Optimizing Energy Efficiency in Wireless Sensor Networks via Cluster-Based Routing and a Hybrid Optimization Approach

Wireless Sensor Networks (WSNs) have garnered substantial interest due to their versatility in deployment and myriad applications [1]. These networks comprise an array of intelligent, multifunctional, low-power sensor nodes that interface with Base Stations (BS) to monitor and interact with the external environment [2, 3]. Sensor nodes are endowed with the capability to self-organize and collaborate, thus facilitating the establishment and maintenance of the network's infrastructure [4]. Each sensor, equipped with storage memory, a processor, and communication channels, undertakes the critical functions of data collection, processing, and transmission [5, 6]. The integration of embedded computing, distributed information processing, sensor technology, and communication technology has accelerated the use of WSNs in a variety of industries, including disaster relief, environmental monitoring, industrial manufacturing, armed forces, traffic management, and healthcare. [7, 8].

The longevity of sensor nodes is constrained by their reliance on non-rechargeable batteries. Despite the extensive utility of WSNs, they are impeded by limitations such as finite storage, battery life, and transmission range [9]. Consequently, energy efficiency is identified as the paramount objective within the WSN paradigm [10]. The current research focuses on developing a cluster-based routing protocol tailored for agricultural applications, where sensor technology underpins production optimization, cost management, and crop growth monitoring. These sensors measure critical agricultural parameters, including pressure, wind speed, temperature, and humidity, and are instrumental in predicting natural disasters and weather conditions [11, 12].

Energy-efficient routing protocols are essential for judicious energy utilization during packet transmission. Clustering-based routing protocols have been observed to outperform their non-clustering counterparts by mitigating energy losses due to idle listening, collisions, and over-hearing [13]. Within clusters, sensors, designated as Cluster Members (CMs), record environmental data and transmit it to their respective Cluster Heads (CHs). The CHs then aggregate and filter the data to reduce redundancy before forwarding it to the Base Station (BS) [14].

Previous methodologies have exhibited shortcomings, including suboptimal selection of fitness function parameters and reliance on single-hop data transfers. The suggested approach includes a fitness function that makes the choice of residual energy, proximity to the BS, and distance to neighboring nodes, thereby facilitating the selection of an optimal CH. This optimal CH selection is critical in reducing node energy consumption and achieving energy-efficient balance within the network [15].

The contributions of this research are as follows:

·Minimization of energy consumption in WSNs is achieved through K-means clustering and optimal CH selection, utilizing the Energy-based Multiobjective Hybrid Optimization Algorithm (E-MHOA). The E-MHOA amalgamates the Cuckoo Search Algorithm (CSA), with its superior global search capability through levy flight, and the Whale Optimization Algorithm (WOA), recognized for its efficiency with fewer parameters required for analysis.

·The E-MHOA is used to create an energy-efficient routing protocol that maximises multi-objective fitness functions, including energy, communication cost, node degree, and coverage.

·The development of an energy-efficient WSN enhances data delivery across the network.

The structure of this document is as follows: The relevant work on cluster-based routing in WSNs is reviewed in Section 2. Section 3 provides further details on the E-MHOA technique. Section 4 presents the E-MHOA approach's findings. The paper is finally concluded in Section 5.

An energy-efficient WSN is created using the E-MHOA technique in order to achieve dependable communication. Three crucial stages make up the E-MHOA method: routing path development, CH selection, and clustering. Here, E-MHOA is used to choose the best CH and route to raise the WSN's energy efficiency. By enhancing the E-MHOA approach with different fitness measures, including energy, communication cost, node degree, and node coverage, the WSN's data delivery is enhanced. The E-MHOA method's architecture is depicted in Figure 1.

3.1 K-means based clustering approach

The nodes are first randomly placed throughout the region of interest. After that, they are grouped using the K-means method, which divides the sensors into an arbitrary number of clusters. Given that the main factor influencing this K-means clustering is the Euclidean distance between the nodes.

One of the main goals of system design in WSN is energy optimisation.

The energy-model uses multipath fading (d⁴ powerloss) or free space (d² powerloss) according to the distance between the transmitter and receiver. Eq. (1) expresses the energy used to transfer a data (qq) over a distance (dd):

$E_{R X}(q, d)=\left\{\begin{array}{l}q \times E_{\text {elec }}+q \times E_{f s} \times d^2, d \leq d_0 \\ q \times E_{\text {elec }}+q \times E_{m p} \times d^4, d>d_0\end{array}\right.$     (1)

where the distance threshold is represented by the symbol d0, and $\mathrm{E}_{\mathrm{Tx}}$ is the amount of energy needed to transmit in free space and multipath, respectively, and is abbreviated as $E_{\text {elec }}$ and $E_{f s}$ and $E_{m p}$ accordingly i.e., $d_{0=} \sqrt{\frac{E_{f s}}{E_{m p}}}$ Eq. (2) expresses the energy consumed by each node while receiving the data, is computed based on Eq. (3).

$E_{R X}(q, d)=q \times E_{\text {elec }}$     (2)

$E_{\text {consumed }}=q \times E_{T X}+E_{R X}$     (3)

where, $E_{\text {consumed }}$ defines the amount of energy that each network sensor uses. The network is first clustered, and then each cluster's optimal CH is chosen using E-MHOA.

1.png

Figure 1. The architecture of the E-MHOA method

3.2 Multi-objective hybrid optimization technique based on energy

To increase energy efficiency, the best CHs from the clusters are chosen in this phase. Here, the E-MHOA with the best number of fitness functions are used to identify the optimal CHs. This hybrid algorithm, which combines WOA and CSA, is utilized for both routing and CH selection. CSA is typically an algorithm inspired by nature that replicates the cuckoo bird's reproductive cycle [26]. Furthermore, WOA imitates the humpback whales' intelligence-seeking behaviours, whereas the bubble-net feeding approach is used to describe the aging behaviour [27].

3.2.1 Multi-objective fitness function definition for E-MHOA

To improve the energy efficiency of the WSN, the best mix of distinct fitness functions, such as residual energy $\left(f f_1\right)$, communication cost $\left(f f_2\right)$, degree of node $\left(f f_3\right)$ and node coverage $\left(f f_4\right)$ in the E-MHOA are employed. E-MHOA's multi-objective fitness function is expressed as in Eq. (4):

Fitness$=\beta_1 \times f f_1+\beta_2 \times f f_2+\beta_3 \times f f_3+\beta_4 \times f f_4$     (4)

where the weighted parameters $\beta 1, \beta 2, \beta 3$, and $\beta 4$ are employed to convert fitness scores from many targets into a single goal. The definition and formulation of these parameters are given as follows:

The proposed CH selection process utilizes a high er energy node as an optimal candidate while selecting the CH. The CH is required to have a better energy budget to balance the energy consumption because the CH has the responsibilities of cluster management and data aggregation from the CM. The residual energy of the CH is derived in Eq. (5):

$f f_1=\sum_{i=1}^m \frac{1}{E_{C H_i}}$     (5)

where $m$ represents the total number of CHs in the WSN and $E_{C H_i}$ represents the residual energy of the cluster head $i$.

Eq. (6) demonstrates the price needed to talk to nearby nodes:

$f f_2=\frac{d_{a v g}^2}{d_0^2}$     (6)

where, $d_{\text {avg }}^2$ indicates the average separation between the node and its neighbor. and the $d_0^2$ denotes the radius of the node.

Node degree is the number of sensors that are accessible from the $\mathrm{CH}$. This is utilized for balancing the load in the $\mathrm{CH}$ which is expressed in Eq. (7):

$f f_3=\sum_{i=1}^m C M_i$     (7)

where, an amount of cluster members in each cluster is denoted as $C M_i$.

Furthermore, network coverage is thought to increase the scalability of the network. The network coverage derived using the radius of the node is expressed in Eq. (8):

$f f_4=\frac{1}{N} \sum_{i=1}^N r(N)$     (8)

where, number of sensors in the WSN is $N$ and radius of the sensors is defined as $(N)$.

3.2.2 Working of E-MHOA based CH selection

E-MHOA combines the Cuckoo Search Algorithm and the Whale Optimization Algorithm to find the best CH. Using the fitness function, the CSA Algorithm finds the best solutions (i.e., the location of the CH). These best solutions are fed into WOA to identify the best CH. The working process of CH section is mainly comprised of two steps such as initialization and iterative process.

3.2.3 Initialization

Every solution in E-MHOA stands for a set of potential sensors that must be chosen to be CHs. Every solution has a dimension equal to the number of $\mathrm{CHs}$. Here, a random node between 1 and $N$ is used to initialize each solution location. Eq. (9) expresses the $i^{\text {th }}$ solution of the proposed E-MHOA:

$X_i=\left(X_i 1(t), X_i 2(t), \ldots, X_i m(t)\right)$     (9)

where, each solution location $X_{i, l}, 1 \leq l \leq m$ denotes the node among the 1 and $N$ in the network.

3.2.4 Iterative process of E-MHOA

The iterative procedure for choosing the best CHs from the clusters uses the initial solutions as input. The following are the key guidelines taken into consideration by the CSA: 1) Every cuckoo lay single egg simultaneously and deposits it in a randomly chosen nest; 2). The best nests with better-quality eggs (i.e., ideal solutions) are passed down to subsequent generations; and 3) there is always a constant supply of host nests that can be accessed, and a host found the alien egg with a probability $P_a \in[0,1]$. In that case, the host either throws the egg or leave the nest to construct the new nest in a new location. According to the aforementioned steps, the processes of the CSA is derived and given as follows:

A $i$-th cuckoo performs the Levy flight to create an updated location $(t+1)$, which is expressed in Eq. (10):

$X_i(t+1)=X_i(t)+\alpha \oplus \operatorname{Le}^{\prime} v y(\lambda)$     (10)

where, the step size is denoted as $\alpha>0 ; X_i(t)$ and $X_i(t+1)$ refers to the current and next position of the population; $\oplus$ denotes the entry wise multiplications. Moreover, the Le'vy used in this E-MHOA is used for effective searching over the search space. An explorative random walk obtained usingLe'vyflights is expressed in Eq. (11): 

$L e^{\prime} v y \sim u=t-\lambda, 1<\lambda \leq 3$     (11)

where, $\lambda$ is the parameter, which represents the average or expected occurrence of the event over the given time period. The best solutions from the CSA (i.e., locations of $\mathrm{CH}$ ) derived using formulated fitness function are given as an input to the WOA to obtain optimal CHs. The best candidate solution from the CSA is considered as the target prey and the other solutions are tried to move towards the best solutions based on the Egs. (12)-(13):

$\vec{D}=\left|\vec{C} \cdot \vec{X}_b-\vec{X}(t)\right|$     (12)

$\vec{X}(t+1)=\vec{X}_b(t)-\vec{A} \cdot \vec{D}$     (13)

where, the current iteration i-st; the location vector of optimal and another population are represented as $\vec{X} b$ and $\vec{X}$ respectively. Moreover, the $\vec{A}$ and $\vec{C}$ are the coefficient vectors which are expressed in Eqs (14)-(15):

$\vec{A}=2 \vec{a} \cdot \vec{r}_{(1-a)}$     (14)

$\vec{C}=2 \cdot \vec{r}_2$     (15)

where, the random vectors among 0 and 1 are denoted as $\vec{r}_1$ and $\vec{r}_2$; the vector $a$ is reduced from 2 to 0 according to the iterations. Next, the shrinking encircling, and spiral updating are done in the exploitation phase. Eq. (14) is responsible for the shrinking encircling approach. In spiral location updating, and updating approach is obtained by formulating the spiral expression as shown in Eq. (16) among the optimal solution and remaining solutions:

$\vec{X}(t+1)=\left|\vec{X}_b(t)-\vec{X}(t)\right|$ ecr.$\cos (2 \pi r)+\vec{X}_b(t)$      (16)

where, the distance between the best solution and the remaining solution is denoted as $\left|\vec{X}_b(t)-\vec{X}(t)\right|$ ; $c$ denotes the constant value and the random number generated among [-1,1] is denoted as $r$. The position updating process of WOA depends mainly on the random number generated during the searching process.

Further, the value of $\vec{A}$ decides whether the remaining solutions are going to process under the exploitation or exploration phase. The search agents accomplished the exploration in E-MHOA, when $\vec{A} \geq 1$; otherwise, it chooses the exploitation phase. The exploration of E-MHOA is shown in the Eqs. (17)-(18):

$\vec{D}=\left|\vec{C} \cdot \vec{X}_{\text {rand }}(t)-\vec{X}(t)\right|$     (17)

$\vec{X}(t+1)=\vec{X}_{\text {rand }}(t)-\vec{A} \cdot \vec{D}$     (18)

where, the randomly selected location vector from the current population is denoted as $\vec{X}$. The previously mentioned EMHOA iterative procedure is carried out up to the point where it achieves the maximum number of iterations 100 . The best way to handle E-MHOA is to employ CH, which collects data from regular nodes. Remaining energy, which is taken into account in the E-MHOA fitness function, is employed to prevent failure nodes, improving data delivery. By utilizing the communication cost, the data broadcasting distance is reduced, hence reducing energy consumption and delay. Energy balance between the nodes is accomplished by using the node degree. Moreover, the network's scalability is enhanced by the node coverage.

3.3 Working on E-MHOA based route discovery

Route discovery phase is started and carry out the routing after the CH selection. The control messages of AODV's, including route request (RREQ), hello (HELLO), route error (RERR), route reply (RREP), and are used in this route discovery process using E-MHOA. First, the source CH broadcasts the RREQ to its neighbouring CH nodes. The same fitness function that was employed in the CH section is now used to assess the routing path's fitness function. Subsequently, an ideal relay node, also known as the next-hop CH, transmits the RREP back via the reverse route. When the source sensor receives the RREP from the nearby nodes, the route discovery process is completed. During this route discovery phase, the RERR and HELLO messages are used to finish route maintenance. Once an ideal path has been found in the WSN, the data message packets are sent throughout the network. Agricultural applications can benefit from the dependable data transfer based on E-MHOA technology.