Sony has built a table tennis robot that has beaten elite human players, a breakthrough that researchers say could mark a major step forward for robotics in tasks that require speed, precision and real-time decision-making. The robot, ‘Ace’, has also outplayed several national level players in controlled matches, underscoring its ability to compete in the fast-paced sport.
Developed by Sony AI, Ace goes far beyond traditional training robots. Featuring an eight-joint robotic arm and paddle, it is designed to handle the speed, spin and unpredictability of real matches. In official competitions, he defeated three out of five players who participated in national tournaments and major competitions.
Matches were played according to standard rules, with equipment certified by the International Table Tennis Federation and supervised by human referees. Each match was played to 11 points. Against two professional-level players, Ace was unable to win a match, showing that the highest level of the sport remains one step ahead.
Table tennis has always been a difficult challenge for machines. The ball can travel at speeds in excess of 70 km/h, often with strong spin that changes its trajectory and bounces. Reading this rotation and reacting in time requires sharp anticipation and precise timing.
The Ace relies on a high-speed vision system using 12 cameras with Sony sensors to monitor ball position, speed and spin. His moves have been trained using deep learning, which allows him to perform a series of 15 serves and maintain a rally. During testing, rallies averaged five hits, with the robot often returning blows faster than its opponents expected.
She has shown the ability to deal with difficult turnovers and awkward net shots, areas where it is difficult to maintain consistency even for experienced players. That consistency has helped him gain an edge over national-level competitors, even as pro players have revealed areas where improvement still needs to be made.
The findings were published in Nature on April 23, making Ace among the most advanced systems at this level in a real sports environment. Earlier table tennis robots were unable to consistently challenge experienced human players under official conditions.
According to Sony AI lead scientist Peter Stone, this work could have applications outside of sports, especially in environments where machines need to react quickly and work with precision alongside humans.
– The end
Issued by:
Saurabh Kumar
Published on:
23 Apr 2026 23:56 IST
