Hybrid Metaheuristic Aided Energy Efficient Cluster Head Selection in Wireless Sensor Network

Clustering is one of the significant techniques for expanding the lifetime of networks in wireless sensor networks (WSNs). It entails combining of sensor nodes (SNs) into clusters and electing cluster heads (CHs) for each and every cluster. CH collects the information from particular cluster nodes and passes the cumulative data to the base station (BS). However, the most important requirement in WSN is to choose a suitable CH with an increased network life span. This work introduces a new CHS model in WSN. The optimal CH is elected by a new hybridized model termed as “Lion Updated Dragonfly Algorithm (LU-DA) that hybrid the concepts of Dragonfly Algorithm (DA) and Lion Algorithm (LA)”. Moreover, the optimal selection of CH is done depending upon constraints like “energy, delay, distance, security (risk) and trust (direct and indirect trust)”. This optimal CH ensures the network lifetime enhancement. At last, the superiority of the developed approach is proved on varied measures like energy and alive node analysis. Accordingly, the proposed model has accomplished higher energy of 0.55 at 1 round, whereas at the 2000 round, the normalized energy value has been dropped to 0.1. Keywords—Cluster head; security; trust; dragonfly algorithm; LU-DA model

 Proposes a Lion Updated Dragonfly Algorithm for optimal CH selection, which integrates the concepts of DA as well as LA models.
The organization of the paper is as follows: Section II reviews CHS models. Section III elaborates the adopted energy aware clustering model in WSN. Section IV depicts the optimal CHS: objective model. Section V explains Lion Updated Dragonfly Algorithm for optimal CHS. The resultants and conclusions are briefed in Section VI and Section VII.

A. Related Work
In 2020, Turki et al. [24] suggested a new clustering model with optimal CHS that considered 4 most important criterions such as security, delay, energy, and distance. Furthermore, for electing the optimum CHs, this work proposed a novel algorithm named as FF-PUD. At last, the performances of developed scheme were performed by evaluating it over other schemes regarding risk, alive nodes, energy and delay.
In 2020, Prachi et al. [25] have employed BOA for choosing an optimal CH from nodes. Accordingly, the developed work aims on lessening the energy usage and for maximizing the life span of network. The path among the CH and BS was determined using ACO and it selected optimal routes depending on node degree, residual energy and distance. At the end, the supremacy of adopted work was proved regarding energy consumption, alive and dead nodes.
In 2019, Reeta and Dinesh [26] have designed a multiobjective model that dependents on distance, traffic rates, energy, cluster densities and delay. Here, energy based routing was carried out based on MOFPL scheme. The implemented model determined the optimum CH from several nodes in WSN. Consequently, the optimal routes were introduced depending on the adopted multi-objective function. Furthermore, effective CHS with high network energy was accomplished by the designed model.
In 2020, Augustine and Ananth [27] have presented an enhanced framework for CHS based on Taylor KFCM that was modified from the KFCM approach in the Taylor series.
The introduced model has chosen the CH by means of "acceptability factor" that was evaluated by the trust, distance, and energy. Further, the advantage of the proposed system was proved in terms of highest energy and high trust.
In 2019, Goswami et al. [28] introduced a cluster-based model by deploying HML and FF model in OWSN for improving the EE and minimizing the costs. Here, the issues in FF model were prevailed over by integrating the theory of HML with it. Furthermore, the distribution of power in nodes was carried out precisely via maximum likelihood property of HML. Finally, the resultants have shown the betterment of presented scheme regarding EE and cost function.
In 2019, Jain and Toor [29] offered a new framework for diverse WSN by considering "MEACBM routing protocol". Accordingly, optimal election of CHs takes place; particularly, the SNs with higher energy were preferred as CH. This model has minimized the energy utilization of SNs while conveying data to BS. The analysis resultants have revealed the enhancement regarding the CH count, network lifetime, throughput and dead node count.
In 2019, Daneshvar et al. [30] have offered a new clustering scheme, which selected CHs by means of GWO. For selecting CHs, the solutions were optimized depending on remaining energy of every node and predicted energy utilization. In addition, for improving the EE, the presented model deployed the similar clustering in numerous successive rounds. This allowed the framework to accumulate the energy, which was necessary for reforming the clustering. Eventually, the outcomes demonstrated that the designed model has ensured effective network lifetime.
In 2018, Tianshu et al. [31] suggested a routing scheme depending upon GECR and GA for expanding the lifetime of networks and improving EE. In addition, while modelling the objective function, the "load balancing factor" was taken into account that balanced the energy usage among SNs. The simulated results have exposed the supremacy of the adopted method with lower variance and improved EE. Table I makes a review of existing cluster-based energyaware CHS models in WSN. Numerous methods have been focused on energy-aware CHS models in WSN. But still, the existing models like FF-PUD [24], BOA + ACO [25], MOFPL [26], Taylor KFCM model [27], FF [28] have some common problems like high convergence, local search issues in FF, high-cost efficiency, there is a need of standard optimizations and need consideration on constraints like security and trust.

C. Objectives
The main objectives of this paper are:  To select an optimal CH depends on certain constraints such as energy consumption, trust, security, delay and distance.
 To propose an improved Algorithm for optimal CH selection for solving the optimization issues.
 And to improve the better convergence rate.

A. Network Model
Assume n M sensor nodes that are randomly deployed in appliance area. Consequently, the clustering process is done by merging the SNs. During clustering, the nodes forms clusters, wherein a CH is elected and the total count of CH is delineated by n CH . Thus, the distances amongst nodes and CHs have to be reduced.
The most important task of WSN is to transfer the information among nodes. Here, the identification of shorter paths is required to enhance the data transmission. Moreover, the energy consumption of node also acts as the most role while transmitting the data. Particularly, a node requires more energy for transmitting massive data. In the clustering based strategy, the CH is responsible for transmitting more data with less energy consumption. However, the security is more crucial for minimizing the overhead and attacks. The architectural depiction of adopted model with varied SNs is illustrated in Fig. 1.

B. Distance Model
In the network, a CH is chosen only if the distance between CH and nodes is minimal. If distance among CH and nodes are higher than distances amid node and BS, the data are transmitted directly to BS by node. By deploying distance matrix ) * ( w g Di , the SNs gets clustered with selected CH as exposed in Eq.  passes the specified data to BS.

C. Energy Model
Energy utilization is a foremost characteristic in WSNs. Actually, additional energy is crucial for conveying data to BS from every SNs. Thereby, the energy model for transmitting data is exposed in Eq. (3), wherein, " ete E symbolizes the electronic energy as given in Eq. (4), wherein agg E refers to the energy utilization during data collection and ) : ( e M E TX signifies the energy necessary for transferring M bytes of packets at distance e ". Eq. (5) shows the essential energy for passing M bytes of packets. Eq. (6) shows the "amplification energy and pr E refers to power amplifier energy and fr E refers to energy required for deploying free space technique" [24].
The whole energy of network is given in Eq. (8), wherein 1 E symbolizes the energy at idle state and ST E symbolizes energy at sensing time.

D. Security Model
The risky mode,  -risky mode and security mode are the factors in security model that are explained below.
"Risky mode: This mode selects an existing CH for facilitating an optimal CHS, for which it takes all the risks. Thus, this mode is considered as an insistent mode while choosing CH [24].
 -risky mode: The CH which could tolerate the utmost  -risk are elected based upon  -risky mode. Accordingly,  signify the probability measure with values, 0   and 1   (i.e., 100%) similar to security and risky mode.
Security mode: This mode prefers the CH that fulfills the needs of security. In Eq. (1), r s and d s denotes the security rank and security needs associated with CHS. If r d s s  , the node are considered as CH". , there will be a delay in the selection process. Still, the CHS process could not be completed, and the corresponding function should be continued for the state refers to value of direct trust computed by A , P as well as P and G in that order.
The fitness function for distance is specified by Eq. (16)

A. Solution Encoding
The proposed work focuses on introducing an efficient CHS in WSN. In WSN, the optimum selection of CH is a versatile task and it is performed based upon criteria like, trust, security energy, delay and distance. The input provided to LU-DA algorithm is shown by Fig. 2.

B. Proposed LU-DA Algorithm
Although the conventional DA [33] model encompasses a variety of enhancements; it suffers from specific limitations like, complexity in solving binary problems etc. Therefore, the theory of LA [34] is mingled with it to introduce a new model named as LU-DA. Hybridized optimization schemes are said to be more appropriate for specific search issues [23] [35] [36] [37]. The steps followed in the proposed LU-DA are as follows. LU-DA model involve two stages such as: "(i) Exploration and (ii) Exploitation". The separation formula is modelled as exposed in Eq. 708 | P a g e www.ijacsa.thesai.org Distraction to enemies is indicated by Eq. (25), wherein, the enemy position is denoted as  S .
Eq. (26) On the other hand, if , the position gets updated as in Eq. (30).

A. Simulation Setup
The adopted LU-DA based CHS in WSN was simulated in MATLAB. The analysis was held by evaluating the alive node count for varied number of round that ranges from 0 t0 2000. Further, log of alive node count was analysed for varied distance that range from 0, 20, 40, 60 and 80. In addition, cost analysis was done for varied iterations that range from 0, 2, 4, 6, 8 and 10. Also, the proposed model was computed over extant approaches such as FF [8], GWO [9], LA [34], DA [23] and FPU-DA [24] and the outcomes were examined in terms of statistical analysis. The simulation parameters in this work are summarized in Table II.

B. Analysis on Alive Nodes
The Analysis on NAN of suggested LU-DA scheme over FF, GWO, LA, DA and FPU-DA models is specified in Fig. 3(a), whereas, the log of NAN analysis is shown in Fig.  3(b). The Analysis on NAN is done for 2000 rounds, while, log of NAN is performed for varied distance that range from 0, 20, 40, 60 and 80. In fact, the transmission of data and clustering operations continues for several rounds up to the death of every node. Thereby, the NAN in cluster gets reduced in every cluster. Thus, the NAN reduces with raise in rounds. Till 700 th round, the NAN for conventional and proposed models is 100, and it is gradually reduced as the round gets improved. Nevertheless, at 2000 th round, the adopted scheme reveals higher NAN than extant models, thus guarantying the enhanced performance of adopted scheme. Particularly, within 75% variation in rounds (i.e. from 500 th round to 2000 th round), the NAN using presented technique has dropped from 100 to 40. On the other hand, for similar variation (75%) in rounds, the NAN using conventional GWO has dropped from 100 to 19. Thus, the analysis established the enhanced efficacy of LU-DA method with the subsistence of more NAN. Fig. 4 describes the examination on normalized energy attained using suggested LU-DA model over traditional models for varied number of rounds that ranges from 0 to 2000. The normalized energy is portrayed depending upon the residual network energy and it have to be high for better system performance. In Fig. 4, the network energy seems to be higher at initial rounds; however, with increase in rounds, the energy starts lessening steadily for both adopted as well as compared extant schemes. Especially, from Fig. 4, the presented model has attained higher energy of 0.55 at 1 st round, while at 2000 th round; the normalized energy value has been dropped to 0.1. However, the adopted model has accomplished a higher energy even at 2000 th round, when distinguished over FF, GWO, LA, DA and FPU-DA models. Thus, the capable performance of developed model is confirmed. Fig. 5 describes the convergence analysis of the adopted model over conventional approaches regarding cost. Here, Analysis is performed for a varied number of iterations that ranges from 0, 2, 4, 6, 8 and 10. On noticing the analysis resultants, the developed LU-DA has a negligible cost for all iterations over conventional approaches. Predominantly, on noticing cost values from Fig. 5, the adopted scheme has attained reduced cost value (0.06) from iteration 3 to 10. At the initial iterations (from 1 to 3), the cost of developed model has accomplished a comparatively higher value, while at further iterations; the developed model has converged to a minimal cost value. Especially, at iteration 2, the adopted model is only 60% enhanced than extant FF model, while at iteration 10, the adopted model is 62.5% enhanced than FF model. Thus, the overall assessment shows the impact of the developed LU-DA on better convergence results with increase in iterations.    Table III describes the statistical analysis of the presented LU-DA model over prevailing approaches regarding alive nodes, energy, cost and time. "As meta-heuristic schemes are stochastic in nature, every algorithm is executed for the number of times to attain the statistic of objective function". On noticing the resultants, the adopted LU-DA model has obtained high NAN and energy values and minimal cost and time values for all scenarios, when compared over the existing schemes. Especially, on noticing the NAN from Table III, the best case scenario using proposed LU-DA model has attained superior values over distinguished schemes. Moreover, the cost and time values of conventional schemes is superior to the presented model. Nevertheless, at specific scenarios, the conventional schemes have exposed its enhanced performance regarding time. Though, the entire evaluation of objectives reveals the impact of improving the adopted scheme, thus ensuring a secured transmission.

F. Analysis on Delay, Security and Trust
The analysis on delay, security and trust attained using implemented model over existing models is tabulated in Table IV. Here, examination is done by altering rounds from 0 to 2000. On analysing the delay, the presented LU-DA model has obtained high values for trust and minimal values for delay and security (risk) for all rounds. Initially, at round 1, the delay of adopted scheme seems to be 0.98, while, as the number of rounds increases, the developed method has acquired a minimal delay value of 0.89 at 2000 th iteration. Simultaneously, while analysing the security (risk), the developed approach at 1 st round has accomplished a minimal risk value of 0.06, whereas, at 2000 th iteration, a comparatively higher risk value of 0.094 has been acquired by adopted model. However, for all rounds, the developed approach has acquired fine outcomes than the compared models as per the desired objectives. Also, while examining the trust values, the adopted method has accomplished a higher trust value of 0.32 at 2000 th round, whereas, the compared models like FF, GWO, LA, DA and FPU-DA has acquired relatively minimal values of 0.24443, 0.21992, 0.17411, 0.23624 and 0.23446. Therefore, the improvement of LU-DA scheme is confirmed from the outcomes.   Fig. 6 shows the parametric analysis of normalized energy, convergence, and alive nodes. In the LU-DA scheme algorithm, there is a random number that varies from 0-1. Here, the analysis has been done by varying the random variable r=0.2, r=0.4, r=0.6, r=0.8, and r=1. On observing the results, it can be noticed that when r=0.4, the proposed LU-DA attains the best results. By setting the random variable r=0.4, the above analysis like normalized energy, convergence, and alive nodes has been portrayed by comparing with other existing models.

H. Discussion
This paper presents a new LU-DA model for optimal CHS. Here, the optimal CH selection is carried out by considering the constraints like "energy, delay, distance, security and trust. Here the analysis is performed for alive nodes, normalized energy, convergence analysis, analysis based on delay, security and trust. The betterment of the proposed LU-DA model is proved on various measures like energy and alive node analysis. From the above analysis, it is evident that the proposed model attains better results when compared with the existing models like FF, GWO, LA, DA and FPU-DA. Thereby attaining improved the network lifetime. www.ijacsa.thesai.org

VII. CONCLUSION
This paper introduced a new LU-DA model for optimal CHS. For optimization, a novel approach termed as LU-DA was developed. Eventually, the primacy of offered method was confirmed over conventional models. The presented model has attained higher energy of 0.55 at 1 st round, while at the 2000 th round; the normalized energy value has been dropped to 0.1. However, the adopted model has accomplished higher energy even at the 2000 th round, when distinguished over FF, GWO, LA, DA and FPU-DA models. At the initial iterations (from 1 to 3), the cost of the developed model has accomplished a comparatively higher value, while at further iterations, the developed model has converged to a minimal cost value. Especially, at iteration 2, the adopted model was only 60% enhanced than the extant FF model, while at iteration 10, the adopted model was 62.5% enhanced than FF model. Therefore, the development of the suggested LU-DA scheme was authenticated over other techniques.