Abstract: A new fault-tolerant estimating approach for unmanned aerial vehicle (UAV) dynamics in the presence of sensor/actuator faults is proposed, which is both adaptive and robust. This study discusses how to choose between adaptive and robust approaches when a sensor or actuator fails. We provide an adaptive method with Q-adaptation as well as a robust method with R-adaptation. The Kalman filter detects faults using the chi-square distribution of the normalized quadratic innovation function (NQI). After detecting a fault, it is recommended to run both R-adaptive and Q-adaptive Kalman filters concurrently and compare their estimation capabilities to distinguish between sensor and actuator faults. As a performance criterion, the mean-squared residuals between estimation and extrapolation values of robust and adaptive filters are proposed.

Keywords: Robust Adaptive Kalman Filter, R-adaptation, Q-adaptation, Unmanned Aerial Vehicle, sensor fault, actuator fault, normalized innovation

Pages: 157-169