Clutter Loss Prediction Models for Satellite-Ground Communication Based on Neural Networks

Satellite communication is considered as one of the key technologies to achieve global seamless coverage and has attracted wide attention. It is crucial to establish an accurate clutter loss model for satellite-ground communication. Clutter loss refers to the extra path loss caused by the obstruction of the terrain and objects on the ground, especially when a satellite has low elevation angle. The clutter loss model proposed by ITU-R P.2108-0 only considers the influence of the elevation angle, and the traditional prediction model for the clutter loss is limited in accuracy and stability. In this work, we used a satellite ground station to carry out the channel measurement at 8.25 GHz on the clutter loss of the X-band satellite-to-ground (S2G) links in the suburban campus environment, and extracted the clutter loss data set involved in the process of satellite inbound and outbound from the received signal strength. We utilize the multi-layer perceptron (MLP), long short-term memory (LSTM), and bidirectional-long short-term memory (Bi-LSTM) neural networks to build channel models to predict the clutter loss based on the measurement data. The model prediction results show that the Bi-LSTM-based model has higher prediction accuracy than the MLP-based and LSTM-based models.