A decomposition-based evolutionary algorithm with multiple reference points strategy for multiobjective optimization

Many real-world optimization problems, including engineering design, can be formulated as multiobjective optimization problems (MOPs) that require finding approximate Pareto optimal fronts (POFs). Decomposition-based evolutionary algorithms have received considerable attention as promising approaches for solving MOPs. However, most existing algorithms utilize the geometric structure of a single point and multiple directions to guide the evolutionary search, which limits their success in dealing with MOPs with irregular POFs. To overcome this limitation, this paper proposes an effective multiobjective evolutionary algorithm that leverages the geometric pattern of multiple reference points and a single direction, thereby preventing solutions from focusing on the same region of the POF to some extent. The algorithm is configured with a multiple reference points strategy that includes the generation and adjustment of reference points. The proposed algorithm is compared with existing state-of-the-art multiobjective evolutionary algorithms on benchmark MOPs with different types of POFs and four real-world MOPs. The experimental results demonstrate the effectiveness of the proposed algorithm.