A Study on the Use of Convolutional Networks for RF Coverage Evaluations in Urban Environments

U-nets are a type of Fully Convolutional Neural Network that has been widely adopted for image segmentation applications. In the present study, U-nets are applied to wide area prediction of radio propagation parameters in urban environment. Coverage maps are generated through simulations in a reference urban environment using a ray tracing tool. The generated dataset is then used to train and test the U-net. Preliminary results look promising.