Show simple item record

dc.contributor.author
Vulin, Nikola
dc.contributor.author
Christen, Sammy
dc.contributor.author
Stevšić, Stefan
dc.contributor.author
Hilliges, Otmar
dc.date.accessioned
2021-03-30T07:17:36Z
dc.date.available
2021-03-28T21:55:09Z
dc.date.available
2021-03-30T07:17:36Z
dc.date.issued
2021-04
dc.identifier.issn
2377-3766
dc.identifier.other
10.1109/LRA.2021.3061308
en_US
dc.identifier.uri
http://hdl.handle.net/20.500.11850/476690
dc.description.abstract
In this letter we address the challenge of exploration in deep reinforcement learning for robotic manipulation tasks. In sparse goal settings, an agent does not receive any positive feedback until randomly achieving the goal, which becomes infeasible for longer control sequences. Inspired by touch-based exploration observed in children, we formulate an intrinsic reward based on the sum of forces between a robot's force sensors and manipulation objects that encourages physical interaction. Furthermore, we introduce contact-prioritized experience replay, a sampling scheme that prioritizes contact rich episodes and transitions. We show that our solution accelerates the exploration and outperforms state-of-the-art methods on three fundamental robot manipulation benchmarks.
en_US
dc.language.iso
en
en_US
dc.publisher
IEEE
en_US
dc.subject
Deep learning in grasping and manipulation
en_US
dc.subject
intrinsic motivation
en_US
dc.subject
reinforcement learning
en_US
dc.subject
tactile feedback
en_US
dc.title
Improved Learning of Robot Manipulation Tasks Via Tactile Intrinsic Motivation
en_US
dc.type
Journal Article
dc.date.published
2021-02-23
ethz.journal.title
IEEE Robotics and Automation Letters
ethz.journal.volume
6
en_US
ethz.journal.issue
2
en_US
ethz.pages.start
2194
en_US
ethz.pages.end
2201
en_US
ethz.identifier.wos
ethz.identifier.scopus
ethz.publication.place
New York, NY
en_US
ethz.publication.status
published
en_US
ethz.leitzahl
ETH Zürich::00002 - ETH Zürich::00012 - Lehre und Forschung::00007 - Departemente::02150 - Dep. Informatik / Dep. of Computer Science::02658 - Inst. Intelligente interaktive Systeme / Inst. Intelligent Interactive Systems::03979 - Hilliges, Otmar / Hilliges, Otmar
ethz.leitzahl.certified
ETH Zürich::00002 - ETH Zürich::00012 - Lehre und Forschung::00007 - Departemente::02150 - Dep. Informatik / Dep. of Computer Science::02658 - Inst. Intelligente interaktive Systeme / Inst. Intelligent Interactive Systems::03979 - Hilliges, Otmar / Hilliges, Otmar
ethz.date.deposited
2021-03-28T21:55:18Z
ethz.source
SCOPUS
ethz.eth
yes
en_US
ethz.availability
Metadata only
en_US
ethz.rosetta.installDate
2021-03-30T07:17:46Z
ethz.rosetta.lastUpdated
2022-03-29T06:06:24Z
ethz.rosetta.versionExported
true
ethz.COinS
ctx_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.atitle=Improved%20Learning%20of%20Robot%20Manipulation%20Tasks%20Via%20Tactile%20Intrinsic%20Motivation&rft.jtitle=IEEE%20Robotics%20and%20Automation%20Letters&rft.date=2021-04&rft.volume=6&rft.issue=2&rft.spage=2194&rft.epage=2201&rft.issn=2377-3766&rft.au=Vulin,%20Nikola&Christen,%20Sammy&Stev%C5%A1i%C4%87,%20Stefan&Hilliges,%20Otmar&rft.genre=article&rft_id=info:doi/10.1109/LRA.2021.3061308&
 Search print copy at ETH Library

Files in this item

FilesSizeFormatOpen in viewer

There are no files associated with this item.

Publication type

Show simple item record