All the Feels: A Dexterous Hand with Large Area Sensing
Raunaq Bhirangi, Abigail DeFranco*, Jacob Adkins*, Carmel Majidi, Abhinav Gupta, Tess Hellebrekers and Vikash Kumar
Paper | Video | Code | Data
Abstract
High cost and lack of reliability has precluded the widespread adoption of dexterous hands in robotics. Furthermore, the lack of a viable tactile sensor capable of sensing over the entire area of the hand impedes the rich, low-level feedback that would improve learning of dexterous manipulation skills. This paper introduces an inexpensive, robust and scalable platform - the D'Manus - aimed at resolving these challenges while satisfying the large-scale data collection capabilities demanded by deep robot learning. Studies on human manipulation point to the criticality of low-level tactile feedback in performing everyday dexterous tasks. The D'Manus comes with ReSkin sensing on the entire surface of the palm as well as the fingertips and we demonstrate the fully integrated setup in a bin picking and sorting tactile task. Code, documentation, design files, detailed assembly instructions, trained models, task videos, and all supplementary materials required to recreate the setup can be found on this website.
Features of the D'Manus
Reliable Actuators: The D'Manus is actuated at joint level using Dynamixel XM430-210 motors. We have run the D'Manus for over 400 hours with no motor breakages.
Large-Area Tactile Sensing: ReSkin sensors are integrated with the fingertips and the palm. Each fingertip sensor is comprised of 8 magnetometers while the palm sensor consists of 32 magnetometers for a total of 56 magnetometers.
Simulated Backend: A MuJoCo-based simulation of the D'Manus is also made available for fast prototyping and debugging.
Open-Source Design: Hardware design as well as build instructions are open-sourced and available online for everybody to build their own D'Manus. Detailed assembly instructions can be found on the Assembly page