D. Aliaga (Japan)
RL Learning POMDP, Macro-Actions, Sequence Finding
This is the first of a series of papers dealing with the prob lems of a Reinforcement Learner Robot in a Partially Ob servable Environment and how Macro-Actions are a good approach to deal with those problems. In this paper we are going to introduce the general problem, and then we focus on the first part of the solution: dealing with the au tomatic generation of sensation-action sequences that will be transformed then into Macro-Actions. The paper ex plains the method the robot uses to autonomously generate useful sequences that then are going to be transformed into appropriate Macro-Actions to help it overcome the partial observability problem.
Important Links:
Go Back
