Automatically generating abstractions for problem solving
Craig A. Knoblock · 1991
A major source of inefficiency in automated problem solvers is their inability to decompose problems and work on the more difficult parts first. This issue can be addressed byemploying a hierarchy of abstract problem spaces to focus the search. Instead of solving a problem in the original problem space, a problem is first solved in an abstract space, and the abstract solution is then refined at successivelevels in the hierarchy. While this use of abstraction can significantly reduce search, it is often difficult to find good abstractions, and the abstractions must be manually engineered by the designer of a problem domain. This thesis presents a completely automated approach to generating abstractions for problem solving. The abstractions are generated using a tractable, domainindependent algorithm whose only inputs are the definition of a problem space and the problem to be solved and whose output is an abstraction hierarchy that is tailored to the particular problem. The algorithm g...