On-the-Fly Training Convolutional Neural Network Models for Radio Wave Propagation in Tunnels

Abstract

The vector parabolic equation (VPE) method has been widely applied to modeling radio wave propagation in tunnels. However, simulation with VPE for long tunnels is still computationally expensive. This paper presents an efficient on-the-fly training convolutional neural network (CNN) model that can provide high-fidelity received signal strength (RSS) prediction without pre-training requirement. Coarse-mesh and dense-mesh VPE simulations are concurrently run for a short distance, while a CNN model which can use only the coarse-mesh VPE data to predict dense-mesh results for the whole tunnel is trained. The accuracy and efficiency of the proposed model have been demonstrated through comparisons with full VPE simulations.