Sony AI trains robot to rally against professional ping-pong players, published in Nature

Sony AI published research in Nature demonstrating that a high-speed robot can learn the real-time perception and rapid movements needed to compete against professional athletes at ping-pong—a task requiring split-second decisions and precise physical control.

The robot combines wheeled and legged locomotion (four individually powered legs on a single circular frame called Ringbot Quad), allowing it to move in ways traditional robots cannot—similar to how a skilled athlete uses their whole body, not just arms, to stay competitive.

For roboticists and AI researchers, this proves autonomous systems can master complex real-world sports involving fast-moving objects and unpredictable opponents, opening the door to robots that handle dynamic, high-pressure environments beyond controlled factory floors—fields like emergency response or collaborative manufacturing.

The research was published on the cover of Nature (a top-tier peer-reviewed journal), signaling that robot-athlete competition has reached a credible milestone; upcoming robotics conferences ICRA 2026 (Vienna, June 1–5) and RSS 2026 (Sydney, July 13–17) will likely showcase similar breakthroughs.