Learning to improve coordinated actions in cooperative distributed problem-solving environments

Coordination is an essential technique in cooperative, distributed multiage nt systems. However, sophisticated coordination strategies are not always c ost-effective in all problem-solving situations. This paper presents a lear ning method to identify what information will improve coordination in speci fic problem-solving situations. Learning is accomplished by recording and a nalyzing traces of inferences after problem solving. The analysis identifie s situations where inappropriate coordination strategies caused redundant a ctivities, or the lack of timely execution of important activities, thus de grading system performance. To remedy this problem, situation-specific cont rol rules are created which acquire additional nonlocal information about a ctivities in the agent networks and then select another plan or another sch eduling strategy. Examples from a real distributed problem-solving applicat ion involving diagnosis of a local area network are described.