基于改进QPSO的单任务Agent联盟生成  

Improved QPSO to solve single task Agent coalition

作　　者：徐桓[1] 孙瑜[1] 

机构地区：[1]云南师范大学信息学院,昆明650500

出　　处：《软件》2015年第2期1-4,共4页Software

基　　金：国家自然科学基金项目(No.60903131);云南省自然科学基金项目(No.2009ZC0052M;2012FB144);云南省高校教育智能信息化科技创新团队项目;云南省中青年学术技术带头人后备人才项目

摘　　要：针对群智能算法求解Agent联盟生成问题过程中易出现粒子过分聚集,导致多样性降低,甚至陷入局部最优现象提出一种基于改进量子粒子群的求解策略,在粒子过分聚集时借鉴实数编码遗传算法中的柯西变异使粒子聚集程度降低,进而维持了粒子的多样性。并采用多种群并行和最优粒子移民策略加快算法收敛。实验表明,该策略可以快速高效的求解Agent联盟,在运行效率上优于同类方法。For quantum particle swarm optimizationalgorithm to solve the Agent coalition generation problems particles which were liable to occur excessive accumulation, results in the decrease of diversity, which may be trapped in local optimal phenomenon, an improved strategy in particle gather too much reference to the Cauchy mutation in the real-coded genetic algorithm make the particles aggregation degree is reduced, thus maintaining the diversity of the particles.Using multiple populations in parallel and the optimal particle migration strategy to speed up the conver-gence.Experiments show that the algorithm can quickly efficient Agent union, on the operation efficiency is better than that of similar algorithms.

关 键 词：AGENT联盟 量子粒子群 多AGENT系统 柯西变异 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]