Learning Contextual Runtime Monitors for Safe AI-Based Autonomy

Researchers introduce a novel framework for learning context-aware runtime monitors for AI-based control systems in autonomous vehicles. The approach uses contextual multi-armed bandits to select the best controller for current conditions rather than averaging outputs, providing theoretical safety guarantees and improved performance in simulated driving scenarios.

• →New framework addresses safety concerns in AI-controlled autonomous systems by using context-aware monitoring instead of traditional ensemble averaging.
• →The approach reformulates safe AI control as a contextual monitoring problem using multi-armed bandit techniques.
• →Framework provides theoretical safety guarantees during controller selection while better utilizing controller diversity.
• →Validation in autonomous driving simulations shows significant improvements in both safety and performance over baseline methods.
• →Research addresses the critical challenge of AI controller accuracy degradation in unfamiliar environments.