See moreA single-arm robot mounted on two-axis linear rails, guided by reinforcement learning, is now capable of playing live table tennis matches against human professional players. Using 9 high-speed cameras and additional event-based vision sensors, the system estimates not only accurate ball trajectories but also spin, which can reach close to 9,000 rpm in high-level play. Its arm and wrist joints agilely control the racket as human opponents vary their shots and tactics. This achievement reflects a highly coordinated integration of sensors, software, hardware, and AI computing, offering a glimpse of a future in which robots can respond to situations that require highly accurate split-second judgment. (Apr 2026)
