Enhanced fault detection and exclusion based on Kalman filter with colored measurement noise and application to RTK

With the development of high-precision safety–critical applications using global navigation satellite systems (GNSS), fault detection and exclusion (FDE) is indispensable to guaranteeing the integrity of a GNSS positioning and navigation system. Many FDE algorithms have been developed based on the standard Kalman filter (KF), assuming that GNSS measurements come with Gaussian uncorrelated white noise. The existence of colored noise in GNSS measurements, which is typical for positioning with low-cost receivers and in challenging environments will, however, degrade the performance of KF-based FDE algorithms. We proposed an FDE scheme based on improved KF considering colored noise (CKF) as a first-order autoregressive model to improve the FDE performance. The performance of the proposed CKF-based FDE algorithm was evaluated with an application to real-time kinematic positioning using a low-cost receiver. A CKF-based fault detection test, a fault identification test, a minimum detectable bias (MDB), error distribution, and positioning results were examined. The results showed that the CKF-based FDE can obtain realistic statistical information to improve integrity monitoring reliability. The fault detection test achieved a 17.83% improvement in FDE performance and a reduction in the false alarm rate, from 23.33 to 5.50%, compared with KF-based FDE. The tests also indicated that the CKF-based FDE can detect multiple faults with zero-miss detection. The fault identification test had an average improvement of 32.14%, and a more realistic MDB was obtained. The results of this study contribute to making objective decisions for the integrity monitoring of practical, precise GNSS positioning.