GS10-2 Tell Agent Where to Go: Human Coaching for Accelerating Reinforcement Learning

Title:	GS10-2 Tell Agent Where to Go: Human Coaching for Accelerating Reinforcement Learning
Publication:	ICAROB2017
Volume:	22
Pages:	567-570
ISSN:	2188-7829
DOI:	10.5954/ICAROB.2017.GS10-2
Author(s):	Nakarin Suppakun, Suriya Natsupakpong, Thavida Maneewarn
Publication Date:	January 19, 2017
Keywords:	learning from demonstration, reinforcement learning, human assisted learning, semi-supervised learning, robot coaching
Abstract:	In this work, we proposed a method to accelerate learning by allowing a human to coach a robot behavior by inserting an intermediate target at the early phase of the reinforcement learning. By using an intermediate target, the different pair of policy and reward function was temporarily used to select an action that most likely to drive the robot toward the intermediate location, while the global reward function is still used for updating the state-action value. Q learning algorithm was used to test with the proposed method on three learning tasks: ball following, obstacle avoidance, and mountain car. The proposed technique resulted in better learning performance than the traditional RL.
PDF File:	https://alife-robotics.co.jp/members2017/icarob/data/html/data/GS_pdf/GS10/GS10-2.pdf
Copyright:	© The authors. This article is distributed under the terms of the Creative Commons Attribution License 4.0, which permits non-commercial use, distribution and reproduction in any medium, provided the original work is properly cited. See for details: https://creativecommons.org/licenses/by-nc/4.0/

HOME