The Ping-Pong Robot is Guaranteed to Beat You!

Robot Player

While we’ve seen AI systems dominate virtual competitions in games like chess, Go, and even StarCraft II, mastering physical games poses a much greater challenge. A robot must adapt to unpredictable changes in its environment, interpret them, decide on a response, and execute that action all in a brief window of time.

This is exactly what Ace, a sophisticated robot made up of three key components, has accomplished. It boasts a perception system that detects the ball’s rotation, which can alter its bounce and flight pattern, along with an AI system capable of real-time decision-making. Finally, it features high-speed robotics: an agile eight-jointed arm that executes quick and precise racket placement.

To evaluate Ace’s capabilities, researchers had the robot compete against five high-level amateur players, achieving three victories out of five matches. However, against two professional players from the Japanese league, Minami Ando and Kakeru Sone, Ace’s performance was less impressive, securing just one win from seven matches. Analyzing Ace’s games indicated that it scored points not primarily through power, but by demonstrating excellent control, successfully returning 75 percent of the balls.

“This research has shown that an autonomous robot can actually win in a sports competition, matching or surpassing human reaction times and decision-making abilities in a physical environment,” stated Peter Dürr, director of Sony AI and project leader on Ace, in a press release.

Beyond Table Tennis

Ace’s achievements signal a breakthrough in robotics. It integrates high-speed sensing, AI for decision-making, and robotic control to compete with human players under real-world conditions, reacting for precision in real time.

“Table tennis is a game of significant complexity requiring split-second decisions alongside speed and strength,” Dürr elaborates. Robots like Ace could pave the way for learning new techniques and enhancing player performance across various fields.