Adaptive splitting mean online expectation-maximization method-based moving object localization

Abstract

This paper introduces positioning techniques for estimating the location of an emitter using range data affected by outliers. In indoor and densely populated metropolitan environments, the presence of non-line-of-sight (NLOS) signals can significantly degrade estimation performance. To mitigate the adverse effects of NLOS signals, robust localization methods are employed. The proposed technique, referred to as the splitting mean (SM) online expectation-maximization (EM)-based two-step weighted least squares (TSWLS) method, is developed from a Bayesian perspective, specifically utilizing the linear minimum mean squared error (LMMSE) criterion. A key element influencing the performance of the SM algorithm is the smoothing factor. Unlike traditional SM methods that use a fixed smoothing factor, the proposed adaptive splitting mean (ASM) bias estimation method dynamically adjusts this factor. Additionally, a theoretical analysis of the mean squared error (MSE) for the proposed measurement bias estimation algorithms is conducted, demonstrating close alignment with simulation results. Simulations further reveal that the proposed method outperforms existing state-of-the-art techniques in localization accuracy across various NLOS bias distributions, including Gaussian, uniform, and exponential distributions.