Non-Line-of-Sight Identification Without Channel Statistics

Identifying non-line-of-sight (NLOS) conditions is important to discard, or improve, any location estimates that have been estimated with NLOS ranges. Typically, NLOS identification relies on channel statistics that have been collected for both LOS and NLOS channels. We investigate NLOS identification using distance residuals instead. The results show that distance residuals can be used to identify location estimates with NLOS ranges with very high accuracy, and that in some cases, individual NLOS ranges can also be identified.

I. Introduction

Accurate localization is useful for a number of applications in industry [1], [2], [3]. Ideally, low power technologies such as Bluetooth [4], [5] or Ultra-wideband (UWB) are used for localization, but several challenges exist [6]. In particular, Non-Line-of-Sight (NLOS) propagation causes ranging errors and reduces location accuracy significantly [7], [8]. To improve location accuracy in NLOS conditions, NLOS identification is an intermediate step in many NLOS mitigation techniques [9]. However, most of the NLOS identification methods investigated in literature rely on channel statistics.