Using relaxation velocity update strategy to improve particle swarm optimization

摘要

In particle swarm optimization (PSO), swarm intelligence is utilized when the velocities of particles are updated depending on their own experience and shared information, which is favorable for avoiding local optima. But frequently updating velocities weaken local exploitation abilities of particles and slow down convergence. In this paper, relaxation-velocity-update (RVU) strategy is incorporated into PSO algorithm to accelerate convergence. RVU strategy suggests that the velocity should be updated only when the particle cannot further improve the fitness with its previous velocity, rather than in every iteration. Standard linearly decreasing weight PSO (LDW-PSO) and LDW-PSO with RVU strategy (LDW-RVU-PSO) are compared on three well-known benchmark functions. The results show that RVU strategy significantly improves the convergence speed of LDW-PSO.