Offspring regeneration driven by finite element mapping for large-scale evolutionary multiobjective optimization

In large-scale evolutionary multiobjective optimization, the size of the search space expands exponentially as the number of decision variables increases, which makes it more difficult to find an optimal search region and generate promising offspring. For this purpose, this paper proposes an offspring regeneration-based algorithm driven by finite element mapping, which can be divided into two stages. The first stage uses a competitive swarm optimizer to perform a pre-section of the initial population and then maps each individual into a location in a finite element region. During evolution, the population searches along these locations. The search region changes from an infinite number of points on the search space to a finite number of points, thus accelerating the convergence to the optimal region. In the second stage, a dual sampling strategy that includes convergence-based sampling and diversity-based sampling is proposed to generate more promising offspring, which balances the exploration and exploitation of the algorithm. By comparing it with five state-of -the-art large-scale multiobjective evolutionary algorithms. Experimental results demonstrate that the proposed algorithm significantly outperforms the compared algorithms on LSMOPs problems with up to 5000 decision variables and achieves satisfactory results on real-world problems.