Variable Resolution Discretization in Optimal Control

Abstract. The problemof state abstractionis of centralimportancein optimalcontrol,reinforcement learning and Markov decision processes. This paper studies the case of variable resolution straction for continuous time and space, deterministicdynamic control problems in which near-optimal policies are required. We begin by de ning a class of variable resolution policy and value function representationsbased on Kuhn triangulationsembeddedin a kd-trie. We then consider top-down approaches to choosing which cells to split in order to generate improved policies.

• Type: Journal
• Year: 2002
• Where: ML
• Authors: Rémi Munos, Andrew W. Moore
• Keywords: Deterministicdynamic Control Problems, Machine Learning, ML 2002, Variable Resolution, Variable Resolution Straction