Reproduce “Mission-Level Runtime Assurance Framework for Autonomous Driving” 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
Chieh Tsai, Salim Hariri
The authors introduce a successor-safe runtime monitor that checks whether autonomous-driving commands preserve platform safety and mission feasibility before execution. In randomized highway-env mission-fault rollouts, it reports mission success of 0.73 ± 0.08 and mission-infeasibility detection of 0.59 ± 0.09, versus 0.44 ± 0.09 and 0.00 ± 0.00 for adapted Simplex platform-RTA. These are simulator benchmark rates, not real-road probabilities or guarantees for arbitrary missions.
Key insights
- Platform-level runtime safety alone is insufficient to guarantee mission success; a dedicated mission-level assurance mechanism is required to prevent strategic failures.
- Proactively evaluating the predicted future evolution of high-level commands enables early detection and rejection of mission-infeasible behaviors before they become unrecoverable.
- The successor-safe RTAA framework effectively integrates with diverse planning sources, including learning-enabled controllers, to enhance both mission success and fault detection.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.