An efficient population diversity measure for improved particle swarm optimization algorithm

By the comparison study on several commonly used population diversity measures, we find that the conclusions derived from these diversity measures are not in keeping with the knowledge accepted by many scholars, and the relationship between population diversity and the evolutionary process of PSO is questionable. To deal with the above mentioned problems, a novel population diversity measure based on the optimal point is proposed in this paper, where the outliers are discarded, and the measure results are easy to understand the evolutionary state of the PSO algorithm. Furthermore, the optimal point-based diversity (O-diversity) measure is studied to control the population diversity and improve the optimization performance of PSO algorithm. The study results verify that the O-diversity measure is effective and useful to assess and control the population diversity, and it is meaningful to guide to better optimization performance of the advanced PSO algorithm.