Watch: Swiss researchers unveil walking robot hand

Swiss researchers at ETH Zurich's Soft Robotics Lab have released videos and a pre-print paper detailing a disembodied robot hand that walks on its fingers. Watch Swiss researchers unveil the Thing-like creation that uses machine learning to stay balanced, recover from falls, push objects and even type on a keyboard.

Key Takeaways

• ETH Zurich's Soft Robotics Lab detailed a robot hand that walks using its fingers.
• Researchers compared the device to Thing from The Addams Family.
• Machine learning in simulation taught the hand to coordinate uneven fingers and stay balanced.
• Demos show it walking on varied surfaces, righting itself, pushing objects and using a keyboard.
• Leaders say adaptable robots could work in confined or disaster spaces humans cannot easily reach.

The October 30 odd-news drop from UPI framed the project as perfect spooky-season fare: a severed-looking hand that moves on its own. For readers who follow bizarre news, the clip hits the same uncanny note as classic pop culture — with real lab science behind it.

What did the Swiss robot hand actually do?

Videos from the Soft Robotics Lab show the hand walking across different surfaces. It can get itself up after falling over, push objects out of the way and operate a computer keyboard.

The team likened the autonomous gripper to Thing, the Addams Family's roaming hand. That visual is why the release landed so hard online — but the underlying claim is technical, not just theatrical.

How did researchers teach the hand to walk?

Doctoral candidate Amirhossein Kazemipour said getting a robotic hand to walk unaided is unusually hard. The fingers are not identical, so the system must coordinate them while remaining balanced.

"Getting a robotic hand to walk is actually pretty challenging," he said in a school-released video. "You know, it's fingers aren't all the same, so it has to learn how to coordinate all of them while remaining balanced."

Instead of hand-coding every motion, the group used machine learning. The robot practiced walking in a computer simulation until it learned which movements best met its goals.

Kazemipour framed a larger aim: not merely how to make a hand walk, but how to make robots adaptable enough to find new ways to use their bodies and interact with the world.

Why does a walking robot hand matter beyond the viral clips?

Assistant professor Robert Katzschmann said the demo may look spooky, yet it carries a practical lesson about purpose-built robots. A large machine can enter one kind of space; a small hand can enter another — possibly a tighter, disaster-related environment.

"It might look a bit spooky, but I think there is actually a pretty meaningful lesson from this. We can have robots made for different purposes," Katzschmann said. "A big robot can go into one space, a small hand can do into a completely different space -- maybe a much more confined, disaster space."

That adaptability pitch is what separates the project from pure novelty. Walking on fingers is the hook; learning new uses for an existing body is the research stake.

For now, the public record is the lab's videos, the pre-print paper and the quotes above. No independent performance benchmarks beyond those materials were included in the UPI report that circulated the story.