A Temporal Attention Unit-Based Machine Learning Model for Efficient Radio Wave Propagation Prediction in Tunnels
Hengyuan Xu, Junqiao Wang, Kunyu Wu, Qiushi Zhao, Hao Qin, Xinyue Zhang, Xingqi Zhang · 2025
The vector parabolic equation (VPE) method is recognized as a high-accuracy approach to simulate radio wave propagation in extensive guiding structures, such as tunnels. Nevertheless, its application to long tunnels still presents considerable computational difficulties. This paper introduces a computationally efficient temporal attention unit (TAU) model that utilizes a small portion of the preceding received signal strength (RSS) data from the tunnel’s entrance to predict a significantly larger sequence extending further along the tunnel. The proposed TAU model incorporates both intraframe static and dynamic attention mechanisms, allowing it to effectively capture complex spatio-temporal dependencies in the RSS data. Validation was performed by comparing the results with full VPE simulations in tunnel scenarios.