Article Summary
On April 22, 2026, the journal Nature published a study detailing “Ace,” a table tennis robot developed by Sony AI, which has achieved human, expert-level performance in the sport. This marks a first for an autonomous robot in a competitive, high-speed physical activity.
Key Technical Specifications
Hardware: Custom-built robotic arm with eight degrees of freedom, enabling aggressive shot execution and broad motion range.
Perception System: Nine high-speed cameras positioned around the court, capable of tracking the ball’s logo in real-time to precisely measure spin and velocity. Three gaze-control systems for tracking.
AI Architecture: Model-free reinforcement learning trained through millions of simulated and physical trials without explicit programming.
Reaction Time: Control policy updates at 31.25 Hz (every 32 milliseconds), matching or surpassing human reaction times.
Performance: Achieved 75% serve return rate against professional players.
Performance Milestones
- Early 2025: Initial challenges against top-tier professionals
- December 2025: Successfully defeated professional Japanese league players including Minami Ando and Kakeru Sone
- March 2026: Continued dominance against elite players
- Spin Mastery: Key breakthrough in reading and returning complex ball spins
Research Significance
Project Lead Peter Dürr: Validates potential for autonomous robots to match or surpass human reaction times and decision-making in complex physical settings.
Former Olympian Kinjiro Nakamura: Ace performed shots previously thought impossible, suggesting robotic systems could aid humans in discovering new athletic techniques.
Future Applications: Technology extends beyond sports into advanced manufacturing and service robotics where high-speed, real-time interaction is crucial.
Sources
- ScienceAlert
- Strait Times
- Robohub
- Nature Journal
- YouTube (Sony AI channel)