A novel global position system signal tracking loop under dynamic environments

In this paper, a neural-adaptive cubature Kalman filter (NACKF) algorithm is proposed for tracking the global position system (GPS) signal under dynamic environments. Although the adaptive cubature Kalman filter (ACKF)-based tracking loops overcome the inherent limitation of traditional cubature Kalman filter (CKF)-based tracking loops, which require the prior statistics of the measurement noise, the performance of ACKF-based tracking loops may degrade under dynamic environments due to the uncertainty of the system model such as mismodelling. To improve the performance of the ACKF-based tracking loop, a multilayer feed-forward neural network (MFNN) is embedded in the ACKF algorithm. In the proposed NACKF-based tracking loop, the multilayer feed-forward neural network (MFNN) is used to approximate the uncertainty of the system model; and the CKF is used for both MFNN online training and state estimation simultaneously. To effectively evaluate the performance of NACKF-based tracking loop this paper deeply compares the proposed method with the CKF-and ACKF-based tracking loops in medium dynamic environments and high dynamic environments, respectively. The analytical and experimental results showed that the performance of the NACKF-based tracking loop outperforms those of the CKF-and ACKF-based tracking loops under dynamic environments.