Most of us don't specifically remember learning how to walk. It was a process that involved slips and falls; trips straight to your, well, bottom. But over time, your body learned to control your movements, learned how far over was too far to lean, learned how to stay upright while walking.
If you are someone who now takes this miracle for granted, there's no better reminder of how difficult it actually was than by learning to ice skate. Sure, there are similarities to walking, but the process is quite different. From balancing to stopping to turning, skating is a unique challenge to master, full of trips straight to your, well ...
Anyway, it's not easy. Now imagine teaching a robot to do it.
Hit the ice!
This was the challenge that the Computational Robotics Lab at ETH Zürich set for themselves with Skaterbot. And to be clear, we're not talking about writing a program that tells a robot to make skating motions. We're talking about AI (artificial intelligence) that allows the robot to learn for itself how to skate.
That is what Skaterbot is all about. Its movements are certainly very careful and basic. And it has the benefit of being a four-legged machine, meaning that it doesn't have to worry about falling over on its, well ... (Hey! No fair!) But it is indeed able to move across the ice as it pleases, and can even track the motion of other objects and skaters on the ice.
Learn more about Skaterbot and watch it play with a pair of professional Swiss hockey players in the video below!