Researchers at ETH Zürich’s Robotic Systems Lab have achieved a remarkable breakthrough by teaching four-legged robots to play badminton, demonstrating unprecedented whole-body coordination. The footage shows these legged manipulators using reinforcement learning to master the complex challenge of tracking and striking shuttlecocks whilst maintaining perfect balance and mobility. What makes this achievement particularly impressive is the unified control policy that simultaneously coordinates perception, locomotion and precise arm movements—a challenging combination that has long baffled roboticists. This research proves that robots can now predict object trajectories, navigate dynamically and interact with human players in fast-paced sporting scenarios, opening new possibilities for robots in dynamic environments requiring dexterity and coordination.
