A Modular Structured Approach to Conditional Decision-Theoretic Planning

A realistic system for planning with unccrtaln information in partially observable domains must be able to reason about sensing actions and to condition its further actions on the sensed information. Among implemented planning systems, we can distinguish two approaches to contingent decision-theoretic planning. The first is characterized by a highly unconstrained plan space, while the second is characterized by a constrained and inflexible specification of plan space. In this paper, we take a middle ground between these two approaches that we consider to be more practical. We perm/t ttLe nser to specie" the. structure of the space of possible plans to bc considered but to do so in a flexible manner. This flexibility is obtained through the use of a modular representation. VV~ separate the representation of actions from the representation of domain relations and we separate those from the representation of the plan space. Acti,ms and domain relations are represented with schematic Ba~s net fragments and plan space is represented using programming language constructs. We present a planning system that can find optimal plans given this rcpresentation.