A Particle Swarm Optimizer with Lifespan for Global Optimization on Multimodal Functions

Abstract

The particle swarm optimizer (PSO) is a popular computing technique of swarm intelligence, known for its fast convergence speed and easy implementation. All the particles in the traditional PSO must learn from the best-so-far solution, which makes the best solution the leader of the swarm. This paper proposes a variation of the traditional PSO, named the PSO with lifespan (LS-PSO), in which the lifespan of the leader is adjusted according to its power of leading the swarm towards better solutions. When the lifespan is exhausted, a new solution is produced and it will conditionally replace the original leader depending on its leading power. Experiments on six benchmark multimodal functions show that the proposed algorithm can significantly improve the performance of the traditional PSO.