A progression of MuJoCo environments and deep reinforcement-learning tools working toward humanoid tennis. Moves from simple ball control through racket-and-wall tasks to an experimental two-humanoid curriculum.

Visiter le site du projet