Environment cross validation of NLOS machine learning classification/mitigation in low-cost UWB positioning systems

Indoor positioning systems based on radio frequency systems such as UWB inherently present multipath related phenomena. This causes ranging systems such as UWB}to lose accuracy by detecting secondary propagation paths between two devices. If a positioning algorithm uses ranging measurements without considering these phenomena, it will make important errors in estimating the position. This work analyzes the performance obtained in a localization system when combining location algorithms with machine learning techniques for a previous classification and mitigation of the propagation effects. For this purpose, real cross scenarios are considered, where the data extracted from UWB low-cost devices for the training of the algorithms come from different environments than those considered for the real application and its analysis.