摘要： 提出了一种MAS 环境下基于信用和关系网的Agent 信任体系，给出了模型的构建方法及相关的量化公式。该信任体系可以使系统中的推荐信任和直接信任等同看待，能抑制恶意推荐、协同作弊这些干扰现象，能促进关系网的凝聚性，简化Agent 合作伙伴的选取，从而促进系统中Agent 间的合作。

关键词: Multi-agent system；信任体系；信用；关系网；技术性贸易壁垒

Abstract: This paper presents a novel agent trust model in MAS which is based on credit and relation Web of agent. It gives the constructionmethods of the credit system, of the relation Web system, and of the trust system in MAS. It also gives some quantitative formulas. In this model, the agent recommendation trust can be regarded as the agent direct trust in calculating the trust value. This model can control the malicious recommendation and the cooperative cheating. It can improve the agglomeration of the agent relation Web, and can remarkably simplify the selecting process of the cooperative partner of an agent. So this model can improve the agent cooperation in MAS.

Key words: Multi-agent system; Trust model; Credit; Relation Web; TBT