A Novel Method for Navigation Satellite Clock Bias Prediction Considering Stochastic Variation Behavior

Abstract

Considering that the satellite clock bias (SCB) is composed of a trend part, periodic item and stochastic component, a novel method incorporating the polynomial model adding a few cyclic terms and least squares support vector machines (LS-SVM) for clock bias prediction is proposed in this paper. The trend part and periodic item are firstly modeled by the polynomial model adding a few cyclic terms according to the physical characteristics of satellite clocks. Then, the polynomial fitting residuals, namely the stochastic component, is modeled based on the LS-SVM. Finally, the forecasted results for the trend part, periodic item and stochastic component are aggregated to produce the final prediction value for the clock bias. For verification and testing, the GPS clocks are taken as an example and the short-term prediction experiments are carried out. The simulation results have demonstrated that the proposed prediction method outperforms the IGU-P solutions at least on a daily basis.