GNSS-based real-time orbit determination and prediction for LEO-PNT ephemeris generation

Positioning, navigation and timing (PNT) with low Earth orbit (LEO) constellations requires precise knowledge of the navigation satellites’ positions at the time of signal transmission. For conventional GNSS this is achieved by transmitting broadcast ephemerides via the navigation signals, allowing a user to predict the satellite’s position at the required time. Similar approaches using perturbed Keplerian ephemeris models for LEO satellites have already been discussed in the literature. The ephemeris fitting process always requires a set of predicted satellite positions over the desired validity interval of the ephemeris. In a GNSS-augmented LEO-PNT architecture, the navigation satellites determine their position on-board in real-time using GNSS. Based on these orbit determination results, the satellites can autonomously predict their trajectory for fitting of on-board ephemeris parameters. This study assesses the real-time orbit prediction accuracy in LEO using real GNSS data from four satellites across different altitudes. With focus on navigation message generation for future LEO-PNT systems, predicted trajectory arcs of 20 minutes are assessed using GNSS broadcast ephemerides. At high atmospheric density near solar maximum the prediction error for altitudes around 500 km may well grow over 1 m, while for higher altitudes the error remains at low decimeter level. Furthermore, the effects of applying Galileo High Accuracy Service (HAS) corrections in a GPS/Galileo navigation filter are assessed for both orbit determination and orbit prediction. Additional considerations on the prediction performance are made, including the choice of atmospheric density model as well as navigation filter tuning and the consequences on real-time and predicted orbits. For LEO-PNT broadcast ephemeris generation, a representative LEO ephemeris model is adopted in the parameter fitting of the predicted trajectory arcs. The results demonstrate that the prediction error remains the dominant error source in an autonomous on-board ephemeris generation and exceeds the fitting error of adequate LEO-PNT ephemeris models. Depending on the LEO altitude, 1D line-of-sight user ranging errors of less than 8 cm over 20 min ephemeris validity intervals can be achieved using real-time orbit prediction.