ETH researchers have used reinforcement learning to teach the in-house four-legged robot ANYmal how to play badminton.