Particle.news
Download on the App Store

Sony Says Ace Robot Reaches Expert-Level Table Tennis, Beating Top Pros

Fast vision plus a custom 8‑joint arm, trained in simulation, powered the result.

Overview

  • Sony AI reports Ace beat professional-level human players in December 2025 and again in March 2026 after winning three of five matches against elite opponents in April 2025.
  • Project lead Peter Diel describes Ace as the first robot to achieve expert-level performance in a competitive physical sport.
  • The system uses nine synchronized cameras and three vision pipelines to track fast, spin-heavy shots with high precision in a sport that pushes perception and control to the limit.
  • A purpose-built platform with eight joints manages paddle position, angle, and swing speed to time returns under tight reaction windows.
  • Ace trained in simulation rather than by copying human play, which the team says produced faster reactions and sharper spin reads, with tactical adaptation still under development.