A Combined Numerical-Empirical Orbit Propagation Algorithm for Satellite Tracking and Backup Navigation System

Abstract

In the recent years, it is very common using GNSS receiver for the LEO satellite navigation and tracking. The system cost of navigation equipment for satellite becomes very low compare to other satellite navigation and tracking system. The accuracy of GNSS receiver is quite higher than other satellite navigation system. But there must be a backup navigation system for situations the GNSS receiver does not working or the GNSS system does not service. When these situations happened satellite cannot operate properly any more even worst case lost its own position. Satellite rotates around the Earth in the space, so we can predict satellite position and velocity for some period. But the errors of numerical orbit propagation method grow enormously with time. The reason of growing error of numerical orbit propagation method is due to that the space model we use is not fully describes real space. Another reason of growing error of orbit prediction is due to the limited accuracy of satellite. This is the main error source of orbit propagation; it is quite difficult work to modeling the dynamics of satellite. In this paper, we purposed combined numerical-empirical orbit propagation method to improve this error. The orbit prediction error was analyzed by empirical component parts in the CW frame (Local Vertical Local Horizontal frame). Residual error show very periodic characteristic in CW frame. We use this character when numerical orbit propagation. First, we generated orbit using numerical orbit propagation method for future time which was assumed GNSS receiver did not work. Residuals are generated using numerical simulated orbit and true orbit data. The residuals are analyzed in CW frame and compressed into Fourier series coefficients. This coefficient are used when predict the residuals for the prediction time we assumed future. We did some simulation to verify our purposed algorithm, and the results show that using one day of orbit data we can reduce the orbit prediction error under one meter.