- Abstract:
-
Due to the unavoidable fact that a robot's sensors will be limited in some manner, it is entirely possible that it can find itself unable to distinguish between differing states of the world. This confounding of states, also referred to as perceptual aliasing, has serious effects on the ability of reinforcement learning algorithms to learn stable policies. Using simple grid world navigation problems we demonstrate experimentally these effects. Although 1-step backup reinforcement learning algorithms performed surprisingly better than expected, our results confirm that algorithms using eligibility traces should be preferred.
- Copyright:
- 2003 by The University of Edinburgh. All Rights Reserved
- Links To Paper
- No links available
- Bibtex format
- @InProceedings{EDI-INF-RR-0176,
- author = {
Paul Crook
and Gillian Hayes
},
- title = {Learning in a State of Confusion: Perceptual Aliasing in Grid World Navigation},
- book title = {Proceedings of Towards Intelligent Mobile Robots (TIMR 2003)},
- year = 2003,
- month = {Aug},
- }
|
