AI-Based Prediction of Optimal Operating Frequency and Transmit Elevation-Angle Ranges for HF Communication

This is the authors' abstract. We don't add key points for this paper.

This study proposes a deep learning-based model for predicting the optimal operating frequency and elevation angle for high-frequency (HF) skywave communication. The model used precipitation conditions in the troposphere, the ionospheric refractive index profile, and distance between the transmitter and receiver as input variables. Training data were generated through radio-wave propagation simulations. A one-dimensional convolutional neural network (1D-CNN) was employed to extract altitude-dependent features from the ionospheric refractive index profile. A mixture density network (MDN) was integrated to model the multisolution characteristics, where multiple combinations of frequency and elevation angle may exist under the same propagation environment. The proposed 1D CNN-MDN model can simultaneously predict candidate solutions of frequency and elevation angle along with their probability distributions, making it applicable for the prediction of operational parameters for HF communication.