A many-objective whale optimization algorithm to perform robust distributed clustering in wireless sensor network

Abstract

The substantial increase in the usage of wireless sensor networks (WSNs) encourages to develop data clustering in event monitoring applications. Many centralized algorithms with single objective optimization are employed to solve this problem. However privacy, security and technical constraints are key issues in traditional centralized approach. Moreover, many WSN applications like condition monitoring and target tracking require more than three objectives for effective partitioning of dataset. This paper proposes many-objective whale optimization algorithm to handle robust distributed clustering in WSN. Initially, a swarm based many-objective whale optimization (MaOWOA) is discussed where reference point based leader selection method is utilized in updating the solutions instead of grid based leader selection as in multi-objective approach. This method gives better convergence and diversity. The simulation result of proposed approach is evaluated on many-objective DTLZ test problems against existing many-objective methods which is faster in terms of simulation time and gives competitive results in terms of generational distance (GD), inverse generational distance (IGD), spacing (SP) and hyper volume difference (HVD). Further, the encouraging results of the proposed MaOWOA are applied to perform robust distributed clustering in WSNs which is termed as distributed many-objective clustering using whale optimization algorithm (DMaOWOA). In this approach, a weight based method is incorporated to detect and remove the outliers and diffusion method of cooperation is used for distributed clustering. The proposed DMaOWOA is tested on one synthetic and three practical WSN datasets. It is observed that DMaOWOA based clustering performs up to 6% and 8% improvement in terms of Silhouette index as compared to particle swarm optimization based many-objective distributed clustering (DMaOPSO) and distributed K-Means (DK-Means) clustering algorithm, respectively.