ETH Zurich's Anymal Robot Learns To Play Badminton Using Reinforcement Learning

September 5, 2025 by Mai Tao

ETH Zurich researchers have pushed their quadruped robot, ANYmal, into new territory: the badminton court.

A new video released by the university shows the four-legged robot attempting to return shots using a racket, advanced vision systems, and reinforcement learning-based controls. (See video below.)

ANYmal, built by ANYbotics , developed as a versatile quadruped platform, uses dual cameras to track the shuttlecock in flight. The system looks for the shuttle's color, calculates its trajectory, and then directs the robot's motors to move into position and execute a return shot.

“The control of the robot is based on reinforcement learning,” a researcher explains in the video.“That means taking a robot and throwing it into a simulator, where it can do thousands of iterations of trial and error, slowly learning what is the best strategy.”

The project highlights the difficulty of teaching robots to play fast-paced sports. Badminton requires rapid reactions, precise limb coordination, and high-speed dynamic control – all challenges for current robotics systems.

Researchers compare the robot's ability to that of“a seven-year-old kid playing with their parent”.

While the robot is far from competitive against human players, the exercise is more than a novelty. ETH Zurich says the lessons learned from controlling ANYmal under such demanding conditions feed back into broader robotics research, improving locomotion, coordination, and system design.

For now, it's just“friendly matches” – but the team is working toward a future where a robot could be a serious opponent on the court.

The research is published in Science Robotics .

MENAFN05092025005532012229ID1110023127