On-Line Learning of Coordination Plans

Coordination is an essential technique in cooperative distributed problem solving (CDPS). However, sophisticated coordination strategies are not always costeffective in all problem-solving situations. This paper presents a learning method to acquire coordination plans for specific problem-solving situations so that the appropriate type of coordination strategy is used. This learning is accomplished by recording and analyzing traces of inferences after problem solving. The analysis results in identification of situations where inappropriate coordination strategies have caused redundant activities or the lack of timely execution of important activities, thus degrading system performance. The analysis is also used to create situation-specific coordination plans, which use additional non-local information about activities in the networks, that are added to the system to remedy the problem. These situation-specific plans have the effect of introducing different levels of coordination into this system. We present this approach for coordination using two examples from a real distributed problem-solving application involving diagnosis of a local area network. This research was conducted at the University of Massachusetts at Amherst. Toshiharu Sugawara’s stay at the University of Massachusetts was supported by Nippon Telegraph and Telephone Corporation (NTT). This research was also supported in part by a grant from Network General Corporation, ARPA contract N00014-92-J-1698 and ONR contract N00014-92-J-1450. The content of the information does not necessarily reflect the position or the policy of the organizations supporting this research, and no official endorsement should be inferred. This is an extended version of a paper with the same title that will appear in the proceedings of 12th workshop on Distributed Artificial Intelligence.