When solving complex constrained multi-objective optimization problems (CMOPs), it's often a struggle to obtain a complete constrained Pareto front (CPF). To directly address the problems, a novel space division constrained multi-objective optimizer (SDCMO) was developed. SDCMO employs a space division mechanism that comprising uniform and non-uniform division strategies to partition the objective space into multiple subspaces. Within each subspace, only the optimal solution is preserved for subsequent evolution. Additionally, to strengthen the algorithm's capacity to traverse infeasible regions and further balance optimization objectives with constraint satisfaction, SDCMO adopts a multi-population multi-stage framework. During the convergence phase, to prevent convergence stagnation caused by premature convergence, SDCMO employs a multipopulation hybrid expansion mechanism. This enlarges the candidate solution pool, enhances informational diversity, and facilitates acquisition of the complete CPF. Experimental results on three CMOP test suites demonstrate SDCMO's competitive performance against state-of-the-art constrained multi-objective optimizers.
Huang et al. (Thu,) studied this question.