Effect of Physical Characteristics on Artificial Neural Network Error Reduction for Indoor Propagation Modeling

In this paper, we examine the effect of physics-based input data in artificial neural network (ANN) for indoor propagation. Our prior knowledge of electromagnetic propagation forms, enables us to predict path loss for a specified geometry. There is a good agreement between ANN predicted output and ray-tracing results for a simple room and the error reduces notably compared to related works. Additionally, we explore feature extraction for more complex scenarios.