What Ace is like inside, the AI-trained robot that already beats professional table tennis players

It is about a robot with nine cameras and gaze control of the opponent that is capable of responding to a play in microseconds without departing from the official regulations of that sport, according to the British magazine Nature in his publication this Wednesday.

Nature took out the Ace robot like protagonist of the cover. And specialists agree that advances in the field of artificial intelligence (AI) have already surpassed the stage of data processing and have moved towards the execution of complex movements, and with commendable precision achieved in a matter of microseconds.

Ace is a creation of Sonythe Japanese technology company, who announced the launch explaining that it is a robotic system to play ping-pong matches against professional table tennis players.

And its creators explain why Ace is dedicated to table tennis. “This sport requires a extreme precision: the ball moves at linear speeds greater than 20 m/s and with a rotation that exceeds 160 revolutions per second. The effect is used to make it difficult to return blows or to obtain a tactical advantage,” they explained.

“To respond, expert players must master a series of skills that allow them to follow, react and generate high-speed, high-spin hits“, they point out about the development of the new robot.

“Through a novel control algorithm based on the reinforcement learningAce can adapt to unpredictable physical dynamics of a live exchange of blows, continually updating its trajectory and strategy based on the observed flight and spin of the ball,” they describe.

Ace consists of a system of high speed perception which uses a network of nine cameras with image sensors and gaze control, an AI-based control systemas well as a high-speed robotic arm with eight joints for agile physical interaction.

Initially, he was tested in matches in the Japanese professional league following the rules of ping-pong, but in a short time he acquired a speed of learning that now matches elite players.

Among its rivals described in the article are five elite players (each with more than 10 years of experience and an average of 20 hours of weekly training) and two professionals (Minami Ando and Kakeru Sone).

The robot won three of the five matches against elite players, and lost both matches against professionals although he managed to snatch a set from one of them.

“Achieve unprecedented ball tracking thanks to a system of hybrid vision which uses a total of 12 high-speed sensors. The system consists of three vision sensors that capture motion with sub-millisecond precision by only recording brightness changes at the pixel level. This allows Ace to track the 40mm ball and measure spins exceeding 9,000 revolutions per minute and can locate the ball in three-dimensional space (3D) at 200 Hz with an error of 3.0 millimeters”, Sony expands on its robotic development.

Nature He also assured that Ace is capable of not only rallying with players and responding with spin hits, but also can serve and initiate points competitively.

In this regard, the creative company says: “In each hit, you test a different skill, which is then transmitted to the control policy for execution. It is not systematized. “This ensures variety in gameplay and keeps human opponents guessing about what the robot’s next move will be.”

“It is very significant that athletes can learn new skills simply by playing against Ace and observing him, which indicates that AI-controlled robotic systems could constitute an area for human development beyond table tennis,” say Carlos Ribeiro and Esther Colombini, Brazilian researchers, in a reaction to the study reported in Nature.

With improvements, “the robot achieved even higher hitting speeds, more aggressive placement near the edge of the table and faster blow exchanges, reflecting continuous improvement in performance in competitive conditions,” they detail.

Ribeiro and Colombini agree: “Ace is an important milestone, showing the potential of the next generation of highly capable competitive robots that interact with physical environments.”