An efficient multi-objective optimization algorithm based on level swarm optimizer

Abstract

In the past few decades, evolutionary multi-objective optimization has become a research hotspot in the field of evolutionary computing, and a large number of multi-objective evolutionary algorithms (MOEAs) have been proposed. However, MOEA is still faced with the problem that the diversity and convergence of non-dominated solutions are difficult to balance. To address these problems, an efficient multi-objective optimization algorithm based on level swarm optimizer (EMOSO) is proposed in this paper. In EMOSO, a sorting method is introduced to balance the diversity and convergence of non-dominated solutions in the whole population, which is based on non-dominated relationship and density estimation. Meanwhile, a level-based learning strategy is introduced to maintain the search for non-dominated solutions. Finally, DTLZ, ZDT and WFG series problems are utilized to verify the performance of the proposed EMOSO. Experimental results and statistical analysis indicate that EMOSO has competitive performance compared with 6 popular MOEAs. The source code of EMOSO is provided at: https://github.com/xuweizhang163/EMOSO.