Impact of on-board satellite orbits and clocks estimation for LEO-PNT ground positioning

Low Earth Orbit Positioning Navigation and Timing (LEO-PNT) is an emerging satellite navigation concept to augment current Global Navigation Satellite Systems by placing satellites close to Earth, at around 600-1200 km altitude. This proximity leads to a rapid change in satellite geometry, which is mainly expected to reduce the convergence time of real-time precise positioning. To realize the benefits of LEO’s faster dynamics, the positions and clock offsets of the LEO-PNT satellites must be known with a high accuracy and low latency. In existing GNSS constellations deployed in Medium Earth Orbit (MEO), global networks of ground stations are generally used to estimate the satellite positions and clock offsets, which are then uplinked to the satellites for broadcast to users. This same approach for LEO-PNT systems would require an extensive ground network due to their closer proximity to Earth. Instead, on-board GNSS-based Precise Orbit Determination (POD) for LEO-PNT satellites offers a feasible alternative. 

This study investigates the impact on ground positioning users when performing on-board POD for LEO-PNT satellites. The numerical assessment consists of two parts: in the first part we focus on the on-board POD results by using Sentinel-6A real-world data from DOY 118-124 in 2024 including both GPS and Galileo observations. A reduced-dynamics extended Kalman filter POD approach with degraded dynamical models is used to replicate on-board processing conditions. Various types of GNSS corrections are tested to assess the POD accuracy achievable on board. 3D RMS orbit errors of 2.8 cm, 4.8 cm, 9.9 cm, and 15.2 cm are obtained in the numerical POD computations respectively based on the CODE MGEX final products (COD), the CNES Real-Time products (CRT), the Galileo High Accuracy Service corrections (HAS), and the broadcast ephemerides (BRD). Moreover, we compare the estimated receiver clock offsets with respect to a precise reference clock solution computed in a batch-least squares approach without orbital model degradation. 

In the second part, we focus on the impact of these LEO orbit and clock errors in an end-to-end simulation of kinematic float-PPP for a ground user. A LEO space segment of 28 satellites was simulated to augment the cases of GPS only, Galileo only, and GPS+Galileo, while considering different product configurations. The results showed that a LEO space segment with CRT-level orbit and clock errors could consistently improve the convergence time as compared to each corresponding stand-alone MEO case. For a HAS ground user using GPS and Galileo, the LEO with HAS-level orbit errors achieved 20 cm horizontal convergence under 3 minutes when clock errors were neglected. At the same time, the overall positioning accuracy results did not show significant improvement nor degradation from including the LEO space segment. Based on our preliminary findings, the expected benefits of LEO-PNT augmentation are only possible when sufficiently accurate orbits and clocks are estimated and provided to users. Still, the impact of the correction latency and availability shall be further investigated in future works.