基于机器学习的超宽带NLOS鉴别方法

Research on NLOS identification method based on machine learning using IR-UWB signals

在超宽带测距定位中,实现非视距(non-line of sight,NLOS)传播的正确鉴别与消除具有非常重要的意义。提出以实测数据为基础,采用多参数的机器学习方法对超宽带信号进行NLOS鉴别和测距误差消除研究,并对波形特征参数的选取进行了分析和优化。数据结果表明,相比较传统NLOS鉴别方法,该方法不需要先验知识、实用性强,在鉴别的准确度和误差消除方面有较大的提升。

Non line of sight （NLOS） error identification and mitigation is of great importance in ultra wideband （UWB） ranging and localization. Based on the features extracted from the received waveform in practical experiments, a machine learning method is proposed for UWB NLOS identification. Corresponding NLOS error mitigation method is also given based on the identification results. Compared with the traditional NLOS identification methods, the ML-based method is able to achieve better results with less priori knowledge, which makes it practical in universal applications.