GNSS Pseudorange Correction Using Machine Learning in Urban Areas

GNSS signals are easily blocked and reflected by high buildings in urban areas, causing non-line-of-sight (NLOS) and multipath errors. These errors deteriorate the accuracy of position. In this chapter, machine learning based correction method is proposed to mitigate the NLOS/multipath errors in pseudorange. The results of a static and a dynamic experiments demonstrate the effectiveness of the proposed method. In the static experiment, the improvements of positioning accuracy in horizontal were 75.6 and 75.6%, and in 3D were 71.4 and 70.9%, compared with two conventional positioning methods. In the dynamic experiment, the two variations of pseudorange error correction model (PBC and GBC) are used to improve positioning accuracy in urban environments. PBC model achieved positional accuracy improvements in horizontal of 42.9 and 41.1%, and in 3D accuracy of 60.1 and 45.7% compared with comparative methods 1 and 2. GBC achieved improvements in horizontal of 40.8 and 38.9%, and in 3D 63.3 and 50.0%, compared with comparative methods, respectively.