[G05-P] Poster

[G05-P-05] Short-Term Prediction of IGS Real Time Service Data for Continuous GNSS Positioning

^○Jeongrae Kim, Mingyu Kim (Korea Aerospace University, Goyang-City, Korea)

For precise global navigation satellite system (GNSS) positioning, GNSS error components such as satellite orbit and clock errors need to be corrected. The international GNSS service (IGS) provides a real-time orbit and clock estimates (real-time service, RTS) for real-time precise positioning. The real-time performance of the RTS, i.e. accuracy and stability, are analyzed with a long-term data. And the RTS correction availability for the GPS satellites observed in East Asia is analyzed. Since the IGS RTS provides the real-time corrections via the internet, intermittent data loss can occur due to software or hardware failures. In order to make up for possible RTS data loss or delays, several prediction algorithms for the RTS data are developed, which include polynomial extrapolation, autoregressive moving average (ARMA), and machine learning algorithms (neural network and genetic algorithm). The prediction interval is set from 10s to 900s, and the performance of those methods are evaluated and compared. The machine learning algorithms yields a substantial improvement, greater than 25%, over the simple extrapolation algorithms. The error reduction is more significant in the clock prediction than in the orbit prediction. The clock correction has a more stochastic variation than the orbit correction, and the machine learning algorithm is efficient for predicting such a stochastic variation.