基于多种群进化算法的多目标并行博弈设计  

MULTI-OBJECTIVE PARALLEL GAME DESIGN BASING ON MULTI-SPECIES

作　　者：谢能刚[1] 潘创业[1] 李锐[1] 王璐[1] 

机构地区：[1]安徽工业大学机械工程学院,安徽马鞍山243002

出　　处：《数值计算与计算机应用》2010年第2期81-91,共11页Journal on Numerical Methods and Computer Applications

基　　金：教育部新世纪优秀人才计划(070003);教育部科学技术研究重点项目(207050);安徽省自然科学基金项目(070414174);安徽高校省级自然科学重点项目(2006kj001A)

摘　　要：针对多目标决策问题,提出一种基于多种群进化算法的多目标并行博弈设计方法,利用Fortran语言编制了相应计算程序.将多目标设计问题描述为博弈问题,在博弈分析中,根据各博弈方在博弈次序和过程上具有同时性和独立性,采用并行计算技术.在以MPI为平台的集群环境中实现数值算例、补偿滑轮组变幅机构、拱型结构的多目标并行博弈设计.计算分析结果显示:该方法的计算精度高、收敛速度快、可以防止早熟、有较好的平衡负载能力,可有效求解多目标问题.For design of multi-objective decision, a new method of multi-objective parallel game design basing on multi-species evolution algorithm has been raised and carried out by For- tran.By describing the issue about multi-objective design into game theory problem and un- der the analysis of the game, the paper adopts the parallel computing technologies according to the synchronism and independence of each player in the game sequence and the process. MPI as a platform in the COW, realized the multi-objective parallel game design of numer- ical count cases, luff mechanism of compensative sheave block and arch structure. Analysis shows parallel game basing on multi-species evolution algorithm can not only converge effi- ciently but prevent precocity; it also has a better capacity of loading balance, resulting in a better robustness. It can resolve the multi-objective decision effectively.

关 键 词：多目标 博弈 并行 多种群进化 NASH均衡 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] O225[自动化与计算机技术—控制科学与工程]