Dual-population two-archive evolutionary framework for constrained multi-objective optimization

Constrained multi-objective optimization problems (CMOPs) constitute a prevalent and ubiquitous class of optimization challenges that are frequently encountered across diverse field within science and engineering. To solve the complementary multi-objective optimization problem with narrow and disconnected feasible regions, dual-population two-archive evolutionary framework for constrained multi-objective optimization with constrained-archive solution phase-transition and auxiliary-population environment selection pause-termination (CAE_2SP) is proposed. The algorithm uses dual-population with different efficacy and two archives with different functions. To improve the problem of lower population diversity, constrained-archive solution phase-transition strategy is proposed. In this strategy, the diversity of solutions is emphasized in the early generation, so the non-dominated infeasible solutions generated by the evolution of main population are stored in the archive. In the late generation, the feasibility of solutions is taken into account, hence, constrained archive is transformed into storing non-dominated feasible solutions. In addition, this paper puts forward auxiliary-population environment selection pause-termination strategy. In this strategy, auxiliary population stop updating in the late generation and uses the optimal population information in the early generation to guide the evolution, to reduce the consumption of computing resources in the late generation and provide more computing resources for main population to help it search for potential feasible regions. The experimental results of nine comparison algorithms in three benchmark function suites demonstrate that CAE_2SP has superior performance in solving CMOPs compared with others. To validate the applicability of the proposed algorithm in solving practical problems, six real-world problems are employed for testing. The experimental results demonstrate that CAE_2SP exhibits competitive performance in addressing practical issues.