Robust Extended Kalman filter for NLOS mitigation and sensor data fusion

This paper presents an algorithm based on the Robust Extended Kalman filter (REKF) for non-line-of-sight (NLOS) mitigation and sensor data fusion. The aim is to locate an off-the-shelf smartphone using only its speaker and its inertial measurement unit (IMU). The target emits inaudible sound signals which are detected by static receivers on the ceiling. Then, its position can be estimated using time differences of arrival (TDOA). The limiting factor of sound localization is the difficulty to identify the signals which travel directly from the target to the anchor nodes (line-of-sight) and the signals which are reflected to walls or other objects (non-line-of-sight). Nowadays, most of the smartphones are provided with an IMU which can be used for localization. However, its accumulative error deteriorates the result after a certain time. Then, a promising approach is to fuse the received sound timestamps and the IMU data in order to estimate the position of the target. The REKF reduces the effect of the non line-of-sight (NLOS) measurements by assigning them lower weights while the fusion with the IMU improves the accuracy, specially when only a reduced number of line-of-sight (LOS) signals are available.