Prediction Method by Deep-Learning for Path Loss Characteristics in an Open-Square Environment

Abstract

We describe a path loss characteristic prediction method we propose using a convolution neural network (CNN). In high traffic environments such as the public squares in front of large stations, various obstacles greatly affect radio wave propagation characteristics in high frequency bands. Conventional methods find it very difficult to deal with this problem. Therefore, we determined whether a model we propose can estimate propagation loss characteristics by taking the problem into account. Experiments to verify this were performed using simulation data obtained by ray-tracing.