An Archive-assisted Two-stage and Dual-population Evolutionary Algorithm for Constrained Multi-Objective Optimization Problems

The key to solving constrained multiobjective optimization problems (CMOPs) is to balance objective optimization and constraint satisfaction. But the existing evolutionary algorithms suffer from great difficulties in striking a balance between them, especially in dealing with CMOPs with complex constraints. To solve the issue, an archive-assisted two-stage and dualpopulation evolutionary algorithm, referred as ATD-CMOEA, is proposed. The aim of Stage 1 is to explore the entire objective space and discover the promising regions, which is achieved by two populations with different priorities of constraints and objectives. Meanwhile, feasible solutions discovered during the evolutionary process are archived in normal archive. The purpose of Stage 2 is to get a series of well-distributed Pareto optimal solutions by the simultaneous evolution of the main population, the assistant population and normal archive. Experimental results on two test suites demonstrate the competitiveness of the proposed ATD-CMOEA when compared with other seven state-of-art algorithms.