Simon Guist, Jan Schneider, Hao Ma, Le Chen, Vincent Berenz, Julian Martus, Heiko Ott, Felix Grüninger,
Michael Muehlebach, Jonathan Fiene, Bernhard Schölkopf and Dieter Büchler
Max Planck Institute for Intelligent Systems
Accepted for RSS 2024
Abstract
Operating robots precisely and at high speeds has been a long-standing goal of robotics research. Balancing these competing demands is key to enabling the seamless collaboration of robots and humans and increasing task performance. However, traditional motor-driven systems often fall short in this balancing act. Due to their rigid and often heavy design exacerbated by positioning the motors into the joints, faster motions of such robots transfer high forces at impact. To enable precise and safe dynamic motions, we introduce a four degree-of-freedom (DoF) tendon-driven robot arm. Tendons allow placing the actuation at the base to reduce the robot’s inertia, which we show significantly reduces peak collision forces compared to conventional robots with motors placed near the joints. Pairing our robot with pneumatic muscles allows generating high forces and highly accelerated motions, while benefiting from impact resilience through passive compliance. Since tendons are subject to additional friction and hence prone to wear and tear, we validate the reliability of our robotic arm on various experiments, including long-term dynamic motions. We also demonstrate its ease of control by quantifying the nonlinearities of the system and the performance on a challenging dynamic table tennis task learned from scratch using reinforcement learning.
Open-Source Hardware
The design of PAMY2 primarily employs off-the-shelf, commercially available components, 3D-printed parts, and only a limited number of custom-machined parts.
To facilitate further research and development in dynamic robotic tasks, we have published our design on GithHub. The PAMY2 GitHub contains:
STL files of all custom made parts
Bill of Materials, including list of all the parts, as well as cost and assembly time estimates
Pneumatic Layout
Wiring Diagram
Coming soon: Two-DoF wrist for PAMY2
Open-Source Software
Besides the Hardware, we have also open-sourced the Software required to run our system.
To control PAMY2, our robot software allows to send commands of desired pressures to the robot, as well as accessing the current and past states of the robot. The software is mostly written in C++ and has Python wrappers. It builds on top of o80 which is a Software framework for synchronising and organising message exchange between processes via simple customised Python APIs. [1]
For the table tennis smashing task, we have built a custom gym environment with only minor modifications to the one used by Büchler et al. [2]. For our experiments with PAMY2, we train in a Hybrid-Sim-and-Real setup (HySR), with the real robot and a simulated ball. The "learning table tennis from scratch"-package builds on top of o80 and our robot software to make training and evaluation in this hybrid setting, or in a completely real or completely simulated settings possible.
[1] Berenz, Vincent, et al. "The o80 C++ templated toolbox: Designing customized Python APIs for synchronizing realtime processes." Journal of Open Source Software 6.66 (2021): 2752.
[2] Büchler, Dieter, et al. "Learning to play table tennis from scratch using muscular robots." IEEE Transactions on Robotics 38.6 (2022): 3850-3860.
Motions Dataset
To collect long-term data, we conducted an experiment where the robot operated continuously for approximately 25 days. The data was recorded at a high sampling rate of 500 Hz.
The dataset contains:
Observed and desired pressures for each muscle (eight muscles)
Position and velocity for each joint (4 degrees of freedom)
Timestamps
Bibtex
@inproceedings{guist2024safe,
title={Safe \& Accurate at Speed with Tendons: A Robot Arm for Exploring Dynamic Motion},
author={Guist, Simon and Schneider, Jan and Ma, Hao and Chen, Le and Berenz, Vincent and Martus, Julian and Ott, Heiko and Gr{\"u}ninger, Felix and Muehlebach, Michael and Fiene, Jonathan and Sch{\"o}lkopf, Bernhard and B{\"u}chler, Dieter},
booktitle={Robotics: Science and Systems (RSS)},
year={2024}
}