GNSS Probabilistic Single Differencing For Non-Parametric State Estimation Based On Spatial Map Data

Abstract

High-precision user position estimation is a necessary prerequisite for various safety-critical applications and location based services in navigation. For vehicular applications, low-cost GNSS receivers are often utilized as they provide a reasonable trade-off between affordability and accuracy. Especially in dense urban environments their performance can be heavily degraded by signal shadowing and non-line-of-sight reception. These conditions challenge state-of-the-art GNSS positioning methods, namely Least Squares Estimation (LSE) and Extended Kalman Filtering (EKF). In this paper, we propose a robust non-parametric state estimation method called GNSS Probability Grid Positioning (PGP) utilizing Between-Satellite Differencing. The proposed method is based on incorporating three-dimensional terrain data providing an a-priori state space. The accuracy of PGP is validated in a dynamic scenario using a GNSS software defined radio simulation setup. We compare its performance to common positioning methods and show that PGP outperforms both LSE and EKF positioning under adverse reception conditions.