Sony's Ace Robot Masters Table Tennis, Defeating Professionals

For years, table tennis has served as a formidable testbed for robotics research, with machines consistently faltering against the sport's high speeds, spinning balls, and unpredictable human opponents. However, Sony AI's 'Ace' system has shattered these limitations. In a demonstration, the robot successfully returned six balls before its opponent, Japanese table tennis athlete Miyu Sasaki, became more aggressive. Sasaki's powerful shots, spinning the ball over 100 times per second and reaching speeds exceeding 70 kilometers per hour, proved too much for the robot, leading to an error and a point for Ace. This scene, from April 2025, highlights the system's progress. By December 2025, Ace defeated professional player Mayuka Taira, and has since added more victories, including one against a player ranked in the Top 25 globally, according to Peter Dürr, project leader at Sony AI. This achievement is particularly noteworthy as it surpasses Google Deepmind's previous best robot, which struggled against players skilled in spin. Ace's innovation lies in its ability to precisely measure ball rotation. Nine cameras capture the ball's trajectory: six determine its position, while three focus on measuring spin speed and direction, aided by the International Table Tennis Federation logo on the ball. This data feeds into two algorithms, one for rapid estimation and another for more accurate analysis, enabling the robot to react effectively. The Neue Zürcher Zeitung views this as a significant leap in AI and robotics, showcasing a level of precision and adaptability previously thought unattainable in such dynamic sports.

“

Sony AI's 'Ace' system has shattered these limitations.