Artificial Intelligence-Based Adaptive Optimization of GNSS Elevation Cut-Off Angles for Static Positioning

Abstract

Abstract

The correct selection of elevation cut-off angle is an important component affecting the accuracy of Global Navigation Satellite System (GNSS) positioning. Traditional processing algorithms usually adopt fixed cut-off angles which cannot adapt to the variations of signal quality and observational conditions, resulting in poor positioning performance. This paper proposes an AI-based adaptive method for optimization of GNSS elevation cut-off angles for static location applications. Static GNSS data were observed for 12 h and processed with elevation cut-off angles ranging from 0° to 25°. A Random Forest classifier was developed to detect and remove low quality observations using signal quality indicators such as elevation angle, Signal to Noise Ratio (SNR), pseudorange residuals and carrier-phase residuals. The results showed that the positioning accuracy improved with the increasing of the cut-off angle up to 15°, which produced the minimal Horizontal Error (0.70 cm), Vertical Error (1.10 cm) and Root Mean Square Error (RMSE) (1.30 cm). Signal quality improved from 34 dB-Hz at 0° to 52 dB-Hz at 25°, however satellite availability decreased from 22 to 7 satellites. The proposed AI-based framework demonstrated the RMSE improvements ranging from 29.17 to 41.39% against the typical fixed cut-off angle approach. From the feature importance analysis elevation angle and SNR were the most important factors affecting observation quality. The results indicate that the quality of GNSS observations and the accuracy of location for static GNSS applications can be improved significantly by adaptive filtering based on machine learning.