Robust Augmented State Extended Kalman Filter for Actuator/Sensor Fault Detection and Isolation in Quadrotor Unmanned Aerial Vehicles

Abstract

Abstract. This study introduces a robust augmented state extended Kalman filter (RASEKF) for sensor and actuator fault detection and isolation (FDI) in quadrotor unmanned aerial vehicles (UAVs) to enhance operational safety and reliability. A general six degrees-of-freedom (6DOF) nonlinear quadrotor model is presented, where actuator faults are modeled as control effectiveness losses and incorporated as augmented system states. RASEKF directly estimates actuator faults while prioritizing sensor FDI to ensure accurate fault estimation, thereby reducing false alarms and missed detections of faults. Unlike the standard extended Kalman filter (EKF) and other robust extended Kalman filters (REKFs), the RASEKF incorporates a dual-threshold mechanism for measurement weighting and adaptive filter gain adjustment, significantly improving sensitivity to slowly growing ramp faults, one of the key challenges in sensor fault detection. The enhanced robustness of the RASEKF enables accurate state estimation even in the presence of sensor faults. Extensive numerical simulations validate their effectiveness in detecting, isolating, and excluding simultaneous step and ramp faults in sensors, alongside precise estimation of single and concurrent actuator faults. The results further indicate that RASEKF achieves a 69% reduction in fault detection time for altitude measurement and a 58% decrease in root-mean-square error (RMSE) for yaw angle estimation compared to EKF and REKF, demonstrating its superior fault detection and robust estimation performance.