Improving GNSS Standard Positioning Service Using Boosting Algorithms
Vladislav O. Zhilinskiy · 2022 IEEE International Multi-Conference on Engineering, Computer and Information Sciences (SIBIRCON) · 2022
Global navigation satellite system performance can be assessed by three key characteristics, namely accuracy, availability, and integrity. Despite the recent development, GNSSs cannot meet the requirements of all users, for instance, autonomous vehicles require high accuracy, availability, and integrity. Improving GLONASS positioning accuracy is one of the main aims of the current modernization stage. The paper describes the approach for calculating reference pseudorange measurements and residual pseudorange error. The article discusses the model building process, including feature selection, and compares the performance of the models that were built using different boosting algorithm implementations– Gradient Boosting, LightGBM, CatBoost, XGBoost. The comparison showed that the performance of the trained models using various implementations has almost identical scores. The experiments showed the ability of the trained and tuned machine-learning model to improve positioning results considerably by compensating for residual pseudorange error. The trained model is able to improve standard positioning service performance by up to 30%.