Integrated Real-Vehicle Fault-Tolerant Path Tracking Framework for Autonomous Vehicles

Abstract

Reliable steering actuation and accurate yaw feedback are critical for maintaining the lateral stability and path-tracking performance of autonomous vehicles. However, simultaneous faults in the steering actuator and global navigation satellite system (GNSS)-based yaw measurement channel can distort the steering input and corrupt the feedback information used for closed-loop lateral control, thereby degrading path-tracking performance and lateral stability. To address this problem, this paper presents an integrated framework for fault detection, fault reconstruction, and active fault-tolerant control in autonomous vehicles subjected to simultaneous steering actuator and yaw measurement faults. The proposed framework is developed based on an extended four-state lateral error model and employs two dedicated adaptive sliding mode observers (ASMOs) to reconstruct steering actuator and yaw measurement faults in parallel. Using the reconstructed fault information, actuator-side compensation is applied to the steering command, while sensor-side correction is applied to the corrupted yaw feedback signal. Consequently, a switching-free active fault-tolerant control structure is achieved without hardware redundancy or controller reconfiguration. To support reliable fault decisions under practical driving conditions, complementary fault-detection schemes are designed for the actuator and sensor fault channels. The proposed method is validated through real-vehicle experiments conducted at three proving grounds, namely KITECH, KIAPI, and JIAT, at speeds ranging from 10 to 70 km/h under simultaneous constant-bias fault conditions. The results demonstrate real-time fault detection and reconstruction in both fault channels and show that the proposed channel-wise compensation framework maintains stable compensated path-tracking performance under practical driving conditions.