A Reference Point and Multi-direction Search Based Evolution Algorithm for Large-scale Multi-objective Optimization

This paper proposes a new algorithm based on a reference point selection mechanism and a multi-direction search strategy for large-scale multi-objective optimization problems. Firstly, a center point symmetry strategy is designed to select uniformly distributed reference points and transform the original problem into several low-dimensional single-objective optimization problems. Based on the reference points, a multi-directional weight variable association strategy is proposed to add search directions for the original problem and to improve the search ability of the algorithm. Then, to solve the transformed single-objective problem effectively, an improved differential evolution algorithm based on center mutation is presented. Finally, the numerical experiments are conducted on the large-scale optimization problem benchmarks LSMOP with 200, 500, and 1000 decision variables and the comparison of the proposed algorithm with four state-of-the-art algorithms is made. The results show that the proposed algorithm significantly outperforms the compared algorithms.