Tactile sensors designed to give robots human-like touch
Next generation robotic perception for object manipulation, manufacturing, and more
Today, general purpose in-hand and object manipulation remains an unsolved challenge in robotics and artificial intelligence.
Common touch sensors do not capture a signal comparable to human sense of touch and robots lack tactile intelligence to perform dexterous tasks and generate hand manipulation skills.
Meta Platforms, Inc. ©2022. All rights reserved. Content used with permission by Meta.
In partnership with Meta AI Research, GelSight is manufacturing DIGIT, a low-cost and miniaturized tactile sensor that gives robots human touch.
Complex Object Manipulation
Assembly Process Efficiency
Material Properties Identification
DIGIT improves upon past tactile sensors by miniaturizing the form factor to be mountable on multi-fingered hands.
DIGIT is built on GelSight’s imaging-based technology platform so that machine learning and AI techniques can be applied immediately to DIGIT’s sensor output.
Engineers and researchers love working with DIGIT because DIGIT is:
Easy to build
Low-cost at $300 per sensor
FREQUENTLY ASKED QUESTIONS
- How do I use DIGIT? DIGIT is designed to be easy to connect and use – DIGIT is equipped with a standard USB-3 port, and its drivers/interface is available here. The sensor is not designed for quantitative measurements.
- Is there a simulator for DIGIT? Yes, an open-source simulator for DIGIT is available at https://github.com/facebookresearch/tacto.
- On what robots can DIGIT be mounted? DIGIT is designed to be natively compatible with the Wonik’s Allegro Hand. In addition, we provide files for 3d-printable adaptors for mounting DIGIT on several common robotic platforms.
- Can I use Machine Learning with DIGIT? Yes, DIGIT is natively supported by the PyTouch open-source library for touch processing and is compatible with Windows and Linux platforms only.