冲突环境下中立Agent群体趋向分析与预测  

Analysis and prediction of behavioral trends in groups of Agents in conflict

作　　者：叶小平[1] 胡苏[1] 

机构地区：[1]中山大学计算机科学系,广东广州510275

出　　处：《智能系统学报》2007年第4期28-33,共6页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金资助项目(60673135);广东省自然科学基金资助项目(05003348)

摘　　要：分布式智能系统基本研究领域之一就是多Agent系统.在MAS协作过程中冲突难以避免,在实际应用的许多情况下,冲突都是在群体层面展开,同时冲突双方都需要特别关注持中立态度Agent的未来发展导向,这在双方势均力敌情况下尤为重要.因此,根据已有信息,建立基于中立Agent群体的智能分析与计算模型,对中立群体可能的趋向进行必要评估和预测,在冲突研究与应用当中具有基本意义.通过对MAS中Agent相互关系的形式描述,引入中立Agent群体模型,建立了对中立Agent未来状态进行智能分析和预测的计算模式,并设计了相应算法和实验.理论分析和实验结果表明了所建立模型的合理性与有效性.One of the basic research fields in distributed intelligence is the multi-Agent system （MAS）. During cooperative interaction of agents in MAS, it is difficult to avoid conflicts among agents. These conflicts take place at the group level. Both sides of a conflict must pay special attention to neutral Agents. Trends in their behaviors are particularly important if they are potential competitors with similar strengths. This paper establishes a computer model of neutral Agents based on group affinity. The model can evaluate and predict the probable tendencies of members of neutral groups. We proposed a formal description of the relationships among all Agents in the MAS, and then built a model of agent groups as well as a predictive and analytical model which may help anticipate future behavior of the neutral agents. To test this, an algorithm and experiment were designed. Both the theoretical analysis and experimental resuits prove the validity of the model.

关 键 词：冲突关系 群体冲突 中立状态Agent群体模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]