ToF-Based Indoor Localization for Harsh Environments: A Continuous Environment Sensing Framework

Non-Line-of-Sight (NLoS) propagation and packet loss are critical factors that influence the accuracy and robustness of Time-of-Flight (ToF)-based indoor localization and navigation systems in harsh environments. In this paper, we propose a continuous environment sensing framework designed for packet loss recovery and NLoS identification in ToF-based localization systems. The proposed framework comprises a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">continuous environment sensing module</i>, a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">packet loss recovery module</i>, a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">coarse localization module</i>, an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">NLoS identification module</i>, and a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">navigation refinement module</i>. Specifically, the continuous environment sensing module captures dynamic environmental characteristics by updating an NLoS table (from the NLoS identification module) and tracking the target’s states (location and velocity). Leveraging the acquired characteristics, we develop a robust low-rank matrix recovery algorithm for packet loss recovery. Subsequently, through coarse localization, we generate localization estimates for different ToF subsets. Based on these localization estimates, we then design a residual analysis-based algorithm for NLoS identification, incorporating an environmentally adaptive residual threshold strategy. Finally, a navigation refinement module is employed to mitigate NLoS errors and refine the trajectory. Experimental results demonstrate that, under a 30% packet loss rate in commodity UWB systems, the proposed framework achieves average localization and navigation accuracies of 17.71cm and 47.42cm, respectively, while reducing the localization error by 51.03% in static scenarios and 50.18% in mobile scenarios compared with state-of-the-art algorithms.