Bio-inspired legged robots have demonstrated the capability to walk and run across a wide variety of terrains, such as those found after a natural disaster. However, the survival of victims of natural disasters depends on the speed at which these robots can travel. This paper describes the need for adaptive gait tuning on an eight-legged robot, which will enable it to adjust its gait parameters to increase the speed at which it navigates difficult and varying terrains. Specifically, we characterize the robot's performance on varied terrains and use the results to inform the implementation of a finite-difference policy gradient reinforcement learning algorithm. We compare the robot's performance under hand-tuned policies with the performance under the reinforcement learning algorithm, and finally, suggest improvements to the presented policy search process.
Title
Reinforcement Learning Methods to Enable Automatic Tuning of Legged Robots
Published
2012-05-31
Full Collection Name
Electrical Engineering & Computer Sciences Technical Reports
Other Identifiers
EECS-2012-145
Type
Text
Extent
17 p
Archive
The Engineering Library
Usage Statement
Researchers may make free and open use of the UC Berkeley Library’s digitized public domain materials. However, some materials in our online collections may be protected by U.S. copyright law (Title 17, U.S.C.). Use or reproduction of materials protected by copyright beyond that allowed by fair use (Title 17, U.S.C. § 107) requires permission from the copyright owners. The use or reproduction of some materials may also be restricted by terms of University of California gift or purchase agreements, privacy and publicity rights, or trademark law. Responsibility for determining rights status and permissibility of any use or reproduction rests exclusively with the researcher. To learn more or make inquiries, please see our permissions policies (https://www.lib.berkeley.edu/about/permissions-policies).
