GNSS Pseudorange Stochastic Model for Urban Classification Scenes and Its Positioning Performance

Objective: There will be many problems such as interruption, attenuation, serious multipath error, and NLOS signals when positioning in harsh urban contexts. It’s difficult to ensure the availability, continuity, and reliability of GNSS positioning services. To improve the GNSS positioning performance in complex urban contexts, this paper proposes a method for constructing GNSS stochastic models adapting to different urban contexts. Methods: First, GNSS signal characteristics in different contexts are analyzed to reveal the significant discrepancy of GNSS signals in varied contexts. Then, dynamic reference benchmarks provided by high-precision integrated navigation equipment are used to extract pseudorange error accurately. In addition, appropriate error statistic (median) and impact factor (C/N0) are selected after tests. Finally, the GNSS stochastic models adapting to different urban contexts are constructed using C/N0 and pseudorange error. Results: The urban vehicle test shows that the stochastic model adapting to different urban contexts can effectively weaken the influence of some gross errors. Compared to elevation stochastic model, the positioning accuracy is improved by 16.76% and 16.18% in horizontal and vertical directions, and by 18.68% and 17.72% compared to the C/N0 stochastic model. Conclusions: Stochastic model reconstructed adapting to different environments can weight observations more realistic, thus improving GNSS positioning performance. This provides a new idea for resilient optimization of stochastic models in complex contexts.