Watch Google DeepMind's Robotic Ping-Pong Player Challenge Human Opponents

Demonstrations - Achieving Human-Level Competitive Robot Table Tennis

Ping-pong has emerged as the go-to sport for tech companies eager to demonstrate their robotic innovations. For instance, Japanese company Omron captured attention years ago with its ping-pong robot, capable of sustaining rallies with human opponents while showcasing its advanced sensor and control technologies.

With recent advancements in artificial intelligence (AI), we are witnessing the rise of more sophisticated robotic table tennis players that could genuinely challenge top human competitors.

A notable achievement comes from engineers at Google DeepMind. Their recent paper, “Achieving Human-Level Competitive Robot Table Tennis,” reveals their creation of an “amateur human-level player” that integrates AI with an industrial robotic arm equipped with a bat. A video showcases the AI-driven robot making rapid decisions to execute a variety of backhand and forehand shots. Impressively, it can return shots served with light spin, indicating its capacity to analyze the ball's spin and adjust its striking technique accordingly. The robot adeptly handles balls at varying speeds, approaches from all angles on the table, and even responds to high-bouncing shots. Its performance is truly remarkable.

“Achieving human-level speed and performance in real-world tasks is a significant goal for the robotics research community,” the researchers noted in their paper. “This work marks a pivotal step toward that objective, presenting the first learned robotic agent to achieve amateur human-level performance in competitive table tennis.”

In a series of test matches, the robot won every game against beginner-level human players, achieving a 100% victory rate, and secured 55% of games played against intermediate players. However, it still has much to improve, as it lost all matches against advanced players. Overall, the robot achieved a 45% win rate across the 29 games it participated in.