A two-stage differential evolution for constrained multi-objective optimization problems
Erping Song, Guodong Han, Guifang Zhang · 2022
The constrained multi-objective optimization problems (CMOPs) are always hard to handle when the constraints are too stringent. In this paper, a two-stage differential evolution algorithm (TSCMODE) is proposed to solve CMOPs, and non-dominated sorting is used to select offspring individuals. Firstly, in the early stage of evolution, a efficient constraint handing technique is proposed to generate more feasible solutions, and differential mutation is used to generate high quality offspring individuals. Then, at a later stage in evolution, the feasibility conditions are used to divide the population into upper and lower sub-populations, the good genes from the upper sub-population are passed to the lower offspring individuals by differential evolution. Besides, the objective values of the lower sub-population are adjusted self-adaptively to avoid the potential infeasibility being discarded. Finally, the proposed algorithm is executed on some recent benchmark functions and compared with four state-of-the art constrained multi-objective evolutionary algorithms. The experimental results show that TSCMODE can solve the CMOPs as well as.