New social-based radius particle swarm optimization

Abstract

Particle Swarm Optimization (PSO) is a swarm intelligence based algorithm to solve the optimization problem. The traditional PSO has disadvantage from the premature convergence when solving the complex optimization problem. We propose the Radius Particle Swarm Optimization (RPSO) which extends the PSO by regrouping the particles within the given radius of the circle. It initializes the group of particles, calculates the fitness function, and finds the best particle in that group. The RPSO takes advantage of group-swarm to maintain the swarm diversity and evolution by sharing information from the agent particles which effectively keep the balance between the global exploration and the local exploitation. We present a radius calibration to measure the search ability of each problem characteristic. The performance of the proposed algorithm is evaluated against ten well-known benchmark functions. In comparison against several PSO variants, the results demonstrate that the RPSO significantly outperform on the complex multimodal problems.