HHO-LPWSN: Harris Hawks Optimization Algorithm for Sensor Nodes Localization Problem in Wireless Sensor Networks

Wireless sensor network (WSN) is a prominent technology for remote area monitoring with the assimilation of the Internet of Things (IoT). Over the past decades, sensor node localization has become an essential challenge of WSNs. The sensor indicates localization challenges related to NP-hard problems. Nature-inspired computational intelligence algorithms are used to solve NP-hard problems efficiently for sensor node localization. After the rigorous advanced search in reputable research journals, efficient newly designed Harris Hawks Optimization (HHO) algorithm has not been used to sensor nodes localization until now. Therefore, this paper does and compares the proposed work from the recently available well-known optimization algorithms such as the Salp Swarm Algorithm (SSA), Equilibrium Optimizer (EO), and Grey Wolf Optimizer (GWO). The simulation results of the proposed work showed that it can outperform in terms of average localization error, the number of localized sensor nodes, and computational cost compared to other computational intelligence algorithms.


Introduction
Today is the era of technological automation [1], where systems are designed with the help of global networks (Internet) in such a way that human intervention would be minimized. Researchers worked with the IoT system to meet all the requirements of technical automation [2][3][4][5]. These types of systems consume a lot of data [6] to solve real-time challenges. A large amount of realistic data can be collected using only WSNs. Researchers are more concerned about the design of WSN-IoT system integration [7]. Real-time data are collected by sensor nodes under the umbrella of a WSN. The collected data of the sensor nodes have no meaning until the WSN knows its actual state. Thus, the localization of sensor nodes becomes an important challenge for WSNs [8][9][10].
The localization algorithm is classified into two parts, such as range-based and range-free-based localization approaches. Range-based localization approaches [11] are designs based on distance or angle calculation between nodes and while range-free-based localization approaches [12] use hop count between sensor nodes to estimate the coordination of sensor nodes. The range-based localization approach is the Received Signal Strength Indicator (RSSI) [13], Time of Arrival (ToA) [14], Angle of Arrival (AoA) [15] and Time Difference of Arrival (TDoA) [16]. Rangefree-based localization approaches are Distance Vector-Hop (DV-Hop) [17], Ad-Hoc Positioning System (APS) [18], and Multi-Dimensional Scaling (MDS) [19]. In range-based localization approaches, anchor nodes information is required to estimate the coordination of sensor nodes. 2 Anchor nodes are nodes whose coordinate information is known in the system. For sensor node localization the needs at least three anchor nodes. [20]. The cost of anchor nodes in the system is higher than the deployment of sensor nodes due to the additional cost of the Global Positioning System (GPS) equipped with anchor nodes [21]. The hardware cost of localization can be solved efficiently using computational intelligence algorithms for sensor nodes. Computational intelligence algorithms are usually designed based on the working principle of nature-induced behavior of humansbeings. Artificial intelligence incorporated into localization modules using computational intelligence algorithms [22]. Thus, there is a need to estimate the location of sensor nodes optimally using computational intelligence algorithms.
Numerous computational intelligence algorithms are available to find the optimal solution for sensor node localization problems but still, there is a need to achieve the fast convergence speed optimization algorithm for sensor node localization by optimally balancing the total number sensor node localization and mean error rate. The newly designed computational intelligence algorithm [23] in which the author claimed that the HHO algorithm outperforms in terms of statistical results compared to other well-known optimizers. Thus, the rigorous advanced search in reputable research journals found that this efficient newly designed HHO algorithm has not been used for sensor node localization until now. Therefore the main contributions of this paper are: 1. A newly designed HHO computational intelligence algorithm by Heidari et al. [23] is used to solve the localization problem of sensor nodes in WSN. 2. The proposed work implementation using the MatLab tool is presented. 3. The design work of this paper is compared with other computational intelligence algorithms such as SSA, GWO, and EO. 4. Performance analysis parameters for the suggested work in terms of mean localization error, computational cost, and the number of localized sensor nodes.
This paper is structured as follows: section two presents computational intelligence algorithms, section three presents the literature survey of esteemed existing works in the field of anchors-based sensor nodes localization, section four provides the proposed approaches model, flowchart and algorithm, section five provides the proposed work evaluation between them in terms of mean localization error, computation cost, the number of localized nodes and section six presents the conclusion of designed paper works.

Computational Intelligence Algorithms
Computational intelligence algorithms are nature-inspired algorithms; nowadays, popularly used in interdisciplinary to achieve optimal results. In this section, various well-known computational intelligence algorithms like SSA, GWO, EO, and HHO are presented in detail below:

Salp Swarm Algorithm (SSA)
Mirjalili et al. [24] proposed the SSA algorithm, which mimics the social interaction behavior of salp swarms.SSA is a population-based Swarm Intelligence (SI) algorithm. The slap has a transparent barrel-like body, and its tissues are like a jellyfish structure. They live underneath the sea and search for their food by the salp chains. The salps chain is divided into two categories as leaders and followers. Leader saps update their location according to equation 1: Where 1 is represented as the leader's location in the j th dimension, and represented as the upper bound and lower bound of the j th dimension of the target region, denotes target localization of food, and 1 , 2 , 3 are random variables.
The follower slaps his place according to equation 2. Mirjalili et al. [25] proposed the GWO algorithm from leadership qualities inspired by Grey Wolves. It is a swarm computational intelligence algorithm similar to PSO, Ant Colony Optimization (ABC) algorithm. This mimics the lead pecking order and the relationship of wolves. The social pecking order is simulated by classifying the population of search agents based on their fitness: • Level 1 (Alpha): This is the leader who is male or female. Alpha is mostly responsible for decision-making (such as hunting, sleeping places, etc.  GWO Search Process: The model demonstrated mimic hunting behavior of grey wolves to use three stages, searching, circling, and attacking prey. The first two stages are given to the exploration process, and the last one presents the exploitation process. • Searching (Exploration): Grey wolves typically detect the search process according to alpha, beta, and delta positions. They distributed themselves from one another to exploit to locate prey and attack prey. The GWO algorithm uses the A constraint, in which A is a random value, and its value is greater than 1 or less than -1. The search agents may diverge from the prey when |A| > 1, and they force to diverge for finding a better one. • Encircling (Exploration): Grey wolves encircling the prey before hunting. The encircling behavior calculated by using mathematical equations (3) and (4) are as follows: Where t represents the current iteration, ⃗ and ⃗ are coefficient vectors, ����⃗ is the prey position vector, ⃗ presents the Grey Wolves position vector and ⃗ ( + 1) is the next position vector of Grey Wolves.
• Attacking Prey (Exploitation): Grey wolves end the hunt when the prey stops moving. In the GWO algorithm, when |A| < 1, then the wolves attack the prey.

Equilibrium Optimizer (EO)
Faramarzi et al. [26] proposed an optimization technique to induce control volume mass prototyping. In EO, each particle denotes solution and concentration as position. The concentration acts as a search agent in EO and is updated according to the best-so-far solution. The best solution obtained is known as the final equilibrium state. The EO algorithm is modeled in equation 5 by updating the rules.
⃗ is represented as a concentration vector, ⃗ is presented as an equilibrium candidates vector, ⃗ is represented as the exponential term vector for the concentration update rule, λ �⃗ denotes a random vector between [0, 1], ⃗ is represented as a generation rate vector, is represented as the control volume of .

Harris Hawks Optimization (HHO)
Heidari et al. [23] proposed a nature-inspired computational intelligence algorithm adopting the harris hawk's behavioral style of prey pursuit. The several hawks cooperatively pounce to surprise the prey. Harris Hawks has a unique cooperative pursuit strategy based on conditions of dynamic nature and escape strategies of prey. The hawks show innovative team spirit to chase strength in terms of hunting, encircling, and getting out of the hunt. The exploration and exploitation steps of the HHO algorithm are as follows: • Exploration Phase: In exploration, the harris hawks use their powerful eyes to locate prey. Harris Hawks is randomly perched in several locations, and they explore the possibility of hunting on two occasions based on q value. If q> 0.5, they are close enough to attack prey, and they sit on the random tallest tree, which is modeled in the equation. In the exploitation phase, there is a chance to attack an already identified prey.

Literature Survey
This section provides a critical analysis of the latest research works available in the field of anchor-based localization in WSNs using computational algorithms. The salp swarm optimization algorithm is proposed by Kanoosh et al. [27] for localizing sensor nodes in WSNs. In WSN, the location accuracy of sensor node localization is greatly affected by the salp swarm algorithm compared to particle swarm optimization, butterfly optimization algorithm, firefly algorithm, grey wolf optimizer. The simulation result shows that the performance of the proposed algorithm is much better than other localization algorithms in terms of the number of localized nodes, localization error, and computing cost.
Rajkumar et al. [28] proposed work by incorporating the Grey Wolf Optimization (GWO) algorithm to detect the accurate geographic location of unknown sensor nodes with the help of anchor nodes in WSNs. The GWO algorithm mimics the social behavior of a grey wolf leadership to attack targets. The suggested work is implemented using the MatLab tool for randomly deployed sensor nodes in the target region. Parameters such as computation cost, localized node percentage, the minimum number of error measures for analysis of GWO's ability, and other types of metaheuristic algorithms. The result of faster convergence and the success rate of the GWO algorithm is better than other PSO and other metaheuristics algorithms like the Modified BAT Algorithm (MBA). Tuba et al. [29] proposed two-stage sensor node localization using a firefly algorithm. In the WSN, the RSSI (Received Signal Strength Signal) propagation model is used to estimate the distance between the anchor nodes and the semi anchor nodes. The proposed algorithm for the localization of the sensor node follows a two-part: first, four anchor nodes are placed at the corners of the target area coverage, and secondly the estimation of the optimal distance using distance calculation. The future direction of this work for an optimal approach for the localization of sensors with firefly algorithm modification and adjustment.
Monarch Butterfly Optimization (MBO) algorithm used by Strumberger et al. [30] to solve the NP-hard problem of WSN localization. The novel Monarch Butterfly SI approach uses multi-phase localization for sensor nodes. MBO is implemented and tested on several problem examples that are found in the literature.  Experimental result analysis of the proposed work from other approaches has been successfully presented and has shown considerable potential in terms of solving the NPhard problem of WSN localization.
A location-aware Mobile Anchor (MA) uses path planning to optimize mobile nodes. The work of MA to traverse into the target region of interest to minimize localization error and maximize localization of the successful node. Alomari et al. [31] proposed two novel dynamic movement approaches that provide the obstacle avoidance path planning for mobile node localization in WSN. Movement planning of mobile nodes designed based on two SI-based algorithms, i.e., GWO and Whale Optimization Algorithm (WOA). Comparing this proposed approach to the snake-like and z-curve models, it has shown remarkable results in terms of localization ratio, localization accuracy, and computation time.
An Elephant Herring Optimization (EHO) algorithm is adopted by Strumberger et al. [32] to solve localization problems in WSN. New metaheuristic computational intelligence approach dealing with NP-hard problems to achieve a near-to-target coordination value. The purpose of this approach is for the localization of randomly deployed sensor nodes in the monitoring area. The implementation of EHO for node localization in a WSN and results in efficient metaheuristic approaches to deal with sensor nodes localization. The work presents a future direction of the EHO algorithm that can apply efficient solutions to the superset problem of node localization, i.e., the coverage problem in WSNs.
An improved version of metaheuristic algorithms, such as the tree development algorithm and the EHO algorithm, is proposed by Strumberger et al. [33] to solve the localization problem of WSNs. The improvement of the proposed algorithm is analyzed by varying the size of the sensor network from 25 to 150 target nodes. The state of the art of some SI algorithms is tested in comparison to the proposed algorithm. Simulation results indicate that the proposed algorithm achieves very efficient results in terms of accurate location estimation of the coordinate of the unknown sensor node.
A Distance Mapping Algorithm (DMA) is proposed by Tan et al. [34] to overcome the node localization problem in WSN. To detect node position with high accuracy using the Figure 2. Anchor-based sensor nodes localization approach using harris hawks optimization algorithm WSNs estimation matrix, distance matrix, and optimized linear transformation function.GA is employed for the optimal detection coordinate value of nodes during the calculation of the proposed algorithm. The node localization approach was simulated using three anchor nodes by the researcher in the laboratory. The results of the proposed algorithm perform well in terms of localization accuracy and energy consumption other than the localization algorithm. Current important works of literature in the field of anchor-based localization WSNs are based on various parameters such as authors' publication, design approach, the technique used, comparison approaches, target parameters, and simulation tools using computational intelligence algorithms, as shown in Table 1.

Proposed Model Formulation
The proposed work presented for sensor nodes location estimation challenges using an anchor-based localization approach with the computational intelligence algorithms. The localization proposed model formulation is further classified into a subsection of the proposed model, proposed flow chart, and proposed algorithm.

Proposed Model
The proposed model designed with the components of anchors node ((x1, y1), (x2, y2), (x3, y3)), sensor node (x4, y4), HHO is used as a computational intelligence algorithm and measuring techniques (RSSI) as the inputs for the positioning estimation of unknown sensor nodes. The traditional optimization-based localization model using GWO, SSA, and EO is depicted in Figure 1. The newly smart localization model for anchor-based localization using the HHO algorithm, as shown in Figure 2.

Proposed Flow Chart
The working principles of the proposed work are depicted in the form of the flow chart in Figure 3, which shows the flow control of a designed framework for sensor nodes localization in an anchors-based approach using HHO computational intelligence algorithms. The computational intelligence algorithms are used SSA, GWO, EO, and HHO algorithms to finding optimal localization.

Proposed Algorithm
The proposed work is designed for anchor-based localization using HHO computational intelligence algorithms. The algorithm for anchor-based localization of sensor nodes using the HHO algorithm is presented below: Inputs: Targetarea is a given target area where sensor nodes are to deploy randomly, l is a length and b is a breath of the target area, AN (x, y) in anchor nodes coordinate, centroid (a, b, c, d) is a function to calculate the centroid of the given area and a, b, c, d are the sides of the given target area, SN (x, y) is a current location of sensor nodes, SNtotal is a total number of sensor nodes, dim is represent the dimensional of the target area, i is denoted the index of sensor nodes, SNref calculates the total number of anchor nodes are in their range, disti is estimating the distance between sensor nodes and anchor nodes, the position is to save the best location of optimization algorithm in each iteration, Maxiter represents the maximum of iteration to position refinement, SearchAgent is agents are required to finding an optimal position, lb is a lower bound and ub is an upper bound of the given target area.

9.
Distance between anchor nodes and sensor node is calculated using the below equation:

11.
Estimate the coordinate value of SN (x, y,z) using below equations: 12.

Simulation Results and Analysis
Performance analysis of the proposed HHO algorithm along with comparative analysis of SSA, GWO, and EO algorithms in an anchor-based localization approach. The performance is analyzed with the help MatLab tool   on the PC with an intel core i7 processor, 3.40 GHz CPU, and 4 GB RAM. This section is divided into two parts such as the simulation scenario and performance evaluation criteria.

Simulation Scenario
In the simulation configuration, the transmission range of anchor and sensor nodes is fixed at 20 m. The random deployment of sensor nodes in the target area of 50 x 50 m 2 . Each simulation setup of up to 100 has randomly deployed anchor nodes in the target region with a variation of 10, and a free space path loss & fading model are considered. The RSSI measurement technique is used to distance estimation between sensor nodes and the anchor node in a range-based localization approach. The optimization algorithms are taken by SSA, GWO, EO, and HHO to the simulation of a single localization approach. In optimization algorithms, the search agents are ten and  the maximum iteration is set 10 times for estimated position refinement.

Performance Evaluation Criteria
The performance evaluation criteria for the anchors-based localization approach using the HHO algorithm are mean localization error, computation cost, and the number of sensors localized with the variation of the number of randomly deployed sensor nodes. The number of randomly deployed anchor nodes by varying from 10 to 100, with a difference of 10 in each simulation. The anchor-based localization approach using HHO, SSA, GWO, and EO algorithms shown in Figure 4, Figure 5, Figure 6, and Figure 7 for randomly deployed of 200 sensor nodes.  • Mean Localization Error: The average difference between actual sensor nodes and estimated sensor nodes coordinate values. The mean localization error for sensor node localization for each randomly deployed anchor node from the variation 10 to 100 with a difference of 10 is shown in Table 2 and Figure 8. The resultant graph shows that the HHO algorithm is much better than the SSA, GWO, and EO algorithms for the anchors-based localization approach. • Computational Cost: The total time is required to complete the localization process for randomly deployed sensor nodes is known as computation cost, and it is generally measured in terms of seconds (sec) unit. The computational cost of anchor-based localization using the HHO algorithm approximates better compared to SSA, GWO, and EO algorithms. By variation of 10 to 200 anchor nodes deployment with a difference of 10, the computation cost is calculated as shown in Table 3 and Figure 9.   The number of localized sensor nodes over the number of randomly deployed anchor nodes by the variation of 10 to 100 sensor nodes with a difference of 10. The number of localized sensor nodes in an anchor-based localization approach using the HHO algorithm performs better than SSA, GWO, and the EO algorithm is shown in Table 4 and Figure 10.

Conclusion
The sensor node's localization became a crucial challenge for WSN. The technology advancement leads to WSN-IoT integration in order to reduce human intervention.
Reduce the extra cost of GPS components is also minimized using an anchor-based localization approach. The optimal coordinate value calculation of the sensor nodes is done using the newly designed HHO algorithm in this paper. The simulated results and analysis of the HHO algorithm are compared with the SSA, GWO, and EO algorithms in an anchors-based localization approach. The percentage improvement of the HHO algorithm for localization problems over the SSA, GWO, and EO in terms of mean localization error, computational cost, and the number of localization nodes is presented in Table 5.  Table 5 shows the HHO algorithm's overall performance analysis parameters for the efficiently estimated location of sensor nodes in WSN compared to other computational algorithms. The future direction of this proposed work can be implemented for the three-dimensional target area.