Vector Parabolic Equation Driven CNN Models for Radio Wave Propagation in Tunnels

The vector parabolic equation (VPE) method has proven to be a powerful numerical technique for high-fidelity propagation modeling in long guiding structures such as tunnels. However, applying VPE to realistic large-scale environments remains a difficult task due to the requirement of high-performance computing resources. This paper proposes a computationally efficient convolutional neural network model, which can compensate numerical dispersion and provide high-fidelity received signal strength prediction based on results from coarse-mesh VPE simulations. The performance of the proposed model has been validated in various tunnel cases, with significantly reduced simulation time.