多Agent主从粒子群分布式计算框架  

Multi-Agent Based Distributed Computing Framework for Master-Slave Particle Swarms

作　　者：郑宇军[1] 陈胜勇[1] 凌海风[2] 徐新黎[1] 

机构地区：[1]浙江工业大学计算机科学与技术学院,浙江杭州310023 [2]解放军理工大学机械工程系,江苏南京210007

出　　处：《软件学报》2012年第11期3000-3008,共9页Journal of Software

基　　金：国家自然科学基金(61105073;61173096;61103140;61020106009;61070043);浙江省自然科学基金(R1110679)

摘　　要：面向大规模复杂优化问题,提出了一个基于并行粒子群优化的分布式Agent计算框架.框架中使用一个主群(master swarm)来演化问题的完整解,并使用一组从群(slave swarm)来并行优化一组子问题的解,主群和从群通过交替执行来提高问题的求解效率.采用异步组结构,主群/从群中的各类Agent共享一个解群,并通过相互协作,对解群进行构造、改进、修补、分解和合并等演化操作.该框架可用于求解复杂的约束多目标优化问题.通过一类典型运输问题上的实验,其结果表明,所提出的方法明显优于另外两种先进的演化算法.To effectively solve large-scale optimization problems, the paper proposes a distributed agent computing framework based on the parallel particle swarm optimization （PSO）. The framework uses a master swarm for evolving complete solutions of the problem, and uses a set of slave swarms for evolving sub-solutions of the subproblems concurrently. The master swarm and slave swarms alternatively implement the PSO procedure to improve the problem-solving efficiency. Using the asynchronous team based agent architecture, a master/slave swarm consists of different kinds of agents, which share a population of solutions and cooperate to evolve the population, such as initializing solutions, moving particles, handling constraints, and decomposing/synthesizing sub-solutions. The framework can be used to solve complicated constained and multiobjective optimization problems efficiently. Experimental results demonstrate that this approach has significant performance advantage over two other state-of-the-art algorithms on a typical transportation problem.

关 键 词：AGENT 粒子群优化 主从模型 协同进化 分布式计算 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]