摘要： 针对各种进化算法在解决PS问题上表现出来的脆弱性，提出一种解决复杂PS问题的自适应多目标差分进化算法SA-MODE。根据随机选择的父个体X与当前种群中的个体Y的支配关系，通过改变缩放因子的大小来控制新个体和父个体的距离。当X支配Y则新个体接近X，反之远离X，当X与Y互相不支配则产生2个新个体，一个接近X一个远离X。实验结果表明，在处理复杂PS问题时，SA-MODE与GDE3和NSGA-II相比有更理想的效果。

关键词: 多目标优化问题, 多目标差分进化算法, 复杂Pareto解集问题, 变量变换, 变异算子

Abstract: After a deep analysis of the faults of traditional MOEAs on solving the complex Pareto Set(PS) problems, this paper proposes a multi-objective evolutionary algorithm on solving the complex PS problems(SA-MODE). According to the dominated relationship between individual X which is chosen in the parent population and the individual Y in the current population, control the distance between new individual and the parent one by altering the size of scaling factor. The new individual is close to X when X dominates Y, the other hand away from X. When the X and Y do not dominate each other then create two new individuals, one near the X and the other from X. Experimental results demonstrate that SA-MODE can deal with complex PS problems more effectively compared with GDE3 and NSGA-II.

Key words: multi-objective optimization problem, multi-objective differential evolutionary algorithm, complex Pareto Set(PS) problem, variable linkages, mutation operator

中图分类号: 

• TP393.02