Solving constrained multi-objective optimization problems (CMOPs) is important in many engineering and scientific domains. However, it remains a challenging task especially when the feasible region is discontinuous, narrow, or structurally complex. Although multi-population constrained multi-objective evolutionary algorithms have been proposed to address such difficulties through diversity enhancement, their performance is often limited by imprecise population partitioning and insufficient coordination among sub-populations. Inspired by the concept of ecological niches, this paper proposes a niche-based multi-objective evolutionary algorithm (NMOEA). The algorithm incorporates two key strategies: (1) a dynamic niche formation mechanism, which partitions the population into sub-groups according to spatial distribution and constraint satisfaction levels, and (2) an adaptive interaction mechanism that facilitates efficient knowledge transfer among niches. This combined framework supports a balanced and well-coordinated search across the entire feasible space. Comprehensive experiments on widely-used CMOP benchmarks and a real-world power dispatch problem validate the effectiveness of NMOEA. The results show that NMOEA outperforms several state-of-the-art constrained multi-objective evolutionary algorithms, achieving better convergence to the true Pareto front while maintaining superior diversity.
Yang et al. (Mon,) studied this question.