Noise-Immune Localization for Mobile Targets in Tunnels via Low-Rank Matrix Decomposition

Abstract

Accurate and robust mobile targets localization is one of the important prerequisites for tunnel safety construction in complex environments. Currently, the mainstream localization methods for tunnel mobile targets are almost range-based, which usually suffer from low accuracy and instability due to the complex natural conditions and geographical environment. The reason for the low accuracy and instability is the existing methods mostly assume that the sampled partial Euclidean distance matrices are corrupted by Gaussian noise or/and outlier noise. However, in real applications, the noise is more likely to be unpredictable compound noise. Therefore, in this paper, we propose a noise-immune LoCalization algorithm via low-rank Matrix Decomposition (LoCMD) to address this challenge. Specifically, we adopt the Mixture of Gaussians to model the unpredictable compound noise and employ the popular Expectation Maximization technique to solve the constructed low-rank matrix decomposition model, and thus a complete and denoised Euclidean distance matrix can be obtained. Finally, the extensive experimental results show that the proposed LoCMD achieves better positioning performance than the existing algorithms in the complex environment.