Improved African Buffalo Optimization-Based Energy Efficient Clustering Wireless Sensor Networks using Metaheuristic Routing Technique

Wireless sensor network (WSN) plays a crucial role in the Internet of Things (IoTs), which assist to produce seamless information that have a great impact on the network lifetime. Despite the substantial application of the WSN numerous challenges like energy, load balancing, security, and storage exist. Energy efficacy is regarded as an integral part of the design of WSN; this can be achieved by clustering and multi-hop routing technique using metaheuristic optimization algorithm. This paper concentrates on design of Metaheuristics Cluster-based Routing Technique for Energy-Efficient WSN (MHCRT-EEWSN). The presented MHCRT-EEWSN technique mainly concentrates on the improvements of energy efficiency and lifespan of the WSN via clustering and routing process. For effectual clustering process, the MHCRT-EEWSN model utilizes Whale Moth Flame Optimization technique and can be utilized by the use of fitness function involving intra-cluster distance, inter-cluster distance, energy, and balancing factor. Besides, the MHCRT-EEWSN model employs Improved African Buffalo Optimization (IABO) based routing technique. To select optimal routes in WSN, the IABO algorithm designs a fitness function comprising multiple parameters like residual energy and distance factor. The experimental validation of the MHCRT-EEWSN model can be tested by making use of a series of simulations. A wide-ranging comparative study shows the promising performances of the MHCRT-EEWSN model than other recent methods. The experimental validation of the MHCRT-EEWSN model can be tested by making use of a series of simulations. A wide-ranging comparative study shows the promising performances of the MHCRT-EEWSN model than other recent methods.


Introduction
Wireless sensor networks (WSN) include numerous low power-sensing and low-cost tools. A large quantity of sensor nodes (SNs) self-organized for forming a large-scale network, and SNs could monitor data in the physical atmosphere [1]. Because of the weak communication capabilities and energy of the energy utilization severely limits the development and application of WSNs. SNs contains a confined energy supply and were distributed arbitrarily all over area which the network covers [2]. This non-uniformity of distribution density generates variations in volume of data transferred from every sector. High-density sectors gather highly correlated data while lower one has coverage issues because of premature expiry of isolated nodes [3]. A concern in every WSNs was sensor battery lifespan that relies on energy utilization by data transferring in network. The model of low-energy-utilization routing protocols becomes a major zone of research & development [4]. The utmost benefits in network longevity still attained by advancing clustering routing protocols that permits nodes for communicating directly with adjacent cluster-heads (CHs) instead of forwarding every data to more distant base station (BS). Clustering routing protocols should search a negotiation among outspreading network lifespan and ensuring efficient coverage of area of interest [5].
Clustering methods segregates the network as several independent clusters, where all cluster has a CH node and numerous CM nodes [6]. The CH could avoid the redundant or incorrect data. Therefore, vast amount of collected data was merged into meaningful data and has small quality depends on data aggregating techniques. Therefore, the topology optimization and data traffic are minimized by utilizing efficient clustering techniques that enhances the energy efficiency in WSN [7]. Cluster-based routing protocols or clustering might solve some promising difficulties due to its reliability, energy efficiency, and scalability of data delivery. In WSN, energy saving was taken as an important issue. Routing processes becomes a risky and crucial task in WSNs because of the SNs mobility, energy limitations, security problems, and distance of sensors from sink node [8]. Usually, the nodes in WSNs were placed in long distance from destiny node must route its data packets through multihop because of the coverage and distance issues. Now, rises the communications for reaching the distant nodes and therefore it requires support of others for routing effectively [9]. Additionally, by not having a centralized control, if every sensors were interacting within themselves for reaching a destiny without an appropriate routing method, the number of messages interchanged would rise that result in the existence of congestion in network [10]. Thus, it becomes essential for offering an optimal clustering method that could constitute clusters with unequal number of sensors, rotates the CHs and do CH-based routing for efficient and reliable transmission. This paper focuses on design of metaheuristics cluster related routing technique for energy-efficient WSN (MHCRT-EEWSN). The presented MHCRT-EEWSN technique mainly concentrates on the improvements of energy efficiency and lifespan of the WSN via clustering and routing process. For effectual clustering process, the MHCRT-EEWSN model utilizes Whale Moth Flame optimization (WMFO) technique was utilized by the use of fitness function (FF) involving intra-cluster distance, inter-cluster distance, energy, and balancing factor. Besides, the MHCRT-EEWSN model employs Improved African Buffalo Optimization (IABO) based routing technique. To select optimal routes in WSN, the IABO algorithm designs a fitness function comprising multiple parameters such as residual energy and distance factor. The simulation results of the MHCRT-EEWSN model is examined under distinct measures. The work proposed in this paper helps in achieving maximum energy efficiency and lifespan of the WSN than the currently available routing algorithm.

Literature Review
In [11], the Butterfly Optimization Algorithm (BOA) was employed for opting an optimum CH in a set of nodes. The CH selection (CHS) is optimizing with residual energy (RE) of nodes, distance to neighbors, distance to BS (DBS), node degree (ND), and node centrality. The route amongst the CH and BS are recognized with utilizing ACO, it chooses an optimum route dependent upon distance, RE, and ND. Shafiq et al. [12] examines the Robust Cluster Based Routing Protocol (RCBRP) for recognizing routing paths whereas lesser energy was utilized for enhancing the network lifetime. This mechanism was presented in six phases for exploring flow and communication. Al-Otaibi et al. [13] progresses a hybridization of meta-heuristic cluster-related routing (HMBCR) approach for WSN. The HMBCR system primarily contains a Brain Storm optimization with Levy Distribution (BSO-LD) related clustering procedure employing a fitness function integrating four parameters namely energy, network load, DBS, and distance to neighbors.
In [14], a novel clustering and routing approach was presented. It was mostly depending on Genetic Algorithms (GAs) and Equilibrium Optimization (EO) techniques. For improving scalability, sensor node was clustered from the primary phase utilizing the GA and then optimum CHs were chosen. It follows by the next step, whereas all the nodes drive their gathered information to CH. In detail, the core contribution of this work was reducing energy utilization from the WSN utilizing a GA to cluster and EO technique for choosing the path amongst the CHs and BS. In [15], the CHS and sink mobility-related data transmission, both enhanced with hybrid method which assumes the GA and PSO technique correspondingly for all the tasks. The robust behavior of GA supports from the optimizing the CHS, but the PSO supports in determining an optimizing route for sink mobility.
In [16], a new secure energy aware cluster-based routing technique termed as Trusted Energy Efficient Fuzzy Logic related Clustering algorithm (TEEFCA) was presented. This scheme comprises of two main objectives. Primarily, the trustworthy node was recognized that can perform as candidate nodes to cluster related routing. Secondarily, the fuzzy inference system was utilized in two circumstances like selective of optimum Cluster Leader (CL) and cluster formation procedure with assuming the subsequent three parameters namely Distance Node BS, nodes RE level, and Cluster Density. In [17], it is an analysis of cluster routing protocol from the WSN, and analyses the benefits and drawbacks of LEACH protocols, indicating the problems. It is purposing at the problems present in the original LEACH protocol, the CHS, the different node processing, and inter-cluster routing problems are enhanced correspondingly, after that an improved protocol named as LEACH-Impt is presented in (Table 1).
An optimal position for the placement of relay nodes in WSNs has been proposed Saunhita Sapre et al. [28] using Moth Flame Optimizer (MFO) algorithm, Interior Search Algorithm (ISA), and Bat Algorithm (BA). Owing to breakdown of one or more sensor nodes in the disjoint segments, it leads to failure of the entire network. The algorithm uses heuristic fully connected network to evaluate the connectivity of the network. The connectivity is restored among the disjoint segments by positioning the relay nodes in their correct position and makes the network fully connected. The algorithm identifies the optimal relay nodes to design a fully connected WSN. The proposed method makes the relay nodes to help in improving the lifetime of the network and reduces data latency. Armin Mazinani et al. [29] developed a Fuzzy Multi Cluster-based effective routing algorithm for WSN. The authors used a Constant Threshold model for optimizing the energy to be consumed in the routing process. They also approached the solution to the routing problem through clustering and multi-hop routing. The main advantage of their model is the improvement in energy efficiency.

The Proposed Model
In this study, a new MHCRT-EEWSN algorithm was devised for WSN. At the primary level, the MHCRT-EEWSN model selects CHs using the WMFO algorithm with the fitness function involving intra-cluster distance, inter-cluster distance, energy, and balancing factor. For route selection process, the IABO algorithm designs a FF comprising multiple parameters like residual energy and distance factor. The detailed working of these processes are discussed in the following. Figure 1 showcases the overall process of MHCRT-EEWSN approach.

Algorithmic design of WMFO-based Clustering Technique
The MHCRT-EEWSN model selects CHs using the WMFO algorithm in this study. To maintain balance between exploitation and exploration via linearly reducing number of flames all over the iterations, the MFO algorithm was regarded as an effective method. But MFO integrally suffer from ineffective exploitation capability that leads to premature convergence into local optima or stagnating in far from potential area. At the same time, the experiment result reveals that the WOA benefitted from effectual exploitation capability, while the balance between search mechanisms and its exploration are insufficient to manage complicated real-time challenges, particularly in the Optimal Power Flow (OPF) problems. As a result, the study presents a hybridization of Whale and Moth Flame Optimization (WMFO) to resolve the OPF problems effectively [30]. Movement strategy, population partitioning strategy, greedy selection operator, and a randomized boundary handling were introduced in this algorithm. Figure 2 portrayed the steps involved in MFO technique.
whereby N characterizes the number of population. Here, every subpopulation independently progresses that causes individual to find searching space from different perceptions. Thus, decreases the risk of premature convergence and improper data flow is slowed down within the population.
Movement strategy: The WMFO applies two movement strategies to evolve Pop WOA and Pop MFO sub-populations. The Pop MFO subpopulation is upgraded by adapted MFO movement strategies whereas Pop WOA subpopulation is upgraded using the WOA movement strategies.
WOA movement strategy: The WMFO applies canonical WOA movement strategy to upgrade the position of Pop WOA sub-population, whereby X i (t + 1) signifies the upcoming location of ith searching agent and i ∈ PopvVOA.
Modified MFO movement strategy: The presented WMFO evolve the Pop MFO subpopulation, whereby b indicates the constant value, k refers to a random value lies within [− 1, 1], and F j represents the ith flame so that index j is calculated. j (t) is calculated by Eq. (4), whereby Xbest j denotes the location of self-memory model determined by Definition 2.
Definition 2 (Self-Memory model): Given SM = SM 1 … SM i … SM N as a fixed set of N searching agent memory. The SM i is represented as M i = Xbest i , Fbest i , whereby Xbest i characterizes the optimal location of X i obtained until now, and Fbest i symbolizes the fitness of Xbest i . Initially,Xbest i (t = 1) ← X i (t = 1) and st i (t = 1) ← OX i (t = 1) , For the residual iteration (t > l),X i and Fbest i are upgraded according to optimally attained solution so far by every X i .

Randomized boundary handling:
The canonical WOA and MFO uses a simplest mechanism for limiting the boundary that assign a value equivalent to its respective lb d lower bound if the dth dimension of searching agent was lesser than value of lb d . For avoiding stagnation, a randomly assigned parameter limiting boundary is initiated in the presented algorithm using Eq. (5), whereby x id represents the value of dth dimensions of ith searching agent, and r indicates a random number among zero and one.
Greedy selection operator : WMFO applies the selection operator to assess acceptance conditions of novel solution by comparing the fitness of novel solution OX(t + 1) with the fitness of preceding population OX(t) based on the Eq. (6). (2) Encircling prey determined in Eq. (12) p i < ��A� < 1 Search for prey determined in Eq.17) The WMFO algorithm computes fitness function involving intra-cluster distance, intercluster distance, energy, and balancing factor.
A fitness function can be computed for selecting the CHs. This FF assures that the node with the maximum energy and the node situated adjacent to the BS has a high chance of selecting as CH. The intracluster space can be calculated as a sum of spaces among their respective CH and the SNs. This intracluster distance should be decrease the usage of network energy. It can be supplied due to sensors waste certain energy once interacting with the separate CH: (2) Average Sink Distance (f 2). : The distance between the BS and CH to the overall amount of sensors present in the CH was utilized for computing average sink distance. Space contains a substantial effect on energy usage; therefore, this aspect was taken into account. Therefore, it is necessary to decrease the distance for saving energy.
(3) Residual Energy (f 3): Network lifetime is based on the energy consumption, it is urgency to decrease energy usage. Therefore, this parameter is considered. It can be evaluated as sum of particular channel existing energy. Since overall energy should be optimized, every objective function can be balanced by the opposite.
(4) CH Balancing Factor (f 4).: The cluster should be balanced; it was possible that certain small and large groups would form due to the random grouping of sensors. Consequently, this characteristic is considered while balancing energy consumption. .

Fig. 2 Steps involved in MFO technique
The abovementioned fitness functions are in seamless sync with others and it given in the following: where q , and r represent constant value and p + q + r = 1.

Process involved in IABO Based Routing Technique
Once the CHs are chosen and clusters are organized, the next stage is the route selection process using the IABO algorithm. ABO is a new metaheuristic approach proposed that imitate and exploit the management structure of herd and efficient communication in the migration lifestyle [31]. During the movement, they utilize two sounds or vocalizations "waaa" and "maaa" for exploitation and exploration. The "waaa" sound is utilized for exploring other position since the existing position might have insufficient pasture whereas the "maaa" sound summons the buffalo to stay on for exploiting the existing position because it has enough pasture and is secure. With this sound, they can improve the search to find food and is arithmetically expressed in the following equation.
Here m k refers to a "maaa" sound with a certain reference to a buffalo k k = 1 , 2 , … , n , bg max,k indicates the best position of buffalo in the herd, bg max.k denotes optimal position discovered by an buffalo k, lp1 and lp2 refers to learning parameter ∈ 0 , 1 . m k+1 denotes relocation of buffalo from present position m k to a novel position that reflects the wide memory capability in migration lifestyle and it is given as follows.
In Eq. (13), w k+1 signifies the movement to a novel position, w k represent the existing exploration value that characterize "waaa" sound whereas m k denotes the existing exploitation value and 1 is a variable that determines the time interval over movement of buffalo and is generally fixed to 1.
The algorithm below defines the ABO approach by insertion of random of kth buffalo. The optimum solution attained depends on altering the movement of buffalo. In every iteration, the fitness value attained and the optimum one amongst each individual is allotted to bg max (best global one) whereas the best for all was allocated to bg max.k (best local one). Every buffalo upgrades the position and move according to of best adjacent buffalo and it allows the movement of buffalo towards the optimal solution and tracking it. (11) Fitness function = (p × f 1) + (q × f 2) + (f 3 × r) The IABO algorithm was derived by the use of Oppositional Based Learning (OBL) concept to improve convergence rate and global optima. In IABO algorithm, the opposite position of solution is produced as per the concept of opposite number [32]. For describing the novel population initialization, it is essential for defining the concepts of opposite number. Assume an N-dimension vector X: To employ the concept of opposite number in the initial population, assume x i as a randomly produced solution in N-dimension problem space. For this random solution, the opposite is produced and represented as x i . Next, x j andx i solution is estimated by the objective function f . Consequently, if f x i is greater than f x i (i.e., f x i < f x i ) , the agent x i is replaced with x i ; or else, we continue with x i . Therefore, in the initial iteration, the first solution and the opposite are assessed concurrently to continue with good (fitter) starting agent.
The IABO algorithm designs a FF comprising multiple parameters like residual energy and distance factor. The objective function is given as follows.
The procedure aim was determining the best group of routes in CH to the BS through an FF that encompasses two parameters i.e., distance and energy. At first, the RE of next-hop node was defined, and the node having maximal energy acts as the RN. Thus, the node having greater RE become the next-hop node as follows: Moreover, the Euclidean distance can be implemented to define distance in a CH to BS . With the minimal distance, the energy usage is kept significantly lower. Once the distance is increased, addition energy is expanded. Therefore, the node with minimum distance is preferred as optimum route.
The above mentioned sub-objective is considered as FF , where 1 and 2 denotes the weight allocated to every sub-objective.

Results and Discussion
In this section, a wide ranging comparison study of the model is carried out under varying numbers of sensor nodes. Table 2 provides detailed results of the MHCRT-EEWSN model with existing models interms of End-to-End Delay (ETED), Packet Delivery Ratio (PDR), and Packet Loss Rate (PLR) [32][33][34][35][36]. Figure 3 inspects a comprehensive ETED examination of the proposed MHCRT-EEWSN model and existing techniques under diverse nodes. The Fig. 3, represents that the MHCRT-EEWSN model has shown enhanced performance with least values of ETED. For instance, with 100 SNs, the MHCRT-EEWSN model has attained effectual outcomes with minimal ETED of 1.36 ms whereas the LEACH, HEED, MBC, FRLDG, and F-GWO models have obtained maximum ETED of 6.15 ms, 5.18 ms, 4.09 ms, 3.39 ms, and 2.14 ms respectively. Moreover, with 500 SNs, the MHCRT-EEWSN model has obtained improved results with lower ETED of 6.43 ms whereas the LEACH, HEED, MBC, FRLDG, and F-GWO models have gained higher ETED of 9.86 ms, 9.31 ms, 8.96 ms, 8.69 ms, and 7.87 ms respectively.
A detailed PDR investigation of the proposed MHCRT-EEWSN with recent approaches is provided in Fig. 4 Table 3 provides detailed results of the proposed MHCRT-EEWSN model with existing models interms of Throughput (THROP), Energy Consumption (ECOM), and Network Lifetime (NLFT). A detailed THROP analysis of the MHCRT-EEWSN with recent approaches is offered in Fig. 6. The experimental values inferred that the MHCRT-EEWSN model has exhibited better results with maximum THROP values under each SN. For example, with 100 SNs, the MHCRT-EEWSN approach has exhibited higher THROP of 0.9969 bps whereas the LEACH, HEED, MBC, FRLDG, and F-GWO models have demonstrated lower THROP of 0.6538 bps, 0.7403 bps, 0.7877 bps, 0.8937 bps, and 0.9635 bps correspondingly. Besides, with 500 SNs, the MHCRT-EEWSN method has exhibited better performance with increased THROP of 0.7207 bps whereas the LEACH, HEED, MBC, FRLDG, and F-GWO models have attained reduced THROP of 0.3664 bps, 0.4111 bps, 0.4473 bps, 0.5450 bps, and 0.5980bps correspondingly. Figure 7 reports a comprehensive ECOM check of the proposed MHCRT-EEWSN model and existing techniques under diverse nodes. The Fig. 7  A detailed NLFT scrutiny of the proposed MHCRT-EEWSN with recent approaches is rendered in Fig. 8. The experimental values implicit the MHCRT-EEWSN model has displayed better results with maximum NLFT values under each SN. For instance, with 100 SNs, the MHCRT-EEWSN approach has exhibited higher NLFT of 5442 rounds whereas   Table 4 and Fig. 9 inspects a comprehensive Bit Error Rate (BER) examination of the proposed MHCRT-EEWSN model and existing techniques under diverse nodes. The figure  By looking into the above mentioned tables and discussion, it is assured that the presented model has gained increased energy efficiency and lifespan in WSN.

Conclusion
In this study, a new MHCRT-EEWSN algorithm was devised for WSN. At the primary level, the MHCRT-EEWSN model selects CHs using the WMFO algorithm with the fitness function involving intra-cluster distance, inter-cluster distance, energy, and balancing factor. For route selection process, the IABO algorithm designs a fitness function comprising multiple parameters such as residual energy and distance factor. The simulation results of the MHCRT-EEWSN model are examined under distinct measures. A wide-ranging comparative study shows the promising performances of the MHCRT-EEWSN model than other recent methods. Thus, the MHCRT-EEWSN model can be appeared as an effectual tool for achieving maximum energy efficiency and lifespan of the WSN. In future, energy efficiency of the presented model can be boosted by the data aggregation and data compression approaches at the CHs of the network.