Publication

Towards transferring tactile-based continuous force control policies from simulation to robot

Conference Article

Conference

Workshop on Touch Processing: a new Sensing Modality for AI at NeurIPS 2023 (WSTP)

Edition

2023

Pages

10

Doc link

https://neurips.cc/virtual/2023/workshop/66515

File

Download the digital copy of the doc pdf document

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 that of 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 have 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. An evaluation in which we compare against a baseline and perform an ablation study shows that our approach outperforms the hand-modeled baseline and that our 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

Categories

learning (artificial intelligence).

Author keywords

tactile sensing, object manipulation, object grasping, neural networks, continuous control, reinforcement learning

Scientific reference

L.M. Lach, R. Haschke, D. Tateo, J. Peters, H. Ritter, J. Borràs and C. Torras. Towards transferring tactile-based continuous force control policies from simulation to robot, 2023 Workshop on Touch Processing: a new Sensing Modality for AI at NeurIPS 2023 , 2023, New Orleans (LA), USA, pp. 10.