Multi-Objective Particle Swarm Optimization Based on Gaussian Sampling

Abstract

This paper proposes a multi-objective particle swarm optimization algorithm based on Gaussian sampling (GS-MOPSO) to locate multiple Pareto optimal solutions for solving multimodal multi-objective problems. In the proposed method, the Gaussian sampling mechanism is used to form multiple neighborhoods by learning from optimal information of particles. And particles search their own neighborhoods to obtain more optimal solutions in the decision space. Moreover, an external archive maintenance strategy is proposed which allows the algorithm to maintain an archive containing better distribution and diversity of solutions. Meanwhile, nine new multimodal multi-objective test problems are designed to evaluate the performance of algorithms. The performance of GS-MOPSO is compared with twelve state-of-the-art multi-objective optimization algorithms on forty test problems. The experimental results show that the proposed algorithm is able to handle the multimodal multi-objective problems in terms of finding more and well-distributed Pareto solutions. In addition, the effectiveness of the proposed algorithm is further demonstrated in a real-world problem.

Keywords

• Particle swarm optimization
• evolutionary algorithm
• multimodal
• multi-objective optimization