A Hybrid Cross Layer with Harris-Hawk-Optimization-Based Efficient Routing for Wireless Sensor Networks

Abstract

Efficient clustering and routing is a main challenge in a wireless sensor network (WSN). To achieve better quality-of-service (QoS) performance, this work introduces k-medoids with improved artificial-bee-colony (K-IABC)-based energy-efficient clustering and the cross-layer-based Harris-hawks-optimization-algorithm (CL-HHO) routing protocol for WSN. To overcome the power-asymmetry problem in wireless sensor networks, a cross-layer-based optimal-routing solution is proposed. The goal of cross-layer routing algorithms is to decrease network-transmission delay and power consumption. This algorithm which was used to evaluate and select the effective path route and data transfer was implemented using MATLAB, and the results were compared to some existing techniques. The proposed CL-HHO performs well in packet-loss ratio (PLR), throughput, end-to-end delay (E2E), jitter, network lifetime (NLT) and buffer occupancy. These results are then validated by comparing them to traditional routing strategies such as hierarchical energy-efficient data gathering (HEED), energy-efficient-clustering routing protocol (EECRP), Grey wolf optimization (GWO), and cross-layer-based Ant-Lion optimization (CL-ALO). Compared to the HEED, EECRP, GWO, and CL-ALO algorithms, the proposed CL-HHO outperforms them.

1. Introduction

A wireless sensor network (WSN) is a collection of specialized sensor nodes uniformly distributed in space to monitor and record physical environmental factors and organize the collected information in a centralized site. WSN detects and records ecological conditions such as sound, temperature, pollution, waves, wind, etc. The WSN are mostly used in some real-time applications such as agriculture, weather monitoring, PV-plant monitoring and industrial applications. WSN are appropriate for harsh environments where wired networks cannot be installed. They are helpful when manual data collection is difficult or hazardous for sensors and an AI-based node position can be calculated [1,2]. A WSN comprises a base station and many randomly distributed SNs in the area of interest. These BSs aggregate information from many nodes for meaningful analysis in the deployed environment. Routing is critical in WSN, because if a sensor node fails during data transmission, the wireless link becomes uncertain, and routing protocols must meet the SN’s power demands [3,4]. However, some additional challenges of the WSN are limited memory, security, connectivity and localization, and time delay. The security of WSNs has been addressed in several ways. However, a few of these security protocols could be more effective because wireless sensor networks have limited sensor resources. Because the network’s respective sensors need more resources, WSNs are challenging to secure. A dependence on network technology also leaves a wide range of security threats and assaults open. Two contentious limitations on sensor networks will vanish as processing technology develops: restricted processor bandwidth and small memory. In order to make the proper observations and decisions, sensor data must frequently be delivered within strict deadlines. Results have yet to be found so far that satisfy the real-time demands of WSNs.

Various kinds of research work have been carried out to improve the above problem in WSN. But the primary concern of WSN is the power consumption of SN and QoS performance. To resolve the above problem, efficient clustering and routing has been proposed in this work. The primary goal of WSN is to collect data from the environment, but, due to resource constraints, data collected by sensor nodes for BS will necessitate another efficient solution. Most network parameters are additive, since the optimization will also be NP-hard. Local optima exist in metaheuristic algorithms, which can be solved using hybrid techniques related to CH selection [5,6].

The sensor nodes must be powered by batteries, which are frequently difficult to replace. Furthermore, latency, energy usage and network coverage are critical in maintaining QoS performance. As a result, the implementation of efficient QoS solutions must consider the challenging factors of energy conservation within the network. However, significant research must be carried out to address the limitations that exist in this region. The most difficult challenge of a WSN operation is energy conservation. Energy consumption is essential in WSNs because extending a WSN’s lifetime increases its productivity. Nodes are outfitted with limited batteries. Because it reports network performance degradation, network lifetime is an important attribute. When a node dies, the next node also dies. As a result, an urgent problem to be solved is how to improve network lifetime and save node power in WSN.

However, energy conservation is a significant challenge for WSNs because the sustained capacity of WSNs is primarily determined by the lifetime of SNs. Because batteries power SNs, their lifespan is entirely dependent on their power source, which cannot be replaced. Over time, the SN’s ability to detect the relevant area and transmit data to the BS depletes. As a result, the only way to achieve long WSN life is to conserve battery power. Obtaining these facts, this study proposes several algorithms to solve these problems [7,8]. The main purpose of the proposed research is to investigate and propose new ways to increase cluster formation to improve energy consumption and the QoS performance of WSN.

Major Contribution

Thus, we proposed k-medoids with an improved-artificial-bee-colony (IABC)-based clustering approach and hybrid cross-layer-based Harris-hawks-optimization algorithm (CL-HHO) for effective routing in WSN. Because k-medoids with the improved artificial bee colony operate and calculate quickly, we used k-medoids with IABC for cluster formation and CH selection. K-medoids provide high efficiency, while defining the behavior of complex nodes and speeding up the procedure with IABC. Therefore, we chose CH using k-medoids and IABC. We used CL-HHO for the routing process because it has the advantage of HHO’s quick search speed, which provides a quick response and convergence. The primary concerns of the introduced research work are:

• Cluster formation and CH selection are mediated by k-medoids and IABC.
• A hybrid CL-HHO algorithm is presented for choosing the shortest path.
• A novel, well-known technique is used to compare the methodology’s performance.
• The effectiveness of the proposed CL-HHO improves the overall QoS performance.

The remaining article is explained below, and the following section reviews the related research. Section 3 describes the cluster formation, CH selection and routing, in detail. Section 4 describes the proposed work performance using MATLAB. The final section summarizes the research paper’s findings and makes recommendations for future research.

2. Literature Survey

This chapter briefly overviews various works related to WSN, including WSN-based clustering, various optimization techniques, and hybrid optimization in WSN. Local optima, NP-hard problems, and so on can be solved by optimizing CHs using various meta-heuristic optimization algorithms. The need for hybrid optimization arises from improving performance, evolving quality algorithm solutions, and integrating algorithms as part of larger systems. Satyajit Pattnaik et al. (2021) proposed a hybrid moth-flame-cuttlefish-optimization (MFO-CFO) technique for routing. Initially the clusters are formed by ant lion optimization (ALO) based on fuzzy-clustering means (FCM). A particle-swarm-optimization (PSO) algorithm based on an adaptive neuro-fuzzy inference system (ANFIS) is proposed for optimal CH selection. The results are compared to conventional methods such as the fuzzy, PSO and genetic algorithm (GA) [9].

Sina Einavi Pour et al. (2021) presented a CH-selection algorithm that chooses CHs based on the remaining energy, location, and centrality of the node. The simulation outcomes show that the suggested method outperforms existing methods such as LEACH, multi-hop LEACH networks based on reliability, network lifetime and energy consumption [10]. Satyajit Pattnaik et al. (2020) revealed the possibility of combining the fuzzy-clustering approach and the elephant-herding-optimization (EHO)-Greedy algorithm for WSN. This method considers individual sink nodes for fixed and mobile sinks to reduce power consumption. Finally, the EHO-Greedy hybrid-algorithm routing protocol has been proposed for data transmission. The simulation outcomes demonstrate that the suggested technique outperforms other existing energy utilization methods and system lifetimes [11].

Koyuncu et al. (2020) revealed a routing method based on multi-objective energy-efficiency-criterion optimization and clustering. A multitier-deterministic routing protocol is employed to determine the optimal path for data transmission to the base station, thereby improving the WSN network’s performance [12]. Jagadeesh et al. (2021) proposed a multi-objective particle swarm optimization with the Levy-distribution (MOPSO-L)-based dynamic-clustering algorithm. The experimental outcomes show that the suggested model outperforms its competitors regarding network lifetime and energy efficiency [13]. Soleymani et al. (2020) proposed a clustering method with self-diagnosis data fault detection and prediction for WSNs. The proposed approach nearly doubles the lifetime improvement compared with LEACH [14].

Aditya Tandon et al. (2021) proposed a bio-inspired cross-layer routing (BiHCLR) algorithm for efficient routing. A fuzzy logic method is used to choose a CH for each node. The routing paths are then chosen using a hybrid bionic algorithm. The hybrid algorithm combines salp-swarm-optimization and moth-search techniques. The proposed BiHCLR’s result is assessed using a quality-of-service (QoS) analysis [15]. Ramalingam et al. (2021) propose an efficient CORP protocol for grouping and routing in WSNs. In this paper, the k-medoid algorithm with HHO was used for the election of CH, and the CORP algorithm was used for routing. The suggested work’s results were compared to the existing FRLDG and HEED algorithms [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38]. Various communication and IoT routing protocols are also discussed in [39,40,41,42,43,44,45].

Azharuddin et al. (2015) revealed distributed fault-tolerant clustering and routing (DFCR). The algorithm is fault-tolerant and uses little energy. When CH fails unexpectedly, DFCR employs SN’s distributed runtime recovery. SNS that do not have CH within their communication range are responsible for this. Various network scenarios are considered, and extensive experiments are carried out. The algorithm’s robustness has been demonstrated in terms of various performance metrics after comparing empirical results with those of existing algorithms [17]. Amer et al. (2018) presented a balanced-energy-efficient-grid-based clustering protocol (BEEG). LEACH, k-medoids-LEACH (K-LEACH), distributed cluster head scheduling (DCHS), modified-LEACH and energy-LEACH (a clustering protocol based on a balanced network of (BEEG)) have been shown to improve network performance and longevity when compared to other protocols [18]. Bouyer et al. (2015) developed an intra-sensor hybrid cluster protocol based on the distance between the base station, node, and residual energy, using a combination of FCM and LEACH. As a result, the NLT is extended. It has been demonstrated that hybrid protocols are more efficient [19]. Fei et al. (2017) proposed a multi-objective-optimization (MOO) technique and survey of recent R&D efforts. Several popular MOO techniques have been discussed, including heuristic/metaheuristic optimization algorithms and various other advanced optimization techniques [20].

Vimalarani et al. (2016) have proposed an improved PSO-based clustering-energy-optimization (EPSO-CEO) algorithm for WSNs. In this case, the PSO algorithm is used for cluster-head selection and clustering to reduce the amount of power consumed by WSN [21]. Lawal Samira et al. (2015) have suggested a deployment strategy for WSNs which uses the artificial fish swarm algorithm (AFSA). To achieve an adequate QoS in WSNs, the sensor nodes must be deployed as optimally as possible [22]. Banka et al. (2016) suggested the multiple-sink placement algorithm based on PSO, called PSO-MSPA. The algorithm is extensively tested in different WSN scenarios by changing the total number of gateways and SNs; and the algorithm’s efficiency is demonstrated by analyzing the results [23].

Djedouboum et al. (2020) proposed that the optimal path in routing be used to enhance network lifetime and energy efficiency. Although several metaheuristics, especially PSO, have been widely used, they lead to poor-local-optimum problems [24]. Kaur et al. (2016) evaluated the effectiveness of various optimization techniques, including the hybrid GA-PSO approach, bacterial foraging optimization (BFO), GA, and PSO. First, optimization methods such as BFO, PSO, and GA are individually applied to WSN, and then GA and PSO are combined. Existing research integrates load balancing and GA optimization. However, other methods apert from mixing GA and PSO are considered here. This will also help to extend the network’s life [25].

Huadong Wang et al. (2017) described the swarm-intelligence method in a clustering approach with a consistent-clustering approach dependent on the artificial bee colony (ABC). The nodes’ remaining energy, location, and density are all factors that reduce energy utilization. Greater network-topology flexibility is achieved through proper cluster-head selection, and node-power utilization is reduced by extending the network lifetime. The swarm-intelligence approach does not address the load-balancing issue [26]. Chuan zhu et al. (2015) presented a tree-cluster that is based on a data-gathering algorithm (TCBDGA). A weight-dependent tree-building method was developed. As the network ages, TCBDGA maintains load balancing and reduces power consumption. Using TCBDGA lengthens the classification process [27].

Ramnik Singh et al. (2017) proposed a hybrid routing algorithm for heterogeneous networks. Weighted probabilities are used to assign cluster heads to network clusters. They consider combining active and passive networks to achieve well-organized data transfer and increase network lifespan. The energy optimization could have been carried out more efficiently [28]. Amir et al. (2016) created battery-operated energy-harvesting WSN. The overhead issues were not addressed in BPSNs and EHSNs [29].

Ying Song et al. (2015) proposed the genetic algorithm (GA), which provides a low-energy method for multipath routing. For scheduling, data were collected from multiple nodes. As a result, the optimum solution was discovered. Routing stability was not considered for efficient data transfer [30]. Hui Li et al. (2016) proposed the double-level-LEACH (DL-LEACH). The DL-LEACH algorithm predicts the node’s remaining energy, the distance from the BS, and the adjacent nodes, to select the group leader. The DL-LEACH algorithm extends the NLT and solves the problems of other LEACH methods. The use of DL-LEACH does not affect the network lifetime [31]. Difference machine learning and soft computing and the deep-learning algorithm are proposed in [32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56]. An effective, reliable data transmission and prediction-based data transfer has been processed using soft computing and a regression algorithm [57,58,59,60,61,62,63,64,65]. Using network partitioning and an evolutionary protocol to select the best CH, location information, and residual power decomposed by WSN can improve power consumption. CH optimization is a complex non-deterministic polynomial (NP) problem in QoS. Such problems will be solved using polynomial-time general-approximation algorithms. Local optima are a problem for metaheuristic problems, and this study used a hybrid technique to address the local optima for CH selection. Efficient WSN protocols that use cross-layer techniques are required to enhance system stability and performance. However, the aforementioned methods strongly emphasize node clustering, routing, and low-power cluster-head selection. As a result, it is essential to create a CL protocol that can use a novel hybrid algorithm for CH selection, eliminate hotspots by using different clusters and apply a suitable optimization algorithm to perform routing.

Due to density, farm area, long-distance communication, high latency/latency, and packet loss, these approaches are challenging for IoT-driven WSNs. WSN applications require low data-transfer latency, low latency, and energy efficiency. This work encourages us to suggest an IMSF solution to deal with energy asymmetry in WSN applications by considering different layers’ requirements when choosing CHs and forming stable clusters. Energy efficiency and the solution to the energy-asymmetry issue can both be attained by choosing the best CH node.

Table 1. Comparison of existing algorithm and its limitations.

References	Proposed Algorithm	Limitation
[9]	MFO-CFO	Jitter and buffer occupancy are considered for QoS perfromance
[11]	(EHO)-Greedy algorithm	CH-selection time is high
[13]	MOPSO-L	Poor energy consumption
[15]	BiHCLR algorithm	High computational complexity
[18]	Balanced-energy-efficient-grid-based clustering protocol	Poor QoS performance
[21]	EPSO-CEO	Communication cost is high
[25]	Hybrid GA-PSO approach	Energy consumption cannot be minimized sufficiently
[27]	TCBDGA	High computational complexity
[29]	BPSNs and EHSNs	The overhead issues are not addressed

Comparative analysis of existing algorithm and its limitations.

The main contribution proposed is to improve the overall QoS performance of WSNs using an optimization-based hybrid routing and clustering algorithm.

3. Materials and Methods

The k-medoids with improved-artificial-bee-colony (K-IABC)-based energy-efficient clustering and the cross-layer Harris-hawks-optimization-algorithm (CL-HHO) routing protocol for WSN were introduced in this work. A cross-layer-based optimal-routing solution is put forth to address the power-asymmetry issue in wireless sensor networks. Cross-layer routing algorithms aim at reducing power consumption and network-transmission delay.

3.1. System Model

This system model assumes that a particular combination of sensor nodes (SN) is distributed randomly in a rectangular area. This system model is based on sensor-network characteristics:

Figure 1 depicts the architecture of clustering in WSN.

• Each SN is unaware of its position.
• A set of SNs distributed throughout the network has homogeneous characteristics.
• The CH and BS distance is considered non-uniform, because it directly impacts SN’s power consumption.
• Sensor nodes are not observed after deployment, because they are considered naturally organized.

3.2. Network Model

Wireless channels are interchangeable, and both sending and receiving sides are multidirectional. The transmitter’s energy consumption is calculated by Equation (1):

$$E_{tx}\left(j,d\right)=\{\begin{array}{c}{E}_{tx-elec}\times j+{\epsilon }_{fs}\times j\times d^2, d<d_0\\ E_{tx-elec}\times j+{\epsilon }_{amp}\times j\times d^4, d\ge d_0\end{array}$$

(1)

The energy consumed by the transmission circuit when transmitting l bit of data is denoted by $E_{tx-elec}$. The energy amplification parameters ${\epsilon }_{fs}$ and ${\epsilon }_{amp}$ are linked to the chosen transmission-energy- consumption approach. ${\epsilon }_{fs}$ represents free-space transmission, whereas ${\epsilon }_{amp}$ multipath loss represents attenuated transmission.

The boundary condition that distinguishes the two models is $d_0$, $d_0$ = $\sqrt{{\epsilon }_{fs}/}.$ When the transmission distance, d, is more significant than d_0, as transmission distance increases, so does the transmitter’s energy consumption. The energy consumed by the receiving side when receiving l bit data at the receiving end is as follows in Equation (2):

$$E_{rx}\left(j,d\right)=E_{tx-elec}\times j$$

(2)

3.3. K-Medoids Clustering Algorithm

The proposed clustering algorithm is based on k-medoids (KCA) to obtain a general clustering method to improve the overall QoS performance. First, we collect the nodes’ coordinates and remaining energy information, and then compute the group number k. We optimized the k-medoids algorithm by computing the central circle’s midpoint and residual energy to reduce iteration time [32,33].

Set-Up Phase

All nodes are divided into appropriate groups during the setup phase, and CH nodes are identified. Let k be the number of groups defined by Equation (3).

$$\mathrm{k}=\sqrt{\frac{N}{2}}$$

(3)

where N represents the number of nodes. Let L be the node’s central position. Equation (4) can be used to calculate L.

$$\mathrm{L}=\frac{{{\displaystyle \sum }}_{i=1}^N{X}_n}{N}$$

(4)

where ${\mathrm{X}}_{\mathrm{n}}$ is the sensor I coordinate and d represents the average distance between the node’s central position, L, and all SNs. Equation (5) can be used to calculate d.

$$\mathrm{d}=\frac{{{\displaystyle \sum }}_{i=1}^N{X}_n-L|}{N}$$

(5)

A central circle can be computed using L and d. We randomly choose k points on the central circle as the initial mean point l_i, which can be calculated using Equation (6).

$$\{\begin{array}{c}{I}_{ix}=d\times \cos \left(\frac{360}{k}\times \left(i-1\right)\times \frac{\pi }{180}\right)+L_x\\ I_{iy}=d\times \sin \left(\frac{360}{k}\times \left(i-1\right)\times \frac{\pi }{180}\right)+L_x\end{array}$$

(6)

where $I_{ix}$ and $I_{iy}$ denote the mean point coordinates of cluster I. Since the k-medoids algorithm’s initial midpoints are chosen at random, when two or more midpoints are very near, we often obtain incorrect results, resulting in a huge amount of iteration time. Equation (7) accurately represents the absolute-error criterion, E.

$$\mathrm{E}={{\displaystyle \sum }}_{i=1}^k{{\displaystyle \sum }}_{x\in C_i}dist (x,r_i$$

(7)

where x is a common node in cluster $C_i$ and $C_i$ represents cluster i. E is the sum of the distances between all represent nodes and common nodes. The replace method is illustrated in Equation (8).

$$I^{\left(t+1\right)}=\{\begin{array}{c}{I}^{*}, E^{*}-E^{\left(t\right)}<0\\ I^{\left(t\right)}, otherwise\end{array}$$

(8)

Equation (7) and new clusters can be used to calculate the new absolute-error criterion, $E^{*}$. If $E^{*}$ is less than $E^{\left(t\right)},$ $I^{\left(t+1\right)}$ will be replaced by $I^{*}$.

3.4. Artificial-Bee-Colony (ABC) Algorithm

The ABC algorithm is a tolerably improved method that mimics the interlinking behavior of honeybees. Figure 2 depicts the flow chart that corresponds to the presented ABC. The ABC algorithm consists of three strides: onlooker-bee stride, employed-bee stride, and scout-bee stride. The following describes the functionality of each stride: (1) The employed bees’ function is to exploit food resources while disseminating data about food-resource abundance and direction via onlooker bees in a wave movement. (2) The function of the onlooker bee is to select the food-resource probability measure. (3) Scout bees repeatedly examine currently produced food resources randomly assigned to the bee’s region, and, once identified, it becomes an employed bee [34,35,36].

ABC employs three distinct bee entities. While scout bees emphasize the attributes of negative feedback and variance, employed bees and onlooker bees favor positive feedback and numerous interactions. By introducing diversity into the population, fluctuation effectively prevents trapping in less-than-ideal locations. Scout bees carry out this task by bringing solutions from remote, completely random locations while searching. By improving the onlooker-bee component’s global-search efficiency, it is possible to enhance ABC’s results for challenging optimization. Measurements of exploration–exploitation during iterations also reveal that repeated scout-bee introductions cause abrupt changes in the exploration–utilization ratio during searches for the ABC algorithm. Determining promising areas for the entire group requires a smooth transition from exploration to exploitation. A scout bee being introduced unexpectedly during a search is similar to introducing a stranger to a group about to reach a decision.

ABC begins by dispatching bees to a variety of random locations. Bees that discover the best food source among all bees are most likely to follow other bees and converge to the best location, and each bee represents a D-dimensional solution. This accelerates convergence. It is noteworthy that the two components of metaheuristic performance are exploration and exploitation. Exploration is the process of finding new neighbors in the search space, and exploitation is the process of looking harder for already-found potential neighbors. In contrast to power-rich algorithms, which lack the diversity of solutions and frequently fall short of the local optimum, powerful meta-heuristic algorithms may avoid potential solutions by jumping far into the search space. Maintaining trade-offs between exploration and development can be significantly aided by measuring these two properties.

At first, one kind of hired bee will make up half the colony, and another, the spectator bees, will make up the other half. Employed bees scour the area for food sources and inform other bees nearby of new ones. Scouting bees are bees that search at random. The hired bees will switch to a different feed ingredient if it has more nectar than the current one. After exhausting all possible food sources, the bystander bees choose the most likely one and change the location of the food source. The optimum is saved until the enormous loops are reached as the new cost function competes with the old cost function.

3.5. Hybrid K-Medoids with IABC-Based CH-Selection Algorithm

The proposed K-IABC method employs the ABC algorithm with k-medoids to improve exploitation and exploration rates during CH selection. Algorithm 1 shows algorithm for clustering using k-medoid with IABC. The k-medoid-based usage parameter in the suggested K-IABC method avoids the ABC algorithm’s traditional drawback of rapidly falling into the local optimum for CH selection, and is depicted in Figure 3.

Algorithm 1: Algorithm for clustering using k-medoid with IABC

Initialization of network: Step 1: Initialize the sensor network. Step 2: Locate BS at coordinate. Step 3: Place all the SNs randomly. Cluster formation using k-medoid and selection of CH using IABC: Step 4: Number of nodes, N, is divided into number of clusters, C. Step 5: Every cluster has N number of nodes and each node is related to its nearest CH. Step 6: Randomly select the first CH by selecting a random first medoid from N in C cluster. Step 7: A three-dimensional coordinate (x, y, z) is generated by every normal node for CH. Step 8: K-means distance calculation is performed by CH. Step 9: Improved-ABC-optimization algorithm is used to select the new CH. Step 10: Repeat steps 7 to 9 until the node in the real center is found. End

The proposed IABC scheme plays an essential role in optimizing the election of the CH process by leveraging properties derived from intelligent-bee-hunting behavior. To identify the viable CH, the employee, onlooker, and scout-bee stages were integrated into the suggested K-IABC protocol.

We used search-space partitioning to generate the initial population, and used new search equations in the employed-bee and onlooker-bee stages to increase the chances of the onlooker bee finding a better solution, and some worst-case positions. With a new position based on the best non-updated position, Scoutbee has been improved. For initialization, the population is generated based on the search space partitioning (SSD) proposed by He et al. [24]. In the employed-bee stage, we enhance the ABC search formula and gradually use the information of the optimal solution to speed up the search. In the observation-bee phase, 25% of recruited bees were selected with equal probability to perform additional search actions. When many optimal solutions are found to be multidisciplinary, the current optima- solution information and the number of optimal solutions are used to provide a long-distance move, replacing the 10 worst positions with 5% new positions to build the function.

3.6. CORP

The cross-layer based opportunistic-routing-protocol (CORP) algorithm increases the stability of communication networks and reduces delays. Conventional cross-layer routing protocols have some flaws. These are as follows: (i) because WSNs have limited computing power, the computational complexity of data transmission increases with the number of limitations; (ii) they are challenging to integrate, and (iii) they consume a lot of power. The proposed CORP technique addresses several existing cross-layer routing-protocol shortcomings in communication delay, energy consumption, throughput, and packet-loss rate. Initialization, deployment, CH selection, and transmission of data to the receiver are the steps required in this work to coordinate the proposed protocol’s algorithm [16].

Because CH nodes are based on residual-power data, the network layer chooses nodes with higher energy. In routing, the network system uses the physical layer’s remaining power data and the link layer’s link-quality instructions to generate routing options and select the best path. The network layer uses routing to identify the next hop node to the physical layer before forwarding data. To ensure network connectivity, the physical layer calculates the effective transmission energy and transfers this data with a lower transmit power. The link layer employs error-control techniques to confirm data transmission based on the channel status and link quality.

3.7. Harris Hawk’s Optimization Algorithm (HHO)

HHO is a new nature-inspired algorithm. The HHO was inspired by the hunting behavior of Harris’s hawk. These birds perch in the air, spot prey from afar, and then pounce on it in a coordinated attack. The perching behavior of hawks is modeled as the scouting phase in HHO, while their foraging behavior is modeled as the exploiting phase. The mathematical model of the HHO algorithm is presented in this section. In HHO, a candidate solution is referred to as a hawk, (x), while the best solution is referred to as the prey $(x_{prey})$ [37].

A.

EXPLORATION PHASE

Optimization methods require a thorough search of the problem landscape to find the best widely available solution. The metaheuristic algorithm begins the search in the exploration phase to find the best position between hills and valleys in the search space. At this point, an exhaustive search is conducted in the most remote locations. HHO begins by randomly placing N search agents (hawks) at a random location, x_n^0, i = {1, 2, …, N}, across the search space, using Equation (9):

$$x_n^0=l{b}_n+r_1\times \left(u{b}_n-l{b}_n\right),r_1=rand(),$$

(9)

After the population is initialized, the exploration phase continues until the escaping energy of prey |E| ≥ 1, is reached, at which point the value of E is calculated as (10):

$$\mathrm{E}=2E_0\left(1-\frac{t}{N}\right), t=\left\{1,2,\dots ,N\right\},$$

(10)

E₀ and N are the prey’s initial energy and the maximum number of iterations, respectively. This phenomenon is governed by a random variable, q, which is calculated as follows (11)–(13):

$$x_{new}=\{\begin{array}{c}{x}_{rand}-r_2 |x_{rand}-2r_2{x}_n| q\ge 0.5\\ \left(x_{prey}-x_m\right)-r_3\left[l{b}_n+r_4\left(u{b}_n-l{b}_n\right)\right] q<0.5\end{array}$$

(11)

$$r_2=rand(), r_3=rand(), r_4=rand(), r_5=rand(),$$

(12)

$$x_m=\frac{1}{N}{\displaystyle \sum }_{n=1}^N{x}_n$$

(13)

where $x_m, x_{new}$ and $x_{rand}$ are the dimension-wise average of the population, the new position, and a randomly selection position, respectively.

B.

EXPLOITATION PHASE

The convergence of many candidate solutions to previously identified promising locations in the search space is called exploitation. This phase is activated after a few iterations to explore the problem landscape. Once a potential neighborhood has been identified through the collective experience of the search agents, strategies are developed to make candidate solutions gradually adopt information from the best unique global solution. It is important to note, however, that approaching local regions too early can result in premature convergence, which leads to suboptimal solutions. To address this issue, HHO employs various exploitation strategies aimed at various eagle-hunting situations.

A system that chooses precise and effective data-transmission routes using an algorithm inspired by nature and a new probabilistic decision-rule function is proposed. A CL-IoT protocol is suggested (QoS) to balance CH energy usage and latency. By considering the exploitation and exploration phases, the suggested optimization algorithm chooses CHs. The remaining energy, distance, and SN link quality are just a few variables in choosing a CH. We should create energy-consumption graphs for the different clusters in the performance-analysis section.

4. Results and Discussion

4.1. Simulation-Parameter Setup

The MATLAB 2018 platform was used to implement the CL-HHO algorithm. Nodes in a WSN are organized based on system-domain standard transmissions. The simulation parameter setup is shown in

Table 2. Parameter Setup.

PARAMETER	VALUE
Total Clusters	6
Number of nodes	500
Initial energy	0.1 J
Communication-Tx Energy	20.5 mW
Communication-Rx Energy	14 mW
Deployment area	500 $\times$ 500
Packet size	512 bytes
Packet-sending rate	1 packet/s
Node distribution	Random

. The proposed method’s outcomes are compared with existing algorithms, including HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO strategy [38].

4.2. Evaluation of Clustering and Routing

CH receives data from SNs in the surrounding environment. Figure 4 and Figure 5 depict data transmission from the source node to CH.

This routing protocol optimizes the transmission power of different sensor nodes and provides higher reliability for network or other transmissions by utilizing the network’s remaining power. The communication time between the source node and the destination node is reduced as data-transmission performance is improved. The K-IABC and CL-HHO algorithms form clusters and paths more efficiently than existing methods. The traditional method has a high packet-loss rate, transmission delay, and unstable power when sending data. The following section evaluates QoS metrics.

4.3. Performance Analysis

4.3.1. Energy Consumption

The K-IABC must find the best energy-based cluster head during CH selection. As a result, the CL-HHO method is used to evaluate the optimum routing path, and achieves better performance. This proposed CL-HHO is better suited than the existing algorithm.

Table 3. Comparison table for Energy consumption.

Methods	Energy Consumption
No. of nodes	100	200	300	400	500
CL-HHO	0.1	0.25	0.33	0.47	0.55
CL-ALO	0.15	0.29	0.48	0.6	0.72
GWO	0.21	0.42	0.55	0.71	0.88
EECRP	0.61	0.8	0.95	1.2	1.4
HEED	0.75	0.95	1.2	1.4	1.6

compares existing technologies and shows the energy consumption of the proposed working network. Figure 6 compares the energy consumption of the suggested technique to that of conventional methods. The suggested CL-HHO uses less energy, 0.1 mJ for 100 nodes, whereas HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO consume 0.75 mJ, 0.61 mJ, 0.21 mJ, 0.15 mJ, and 0.1 mJ, respectively. As a result, the suggested CL-HHO method is extremely efficient in terms of energy consumption.

4.3.2. Network Lifetime (NLT)

NLT predicts the time between sensor-node locations and a dead network in this process. Figure 7 compares the network life spans of the presented CL-HHO technology and the existing technology.

Table 4. Comparison table for network lifetime.

Methods	NLT
No. of nodes	100	200	300	400	500
CL-HHO	5600	5300	4900	4600	4100
CL-ALO	5100	4800	4600	4100	3800
GWO	4700	4500	4300	3800	3300
EECRP	4400	4100	3700	3400	2900
HEED	4100	3800	3300	3200	2800

compares current technologies and displays the NLT of the proposed working network. In a WSN, data packets must travel longer distances and consume more power. The efficiency of data transmission improves as the network time increases. The CL-HHO has a network life cycle of 5600 (rounds) at 100 nodes, whereas the latest generation of HEED, EECRP, GWO, CL-ALO, and CL-HHO have life cycles of 4100, 4400, 4700, 5100, and 5600 rounds, respectively.

4.3.3. Throughput

The rate at which data is fully transferred across the network is referred to as throughput. Throughput is a measure of the total amount of information that a system can process in a given amount of time.

$$\mathrm{Throughput}=\frac{\mathrm{Amount}\mathrm{of}\mathrm{packet}\mathrm{send}}{\mathrm{time}\mathrm{taken}\mathrm{to}\mathrm{transmit}\mathrm{the}\mathrm{packet}}$$

(14)

Figure 8 and Table 4 compare the suggested CL-HHO method to existing methods in terms of throughput. Table 4 shows that, compared to CL-HHO and other methods, the proposed method achieves high throughput (0.98 Mbps) at 100 nodes. HHOs proposed by HEED, EECRP, GWO, CL-ALO, and CL- represent 0.71 Mbps, 0.75 Mbps, 0.85 Mbps, 0.93 Mbps, and 0.98, respectively. The number of packets received is heavily influenced by the WSN’s lifetime and remaining power. As the number of nodes grows, performance will decrease.

Table 5. Comparison table for Throughput.

Methods	Throughput
No. of nodes	100	200	300	400	500
CL-HHO	0.98	0.92	0.89	0.85	0.79
CL-ALO	0.93	0.85	0.77	0.72	0.69
GWO	0.85	0.78	0.71	0.65	0.60
EECRP	0.75	0.70	0.68	0.60	0.55
HEED	0.71	0.65	0.61	0.54	0.49

demonstrates that the implementation strategy is significantly more efficient than other strategies.

4.3.4. End-To-End Delay (E2E)

Figure 9 and

Table 6. Comparison table for E2E delay.

Methods	E2E Delay
No. of nodes	100	200	300	400	500
CL-HHO	1.8	2.4	3.3	3.8	4.0
CL-ALO	2.9	3.8	4.3	4.7	5.2
GWO	4.2	4.8	5.4	6.3	5.5
EECRP	4.6	5.2	6.1	7.0	8.2
HEED	5.8	6.5	7.0	8.2	9.0

show the E2E delay of the suggested CL-HHO and its comparison with the conventional method. The proposed CL-HHO method achieves lower end-to-end latency (1.8 s) within 100 s node than other conventional methods shown in Figure 9. In 100 nodes, the end-to-end delays of conventional methods such as HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO are 5.8 s, 4.6 s, 4.2 s, 2.9 s, and 1.8 s, respectively.

4.3.5. PDR

Table 7. Comparison table for PDR.

Methods	PDR
No. of Nodes	100	200	300	400	500
CL-HHO	99.4	98.4	97.8	96.5	95.5
CL-ALO	98.5	97.0	96.5	95.4	94.6
GWO	96.8	96.5	95.6	95.0	93.8
EECRP	95.8	95.3	94.8	94.1	93.2
HEED	95.1	94.2	93.2	92.7	91.8

and Figure 10 show the performance of the PDR. Compared to other conventional algorithms, the CL-HHO achieves high PDR (99.4%) energy efficiency at 100 nodes. The proposed method outperforms previous methods. At 100 nodes, the packet-delivery rates of the old methods HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO are 95.1%, 95.8%, 96.8%, 98.5%, and 99.4%, respectively. To transmit data efficiently, PDR must be high. As a result, the suggested model outperforms existing methods in terms of performance and efficiency.

4.3.6. PLR

Table 8. Comparison table for PLR.

Methods	PLR
No. of Nodes	100	200	300	400	500
CL-HHO	0.5	1	2	2.5	3.0
CL-ALO	1	3	4	4.5	5.0
GWO	2.5	4	5	6	7.5
EECRP	3	4.5	6.5	8.0	9
HEED	6	7.5	8.5	10	11

and Figure 11 show the PLR of the CL-HHO and the traditional algorithms. Compared to previous solutions, the proposed CL-HHO achieves a lower PLR (0.5%) among 100 nodes. The PLR rises as the number of nodes increases. Existing algorithms, such as HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO, have packet-loss rates of 6%, 3%, 2.5%, 1%, and 0.5%, respectively.

4.3.7. Jitter

This calculates the time difference between the first packet sent and when it is received. Figure 12 and

Table 9. Comparison table for Jitter.

Methods	Jitter
No. of Nodes	100	200	300	400	500
CL-HHO	0.45	0.41	0.39	0.36	0.35
CL-ALO	0.53	0.48	0.45	0.41	0.39
GWO	0.60	0.58	0.56	0.53	0.50
EECRP	0.68	0.64	0.61	0.56	0.52
HEED	0.73	0.69	0.66	0.61	0.57

depicts the proposed algorithm’s jitter comparison. On 100 nodes, the suggested algorithm has a lower jitter value (0.45) than existing methods such as HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO, which have values of 0.73 ms, 0.68 ms, 0.60 ms, 0.53 ms, and 0.45 ms, respectively. The CL-HHO has a jitter value of 0.41 ms for 200 nodes; however, existing strategies such as HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO have values of 0.69 ms, 0.64 ms, 0.58 ms, 0.48 ms, and 0.41 ms, respectively.

4.3.8. Buffer Occupancy

Figure 13 and

Table 10. Comparison table for buffer occupancy.

Methods	Buffer Occupancy
No. of Nodes	100	200	300	400	500
CL-HHO	4	5	8	10	13
CL-ALO	5	7	10	11	14
GWO	9	11	13	16	17
EECRP	12	14	16	18	19
HEED	14	15	17	20	23

depicts the buffer-occupancy performance of the CL-HHO and conventional methods. In comparison to existing methods, the CL-HHO-method simulation results achieve lower buffer occupancy (4%) at 100 nodes. As a result, the proposed CL-HHO protocol performs better than conventional methods. The buffer-occupancy rates of the proposed HEED, EECRP, GWO, CL-ALO, and CL-HHO methods are 14%, 12%, 9%, 5%, and 4%, respectively.

4.3.9. Communication Cost

Figure 14 and

Table 11. Comparison table for communication cost.

Methods	Communication Cost
No. of nodes	100	200	300	400	500
CL-HHO	0.04	0.05	0.08	0.12	0.14
CL-ALO	0.07	0.10	0.12	0.15	0.19
GWO	0.09	0.12	0.15	0.18	0.22
EECRP	0.10	0.13	0.16	0.18	0.24
HEED	0.12	0.16	0.21	0.23	0.30

depict the communication-cost performance of the proposed (CL-HHO) and existing methods. In comparison to existing methods, the CL-HHO-method simulation results achieve lower communication cost (0.04%) at 100 nodes. The communication cost increases as the number of nodes grow. As a result, the proposed CL-HHO protocol outperforms other existing methods. The communication cost rates of the proposed HEED, EECRP, GWO, CL-ALO, and CL-HHO methods are 0.12%, 0.10%, 0.09%, 0.07%, and 0.044%, respectively.

4.3.10. CH-Selection-Time Analysis

Figure 15 depicts the time performance of the CH-selection method. Compared to existing methods, the proposed CH selection results in a shorter execution time. The time will be extended if the number of CHs increases. Compared to the conventional algorithm, the introduced method achieves a faster execution time (37 s) in the first cluster head.

4.3.11. Energy Consumption vs. Cluster

Figure 16 depicts energy- consumption comparison for different clusters. The energy usages of CL-HHO provide minimum value compared to other conventional algorithms in different clusters. It achieves the minimum energy usage for different CH when compared to other algorithms.

4.3.12. Hop Count

The average number of hops as the network grows is depicted in Figure 17. The quantity of intermediate nodes involved in the packet-relay process affects the number of hops. The results offer a more precise relationship between node count and average hop count. CL HHO exploits the SN that can send packets over great distances. In other words, a node with a more excellent transmission range can travel fewer hops to reach its destination. On the other hand, the number of hops required to relay transmitted packets increases for an existing algorithm. Implementing the HHO method when energy-hole situations arise enhances network restructuring.

5. Conclusions

The k-medoid with IABC and CL-HHO has been proposed for clustering and routing in WSN. The CL-HHO hybrid technique determines the RP with the lowest energy consumption. MATLAB is used to implement all of the work, and the results are compared to some existing techniques. Multiple nodes are used to calculate QoS-performance metrics. When comparing HEED, EECRP, GWO, CL-ALO, and the proposed CL-HHO, the latter works well. Finally, the proposed k-medoids-IABC technique combined with the CL-HHO algorithm will be a better alternative to power-based routing in WSNs. As a result, more iteration may be required to select the most appropriate path. As a result, in the future, a better strategy should be proposed instead of distance estimation to improve the execution speed of the route search. In the future, a new CL routing protocol based on optimization could extend network lifetime and improve WSN performance. Compared to the HEED, EECRP, GWO, and CL-ALO, the introduced CL-HHO, increased network lifetime by 36.58%, 27.27%, 19.14%, and 9.80%, respectively. According to the experimental data, it outperforms HEED, EECRP, GWO, and CL-ALO in terms of QoS performance. Furthermore, adaptive hybrid optimization and deep learning will be proposed to investigate application-specific routing-data schemes to maximize overall routing network performance.