Social class particle swarm optimization for variable-length Wireless Sensor Network Deployment

Abstract

Listen

Translate article

Wireless Sensor Network Deployment (WSND) is an active research topic. This mechanism involves optimal placement of a wireless sensor network in a 2D environment for optimizing a set of metrics, such as coverage and cost. The topic of WSND has two challenging parts. First, it has a Multiobjective Optimization (MOO) nature instead of a single optimal solution due to the existing set of nondominated solutions. Second, its Variable length (V-length) decision space obtains nonhomogeneous solutions in terms of length. These challenging concepts cause traditional MOO algorithms to become insufficient to solve WSND; thus, developing an MOO algorithm with a V-length nature is required. In this study, Social Class Multiobjective Particle Swarm Optimization (SC-MOPSO) was developed for solving difficult optimization problems with MOO and V-length nature. The algorithm extends the concept of social interaction of Particle Swarm Optimization by decomposing the solution space into classes on the basis of their dimension. Furthermore, it incorporates intra and inter class operators for assuring the required dynamics of solution changes to reach the Pareto front. A set of mathematical optimization problems with two and three objectives based on different dimensions of mathematical functions was tested for evaluation. In addition, SC-MOPSO and the benchmarks were evaluated for accomplishing WSND. Experimental results show that SC-MOPSO outperforms all benchmarks in terms of domination for WSND with maximum percentage of 100% for Weighted Sum Variable Length Particle Swarm Optimization (WS-VLPSO) and minimum percentage of 68% for Nondominated Sorting Genetic Algorithm (NSGA-II).