LEAP Hand: Low-Cost, Efficient, and Anthropomorphic Hand for Robot Learning

TL;DR

LEAP Hand introduces a low-cost, open-source anthropomorphic dexterous hand designed for robot learning, addressing hardware barriers to real-world dexterity. Its universal abduction-adduction mechanism preserves all finger DoF across positions, while thumb opposability and high manipulability enable versatile grasps. The hand demonstrates robust performance across teleoperation, video-based learning, behavior cloning, and sim2real tasks, outperforming the Allegro Hand in many settings. By releasing hardware, simulation tools, and APIs, the work aims to accelerate democratized research in dexterous manipulation and real-world robot learning.

Abstract

Dexterous manipulation has been a long-standing challenge in robotics. While machine learning techniques have shown some promise, results have largely been currently limited to simulation. This can be mostly attributed to the lack of suitable hardware. In this paper, we present LEAP Hand, a low-cost dexterous and anthropomorphic hand for machine learning research. In contrast to previous hands, LEAP Hand has a novel kinematic structure that allows maximal dexterity regardless of finger pose. LEAP Hand is low-cost and can be assembled in 4 hours at a cost of 2000 USD from readily available parts. It is capable of consistently exerting large torques over long durations of time. We show that LEAP Hand can be used to perform several manipulation tasks in the real world -- from visual teleoperation to learning from passive video data and sim2real. LEAP Hand significantly outperforms its closest competitor Allegro Hand in all our experiments while being 1/8th of the cost. We release detailed assembly instructions, the Sim2Real pipeline and a development platform with useful APIs on our website at https://leap-hand.github.io/

Figures (10)

• Figure 1: (a) LEAP Hand is an anthropomorphic dexterous robot hand designed for robot learning research. It can be assembled in under 4 hours for 2000 USD, is composed of readily available parts, and is robust. (b) to-scale comparison of LEAP Hand and a human hand (c-h) LEAP Hand in different power and precision grasps holding common objects. The hand design and code will be open-sourced to democratize access to hardware for anthropomorhic dexterous manipulation. Video, assembly instructions, and sim2real pipeline at https://leap-hand.github.io/
• Figure 2: Relative size of popular robot hands to scale.Left to right, adult human hand, Allegro Hand allegro, LEAP-C Hand, LEAP Hand, Inmoov inmoov, D'Manus bhirangi2022all. LEAP Hand is similar in size to Allegro and $\sim30\%$ larger than a human hand. D'Manus is considerably larger than the rest. Because of the tendon-driven nature, Inmoov is the smallest robotic hand. The hands are accurate to scale.
• Figure 3: Comparison of MCP joints in different robot hands and their dexterity and two different positions. (A) In LEAP-C Hand there is a large range of motion at extended but not flexed position (B) In LEAP Hand, at flexed and extended positions, the fingertip has a large range of motion. (C) In Allegro, there is a large of motion at flexed but not extended position.
• Figure 4: The human hand kinematics above has ball joints at the MCP and CMC joints. These are difficult joints for low-cost hands to include. Left Figure from cerulo2017teleoperation. Comparison of MCP joints in different robot hands. (A) In LEAP-C Hand there is a large range of motion at extended but not flexed position (B) In LEAP Hand, at flexed and extended positions, the fingertip has a large range of motion. (C) In Allegro, there is a large of motion at flexed but not extended position.
• Figure 5: We compare the possible positions of opposability of the thumb and each of the other fingers on each of the three hands. We find that LEAP Hand has the best even spread on top of the palm and a very large contact area.