Path Loss Prediction in Urban Areas using Convolutional Neural Networks

Urban area path loss prediction is becoming more important in cellular networks that operate at frequencies that are affected more by the loss of line-of-sight such as 5G and beyond 5G. The general move of the industry towards higher frequencies stresses the need to have optimal path loss prediction stronger. In this paper, we present a convolutional neural network based solution to predict path loss in an urban area using publicly sourced GIS data. The outcome of this work is not restricted to path loss prediction in urban areas only, but it is also applicable to disaster struck areas in which emergency cellular services are deployed.