A Multi-population Evolutionary Algorithm Based on Strong Coevolution for Constrained Multi-objective Optimization

When using evolutionary algorithms to handle constrained multi-objective optimization problems (CMOPs), it is of significance to balance objective optimization and constraints satisfaction, which poses a severe challenge for solvers. As a remedy for this issue, this paper proposes a multi-population evolutionary algorithm based on strong coevolution for CMOPs. Specifically, the proposed method consists of one main population and two auxiliary populations. The main population takes constraints into account to find feasible Pareto optimal solutions to guarantee the feasibility. The first auxiliary population is used to preserve solutions with superior objective function values, which is beneficial to break through infeasible barriers. Meanwhile, a mating selection strategy is designed to coordinate the interaction between main population and the first auxiliary population. Furthermore, the second auxiliary population evolve towards CPF from the infeasible sides by employing the improved method, aiming to provide some promising search directions. Similarly, another mating selection strategy is designed to coordinate the information exchange between the main population and the second auxiliary one. The effectiveness of proposed algorithm is demonstrated on 47 benchmark CMOPs compared with four state-of-the-art methods.