An autonomous ping-pong-playing robot nicknamed Ace achieved a milestone for AI and robotics in Tokyo by competing with and, in some cases, defeating top-level human players at table tennis, a feat that could foreshadow a host of other applications for equally skilled robots.
Ace, created by Japanese company Sony’s AI research division, is the first robot to achieve professional-level performance in a competitive physical sport that requires quick decisions and precise execution, the project leader said. The Ace achieved this by employing high-speed perception, AI-based control and a state-of-the-art robotic system.
There have been several table tennis-playing robots since 1983, but until now they have been unable to rival highly skilled human competitors. Ace changed that with its performance against elite-level and professional human players in matches following the rules of the International Table Tennis Federation, the sport’s governing body, and refereed by licensed judges.
“Unlike computer games, where previous AI systems outperform human experts, physical and real-time sports such as table tennis remain a huge open challenge due to the need for fast, accurate and competitive interactions near obstacles and at the limit of human reaction time,” said Peter Dürr, director of Sony AI Zurich and leader of Sony AI’s Ace project.
The goal of the project was not just to compete in table tennis, but also to develop insights into how robots can perceive, plan and act with human-like speed and precision in dynamic environments, Dürr said.
“The success of Ace, with its perception system and learning-based control algorithm, suggests that similar techniques could be applied to other areas that require rapid real-time control and human interaction — such as manufacturing and service robotics, as well as applications in sports, entertainment and safety-critical physical domains,” said Dürr, lead author of a study describing Ace’s achievements published Wednesday in the journal Nature.
In matches detailed in the study, Ace won three of five matches against elite players in April 2025 and lost two matches against professional players, the highest level of skill in the sport. Sony AI said that Ace has since defeated professional players in December 2025 and last month.
Companies around the world are moving forward with robots. On Sunday, for example, robots outperformed human runners in a half-marathon in Beijing.
“A blur to the human eye”
AI systems have already excelled in digital domains in strategy games like chess and Go and in complex video games.
While video games take place in simulated environments, table tennis requires quick decision-making, precise physical execution and continuous adaptation to an unpredictable opponent, Dürr said. The ball moves at high speeds with complex rotations and trajectories, pushing humans and robots to operate at the limits of perception, prediction and motor control, Dürr said.
The Ace architecture integrates nine synchronized cameras and three vision systems to track a spinning ball with exceptional accuracy and fast processing time.
“This is fast enough to capture movements that would be a blur to the human eye,” said Dürr.
The researchers developed a custom robotic platform with eight joints. This was, said Dürr, the minimum number needed to execute competitive strokes: three for the position of the racket, two for its orientation, and three for the speed and power of the stroke.
Mayuka Taira, a professional table tennis player who lost a match to Ace last December, said in comments provided by Sony AI that the robot’s strengths “are the difficulty of predicting its movements and the absence of emotions.”
“As you can’t read his reactions, it’s impossible to tell what kind of shots he doesn’t like or have difficulty with, which makes playing against him even more difficult,” said Taira.
Rui Takenaka, an elite player who has won and lost matches against Ace, commented: “When it came to my serve, if I used a complex serve, Ace would also return the ball with a complex effect, which made things difficult for me. But when I used a simple serve — what we call a dead serve — Ace would return a simpler ball. This made it easier for me to attack on the third hit, and I think that was the main reason I was able to win.”
The Ace has room for improvement, Dürr said.
“Ace has a superhuman ability to read the rotation of incoming balls and a superhuman reaction time. Because he learns to play not by watching humans play, but by being trained alone in simulation, he also reacts differently than human players and creates surprising situations,” said Dürr.
“At the same time, professional human athletes are very good at adapting to their opponent and finding weaknesses, which is an area we are working on.”
