UWB定位中基于XGBoost的NLOS识别方法  

XGBoost-based NLOS Identification for UWB Indoor Localization

作　　者：袁驰 黄波[1] YUAN Chi;HUANG Bo(College of Telecommunications&Information Engineering,Nanjing University of Posts and Telecommunications,Naiijing 210003,China)

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003

出　　处：《移动通信》2023年第5期83-89,共7页Mobile Communications

摘　　要：NLOS场景下UWB室内定位精度下降,识别出NLOS有助于进一步提升定位精度。针对超宽带定位中非视距识别的问题,提出基于XGBoost的NLOS识别方法。基于XGBoost决策树算法,采用两个层面的数据源对UWB定位中NLOS进行识别:元数据层面,用真实定位场景中采集的UWB CIR数据作为数据源;统计数据层面,将CIR数据经过一定的加工,得到几个关键的信号特征参数作为数据源。两组实验结果表明,XGBoost算法在元数据和统计数据上识别精确度分别达到91%和92%。UWB定位中XGBoost算法能显著提高识别NLOS信号的准确度。Non-line-of-sight(NLOS)scenarios can lead to a decrease in the accuracy of ultra-wideband(UWB)indoor localization,and identifying NLOS can help to improve the accuracy of localization.To address the issue of NLOS identification in UWB positioning in such scenarios,an NLOS identification method based on XGBoost is proposed.The method uses two levels of data sources to identify NLOS in UWB positioning:the metadata level,which employs UWB CIR data collected in real positioning scenarios as the data source,and the statistical data level,which processes the CIR data to obtain several key signal feature parameters as the data source.Two sets of experimental results demonstrate that the XGBoost algorithm achieves an identification accuracy of 91% and 92% respectively on the metadata and statistical data.The proposed method can significantly improve the accuracy of identifying NLOS signals in UWB positioning.

关 键 词：室内定位 超宽带 非视距 XGBoost 

分 类 号：TN929[电子电信—通信与信息系统]