An improved particle swarm optimizer with attraction and repulsion

The particle swarm optimization (PSO) is a population-based strategy for global optimization. A rapid decrease of diversity in the iterative procedure leads the PSO to suffer from premature convergence on multimodal problems. This paper presents a novel variant of the PSO (PSO-CAR). It uses an animal foraging strategy with attraction and repulsion. This strategy guarantees a high diversity of the swarm to protect the PSO from premature convergence. The basic PSO (bPSO), genetic algorithm (GA) and two other variants of the original PSO are employed here to evaluate the effectiveness of the PSO-CAR using 6 standard numerical functions. Simulation results reveal that the proposed PSO-CAR outperforms the other approaches in solving multimodal problems.