Zero-shot transfer of a tactile-based continuous force control policy from simulation to robot
Lach, Luca ; Haschke, Robert ; Tateo, Davide LU
; Peters, Jan
; Ritter, Helge
; Borràs, Júlia
and Torras, Carme
(2024)
2024 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2024
In IEEE International Conference on Intelligent Robots and Systems
p.725-732
- Abstract
The advent of tactile sensors in robotics has sparked many ideas on how robots can leverage direct contact measurements of their environment interactions to improve manipulation tasks. An important line of research in this regard is grasp force control, which aims to manipulate objects safely by limiting the amount of force exerted on the object. While prior works have either hand-modeled their force controllers, employed model-based approaches, or not shown sim-to-real transfer, we propose a model-free deep reinforcement learning approach trained in simulation and then transferred to the robot without further fine-tuning. We, therefore, present a simulation environment that produces realistic normal forces, which we use to train... (More)
The advent of tactile sensors in robotics has sparked many ideas on how robots can leverage direct contact measurements of their environment interactions to improve manipulation tasks. An important line of research in this regard is grasp force control, which aims to manipulate objects safely by limiting the amount of force exerted on the object. While prior works have either hand-modeled their force controllers, employed model-based approaches, or not shown sim-to-real transfer, we propose a model-free deep reinforcement learning approach trained in simulation and then transferred to the robot without further fine-tuning. We, therefore, present a simulation environment that produces realistic normal forces, which we use to train continuous force control policies. A detailed evaluation shows that the learned policy performs similarly or better than a hand-crafted baseline. Ablation studies prove that the proposed inductive bias and domain randomization facilitate sim-to-real transfer. Code, models, and supplementary videos are available on https://sites.google.com/view/rl-force-ctrl
(Less)
- author
- Lach, Luca
; Haschke, Robert
; Tateo, Davide
LU
; Peters, Jan
; Ritter, Helge
; Borràs, Júlia
and Torras, Carme
- publishing date
- 2024
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2024
- series title
- IEEE International Conference on Intelligent Robots and Systems
- pages
- 8 pages
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- conference name
- 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2024
- conference location
- Abu Dhabi, United Arab Emirates
- conference dates
- 2024-10-14 - 2024-10-18
- external identifiers
-
- scopus:85216488839
- ISSN
- 2153-0866
- 2153-0858
- ISBN
- 9798350377705
- DOI
- 10.1109/IROS58592.2024.10802386
- language
- English
- LU publication?
- no
- id
- 307da758-c698-4bbc-ac44-1760861cf1ec
- date added to LUP
- 2025-10-16 14:05:39
- date last changed
- 2025-11-27 17:28:00
@inproceedings{307da758-c698-4bbc-ac44-1760861cf1ec,
abstract = {{<p>The advent of tactile sensors in robotics has sparked many ideas on how robots can leverage direct contact measurements of their environment interactions to improve manipulation tasks. An important line of research in this regard is grasp force control, which aims to manipulate objects safely by limiting the amount of force exerted on the object. While prior works have either hand-modeled their force controllers, employed model-based approaches, or not shown sim-to-real transfer, we propose a model-free deep reinforcement learning approach trained in simulation and then transferred to the robot without further fine-tuning. We, therefore, present a simulation environment that produces realistic normal forces, which we use to train continuous force control policies. A detailed evaluation shows that the learned policy performs similarly or better than a hand-crafted baseline. Ablation studies prove that the proposed inductive bias and domain randomization facilitate sim-to-real transfer. Code, models, and supplementary videos are available on https://sites.google.com/view/rl-force-ctrl</p>}},
author = {{Lach, Luca and Haschke, Robert and Tateo, Davide and Peters, Jan and Ritter, Helge and Borràs, Júlia and Torras, Carme}},
booktitle = {{2024 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2024}},
isbn = {{9798350377705}},
issn = {{2153-0866}},
language = {{eng}},
pages = {{725--732}},
publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}},
series = {{IEEE International Conference on Intelligent Robots and Systems}},
title = {{Zero-shot transfer of a tactile-based continuous force control policy from simulation to robot}},
url = {{http://dx.doi.org/10.1109/IROS58592.2024.10802386}},
doi = {{10.1109/IROS58592.2024.10802386}},
year = {{2024}},
}