RUKA: Rethinking the Design of Humanoid Hands with Learning

TL;DR

RUKA addresses dexterous manipulation by pairing a tendon-driven humanoid hand with data-driven control learned from autonomous MANUS glove data. The approach yields a human-sized, affordable hand with $15$ DOFs and $11$ actuators, assembled from 3D-printed parts for under $1300$, and demonstrates 29/33 grasps, 20 hours of durability, and strong strength metrics compared to LEAP, Allegro, and Inmoov. The learned controllers map fingertip/joint positions to motor commands using an LSTM+MLP architecture, trained via autonomous data collection and validated on both robot and human data, enabling teleoperation and policy-learning demonstrations like Cube Flipping and Bread Pick-and-Drop. The work provides an open-source, scalable tool for researchers to study dexterous manipulation and emphasizes data-driven strategies to compensate tendon-driven control challenges, with practical implications for rapid prototyping and learning-based manipulation in real environments.

Abstract

Dexterous manipulation is a fundamental capability for robotic systems, yet progress has been limited by hardware trade-offs between precision, compactness, strength, and affordability. Existing control methods impose compromises on hand designs and applications. However, learning-based approaches present opportunities to rethink these trade-offs, particularly to address challenges with tendon-driven actuation and low-cost materials. This work presents RUKA, a tendon-driven humanoid hand that is compact, affordable, and capable. Made from 3D-printed parts and off-the-shelf components, RUKA has 5 fingers with 15 underactuated degrees of freedom enabling diverse human-like grasps. Its tendon-driven actuation allows powerful grasping in a compact, human-sized form factor. To address control challenges, we learn joint-to-actuator and fingertip-to-actuator models from motion-capture data collected by the MANUS glove, leveraging the hand's morphological accuracy. Extensive evaluations demonstrate RUKA's superior reachability, durability, and strength compared to other robotic hands. Teleoperation tasks further showcase RUKA's dexterous movements. The open-source design and assembly instructions of RUKA, code, and data are available at https://ruka-hand.github.io/.

Figures (11)

• Figure 1: Ruka is a tendon-driven humanoid hand that is simple, affordable, and capable. Its size and morphology closely match those of a human hand, enabling it to perform diverse human-like power, precision, and fine-grained grasps.
• Figure 2: (A) A Venn diagram of a variety of robotic hands allegroshaw2023leapshadowhandetukuru2024robotinmoovclone demonstrates Ruka's unique combination of low cost, anthropomorphism and usability. (B) An illustration of the sizes of different hands that are commonly used by the robotics community. Ruka is designed to closely match the average human hand.
• Figure 3: (A) Joints enable 15 degrees of freedom of Ruka labeled with their corresponding joint names. (B) The splay of the fingers allow for natural abduction-adduction movement without an active degree of freedom. (C) The MCP and PIP / DIP coupled tendons (light blue and dark blue respectively) are responsible for flexion, while the springs are responsible for extension.
• Figure 4: The intersection space of the thumb fingertip and each of the fingertips overlayed on the hand. This demonstrates the large set of opposable grasps possible with Ruka.
• Figure 5: We run the hand continuously for 90 minutes, repeatedly doing a full range of motion. Here we show the temperature of each motor during. Note how the temperature stabilizes after some time.