When Delay Meets Drift: Physical Safety of Robot Foundation Models Is a Cross-Layer Problem
When Delay Meets Drift: Physical Safety of Robot Foundation Models Is a Cross-Layer Problem
This research paper argues that physical safety in robot foundation models is not an isolated property but emerges from cross-layer interactions between environmental variations, model inference delays, and execution conditions.
The study demonstrates that physical safety risks in robot foundation models emerge from cross-layer interactions between environmental variations, model inference delays, and execution conditions, quantified through closed-loop experiments on the pi0.5 model.
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
inference delay duration
400
View original evidence
l environment, the model, the computing platform, and execution interact in closed loop. We formalize this as cross-layer variation propagation across these system layers and use an interaction contrast to measure when joint variation creates more risk than the sum of its parts. In 4,000 paired closed-loop episodes of the pi0.5 vision-language-action model on LIBERO, a 400 ms inference delay raises the safety-violati
Open source S3paired closed-loop episodes count
4,000
Reported trials
View original evidence
l environment, the model, the computing platform, and execution interact in closed loop. We formalize this as cross-layer variation propagation across these system layers and use an interaction contrast to measure when joint variation creates more risk than the sum of its parts. In 4,000 paired closed-loop episodes of the pi0.5 vision-language-action model on LIBERO, a 400 ms inference delay raises the safety-violati
Open source S3
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
on rate by 3.5 percentage points under the training viewpoint but by 37 percentage points under a 60-degree camera shift: 33.5 percentage points (95% CI: 25-43) more than the two effects combined. We close with a research agenda toward predictable physical intelligence: predicting when variation becomes risk, and adapting compute, power, and control accordingly.
Open source S4
l environment, the model, the computing platform, and execution interact in closed loop. We formalize this as cross-layer variation propagation across these system layers and use an interaction contrast to measure when joint variation creates more risk than the sum of its parts. In 4,000 paired closed-loop episodes of the pi0.5 vision-language-action model on LIBERO, a 400 ms inference delay raises the safety-violati
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org