Toward Uncertainty-Aware, Physics-Grounded Human World Models
Toward Uncertainty-Aware, Physics-Grounded Human World Models
This paper proposes a framework to couple two existing components for building uncertainty-aware, physics-grounded human world models. The first component, KNOWN, reconstructs 3D human bodies in real time while estimating aleatoric and epistemic uncertainty, reducing error on low-density inputs from 81.3 mm to 70.3 mm. The second, PhysMoP, predicts future body configurations by solving Euler-Lagrange equations with learned dynamics, reducing prediction error from 9.6 mm to 2.1 mm on Human3.6M and from 12.1 mm to 0.7 mm on 3DPW. The authors define an interface between the two components, propose three mechanisms for integrating KNOWN's uncertainty into PhysMoP, and specify an evaluation protocol on in-the-wild video. No new experimental results are reported.
The authors propose a coupling of two existing components, KNOWN for uncertainty-aware 3D human reconstruction and PhysMoP for physics-grounded motion prediction, defining an interface and evaluation protocol without reporting new experimental results.
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.
Reported numbers
uncertainty-aware inference time
11.6
View original evidence
certainty in real time (11.6 ms per frame); its aleatoric term identifies unreliable observations, its epistemic term identifies inputs from low-density regions of the training data, and refinement driven by the epistemic term reduces the error on such inputs from 81.3 to 70.3 mm. PhysMoP [2] predicts future body configuration by solving the Euler-Lagrange equations of the body with learned dynamics terms, reducing t
Open source S3prediction inference time
80
View original evidence
he 80 ms prediction error from 9.6 to 2.1 mm on Human3.6M and from 12.1 to 0.7 mm on 3DPW. We define the interface between the two, propose three mechanisms by which KNOWN's uncertainty enters PhysMoP, and specify an evaluation protocol on in-the-wild video. No new experimental results are reported here.
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
he 80 ms prediction error from 9.6 to 2.1 mm on Human3.6M and from 12.1 to 0.7 mm on 3DPW. We define the interface between the two, propose three mechanisms by which KNOWN's uncertainty enters PhysMoP, and specify an evaluation protocol on in-the-wild video. No new experimental results are reported here.
Open source S4
certainty in real time (11.6 ms per frame); its aleatoric term identifies unreliable observations, its epistemic term identifies inputs from low-density regions of the training data, and refinement driven by the epistemic term reduces the error on such inputs from 81.3 to 70.3 mm. PhysMoP [2] predicts future body configuration by solving the Euler-Lagrange equations of the body with learned dynamics terms, reducing t
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org