RSRP Prediction In Low Altitude by Deep Learning-Based Method and Geographic Information

With the gradual maturation of aircraft technology and the increasing number of drones, communications in low-altitude airspace will become more and more complex. In order to provide high quality low-altitude communication services, path loss or RSRP prediction of the celluar network is a particularly critical step for base station location and network optimization. In this paper, Digital Elevation Map is employed to provide the geospatial information of the path for RSRP prediction. ResMLP as well as MLP network are utilized to perform feature extraction for the two types of data respectively. Then RSRP values are regressed by the concatenated features. The algorithm achieves an RMSE of 2.94 dB for RSRP prediction and reduces the error by 22.6% compared to MLP-only structure.