基于多种群进化算法的多目标并行博弈设计.pdfVIP

2010 年 6 月 数值计算与计算机应用 第 31 卷第 2 期 Jun. , 2010 Journal On Numerical Methods and Computer Applications Vo1.31 , No.2 基于多种群进化算法的多目标并行 博弈设计划 谢能刚潘创业李锐王璐 (安徽工业大学机械工程学院，安徽马鞍山 243∞2) 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奔设计方法，利用 Fortran 语言编制 f相向计算程序.将多目标设计问题描述为博弈问题，在博弈分析中，根据 各博奔方在博弈次序和过程上具有同时性和抽主性，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奔设计.计算 分析结果显示:该方法的计算精度高、收敛速度快、可以防止单熟、有较好的平衡负载能力， 可有放求解多目标问题. 关键词:多目标;博弈;并行;多种群进化; Nash 均衡 MR (2000) 主题分类: 9OC29 如IULTI-OBJECTIVE PARALLEL GAME DESIGN BASING ON 岛1ULTI-SPECIES Xie Nenggang Pan Chuangye Li Rui Wang Lu (AnHui Unit旷sity 01 Technol咽宫， Maanshan 243002, China) Abstract For desi胆 of multi-objective decision , a new method of multi-objective parallel game d回ign b咽ing on multi-sp回阳 evolution algorithm h眉 beenr凶sed and carried out by For- tran.By describing the 回ue about multi-objective d臼ign into game th四ry problem and un- der the analysÍ8 of 也，e game,the paper 时opts the p创allelωmputi吨technologies 配ωrding to the synchronism and independen四 of each pla;归r in the game sequence and the proc回s. MPI 副 a platform in the COW, realized the multi-objective parallel game design of nUmer- ical count c倒锢， luff mechanism of ωmpensati四 sheave block and 缸由 structure. Anal归is shows parallel g缸ne basing on multi-speci创 evolution algorithm can not only ∞nverge effi- ciently but pre四nt pr，配ocitYi it also has a. better cap配ity of loading balance , resulting in a better robustness. It can r，惕。lve the multi-objective decision effectively. Keywords: multi-objective; g，创ne; parallel; multi-species evolution algorithm; Nash