Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation

Wenyan Yang; Alexandre Angleraud; Roel S. Pieters; Joni Pajarinen; Joni-Kristian Kämäräinen

doi:10.1109/ICRA48891.2023.10161145

Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation

Wenyan Yang, Alexandre Angleraud, Roel S. Pieters, Joni Pajarinen, Joni-Kristian Kämäräinen

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

2 Citations (Scopus)

15 Downloads (Pure)

Abstract

Robot control for tactile feedback based manip-ulation can be difficult due to modeling of physical contacts, partial observability of the environment, and noise in perception and control. This work focuses on solving partial observability of contact-rich manipulation tasks as a Sequence-to-Sequence (Seq2Seq) Imitation Learning (IL) problem. The proposed Seq2Seq model first produces a robot-environment interaction sequence to estimate the partially observable environment state variables, and then, the observed interaction sequence is transformed to a control sequence for the task itself. The proposed Seq2Seq IL for tactile feedback based manipulation is experimentally validated on a door-open task in a simulated environment and a snap-on insertion task with a real robot. The model is able to learn both tasks from only 50 expert demonstrations while state-of-the-art reinforcement learning and imitation learning methods fail.

Original language	English
Title of host publication	Proceedings - ICRA 2023
Subtitle of host publication	IEEE International Conference on Robotics and Automation
Publisher	IEEE
Pages	5829-5836
Number of pages	8
ISBN (Electronic)	979-8-3503-2365-8
ISBN (Print)	979-8-3503-2366-5
DOIs	https://doi.org/10.1109/ICRA48891.2023.10161145
Publication status	Published - 2023
Publication type	A4 Article in conference proceedings
Event	IEEE International Conference on Robotics and Automation - London, United Kingdom Duration: 29 May 2023 → 2 Jun 2023

Conference

Conference	IEEE International Conference on Robotics and Automation
Country/Territory	United Kingdom
City	London
Period	29/05/23 → 2/06/23

Publication forum classification

Publication forum level 2

ASJC Scopus subject areas

Software
Control and Systems Engineering
Electrical and Electronic Engineering
Artificial Intelligence

Access to Document

10.1109/ICRA48891.2023.10161145

Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Accepted author manuscript, 4.26 MBLicence: Unspecified

https://urn.fi/URN:NBN:fi:tuni-2023112110147Licence: Unspecified

Cite this

@inproceedings{92a0780608c9488e98e040ad379e7897,

title = "Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation",

abstract = "Robot control for tactile feedback based manip-ulation can be difficult due to modeling of physical contacts, partial observability of the environment, and noise in perception and control. This work focuses on solving partial observability of contact-rich manipulation tasks as a Sequence-to-Sequence (Seq2Seq) Imitation Learning (IL) problem. The proposed Seq2Seq model first produces a robot-environment interaction sequence to estimate the partially observable environment state variables, and then, the observed interaction sequence is transformed to a control sequence for the task itself. The proposed Seq2Seq IL for tactile feedback based manipulation is experimentally validated on a door-open task in a simulated environment and a snap-on insertion task with a real robot. The model is able to learn both tasks from only 50 expert demonstrations while state-of-the-art reinforcement learning and imitation learning methods fail.",

author = "Wenyan Yang and Alexandre Angleraud and Pieters, \{Roel S.\} and Joni Pajarinen and Joni-Kristian K{\"a}m{\"a}r{\"a}inen",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; IEEE International Conference on Robotics and Automation ; Conference date: 29-05-2023 Through 02-06-2023",

year = "2023",

doi = "10.1109/ICRA48891.2023.10161145",

language = "English",

isbn = "979-8-3503-2366-5",

pages = "5829--5836",

booktitle = "Proceedings - ICRA 2023",

publisher = "IEEE",

address = "United States",

}

Yang, W , Angleraud, A , Pieters, RS, Pajarinen, J & Kämäräinen, J-K 2023, Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation. in Proceedings - ICRA 2023: IEEE International Conference on Robotics and Automation. IEEE, pp. 5829-5836, IEEE International Conference on Robotics and Automation, London, United Kingdom, 29/05/23. https://doi.org/10.1109/ICRA48891.2023.10161145

TY - GEN

T1 - Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation

AU - Yang, Wenyan

AU - Angleraud, Alexandre

AU - Pieters, Roel S.

AU - Pajarinen, Joni

AU - Kämäräinen, Joni-Kristian

PY - 2023

Y1 - 2023

N2 - Robot control for tactile feedback based manip-ulation can be difficult due to modeling of physical contacts, partial observability of the environment, and noise in perception and control. This work focuses on solving partial observability of contact-rich manipulation tasks as a Sequence-to-Sequence (Seq2Seq) Imitation Learning (IL) problem. The proposed Seq2Seq model first produces a robot-environment interaction sequence to estimate the partially observable environment state variables, and then, the observed interaction sequence is transformed to a control sequence for the task itself. The proposed Seq2Seq IL for tactile feedback based manipulation is experimentally validated on a door-open task in a simulated environment and a snap-on insertion task with a real robot. The model is able to learn both tasks from only 50 expert demonstrations while state-of-the-art reinforcement learning and imitation learning methods fail.

AB - Robot control for tactile feedback based manip-ulation can be difficult due to modeling of physical contacts, partial observability of the environment, and noise in perception and control. This work focuses on solving partial observability of contact-rich manipulation tasks as a Sequence-to-Sequence (Seq2Seq) Imitation Learning (IL) problem. The proposed Seq2Seq model first produces a robot-environment interaction sequence to estimate the partially observable environment state variables, and then, the observed interaction sequence is transformed to a control sequence for the task itself. The proposed Seq2Seq IL for tactile feedback based manipulation is experimentally validated on a door-open task in a simulated environment and a snap-on insertion task with a real robot. The model is able to learn both tasks from only 50 expert demonstrations while state-of-the-art reinforcement learning and imitation learning methods fail.

U2 - 10.1109/ICRA48891.2023.10161145

DO - 10.1109/ICRA48891.2023.10161145

M3 - Conference contribution

AN - SCOPUS:85168710362

SN - 979-8-3503-2366-5

SP - 5829

EP - 5836

BT - Proceedings - ICRA 2023

PB - IEEE

T2 - IEEE International Conference on Robotics and Automation

Y2 - 29 May 2023 through 2 June 2023

ER -

Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation

Abstract

Conference

Publication forum classification

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this