Reproduce “ROVER: Regulator-Driven Robust Temporal Verification of Black-Box Robot Policies” 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
Kristy Sakano, Jianyu An, Dinesh Manocha, Huan Xu
Researchers from the University of Maryland, College Park developed ROVER, an iterative framework for verifying black-box robot policies against formal Signal Temporal Logic (STL) specifications, enabling quantitative evaluation of compliance and targeted policy improvement. The method achieved an average 43.8% increase in satisfaction rates for safety rules across virtual racing and mobile robot navigation scenarios, with improvements translating to real-world robot behavior.
Key insights
- STL and its quantitative robustness semantics can effectively bridge the gap between abstract regulatory requirements and observable black-box robot behavior, providing detailed, trajectory-level insights.
- Decomposing robustness into TRV, LRV, and AVRV allows regulators to differentiate between common minor deviations, rare catastrophic failures, and overall compliance, guiding targeted policy improvements.
- An iterative feedback loop between regulators (defining safety rules and evaluating) and designers (updating policies based on quantitative scores and qualitative recommendations) systematically enhances robot safety and compliance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.