A Machine Learning Based Channel Parameter Prediction Method Utilizing Ray Tracing Simulation Dataset

Accurate channel models are essential to the evaluation and optimization deployment of mobile communication systems, especially with the development of next generation of mobile communication, which has raised the accuracy demand for channel modeling. This paper proposes a path loss prediction method based on spatial propagation environment characteristics with a combined application of convolutional neural network and deep neural network. Especially, the latest empirical path loss model is added in the proposed method as prior information for predicting path loss. The accuracy and efficiency of the proposed path loss prediction model are verified in this work with its application to ray tracing simulation examples that combine different propagation environment characteristics, transmitter (Tx) and receiver (Rx) coordinates, Tx-Rx distances, and the potential of this method for channel prediction in different urban communication scenarios is demonstrated.