DeepMind is testing out a highly durable robotic hand

The new robotic hand offers extremely fast and flexible finger movements, while being robust enough to withstand severe damage. This durability will aid the hands already used in Google DeepMind's robotics experiments during the trial-and-error learning required to train artificial intelligence.

Developed by the UK-based Shadow Robot Company, this latest robotic hand can go from fully open to closed in less than 500 milliseconds and can pinch fingertips with up to 10 newtons of force. It can also withstand repeated punishment, such as hitting your finger with a piston from multiple angles or a person destroying the device with a hammer.

The new hand's robust design makes it well-suited for AI-powered robotics experiments based on reinforcement learning, where the robot learns how to interact with its environment by fumbling through tasks through trial and error. You can learn it gradually.” Ram Ramamoorthy at the University of Edinburgh, UK.

“Every interaction with the world involves the risk of collision damage,” he said. rich walkerDirector of Shadow Robot Company during a press conference.

One trade-off is that the hand is “heavier than other options because the design decisions are made for reliability in long-term use,” Ramamoorthy says. The new Shadow Hand's chunky three-finger setup weighs a total of 4.1 kilograms, or 1.2 kilograms per finger.

This structure makes the hands less anthropomorphic than other robot limbs, but more versatile. You can add and change fingers as needed, and each finger is a modular component that can be quickly replaced in case of emergency. of damage.

Each robot finger has hundreds of sensors on the fingertip and dozens of sensors on the other parts of the finger. A miniature camera focuses on the inner surface of the silicone skin of each robot finger. This internal view can indicate the object's hardness and shape, as the robot's skin can deform when it touches the object.

“It's great in terms of sensing and very robust,” he says. Ingmar Posner at Oxford University. “It's also designed to be easily modified.”

Some research institutions and companies outside of Google DeepMind could benefit from talented people, Posner says. But it's likely to be expensive, and Shadow Robot Company hasn't announced its price yet. Other researchers may prefer cheaper robotic hand options, even if they lack some or all of the advanced sensing and object processing capabilities that shadow hands are combined with. 1 package.