一种Multi-agent System的信任模型

Trust model of multi-agent system

Multi-agent技术已广泛用于大型分布式管理系统的开发.多agent间合作的信任问题会严重影响系统的效率.现有的信任模型基本上都是基于WEB的peer-to-peer环境给出的,它们不太适于multi_agent system(MAS)环境.针对MAS环境中信任的特点,提出了一种基于信誉和关系网的agent信任体系.分析实验表明该模型能促进agent关系网的凝聚性,能较大的提高agent合作伙伴选取的效率,并在重复交往环境下agent的合作成功率,以及在抑制恶意推荐、协同作弊、偏心这些干扰现象上较已有模型有较大改进.该模型在供应链管理和电子商务合作伙伴选择方面,具有一定的理论意义和实用价值.

The technique of muhi-agent has been largely used in the development of large-scale and distributed management systems. Trust between agents in multi-agent cooperation can severely affect the efficiency of multiagent system （MAS）. Most trust models presented are based on the Internet peer-to-peer environment. They do not fit for the environment of MAS. Considering the trust characteristics in MAS, this paper gives a novel agent trust model in MAS which is based on credit and relationship web of agent. Theoretical analysis and simulations show： this model can improve the agglomeration of the agent relationship web, and can improve the efficiency of the selecting of cooperative agent; furthermore, compared with the current global model, in agent successful cooperation ratio in the environment which needs agents repeatedly cooperating with each other, in controlling the agent malicious recommendation, the agent cooperative cheating, and in the partiality for the new agent when an agent selects its cooperative partner, this model is more compete. It has both theoretical and practical value on the copartner＇s selection in supply chain management system and e-commerce system.