Sony's 'Ace' Humanoid Outperforms Professional Table Tennis Players

Sony AI has unveiled "Ace," a state-of-the-art humanoid robot that has achieved professional-level proficiency in table tennis. In recent matches regulated by International Table Tennis Federation (ITTF) standards, Ace successfully defeated several high-level human competitors, showcasing a leap in robotic agility and precision.

Advanced Perception and Control

The robot's performance is driven by a sophisticated high-speed perception system. Utilizing an array of nine specialized cameras and three distinct vision systems, Ace can track and predict the trajectory of a table tennis ball with millisecond accuracy. This data is fed into an eight-joint motor control system, allowing the robot to execute rapid, precise movements and react to the dynamic spin and speed of competitive play.

Simulation-to-Reality Breakthrough

Perhaps the most significant technical achievement of the Ace project is its training methodology. The robot was trained entirely within a simulated environment using advanced self-training techniques. By mastering complex physical interactions in simulation before moving to the real world, Sony’s researchers have demonstrated that Physical AI models can achieve elite performance without the need for extensive human demonstration or watching human gameplay.

The Future of Physical AI

The success of Ace serves as a powerful demonstration of the progress in Physical AI. Beyond sports, the technologies developed for Ace—including high-speed perception and precise motor coordination—have broad implications for the future of robotics in industries requiring high-speed interaction and safety in complex, fast-paced environments.