Prediction of Radio Wave Propagation on Complex Terrains Using Deep Learning Models

In this paper, we propose an innovative model to predict electromagnetic wave propagation on complex terrain, using the Dominant Path Model (DPM) algorithm to generate a dataset, and training a ResNet based wave propagation model. The key innovation is to use stepwise training techniques to apply the progress made by deep learning in the field of images to the prediction of radio wave propagation, in order to improve fitting performance. This method helps to improve the accuracy and adaptability of electromagnetic wave propagation prediction, providing new ideas for the prediction research of large-scale radio wave propagation.