Abstract: This paper analyzes the characteristics and shortcomings of Multi-Objective Evolutionary Algorithm(MOEA) methods, and introduces a method to evaluate the diversity of non-dominated solutions in new coordinates, which avoids the influence because of distinct extent of convergence to the diversity evaluation, and the new objective space is divided into some equal regions. The solution is evaluated by using the number of individual in these regions. Experimental results and theoretical analysis show it can accurately evaluate the solution set of distribution.

Key words: Multi-Objective Evolutionary Algorithm(MOEA), diversity metric method, coordinate transformation

摘要： 分析现存多目标进化算法分布度评价方法的特点和不足，提出一种在新的坐标下对解集进行分布度评价的方法。该方法把直角坐标系下的解集映射到另一个基于角度的坐标下，以避免算法因收敛性不同对分布性评价造成影响，把新的坐标空间划分成若干相等的区域，利用区域内的个体数评价解集的均匀性。理论分析与实验结果证明该方法能精确地评价解集的分布情况。

关键词: 多目标进化算法, 分布度评价方法, 坐标变换

CLC Number: 

• TP18