Surpassing Human Wrist Constraints: Dexterity-Enhanced Techniques for Teleoperated Robotic Manipulation

Abstract

Free-hand teleoperation enables intuitive robot control by mapping human hand and arm poses to end-effector motion. However, its effectiveness in complex tasks is constrained by the mismatch between human wrist biomechanics and robotic arm capabilities. From 60 demonstration videos across four task categories, we identified three end-effector motion capabilities that surpass human constraints: unbounded axial rotation, omnidirectional alignment, and path-constrained orientation maintenance. Inspired by enhancement techniques in virtual reality (VR), we propose two interaction strategies—finger-based motion extension and non-isomorphic rotational gain—instantiated in four methods: FingerCircle, FingerRaise, WristSpin, and WristDNR. We evaluated these methods in VR using three representative teleoperation tasks, with a clutch-based control mode as the baseline. Results showed that finger-based methods improve efficiency and agency in continuous rotation tasks, whereas wrist-based methods enhance alignment and orientation control, at the expense of predictability. We further deployed them on a UR5 robotic arm in two application scenarios. Overall, these techniques better leverage robotic dexterity while reducing operator workload.