HLO-AMP: Hierarchical Loco-Manipulation from Object-Aware Motion Priors
HLO-AMP: Hierarchical Loco-Manipulation from Object-Aware Motion Priors
This research introduces HLO-AMP, a hierarchical pipeline for humanoid loco-manipulation that trains a low-level controller as an object-aware motion prior using retargeted motion-capture references with the manipulated object physically present in the scene. A reference-free high-level policy is trained on top to generalize these priors to long-horizon multi-stage tasks while preventing off-distribution queries. The framework is instantiated on a Unitree G1 for bimanual box pick-and-place, showing generalization across object masses and dimensions, initial and goal poses, and reactive recovery from intentional drops mid-carry.
The authors present a hierarchical loco-manipulation framework trained with object-aware motion priors on a Unitree G1 humanoid, demonstrating generalization across object masses and dimensions as well as reactive recovery from intentional drops mid-carry.
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.
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
Source：arXiv Robotics — research abstracts · arxiv.org