Conclusions

Abstract

The overall objective of this research has been to develop, implement, and evaluate an approach for coordinating cooperating problem solvers, and this was done in two major phases. In the first phase (Chapters 3–5), the control mechanisms for an individual problem solver were modified to improve its self-awareness of goals and plans. It uses simplified domain knowledge to cheaply generate an abstract view of the problem situation so that it can recognize potential solutions and predict their long-term significance and cost. By reasoning about how the potential solutions are related, the problem solver plans actions that work toward one or more of them, and that at the same time generate information that helps resolve uncertainty about which potential solutions to pursue. Because the results of earlier actions may affect how (and whether) a plan should be pursued, the planning mechanisms interleave plan generation, monitoring, and repair with plan execution, leading to more versatile planning where actions to achieve problem solving goals and actions to resolve uncertainty are integrated into a single plan. The experiments indicate that this self-awareness improves the problem solver’s control decisions, and the additional control overhead (computation and storage) is generally compensated for by resources saved in problem solving since fewer incorrect actions are taken.

If to do were as easy as to know what were good to do, chapels had been churches, and poor men’s cottages princes’ palaces. It is a good divine that follows his own instructions; I can easier teach twenty what were good to be done, than to be one of the twenty to follow mine own teaching. -Shakespeare (The Merchant of Venice)