Energy Aware Seagull Optimization-Based Unequal Clustering Technique in WSN Communication

Wireless sensor network (WSN) becomes a hot research area owing to an extensive set of applications. In order to accomplish energy efficiency in WSN, most of the earlier works have focused on the clustering process which enables to elect CHs and organize unequal clusters. However, the clustering process results in hot spot problem and can be addressed by the use of unequal clustering techniques, which enables to construct of clusters of unequal sizes to equalize the energy dissipation in the WSN. Unequal clustering can be formulated as an NP-hard issue and can be solved by metaheuristic optimization algorithms. With this motivation, this paper presents a novel seagull optimization (SGO) based unequal clustering (SGOBUC) model to attain energy efficiency in WSN. The SGO algorithm is mainly inspired by the migrating and attacking behaviour of seagulls. They are formulated in a mathematical way and designed to highlight exploration as well as exploitation in a provided searching area. The SGOBUC technique derives a fitness involving different parameters in such a way that energy efficiency can be accomplished. A comprehensive simulation analysis takes place to showcase the enhanced outcomes of the SGOBUC technique. The simulation outcomes highlighted the betterment of the SGOBUC technique over the recent techniques interms of different dimensions.


Introduction
CHs far from the BS. It implies CHs near to the BS are burdened with more traffic load owing to the intracluster traffic from its data aggregation, CMs, and intercluster traffic from other CHs for transmitting data to the BS [8]. It interrupts the network connection, and the clusters near the BS create connectivity problems that are named hot spot problems [9]. The unequal clustering method is the effective method for handling hot spot issues since it could be utilized for load balancing amongst the CHs [10].
Unequal clustering organizes the clusters based on the sizes where the sizes gets decreased nearer to the BS. It implies that the cluster sizes are equivalent to the intercluster distance. A small cluster nearby the BS represents a small amount of CMs and lesser intracluster traffic [11]. Thus, the small clusters could pay considerable attention to intercluster traffic, and the CH doesn't exhaust energy more rapidly. When the DBS is high, then, the cluster size also gets increased. When the group holds an essential amount of CMs, they would expend high energy on intracluster traffic. Since the cluster becomes far away from the BS, they hold lesser intercluster traffic and don't have to spend extra energy on intercluster routing [12]. Unequal clustering enforces each CH to expend similar quantity of energy utilization, and hence the CHs nearby the BS and the CHs far away from the BS spend equivalent amounts of energy. Also, the cluster construction procedure might create 2 level hierarchies with lower and higher levels [13,14]. The CHs create the high-level cluster, and the CMs creates the low-level cluster. The sensors regularly transfer the data to the respective CHs. The CH carries out data aggregation of each data attained from the CM and forwards the aggregated data to the BS through single/multi-hops. The CH has the data of its CMs like energy level, node id, and position. If a node moves/dies to other clusters, the modifications are instantly recorded and the CH notifies the BS, and re-clustering takes place to preserve the network topology efficiently. The amount of transmission and the overall network load gets decreased considerably.
To resolve the hot spot problem in WSN, this paper designs a new seagull optimization (SGO) based on unequal clustering (SGOBUC) technique to accomplish energy efficiency and maximum lifetime. The SGOBUC technique aims to construct clusters of unequal sizes and elect proper set of CHs. In addition, the SGOBUC technique involves three phases namely initialization phase, neighboring data collection phase, and finally, cluster construction phase. The SGOBUC technique constructs clusters based on a fitness function involving different input parameters. The application of unequal clustering technique assists to uniformly balance the energy dissipation throughout the network. An Wireless Sensor Networks (WSN) that self-configures, making use of less infrastructure, and is deployed via wireless sensor networks is referred to as a WSN. When you utilize a sink or base station, you are establishing an interface with the network. Injecting queries and receiving results from the sink allow one to access necessary information from the network. Sensor networks can range from hundreds of thousands of sensor nodes to millions of sensor nodes. They can communicate via radio signals, which enables them to monitor each other. An Internet of Things sensor node has a radio transceiver, a power supply, and sensing and processing components.

Literature Survey
Guleria et al. [15] proposed a new unequal clustering technique using ant colony optimization algorithm. The fusion of data from the CHs to the intermediary node is known as Rendezvous node which decreases the packet broadcasts and therefore the energy spent by the nodes is minimal. The neighbour discovery stage and the link maintenance via ACO method chooses optimum paths among the nodes rises the packet delivered to the end. The estimation of the optimum paths and the CH election utilizing ACO algorithm and unequal clustering decreased the energy utilization efficiency. Arikumar et al. [16] proposed an EELTM method that utilizes smart approaches like PSO and FIS. Moreover, they proposed an optimum CH-CR selection technique in this method utilizes the fitness values evaluated by the PSO method to define 2 optimum nodes in all the clusters for performing CH and CR. The chosen CH completely collects the data from its CM, while the CR is responsible for getting the collected data from the CH and transmitting to the BS. Moussa et al. [17] proposed an ECRP-UCA method, which splits the network into unequal clusters on the basis of distance to BS, neighboring node count, RE, and a novel variable called amount of backward relay nodes in an earlier rounds to appropriately manage the load between CHs. Also, they proposed a batch-based clustering technique which permits the WSN to operate various iterations with no control overhead to configure it. Rao et al. [18], proposed CSO based method, also known as CSO-UCRA. Firstly, the CH selection method was introduced that depends on CSO based method, then allocates the non-CH sensor nodes to CHs on the basis of elected CH proficiency function. Lastly, a CSO-based routing method was introduced. Effective particle encoding systems and new FF were established for this method.
Pan et al. [19] proposed a new optimization method such as cBA, for using the class of optimization issues comprising devices that have constrained hardware resources. A real-valued prototype vector is utilized in the probabilistic operation to generate candidate solutions of cBA optimization. Anand et al. [20] proposed a method to improve lifespan of networks with the help of PSO based clustering and Harmony Search (HS) based routing technique in WSN. Here, the global optimum CH are chosen and gateway nodes are presented for reducing energy utilization of the CH when transmitting accumulated data to the BS. Then, the HS based local searching approach detects an optimal routes for gateway node to the BS. In Sefati et al. [21], the optimized black hole method is applied for selecting an optimum CH from the sensors. The choice of CHs can be optimized using distance, free buffer of nodes, and remaining energy level. The route from the CHs to BS are recognized with the help of ACO method.
Zhu et al. [22] proposed an energy effective routing protocol that adapts unequal clustering techniques for solving the hot spot problems and proposed double CH approach for reducing the energy utilization of CHs in the cluster. Additionally, to uniformly manage the energy utilization among CMs and CHs, a hybridized CH rotation approach depends on energy driven and time driven is projected that could create the timing of rotation highly moderate and the energy utilization highly effective. Arjunan et al. [23] present an method called FL based unequal clustering, ACO based Routing, and Hybrid protocol for WSN to remove hot spot problems and also enlarged the lifespan of network. This process includes cluster maintenance, CH selection, and intercluster routing. The FL selects CH effectively and partitions the network into unequal clusters on the basis of DBS, RE, node centrality, distance to its neighbours, and node degree. It utilizes ACO based routing method for reliable and effective intercluster routing from CH to BS.

System Model
Consider an unequal clustering in WSN with several nodes effectively to organize the sensors, balances load distribution and avoids hot spot issue [24]. But, the equal clustering method frequently has cluster size i.e., the similar over the network. Indeed, in unequal clustering, the size of the cluster is established on the basis of DBS. The cluster size is shorter in case of smaller DBS and the size gets increased with a rise in DBS. The wireless radio transceiver in the network is based on distinct variables, for example, distance and energy utilization. The description is interrelated to the energy utilization and the communication range of the nodes. The distance among the receiver and transmitter followed on the reduced trans-receiving power exponentially reduces with increased distance. A threshold splits the multipath and free space models. Fig. 2 illustrates the first order radio energy model.
Whereas ε_fs, ε_mp denotes power loss for free space and multipath modules, correspondingly, and d_0 denotes a threshold of space module. The broadcasting l bits for energy spent on distance d is expressed by: Whereas E denotes the power utilization of the node, Tx indicates broadcasting subscript elec, amp indicates electronic, and amplify for modulating, digital coding, spreading signal, and filtering. The energy utilization for getting the messages could be formulated by: WSN considered a set of N nodes in the 2D region of M^2 using k clusters. Where d denotes the intracluster distance and D indicates the intercluster distance. The expended energy of the cluster for a data packet is modeled by: Whereas N/k-1 denotes the average CMs in a cluster. E_CH and E_members indicates the dissipated spent for CH and CMs. The estimated energy consumption for the CHs and CMs are given in Eqs. (6) and (7).
The overall expended energy for a WSN can be expressed from the power transceiver and energy procedure as: Whereas E_p denotes the power utilization of the microcontroller. It doesn't affect the optimizing procedures. Therefore, the power utilization of E_frame optimized is depends only on distance for the cluster optimization.

The Proposed SGOBUC Technique
The workflow involved in the SGOBUC technique is exhibited in Fig. 3. The SGOBUC technique involves three phases namely initialization phase, neighboring data collection phase, and finally, cluster construction phase. At first, the nodes are arbitrarily placed in the WSN and are initialized together. Then, the information about the neighboring nodes is fathered in the network. Followed by, the SGOBUC

Initialization Phase
The BS in the WSN starts the network setup message using the timing information, location details, broadcasting signal strength, and power. The node existing in the network calculates its distance from the BS according to the Haversine distance equation. The position of the nodes can be anticipated by the analyses of signals received and transmitted. The Global Positioning System (GPS) discovers the node position by gathering the longitude and latitude data. The latitude data from the GPS is presented in seconds, degrees and minutes as given below: The main concept of the Haversine performance is that the earth is in sphere shape instead of ellipse. The numerical formulation of Haversine distance calculation is given below, Whereas ∂ -distance, d lt , d lg -Deviation values of longitude and latitude, α n ,-Nodes latitude data, and α req -requested latitude data. The distance calculation from the coordinates (11) is arithmetically given by, Whereas R-Radius of the earth. The energy utilization of the hops near the BS rises to the maximal level.
Hence the amount of packet communication between the nodes rise and entire efficiency of the WSN reduces.

Neighboring Data Collection Phase
The node exploits the Medium Access Control (MAC) protocol for finding the nearby nodes. Initially, the node transmits its information to the network via the node details message which contains Node_ID, layer_ID, distance to BS, and RSSI. Because of the non-functioning of nodes and link failures, the neighbour node manually alters on the basis of network topology. In order to create the presented method as dynamic to node failures, this stage is continued on the particular amount of simulations. The node termination is because the energy consumption creates the neighbour finding stage to be repeated and thus the packet transmission gets increased.

Unequal Cluster Construction Phase
Once the sensor nodes gathered the information, the SGOBUC technique gets executed to elect CHs and cluster sizes. Seagull perform migration process, where they migrate from one position to other for finding most abundant and richest food sources that would provide sufficient energy. Such behaviors are given in the following section: a) At the time of migration, they move within a group. The early position of seagulls is distinct to evade the collisions among them. b) With their groups, they move in the course of optimal survival fittest seagull, viz., when a fitness value is lower than others. c) According to the fittest seagull, another seagull could upgrade their early position.
They often attacks migrating birds on the ocean while migrating from one position to other. It could create their spiral natural shape motion at the time of attacking. Such behaviours are related to the objective function to be optimized. This can enable to development of a novel optimization method. This study emphases on 2 natural behaviours of the seagull. The mathematical model of migrating and prey attacking processes are mentioned below. At the time of migration, this method inspires the way that the seagulls travel from one place to other. A seagull must fulfil 3 criteria as given below: In order to evade the collision among neighboring ones (viz., another seagull), a further parameter A is applied for calculating novel searching agent location.
WhereasC s denotes the location of searching agent that doesn't collide with another searching agent,P s indicates the present location of searching agent, x signifies the present round, and A indicates the migration nature of searching agent in the provided searching region.
where x ¼ 0; 1; 2; . . . ; Max iteration . Let f c is presented for controlling the frequency of employed parameter A i.e., reduced linearly from f c to zero. In this study, the value of f c is fixed to two. The thorough sensitivity analyses of f c has been deliberated.
Afterward evading the collision among neighboring ones, the searching agent is traveling in the direction of optimal neighbor. M s ¼ B Â ðP bs ðxÞ ÀP s ðxÞÞ (15) WhereasM s denotes the place of searching agentP s to the optimal fit searching agentP bs (viz., fittest seagull). The behaviour of B is randomized i.e., in charge for balancing properly between exploitation and exploration. B is evaluated by: Whereas rd denotes an arbitrary amount in the range of zero & one. Finally, the searching agent could upgrade its location regarding optimal searching agent.
In whichD s indicates the distance among the searching agent and optimal fittest searching agent (viz. optimal seagull fitness value is lesser). The exploitation aims to use the experience and history of the search procedure. They are capable of changing the speed and angle of attack continually at the time of migration [25]. They preserve their altitude with their weight and wings. During the attacking process, the spiral motion happens in the air.
These behaviors in, y, and z plane is given below.
r ¼ u Â e kv (21) Whereas r denotes the radius of every turn of the spiral, k indicates an arbitrary amount in range [0 ≤ k ≤ 2π]. u and v represent constant to determine the spiral shape. The upgraded location of searching agent is evaluated by Eqs. (17)- (21).
whereP s ðxÞ stores the optimal solution and upgrades the location of other searching agents. The projected SGO begins at an arbitrary created population. The searching agents could upgrade their locations regarding their optimal searching agent at the time of iteration procedure. A is reduced linearly from f c to zero. For smooth transition among exploitation and exploration, parameter B is liable. Thus, SGO is deliberated as a global optimizer (Algorithm 1) due to its optimal exploitation and exploration ability. A fitness function is derived using five parameters as defined in the following.
At time t, the amount of adjacent nodes, for instance, the node degree is signified as: where N(x) = {n y /dist(x, y) < trans range } x ≠ y, and dist(x, y) signifies the distance amongst 2 nodes n x and n y , trans range refers the coverage area.

(4) Residual energy (RE)
The RE of sensors on broadcast of k bit data to node n y in distance d is signified as: where E signifies the present energy level of nodes, E T implies the energy spent for transferring data.
(27) where E e refers the energy of electrons and E a implies the required amplified energy, E R ðkÞ signifies the energy spent for reception and is referred to as: The centrality of the node N c is defined below: (i t , j t ) and (i t , j t , j t−1 ) represents the coordinates at the time t and t − 1 correspondingly.

Experimental Validations
The proposed model is simulated using MATLAB tool. The results are examined interms of different measures namely average RE, average delay, packet delivery ratio (PDR), number of alive nodes, and network lifetime. Tab. 1 and Fig. 4 Fig. 6 depict the PDR analysis of the SGOBUC approach with other existing methods. From the attained outcomes, it can be clear that the SGOBUC method has attained maximal PDR over the other existing methods. For instance, with round 100, the SGOBUC manner has offered a superior PDR of 0.9850% whereas the IPSO, KHA, F5NUCP, FUCHAR, and SUCID manners have obtained a lower PDR of 0.8000%, 0.8540%, 0.8700%, 0.9000%, and 0.9700%. Also, with round 400, the SGOBUC algorithm has offered a higher PDR of 0.9540% whereas the IPSO, KHA, F5NUCP, FUCHAR, and SUCID techniques have obtained a lower PDR of 0.7300%, 0.8000%, 0.8000%, 0.8500%, and 0.9435% correspondingly. Along with that, with round 700, the SGOBUC method has existing an improved PDR of 0.9279% whereas the IPSO, KHA, F5NUCP, FUCHAR, and SUCID methods have reached a least PDR of 0.6600%, 0.7300%, 0.7500%, 0.8200%, and 0.9135% correspondingly. Eventually, with round 1000, the SGOBUC technique has offered a higher PDR of     Fig. 8. The experimental outcomes pointed out that the SGOBUC technique has accomplished better network lifetime over the other methods with the maximum FND, HND, and LND. For instance, the proposed SGOBUC technique has obtained a higher FND of 3184 rounds whereas the IPSO, KHA, F5NUCP, FUCHAR, and SUCID techniques have demonstrated a lower FND of 485, 685, 898, 1114, and 1985 rounds respectively. Next to that, the presented SGOBUC manner has gained a maximum HND of 13958 rounds whereas the IPSO, KHA, F5NUCP, FUCHAR, and SUCID algorithms have outperformed a minimum HND of 9840, 6253, 7115, 5496, and 13049 rounds correspondingly. At last, the projected SGOBUC method has achieved an improved LND of 19876 rounds whereas the IPSO, KHA, F5NUCP, FUCHAR, and SUCID methodologies have showcases a lesser LND of 17021, 15049, 13296, 11192, and 18974 rounds correspondingly.
By looking into the above-mentioned results analysis, it is ensured that the SGOBUC technique has accomplished improved outcome over the other existing techniques with the maximum energy-efficiency and network lifetime.

Conclusion
This paper has devised an effective SGOBUC technique to eradicate the hot spot issue in WSN. The SGOBUC technique involves three phases namely initialization phase, neighboring data collection phase, and finally, cluster construction phase. The SGOBUC technique derives a fitness involving different  parameters in such a way that the energy efficiency can be accomplished The application of unequal clustering technique assist to uniformly balance the energy dissipation throughout the network. An extensive experimental validation process is performed to ensure the enhanced outcomes of the SGOBUC model. The experimentation values highlighted the better performance of the SGOBUC technique over the other techniques interms of different evaluation parameters. In future, novel routing discovery and route planning strategies can be designed for WSN. Moreover, effective data aggregation and MAC scheduling techniques can be designed to further optimize the resource utilization in the network.
Funding Statement: The authors received no specific funding for this study.
Conflicts of Interest: The authors declare that they have no conflicts of interest to report regarding the present study.