Tradeoff strategy between exploration and exploitation for PSO

Abstract

Particle Swarm Optimization (PSO) is a class of stochastic search algorithms based on population. Due to the simplicity of implementation and promising optimization capability, PSO is successfully applied to solving a wide class of scientific and engineering optimization problems. However, PSO has some drawbacks such as high computational complexity and premature convergence. Inspired by the tradeoff strategy between exploration and exploitation in reinforcement learning, we propose an improved PSO. The sigmoid function is incorporated into the velocity update equation of PSO to tackle these drawbacks of PSO. The comparison with inertia weight PSO, constriction factor PSO and Tribe PSO using classic benchmark functions demonstrates that our approach achieves a good tradeoff between exploration and exploitation, and thus obtain better global optimization result and faster convergence speed.