Abstract
Indoor location systems based on ultra-wideband (UWB) technology have become very popular in recent years following the introduction of a number of low-cost devices on the market capable of providing accurate distance measurements. Although promising, UWB devices also suffer from the classic problems found when working in indoor scenarios, especially when there is no a clear line-of-sight (LOS) between the emitter and the receiver, causing the estimation error to increase up to several meters. In this work, machine learning (ML) techniques are employed to analyze several sets of real UWB measurements, captured in different scenarios, to try to identify the measurements facing non-line-of-sight (NLOS) propagation condition. Additionally, an ulterior process is carried out to mitigate the deviation of these measurements from the actual distance value between the devices. The results show that ML techniques are suitable to identify NLOS propagation conditions and also to mitigate the error of the estimates when there is LOS between the emitter and the receiver.
Highlights
Indoor location systems based on ultra-wideband (UWB) are very popular nowadays due to their ability to provide accurate range estimations based on the signal time of arrival (TOA) or time difference of arrival (TDOA)
Since we are considering low-cost UWB devices, statistic parameters computed from received signal strength (RSS) and ranging estimates are considered as input features to the machine learning (ML) algorithms instead of channel impulse response (CIR) estimates
In this paper we analyzed the performance of ML techniques applied to classification and error mitigation in low-cost UWB systems when NLOS propagation conditions are present
Summary
Indoor location systems based on ultra-wideband (UWB) are very popular nowadays due to their ability to provide accurate range estimations based on the signal time of arrival (TOA) or time difference of arrival (TDOA). The main idea behind these solutions is that the energy of the first path is noticeably greater than the energy of the delayed paths in LOS conditions, whereas this difference tends to shorten in a NLOS scenario In addition to these methods, NLOS detection can be performed at a higher logic level, that is, in the location algorithm that uses the range estimation [11]. The main contribution of this work is an approach to identify and mitigate the effects caused by NLOS propagation conditions This contribution corresponds to the boxes labeled as Classification and Mitigation, which shows the typical flow diagram of a location system based on ranging and RSS measurements, where the measurements between multiple anchors (devices placed at known fixed positions) and a target are classified and corrected before being processed by a location algorithm.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
