An Adaptive Heuristic‐Information‐Guided Co‐Evolutionary Algorithm for Constrained Multiobjective Optimisation

ABSTRACT With the rising complexity of practical application scenarios, conventional evolutionary algorithms face great challenges in handling constrained optimisation problems. These difficulties include premature convergence to local optima, subpar solution quality and excessive computational overhead; in particular, acquiring Pareto optimal solutions for multiobjective optimisation problems has become increasingly arduous. Due to the excellent adaptive search performance, the single‐objective covariance matrix adaptation evolution strategy (CMA‐ES) has gained recognition as an effective algorithm to address continuous optimisation problems. However, when extended to constrained multiobjective optimisation scenarios, CMA‐ES still reveals its shortcomings including high computational complexity and inefficient constraint handling. To address these issues, this paper proposes a co‐evolutionary algorithm with three primary contributions: First, an adaptive co‐evolutionary framework is developed, which dynamically selects crossover operators based on individual information at different evolution stages, thereby rationally allocating computational resources and balancing global exploration and local exploitation. Then, a CMA‐ES operator tailored for constrained multiobjective problems is designed, incorporating an adaptive step size and a multicriteria acceptance criterion. In addition, a dynamic elite selection strategy is presented, which uses densities in both the objective space and decision space, and leverages these metrics according to the evolutionary stage to preserve population diversity. A variety of comparative tests are conducted on standard benchmarks and real‐world engineering cases. The results verify the superiority of the proposed method over seven comparative algorithms.