LEAP Hand V2: Dexterous, Low-cost Anthropomorphic Hybrid Rigid Soft Hand for Robot Learning

Kenneth Shaw, Deepak Pathak

Track: Paper Submission Track

Keywords: Dexterous Manipulation, Robot Learning

TL;DR: Low-Cost Robot Hand for Robot Learning

Abstract: The human hand is a remarkable feat of biology, offering remarkable versatility and precision with many joints and muscles working in tandem. It enables us to adeptly handle intricate tools with great strength. Yet, it retains a soft, secure, and yielding nature for delicate objects. This fusion of formidable strength and gentle compliance renders it an unparalleled manipulative instrument. However, robot hands attempting to emulate this have often fallen into one of two categories: soft or rigid. Soft hands, while compliant and safe lack the precision and strength of human hands. Conversely, while rigid robot hands can match the precision and power of human hands, they are brittle and do not conform to their environment. Our proposed solution is to build a robotic hand that bridges the gap between these two categories. We call this hand DLA Hand, a dexterous, $3000, simple anthropomorphic soft hand that is extremely dexterous and versatile. First, it achieves a balance of human-hand-like softness and stiffness via a 3d printed soft exterior combined with a 3d printed internal bone structure. Next, DLA Hand incorporates two powered articulations in the foldable palm: one spanning the four fingers and another near the thumb—mimicking the essential palm flexibility for human-like grasping. Lastly, DLA Hand boasts a dexterous Metacarpophalangeal (MCP) kinematic structure, making it highly human-like, easy to assemble, and versatile. Through thorough real-world experiments, we show that DLA Hand exceeds the capabilities of many existing robot hands for grasping, teleoperated control, and imitation learning. We plan to release 3D printer files and assembly instructions on our website for the dexterous hand research community to use.

Submission Number: 203