关键词: 多目标进化算法, NSGA-II, 外部归档集, 归档策略

Abstract: When calculating the crowding coefficient, NSGA-II does not consider the individual distribution of the parent population and offspring population, so that some individuals with excellent distribution in the global space are eliminated. Aiming at the poor distribution of NSGA-II convergence results, an improved algorithm（UAA-NSGA-II） is proposed. In the iterative process, the non-dominated individuals of a certain quantum generation are randomly filled into the external archive, and the crowding coefficient operator is used to prune the archive. At the same time, to solve the problem that binary encoding is trapped in local optimum, the search speed and width in solution space are enhanced by Gray code and dynamic mutation operator. Tests are performed on ZDT series problems and compared with two typical algorithms and three improved NSGA-II algorithms. The results show that UEA-NSGA-II algorithm is superior to the contrast algorithm in terms of stability and optimization effect.

Key words: multi-objective evolutionary algorithm, NSGA-II, external archive, archiving strategy