We have already covered plenty of robots that can play soccer, basketball, and even tennis. Most of them are in humanoid form. This robot dog can play badminton. Researchers came up with a reinforcement learning-based control policy for “for whole-body visuomotor skills involving all degrees of freedom to achieve effective shuttlecock tracking and striking.” In other words, the robot can coordinate its full body movement, track the shuttlecock, and hit it accurately.
These robots learn to adapt to imperfect visual input. This approach also has a prediction model for the shuttlecock’s trajectory. The above video shows this robot in action.
[HT]