Multiregional co-evolutionary algorithm for dynamic multiobjective optimization

Abstract

Dynamic multiobjective optimization problems (DMOPs) require Evolutionary algorithms (EAs) to track the time-dependent Pareto-optimal front (PF) or Pareto-optimal set (PS), and provide diversified solutions. Thus, a multiregional co-evolutionary dynamic multiobjective optimization algorithm (MRCDMO) is proposed based on the combination of a multiregional prediction strategy (MRP) and a multiregional diversity maintenance mechanism (MRDM). To accurately predict the moving trend of PS, a series of center points in different subregions is used to build a difference model to estimate the new location of center points when an environmental change is detected. To promote the diversity of the population, some diverse individuals are generated within the subregion of the next predicted PS. These two parts of solutions make up the population under a new environment. The performance of our proposed method is validated by comparison with four state-of-the-art EAs on 12 test functions. Experimental results demonstrate that the proposed algorithm can effectively cover the changing PF and efficiently predict the location of the moving PS.