12K views · 165 reactions | Learning Robust Autonomous Navigation and...

10K views · 126 reactions | Discrete-Time Hybrid Automata Learning:...

Researchers at ETH Zürich have taught ANYmal-D, a four-legged manipulator robot, to play badminton against humans. The system uses a reinforcement-learning controller that integrates vision, locomotion, and arm manipulation to intercept and return shots. While the robot isn’t yet at human-competitive level, the work demonstrates how sport can serve as a proxy task for developing robots with dynamic perception-motion coordination—a capability relevant for applications such as search and rescue or