, such as Particle Swarm Optimization (PSO),
Ant Colony Optimization (ACO), and African Vultures Optimization Algorithm
(AVOA), are widely used for solving real-world optimization problems in areas like
cloud computing, wireless sensor networks, and machine learning due to their
flexibility, scalability, and ability to avoid local minima.

7
1.2.2 CLASSIFICATIONS OF SWARM OPTIMIZATION TECHNIQUES:
Swarm Optimization techniques can be classified based on the natural phenomena
they are inspired by. The main categories include:

a) Particle-Based Algorithms: Inspired by the social behavior of bird flocking or fish

schooling.
Example: Particle Swarm Optimization (PSO) – agents (particles) adjust their
positions based on personal and global best positions.
b) Ant-Based Algorithms: Modeled after the foraging behavior of ants using
pheromone trails.
Example: Ant Colony Optimization (ACO) – artificial ants build solutions and
reinforce good paths with virtual pheromones.
c) Bee-Inspired Algorithms: Based on the food foraging and communication of
honeybee colonies.
Example: Artificial Bee Colony (ABC) – uses employed, onlooker, and scout bees to
explore and exploit food sources.

d) Mammal/Bird-Inspired Algorithms: Derived from the hunting or migration

behavior of specific animals [25] .
Examples:
Grey Wolf Optimizer (GWO) – inspired by leadership and hunting in grey wolves.
African Vultures Optimization Algorithm (AVOA) – mimics the soaring and
scavenging behavior of vultures.

1.2.3 CHARACTERISTICS OF SWARM OPTIMIZATION TECHNIQUE:

a) Population-Based Search:
These techniques use a group of agents (particles, ants, etc.) that explore the solution
space collectively, increasing the chances of finding global optima [26].

b) Decentralized Control:
There is no central authority; each agent operates independently based on local
information and simple rules, mimicking natural swarms.

8
c) Self-Organization:
Agents adapt and evolve their behavior over time through local interactions, leading
to the emergence of intelligent global behavior.

d) Stochastic and Adaptive:

Swarm algorithms incorporate randomness to explore diverse solutions and adapt
dynamically to the problem landscape.

e) Flexible and Robust:

They perform well in dynamic and uncertain environments and can be applied to a
wide range of complex optimization problems without needing gradient information
[27].

1.2.4 CHALLENGES IN IMPLEMENTING SWARM OPTIMIZATION

TECHNIQUES:
While swarm optimization techniques offer significant advantages, their
implementation in WSNs is not without challenges. These include:

1.2.4.1 Scalability: Adapting algorithms to large-scale networks with hundreds or

thousands of nodes. Swarm optimization techniques must balance computational
efficiency with solution quality to handle the complexity of large networks.

1.2.4.2 Dynamic Environment: Handling changes in network topology due to node

failures, mobility, or environmental factors [28]. Algorithms must be robust and
adaptive to maintain performance in dynamic conditions.

1.2.4.3 Computaional Complexity: Reducing the computational overhead to match

the limited processing power of sensor nodes. Lightweight implementations of swarm
optimization algorithms are essential for practical deployment in WSNs.

1.2.4.4 Convergence Speed: Ensuring algorithms converge to an optimal solution

within a reasonable timeframe. Slow convergence can lead to suboptimal solutions
and increased energy consumption.

9
1.3REVIEW OF SWARM OPTIMIZATION TECHNIQUES:
1.3.1 PARTICLE SWARM OPTIMIZATION (PSO):
Particle Swarm Optimization (PSO) is a nature-inspired, population-based
optimization algorithm that draws inspiration from the collective and social behaviors
observed in bird flocking, fish schooling, and swarming insects [29]. It was
introduced by Kennedy and Eberhart in 1995 and has since gained wide popularity for
solving complex optimization problems due to its simplicity, adaptability, and
effectiveness. In PSO, each individual in the population, known as a "particle,"
represents a potential solution to the optimization problem. These particles move
through the solution space by updating their positions based on their own experience
(personal best) and the experience of their neighbors or the entire swarm (global best)
[30]. This behavior enables the swarm to converge toward an optimal or near-optimal
solution over successive iterations. The movement and decision-making of the
particles are influenced by social sharing of information, which mimics the group
intelligence found in natural swarms.

Fig. 1.4 Particles (representing candidate solutions) movement

a) Grey Circle (ith Particle – Current Position): This is the current position of a
particle in the search space. Think of it as a possible solution that the algorithm is
evaluating.

b) Yellow Circle (Best Previous Position – Piᵏ): This is the personal best position
that this particle has visited so far. It remembers where it previously achieved the best
performance (like finding the best food source in a flock) [31].

c) Green Circle (Best Global Particle – Pgᵏ): This represents the global best
10
position found by the entire swarm. It's the best solution any particle has found up to
this point.

d) Blue Circle (New Position – Xiᵏ⁺¹): This is where the particle will move next. The
position is updated based on a combination of three factors:

Inertia (ωVᵏᵢ): The tendency to keep moving in the same direction.

Cognitive Component (c₁r₁(Pᵏᵢ − Xᵏᵢ)): The pull toward the particle's personal best.

Social Component (c₂r₂(Pgᵏ − Xᵏᵢ)): The pull toward the global best [32].

e) Blue Birds: These visually represent particles "flying" toward better solutions,
similar to how birds flock together in nature based on shared information.

1.3.1.1 Glide Toward Optimization: How PSO Learns and Adapts:

Particle Swarm Optimization (PSO) mimics the social behavior of birds flocking or
fish schooling to find optimal solutions in a search space [33]. Each particle
represents a potential solution and "flies" through the space, influenced by its own
experience and that of its peers. The working of PSO can be broken down into several
key phases:

a) Initialization Phase: A swarm of particles is randomly initialized with positions

and velocities. Each particle has knowledge of its position and evaluates its quality
using a fitness function [34].

b) Memory of the Best: As particles move, they remember two key pieces of
information, i.e., their personal best position (Pᵢᵏ) – the best solution they have
encountered and the global best position (Pgᵏ) – the best solution found by any
particle in the swarm.

c) Velocity Update – The Driving Force: Each particle updates its velocity
considering, i.e., its current direction (inertia), its attraction toward its own best
position (cognitive component), and its attraction toward the swarm’s best position
(social component) [35]. This blend helps balance exploration (searching new areas)

11
and exploitation (refining known good solutions).

d) Position Update – The Next Move: The new velocity guides the particle to a new
position in the search space [36]. This movement simulates learning from experience
and peer influence.

e) Evaluation and Iteration: The new position is evaluated, and if it improves the
fitness, it updates the personal and/or global best. The process continues for a number
of iterations until an optimal or satisfactory solution is reached.

1.3.2 CRAYFISH OPTIMIZATION ALGORITHM (COA):

The Crayfish Optimization Algorithm (COA) is a relatively new bio-inspired
metaheuristic optimization method, developed by mimicking the natural behavior of
crayfish in their aquatic environments. Crayfish exhibit intelligent strategies such as
searching for food, escaping predators, and locating shelters by following dynamic
and memory-based movement patterns. These behaviors have been mathematically
modeled to help solve optimization problems effectively, particularly those involving
high-dimensional and complex search spaces [37]. COA models a population of
candidate solutions as "crayfish," each navigating the search space by adapting its
direction based on personal experience and the influence of stronger individuals in the
group. The goal is to strike a balance between exploration (searching new regions)
and exploitation (refining good solutions). Crayfish use memory to recall successful
positions and adjust their path toward those, while also considering repulsion from
less favorable areas.

a) Crayfish (Candidate Solutions): Each crayfish represents a potential solution in

the search space. Their positions symbolize different points being evaluated for
optimality.

b) Current Position (Xi): This is the current location of a crayfish in the search space
—analogous to the present solution being explored.

c) Best Personal Position (Piᵏ): It represents the best position this crayfish has
visited so far, based on past fitness evaluations. The crayfish uses memory to recall

12
and gravitate toward this position.

d) Best Global Position (Pgᵏ): This is the best solution found by the entire group,
and it acts as a social guide that influences all other crayfish [38].

e) Motion Vector: The movement of a crayfish toward a new position is determined

by three key components:

 Memory Influence: A pull toward its own best-known position.

 Social Influence: A pull toward the best-known position in the group (elite
crayfish).
 Random Perturbation: A stochastic component that helps avoid local optima and
encourages exploration.


f) Boundary Constraints: If a crayfish exceeds the boundaries of the search space,

its position is corrected and pulled back within the feasible region.

g) New Position (Xiᵏ⁺¹): The blue crayfish or arrow endpoint shows where the
crayfish will move next, based on the combined influences.

h) Convergence Behavior: Over multiple iterations, the crayfish group is seen

swimming toward more optimal regions, clustering around promising solutions [39].
This convergence is visualized by the gradual tightening of crayfish paths toward the
global best region.

1.3.2.1 Swimming Through Complexity: How COA Navigates the Search Space:

a) Initialization Phase: A group of crayfish is randomly placed in the search space.

Each has an initial velocity and position.
b) Memory and Social Update: Each crayfish remembers its best position and is
influenced by better-performing crayfish in its neighborhood or the globally best
individual.
c) Movement Simulation: Crayfish move based on their own memory, the influence
of the elite crayfish, and a random term that helps prevent premature convergence

13
[40].
d) Boundary Check and Position Correction: If a crayfish goes beyond the search
boundary, it's brought back to a feasible region to maintain solution validity.
e) Evaluation and Iteration: New positions are evaluated for fitness. If better, the
personal and global bests are updated. The process repeats for a set number of
iterations or until a stopping criterion is met.

1.3.3 GREY WOLF OPTIMIZATION ALGORITHM (GWO):

The Grey Wolf Optimization (GWO) algorithm is a nature-inspired metaheuristic
optimization algorithm. Grey wolves hunt in packs, with an alpha wolf leading the
pack and coordinating the group’s strategy to capture prey. The GWO algorithm
mimics this social hierarchy and hunting strategy, dividing the population of
candidate solutions into four categories: alpha, beta, delta, and omega wolves, where
each represents different levels of leadership and contribution to the optimization
process [41]. The main goal of GWO is to find the optimal solution by adjusting the
positions of the wolves in the search space based on their social hierarchy and prey-
capture strategy.

Fig 1.5: Grey Wolf Optimization

a) Alpha Wolf (α): Represented by the orange circle (α₁). This is the leader and the
fittest wolf in the pack. Its position corresponds to the best solution found so far,
guiding the pack toward the optimal solution.

b) Beta Wolf (β): Represented by the blue circle (β₂). It is the second-best wolf and
assists the alpha in decision-making. Its position corresponds to the second-best

14
solution found by the pack.

c) Delta Wolf (δ): Represented by the green circle (δ₃). This is the third-best wolf,
which dominates the omega wolves. Its position is the third-best solution found so far.

d) Candidate Position of the Wolves: Represented by the grey wolf image. This
symbolizes a regular search agent (an omega wolf or potentially a new position for
alpha, beta, or delta in the next iteration).

e) Estimated Position of the Prey: Represented by the image of the prey (deer). This
marks the target optimal solution that the wolves are attempting to locate.

f) Dα, Dβ, Dδ: These arrows indicate the distance vectors between the candidate wolf
and the alpha, beta, and delta wolves, respectively. Mathematically, these are
calculated as:
Dα =¿ C ₁⋅ X α −X ∨¿(1.1)
D β=¿ C ₂⋅ X β −X ∨¿
(1.2)
Dδ ¿∨C ₃ ⋅ X δ −X∨¿(1.3)
Where X α, X β , and X δ are the positions of the alpha, beta, and delta wolves,
respectively, and X is the position of the candidate wolf. C₁, C₂, and C₃ are
coefficient vectors that define the encircling behavior.

g) Circles around α₁, β₂, and δ₃: These dashed circles, along with the C₁, C₂, and
C₃ labels, visually represent the "encircling behavior." The radius of these circles is
influenced by the coefficient vectors C₁, C₂, and C₃, which have random components
that allow the wolves to move within this range around the prey (in this analogy,
around the alpha, beta, and delta wolves) [42].

h) "Move" Arrow: This arrow shows the resulting movement of the candidate wolf
towards a new position. The new position is determined by averaging the influence of
the alpha, beta, and delta wolves. The mathematical formulation for updating the
position of a candidate wolf is:

15
X ₁= X α − A ₁ ⋅ Dα (1.4)
X ₂= X β − A ₂ ⋅ Dβ (1.5)
X ₃= X δ − A ₃ ⋅ Dδ (1.6)
X (t+1)=(X ₁+ X ₂+ X ₃)/3 (1.7)
Where A₁, A₂, and A₃ are other coefficient vectors that help control the step size and
balance between exploration and exploitation.

1.3.3.1 Hunting and Leadership: How GWO Navigates the Search Space:
a) Initialization Phase: A group of grey wolves is randomly initialized within the
search space, each with a random position and velocity. The positions represent
potential solutions to the optimization problem.

b) Social Hierarchy Update: Wolves are classified into alpha, beta, delta, and omega
wolves based on their fitness. The alpha wolf represents the best solution, while the
beta and delta wolves represent the next best solutions, and the omega wolves are the
least effective.

c) Hunting Process (Position Update): Wolves move toward the prey (optimal
solution) by adjusting their positions according to the position of the alpha, beta, and
delta wolves. This process is inspired by the wolves' social behavior during hunting,
where they coordinate and work together to capture prey.

d) Exploration and Exploitation: The algorithm strikes a balance between

exploration (searching new regions) and exploitation (refining existing good
solutions) by adjusting the position based on the distances from the alpha, beta, and
delta wolves, combined with random vectors that guide exploration.

e) Boundary Check and Position Correction: If a wolf's position goes beyond the
defined search boundaries, it is corrected to stay within the feasible region.

f) Evaluation and Iteration: The fitness of the new positions is evaluated. If the new
position is better, the personal bests and global bests are updated. This process repeats
iteratively until a stopping criterion is met, such as reaching a maximum number of

16
iterations or converging to an optimal solution.

1.3.4 AFRICAN VULTURES OPTIMIZATION ALGORITHM (AVOA):

The African Vultures Optimization Algorithm (AVOA) is a bio-inspired optimization
technique that simulates the scavenging behavior of African vultures. These birds are
known for their remarkable ability to locate and scavenge carcasses across vast areas,
relying on their keen sense of sight and dynamic flock behavior. AVOA models the
vultures' movement patterns, which involve both local search and global exploration,
allowing the algorithm to efficiently navigate complex search spaces [43]. It
incorporates the concept of food sources (solutions) and the role of different vultures
in the search process. Vultures in the algorithm are metaphorically represented as
candidate solutions, and the algorithm works by adjusting the positions of these
solutions to converge toward the optimal solution. Fig. 1.10 shows the working of
GWO:

a) Selection of the Leader: For each vulture P(i), a probability value pᵢ determines
whether it will follow the first best vulture or the second best vulture. This mechanism
introduces a balance between exploitation (focusing on the best solution) and
exploration (searching around the second-best solution).

Fig 1.6: African Vulture Optimization

17
b) Movement Towards the Leader:
If pᵢ = L₁: The vulture P(i) moves toward the First Best Vulture. The updated position
P(i+1) lies along the line connecting the current position P(i) and the first best vulture.
This movement is influenced by the distance function d(t) and the feeding rate F.
If pᵢ = L₂: The vulture P(i) moves toward the Second Best Vulture in a similar
manner, with the new position P(i+1) influenced by d(t) and F, but guided by the
second-best vulture instead.

c) Influence of Feeding Rate (F): The feeding rate F determines how far a vulture
can move toward the leading vultures. A higher feeding rate allows larger exploratory
steps, whereas a lower feeding rate encourages more refined, localized search. The
dashed circles around the best vultures visually depict their zones of influence, which
are modulated by the value of F.

1.3.4.1 Scavenging and Flocking: How AVOA Navigates the Search Space:
a) Initialization Phase: A population of vultures is randomly initialized within the
search space. Each vulture has an initial position and velocity, representing potential
solutions to the problem.

b) Food Source Search: Each vulture in the population evaluates the fitness of its
current position and searches for food sources (better solutions) using both
exploration (scavenging in wide regions) and exploitation (focusing on previously
known good solutions).

c) Social Interaction and Memory Update: Vultures are influenced by the best food
sources found by others (social interaction), and they update their memory to retain
the best solution they have encountered. This allows vultures to converge to
promising areas of the search space while exploring others.

d) Movement Simulation: The movement of each vulture is determined by a

combination of the global best solution (food source) and its own personal best
solution. The vultures move through the search space, adjusting their positions to

18
improve their fitness by balancing exploration and exploitation.

e) Boundary Check and Position Correction: Similar to other optimization

algorithms, AVOA ensures that the vultures stay within the search boundaries. If a
vulture moves out of the feasible region, its position is corrected to stay within valid
search space limits.
f) Evaluation and Iteration: The fitness of the vultures' new positions is evaluated. If
a better solution is found, the global and personal bests are updated. The process
repeats for a set number of iterations or until convergence is achieved.

1.3.5 RESEARCH GAPS AND MOTIVATION FOR THIS WORK:

A thorough review of the literature reveals several gaps. Firstly, despite the success of
algorithms like PSO and GWO, they often suffer from premature convergence and
lack adaptability in rapidly changing environments. Algorithms like COA and AVOA,
although promising, are still underexplored in cloud-specific contexts such as
federated or hybrid cloud systems. Additionally, there's limited research on
hybridizing these algorithms or enhancing them with real-time feedback mechanisms
for dynamic resource scheduling.

Furthermore, multi-objective optimization, which is critical for balancing cost,

energy, latency, and performance in the cloud, remains a challenge. Most studies still
focus on single-objective formulations [44]. This work aims to explore and improve
upon newer swarm algorithms such as AVOA, customizing them for real-time, multi-
objective resource allocation in cloud environments.

1.3.6 ADVANTAGES OF SWARM OPTIMIZATION TECHNIQUES:

Simplicity and Flexibility: Easy to implement with simple rules governing agent
behavior. It can be adapted to a wide range of optimization problems (discrete,
continuous, multi-objective).
Scalability: Performs well with increasing problem size or dimensionality. Suitable
for large-scale distributed systems like cloud computing and IoT networks.
Robustness and Fault Tolerance: The decentralized approach ensures that failure of
a few agents doesn’t crash the system. Robust against noisy or incomplete data.
19
Global Optimization Capability: Capable of escaping local optima due to stochastic
search mechanisms. Good exploration and exploitation balance (especially in GWO,
AVOA, etc.).

1.4 OBJECTIVE:
This study aims to explore and evaluate the application of swarm optimization
techniques in minimizing energy expenditure in WSNs. The specific objectives
include:
a) Developing clustering and routing strategies using PSO, GWO, and COA.
b) Comparing the performance of these algorithms based on energy consumption,
network lifetime, and scalability.
c) Identifying the most effective algorithm for energy-efficient WSN operation.

1.5 CONTRIBUTION:
The thesis comprises of the five chapters. The first chapter includes the introduction
to wireless sensor networks and swarm optimization techniques In second chapter
literature review regarding Wireless Sensor Networks and the existing algorithms and
the problems faced in maintaining data transmission time and efficiency in energy
minimization is shown. In third chapter discussions about the proposed system model
is shown. In this chapter we see that the energy efficiency, network lifetime, and
overall performance have been shown to be improved significantly. Fourth chapter
analyze the referenced system model and its energy usage in different algorithms.
Fifth chapter shows the result where we have achieved that COA has a better network
lifetime than other clustering algorithms.

1.6 THESIS OUTLINE:

The following study presents an improved approach to routing and clustering in
WSNs using swarm optimization techniques. PSO, GWO, AVOA, and COA are
compared, with results showing that COA and AVOA achieve better network lifetime
and energy balancing. The results demonstrate a reduction in energy consumption

20
while ensuring better data delivery in terms of rate and network operation. Section 2
summarizes the literature review. Section 3 discusses the proposed algorithm with its
flowchart. Section 4 discusses the referenced energy dissipation model. Section 5
discusses the final result. Section 6 gives the conclusion, and Section 7 includes
references.

CHAPTER 2

LITERATURE REVIEW
Wireless Sensor Networks (WSNs) face critical challenges in energy optimization due
to limited battery life and high energy consumption during data transmission. Various
swarm optimization techniques, including Particle Swarm Optimization (PSO),
Crayfish Optimization, and Grey Wolf Optimization (GWO), have been explored to
address these issues. These methods have shown promise in improving routing
efficiency, clustering, and load balancing, leading to reduced energy consumption.
Recent studies highlight the integration of hybrid swarm algorithms with machine
learning to further enhance energy efficiency and network performance.
S. No REFERENCES FINDINGS LIMITATIONS
1. P. Kuila and P. K. Jana, “Energy Energy savings High computational
efficient clustering and routing through PSO- complexity, limiting
algorithms for wireless sensor based clustering adaptability for
networks: Particle swarm and routing dynamic network
optimization approach”, algorithms, scenarios.
Engineering Applications of effectively
Artificial Intelligence, vol. 33, extending network
pp. 127 - 140, 2014.[1] lifetime.

21
 2. F. Bajaber & I. Awan, “An The protocol The protocol
efficient cluster-based enhanced energy enhanced energy
communication protocol for efficiency, efficiency, extended
wireless sensor extended network network
networks,” Telecommunication lifetime through lifetime through
Systems, vol. 55, no. 3, pp. 387– adaptive clustering, adaptive clustering,
401, 2014.[2] reduced reduced
communication communication
overhead. overhead, and
improved
scalability.
Table 2.1: References, findings, and limitations.

3. S. Mirjalili, S. M. Mirjalili, and Global Premature

A. Lewis, “Grey Wolf optimization with convergence in
Optimizer”, Advances in fast convergence complex, high-
Engineering Software, vol. 69, and robustness dimensional
pp. 46-61, 2014.[3] across various problems and
benchmark lacked mechanisms
functions and to maintain search
engineering diversity.
applications.

Energy minimization in Wireless Sensor Networks (WSNs) is a critical challenge due

to the limited power resources of sensor nodes, necessitating efficient algorithms for
energy-efficient routing and clustering. Kuila and Jana [1] proposed a Particle Swarm
Optimization (PSO)-based approach for clustering and routing, which resulted in
significant energy savings and prolonged network lifetime by optimizing cluster
formation and routing paths. However, their method faced limitations related to high
computational complexity and lack of real-world validation, as it was primarily tested
in simulation environments. In a similar vein, Bajaber and Awan [2] introduced a
cluster-based communication protocol that leveraged adaptive clustering to reduce
energy consumption and improve scalability. While this approach showed promise in

22
larger networks, it was constrained by its static clustering mechanism and the absence
of consideration for energy heterogeneity among nodes. Furthermore, the study was
validated mainly through simulations, which could affect the generalizability of the
results. On the other hand, Mirjalili et al. [3] presented the Grey Wolf Optimizer
(GWO), a nature-inspired metaheuristic algorithm that demonstrated effective global
optimization in various applications, including WSNs. Their approach was
particularly noted for its fast convergence and robustness, but challenges such as
premature convergence in high-dimensional problems and limited search diversity
were identified as drawbacks. These studies underline the potential of swarm
intelligence and bio- inspired algorithms for enhancing energy efficiency in WSNs,
though issues related to adaptability, scalability, and real-world feasibility remain
areas for further investigation. Combining these approaches could potentially offer
more robust solutions to the challenges of dynamic and large-scale network
environments.
S. No REFERENCES FINDINGS LIMITATIONS
4. A. Norouzi, & A. H. Zaim, Optimize energy High computational
“Genetic algorithm consumption, node overhead and was
application in optimization of deployment, and sensitive
wireless sensor routing in to parameter
networks,” The Scientific WSNs, improving tuning,
World Journal, pp. 1–15, network coverage limiting its real-time
2014.[4] and performance. applicability in
large-scale networks.

23
 5. M. I. Chidean, E. Morgado, Improved energy High computational
M. S. Junquera, J. R. efficiency and data complexity and was
Bargueno, J. Ramos, & A. J. reconstruction sensitive to the
Caamaño, “Energy efficiency quality in large- initial network
and quality of data scale WSNs configuration,
reconstruction through data- through a data- affecting scalability
coupled clustering for self- coupled clustering and adaptability.
organized large-scale method.
WSNs,” IEEE Sensors
Journal, vol. 16, no. 12, pp.
5010–5020, 2016.[5]
6. S. Noureddine, B. Khelifa, Optimizing data Relied heavily on
and B. Mohammed, transmission and specific network
“Approach to minimizing node scheduling, conditions and was
consumption of energy in leading to not tested in large-
wireless sensor significant energy scale, dynamic
networks,” International savings and environments,
Journal of Electrical and extended network limiting its
Computer Engineering, vol. lifetime. generalizability.
10, no. 3, pp. 2551, 2020.[6]

Minimizing energy consumption in Wireless Sensor Networks (WSNs) has been a

primary focus of research to extend network lifetime and optimize resource usage.
Noureddine et al. [4] proposed a method for minimizing energy consumption through
optimized data transmission and node scheduling, achieving significant energy savings
and prolonging the network's operational lifespan. However, their approach was
tailored to specific network conditions and not tested in large-scale or dynamic
environments, limiting its applicability. Similarly, Chidean et al. [5] introduced a data-
coupled clustering method that improved both energy efficiency and data
reconstruction quality in large-scale self-organized WSNs. Their approach, while
effective, faced challenges related to computational complexity and sensitivity to the
network's initial configuration, which impacted its scalability and adaptability. In

24
contrast, Norouzi and Zaim [6] employed Genetic Algorithms (GAs) to optimize
energy consumption, node deployment, and routing in WSNs, demonstrating
improvements in network coverage and performance. However, the high
computational overhead and the need for careful parameter tuning were notable
limitations, affecting real-time applicability, especially in large-scale networks. These
studies highlight the potential of various optimization techniques, such as node
scheduling, data aggregation, and metaheuristic algorithms, in addressing energy
efficiency challenges in WSNs. However, issues related to scalability, computational
complexity, and adaptability in dynamic environments remain important areas for
further research.

S. No References Findings Limitations

7. S. Loganathan, J. Arumugam, Improved energy Performance is
“Energy Efficient Clustering efficiency in sensitive to network
Algorithm Based on Particle WSNs by using parameters and may
Swarm Optimization Technique an enhanced need further
for Wireless Sensor Algorithm for refinement for
Networks,” Wireless Pers optimal cluster dynamic
Commun, vol. 119, pp. 815– head selection, environments.
843, 2021.[7] reducing energy
consumption and
extending network
lifetime.

8. H. Jia, H. Rao, C. Wen, and Superior Potential premature

S.Mirjalili,“Crayfish optimization performance in convergence in
algorithm”, Artificial Intelligence global optimization requires careful
Review, vol. 56, pp. 1919 tasks, including parameter tuning to
- 1976, WSN achieve optimal
2023.[8] applications, by results.
efficiently
balancing
exploration and
exploitation.

25
 9. V. Chandran and P. Mohapatra, Improving its Computational
“Enhanced opposition-based global complexity and
grey wolf optimizer for global optimization premature
optimization and engineering performance and convergence when
design problems”, Alexandria providing better applied to highly
Engineering Journal, vol. 76, solutions for complex, high-
pp. 429- 467, 2023.[9] engineering dimensional
design problems, optimization
including WSN problems.
applications.

Recent advancements in optimization techniques for Wireless Sensor Networks

(WSNs) have shown promising results in improving energy efficiency and network
performance. S. Loganathan, J. Arumugam [7] introduced an energy-efficient cluster
head selection method using an improved algorithm for particle swarm optimization,
which enhanced network lifetime by reducing energy consumption during cluster
formation. However, the method was sensitive to network parameters and required
further refinement for dynamic environments. Jia et al. [8] proposed the Crayfish
Optimization Algorithm (COA), a nature-inspired metaheuristic that effectively
balanced exploration and exploitation, leading to better global optimization
performance in WSNs. Despite its potential, the COA faced challenges with premature
convergence in high-dimensional problems and required careful parameter tuning.
Additionally, Chandran and Mohapatra [9] enhanced the Grey Wolf Optimizer (GWO)
by incorporating opposition-based learning, significantly improving its optimization
capabilities for engineering design problems, including WSN applications. This
enhancement showed improved results but still struggled with computational
complexity and premature convergence in complex optimization scenarios. These
studies highlight the growing role of metaheuristic and nature-inspired algorithms in
optimizing energy consumption and network efficiency in WSNs, though scalability
and adaptability to dynamic environments remain key challenges.

S. No REFERENCES FINDINGS LIMITATIONS

26
 10. H.R.H. Al Dallal and D.A. Optimized The method
Sultan, “An optimized algorithm algorithm for struggles to adapt to
design for a target tracking in target tracking in dynamic
wireless sensor networks”, Spec. WSNs, improving environments with a
Jour. of Inno. Ref. and Dev., vol. accuracy and large number of
12, pp. 76–94, 2023.[10] energy efficiency. targets.

11. P. Rawat, P. Kumar, and S. Their fuzzy logic The approach is

Chauhan, "Fuzzy logic and and PSO-based sensitive to
particle swarm optimization- clustering protocol parameter tuning,
based clustering protocol in enhances energy affecting real-time
wireless sensor network," Soft efficiency and performance.
Computing, vol. 27, no. 9, pp. network
5177-5193, 2023.[11] reliability.

12. S.D. Mishra, S. Dutta, and D. Energy-efficient Scalability

Verma. "Energy-Efficient and clustering method
Reliable Clustering with with optimized and computational
Optimized Scheduling and scheduling and complexity in large
Routing for Wireless Sensor routing to networks
improve reliability
Networks," Multimedia Tools
and
and Applications, pp. 1-27. 2024. remain challenging.
reduce energy
[12]
consumption.

Recent studies in optimizing energy consumption and improving performance in

Wireless Sensor Networks (WSNs) have explored novel algorithms and hybrid
techniques. Al Dallal and Sultan [10] designed an optimized algorithm for target
tracking in WSNs, focusing on improving the accuracy and energy efficiency of
tracking tasks, though it was limited by challenges in adapting to dynamic
environments with a large number of targets. Rawat et al. [11] proposed a fuzzy logic
and Particle Swarm Optimization (PSO)-based clustering protocol, which aimed to
enhance network lifetime and reliability by dynamically adjusting cluster formations
based on environmental conditions. While this hybrid approach showed promising

27
results in energy efficiency and load balancing, its reliance on fuzzy logic and PSO
made it sensitive to parameter tuning, which could affect real-time performance.
Additionally, Mishra et al. [12] presented an energy-efficient clustering method with
optimized scheduling and routing, aiming to enhance both reliability and energy
efficiency in WSNs. Their approach focused on reducing energy consumption while
ensuring reliable data transmission, but scalability and computational complexity in
large networks remained a concern. These studies underscore the effectiveness of
hybrid algorithms in optimizing energy usage and improving the overall performance
of WSNs, yet issues related to scalability, adaptability, and computational overhead
remain significant challenges.

S. No REFERENCES FINDINGS LIMITATIOS

13. B.H. Goud, T.N. Shankar, B. Energy Scalability issues in
Sah, R. Aluvalu, “Energy optimization larger networks.
optimization in path arbitrary reduces
wireless sensor network,” consumption in
Expert Systems, vol. 41, no. 2, arbitrary path
pp. 13282, 2024.[13] WSNs.
14. I. Surenther, K.P. Sridhar, M.K. Machine High
Roberts, “Enhancing data learning implementation
transmission efficiency in enhances data complexity
wireless sensor networks transmission
through machine learning- efficiency and and
Sensitivity to
network

enabled energy optimization: A energy conditions.

grouping model approach,” Ain optimization.
Shams Engineering Journal, vol.
15, no. 4, pp.102644, 2024.[14]

28
 15. V. Prakash, D. Singh, S. Pandey, The combination Scalability issues
S. Singh, and P.K. Singh, of M-PSO and when applied to
“Energy-Optimization Route and GA successfully large-scale wireless
Cluster Head Selection Using M- enhances energy sensor networks
PSO and GA in Wireless Sensor efficiency and with a high number
Networks”, Wireless Personal improves the of nodes.
Communications, pp.1-26, selection of
2024.[15] optimal cluster
heads in wireless
sensor networks.

Energy optimization in Wireless Sensor Networks (WSNs) has been widely studied
due to the limited battery life of sensor nodes. Goud et al. [13] explore energy
optimization in networks with arbitrary paths and dynamic topologies, emphasizing the
importance of multi-path routing and adaptive energy-saving protocols to minimize
energy consumption. In contrast, Surenther et al. [14] introduce machine learning
(ML)- enabled energy optimization, proposing a grouping model approach where
nodes are dynamically grouped based on their energy levels and proximity, enhancing
data transmission efficiency. This ML-based approach provides real-time adaptability,
which is especially useful in large, dynamic networks. On the other hand, Prakash et
al. [15] present a hybrid optimization approach combining Modified Particle Swarm
Optimization (M-PSO) and Genetic Algorithms (GA) to improve cluster head
selection and routing, showing that such hybrid methods outperform traditional
approaches in terms of energy efficiency, particularly in large-scale networks. These
studies collectively highlight the shift towards more dynamic, adaptive, and hybrid
methods for optimizing energy consumption in WSNs, though challenges like
scalability, computational complexity, and real-time adaptability remain.

S. No REFERENCES FINDINGS LIMITATIOS

29
16. Y. Razooqi, M. Al-Asfoor, and Reduces energy Scalability issues in
M.H. Abed, “Optimise Energy consumption in large networks due
Consumption of Wireless Sensor Wireless Sensor to the
Networks by using modified Ant Networks computational
Colony Optimization”, Acta (WSNs) by complexity of path
Technica Jaurinensis, vol. 17, optimizing optimization in
pp.111-117, 2024.[16] routing paths, dynamic
leading to environments.
improved
network lifetime.

17. J.S. Abdullah, and M.I. Aal- Improves node Challenges in

Nouman, “Efficient Deployment deployment highly dynamic
of Nodes in a WSN Using a efficiency in environments where
Modified PSO Algorithm”, 1st WSNs by node
International Conference on ensuring optimal movement or
Emerging Technologies for distribution, failure occurs
Dependable Internet of Things, reducing energy frequently.
pp. 1-6, 2024.[17] consumption and
communication
overhead.

18. P. Mishra, R.K. Dash, T. Effectively Difficulties in real-

Choudhury, and K. Kotecha, optimizes both time adaptability,
“Optimizing Residual Energy residual energy especially in highly
and Delay in WSN Routing using and delay in dynamic WSNs
Particle Swarm Optimization” WSN routing, with fluctuating
Journal homepage: http://iieta. leading to network conditions.
org/journals/isi, vol. 29, pp.761- enhanced
770, 2024.[18] network
performance and

30
Energy optimization in Wireless Sensor Networks (WSNs) is a critical research area,
with various algorithms being developed to enhance performance and extend network
lifetime. Razooqi et al. [16] propose a modified Ant Colony Optimization (ACO)
algorithm, which reduces energy consumption by optimizing the routing paths, thus
improving the overall lifetime of the network. However, their method faces scalability
issues in large, dynamic networks due to the computational cost of path optimization.
Abdullah and Aal-Nouman [17] focus on efficient node deployment using a Modified
Particle Swarm Optimization (PSO) algorithm, which ensures optimal node
distribution, reduces energy usage, and minimizes communication overhead. Despite
its effectiveness, the method faces challenges in dynamic environments, where nodes
may frequently move or fail. Mishra et al. [18] optimize residual energy and delay in
WSN routing through PSO, which enhances network performance and prolongs its
lifetime by balancing energy consumption and routing delay.

S. No REFERENCES FINDINGS LIMITATIOS

19. M.A. Lilo, A.K. Yasari, M.M. Effectively The enhanced PSO
Hamdi, and A.D. Abbas, reduces approach may have
“Transmission Power Reduction transmission limited applicability
Based on an Enhanced Particle power in in
Swarm Optimization Algorithm Wireless Sensor highly dynamic
in Wireless Sensor Network for Networks environments where
Internet of Things”, ARO-THE (WSNs) for network
SCIENTIFIC JOURNAL OF Internet of topologies and
KOYA UNIVERSITY, vo. 12, Things (IoT) traffic patterns
pp.61-69, 2024.[19] applications, change rapidly.
improving energy
efficiency and
network lifetime.
20. J. Dev and J. Mishra, “Energy Improves energy- Scalability issues
Efficient Routing in Cluster efficient routing and complexity
Based Heterogenous Wireless in cluster-based when dealing with
Sensor Network Using Hybrid heterogeneous large, dense, or
GWO and Firefly Algorithm”, WSNs, extending highly

31
 Wireless Personal the network’s heterogeneous
Communications, vol. 132, pp. lifetime and networks.
997-1028, 2024 [20] reducing energy
consumption

Recent studies have focused on improving energy efficiency in Wireless Sensor

Networks (WSNs) using optimization algorithms. Lilo et al. [19] propose an enhanced
Particle Swarm Optimization (PSO) algorithm to reduce transmission power in WSNs
for Internet of Things (IoT) applications. This approach significantly improves energy
efficiency and extends network lifetime, though it may not perform optimally in highly
dynamic environments with rapidly changing topologies. In another approach, Dev
and Mishra [20] present a hybrid solution combining Grey Wolf Optimization (GWO)
and Firefly algorithms for energy-efficient routing in cluster-based heterogeneous
WSNs. Their method outperforms traditional routing protocols by reducing energy
consumption and prolonging network lifetime, but it faces scalability and complexity
issues, particularly in large-scale or dense networks. Both studies highlight the
potential of optimization algorithms in enhancing energy efficiency and extending the
life of WSNs, though challenges related to scalability, real-time adaptability, and
dynamic environments remain significant.

Chapter 3

32
 ENERGY MINIMIZATION USING SWARM
OPTIMIZATION TECHNIQUE PROPOSED MODEL

3.1 ENERGY MODEL:

In this model, both the free space and multi-path fading channels are used depending
on the distance between the transmitter and receiver. When the distance is less than a
threshold value d₀, the free space (fs) model is used; otherwise, the multipath (mp)
model is used. Let Eₑₗₑc, εfs, and εmp be the energy required by the electronics circuit
and by the amplifier in free space and multipath respectively. Then the energy
required by the radio to transmit an l-bit message over a distance d is given as
follows:

{
2
ET ( l , d )= l Eelec +l ε fs d 4,∧d <d 0 (3.1)
l E elec + l ε mp d ,∧d ≥ d 0

The Eelec depends on several factors such as digital coding, modulation, filtering, and
spreading of the signal, whereas the amplifier energy, εfsd²/εmpd⁴, depends on the
distance between the transmitter and the receiver and also on the acceptable bit-error
rate. It should be noted that this is a simplified model. In general, radio wave
propagation is highly variable and difficult to model.

Fig 3.1: Energy dissipation model transmitter

33
 Fig 3.2: Energy dissipation model receiver

3.2 NETWORK MODEL:

We assume a WSN model where all the sensor nodes are deployed randomly along
with a few gateways, and once they are deployed, they become stationary. A sensor
node can be assigned to any gateway if it is within the communication range of the
sensor node [45] . Therefore, there are some pre-specified gateways onto which a
particular sensor node can be assigned. Thus, each sensor node has a list of gateways
and it can be assigned to only one gateway amongst them. Similar to LEACH, the
data-gathering operation is divided into rounds. In each round, all sensor nodes collect
the local data and send it to their corresponding CH (i.e., gateway). On receiving the
data, the gateways aggregate them to discard the redundant and uncorrelated data and
send the aggregated data to the base station via another CH as a next hop relay node.
Between two adjacent rounds, all nodes turn off their radios to save energy. All
communication is over a wireless link. A wireless link is established between two
nodes only if they are within the communication range of each other. The current
implementation supports the TDMA protocol to provide MAC layer communication.
Gateways use a slotted CSMA/CA MAC protocol to communicate with the base
station.

34
3.2.1 LP FORMULATION FOR ROUTING PROBLEM:
Now, we address the routing problem where our main objective is to minimize the
maximum transmission distance between two nodes in the routing path and maximum
hop count. Let aij be a Boolean variable defined as:

{
Aij = 1 ,∧If NextHop ( gi )=g j
0 ,∧otherwise
(3.2)

Then the Linear Programming (LP) of the routing problem is formulated as follows:
Minimize W=α × MaxDist + β×MaxHop (3.3)
where α = 1 – β and 0 < β < 1

α and β are two control parameters. α controls the total path distance and β controls the
total hop count. The constraint defines the range of α and β.

We represent the particles in such a way that each particle provides the route from
each CH to the BS. The dimension of the particles is same and equal to the number of
gateways (i.e., M) in the network. We initialize each component, i.e., X i,d, 1 ≤ i ≤ Np,
1 ≤ d ≤ M with a randomly generated uniformly distributed number Rand(0,1), 0 <
Rand(0,1) ≤ 1. The value of the dth component (i.e., Xi,d) maps a gateway (say gk) as
a next hop relay towards BS from gd. Therefore, Xi,d = Rand(0,1) maps a gateway
(say gk), indicating that gd send data to gk. The mapping is done as follows:
𝑔𝑘 = 𝐼𝑛𝑑𝑒𝑥(𝑃𝑁𝑒𝑥𝑡𝐻𝑜𝑝𝑠(𝑔𝑑), 𝑛)
(3.4)

where 𝐼𝑛𝑑𝑒𝑥(𝑃𝑁𝑒𝑥𝑡𝐻𝑜𝑝𝑠(𝑔𝑑), 𝑛) is an indexing function that indexes the nth

gateway from 𝑃𝑁𝑒𝑥𝑡𝐻𝑜𝑝𝑠(𝑔𝑑) and
𝑛 = 𝐶𝑒𝑖𝑙𝑖𝑛𝑔(𝑋𝑖,𝑑 × |𝑃𝑁𝑒𝑥𝑡𝐻𝑜𝑝𝑠(𝑔𝑑)|)
(3.5)

3.2.2. FITNESS FUNCTION:

It is noteworthy that the above two objectives conflict each other, i.e., lower distance
of next-hop relay node increases the hop count and vice versa. Therefore,
optimization of one objective hampers the optimization of the other. Our proposed

35
work constructs the fitness function in such a way that a trade-off can be built with
these conflicting objectives [46]. We have used the weight sum approach (WSA) for
the construction of the multi-objective fitness function. WSA is a classical approach
for solving the multi- objective optimization problem. In this approach, a weight value
Wi is multiplied with each objective. Finally, all the multiplied values are added to
convert the multi- objectives into a single scalar objective function as follows:

Fitness =𝑊1 ×MaxDist + 𝑊2×MaxHop

(3.6)

36
Fig 3.3: Flowchart

37
In our approach we have taken W2=1− W1 and 0<W1<1. Our objective is to
minimize the fitness value. In other words, lower the fitness value, the better is the
particle position.

3.2.3 VELOCITY AND POSITION UPDATE:

The velocity and the position are updated in each iteration using Eqs. (3.1) and (3.2)
respectively. It is noteworthy that, the algebraic steps of addition and subtraction
operation in Eqs. (3.1) and (3.2) may cause the new position of the particle to be
negative or greater than one. In our scenario the position of the particle must satisfy the
range [0,1]. Therefore, our algorithm should generate the positions of the particles in
such a way that it can satisfy the range. This can be made if we choose the positions as
follows: If new position is negative or zero, then replace the position value by a newly
generated random number which tends to zero or if new position is greater than one,
then replace the position value by one.

After getting the new position, the particle Pi is evaluated by the fitness function.
Now, its personal best (Pbesti ) is replaced by itself, only if its current fitness value is
better than its Pbesti fitness value.

The updating process is as follows:

{
Pbest i= P i ,∧if fitness ( Pi)< fitness(Pbest i)
Pbest i ,∧otherwise
(3.7)

Now, the global is also updated as follows:

{
Gbest = P i ,∧if fitness ( Pi)< fitness(Gbest i)
Gbest i ,∧otherwise
(3.8)

The velocity and the positions are iteratively updated until the termination criteria are
fulfilled [47]. In our approach, the termination criterion is a predefined iteration
number. After termination of the PSO-based routing algorithm, the particle Gbesti
represents the final routing solution.

38
3.2.5 AFRICAN VULTURES OPTIMIZATION ALGORITHM MODEL:
The African Vultures Optimization Algorithm (AVOA) is a nature-inspired
metaheuristic algorithm developed to solve complex optimization problems, it models
the dynamic interaction between exploration (global search) and exploitation (local
search) [49]. The algorithm maintains a population of candidate solutions, each
representing a vulture, and dynamically classifies them into leaders and followers
based on their fitness values. Leaders guide the search toward promising regions,
while followers update their positions by learning from the leaders and exploring the
search space. This balance enables AVOA to avoid premature convergence and
effectively navigate complex, multidimensional solution spaces. Fig 3.5 depicts the
steps involved in the algorithm.

The probability of selecting vultures to move other vultures toward one of the best
solutions is calculated, where L 1 and L2 are the parameters to be measured before
searching:

R ( i )=¿ {BestVulture 1 if pi=L1

BestVulture 2 if pi=L2
(3.20)

The probability of choosing best solution is measured with Roulette wheel to select
each of the best solutions for each group:

n
pi=∑ Fi (3.21)
i =1

Rate of starvation of vultures is given by the following equation:

T =h × sin
( ( (
w π
2
×
maxiterations ) (
iteration ( i ) π
+ cos ×
2 maxiterations ) ))
iteration ( i )
−1 (3.22)

( (
F=( 2 ×rand 1 +1 ) × z × 1 –
iteration
maxiterations ))
+t (3.23)

39
Fig 3.4 Flowchart Based on AVOA

40
In Eqs. (3.22) and (3.23), F denotes the vultures that are satiated, iteration(i) gives the
current iteration number, maxiterations gives the total number of iterations, and z is a
random number between -1 and 1, h is a random number between -2 and 2, rand 1 has a
random value between 0 and 1.

F ( i+1 )=¿
{ Equation ( 6 ) if P1 ≥ rand ( p 1)
Equation ( 8 ) if P1 <rand ( p1 )

(3.24)
P ( i+1 )=R ( i ) – D ( i ) × F
(3.25)
D ( i )=|X × R ( i ) – P ( i )| (3.26)
P ( i+1 )=R ( i ) – F +rand 2 × ( ( ub – lb ) × rand 3+ lb ) (3.27)

A random number is generated between 0 and 1 in the rand(p1) exploration phase. If

this number is greater than or equal to P1, then Eq. (3.25) is used, else Eq. (3.27) is
used. This procedure is shown in Eq. (3.24).

P ( i+1 )=¿
{ Equation ( 10 ) if P2 ≥rand ( p2 )
Equation ( 13 ) if P 2< rand ( p 2 )

(3.28)

Eq. (3.28) gives the exploitation value when AVOA enters the first phase when |F| is
between 0.5 and 1. Rotating flight and siege-fight strategies are carried out. P2 is for
determining each strategy’s choice, with rand( p2) being a random number between 0
and 1. Siege-fight is implemented only when this number is greater than or equal to
P2, else, rotating flight is implemented.

P ( i+1 )=D (i ) × ( F +rand 4 ) – d ( t ) (3.29)

d ( t )=R ( i ) – P ( i ) (3.30)

Eq. (3.29) and Eq. (3.30) model the process that weaker vultures follow to tire and get
food from healthy vultures. D(i) is calculated using Eq. (7) and F is the satiation rate,

41
calculated by Eq. (3.23), with rand4 is a random number between 0 and 1 which is to
increase random coefficient.

S1=R ( i ) × { rand 5 × P ( i )
2π }
×cos ( P ( i ) ) (3.31)

S2=R ( i ) × { rand 6 × P ( i )
2π }
×sin ( P ( i ) ) (3.32)

P ( i+1 )=R ( i )−( S1 +S 2 ) (3.32)

In Eqs. (3.31) and 3.32), are used to represent position vector of the two best vultures.

P ( i+1 )=
{
Equation ( 16 ) if P3 ≥rand ( p3 )
Equation (17 ) if P 3< rand ( p 3 )

(3.33)

Eq. (3.33) is used to carry out the second phase exploitation.

A1=BestVulture1 ( i ) –
{( BestVulture1 (i ) × P ( i )
BestVulture1 (i ) – P ( i )
2
) } ×F

A2=BestVulture2 ( i ) –
{ BestVulture 2 ( i ) × P ( i )
BestVulture2 ( i ) – P ( i )
2
} ×F (3.34)

In Eq. (3.34), BestVulture1 (i) is the best vulture of first group and BestVulture 2 (i) is
the best vulture of the second group, both in the current iteration. F is the vulture
satiation rate, and P(i) is current vector position. Eq. (3.35) shows the process of
aggregation carried by all vultures.

P ( i+1 )=¿ ¿2)/2 (3.35)

Eq. (3.36) shows the aggressive competition for food when |F| < 0.5.

P ( i+1 )=R ( i ) – d ( t ) × F × Levy ( d ) (3.36)

42
3.2.4 CRAYFISH OPTIMIZATION ALGORITHM MODEL:

Fig 3.5: Flowchart Based on COA

The Crayfish Optimization Algorithm (COA) is a swarm intelligence-based

metaheuristic inspired by the social and survival behaviours of crayfish, including
foraging, shredding food, temperature-driven movement, and cave competition [48].
43
The algorithm begins with population initialization and fitness evaluation to
determine the global best (XG) and current (XL) solutions. Fig 3.4 depicts the steps
involved in algorithm.

The temperature influences behaviour and phase transitions and it is given by:
Temp = rand × 15 + 20
(3.9)

The food intake probability of each crayfish is determined by:

( )
2
1 −( temp−μ )
p=C 1 × × exp (3.10)
σ √2 π 2σ
2

where µ is the most suitable temperature for crayfish, σ and C1 controls the intake of
the crayfish at various temperature.

The size of the shredded food is calculated as:

Q=C 3 ×rand ׿ fitness food ¿ (3.11)
where C3 is the food factor.

If the food size Q > 2, crayfish shred food and update their positions using:

X food =exp ( −1Q ) × X food (3.12)

( t +1)
X i , j = X i , j + X food × p × ( cos ( 2 π × rand ) – sin ( 2 π × rand ) ) (3.13)
t

Otherwise, they forage in the opposite direction:

j = ( X i , j – X food ) × p+ p × rand × X i , j
X i(t, +1 ) t t
(3.14)

When the temperature exceeds 30°C, crayfish enter the summer resort stage. In this
phase, cave location is defined as:
( XG+ X L ) (3.15)
X shade =
2

The crayfish position is then updated toward the cave:

44
 t +1 t t
X i , j = X i , j+ C 2× rand ×(X ¿¿ shade – X i , j)¿
(3.16)
where C2 is given by

C 2=2 – ( Tt ) (3.17)

Cave competition among crayfish is modeled by the following update:

t +1 t t
X i , j =X i , j – X z , j +¿ X shade
(3.18)
Where z represents the random individual crayfish and it is given by:
z=round ( rand × ( N – 1 ) ) +1(3.19)

These behaviours are repeated iteratively to improve the population until the stopping
condition is met.

45
 Chapter 4

REFERENCED MODEL

4.1 SYSTEM MODEL:

Wireless sensor networks adopted for the study consisted of one sink node and several
sensor node numbers. The sensor nodes and sink nodes interacted via multi-hop
communication. The network is divided into the number of clusters, the sink node
assigns one cluster head, assistant cluster head, and super cluster head from the sensor
nodes for each cluster. Every sensor node in the cluster transmits the sensed data via a
cluster head node to the sink node [50]. The head node of the cluster is responsible for
collecting data from the cluster members. The collected data is then transferred for
aggregation to the assistant cluster head node and sent to the sink through the head
node of the super cluster. The new position of the particle is calculated using the
following Eq. 4.1.

x i ,d ( t +1 )=x i ,d ( t ) +v i , d ( t+1 )

Each particle has the vector of position, and the velocity vector in the space or field
dimension. particle’s fitness value is measured and local best tracked, and each
iteration determines the best global position by Eq. 4.2.

Globalbest =min { pi ,1 , p i ,2 ,… pi , d } where i ∈ M

To reach the best solution, particle utilizes the local best and global best to update the
velocity and position of the particle. The Eq. 3 defnes the velocity update. The
updating process repeated until it reaches to an acceptable value of Global best.

v i, d =v i , d (t−1 )+ c1 r 1 ( localbest ( t )−x i , d ( t−1 )) +c 2 r 2 ( globalbest ( t ) )

46
Equations 4.4 and 4.5 calculates the transmitter’s energy demand to send a ‘l’ bit
message over a distance of ’d’ meters where the Eelec is the initial energy to run the
transmitter electronics, and Etx−amp(l, d) is the energy required for the transmitter
amplifier electronics.

Etx ( l , d )=E elec ×l+ E tx−amp ( l , d )

{
2
ET ( l , d )= l Eelec +l ε fs d 4,∧d <d 0 (4.5)
l E elec + l ε mp d ,∧d ≥ d 0

Equation 4.6 computes the receiver energy requirement for the receiver to receive ‘l’
bit message.
Erx =E elec ×l

Average communication distance: Average contact distance is the distance from the
sink node and distance between the sensor nodes and the cluster head node which is
given by Eq. 4.7.

F 1=Σ ( 1n ) ∑ dist ( S i,−CH )+dist (CH , sink node)

k k

Residual energy: The residual energy of the sensor nodes is the remaining energy
available for sensing and communication operation. The sensor node with the higher
residual energy selected as the head of the cluster when decreasing the parameter F2.

F 2= ∑
( 1
residual ( CH k ) )
Distance centroid: Distance centroid is the distance between sensor nodes and the
position of the centroid distance, calculated using (4.7) equation.

F 3= ∑ ( n1i ) (∑ S i−distancecentriod ) k

47
Clustering coefficient: The clustering coefficient of the sensor nodes is the measure of
the degree to which sensor nodes tend to cluster together. While maximizing the
cluster coefficient parameter, the sensor node with better connectivity with the
neighboring node is selected as the cluster head.

F 4=∑ Clustercoefficient ( CH k )

The solution to the problem of optimization is to minimize the ftness function as

indicated in Eq. 4.11.

fitness=α × F 1+ β × F 2+ γ × F 3+ δ × F 4

These are the control parameters in the span of [0,1] with α + β + γ + δ = 1. α + β > γ
+ δ to elect the cluster heads with the highest residual, closer to the sink node. The
selection of the cluster head process indicated in pseudocode3.

Equation 4.12 measures the sensor node cost factor. Depending on the cost factor
sensor nodes are joined to the head node of the cluster.

Clustercost ( S i ,CH k ) =
{residualenergy ( CH k ) ×cluster coefficient ( CH k ) }
{ node degree ( CH k ) ×dist ( Si ,CH k ) }
(4.12)

The node second minimum value of residual energy and distance is chosen as
assistant cluster head. Assistant cluster head and super cluster head cost value is
calculated using 4.13 and 4.14.

residualenergy ( S i )
SCH cost ( Si ∈cluster k )=
dist ( Si , CH k ) × dist ( Si , sinknode )
residualenergy ( S i )
ACH cost ( S i ∈ cluster k )=
dist ( Si , CH k ) × dist ( S i , S CH k ∈ cluster k )
48
(4.14)
For each cluster, the energy brink value is calculated using the Eq. 4.15 below.
2
Ebrink =l E elec (N /optc−1)+l EDA ( N /optc)+l E elec +l εfs d toBS

Cluster member nodes calculate the hop distance value for all the neighbor sensor
nodes using Eq. 4.16.

hopdistance={dist (S i , CH k )−dist (nexthopi(Si , CH k ))}/{dist (Si , nexthopi(Si ))}

49
 Chapter 5

SIMULATION RESULTS AND DISCUSSIONS

5.1 INTRODUCTION:
The performance of a Wireless Sensor Network (WSN) depends critically on its
energy consumption pattern, clustering efficiency, and overall routing strategy. Due to
the constrained battery power of individual sensor nodes, the choice of clustering and
routing algorithms significantly affects the network lifetime, stability period, and data
throughput. This study proposes and evaluates the performance of Cray Optimization
Algorithm (COA) for clustering in WSNs, comparing its effectiveness with other
popular swarm-based algorithms such as Particle Swarm Optimization (PSO), Grey
Wolf Optimization (GWO), and the African Vultures Optimization Algorithm
(AVOA). Additionally, a traditional distance-based clustering method is used as a
baseline to establish the relative advantages provided by the aforementioned
optimization methods.

Swarm optimization techniques have been widely employed in Wireless Sensor

Networks (WSNs) with the primary objective of minimizing energy consumption,
thereby enhancing the overall efficiency and lifetime of the network. The empirical
results obtained from these optimization approaches demonstrate a notable
improvement in energy utilization, network lifetime, and overall network
performance. All experiments and simulations were executed using MATLAB.

The performance evaluation was conducted under two distinct experimental cases. In
the first case, the number of sensor nodes was varied from 100 to 700, while the
number of gateway nodes remained fixed at 60. In the second case, the number of
sensor nodes was fixed at 200, and the number of gateway nodes was varied from 60
to 140. Each gateway node was initialized with a starting energy of 10 joules. The

50
simulation parameters adopted throughout these tests are summarized in Table I. The
optimization algorithms were tested for both routing and clustering functionalities
under varying network densities and configurations.

TABLE I : Simulation Parameters

Parameter Value
Area 500×500 m2

Sensor nodes 100-700

Gateways 60-120
Initial energy of sensor nodes 10 J
Number of simulation iterations 700
Communication range 100
do (threshold distance) 75 m
Message 525 bit

ℇfs (free space constant)

Eelec 50 nJ/bit

ℇmp (multipath constant)

10 pJ/bit/m2
0.0013 pJ/bit/m4

Two different network scenarios were considered for evaluation: WSN1 and WSN2.
Both configurations utilized a sensing field area of 500 × 500 m². The base station
(BS) in WSN1 was positioned at coordinates (500, 500), simulating a scenario where
the BS is located at the edge of the deployment region. In contrast, the BS in WSN2
was located at (250, 250), representing a central placement within the sensing region.
This variation was intended to observe the influence of BS location on the efficiency
of the routing and clustering mechanisms.

The data reveal that the COA-based approach consistently outperforms the other
methods in terms of prolonging network lifetime. Specifically, COA exhibited
superior clustering efficiency, which translated into reduced energy dissipation and
more effective load balancing among gateway nodes. This performance edge was
more pronounced in larger sensor deployments and when the BS was centrally
located, as in WSN2. These findings underscore the potential of COA and similar
swarm-based algorithms in achieving scalable and energy-efficient WSN deployment.

51
5.2 SIMULATION RESULTS:
The simulation results are discussed in three parts: first, a comparison among
distance-based, PSO, GWO, and AVOA; second, a comparison among distance-
based, PSO, GWO, and COA; and third, an integrated comparison including all five
algorithms.

5.2.1 COMPARATIVE ANALYSIS I: DISTANCE-BASED CLUSTERING VS

PSO VS GWO VS AVOA

Fig. 5.1 Comparison of algorithms (PSO, GWO,AVOA) using the number of rounds before the first
node die w.r.t number of sensors with the base station at [250,250]

In the first part of the comparative study, four clustering strategies—distance-based,

PSO, GWO, and the proposed AVOA—are examined. The aim of this evaluation is to
analyze how the bio-inspired AVOA algorithm enhances the performance of WSNs,
particularly in prolonging the network lifetime, reducing energy consumption, and
increasing throughput. The experiments were carried out by sensor nodes randomly
deployed in a two-dimensional sensing field of 500×500 meters. Several simulation
scenarios were executed by varying the number of nodes to assess algorithm behavior
under different network densities.
52
Fig. 5.2 Comparison of algorithms (PSO, GWO,AVOA) using the number of rounds before the first
node die w.r.t number of sensors with the base station at [500,500]

Fig 5.1 and fig 5.2 shows the number of rounds before first node dies with the
application of different algorithms in WSN2 and WSN1 respectively. The number of
alive nodes over time is a crucial metric to evaluate the energy efficiency and
reliability of a clustering algorithm. The results indicate that the traditional distance-
based clustering approach, which selects cluster heads purely based on proximity to
the base station, performs poorly. It leads to rapid energy depletion and node deaths
because it does not account for residual energy or load balancing. As observed in the
simulation, the first node in distance-based clustering dies within the first 250 to 300
rounds, and the number of alive nodes falls sharply thereafter. PSO, on the other hand,
demonstrates improved performance owing to its nature of optimizing CH selection
based on both residual energy and distance. However, due to its tendency to converge
early, PSO is still limited in its ability to maintain an evenly distributed energy
consumption across the network, resulting in an FND (First Node Dead) around 1000

53
rounds with number of sensors 100 and gateway 60 in both cases. GWO improves
upon this by simulating a hierarchical leadership mechanism inspired by grey wolves.
This approach ensures better exploration and exploitation, enabling improved CH
selection. GWO’s performance is more robust than PSO, achieving a higher number
of rounds before the first node dies.

The AVOA algorithm significantly outperforms all three, especially in scenarios

involving a larger number of nodes. Inspired by the strategic foraging and attacking
behavior of African vultures, AVOA incorporates both local and global search
mechanisms that adapt dynamically based on the fitness of candidate solutions. This
dual-phase behavior allows it to avoid premature convergence and ensures that CHs
are chosen in a way that balances energy load across the network. Consequently,
AVOA delays the FND until around 400 rounds and maintains a higher number of
alive nodes throughout the simulation, even in networks with 600 sensor nodes in
WSN1. Its adaptive strategy ensures more stable cluster formation and better resource
distribution, which is evident in the energy consumption graphs. The total residual
energy over rounds declines most rapidly for the distance-based method, followed by
PSO and GWO. AVOA demonstrates a more gradual decline, suggesting better
energy management. It ensures that energy usage is evenly distributed, preventing
early node failures and maintaining network connectivity for a longer time.

Fig. 5.3 Comparison of algorithms (PSO, GWO, AVOA) using the number of rounds before the first
node die w.r.t numberof gateways with the base station at [250,250]
54
Fig. 5.4 Comparison of algorithms (PSO, GWO, AVOA) using the number of rounds before the first
node die w.r.t number gateways with the base station at [500,500]

In addition to the comparative study based on varying sensor nodes, an alternate

evaluation was carried out by keeping the number of sensor nodes constant while
varying the number of gateways. This analysis is crucial in understanding how the
clustering and routing algorithms adapt to changes in infrastructure support (i.e.,
gateways) within a WSN environment. Fig. 5.3 and Fig. 5.4 represent the outcomes of
this study for WSN1 and WSN2, respectively.

In these figures, the x-axis represents the number of gateway nodes, and the y-axis
denotes the number of rounds completed before the first node dies (FND). The graphs
clearly show an increasing trend in the number of rounds before FND as the number
of gateways increases for all algorithms under consideration—Distance-based, PSO,
GWO, AVO.

This increasing trend can be attributed to the fact that the introduction of additional
gateways reduces the communication load on each sensor. More gateways offer more
options for cluster heads to forward their data, resulting in lower transmission
distances, reduced energy consumption, and ultimately, longer node lifespans.
55
Consequently, the energy burden is more evenly distributed among the nodes,
delaying the first node death and thus extending the operational lifetime of the
network.

Among all the algorithms tested, AVOA consistently outperformed the rest, showing
the highest number of rounds before FND at every gateway increment. This indicates
that AVOA’s adaptive exploration–exploitation behavior is especially effective in
utilizing additional gateways for optimizing data routes and balancing the energy load
more efficiently across the network. Its flexibility in adapting cluster formations to the
increased infrastructure allows it to achieve greater energy efficiency and reliability in
WSN deployments.

PSO and GWO also displayed consistent improvements with the increasing number of
gateways but did not match the performance level of AVOA. These results validate
that increasing the number of gateways while keeping sensor count constant has a
direct and positive impact on the FND, and that advanced algorithms like AVOA can
significantly capitalize on this configuration to extend network life. Hence, optimizing
gateway placement and count, in conjunction with advanced clustering algorithms, is
a promising approach for enhancing energy efficiency in WSNs.

5.2.2 COMPARATIVE ANALYSIS II: DISTANCE-BASED CLUSTERING VS

PSO VS GWO VS COA
The second part of the study introduces the Cray Optimization Algorithm (COA) into
the comparative framework, analyzing its performance alongside distance-based,
PSO, and GWO clustering techniques. COA is inspired by the behavior of crayfish in
nature and employs a bio-inspired optimization mechanism combining exploration
and exploitation. It differs from PSO and GWO in its strategy for population
movement and avoidance of local optima. The objective of this comparison is to
evaluate how COA performs in comparison to existing techniques when it comes to
clustering efficiency, energy usage, and node survival.

56
Fig. 5.5. Comparison of algorithms (PSO, GWO, COA) using the number of rounds before the first
node die w.r.t number of sensors with the base station at [500,500] by taking average over multiple
trials.

Fig. 5.6. Comparison of algorithms (PSO, GWO, COA) using the number of rounds before the first
node die w.r.t number of sensors with the base station at [250,250] .

Fig 5.5 and 5.6 shows the comparison of different algorithms with COA. Similar to
the previous experiments, the network area, initial energy of nodes, and packet size
were kept constant. COA demonstrates considerable improvement over the traditional

57
distance-based approach. In scenarios with 100 nodes, COA significantly extends the
stability period and the overall network lifetime. It achieves this by smartly selecting
cluster heads that not only minimize intra-cluster distances but also balance energy
consumption among nodes. The first node death in COA occurs after nearly 2000
rounds, indicating better energy conservation than PSO and comparable to GWO.

In terms of energy consumption, COA achieves a more gradual depletion than

distance-based and PSO. It shows patterns similar to GWO, especially in early and
mid rounds. COA’s use of modified Lévy flight and random walk behaviors in its
candidate solution updates helps prevent early convergence and results in better
energy distribution. However, in high-density networks, COA’s performance starts to
decline slightly. This may be attributed to increased communication overhead among
more clusters and a higher complexity in selecting optimal CHs.

Fig. 5.7 Comparison of algorithms (PSO, GWO, COA) using the number of rounds before the first
node die w.r.t number of gateway with the base station at [500,500] .

58
Fig. 5.8 Comparison of algorithms (PSO, GWO, COA) using the number of gateways with the base
station at [250,250] .

Beyond the evaluation of algorithm performance with varying numbers of sensors, the
second dimension of analysis involved observing the behavior of each algorithm—
Distance-based, PSO, GWO, and COA—under conditions where the number of
sensor nodes remained constant while the number of gateway nodes was
systematically varied. This evaluation was critical to understand the responsiveness
and scalability of each algorithm with respect to infrastructure enhancements,
particularly in scenarios where increasing the number of gateways can potentially
enhance network lifetime by alleviating data transmission loads on sensor nodes.

Fig. 5.7 and Fig. 5.8 depict the performance of these four algorithms in WSN1 and
WSN2 respectively under this modified scenario. Here, the x-axis represents the
varying number of gateways, while the y-axis shows the number of rounds before the
First Node Dies (FND). In both WSN1 and WSN2 environments, a clearly increasing
trend was observed for all four algorithms as the number of gateways increased,
confirming the hypothesis that more gateways contribute to better energy balancing
and longer network longevity.

59
Among the algorithms evaluated, COA demonstrated the best performance in both
WSN1 and WSN2 under increasing gateway configurations. Its biologically inspired
optimization process and ability to explore a wide range of solutions made it highly
adaptive to new gateway positions. This adaptiveness allowed COA to optimize the
clustering process more effectively, resulting in a higher number of rounds before
FND. In both WSN1 and WSN2, COA’s FND values increased sharply with
additional gateways, outperforming PSO and GWO.

The GWO algorithm also showed significant improvements, particularly in WSN2,

which had a denser node distribution. Its pack-based leadership hierarchy and
exploitation mechanism helped maintain consistent performance improvements as
gateways increased, although its performance remained slightly lower than that of
COA.PSO, while generally effective in optimizing clustering, demonstrated a more
modest improvement compared to COA and GWO. Its convergence behavior was
influenced by its tendency to sometimes prematurely settle on local optima, which
became evident in scenarios where complex gateway distributions required more
dynamic adjustment of clusters. Nevertheless, PSO still performed better than the
Distance-based method, which served as a baseline and consistently exhibited the
lowest performance due to its static routing behavior and lack of optimization
strategy.

The Distance-based approach, although showing a positive trend with increasing

gateways, failed to match the performance of the metaheuristic algorithms. Its rigid
selection of the nearest gateway or cluster head did not adapt well to changes in
gateway distribution, resulting in suboptimal routing paths and quicker energy
depletion in certain regions of the network.

Notably, the performance gap between COA and the other algorithms widened with
the increase in the number of gateways, especially in WSN2. This indicates COA’s
superior capability in leveraging infrastructure expansion to optimize energy usage.
WSN2, with its denser topology, provided more opportunities for cluster adjustment
60
and energy load balancing, which COA utilized more effectively than the other
algorithms.

In conclusion, this second evaluation clearly demonstrates that increasing the number
of gateways while keeping the number of sensor nodes constant significantly benefits
all algorithms in terms of delaying the first node death. However, among the tested
techniques, COA emerges as the most efficient algorithm in both WSN1 and WSN2
configurations under this scenario. These findings underscore the importance of
selecting intelligent optimization strategies like COA when designing WSNs that rely
heavily on gateway configurations for efficient data transmission and extended
network lifetime.

5.2.3 COMPARATIVE ANALYSIS III: DISTANCE-BASED CLUSTERING VS

PSO VS GWO VS AVOA VS COA
In this third and final part, a complete comparative analysis is conducted by
evaluating all five clustering techniques: distance-based, PSO, GWO, AVOA, and
COA. The objective is to offer a holistic understanding of how each algorithm
performs under the same network conditions and identify which method delivers the
best trade-off between energy conservation, network stability, and data throughput.

61
Fig. 5.9. Comparison of algorithms (PSO, GWO, AVOA, COA) using the number of rounds before the
first node die w.r.t number of sensors with the base station at [500,500]

Fig. 5.10. Comparison of algorithms (PSO, GWO, AVOA, COA) using the number of rounds before
the first node die w.r.t number of sensors with the base station at [250,250]
62
From the perspective of node survival, the simulations clearly reveal a performance
hierarchy. Distance-based clustering consistently results in the shortest network
lifetime, with nodes beginning to die as early as 250 rounds. PSO delays the FND to
around 500 rounds, and GWO pushes it to approximately 600 rounds. AVOA further
extends it to around 650–700 rounds. However, COA leads with the most significant
improvement, delaying the FND until 800–850 rounds. This shows that AVOA not
only selects optimal CHs initially but continues to update them dynamically to adapt
to energy levels and node distribution.

From the combined results, it is evident that COA offers the most balanced
performance across all metrics. Its novel approach to CH selection, inspired by
scavenging and foraging behaviors of crayfish, allows it to achieve optimal results in
network longevity, energy distribution, and data transmission. AVOA and GWO
stand as competent alternatives, especially in medium-density scenarios, while PSO
and distance-based methods are outperformed across most metrics.

63
 Chapter 6

CONCLUSION

This thesis explored energy-efficient strategies for Wireless Sensor Networks

(WSNs) by implementing and evaluating swarm optimization techniques aimed at
minimizing energy expenditure and extending network lifetime. The primary focus
was on optimizing clustering and routing using four algorithms—Particle Swarm
Optimization (PSO), Grey Wolf Optimizer (GWO), African Vultures Optimization
Algorithm (AVOA) and Cray Optimization Algorithm (COA)—and comparing
them with a baseline Distance-based approach. Performance was assessed based on
the number of rounds before the First Node Dies (FND), a key indicator of energy
efficiency.

Simulations were conducted in two distinct WSN scenarios, WSN1 and WSN2,
representing different network topologies. Two experimental cases were
considered: (i) varying the number of sensors with a fixed number of gateways,
and (ii) varying the number of gateways while keeping sensor count constant. In

64
both scenarios, swarm optimization techniques consistently outperformed the
Distance-based method.

COA delivered the best overall performance when the number of sensors varied,
demonstrating excellent exploration and exploitation capabilities that extended the
network’s lifetime. GWO and PSO also improved FND compared to the Distance-
based method but were slightly less efficient than AVOA and COA.

The study highlights the effectiveness of swarm-based methods in prolonging

WSN lifetime and reveals that increasing gateway nodes, even with constant sensor
density, significantly enhances network performance. These findings provide
valuable insights for designing scalable, energy-efficient WSN architectures.

While this work focused exclusively on the FND metric, future research could
incorporate additional parameters like latency, packet delivery, and dynamic node
behavior. Overall, this study establishes AVOA and COA as promising
optimization strategies for sustainable WSN deployment.

65
 References

[1] P. Kuila and P. K. Jana, “Energy efficient clustering and routing algorithms for
wireless sensor networks: Particle swarm optimization approach”, Engineering
Applications of Artificial Intelligence, vol. 33, pp. 127 - 140, 2014.
[2] F. Bajaber & I. Awan, “An efficient cluster-based communication protocol for
wireless sensor networks,” Telecommunication Systems, vol. 55, no. 3, pp. 387–401,
2014.
[3] S. Mirjalili, S. M. Mirjalili, and A. Lewis, “Grey Wolf Optimizer”, Advances in
Engineering Software, vol. 69, pp. 46-61, 2014.
[4] A. Norouzi, & A. H. Zaim, “Genetic algorithm application in optimization of
wireless sensor networks,” The Scientific World Journal, pp. 1–15, 2014.
[5] M. I. Chidean, E. Morgado, M. S. Junquera, J. R. Bargueno, J. Ramos, & A. J.
Caamaño, “Energy efficiency and quality of data reconstruction through data-coupled
clustering for self-organized large-scale WSNs,” IEEE Sensors Journal, vol. 16, no.
12, pp. 5010–5020, 2016.

66
[6] S. Noureddine, B. Khelifa, and B. Mohammed, “Approach to minimizing
consumption of energy in wireless sensor networks,” International Journal of
Electrical and Computer Engineering, vol. 10, no. 3, pp. 2551, 2020.
[7] P. Kathiroli and K. Selvadurai, “Energy efficient cluster head selection using
improved Sparrow Search Algorithm in Wireless Sensor Networks,” Journal of King
Saud University-Computer and Information Sciences, vol. 34, no. 10, pp. 8564-8575,
2022.
[8] H. Jia, H. Rao, C. Wen, and S. Mirjalili, “Crayfish optimization algorithm”,
Artificial Intelligence Review, vol. 56, pp. 1919 - 1976, 2023.
[9] V. Chandran and P. Mohapatra, “Enhanced opposition-based grey wolf optimizer
for global optimization and engineering design problems”, Alexandria Engineering
Journal, vol. 76, pp. 429- 467, 2023.
[10] H.R.H. Al Dallal and D.A. Sultan, “An optimized algorithm design for a target
tracking in wireless sensor networks”, Spec. Jour. of Inno. Ref. and Dev., vol. 12, pp.
76–94, 2023.
[11] P. Rawat, P. Kumar, and S. Chauhan, "Fuzzy logic and particle swarm
optimization-based clustering protocol in wireless sensor network," Soft Computing,
vol. 27, no. 9, pp. 5177-5193, 2023.
[12] S.D. Mishra, S. Dutta, and D. Verma. "Energy-Efficient and Reliable Clustering
with Optimized Scheduling and Routing for Wireless Sensor Networks," Multimedia
Tools and Applications, pp. 1-27. 2024.
[13] B.H. Goud, T.N. Shankar, B. Sah, R. Aluvalu, “Energy optimization in path
arbitrary wireless sensor network,” Expert Systems, vol. 41, no. 2, pp. 13282, 2024.
[14] I. Surenther, K.P. Sridhar, M.K. Roberts, “Enhancing data transmission
efficiency in wireless sensor networks through machine learning-enabled energy
optimization: A grouping model approach,” Ain Shams Engineering Journal, vol. 15,
no. 4, pp.102644, 2024.
[15] V. Prakash, D. Singh, S. Pandey, S. Singh, and P.K. Singh, “Energy-
Optimization Route and Cluster Head Selection Using M-PSO and GA in Wireless
Sensor Networks”, Wireless Personal Communications, pp.1-26, 2024.

67
[16] Y. Razooqi, M. Al-Asfoor, and M.H. Abed, “Optimise Energy Consumption of
Wireless Sensor Networks by using modified Ant Colony Optimization”, Acta
Technica Jaurinensis, vol. 17, pp.111-117, 2024.
[17] J.S. Abdullah, and M.I. Aal-Nouman, “Efficient Deployment of Nodes in a WSN
Using a Modified PSO Algorithm”, 1st International Conference on Emerging
Technologies for Dependable Internet of Things, pp. 1-6, 2024.
[18] P. Mishra, R.K. Dash, T. Choudhury, and K. Kotecha, “Optimizing Residual
Energy and Delay in WSN Routing using Particle Swarm Optimization”, Journal
homepage: http://iieta. org/journals/isi, vol. 29, pp.761-770, 2024.
[19] M.A. Lilo, A.K. Yasari, M.M. Hamdi, and A.D. Abbas, “Transmission Power
Reduction Based on an Enhanced Particle Swarm Optimization Algorithm in Wireless
Sensor Network for Internet of Things”, ARO-THE SCIENTIFIC JOURNAL OF
KOYA UNIVERSITY, vo. 12, pp.61-69, 2024.
[20] J. Dev and J. Mishra, “Energy Efficient Routing in Cluster Based Heterogeneous
Wireless Sensor Network Using Hybrid GWO and Firefly Algorithm”, Wireless
Personal Communications, vol. 132, pp.997-1028, 2024.
[21] M. Kaddi, M. Omari, K. Salameh, and A. Alnoman, “Energy-Efficient Clustering
in Wireless Sensor Networks Using Grey Wolf Optimization and Enhanced
CSMA/CA,” Sensors, vol. 24, no. 16, p. 5234, 2024.
[22] H. Wang, K. Liu, C. Wang, and H. Hu, “Energy-Efficient, Cluster-Based
Routing Protocol for Wireless Sensor Networks Using Fuzzy Logic and Quantum
Annealing Algorithm,” Sensors, vol. 24, no. 13, p. 4105, 2024.
[23] L. M. Alkwai and K. Yadav, “Energy-efficient clustering algorithm using
distributed fuzzy-logic to prolong the survivability of wireless sensor networks,”
Internet Technology Letters, vol. 7, no. 3, e549, 2024.
[24] H. Koyuncu, G. S. Tomar, and D. Sharma, “A New Energy Efficient Multitier
Deterministic Energy-Efficient Clustering Routing Protocol for Wireless Sensor
Networks,” Symmetry, vol. 12, no. 5, p. 837, 2020.
[25] A. Farzaneh, M.-A. Badiu, and J. P. Coon, “LEAST: a Low-Energy Adaptive
Scalable Tree-based routing protocol for Wireless Sensor Networks,” arXiv preprint
arXiv:2211.09443, 2022.
68
[26] G. P., S. K. K. N., and A. S. R. Murthy, “Clustering Based Energy Efficient
Routing for Wireless Sensor Network Using Particle Swarm Optimization,”
International Journal of Communication Networks and Information Security, vol. 15,
no. 3, pp. 243–260, 2023.
[27] B. M. Sahoo, H. M. Pandey, and T. Amgoth, “Metaheuristic algorithm for energy
efficient clustering scheme in wireless sensor networks,” Microprocessors and
Microsystems, vol. 97, p. 104898, 2023.
[28] J. Vellaichamy et al., “Wireless Sensor Networks Based on Multi-Criteria
Clustering and Optimal Bio-Inspired Algorithm for Energy-Efficient Routing,”
Applied Sciences, vol. 13, no. 5, p. 2801, 2023.
[29] S. J. Jagadeesh and I. Muthulakshmi, “A novel oppositional artificial fish swarm
based clustering with improved moth flame optimization based routing protocol for
wireless sensor networks,” Energy Systems, pp. 1–21, 2022.
[30] P. V. S. Reddy, “Fuzzy conditional inference and application to wireless sensor
network fuzzy control systems,” in Proceedings of the 2015 IEEE 12th International
Conference on Networking, Sensing and Control, pp. 1–6, 2015.
[31] F. Fanian and M. K. Rafsanjani, “A new fuzzy multi-hop clustering protocol with
automatic rule tuning for wireless sensor networks,” Applied Soft Computing, vol. 89,
p. 106114, 2020.
[32] D. Chandirasekaran and T. Jayabarathi, “Cat swarm algorithm in wireless sensor
networks for optimized cluster head selection: A real time approach,” Cluster
Computing, vol. 22, no. 5, pp. 11351–11361, 2019.
[33] W. M. Zheng et al., “Cluster head selection strategy of WSN based on binary
multi-objective adaptive fish migration optimization algorithm,” Applied Soft
Computing, vol. 148, p. 110875, 2023.
[34] E. X. Zheng and R. R. Liu, “Routing Technology in Wireless Sensor Network
Based on Ant Colony Optimization Algorithm,” Wireless Personal Communications,
vol. 95, no. 4, pp. 1911–1925, 2017.
[35] N. Muruganantham and H. El-Ocla, “Routing Using Genetic Algorithm in a
Wireless Sensor Network,” Wireless Personal Communications, vol. 111, no. 4, pp.
2703–2732, 2020.
69
[36] P. Maheshwari, A. K. Sharma, and K. Verma, “Energy efficient cluster based
routing protocol for WSN using butterfly optimization algorithm and ant colony
optimization,” Ad Hoc Networks, vol. 110, p. 102317, 2021.
[37] M. M. Asiri et al., “Metaheuristics Enabled Clustering with Routing Scheme for
Wireless Sensor Networks,” Computers, Materials & Continua, vol. 73, no. 3, pp.
5491–5507, 2022.
[38] X. Yang et al., “THSI-RP: A two-tier hybrid swarm intelligence based node
clustering and multi-hop routing protocol optimization for wireless sensor networks,”
Ad Hoc Networks, vol. 149, p. 103129, 2023.
[39] P. Srinivas and P. Swapna, “Quantum tunicate swarm algorithm based energy
aware clustering scheme for wireless sensor networks,” Microprocessors and
Microsystems, vol. 94, p. 104633, 2022.
[40] S. M. Abas et al., “Quantum Computing-Inspired Genetic Algorithm for Network
Optimization in WSN,” International Journal of Intelligent Systems and Applications
in Engineering, vol. 12, no. 2, pp. 188–194, 2024.
[41] R. Nikouei et al., “A quantum-annealing-based approach to optimize the
deployment cost of a multi-sink multi-controller WSN,” Procedia Computer Science,
vol. 155, pp. 250–257, 2019.
[42] Z. J. Huang et al., “Evolutionary Method of Sink Node Path Planning Guided by
the Hamiltonian of Quantum Annealing Algorithm,” IEEE Access, vol. 9, pp. 53466–
53479, 2021.
[43] A. T. A. Al-Khayyat and A. Ibrahim, “Energy optimization in WSN routing by
using the K-means clustering algorithm and ant colony algorithm,” in Proceedings of
the 2020 4th International Symposium on Multidisciplinary Studies and Innovative
Technologies (ISMSIT), pp. 1–4, 2020.
[44] M. Kaedi, A. Bohlooli, and R. Pakrooh, “Simultaneous optimization of cluster
head selection and inter-cluster routing in wireless sensor networks using a 2-level
genetic algorithm,” Applied Soft Computing, vol. 128, p. 109533, 2022.
[45] Y. Kongsorot et al., “An enhanced fuzzy-based clustering protocol with an
improved shuffled frog leaping algorithm for WSNs,” Expert Systems with
Applications, vol. 198, p. 116942, 2022.
70
[46] V. K. Alla and M. Mallikarjuna, “Routing protocol based on bacterial foraging
optimization and type-2 fuzzy logic for wireless sensor networks,” in Proceedings of
the 2020 11th International Conference on Computing, Communication and
Networking Technologies (ICCCNT), pp. 1–6, 2020.
[47] R. Guo et al., “Efficient routing algorithms for maximizing network lifetime in
chain-type wireless sensor networks using multiple sinks,” in Third International
Conference on Green Communication, Network, and Internet of Things (CNIoT
2023), vol. 12814, pp. 120–128, 2023.
[48] A. Singh, “Chaotic Bee Colony Optimization for Enhancing Life Span Of
Wireless Sensor Networks,” in 2022 International Conference on Computational
Modelling, Simulation and Optimization (ICCMSO), pp. 158–165, 2022.
[49] S. J. Janakiraman, “Improved bat optimization algorithm and enhanced artificial
bee colony-based cluster routing scheme for extending network lifetime in wireless
sensor networks,” International Journal of Communication Systems, vol. 36, no. 5,
e5428, 2023.
[50] M. F. Suleiman and U. Adeel, “An Energy Efficient Multi-Region Based Mobile
Routing Protocol to Improve Network Lifetime in Wireless Sensor Networks,” in
2023 3rd International Conference on Electrical, Computer, Communications and
Mechatronics Engineering (ICECCME), pp. 1–7, 2023.

71
 LIST OF PUBLICATIONS
[1] A. Jha, A. Maurya, Aprajita, A. Pandey, “Minimization of Energy Expenditure in
Wireless Sensor Networks using Crayfish Optimization Algorithm”, accepted in IEEE
Second International Conference on Networks and Soft Computing, 2025

[2] A. Jha, A. Maurya, Aprajita, A. Pandey, “Energy Expenditure Minimization in

Wireless Sensor Networks using African Vultures Optimization Algorithm”,
communicated in Springer International Conference on Power and Intelligent Control
Systems, 2025

72
73