Satellite selection, satellite orbit interpolation and prediction for GNSS based positioning

Abstract

Modern infrastructure and a myriad of services rely on the ability to generate precise position using GPS. Although, generating precise position using GPS is a complex multi step process involving selecting satellites, interpolating, state estimation and multiple other steps. The positioning process starts by selecting satellites in view. The accuracy of positioning in- creases with increasing GPS satellites used for computing position. However, there are limited numbers of channels on the receiver, and the onboard computer has a limited real-time pro- cessing capability. Thus, it is imperative to select a subset of measurements that provides the best possible geometry. We develop two algorithms Expansion Algorithm and Expansion Quasi Algorithm. They outperform the existing algorithm, with near similar runtime and operate near to the established performance limit of cost function used. After selecting satellites, precise orbit of GPS satellites are required for calculating the receiver position. International GNSS Service (IGS) provides precise orbit solution for GPS satellites. However, the data rate of such precise orbits is usually limited to 15 minutes, making GPS orbit interpolation an essential part of the process. Hence this report further examines the performance and provide a comparison between orbit interpolation techniques, including Univariate Spline, Polynomial fitting, Lagrange Interpolation, Cubic Spline and Nvilli Interpolation from a practical standpoint. We found, on the basis of comparison, including the ability to generalise in sparse data and runtime simulation on actual data, Polynomial fitting outperforms other methods. Precise orbit solution from IGS are provided via internet. However, there is always a possibility of communication outage due to equipment malfunction or operations in remote areas forcing to switch to less accurate single-point positioning. This presses the need for a failsafe system to improve broadcast orbit in real-time in case of a long-term communication break. We present an analysis of GPS Broadcast Orbit error. Further, we propose a method named Back Correction to improve broadcast orbit product in real-time for up to 24 hours of correction outage using the precise position of the previous day. The method reduces the error in broadcast orbit by 30∼40%. Additionally, it is computationally inexpensive and hence could be deployed on a low-power GPS receiver. This proposed method results in a 50∼60% improvement in positioning accuracy when tested the GRACE satellite. Further, we explore the effectiveness of Neural Network, Physics Informed Neural Networks, ARIMA, and LSTM based methods for improving broadcast orbit products and compare the results with the proposed Back correction method for a 24-hour period. Further, we present the software development for in making of Precise Orbit Determination software for ISRO.