Investigations on multi-GNSS constellation precise orbit determination

Abstract

With the rapid development of the multi-constellation Global Navigation Satellite Systems (GNSS), the real-time precise point positioning (PPP) is undergoing dramatic growth and is recognized as the most promising service. One of the critical issues in such a system is to provide precise and reliable ultra-rapid orbits. Although precise orbit determination (POD) is quite mature for GPS, there is still space for improvement and challenges for newly emerging constellations such as BeiDou System (BDS), Galileo, and Quasi-Zenith Satellite System (QZSS). On the one hand, algorithms of high computational efficiency are required to update the solution as fast as possible in order to shorten the prediction time. On the other hand, observation modeling and satellite force models can be further refined, especially for the new GNSS systems. Hence the major objectives of this study is to improve the precise orbit products for real-time positioning service by increasing modeling accuracy and developing more efficient processing procedures. Recently, most of International GNSS Service (IGS) analysis centers (ACs) use the latest 24-hour observations of about 100 stations to update the ultra-rapid orbit products every three hours and the products are employed in the real-time positioning service. However, both shortening the update time and involving longer observations of more stations will enhance the solution for better orbits, which is very critical for generating real-time products and supporting reliable integer ambiguity resolution. Therefore, the whole processing procedure is optimized according to available computer resources by utilizing multi-process and parallel computation techniques and a new processing scheme is proposed to realize the hourly update of multi-GNSS ultra-rapid orbits. In particular, a high-performance computing (HPC) parallel algorithm is developed based on a multi-core processor for improving the efficiency of the batch least square (LSQ) procedure in the GNSS network solution. For the observation modeling, the correction of higher order ionospheric (HOI) delays remaining in the dual-frequency ionosphere-free (IF) observations has been suggested for high-precision GNSS data processing. However, this correction is often ignored in the ultra-rapid POD most likely because a real-time ionospheric model needed for calculating the HOI corrections is hardly available or the HOI impact is believed rather small. In this contribution, the temporal-spatial characteristics of HOI effects on GNSS observables are investigated thoroughly using data collected from a network of globally distributed IGS stations and consequently its impact on the ultra-rapid orbits is also evaluated. The results show that HOI fluctuations could reach up to several centimeters during periods of high ionospheric activity and that owing to the applied HOI corrections, the agreement of overlapping orbits can be improved significantly for all satellites and especially in the radial direction. Among the GNSS satellite force models, the solar radiation pressure (SRP) force is the most difficult one to be accurately modeled, as it depends on the structure and material characteristics of satellites and should be optimized accordingly. Although the CODE's (Center for Orbit Determination in Europe) SRP model ECOM is widely used for GPS, it is further adapted for GLONASS and GALILEO satellites, and the box-wing model is also suggested to consider their particular satellite surface structures and material features. We concentrate on the optimization of the SRP modeling for BDS-3 satellites, as there is already a global network with about 200 stations tracking BDS-3 satellites. From the large disagreement of overlapping orbits of adjacent sessions with ECOM1 or ECOM2, we noticed that their parameterization should be improved by carefully selecting proper periodic terms. Based on our numerical analysis and parameter significance tests, it is confirmed that the cosine terms must be excluded and the fourth- and sixth-order sine terms are significant in the Sun direction for the SRP model of BDS-3 satellites. With the newly proposed SRP model, the large fluctuations of overlapping residuals in the radial component are reduced remarkably, especially from 20 cm to below 10 cm over deep eclipse seasons and Satellite Laser Ranging (SLR) residuals are also reduced by a factor of two compared to that of ECOM1 and ECOM2. Furthermore, the RMS of predicted orbits over the eclipse seasons can be reduced from about 7.3, 13.3, and 21.7 cm to about 3.1, 4.6, and 10.6 cm, in the radial, cross, and along directions, respectively. Concerning the box-wing model for BDS satellite, the latest satellite metadata have been published on March, 2020, whereas the coefficients of optical properties are still uncertain at present for each box-wing surface of BDS-3 satellites. Based on the SLR validation, a priori coefficients can be optimized by the comparative experiments to confirm the main type of reflection for the associated surfaces of BDS-3 satellites. With the optimized coefficients of the box-wing model, the orbit accuracy could be further improved from about 5 cm to 3 cm in terms of the RMS of SLR residuals. Overall, in this study, we have improved the multi-GNSS POD by shortening the orbit prediction time with a computational efficient processing strategy, and refining the observation i.e., HOI corrections, and adapting the ECOM SRP model and optimizing the optical coefficients of the corresponding box-wing model for BDS-3 satellites as well. All the improvements are implemented in the processing software package and validated with a large set of real observations.