摘要
提出了一种基于智能体的多目标社会进化算法用以求解多目标优化问题(multiobjective optimization problems,简称MOPs),通过多智能体进化的思想来完成Pareto解集的寻优过程.该方法定义可信任度来表示智能体间的历史活动信息,并据此确定智能体的邻域、控制智能体间的行为.针对多目标问题的特点,设计了3个进化算子分别体现适者生存、弱肉强食、多样性原则以及自学习的特性.同时采用擂台赛法则构造Pareto解的存储种群.仿真实验结果表明,该算法能够较好地收敛到Pareto最优解集上,并且具有良好的多样性.另外,通过对智能体局部邻域环境建立方式的分析结果表明引入"关系网模型"可有效提高算法的收敛速度,并能在一定程度上提高解的质量.
In this paper, a multi-agent social evolutionary algorithm is proposed for multiobjective optimization problems. It completes the search process by the agent evolution. MOMASEA (multi-agent social evolutionary algorithm for multiobjective) defines the trust degree to denote the historical information of agents, and the neighborhood of agent is confirmed by it. According to the characteristic of multiobjective problems, three evolutionary operators are designed to complete the whole evolutionary process. The experimental results show that MOMASEA has a good convergence to the Pareto set. Furthermore, the analysis of the mode for instructs local environment verified that importing acquaintance net model can speed up the convergence effectively.
出处
《软件学报》
EI
CSCD
北大核心
2009年第7期1703-1713,共11页
Journal of Software
基金
国家自然科学基金Nos.60703107
60703108
60703109
60702062
国家高技术研究发展计划(863)No.2006AA01Z107
国家教育部博士点基金No.20060701007
国家重点基础研究发展计划(973)No.2006CB705700
国家教育部长江学者和创新团队支持计划No.IRT0645
陕西省自然科学基金No.2007F32~~
关键词
多目标优化
多智能体系统
关系网模型
可信任度
擂台赛法则
multiobjective optimization
multi-agent system
acquaintance net
trust degree
arena's principle