摘要
NSGA-Ⅱ-DE算法是在NSGA-Ⅱ算法的基础上利用DE算法的收敛速度快、鲁棒性高的特性得到的改进算法,该算法提高了原算法的收敛速度,同时也降低了原算法对参数的依赖性.然而,原算法的解群分布性却没有得到提高.鉴于此,提出一种基于种群扩张与稀疏化策略的改进型NSGA-Ⅱ-DE算法.该算法利用种群扩张增加候选解的数量,再利用稀疏化策略从候选解中选出使得整体分布尽可能均匀的最优解.种群扩张通过在进化最后的若干代保留每代中的第一非支配面上的个体来实现.在迭代结束后,对种群进行非支配排序,去除第一非支配面以外的个体,以提高解群质量.进行稀疏化处理,即对扩张后的全部个体按目标向量的某一维度排序,再筛选出相邻间距最接近期望距离的个体,以达到改善解群分布性的目的.仿真实验表明,所提出的算法在改善原算法的解群分布性上表现优异,但算法的时间和空间复杂度较原算法有所增加.
The NSGA-Ⅱ-DE algorithm improves the convergence speed and reduces the dependency on the parameters of NSGA-Ⅱ by taking advantage of the fast convergence and high robustness of the DE algorithm. However, the solutions distribution of the original algorithm is not improved. In this study, an advanced NSGA-Ⅱ-DE algorithm based on the population expansion and sparsity strategy is proposed. This algorithm increases the number of candidate solutions by using population expansion operation, and the sparsity strategy is used to get optimum solutions from candidates that make the global distribution as uniform as possible. Population expansion is achieved by preserving individuals on the first non-dominated surface of each generation in the last few generations of evolution. At the end of the iteration, the population is non-dominated sorted, and the individuals outside the first non-dominating surface are removed to improve the quality of the solutions. Then, the sparseness is processed, that is, all the individuals after expansion are sorted according to a certain dimension of the target vector, and the individuals whose adjacent distance closest to the expected distance are selected, so as to improve the distribution of the solutions. The simulation results show that the proposed algorithm is superior to the original algorithm in terms of the distribution of solutions, but inferior to the original algorithm in terms of the time and space complexity.
作者
蒋永华
许妙忠
成刚
JIANG Yong-hua;XU Miao-zhong;CHENG Gang(School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430070,China;State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing,Wuhan University,Wuhan 430070,China)
出处
《控制与决策》
EI
CSCD
北大核心
2019年第1期55-62,共8页
Control and Decision
