Files in this item



application/pdfECE499-Sp2014-deng.pdf (722kB)Restricted to U of Illinois
(no description provided)PDF


Title:Reinforcement Learning Based Track Solving Control on a Humanoid Robot
Author(s):Deng, Dechen
Contributor(s):Levinson, Stephen
Subject(s):Reinforced learning
Humanoid robot
Abstract:For a long time people have debated about whether an artificial intelligence can learn like human beings. Whether it is possible for them to learn a skill or an ability through human or environmental interaction continues to raise discussions. Thus, in a small part of this discussion, I implemented a reinforcement learning based control of a ball on a track using a humanoid robot called iCub. The robot uses a camera as an eyeball to see the world and uses sensory-motor feedback to control the ball based on the policy learned from the feedback of the environment. In the process, I used Q-learning, a reinforcement learning algorithm for the policy learning and I also implemented a simple computer vision method to track the ball. This experiment reveals that a robot has the ability to interact with the environment and even with people.
Issue Date:2014-05
Date Available in IDEALS:2014-09-22

This item appears in the following Collection(s)

Item Statistics