Robust Reference-Free Sim-to-Real Reinforcement Learning for Bipedal Locomotion

Graduate Thesis Or Dissertation

Public Deposited

Citeable URL: https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/bk128k764

Descriptions

Attribute Name	Values
Creator	Siekmann, Jonah A.
Abstract	In recent years, model-free Deep Reinforcement Learning (RL) has become an increasingly popular alternative to more traditional model-based or optimization-based control methods in solving robotic legged locomotion. However, deploying RL in the real world can be a significant undertaking. Constructing reward functions which compel controllers to learn the desired behavior is not straightforward. For example, can a reward function which trains a controller to stand be easily modified to train it to walk or run? This thesis seeks to provide insights into training such controllers in ways that make desired behaviors easier to realize by parameterizing behaviors in a low-dimensional periodic space which covers the entire breadth of legged gaits. In addition, it explores the limits of the approach by training controllers to interact (in the real world) with challenging terrain conventionally assumed to be extremely difficult for bipedal robots.
Contributor	Godse, Yesh Green, Kevin Warila, John Fern, Alan Hurst, Jonathan
License	All rights reserved
Resource Type	Masters Thesis
Date Issued	2023-03-15
Degree Level	Master's
Degree Name	Master of Science (M.S.)
Degree Field	Robotics
Degree Grantor	Oregon State University
Commencement Year	2023
Advisor	Fern, Alan
Committee Member	Lee, Stefan Hurst, Jonathan Arnadottir, Liney
Rights Statement	In Copyright
Publisher	Oregon State University
Peer Reviewed	No
Language	English [eng]

Parents:

This work has no parents.

In Collection:

Thumbnail	Title	Date Uploaded	Visibility	Actions
	SiekmannJonahA2023.pdf	2023-04-03	Public	Download