Multi-Modal Frequency-Selective Channel Modeling for UAV-to-Ground Communication in Future Wireless Networks

Abstract

Unmanned Aerial Vehicles (UAVs) are increasingly used in wireless networks because of their ability to provide mobility and wide range coverage. However,
the reliable UAV-to-Ground (U2G) communication in Future Wireless Networks (FWNs), remains challenging. The main difficulties arise from UAV posture
changes during flight, the presence of static and dynamic scatterers in dense environments, and the frequency-selective nature of fading across different bands.
The existing models focus on simplifying these factors, which limit their accuracy. This paper proposes a posture-aware, multi-modal, and frequency-selective
channel modeling framework for U2G communication. The framework includes UAV yaw, pitch, and roll to capture posture effects on multipath behavior and
Doppler spread. It also combines LiDAR sensing with Radiofrequency (RF) data to distinguish between static and dynamic scatterers. In addition, Graph Neural
Network (GNN) and Grasshopper Optimization Algorithm (GOA) are used to learn spatial-temporal channel features and optimize delay spread, coherence
bandwidth, and angular dispersion across sub-6 GHz and mm Wave bands. The simulation results indicate reduced Path Loss Root Mean Square Error (PLRMSE) and more accurate predictions of delay spread and coherence bandwidth across sub-6 GHz and mm Wave bands.