Path Loss Exponent Prediction for Outdoor Millimeter Wave Channels through Deep Learning

Abstract

In this paper, we propose a new algorithm for predicting the path loss exponent of outdoor millimeter-wave band channels through deep learning method. The proposed algorithm has the advantage of requiring less inference time compared to existing deterministic channel models while concretely considering the topographical characteristics. We used three-dimensional ray tracing to generate the outdoor millimeter-wave band channel and path loss exponent. We trained a neural network with generated path loss exponent. To evaluate the performance of the proposed method, we analyzed the influence of the hyperparameters and environmental features, for example, building density and average distance from the transmitter.