Keywords: Soft Robots, Mechanisms & Design, Robot Learning: Reinforcement Learning
TL;DR: We developed a lightweight, tendon-driven robotic arm with pneumatic muscles, enabling precise, high-speed, and safe interactions, making it possible to generate large datasets of dynamic motions.
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. This configuration, paired with pneumatic muscles, enables high-force, accelerated motions and improves safety through passive compliance. This capacity for rapid motion exploration is particularly valuable for generating extensive datasets of dynamic motions autonomously, without additional direct human oversight. We leverage this feature to generate a proprioceptive dataset of 25 days of diverse robot motions that highlights it's robustness and reliability. 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. We open-source the entire hardware design, which can be largely 3D printed, the control software, and the motions dataset at [link removed to comply with anonymity requirements].
Supplementary Material: zip
Submission Number: 31
Loading