多Agent分布仿真中的环境划分模型研究.pdf

ISSN 11．1777／TP 计算机研究与发展 1000．1239[CN Researchand Journalof Development Computer 多Agent分布仿真中的环境划分模型 叶超群1 马 民1 李 宝1 金士尧1 曹贺锋2 t(国防科学技术大学并行与分布处理国家重点实验室长沙410073) 2(武警总部通信部北京100089) (cqyie@163．COWl) AnEnvironment Modelfor DistributedSimulation PartitioningMulti-Agent Ye Ba01，Jin Cao Minl，Li Shiya01，andHefen92 Chaoqunl，Ma ParallelDistributed 1(National for Proc部sing，NationalUniversityofDefense 410073) Laboratory Technology，Changsha 2(Communication Police 100089) Department，ArmedHeadquarters，BeOing Abstract distributedenvironmentmodelcan the of distributed Utilizing improvescalabilitymulti—agent distributed simulation．Inthis the of the environmentmodeltobe paperproblemoptimallypartitioning executed isaddressed．Thetotalcostofthe amongmultiplecomputers partitioning，includingcomputational costandcommunicationtobeminimized．Itisshownthatthis isNP—hardandhasno cost，has problem time withfinite factorunlessP≠NP．An algorithm polynomialapproximation approximation is anditstime is showsthe of approximationalgorithmproposed complexityanalyzed．Evaluationefficiency the approximationalgorithm． words Key multi—agentsimulation；distributedsimulation；environment algorithm 摘要在多Agent分布仿真中，采用分布式环境模型有利于提高系统的扩展性．把分布式环境模型划 分到多机上执行需要考虑计算和通信两方面的代价，求解总代价最小的划分方案的问题是NP难的，且 当P≠NP时不存在具有有限近似比率的多项式时间复杂性近似算法．提出了一种基于准贪心策略和