This chapter presented the replay mechanism. It motivates the problem and describes the approach developed. The replay algorithm is designed as an extension to the base-level planner. It involves a complete reinterpretation of the justifications structures in the new problem solving context, as well as the development of appropriate actions to be taken when transformed justifications are no longer valid.
The base-level problem solver alternates between generating alternatives to solve a problem and searching the space created by these alternatives. In contrast, the analogical reasoner tests previous alternatives, attempting to pursue the successful ones. The branching factor of the search space may also be reduced when the replay mechanism validates previous failures and prunes them from the new search space.
The replay mechanism can integrate guidance from multiple past similar cases. The chapter also discusses different merging strategies.
Unable to display preview. Download preview PDF.