A new white paper from IEEE Spectrum and Wiley, sponsored by Noitom Robotics, describes an adaptive motion planning and tracking framework that lets humanoid robots learn tennis serving and rally styles from broadcast video and motion-capture data. The authors note that preserving a professional athlete's motion style while maintaining reliable task performance remains an open challenge in humanoid ball-sports robotics.
The framework uses a hierarchical design that separates a high-level planner, which generates stylistic kinematic motion, from a low-level tracker, which executes that motion with minimal interference. Because tracking performance degrades when a policy trained in simulation moves to physical hardware, the tracker is trained on randomized execution speeds and the planner is conditioned on a learned motion-speed representation. This reduces the compounding errors introduced by autoregressive planning and noisy perception.
The white paper documents real-world validation on a research humanoid platform and transfer to a full-size humanoid robot, including serving in an uncontrolled environment without any motion-capture system. It closes with practical lessons for engineers building future humanoid ball-sports systems, though the details are presented as part of a sponsored technical overview rather than a peer-reviewed study.