Coalition formation based on fuzzy trust relationship in MAS

Coalition formation is an important way for agents to cooperate and coordinate. By forming coalitions, agents can execute tasks efficiently and maximize their payoff in multi-agent systems. Most of existing work focuses on designing mechanisms that are based on coalition structure generation or game theory. However they ignore that as a system composed of autonomous agents, relations will play an important role in coordination among agents. Specifically, trust and reputation among agents are crucial to this issue. In this paper, a novel approach is proposed which takes advantage of two relationships: trustworthiness among agents and similarity between tasks. The aim is by taking both relations into account, the trustworthiness to the potential transaction can be estimated, which will serve as valuable reference information to new coalition formation. Another contribution of this work lies in that as ambiguous concepts, these two relationships are denoted by fuzzy relations. In addition, the methods for computing membership function are presented. Based on the composition of fuzzy relation, trust among agents to the new task can be calculated. Then agents who are more trustworthy are selected to form a coalition. In empirical simulations, the results demonstrate that the performance of our mechanism outperforms a random strategy.