Planning in Prodigy

Planning in Prodigy

Planning is implemented using the general purpose problem-solver. A plan is simply the trace of a problem solution generated by the problem solver. The PLANNER/PROBLEM SOLVER uses domain knowledge and applicable control knowledge to make efficient plans. Control knowledge may include stored plans.

Prodigy is able to do hierarchical planning, by selecting an appropriate abstraction hierarchy. The planner uses the ALPINE abstraction module to build a series of abstract solutions to problems. The abstraction hierarchy is traversed and increasingly detailed plans built during this traversal. Thus the approach Prodigy uses is top-down, which the authors contend produces more elegant and concise plans than an opportunistic (bottom-up) approach.

The planner uses the definition of a problem as an initial state and a goal state(s). Then the planning/problem solving module, directed by knowledge from ALPINE, traverses the search space at each level in the abstraction hierarchy. When a solution is found, the plan is refined at the next lower level as described above.


Return to the top of this architecture.

Go to a discussion of this capability for multiple architectures.


Current Location: Prodigy-Capabilities-Planning

Go to NEXT page.