机构地区: 遵义师范学院数学系
出 处: 《计算机应用研究》 2011年第1期72-74,共3页
摘 要: 为了提升粒子群算法求解多目标问题的能力,通过分析初始种群的方法对算法的影响,提出一种基于正交设计的多目标粒子群算法(ODMOPSO)。在算法运行过程中,通过正交设计来产生初始种群,使得种群均匀分布在可行区域,进而使得算法能够在整个可行解空间上进行均匀搜索;同时,引入广义学习策略提升粒子向Pareto前沿飞行的概率。在基准函数的测试中,结果显示ODMOPSO算法获得了质量更高的解。 In order to improve the ability of PSO in solving multi-objective optmization, by discussing the relationship between method of initialization swarm and algorithm performance, this paper proposed a multi-objective PSO based on orthogonal design (ODMOPSO for short). In the proposed algorithm, firstly, used orthogonal design to initialize the population, which made the algorithm can search in the whole feasible space;next, introduced the comprehensive learning strategy to improve the probabili- ty of flying to Pareto front. In benchmark functions, simulation results show that the ODMOPSO algorithm can find better solutions.
领 域: [自动化与计算机技术] [自动化与计算机技术]