Repairing Decision-Theoretic Policies Using Goal-Oriented Planning

In this paper we address the problem of how decision-theoretic policies can be repaired. This work is motivated by observations made in robotic soccer where decisiontheoretic policies become invalid due to small deviations during execution; and plan repair might pay off rather than replan from scratch. Our policies are generated with Readylog, a derivative of Golog based on the situation calculus, which combines programming and planning for agents in dynamic domains. When an invalid policy was detected, the world state is transformed into a PDDL description and a state-of-the-art PDDL planner is deployed to calculate the repair plan.

• Type: Journal
• Year: 2008
• Where: KI
• Authors: Christoph Mies, Alexander Ferrein, Gerhard Lakemeyer
• Keywords: Artificial Intelligence, Decision-theoretic Policies, Decisiontheoretic Policies, KI 2008, Policies