Tennis is a demanding sport even for humans. It’s even more so for robots to master. LATENT is a system that enables humanoid robots to learn the game with imperfect data. It learned how to play tennis using only 5 hours of motion capture data. It can sustain multi-shot rallies, traveling balls at speeds over 15 m/s.
Researchers trained a humanoid robot to play tennis using only 5 hours of motion capture data
The robot can now sustain multi-shot rallies with human players, hitting balls traveling >15 m/s with a ~90% success rate
AlphaGo for every sport is coming pic.twitter.com/R5OhqQOT8A
— Andrew Kang (@Rewkang) March 15, 2026
LATENT features reactive footwork. Its imperfect data “consists only of motion fragments that capture the primitive skills used when playing tennis rather than precise and complete human-tennis motion sequences from real-world tennis matches.”
[HT]
