Planning with Abstraction

Intelligent problem solvers for complex domains must have the capability of reasoning abstractly about tasks that they are called upon to solve. The method of abstraction presented here allows one to reason analogically and hierarchically, making both the task of formalizing domain theories easier for the system designer, as well as allowing for increased computational efficiencies. It is believed that reasoning about concepts that share structure is essential to improving the performance of automated planning systems by allowing one to apply previous compuational effort expended in the solution of one problem to a broad range of new problems.