Vector evaluated particle swarm optimization exploration behavior part I: Explorative analysis

Abstract

An explorative analysis in low dimensional objective space of the vector evaluated particle swarm optimization (VEPSO) algorithm is presented. Results indicate that the VEPSO algorithm continues to explore, and does not focus enough on exploitation. The resulting Pareto optimal fronts (POFs) have a poor spread and in some cases provide a poor estimate of the real POFs. It is hypothesized that the poor results can be attributed to the fact that the VEPSO algorithm is not exploiting enough. The multi guided VEPSO with random archive selection (MGVEPSOa) is introduced to address the lack of exploitation of the VEPSO algorithm. Results indicate that the MGVEPSOa algorithm outperforms the VEPSO algorithm leading to better POFs. Analysis of the candidate solutions of the MGVEPSO a algorithm confirm that the algorithm exploits existing solutions more than the VEPSO algorithm. The improved performance is attributed to the increased exploitation confirming the hypothesis that VEPSO does not exploit enough.