Plan-space vs. State-space Planning in Reuse and Replay

The aim of case-based planning (CBP) is to improve the e ciency of plan generation by taking advantage of previous problem-solving experience. This paper explores the relative utility of placing CBP within plan-space vs. state-space planning. We argue that it is in the ability to adapt the previous case to the new situation that plan-space planning proves to be more e cient. To be able to extend a previous episode to solve the current problem situation, a planner needs the ability to splice in new steps into an existing sequence. The planspace planner, which decouples the order of derivation of plan steps from their execution order, provides this capability. We will present controlled empirical studies that support our hypothesis regarding the relative advantage of plan-space planning in CBP. Our experiments demonstrate that this advantage holds whether we employ either of two CBP methods, which we call plan reuse and derivation replay.