基于聚集密度的约束多目标进化算法

Constrained Multi-objective Evolutionary Algorithm Based on Crowding Density

对基于群体聚类的约束多目标进化算法进行了改进,引入了聚集密度以度量群体中个体间的关系,保持种群的多样性。其基本思想为:首先将初始群体按多判据聚类方法分为适应度值不同的四类,然后计算类内群体中个体的聚集密度,根据适应度值和聚集密度定义一个偏序集,最后采用比例选择原则依次从偏序集中选择个体,更新精英集。通过数值实验用量化指标研究了改进算法的收敛性和分布性,结果表明:改进算法的收敛性与常规约束多目标进化算法相当,但分布性有了明显的提高。

The constrained multi- objective evolutionary algorithm based on group clustering was improved,and crowding- density was introduced to measure the relationship among individuals and maintain the diversity of population. The basic idea is that the initial population is divided into four groups with different fitness by multi- criterion clustering method,and the crowding- density of each group is calculated. A poset is defined according to the objective function value and crowding- density,and the individuals are selected from poset by the principle of proportion selection,then the elite set is updated. The convergence and distribution of improved algorithm were studied by means of numerical experiments,and the results showed that the convergence of improved algorithm is roughly equal to the conventional multi- objective evolutionary algorithm,but the distribution of improved algorithm is significantly improved.