基于维数划分策略和免疫的多任务联盟并行生成算法  

Multi-task coalition parallel generation algorithm based on dimension partition strategy and immunity

作　　者：苏兆品[1] 蒋建国[1] 夏娜[1] 张国富[1] 

机构地区：[1]合肥工业大学计算机与信息学院

出　　处：《系统工程理论与实践》2008年第1期118-123,共6页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(60474035);国家教育部博士点基金(20060359004);安徽省自然科学基金(070412035)

摘　　要：设计了一种基于维数的Agent能力划分策略,提出"子Agent"概念;在此基础上设计了一种基于三维二进制编码的免疫算法求解多任务联盟并行生成问题,并对疫苗采取了自适应提取的策略.实验结果证明了该算法的有效性.Coalition generation, especially multi-task coalition parallel generation, is a key topic in Multi-Agent System. It mainly researches how to generate several optimal task-oriented coalitions parallel in dynamic manner. But existing researches are restricted in the condition that multi-task coalitions are generated serially and each Agent can only take part in a coalition. To solve the problem, an ability partition strategy based on dimension and a novel Child Agent are proposed to ensure that an agent can take part in several different coalitions synchronously. A novel three- dimensional binary encoding approach is designed to solve coalition parallel generation based on Immune Algorithm. And a novel method of vaccine adaptive obtaining is used to improve the searching effect of the Immune Algorithm. The experimental results show that the proposed algorithm is effective and can obtain a reasonable solution in an acceptable time.

关 键 词：并行生成 维数划分策略 子Agent 免疫算法 三维二进制编码 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]