Focusing Attention in Anytime Decision-Theoretic Planning

Planning with Incomplete Information for Robot Problems

Volume

Planning with Incomplete Information for Robot Problems

Contents

Any anytime algorithm used for decision-making should have the property that it considers the most important aspects of the decision problem first. In this way, the algorithm can first eliminate disastrous decisions and recognize particularly advantageous decisions, considering finer details if time permits. We view planning as a decision-making process and discuss the design of anytime algorithms for decisiontheoretic planning. In particular, we present an anytime decision-theoretic planner that uses abstraction to focus attention first on those aspects of a planning problem that have the highest impact on expected utility. We discuss control schemes for refining this behavior and methods for automatically creating good abstractions. We present an intelligent agent architecture for integrating our anytime planner with an execution module and describe the status of the implementation.