An autonomous desk tennis robotic developed by Sony AI has competed towards and defeated high-level human gamers in regulated matches, in keeping with Reuters. The system is a part of a broader class sometimes called “physical AI,” the place synthetic intelligence is utilized to machines working in real-world environments.
The robotic, named Ace, was designed to function in a aggressive sport surroundings that requires fast decision-making and exact motor management. In response to the venture group, it combines high-speed notion programs with AI-driven management to execute photographs beneath match circumstances.
Ace competed in matches performed beneath Worldwide Desk Tennis Federation guidelines and officiated by licensed umpires. In trials documented in April 2025, the system gained three out of 5 matches towards elite gamers and misplaced two towards professional-level opponents. Sony AI reported that subsequent matches in December 2025 and early 2026 included wins towards skilled gamers.
Earlier desk tennis robots have existed for the reason that Nineteen Eighties, however they weren’t capable of match the efficiency of superior human gamers. “Not like pc video games, the place prior AI programs surpass human specialists, bodily and real-time sports activities like desk tennis stay a serious open problem,” mentioned Peter Dürr, director at Sony AI Zurich and lead of the venture.
AI programs have achieved sturdy leads to digital environments like chess and video video games, the place circumstances are totally simulated, Dürr mentioned.
Dürr mentioned the system was developed to review how robots can reply with pace and accuracy in dynamic environments. The work was detailed in a examine printed within the journal Nature.
The game presents technical challenges as a result of pace and variability of the ball, together with advanced spin and altering trajectories, which require fast sensing and coordinated motion in tight time constraints, Dürr mentioned. Ace’s structure contains 9 synchronised cameras and three imaginative and prescient programs, which observe the ball’s motion and spin. The system processes visible information at a pace enough to seize movement that’s troublesome for the human eye to resolve. “That is quick sufficient to seize movement that will be a blur to the human eye,” Dürr mentioned.
The robotic platform makes use of eight joints to regulate the racket. Three management positioning, two management orientation, and three handle shot pressure and pace. The configuration was designed to fulfill the minimal mechanical necessities for aggressive play.
Not like many AI programs skilled by human demonstration, Ace was skilled in simulation. The strategy allowed it to develop its personal methods, leading to play patterns that differ from human opponents. Dürr mentioned the system “learns to play not from watching people” however by self-training in simulated environments.
Skilled participant Mayuka Taira, who misplaced a match to the system, mentioned the robotic was troublesome to foretell as a result of it exhibits no seen cues throughout play. Rui Takenaka, an elite participant who each gained and misplaced towards Ace, mentioned it dealt with advanced spins nicely however was extra predictable on easier serves. Taira mentioned the system’s lack of emotional alerts made it more durable to anticipate its responses. “As a result of you possibly can’t learn its reactions, it’s inconceivable to sense what sort of photographs it dislikes or struggles with,” she mentioned.
Dürr mentioned the system demonstrates sturdy skill in studying ball spin and reacting rapidly, whereas ongoing work focuses on enhancing adaptability throughout matches. The venture group mentioned related notion and management methods might be utilized to areas like manufacturing and repair robotics.
Humanoid robots examined in long-distance race
On the 2026 Beijing E-City Humanoid Robotic Half Marathon, humanoid robots competed over a 21-kilometre course in Beijing. The occasion included greater than 100 robots and roughly 12,000 human contributors, who ran on separate tracks.
A robotic named Lightning, developed by Honor, accomplished the race in 50 minutes and 26 seconds. The time was sooner than Olympic runner Jacob Kiplimo’s 57 minutes and 20 seconds recorded on the Lisbon Half Marathon in March. Lightning collided with a barricade in the course of the race however continued and completed first. Honor robots additionally positioned second and third within the competitors. Efficiency improved in comparison with the earlier 12 months’s occasion, the place the quickest robotic accomplished the course in two hours, 40 minutes and 42 seconds. Organisers mentioned the occasion was supposed to check humanoid robots in large-scale, real-world circumstances.
In response to Related Press, one other Honor robotic accomplished the course in 48 minutes beneath distant management. Nevertheless, race guidelines prioritised autonomous navigation, and Lightning was recognised because the official winner.
Honor engineers mentioned applied sciences developed for the robotic, together with structural reliability and liquid-cooling programs, might be utilized in industrial situations.
(Picture by Mattias Banguese)
See additionally: Cadence expands AI and robotic partnerships with Nvidia, Google Cloud
Need to study extra about AI and large information from business leaders? Try AI & Big Data Expo happening in Amsterdam, California, and London. The great occasion is a part of TechEx and co-located with different main know-how occasions. Click on here for extra info.
AI Information is powered by TechForge Media. Discover different upcoming enterprise know-how occasions and webinars here.
