File(s) under permanent embargo

Agent-advising approaches in an interactive reinforcement learning scenario

conference contribution
posted on 01.01.2017, 00:00 authored by F Cruz Naranjo, P Wuppen, S Magg, A Fazrie, S Wermter
Reinforcement learning has become one of the fundamental topics in the field of robotics and machine learning. In this paper, we expand the classical reinforcement learning framework by the idea of external interaction to support the learning process. To this end, we review a number of proposed advising approaches for interactive reinforcement learning and discuss their implications, namely, probabilistic advising, early advising, importance advising, and mistake correcting. Moreover, we implement the advice strategies for interactive reinforcement learning based on a simulated robotic scenario of a domestic cleaning task. The obtained results show that the mistake correcting approach outperforms a purely probabilistic advice approach as well as the early and importance advising approaches allowing to collect more reward and also to converge faster.

History

Event

IEEE Computational Intelligence Society. Conference (2017 : Lisbon, Portugal)

Series

IEEE Computational Intelligence Society Conference

Pagination

209 - 214

Publisher

Institute of Electrical and Electronics Engineers

Location

Lisbon, Portugal

Place of publication

Piscataway, N.J.

Start date

18/09/2017

End date

21/09/2017

ISBN-13

9781538637159

Language

eng

Publication classification

E1.1 Full written paper - refereed

Copyright notice

2017, IEEE

Editor/Contributor(s)

[Unknown]

Title of proceedings

IEEE ICDL-EPIROB 2017 : Proceedings of the 7th Joint International Confernce on Development and Learning and on Epigenetic Robotics 2017