Multimodal Multiobjective Particle Swarm Optimization with Space Information Aggregation

Multimodal multiobjective particle swarm optimization conducts diversity-driven evolution within the decision space to obtain equivalent solutions. However, due to the inconsistency between the diversity in the decision and objective spaces, it is difficult for evolutionary directions to reconcile the performance in both spaces. Targeted to this issue, a multimodal multiobjective particle swarm optimization with space information aggregation (MMPSO-SIA) is proposed in this paper. First, an information aggregation is proposed based on the local sensitive hash. Then, the diversity information of particles in decision space and objective spaces can be integrated into diversity metrics comprehensively. Second, a direction evaluation strategy is developed to capture the potential solutions in both search spaces. Then, effective directions of potential solutions are supplemented to obtain evolutionary guidance for particle swarm. Third, a variable learning intensity is integrated into MMPSO-SIA. Then, the evolutionary direction provided by information aggregation methods can be efficiently utilized. Finally, the benchmark function test results prove the superiority of MMPSO-SIA over other multimodal multiobjective optimization methods.