Methods for Long-Term GNSS Clock Offset Prediction

Abstract

Clock offset predictions along with satellite orbit predictions are used in self-assisted GNSS to reduce the Time-to-First-Fix of a satellite positioning device. This paper compares three methods for predicting GNSS satellite clock offsets: polynomial regression, Kalman filtering and support vector machines (SVM). The regression polynomial and support vector machine model are trained from past offsets. The Kalman filter uses past offsets to estimate the clock offset coefficients. In tests with GPS and GLONASS data, it is found that all three methods significantly improve the clock predictions relative to extrapolation with the basic clock model of the last obtained broadcast ephemeris (BE). In particular, the 68% quantile of 7 day clock offset errors of GPS satellites was reduced by 66% with polynomial regression, 69% with Kalman filtering and 56% with SVM on average.