# Learning Contextual Runtime Monitors for Safe AI-Based Autonomy

arXiv: https://arxiv.org/abs/2601.20666 · HTML: https://mlsanity.com/paper/2601.20666/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 49 references that are in the map.
- First posted on arXiv 28 Jan 2026.

## Builds on (most cited first, 22 of 22)

- [CARLA: An Open Urban Driving Simulator](https://mlsanity.com/paper/1711.03938/) (2017)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [End to End Learning for Self-Driving Cars](https://mlsanity.com/paper/1604.07316/) (2016)
- [Safe Reinforcement Learning via Shielding](https://mlsanity.com/paper/1708.08611/) (2017)
- [Deep Bayesian Bandits Showdown: An Empirical Comparison of Bayesian Deep Networks for Thompson Sampling](https://mlsanity.com/paper/1802.09127/) (2018)
- [Improved Optimistic Algorithms for Logistic Bandits](https://mlsanity.com/paper/2002.07530/) (2020)
- [The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems](https://mlsanity.com/paper/2309.05837/) (2023)
- [Scenic: A Language for Scenario Specification and Scene Generation](https://mlsanity.com/paper/1809.09310/) (2018)
- [Neural Contextual Bandits with Deep Representation and Shallow Exploration](https://mlsanity.com/paper/2012.01780/) (2020)
- [Randomized Exploration in Generalized Linear Bandits](https://mlsanity.com/paper/1906.08947/) (2019)
- [Active Preference Optimization for Sample Efficient RLHF](https://mlsanity.com/paper/2402.10500/) (2024)
- [Jointly Efficient and Optimal Algorithms for Logistic Bandits](https://mlsanity.com/paper/2201.01985/) (2022)
- [Towards Verified Artificial Intelligence](https://mlsanity.com/paper/1606.08514/) (2016)
- [Improved Regret Bounds of (Multinomial) Logistic Bandits via Regret-to-Confidence-Set Conversion](https://mlsanity.com/paper/2310.18554/) (2023)
- [Formal Synthesis of Stochastic Systems via Control Barrier Certificates](https://mlsanity.com/paper/1905.04585/) (2019)
- [Formal Analysis and Redesign of a Neural Network-Based Aircraft Taxiing System with VerifAI](https://mlsanity.com/paper/2005.07173/) (2020)
- [Pure Exploration in Bandits with Linear Constraints](https://mlsanity.com/paper/2306.12774/) (2023)
- [SMARLA: A Safety Monitoring Approach for Deep Reinforcement Learning Agents](https://mlsanity.com/paper/2308.02594/) (2023)
- [Autonomous Driving With Perception Uncertainties: Deep-Ensemble Based Adaptive Cruise Control](https://mlsanity.com/paper/2403.15577/) (2024)
- [Dynamic-Weighted Simplex Strategy for Learning Enabled Cyber Physical Systems](https://mlsanity.com/paper/1902.02432/) (2019)
- [Sketch and Refine: Towards Fast and Accurate Lane Detection](https://mlsanity.com/paper/2401.14729/) (2024)
- [An Efficient Risk-aware Branch MPC for Automated Driving that is Robust to Uncertain Vehicle Behaviors](https://mlsanity.com/paper/2403.18695/) (2024)

## Built on by (most cited first, 1 of 1)

- [Optimizing Minimax Regret in Uncertain MDPs with Small Sets of Policies](https://mlsanity.com/paper/2608.02509/) (2026)

## Abstract (arXiv)

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.