Sony's paddle-wielding robot, called Ace, is adept at playing table tennis and sometimes defeats elite human players, according to multiple reports. The robot plays by official table tennis rules on a typically sized court. Sony claims this is the first time a robot has achieved human, expert-level play in a commonly played competitive sport in the physical world.
The robot has nine camera eyes positioned around the court and can follow the ball's logo to measure its spin. It has eight joints, or degrees of freedom, enabling it to position the racket, execute shots, and respond swiftly. The robot learned to play table tennis using reinforcement learning, an AI method. According to Peter Dürr, a Sony AI researcher, there is no way to program a robot by hand to play table tennis; it must learn from experience. The robot's eight joints allow it to move the racket in various ways, and the cameras track the ball's logo to determine spin, which is crucial for returning shots.
Sony built an Olympic-sized table tennis court at its headquarters in Tokyo for the experiments. The technology could be used in manufacturing and other industries, and potentially in war, according to Michael Spranger, president of Sony AI. Spranger described speed as one of the fundamental issues in robotics today, especially in environments that are not fixed. He noted that while many factory robots are very fast, they repeat the same trajectory over and over. With this technology, Sony shows it is possible to train robots to be adaptive, competitive, and fast in uncertain environments that constantly change. Spranger also said it is very easy to build a superhuman table tennis robot. The experiments took place on a full-sized Olympic court, providing a realistic setting for the robot to learn and compete.
In a related development, a humanoid robot ran faster than the human world record in a half-marathon race for robots in Beijing on Sunday, according to multiple reports.