Sony's artificial intelligence company has announced a new achievement in the field of robotics through its Ace project, a self-operating system capable of competing with elite table tennis players
This development represents a significant shift in the transition of artificial intelligence from digital environments to the real world, after years of achieving success in games such as chess and racing simulations. However, interacting with real-world sports has remained a greater challenge due to the high speed and the need for precise and unpredictable reactions.Ace relies on advanced technologies that combine high-precision sensors, reinforcement learning, and microrobots. The project builds upon previous successes, such as the Gran Turismo Sophy system, which excelled at driving virtual cars at high speeds, but Ace brings these capabilities to a fully realistic environment.
The system uses nine high-speed cameras to accurately track the movement of a table tennis ball in 3D, along with sensors capable of measuring the ball's spin and speed in real time. This data is integrated into a machine learning system that allows the robot to adapt during gameplay rather than relying on fixed rules.
Table tennis requires quick reactions and high accuracy in dealing with rotation and speed changes, which made it a difficult test for the robot's capabilities.
Peter Dorr, director of artificial intelligence at Sony Zurich and project leader, said: "The robot has demonstrated the ability to compete with humans in a competitive sport, in terms of reaction speed and decision-making in a realistic environment."
He also explained that table tennis presents a complex challenge that requires a combination of speed and accuracy, noting that this achievement opens the door to wider uses of artificial intelligence in interacting with humans in real time.
During the tests, Ace faced professional players according to the rules of the International Table Tennis Federation, and achieved victory in three out of five matches against elite players, and also delivered a competitive performance against professional players.
The system demonstrated the ability to handle highly spin balls at speeds up to 450 radians/second (a unit of measurement for spin speed (also known as angular velocity)), with a response rate exceeding 75%, in addition to scoring points directly from serves.
Sony asserts that these results represent, for the first time, the ability of physical AI systems to excel in direct interactive tasks against human experts.
Researchers at the company point out that this development is not limited to games, but paves the way for broader applications in robots capable of working in rapidly changing, realistic environments.
This research, titled "Outperforming elite table tennis players using a self-operating robot," was published on the cover of the journal Nature.
