DKWBC: Decoupling Motion Tracking and Balancing for Humanoid Whole-Body Control
DKWBC: Decoupling Motion Tracking and Balancing for Humanoid Whole-Body Control
This paper introduces DKWBC, a dynamics-informed whole-body control framework for humanoids that separates motion tracking from balancing. By using inverse-dynamics feedforward for tracking torques and a learned residual policy for balance corrections, the method simplifies learning and reduces high-frequency fluctuations in control outputs. Experiments in simulation and on real hardware show that the single learned balancing policy generalizes across diverse motion datasets while maintaining accurate tracking and balance under various conditions including ground support, attached payloads, and physical interactions.
The authors propose DKWBC, a framework that decouples motion tracking from balancing in humanoid control by combining inverse-dynamics feedforward with a learned residual policy for balance corrections.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- code: not_reported
- weights: not_reported
- dataset: not_reported
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
e that a single learned balancing policy generalizes across motion datasets while maintaining accurate motion tracking and balance across ground support, attached payloads, and physical interaction. Project page: https://robbyant-research.github.io/DKWBC/.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org