Trust values for agent selection in multiagent systems

Abstract

A key step of any distributed system, based on the negotiation paradigm, is the selection of an agent to whom offer a contract to solve a problem. The agents must have social knowledge that may contain a list of acquaintances. If more than one agent match the requirements needed to solve a problem, a conflict resolution criterion is necessary. The paper begins describing the agentUs structure, with the knowledge that intervenes in the selection process. Afterwards, some strategies for agent selection, based on knowledge acquired during previous contracts, are presented, and a generic model is proposed to support them. The adaptation of the proposed model to societies with different features is discussed too. Finally, some extensions for the agentUs selection criteria are explained. The proposed solution is simple and efficient.

Postprint (published version)