Runtime Monitoring and Fault Detection for Neural Network-Controlled Systems

arXiv:2403.16132v1 Announce Type: cross
Abstract: There is an emerging trend in applying deep learning methods to control complex nonlinear systems. This paper considers enhancing the runtime safety of nonlinear systems controlled by neural networks in the presence of disturbance and measurement noise. A robustly stable interval observer is designed to generate sound and precise lower and upper bounds for the neural network, nonlinear function, and system state. The obtained interval is utilised to monitor the real-time system safety and detect …

abstract arxiv control cs.lg cs.sy deep learning detection eess.sy generate interval measurement monitoring network networks neural network neural networks noise paper safety systems trend type