GNSS stochastic model parameter refinement method based on swarm intelligence theory and large-scale datasets

Abstract

A stochastic model is critical for high-precision GNSS data processing, while conventional stochastic models based on elevation angle or SNR are insufficient for multi-frequency, multi-system GNSS. This manuscript presents a GNSS stochastic model refinement method using swarm intelligence theory and large-scale datasets. We first analyze the noise characteristics of pseudo-range and carrier phase observations, and verify the strong correlation between observation noise and elevation angle as well as SNR. Then, a stochastic model combining elevation angle and SNR is proposed, with its parameters fitted using massive measured data. Vehicle-borne dynamic tests show that the proposed model outperforms traditional elevation angle and SNR models. The positioning RMS in E/N/U directions reaches 1.7 cm, 1.2cm, and 4.3 cm, respectively, and the ambiguity fixing success rate is improved to 97.14%. The proposed method realizes adaptive refinement of the GNSS stochastic model and supports high-precision positioning.