Humanoid robot control achieves stable motion from sparse keyframes
PLAT: Sparse Timed Keyframe Motion Tracking for Humanoid Control via Privileged Latent Transition Learning
Robotics
Summary
Robots that walk and move like humans usually need detailed instructions for every moment, which is hard to plan or adjust on the fly. This paper presents a new method where a robot only gets a few key target positions and times, and it figures out how to move smoothly in between. The authors created a system that learns from detailed example movements during training but only needs these sparse instructions later, allowing more flexible and stable whole-body motion. They tested it on a simulated robot and a real humanoid robot, showing it works well even when the goals are set far apart in time.
What this means in practice
- •For robotics engineers: Control humanoid robots using only sparse timed keyframes for smoother and more stable motions over long plans.
- •For robotic system integrators: Deploy efficient humanoid motion controllers that reduce dependency on dense motion references, simplifying integration with high-level planners.
Authors
Zepeng Wang, Jiangxing Wang, Chao Ma, Xiaochuan Shi, Zongqing Lu
Abstract
Humanoid motion tracking policies rely on dense frame-by-frame references, limiting their use as high-level motion controllers for planning and interactive motion generation. We study \emph{Sparse Timed Keyframe Motion Tracking}, where a policy receives only sparse future keyframes and their desired arrival times, and must execute stable whole-body motions that reach successive goals. We propose \textbf{PLAT}, a three-stage sparse timed keyframe motion tracking policy learning framework with \textbf{P}rivileged \textbf{LA}tent \textbf{T}ransition learning. PLAT bridges dense motion tracking and sparse goal-conditioned control by exploiting dense goal sequences as privileged supervision during training while requiring only sparse timed keyframe commands at deployment. A pretrained dense tracking expert first provides robust motion priors. A privileged latent prior is then learned through DAgger-style imitation, followed by latent residual reinforcement learning that refines latent transitions instead of directly optimizing actions. Extensive simulation experiments demonstrate that PLAT maintains accurate and stable sparse timed keyframe tracking across varying planning horizons, with particularly strong performance under long-horizon commands. Successful deployment on a Unitree G1 humanoid robot further demonstrates the effectiveness and practicality of PLAT for sparse humanoid motion control.