Robot Hand with Dexterity Can Withstand Substantial Abuse for AI Research Purposes

A robotics firm, famed for presenting its adept robotic hand with Jeff Bezos at Amazon re: MARS, has introduced a robust new model designed for machine learning research, developed in collaboration with Google’s DeepMind.

Located in London, Shadow Robot brings over twenty years of expertise in robot design and has catered to prestigious research and industry clients, such as NASA, ESA, OpenAI, Google, MIT, and numerous universities.

The Robust and Dexterous Humanoid Hand

The latest iteration of “the world’s most dexterous humanoid robot hand” presents a departure from its predecessors. Developed with insights from the Google DeepMind robotics team, the new Shadow Hand features a gripper-like formation with only three fingers.

The robotics firm emphasizes the challenge of creating hardware that balances dexterity for complex tasks with durability for learning processes, highlighting the importance of real-world testing in robot learning. Such repairs can be both expensive and time-consuming, consequently slowing down experiments.

Additionally, in addition to prioritizing speed, flexibility, and precision, the new robot hand is engineered to withstand a considerable degree of misuse. This includes “enduring aggressive force demands, abrasion, and impacts, ensuring resilience amid challenging learning environments.”

It has dimensions of 350 mm in length, with a width of 165 mm and a height of 160 mm (13.78 x 6.5 x 6.3 in). Each individual finger weighs 1.2 kg (2.6 lb), and the entire hand weighs 4.1 kg (9 lb). Furthermore, it necessitates a power supply of 48-V/200-W.

Advanced Features of the Robot Hand

The robot hand boasts precise torque control, enabling each finger to exert up to 10 N of fingertip pinch force. Driven by motors housed in the base and connected via “tendons,” the four joints of each finger can transition from fully open to closed in just 500 milliseconds.

Each finger operates as an independent unit and features several 3-DOF tactile sensors at the proximal and middle segments. Furthermore, a stereo camera setup, directed at the inner surface of silicone skin covering the fingertip, delivers real-time, high-resolution, wide-dynamic-range tactile feedback. These features collectively enhance the robot’s ability to interact with its surroundings “through the sense of touch.”

In case of damage during AI experiments, a finger module of the robot hand can be replaced to minimize downtime. Additionally, tactile sensors are replaceable, with the finger’s communication network automatically transmitting data. Pricing details for the Shadow Hand are unavailable, but it will debut at ICRA 2024 in Yokohama, Japan, next week.

Read the Original Article on: New Atlas

Read more: How are Robots Helping in Personalized Knee Replacement Surgery