Agile Radio Map Prediction Using Deep Learning

In this paper, we introduce a runtime-efficient radio frequency (RF) map prediction method based on UNet convolutional neural networks (CNNs), trained on a large-scale 3D maps dataset. The proposed method calculates the line-of-sight maps and feeds them as input for the UNet CNN. A special Kullback–Leibler divergence loss function is adopted, enabling the proposed method to minimize both error's mean and variance. The performance of our model is evaluated in the context of the 2023 IEEE ICASSP Signal Processing Grand Challenge, namely, the First Pathloss Radio Map Prediction Challenge. The evaluation results demonstrate that the proposed method achieves an average normalized root-mean-square error (RMSE) of 0.045 with an average of 14 milliseconds (ms) runtime.