Monitor and Recover: A Paradigm for Future Research on Distribution Shift in Learning-Enabled Cyber-Physical Systems

2025-04-18

AI Summary

Learning-based cyber-physical systems (CPS) exhibit insufficient robustness under distributional shift, and conventional “detect-and-abstain” paradigms severely degrade system availability. To address this, we propose a novel “monitor-and-recover” paradigm: formal safety monitoring replaces distributional shift detection, while feedback-driven, lightweight adaptive recovery supersedes passive abstention. Our approach achieves, for the first time, deep integration of safety monitoring with dynamic recovery mechanisms, synergistically incorporating online uncertainty quantification and model recalibration. Evaluated on two representative CPS benchmarks, the method delivers millisecond-scale response times, ensures continuous safe decision-making under shift, and significantly improves both availability and robustness. It overcomes fundamental limitations of existing approaches—namely, their reliance on explicit shift detection and output abstention—thereby advancing the state of safe, adaptive CPS operation.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Safety and Robustness
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

With the known vulnerability of neural networks to distribution shift, maintaining reliability in learning-enabled cyber-physical systems poses a salient challenge. In response, many existing methods adopt a detect and abstain methodology, aiming to detect distribution shift at inference time so that the learning-enabled component can abstain from decision-making. This approach, however, has limited use in real-world applications. We instead propose a monitor and recover paradigm as a promising direction for future research. This philosophy emphasizes 1) robust safety monitoring instead of distribution shift detection and 2) distribution shift recovery instead of abstention. We discuss two examples from our recent work.

Problem

• Addressing neural network vulnerability to distribution shift
• Improving reliability in learning-enabled cyber-physical systems
• Proposing monitor and recover over detect and abstain

Innovation

• Monitor and recover paradigm for reliability
• Robust safety monitoring over shift detection
• Distribution shift recovery instead of abstention