APPLICATION OF WAVELET BASED NEURAL NETWORK TO MODELING OF L1 GPS TIMING ERROR

摘要

The fact that Global Positioning System (GPS) time is based on atomic clocks, plus the fact that the GPS satellite ephemerides are accurately known, leads to some significant national and international time comparison opportunities. Even though GPS is fundamentally a navigation system, accurate time is also available. Wavelet Neural Network (WNN) possesses the best function approximation ability, that is to say it has the ability to identify the model. Because the constricting model algorithm is different from common Artificial Neural Network (ANN) back propagation algorithm, it can effectively overcome intrinsic defect of common ANN. Therefore the better modeling and prediction effect can be reached effectively. This paper proposed a method of prediction and modeling of L1 GPS receiver timing error based on WNN that enables prediction model to have not only wavelet good approximation properly, but also NN self-learning adaptive quality. The experimental results on collected real data demonstrate the suitability of this method in improving of timing error of single frequency GPS receivers. The experiments show that GPS receiver timing RMS error can reduce from 300nsec and 200nsec to less than 183nsec and 51nsec, before and after SA, respectively.