An Organizational Approach to E cient Cooperation in An Agent Society?

A cooperative intelligent agent is a user's personal consultant that takes care of problems the user got, and solve them either by itself or through cooperation. To construct a suitable organization structure which allows individual agents to easily reason about the others' activities, and thus nd the peers, is a key issue for e cient cooperation. In this paper, we suggest an activity-oriented hierarchy organization structure, which uses the twin-base modeling technique for achieving e cient cooperation in an agent society. The twin-base consists of a cooperator-base and a task-base, and is originally developed for e cient cooperation in a small agent group. When facing a task beyond an agent's capabilities, the agent directly retrieves the best quali ed peer from its task-base, and asks the peer to take over the job. In a large agent group, agents sharing similar activities constitute a club, which can be regarded as a small agent group, and the twin-base modeling is employed for cooperation among club members. Each club has a club trader. Club traders with similar activities belong to a community, and adopt the twin-base modeling as well. The community traders are members of an agent society. This club-community-society organization achieves e cient cooperation in the sense that: (1) experimental results indicate that the twin-base modeling achieves more e cient cooperation in a club with high on-line performance and low maintenance cost; (2) with the twinbase modeling, agents (including traders) perform better heuristic searches in their peer locating processes by reducing the search chain; (3) the activity-oriented structure minimizes the computation and communication time of peer locating in the community and society.