Non-Line-of-Sight Multipath Detection Method for BDS/GPS Fusion System Based on Deep Learning

• Published in: Shanghai jiao tong da xue xue bao Vol. 27; no. 6; pp. 844 - 854
• Main Authors: Su, Hong, Wu, Bozhao, Mao, Xuchu
• Format: Journal Article
• Language: English
• Published: Shanghai Shanghai Jiaotong University Press 01.12.2022; Springer Nature B.V
• Subjects: Accuracy; Algorithms; Architecture; BeiDou Navigation Satellite System; Canyons; Classification; Computer Science; Deep learning; Electrical Engineering; Engineering; Extended Kalman filter; Global navigation satellite system; Global positioning systems; GPS; Life Sciences; Line of sight; Machine learning; Materials Science; Neural networks; Recognition; Recurrent neural networks; Satellite navigation systems; Signal classification; Street canyons; Support vector machines; Urban environments; recurrent neural network (RNN); global positioning system (GPS); A; BeiDou navigation satellite system (BDS); TN 967.1; urban canyon environments; non-line-of-sight (NLOS) multipath
• DOI: 10.1007/s12204-022-2430-9
• ISSN: 1007-1172; 1674-8115

Non-line-of-sight (NLOS) multipath effect is the main factor that restricts the application of global navigation satellite system (GNSS) in complex environments, especially in urban canyon. The effective avoidance of NLOS signals can significantly improve the positioning performance of GNSS receiver. In this paper, an NLOS/LOS classification model based on recurrent neural network is proposed to classify satellite signals received in urban canyon environments. The accuracy of classification is 91%, and the recognition rate of NLOS is 89%; the classification performance is better than traditional machine learning classification models such as support vector machine. For BeiDou navigation satellite system/global positioning system (BDS/GPS) fusion system, the least square algorithm and extended Kalman filter are used to estimate the position. The experimental results show that the three-dimensional positioning accuracy after NLOS recognition is improved about 60% on average compared with the traditional methods, and the positioning stability is also improved significantly.