ViSe2: an Experimental Model for E cient Multi-agent Cooperation

The Virtual Secretary 2 project (ViSe2) focuses on the construction of a multi-agent cooperation model. As a research vehicle, we have chosen to build intelligent agents that perform secretarial tasks for their users either by themselves or via cooperation. An individual ViSe2 agent has limited knowledge and problem solving capabilities. To act better on its user, the agent could interact with other agents to solve problems. With this regard, an agent's ability to reason about the other agents' activities and thus nd the right peer becomes a key issue. In this paper, we present a way to help agents attain such abilities and thus achieve e cient cooperation. We propose a twin-base system (cooperator-base U task-base) to monitor the other agents' activities in a small agent group. The cooperator-base collects stable information of the others and acts as an auxiliary base to the task-base. The task-base provides a direct mapping between tasks and relevant expert agents that can perform such tasks. A capability revision process is used for keeping the mapping information updated and consistent. With such twin-base modeling, after receiving a task, an agent could directly retrieve experts to perform the job. Considering a large agent group, we suggest a community{club hierarchy organization structure for the purpose of e cient cooperation. We regard a large agent group as a community and divide it into di erent interest groups called clubs. Agents that share similar interests belong to one club, and employ twin-base modeling. Each club has a club trader. These club traders in the community form a special cooperative group, and adopt the twin-base modeling as well. Results received so far indicate that our method achieves expected e cient cooperation with high on-line performance.