AdaPT robot tennis demo Tao Huang

    yazan Terrymiller: Rafael Nadal

    Researchers built AdaPT to learn pro tennis styles from broadcast videos, capturing the distinct moves of Nadal, Federer, and Djokovic. The system splits motion planning and tracking so humanoid robots can adjust execution speed for real-world errors. Demonstrations on the Unitree G1 and full-size Dobot Atom show forehand and backhand rallies plus professional-style surfs.

    Metin dökümü (en)

    Researchers have developed ADAPT, a system that teaches humanoid robots to play tennis using the distinctive styles of professional players such as Rafael Nadal, Roger Federer, and Novak Djokovic. It learns their movements from broadcast videos, then separates motion planning from tracking, and adapts execution speed to handle real-world errors. The Unitree G1 can perform forehand and backhand rallies and professional-style surfs, while the full-size Dobot Atom also demonstrates these skills.