Vise2 an Application of Intelligent Agents Approaching Eecient Cooperation

ViSe2 agents are intelligent software agents that can solve users' problems either by themselves or through cooperation. In a cooperative multi-agent paradigm, how smart an agent is depends on both its local reasoning capability, and its social ability to e ciently seek assistance. This paper focuses on applying a twin-base agent modeling technique for agents to capture and reason the activities of the others, and thus e ciently locate peer agents for help. To verify the e ciency of cooperation, we have implemented a ViSe2 multi-agent cooperation framework, and compared the system performance with other cooperation approaches. Experimental results obtained indicate that the twin-base modeling achieves the most e cient cooperation with the ideas of pre-evaluation and increased localization.