A Unified Inertial Encoder for Human and Humanoid Motion
A Unified Inertial Encoder for Human and Humanoid Motion
This paper introduces a unified inertial encoder that leverages kinematic IMU synthesis to align human and humanoid motion representations. By jointly training on human datasets and retargeted humanoid motions, the model achieves zero-shot transfer to a physical Unitree G1 robot without physics simulation or fine-tuning. The approach demonstrates superior velocity estimation accuracy (0.099 m/s error) across five datasets and enables language-grounded motion understanding for motion categories not present in the original human training data.
The authors propose a unified inertial encoder that achieves zero-shot transfer to a physical Unitree G1 robot, reporting a velocity error of 0.099 m/s on real-robot data without physics simulation or fine-tuning.
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.
- Data origin
- Robot trajectoriesOpen source S4
Reported numbers
velocity error
0.099
View original evidence
nitree G1 without physics simulation, real-robot training trajectories, or fine-tuning, achieving the lowest velocity error among evaluated methods across five datasets, including 0.099m/s on real G1. Its motion-description retrieval also extends to retargeted G1 motion categories observed only in human training, demonstrating how human motion knowledge can support the generalization of humanoid inertial odometry and
Open source S4datasets evaluated
5
View original evidence
nitree G1 without physics simulation, real-robot training trajectories, or fine-tuning, achieving the lowest velocity error among evaluated methods across five datasets, including 0.099m/s on real G1. Its motion-description retrieval also extends to retargeted G1 motion categories observed only in human training, demonstrating how human motion knowledge can support the generalization of humanoid inertial odometry and
Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
nitree G1 without physics simulation, real-robot training trajectories, or fine-tuning, achieving the lowest velocity error among evaluated methods across five datasets, including 0.099m/s on real G1. Its motion-description retrieval also extends to retargeted G1 motion categories observed only in human training, demonstrating how human motion knowledge can support the generalization of humanoid inertial odometry and
Open source S4
arXiv:2611.10137v1 Announce Type: new Abstract: Learning humanoid state estimation and language-grounded motion understanding from robot recordings ties training coverage to the robot's motion repertoire. Human motion datasets offer broader coverage, yet transferring this knowledge requires overcoming differences in body geometry, sensing, and physical disturbances. To enable this transfer, we propose a unified inert
Open source S2
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Match the buyer’s robot embodiment, action space, sensors and timestamp requirements before planning collection.
- Compare the reported units and scope before using these quantities in a budget. Recording hours, sensor-hours and trajectories are different measures.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：arXiv Robotics — research abstracts · arxiv.org