Ray-Tracing Driven ANN Propagation Models for Indoor Environments at 28 GHz

Ray-tracing is widely used for radio propagation modeling of indoor environments, such as hallways and offices. Shooting and bouncing ray-tracing methods are faster than fullwave methods in such environments, especially as the frequency of operation increases following the new 5G specifications. Still, a machine learning approach can generalize a few ray-traced points into full signal strength maps of arbitrary resolution. In this paper, a feedforward standard artificial neural network is trained by ray-tracing data at 28 GHz to predict signal strength in a Γ -shaped corridor. The network's accuracy in reconstructing the actual signal levels is on par with that of the ray-tracer.