Near Real-Time Positioning of Low Earth Orbit Satellites using Galileo High Accuracy Service

Kurzfassung

Low Earth Orbit satellites increasingly rely on precise positioning for navigation and scientific applications. One approach is to equip LEO satellites with GNSS receivers and use Precise Point Positioning to estimate the position from onboard measurements. The Galileo High Accuracy Service (HAS) provides real-time orbit, clock, and bias corrections through the E6-B signal, enabling high-accuracy PPP without relying on post-processed products. Within the GHASP3 software framework, two estimation strategies are available: a Kalman filter for real-time processing and a batch least-squares solver for near-real-time applications. While the filter has been evaluated in previous work, the batch solver required further analysis, particularly under HAS corrections and for a space user.

This thesis evaluates batch least-squares PPP for the Sentinel-6A (S6) LEO satellite using both final precise products and Galileo HAS corrections, and compares its performance against the Kalman filter. The ground station BRUX is used as a validation reference. The batch under HAS is run as a growing window, where the window length equals the time elapsed since the start of the arc, directly reflecting how it would operate in a near-real-time setting. Batch windows ranging from 10 to 180 minutes are tested, and the sensitivity of the solution to sampling interval, correction product, user type, and disturbances is assessed.

The results show that batch becomes reliable from around 60 minutes for both users under FIN, and between 90min and 120min for S6 under HAS. At 114 minutes - one full orbital period - batch gives its best result and outperforms the filter. An unexpected finding is that 30s sampling outperforms 10s for S6 under HAS, the opposite of what was observed under FIN, which may suggest that denser sampling amplifies correction noise under real-time products rather than improving geometry. Under HAS, the filter takes nearly 3 hours to converge for BRUX and over 3.5 hours for S6, meaning batch delivers a reliable solution faster. DIA is found to be essential under disturbed conditions, and batch with DIA enabled is the most resilient configuration tested.

For FIN products, the filter remains appropriate when low latency is the priority. For HASbased LEO positioning, batch is the better choice.