A Practical Adaptive Clock Offset Prediction Model for the Beidou-2 System

The predicted navigation satellite clock offsets are crucial to support real-time global navigation satellite system (GNSS) precise positioning applications, especially for those applications difficult to access the real-time data stream, such as the low earth orbit (LEO) autonomous precise orbit determination. Currently, the clock prediction for the Chinese BeiDou system is still challenging to meet the precise positioning requirement. The onboard clocks of BeiDou satellites are provided by different manufacturers, and the clocks' switch events are more frequent. Considering the satellite-specified and temporal variation of the BeiDou clocks characteristics, we intend to use an adaptive model for BeiDou clock prediction. During clock prediction, we identify different models for BeiDou clocks' characteristics, and then address the optimal model with a cross-validation procedure. The model achieving the minimum variance in the cross-validation procedure is used for the final clock prediction. We compared the prediction results of our method with two well-recognized BeiDou ultra-rapid clock products, named GBU-P and ISU-P, respectively. The comparison results indicate that the adaptive model achieves about 1-ns precision for 3-h prediction, which corresponds to 47.3% and 32.1% precision improvement compared to the GBU-P and ISU-P products, respectively. The efficiency of the predicted clocks is further validated with the precise point positioning (PPP) data processing. The results indicate that the static PPP solution precision is improved by 21.6%-30.0% compared to the current predicted clock product. The precision improvement in kinematic PPP is even more significant, which reaches 46.7%-53.9% with respect to these GBU-P and ISU-P products. Therefore, the proposed adaptive model is a practical and an efficient way to improve the BeiDou clock prediction.