Fault Tolerant Estimation Under Uncertain Measurement Noise Conditions

The probabilistic-adaptive Kalman filter (p-AKF) algorithm is developed for fault-tolerant estimation of unmanned aerial vehicle (UAV) states in the face of measurement faults. The proposed p-AKF is based on calculating the posterior probability of normal operation of the system for the present measurement. It is suggested that this probability be computed using the posterior probability density of the normalized innovation sequence at the current estimation step. As a result, the system corrects defects in the estimating system while preserving the estimation's good behavior. The developed p-AKF algorithm is used for fault-tolerant UAV state estimation and tested under a variety of measurement failure scenarios, including continuous measurement bias, measurement noise increment, instantaneous abnormal measurements, and sensor zero-output signal.