GNSS Precise Positioning Method and Device Based on Lightweight Machine Learning Model

Abstract

This invention discloses a GNSS precise positioning method and apparatus based on a lightweight machine learning model, belonging to the field of satellite positioning technology. The method includes: obtaining a verification feature vector of the optimal candidate ambiguity fixed solution, the verification feature vector including a state estimation consistency feature, which characterizes the degree of agreement between the state estimate value and the observation residual of the ambiguity fixed solution; inputting the verification feature vector into an ambiguity fixed solution correctness discrimination model to obtain the correctness discrimination probability of the optimal candidate ambiguity fixed solution; and if the correctness discrimination probability of the optimal candidate ambiguity fixed solution is greater than a discrimination threshold, then the optimal candidate ambiguity fixed solution is used as the final positioning result. This method can achieve lightweight and highly generalizable GNSS precise positioning.

Images