GNSS Signal Propagation Error Modelling Method in Train Localization Environments

GNSS is an important source for train location, which provides continuous train location possibilities for train operation. As GNSS positioning principle is radio localization using TOA (Time of Arrival), the radio signals are susceptible to the environment its propagated, location accuracy is dependent on the environment. According to the propagation of the radio signals, they can be divided into three categories: LOS (Line-of-Sight), NLOS (Non-Line-of-Sight) and non-visible. To quantify the ground-related satellite signal propagation error applied to typical train localization environments, this paper addresses two sets of statistical properties-based local pseudorange model for both occluded-signal and non-occluded signal. This paper takes urban canyon as an example for analysis. Firstly, the occluded-signal and non-occluded-signal are divided based on the obstacle boundary shown in the skyplot. Secondly, the difference elimination between user and stations based on CORS network is used to extract the ground-related pseudorange error. The 10-hour static observation data show that the mean and dispersion of the local pseudorange error of the occluded-signal are significantly larger than those of the non-occluded-signal, and arc negatively correlated with the elevation range. The distribution of the occluded-signal is multimodal, and the GMM model is used to describe it. The distribution of non-occluded-signal shows non-Gaussian, and the steepness of the distribution changes with the elevation, which the T Location-Scale distribution model is used for reasonable description.