DeepMind is experimenting with a nearly indestructible robot hand

A new robotic hand enables extremely fast and flexible finger movements while being robust enough to withstand severe damage. This durability helps the hand, already used in Google DeepMind's robotics experiments, with the trial-and-error learning needed to train artificial intelligence.

This latest robotic hand, developed by the UK-based Shadow Robot Company, can go from fully open to closed within 500 milliseconds and can apply fingertip pressure with a force of up to 10 Newtons. It can also withstand repeated abuse, such as when pistons hit fingers from different angles or when a person smashes the device with a hammer.

The new hand's robust design is well-suited to AI-powered robotics experiments based on reinforcement learning, allowing robots to gradually learn how to interact with environments by completing tasks using trial and error, it said Ram Ramamoorthy at the University of Edinburgh in the United Kingdom.

“Every interaction with the world carries the risk of collision damage,” he said Rich WalkerDirector of the Shadow Robot Company, during a press conference.

One downside is that the hand is “heavier than some other options because design decisions are aimed at reliability over long-term use,” says Ramamoorthy. The chunky three-finger structure of the new Shadow Hand weighs a total of 4.1 kilograms and 1.2 kilograms per finger.

This structure makes the hand look much less anthropomorphic than some other robotic limbs, but also makes it more versatile: it can be modified with more fingers if necessary, and each finger is a modular component that can be quickly swapped out for a replacement in the event of damage .

Each robot finger has hundreds of sensors on the fingertips and dozens on the other finger segments. Tiny cameras focus on the inside surface of the silicone skin of each robot finger – touching an object can deform the robot skin, and this inside view can reveal the hardness and shape of the object.

“It’s excellent in terms of sensing and extremely robust,” he says Ingmar Posner at the University of Oxford. “It’s also designed to be easy to repair.”

Some research labs and companies outside of Google DeepMind may find the capable hand useful, Posner says. But it's likely to be expensive — Shadow Robot Company hasn't announced its price yet — and other researchers may prefer cheaper robotic hand options, even if they lack some or all of the sophisticated sensing and object-handling features that the Shadow Hand combines in one package.