A Hybrid Response Strategy for Dynamic Constrained Multi-objective Optimization

Multi-objective optimization problems are widely present in the real world. When the objective functions of these problems change over time and are subject to constrained limitations, they become dynamic constrained multi-objective optimization problems (DCMOPs). As the problems become more complex, multi-objective optimization algorithms face greater challenges. In this paper, a hybrid response strategy for dynamic constrained multi-objective optimization algorithm (HRDC) is proposed to address these challenges. Specifically, a classical constraint coevolutionary framework (CCMO) is used to handle constraints. In addition, two independent response strategies are included. The first strategy is a region-based sampling strategy, which analyzes the previous environment’s optimal population, divides the decision space into uniform grids, and performs random sampling within each grid to improve the distribution of the initial population in the new environment. The second strategy is the classification prediction strategy. Firstly, the CCMO is employed to obtain a feasible priority population and an unconstrained population. Then, prediction is performed separately on each population, with the unconstrained population assisting evolution from the perspective of infeasible solutions. The effectiveness of the algorithm is validated through two sets of test instances. The experimental results demonstrate that compared to several state-of-the-art algorithms, HRDC exhibits strong competitiveness in handling DCMOPs.