Multi-Tag UWB Localization With Spatial–Temporal Attention Graph Neural Network

Sizhen He,Bo Yang,Tao Liu,Hong Zhang

DOI: https://doi.org/10.1109/tim.2024.3462984

IF: 5.6

2024-10-11

IEEE Transactions on Instrumentation and Measurement

Abstract:Accurate ultrawideband (UWB) localization is still a challenging issue in complex environments. In this article, we propose a novel UWB localization method with twofold strategies to further improve its robustness and accuracy: 1) we construct the UWB ranging measurements into a graph structure and design a multiple tags UWB localization with the spatial-temporal attention graph neural network (STA-GNN-M) to handle it and 2) we utilize multiple tags to obtain multiple ranging measurements, comprehensively establish the geometric relationship between these measurements, and design a novel cost function to train the proposed network effectively. Our real-world experiments demonstrate that: 1) compared to the convolutional neural networks (CNNs) and long short-term memory (LSTM), which have been used in UWB localization, the graph neural network (GNN) can better capture the geometric relationships from the UWB ranging measurements, leading to a better estimation for tag location, and in addition, our proposed STA-GNN-M can effectively extract high-level spatial-temporal features from the measurements, which benefit UWB localization; and 2) using multiple tags can provide more information and geometric constraints, which can help improve the accuracy and robustness of UWB localization compare to the single tag-based methods, especially in non-line-of-sight (NLOS) environments.

engineering, electrical & electronic,instruments & instrumentation