Adaptive Neural Network‐Based Control for Vehicle Under Actuator Faults and Stealthy Replay Attack Threats

With the development of intelligent and connected vehicles, actuator faults and stealthy replay attacks have become important threats to vehicle control systems. Their coupled effects pose significant challenges to system reliability and security. To address these challenges, this paper proposes a unified adaptive neural‐network‐based fault‐tolerant and intrusion‐tolerant control framework for vehicles. Within this framework, a comprehensive vehicle dynamics model is established by considering both actuator faults and replay cyber attacks. An adaptive neural network‐based fault compensation scheme is developed, in which the radial basis function neural network (RBFNN) is employed to approximate actuator faults, while the nonlinear disturbance observer (NDO) is used to estimate unknown disturbances and approximation errors. In addition, a dynamic watermarking‐based intrusion‐tolerant control strategy is designed to detect replay attacks, isolate compromised signals, and improve the accuracy of Kalman filter based state estimation. By replacing compromised measurements with estimated states, the proposed method achieves intrusion‐tolerant control while preserving control performance. Finally, hardware‐in‐the‐loop (HIL) experiments are conducted to validate the effectiveness of the proposed control strategy.