Humanoid robots learn tennis serves and rallies from pro motion data
Overview
A hierarchical framework trains humanoid robots to play tennis using pro players' motion data, according to IEEE Spectrum.
The system splits control into a planner that generates stylistic tennis motion and a tracker that executes it. The tracker is trained on randomized execution speeds, and the planner uses a learned motion-speed representation to reduce errors when moving from simulation to the real world.
The article says the framework was validated on a research humanoid and on a full-size humanoid serving outdoors without motion capture.
Written by AI from the articles below · updated Oct 12, 6:54 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- IEEE Spectrum · AINewsHumanoid robots learn tennis serves and rallies from pro motion data
AIA hierarchical framework separates a planner that generates stylistic tennis motion from a tracker that executes it, the article says. The tracker is trained on randomized execution speeds and the planner uses a learned motion-speed representation to reduce sim-to-real errors. The framework was validated on a research humanoid and a full-size humanoid serving outdoors without motion capture.
Heat trend
Not enough continuous observations to show a trend yet.