Runtime Monitoring and Fault Detection for Neural Network-Controlled Systems
This work addresses safety concerns for control systems using neural networks, but it is incremental as it builds on existing interval observer methods.
The paper tackles the problem of ensuring runtime safety for neural network-controlled nonlinear systems under disturbances and noise by designing a robust interval observer to generate precise bounds for monitoring and fault detection, demonstrating effectiveness in an adaptive cruise control simulation.
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 faults in the system outputs or actuators. An adaptive cruise control vehicular system is simulated to demonstrate effectiveness of the proposed design.