Robust TDOA-Based Localization for IoT via Joint Source Position and NLOS Error Estimation

Abstract

Accurate localization is critical to facilitate location services for Internet of Things (IoT). It is particular challenging to provision localization based on nonline-of-sight (NLOS) signals. Thus, we actualize source localization based on time difference of arrival (TDOA) derived from NLOS signal propagations. The existing robust least squares (RLS) method exhibits two shortcomings: 1) it is formulated using too large upper bounds on the NLOS errors, and 2) it suffers from the possible inexact triangle inequality problem. Aiming at circumventing the shortcomings of the existing RLS method, we propose two new RLS formulations. On one hand, to reduce the upper bounds on the NLOS errors, we propose to jointly estimate the source position and the NLOS error in the reference path. On the other hand, to avoid using the triangle inequality, we introduce a “balancing parameter” in the first formulation and develop the second formulation by transforming the measurement model. Both formulations are transformed via the S-lemma into optimization problems that are amendable to semidefinite relaxation. The proposed methods achieve superior performance over the existing methods, as validated by using both simulated and experimental data.