ETH Zurich’s badminton-playing robot suggests ESPN for physical AI isn’t coming soon

Scientists at ETH Zurich have developed ANYmal, a quadruped robot that can play badminton using AI-powered perception and movement skills. The robot represents a breakthrough in combining real-time visual processing with physical agility, though its performance against human players reveals significant limitations that highlight ongoing challenges in robotics.

How it works: ANYmal resembles a miniature giraffe holding a badminton racket in its teeth, built on a 50-kilogram industrial platform originally designed for oil and gas applications by ANYbotics, an ETH Zurich spinoff company.

• The robot uses a stereoscopic camera (which creates depth perception like human eyes) for shuttlecock tracking and environmental sensing, paired with an articulated arm that swings the racket.
• Instead of traditional model-based control systems, the team trained ANYmal using reinforcement learning in a simulated badminton court where the robot learned through trial and error.
• During training, the robot learned to predict shuttlecock trajectories and successfully hit returns six times in a row while discovering its physical limitations.

What the robot learned: ANYmal developed strategic behaviors that mirror human badminton players through its AI training process.

• The robot figured out that moving back to center court and toward the backline after successful returns was the optimal strategy.
• It learned to stand on its hind legs to get better views of incoming shuttlecocks.
• ANYmal developed risk assessment skills, avoiding impossible plays that could cause damage while remaining competitive within its capabilities.

The performance gap: When tested against human players, ANYmal’s limitations became apparent, particularly in reaction time and perception accuracy.

• The robot needed roughly 0.35 seconds to process and respond to shuttlecocks, compared to elite human players who react in 0.12–0.15 seconds.
• “The robot localized the shuttlecock with the stereo camera and there could be a positioning error introduced at each timeframe,” explained Yuntao Ma, the roboticist who led the project.
• The camera’s limited field of view restricted how long ANYmal could track shuttlecocks before needing to act.

Why this matters: The research addresses a fundamental challenge in robotics—integrating perception with movement at speeds that approach human reflexes.

• “I wanted to fuse perception and body movement,” Ma said, highlighting the importance of developing robotic equivalents to human reflexes.
• The work demonstrates the trade-offs robots face between movement speed and perception accuracy, as faster movement creates camera shake that reduces tracking precision.

What’s next: The research team has identified several improvements that could enhance ANYmal’s performance in future iterations.

• Reaction times could improve by predicting shuttlecock trajectories based on opponents’ body positions rather than waiting to see the shuttlecock.
• Event cameras with microsecond-range latencies could dramatically improve the robot’s visual processing capabilities.
• Faster, more capable actuators could reduce the physical limitations that currently constrain the robot’s performance.

The bigger picture: While ANYmal won’t be competing in professional badminton tournaments anytime soon, its training framework has broader applications.

• “I think the training framework we propose would be useful in any application where you need to balance perception and control. Picking objects up, even catching and throwing stuff,” Ma suggested.
• The research contributes to the ongoing effort to develop robots with more natural, responsive interactions with dynamic environments.

What they’re saying: Ma acknowledged the robot’s current limitations while remaining optimistic about the underlying approach.

• “Overall, it was suited for more friendly matches—when the human player starts to smash, the success rate goes way down for the robot,” he said.
• When asked about commercializing badminton-playing robots, Ma was realistic: “Would I set up a company selling badminton playing robots? Well, maybe not.”