Ace returned about 75 percent of balls in matches played under official rules, Sony AI reports. The table-tennis robot beat three of five high-level amateurs and won one of seven matches against two Japanese-league professionals by pairing high-speed sensing, on-the-fly decision-making and an eight-jointed robotic arm to track spin and return shots. The work was published in Nature.

Robot design built for speed and spin

Ace is a three-part system: a perception layer that spots the ball and spin; an AI that makes split-second choices; and an eight-jointed high-speed robotic arm that executes precise strokes. The arm is designed to move quickly and accurately enough to meet competitive timing demands.

  • Perception: tracks spin and trajectory in real time.
  • Decision-making: AI selects responses on the fly.
  • Actuation: the arm places the racket where the software directs.

Table tennis forces sensing, interpretation, rapid planning and fast motor execution within fractions of a second, making it a challenging benchmark for robots.

Match results and what they show

Sony AI reports Ace beat three of five high-level amateurs and won one of seven matches against two Japanese-league professionals, Minami Ando and Kakeru Sone. Analysts inside the project found Ace returned about 75 percent of balls it faced.

Those figures point to a robot that relies on control and consistency rather than smash power: it converts human errors into points but has gaps when facing trained professionals who vary pace, spin and placement in complex ways. Project director Peter Dürr framed the achievement as more than a novelty, saying the system matched or exceeded human reaction and decision speed in a physical space.

Why table tennis matters for robotics

Games have long served as benchmarks for machine intelligence, but physical games add perception noise and hardware constraints. Table tennis compresses those challenges into a small playing area and rapid time scale: high speed, heavy spin and tight timing mean milliseconds matter.

Because Ace maintained returns under official match conditions, it demonstrates that coordinated sensing, AI and high-speed actuation can operate at human-competitive speeds in a closed, physical task. That makes it a useful research benchmark and suggests practical uses such as training and skill development.

Related Articles

Published in Nature, the results show Ace returned about 75 percent of balls under official match conditions — a measurable step forward for robots in real-world physical tasks and potential training and skill-development uses.

This article was created with AI assistance.