A channel modeling and estimation study for Sub-THz ultra-massive MIMO UAV links

Official URL: https://dx.doi.org/10.1109/SIU71813.2026.11636535

Sub-Terahertz (sub-THz) 6G networks promise high-capacity unmanned aerial vehicle (UAV) communications; however, rapid mobility and high-frequency propagation complicate ultra-massive MIMO (UM-MIMO) channel estimation. This paper presents a ray-tracing-aided channel modeling and machine learning-based estimation study for sub-THz UAV communications. Realistic channel realizations are generated from ray-tracing-based propagation paths and extended to UM-MIMO representations. To capture channel variability in UAV scenarios, stochastic fading and controlled angular perturbations are incorporated. Estimation is performed using a model-driven deep learning architecture based on a fixed-point network with orthogonal approximate message passing (FPN-OAMP). Results show stable convergence and consistent accuracy improvement with increasing signal-to-noise ratio, highlighting the effectiveness of combining physics-based channel generation with model-driven learning for sub-THz UAV systems.