Synthesizing Solutions from Memory Chunks

An intelligent agent or problem solver is expected to solve a problem correctly and efficiently, communicate the solution and also explain the reasoning behind the solution. Several problem solving paradigms are available, each having its own advantages. Model based planners are very robust but at increased computational cost while case based methods are often more efficient. In this paper we explore a problem solving methodology integrating these two methods in a hybrid architecture, exploiting their advantages and avoiding the drawbacks. Route planning has been chosen as the candidate problem. It has two major subactivities, one in route finding and the other in route communication. Both these problems have been earlier studied from the point of view of path planning and natural language generation. We describe a system, named CaSyn ( Case Synthesizer ), which adopts a hierarchical problem solving approach with the higher level reasoning providing control and a means of explaining the solution.