Learning Contextual Runtime Monitors for Safe AI-Based Autonomy

TL;DR

This paper presents a framework for learning context-aware runtime monitors that select the most suitable AI controller in real-time, enhancing safety and performance in autonomous systems.

Contribution

It introduces a novel contextual monitoring approach using multi-armed bandit techniques to improve safety and controller utilization in AI-based control ensembles.

Findings

Significant safety improvements in autonomous driving simulations.

Enhanced performance by selecting contextually appropriate controllers.

Theoretical safety guarantees during controller selection.

Abstract

We introduce a novel framework for learning context-aware runtime monitors for AI-based control ensembles. Machine-learning (ML) controllers are increasingly deployed in (autonomous) cyber-physical systems because of their ability to solve complex decision-making tasks. However, their accuracy can degrade sharply in unfamiliar environments, creating significant safety concerns. Traditional ensemble methods aim to improve robustness by averaging or voting across multiple controllers, yet this often dilutes the specialized strengths that individual controllers exhibit in different operating contexts. We argue that, rather than blending controller outputs, a monitoring framework should identify and exploit these contextual strengths. In this paper, we reformulate the design of safe AI-based control ensembles as a contextual monitoring problem. A monitor continuously observes the system's context and selects the controller best suited to the current conditions. To achieve this, we cast monitor learning as a contextual learning task and draw on techniques from contextual multi-armed bandits. Our approach comes with two key benefits: (1) theoretical safety guarantees during controller selection, and (2) improved utilization of controller diversity. We validate our framework in two simulated autonomous driving scenarios, demonstrating significant improvements in both safety and performance compared to non-contextual baselines.