Mitigating effects of NLOS propagation in MDS-based localization with anchors

Indoor localization based on radio-frequency (RF) signals usually suffers from multipath effects due to occlusions and obstacles in the deployment region which cause nonline-of-sight (NLOS) propagation. These downgrade the accuracy of localization algorithms when distances are constructed directly from noisy or NLOS RF signals. This work proposes a technique for estimating NLOS biases and measurement noise in distance measurements by exploiting the intrinsic symmetry in a matrix of line of sight (LOS) measurements of pairwise distances. This symmetry allows to compute unknown locations (tags) from known-position nodes (anchors) and then recompute exactly the same anchor positions from the earlier computed tag positions under MDS-based (Multi-Dimensional Scaling) localization. In a NLOS environment, there is no such symmetry; anchor positions can not be reproduced exactly from those of tags. This error in the reproduced anchor positions is used to approximate the overall bias and noise in the matrix of measured distances such that NLOS effects alongside noise can be mitigated. Simulations are used to examine the improvement in localization accuracy of tag positions after the reduction of NLOS bias using the proposed approach. Results show that in a setup with 8 anchors and 22 tags, NLOS effects and noise can be reduced by up to 35%.