Wireless Sensor Networks: Network Life Time Enhancement using an Improved Grey Wolf Optimization Algorithm 


Kanthi M and Ravilla Dilli*


Electronics and Communication Engineering, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, Udupi District, Karnataka 576104, India.

*E-mail: dilli.ravilla@manipal.edu (R. Dilli)



Clustering is an effective technique in a wireless sensor network (WSN) to minimize energy consumption and low-energy adaptive clustering hierarchy (LEACH) is the most popular clustering protocol in WSNs. However, a random selection of cluster head (CH) in LEACH protocol results in poor performance in real network deployments. Dynamic formation of CHs and energy-aware clustering schemes helps in enhancing the lifetime of WSNs. In this paper, we have proposed an improved version of the grey wolf optimization (IGWO) algorithm to overcome the premature convergence of conventional GWO algorithm. The algorithm is applied to optimize the CH selection in WSNs to maximize the network lifetime. The improvements of IGWO algorithm are based on sink distance, CH balancing factor, residual energy, and average intra‐cluster distance. The proposed algorithm has been tested in terms of the number of rounds, number of operating nodes, number of transmissions, and energy levels used for communications. The performance results of IGWO algorithm are compared with a conventional LEACH protocol, it is observed that the total number of operational rounds has been increased by 441.4% and 869.6% for a network size of 50 and 699.8% and 990.8% for a network size of 100 when CH selection probability is 5% and 10%, respectively. The simulation results show that the proposed IGWO based LEACH protocol outperforms the existing state-of-the-art algorithms based on GWO for enhancing the WSNs lifetime.  


Table of Contents


Keywords: Wireless sensor network; Low-energy adaptive clustering; Grey wolf optimization; Sensor network lifetime; Cluster head selection; Improved grey wolf optimization algorithm.


1. Introduction

A wireless sensor network (WSN) is a combination of several sensors installed on physical devices that are geographically distributed in a network field to collect the data based on the application. The scattered nodes can be grouped into clusters to deliver the information to the base station (BS). The sensor nodes are self-organized, having the ability to sense, process, and communicate. In most WSN applications, it is not feasible to recharge or replace the batteries of sensor nodes. The main challenges in WSNs are the optimal number of CHs, energy efficiency, optimal network coverage, network lifetime, stability, reliability. Designing an energy-efficient protocol for optimum selection of CHs is essential for maximizing network lifetime in WSNs. Low energy adaptive clustering (LEACH) is a hierarchical protocol at MAC sublayer used to form clusters in WSNs.[1] The node which has maximum residual energy becomes CH to improve energy efficiency and network lifetime. The functioning of LEACH protocol has two phases: the “setup phase” in which cluster formation is performed and in the next phase called the “steady phase”, information is transferred to the sink node.

1.1 Setup phase

In this phase, every sensor node chooses a random number in the range 0 to 1 and if this number is lower than the threshold of nth node ‘T(n)’, then the sensor node becomes CH for that specific cluster and the remaining nodes will act as cluster members. The CH selection phase and clustering phase are the key parts of the LEACH protocol. Each sensor node is having threshold energy as:

T(n)  =         n ϵ G                             

     =    0               otherwise                                (1)

Initially, CHs are chosen based on a threshold value T(n) and BS transmits the message to every CH of the first round. This strategy ensures that all the sensor nodes uniformly spend equal energy. Once CH selection is completed, the CHs advertise their selection to all sensor nodes in the network. After receiving the advertisements, the sensor nodes choose their nearest CH based on the received signal strength, and then CHs assign a TDMA schedule for the nodes in their respective clusters.

1.2 Steady phase

In this phase, cluster members join their CHs and transmit data based on the TDMA schedule. To achieve energy conservation, CHs perform data aggregation (or fusion) through local computation and the resultant data is further transmitted to BS. A longer steady phase minimizes overhead of cluster formation. After a certain amount of time in the steady phase, the selection of CHs is repeated through the setup phase. Adaptive CH selection based on event occurrence instead of periodic selection significantly minimizes message overheads and computations without compromising for network throughput and end-to-end delay.[2]

1.3 Grey Wolf Optimization (GWO) algorithm

GWO is a meta-heuristic algorithm inspired by the food searching behavior of grey wolves and it mimics the hunting mechanism and leadership hierarchy of grey wolves present in nature. The wolves do attacks in a group and the best position of any wolf can be provided with the best solution. They follow a strict social dominance hierarchy as alphas (α), beta (β), delta (δ), and omega (ω) which is shown in Fig. 1 and the implementation of hunting, searching, encircling, and attacking the prey. The dominance\ decreases from top to bottom of the hierarchy. This algorithm is applied to WSNs in which the position of wolves represents the position of CH in each cluster and the position of the grey wolves updates the position of CH in LEACH protocol.

Fig. 1 Hierarchy of Grey Wolves.


A schematic illustration of the synthesis process of the α wolves (both male and female) are most responsible in the decision making, organization, the discipline of the group and the other category of wolves in a group follows α wolf’s decisions. In the second level, β wolves (either male or female) help, advice the α wolf in decision-making, feedback and pass commands to lower level wolves in the hierarchy. δ category wolf is subordinate of α, β and dominant of ω wolves. δ wolves play a role of scouts, sentinels (guarantee the safety and protection of group), elders (experienced wolves who are used by α and β wolves, hunters (help the α, β wolves in hunting for providing food for the group), caretakers (take care of weak, ill and wounded wolves in the group). Grey wolf, ω is the lowest level animal in a group that follows orders of all other dominant wolves. ω wolf assists the entire group and if ω wolf is absent in a group, there are chances of the internal fight due to frustration of all other wolves. The main steps of group hunting by grey wolves are: 1) Tracing, chasing, and approaching the prey; 2) pursuing, encircling, and harassing the prey till it stops movement; 3) attack towards prey. The social hierarchy and hunting techniques of grey wolves are modeled to design the GWO algorithm and its performance optimization. The GWO algorithm is used in solving real time engineering problems across various fields.[3-4]


2. Performance optimization in WSNs

Optimization of CH selection in WSN greatly improves the network lifetime. As energy is a major constraint in WSNs, optimum energy utilization is an important aspect in WSNs to enhance the lifetime of the network and this feature is essential for IoT and IIoT applications. As a part of energy harvesting in WSNs, there are many algorithms and techniques proposed in the literature for CHs selection based on residual energy, centrality, the position of nodes, etc., to minimize the energy consumption and improve network lifetime without compromising the reliability of the network. Cluster formation and CH selection as well as re-formation and re-selections based on residual energy further improve network efficiency over LEACH protocol. However, the reformation of clusters occurs only if there exists a large separation between CHs and sensor nodes. This approach improves the efficiency of the network in terms of lifetime and decreases complexity over the LEACH protocol.[5] In a normal operation of LEACH protocol, the energy consumption of WSN is equalized by randomly selecting CHs in an iterative fashion and it leads to network instability. Therefore, it is essential to optimize the routing protocol and number of CHs in order to minimize the energy consumption for data transmissions. Decentralized clustering protocols that differentiate intra-cluster and inter-cluster communications reduce the number of transmissions and routing control packets thereby longer lifetime for the network.[6] CDMA-based LEACH protocol uses a different set of codes in each cluster to reduce inter-cluster interference.[7] Rank-based algorithms consider a number of links between nodes and path cost while electing CHs, LEACH protocol based on node rank enhances the lifetime of WSN.[8] Using Voronoi diagrams to form clusters and ant colony algorithms to optimize multi-hop routing can improve the first node death time by 14.5% and 127% compared to SEP and LEACH protocols respectively.[9] A Combination of ARSH-FATI-based CH selection algorithm and rank-based clustering minimizes the consumption of communication energy at sensor nodes and this algorithm switches dynamically between exploration and exploitation of search process during the run-time to have a 25% higher network lifetime over PSO.[10] Fuzzy based clustering provides an efficient data aggregation and enhances the life span of WSNs.[11,12] In heterogeneous WSNs, adaptive clustering routing techniques consider residual energy at a given node and its location for efficient CHs selection. The nodes which are close to BS and have more residual energy become CH with higher probability. The CHs which are far away from BS uses multi-hop routing and the other nodes use single-hop routing. These features balance energy consumption and enhances the network lifetime, network throughput over SEP and DEEC protocols.[13] A hybrid meta-heuristic algorithm with a combination of cuckoo search and Krill herd for homogeneous and heterogeneous WSNs environments enhance their lifetime. Krill herd algorithm is used to compute the optimal cluster centroid positions and a cuckoo search algorithm is applied to select optimal CHs.[14]


3. Performance optimization of WSN based on GWO algorithm

The selection of CH is not optimal in conventional LEACH schemes, therefore usage of GWO improves the optimal selection of CH in LEACH protocol.[15] GWO is a unique metaheuristic algorithm that mimics the social behavior of grey wolves with respect to their leadership hierarchy and, attacking strategy. Localization problems (search for the geographical position of unknown nodes using anchor nodes) in WSNs can be resolved using the GWO algorithm. CHs selection using GWO considers the distance between clusters and sink, current residual energy of each node and predicted energy consumption. This technique considers the same clustering in consecutive rounds to enhance energy efficiency and the saved energy can be utilized for cluster reformation. For CHs that are far away from BS can use dual-hop routing and it ensures optimum energy consumption.[16] GWO is applied in defining objective functions and weights for efficient cluster formation and CH selection.[17] Appropriate fitness function ensures the coverage of WSN and it is fed to the GWO algorithm to determine its optimum. This procedure outperforms the LEACH clustering and routing algorithm in terms of network throughput, lifetime and, residual energy.[18] Energy efficiency and network stability are two typical trade-off parameters in WSNs. Distance-based stable CDS method and clustering algorithms using GWO in WSN minimizes effective transmission distance with deterministic CH selection and achieves a load-balanced, energy efficient and stable network. These algorithms perform much better than GA and LEACH by 70.5% and 74.7% respectively in terms of network stability and energy efficiency.[19] To improve the energy-efficiency of the clustering mechanism, a combination of two meta-heuristic algorithms named whale and GWO can be used to define the objective function for the formation of an optimal number of clusters, dynamic CH selection, and relay nodes selection on a priority basis.[20,21] MLHP with three levels based on the GWO for WSNs shows better performance in terms of maximum residual energy, network lifetime, stability period compared to other algorithms named LEACH, DEEC, and SEP. At level 1, BS selects CHs, at level 2, GWO routing gives the best route to BS for data transfer, and at level 3, distributed clustering is performed based on cost function.[22] In the cluster-based architecture of WSNs, gateways which are far from BS communicate with BS through the gateways that are close to BS. This causes faster energy depletion at gateways closer to BS because of heavy traffic load and causes energy hole problems around BS. GWO can resolve the energy hole issue by distributing traffic load with minimum traversal distance and number of hops. GWO based algorithm performs better than GA, PSO and multi-objective fuzzy clustering.[23] The percentage of localized nodes, quick convergence rate and success rate of GWO algorithm is higher with less computation time and, localization error compared to PSO, and MBA metaheuristic algorithms.[24] Random deployment of sensor nodes causes a high degree of aggregation and a low coverage rate. The simulated annealing method embedded into GWO avoids falling into local optimum, improves convergence rate, accelerates the convergence speed, coverage optimization and this leads to higher network stability and network life cycle in WSNs.[25] GWO is used to compute the threshold for sensor decision rules from the fusion center that is independent on initial values and with lesser complexity. The results of this algorithm demonstrate the reduced buyer’s risk by 15%-20% of fusion system.[26] The Virtual Force-Levy technique embedded with the GWO algorithm effectively enhances the coverage optimization in WSNs in terms of uniformity in nodes distribution, node’s average moving distance, changes in the number of nodes, and coverage rate.[27] The robust deployment of WSNs for industrial applications introduces redundant nodes and causes overhead. Quantum clone GWO algorithm is proposed to design and optimize the sensor duty cycle, also avoid falling into a local optimum. The convergence speed of this algorithm is much faster than GA, SA, and enhances the network lifetime.[28] The combination of crow search optimization with the GWO algorithm for optimal CH selection enhances the network lifetime by reducing delay, energy stabilization, and distance between nodes.[29] Providing energy efficiency in heterogeneous WSNs is much more challenging due to the irrational utilization of energy at nodes. For this kind of network, GWO provides a solution based on the node’s fitness values. The fitness values are treated as weights and they are dynamically updated based on the distance between wolves and their prey. This algorithm ensures the optimal CHs selection and enhances the network lifetime by 55.7%, 46.3%, 27.0%, and 31.9% over SEP, modified SEP, fitness value based-GWO, and DEEC respectively.[30] The combination of GWO and GOA algorithms improves the convergence levels of meta-heuristic algorithms.[31] GWO can be used in the implementation of server-less systems to improve task allocation and minimize runtimes.[32] The performance of intrusion detection systems in WSNs can be enhanced using binary GWO with a support vector machine. These techniques can minimize false alarm rates in the WSN environment thereby detection rate and accuracy of the intrusion detection system are increased with minimum processing and execution times.[33] Energy aware clustering in WSNs based on fuzzy logic improves the formation of clusters and CH selection. This technique uses a neural network to have an optimum training set of energy and density values at all sensor nodes to compute the expected energy for uncertain CHs.[34] “Optimal Clustering in Circular Networks” is proposed where a single-hop communication between the sink and CHs is replaced with optimal multi-hops to enhance the WSNs lifetime by reducing the energy consumption.[35] The precision of the node positions is a very significant parameter for effective data transmission and to enhance the network lifetime in WSNs. Geographic routing based on weighted centroid localization improves the precision of node positions which uses fuzzy logic to compute the position of unknown nodes. This technique helps in the efficient selection of next-hop CH for minimizing the energy dissipation and enhances network lifetime.[36] An improved version of GWO which is based on the fitness value, ensures optimal distribution of CHs and enhances the chances of finding the optimal solution in GWO. To decrease the energy consumption, each sensor node’s communication distance is recomputed based on the distance between BS and CHs. This feature improves the WSN in terms of throughput, energy consumption and, stability by 57.8% compared to the LEACH protocol. The optimum method for selecting CHs and their distribution with balanced cluster structures can be achieved using fitness value based GWO. This optimal solution ensures that the node in a cluster located near to BS and having maximum energy becomes CH. To decrease the energy consumption, the communication range of sensor nodes is recalculated as per the distances between BS and CHs, it means that when a new CH is elected. This approach minimizes the average communication distance, reduces the energy consumption and improves WSN stability by 31.5% and 57.8% compared to SEP, LEACH protocols respectively.[37] IGWO algorithm benefits from DLH search strategy which is inherited from individual wolf’s hunting behavior. DLH uses dimension learning to collect neighbor’s information at every wolf and this information is shared among other wolves. This DLH search strategy maintains diversity as well as improves balancing between global and local search to improve convergence speed. The results of the engineering experiments and statistical reports prove that the applicability, optimal performance, effectiveness, solution stability, convergence speed, robustness, convergence precision, and efficiency of the IGWO algorithm outperforms the other existing metaheuristic algorithms.[38,39]