Sony's Table Tennis Robot Ace Defeats Professional Players, Published in Nature
Sony AI has developed a groundbreaking table tennis robot named 'Ace' that has defeated multiple elite human players while adhering to official ITTF rules. The robot combines a high-speed perception system with 9 cameras tracking the ball at 200fps, an AI-driven control system using deep reinforcement learning, and an 8-DOF robotic arm.
In April 2025, Ace won 3 out of 5 matches against elite players with 10+ years of training and defeated professional Japanese players including Miyuu Kihara (WTT top 25). The system can return 71%+ of high-spin shots and consistently handles balls under 14m/s. Published in Nature (2026), this represents a milestone in embodied intelligence research.
The robot features:
- 9 high-speed Sony cameras at 200fps
- 3 event-based vision units at 400-700Hz for spin detection
- 10ms data processing time
- 8-DOF high-speed mechanical arm
- Deep reinforcement learning in custom simulator
Ace can use 15 different serving techniques and scored 16 aces against elite players compared to just 8 by human opponents. Despite its success, researchers note that the gap to world champion level remains, with challenges in defensive tactics and Sim-to-Real transfer.
