AI robot beats elite table tennis players
A research team involving Sony Group Corp.’s artificial intelligence (AI) units has developed an AI-powered robot capable of competing with and beating skilled human table tennis players.
Although still unable to defeat professional athletes, the autonomous robot “Ace,” developed by Sony AI Inc., demonstrated a high level of performance, including the ability to handle spinning balls, according to an article published in the April 22 edition of British scientific journal Nature.
Ace is equipped with an eight-jointed arm and uses multiple cameras positioned around the table to track the ball’s high-speed movement, with its actions controlled by AI.
Complex trajectories
“For more than 40 years, roboticists have chased a classic challenge: how to build a machine that can rally with a human. And not only that, one capable of perceiving, reacting to, and returning the blistering speed and spin of elite-level table tennis,” Sony AI said on its website. “Ace meets this challenge.”
Matches against human players were conducted in accordance with International Table Tennis Federation rules.
Ace played against five elite players, all with over 10 years of competitive table tennis experience, defeating three of them.
In the best-of-five games, it lost to two active professional players, but won one out of the seven games played.
The robot scored points with a variety of spin types while returning a wide range of spins, according to the article. It also reacted to balls that change trajectory after hitting the net.
‘Fast, real-time control’
Table tennis involves fast-paced rallies and requires players to predict complex ball trajectories, including spin.
Given the sport’s challenging nature for robots, their performance has been tested, including by using equipment that reduces spin and speed.
“Ace challenges elite and professional players using unaltered, professional-level equipment and rules, demonstrating for the first time, to our knowledge, that it is possible for AI systems to outperform human athletes in interactive, physical skill-based games,” the article said.
The success of the benchmark in robot table tennis suggests that “similar techniques apply to other areas featuring fast, real-time control and human interaction, including, for example, manufacturing and service robotics,” it added.
