Shadow Robot Unveils New Robust Model for Machine Learning Research Developed with Google’s DeepMind Collaboration

Shadow Robot, a renowned robotics company known for its dexterous robot hand demo with Jeff Bezos at ICRA 2019, has now unveiled a new robust model designed for machine learning research in collaboration with Google’s DeepMind.

The London-based company, with over 20 years of experience in robot design, has worked with high-profile clients like NASA, ESA, OpenAI, Google, MIT, and various universities. The new Shadow Hand, developed with insights from Google DeepMind, features three gripper-like fingers and is built for speed, flexibility, precision, and durability.

“A key challenge in AI and robotics is developing hardware that is dexterous for complex tasks yet robust for learning,” said the company. The new robot hand is designed to endure misuse, aggressive force demands, abrasion, and impacts, making it ideal for trial and error learning in real-world scenarios.

The hand measures 350 mm in length, 165 mm wide, and 160 mm high, with each finger capable of up to 10 N of fingertip pinch force. The fingers are equipped with tactile sensors and a stereo camera setup for real-time tactile feedback, allowing the robot to interact with its environment through touch.

If a finger module is damaged during experiments, it can be easily replaced for minimal downtime. The company will showcase the Shadow Hand at ICRA 2024 in Yokohama, Japan, providing a glimpse into the future of robotics and machine learning research.

Source: Shadow Robot