文献类型：期刊文献

英文题名：GNSS Signal Labeling, Classification, and Positioning in Urban Scenes Based on PSO-LGBM-WLS Algorithm

作者：Luo, Hui[1];Liu, Ji[1];He, Di[2]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[2]Shanghai Jiao Tong Univ, Sch Sensing Sci & Engn, Sch Elect Informat & Elect Engn, Shanghai Key Lab Navigat & Locat Based Serv, Shanghai, Peoples R China

年份：2023

卷号：72

外文期刊名：IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT

收录：;EI(收录号：20234014842402);WOS:【SCI-EXPANDED(收录号:WOS:001082374600031)】；

基金：This work was supported by the National Natural Science Foundation of China under Grant 61971278, Grant 62231010, and Grant 61973122

语种：英文

外文关键词：Global navigation satellite system (GNSS); line-of-sight (LOS)/nonline-of-sight (NLOS) signal classification; machine learning; multipath (MP); NLOS; particle swarm optimization (PSO)

摘要：In urban canyon areas, the receiver positioning error of the global navigation satellite system (GNSS) is mainly affected by multipath (MP) effects and nonline-of-sight (NLOS) reception, especially NLOS reception issues, which sometimes even cause errors of hundreds of meters. Therefore, detecting and suppressing NLOS reception has become an effective means to improve the positioning accuracy of GNSS, which has been extensively researched and used in the literature. Based on the preferred combination of satellite signal features, a line-of-sight (LOS)/NLOS signal classification method utilizing the light gradient boosting machine (LGBM) is first proposed in this article. And based on this, while calculating the position coordinates by combining the weighted least squares (WLS) algorithm and the particle swarm optimization (PSO) algorithm, the reception of NLOS can be suppressed so that its influence can be significantly reduced, and finally, a novel PSO-LGBM-WLS algorithm model is established in this study. In addition, in view of the difficulty of obtaining labels in NLOS detection using machine learning-based algorithms, we also propose to use the pseudorange error as the basis for label production and use PSO to optimize the classification threshold of pseudorange error, which also helps to reduce the final positioning error from another aspect. In the real experiments, relying on the data collected by the Google team in the San Francisco area, the proposed algorithm model is tested in terms of LOS/NLOS signal classification and positioning accuracy, which are compared with other machine learning models and positioning methods. The experimental results show that the proposed algorithm has the advantages of high classification accuracy and short consumption time, which provides the possibility for practical application. From the positioning results, the proposed PSO-LGBM-WLS algorithm model shows significant effects on improving positioning accuracy by detecting and suppressing NLOS reception.