Differential NLOS Error Detection in UWB-based Localization Systems using Logistic Regression

This paper addresses the NLOS detection and error mitigation of differential time of arrival (TDoA) topologies in localization systems. The localization error of such systems is hard to detect mainly due to the differential structure of the network. The proposed channel detection method is based on the logistic regression method which is very simple and requires low amount of resources for detection of the channel condition. The detection process is based on extraction of the probability of the channel for each anchor. In practical results the overall accuracy of 87% is achieved which is only slightly less than other powerful methods such as support vector machines (SVM) which requires much more resources. The localization engine used for this system is based on the robust H
_∞
filter. For mitigation of the error, two approaches are proposed. The first method is based on the modification of the variance and the second one is based on the modification of innovation term in update equation of the filter. The results of practical experiments indicate that the system can effectively determine and suppress the NLOS error up to 80%.