Publication
Efficient gait generation using reinforcement learning
Conference Article
Conference
International Conference on Climbing and Walking Robots (CLAWAR)
Edition
4th
Pages
411-418
Doc link
http://eu.wiley.com/WileyCDA/WileyTitle/productCd-1860583652.html
File
Authors
Projects associated
Abstract
The use of a free gait becomes necessary when walking over discontinuous terrain or when performing changes in the heading of the robot. If no look-ahead planning is possible, the free gait must be generated reactively. We present a basic gait generation strategy and a heuristic enhancement of it that results in a significant efficiency improvement. In the search for a better strategy, we decided to let the robot adjust the gait by itself attending to a reinforcement signal. Results of the application of a novel reinforcement learning algorithm to the free gait generation task are reported.
Categories
robots.
Scientific reference
J.M. Porta and E. Celaya. Efficient gait generation using reinforcement learning, 4th International Conference on Climbing and Walking Robots, 2001, Karlsruhe, Alemanya, in Climbing and Walking Robots: From Biology to Industrial Applications, pp. 411-418, 2001, Wiley, Hardcover.
Follow us!