A hybrid vertical mutation and self-adaptation based MOPSO

Abstract

Multi-Objective Particle Swarm Optimizers (MOPSOs) are often trapped in local optima, converge slowly, and need more function evaluations when applied to solve Multi-objective Optimization Problems (MOPs). A hybrid Vertical Mutation and self-Adaptation based MOPSO (VMAPSO) is proposed to overcome the disadvantages of existing MOPSOs. Firstly, a hybrid vertical mutation operator is carefully designed, which can escape local optima and conduct a local search by uniform distribution mutation and Gaussian distribution mutation, respectively. Secondly, the adaptation ratio models of two mutations are fully analyzed and compared. Thirdly, the velocity update equations proposed by Clerc are improved to reduce the randomness of MOPSOs, and ϵ -dominance based archive strategy is adopted in the proposed algorithm. Finally, the VMAPSO is tested on several classical MOP benchmark functions. The simulation results show that the VMAPSO can be used to solve both simple and complex MOPs and that the VMAPSO is superior to other MOPSOs in solving complex MOPs. In particular, the self-adaptation VMAPSO can be applied to problems that you have no knowledge about.