Reproduce “Predictive Semantic Safety: From Visual Physical Reasoning to Safety-Critical Control” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Taekyung Kim, Salem Fradi, Yanning Dai, Mateusz Ostaszewski, Jürgen Schmidhuber
The paper presents PSS, a safety filter that uses visual event predictions to estimate future hazard occupancy and test a prescribed braking maneuver against it. A conformal margin accounts for prediction error. In simulated Unitree Go1 navigation, PSS achieved a 99.3% safe episode rate (149/150), versus 43.3% for Backup CBF using geometry; this evidence is limited to simulation and modeled objects.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.