hide
Free keywords:
-
Abstract:
In the Global Navigation Satellite System (GNSS), the satellite clock bias (SCB) is one of the sources of ranging error, and the prediction capability directly affects the users navigation and positioning accuracy. The establishment of a reliable SCB predicting model is important for real-time precise point positioning, precise orbit determination and optimization of navigation message parameters. In this report, we apply a Long Short-Term Memory (LSTM) model for predicting BDS-3 SCB, which uses a multiple single-step predicting method to avoid error accumulation in the process. Short- (0 to 6 hours), medium- (6 hours to 3 days) and long-term (3 days to 7 days) predicting is performed, and the results are compared with those of two traditional models to verify the reliability and accuracy of the LSTM method. In the long-term prediction of BDS-3 SCB, LSTM improves the accuracy about 70% and 60% compared to the autoregressive integrated moving average (ARIMA) and quadratic polynomial (QP) model, respectively. This report also presents the results of predicting GPS and Galileo SCB using the LSTM method.
