This paper presents a comprehensive measurement-based trajectory-aware characterization of low-altitude Air-to-Ground (A2G) channels in a suburban environment. A 64-element Massive Multi-Input Multi-Output (MaMIMO) array was used to capture channels for three trajectories of an Uncrewed Aerial Vehicle (UAV), including two horizontal zig-zag flights at fixed altitudes and one vertical ascent, chosen to emulate AUE operations and to induce controlled azimuth and elevation sweeps for analyzing geometry-dependent propagation dynamics. We examine large-scale power variations and their correlation with geometric features, such as elevation, azimuth, and 3D distance, followed by an analysis of fading behavior through distribution fitting and Rician K-factor estimation. Furthermore, temporal non-stationarity is quantified using the Correlation Matrix Distance (CMD), and angular stationarity spans are utilized to demonstrate how channel characteristics change with the movement of the UAV. We also analyze Spectral Efficiency (SE) in relation to K-factor and Root Mean Square (RMS) delay spread, highlighting their combined influence on link performance. The results show that the elevation angle is the strongest predictor of the received power, with a correlation of more than 0.77 for each trajectory, while the Nakagami model best fits the small-scale fading. The K-factor increases from approximately 5 dB at low altitudes to over 15 dB at higher elevations, indicating stronger LoS dominance. Non-stationarity patterns are highly trajectory- and geometry-dependent, with azimuth most affected in horizontal flights and elevation during vertical flight. These findings offer valuable insights for modeling and improving UAV communication channels in 6G Non-Terrestrial Networks (NTNs).

翻译：本文基于实测数据，对郊区环境中的低空空地信道进行了全面的轨迹感知特性表征。采用64单元大规模多输入多输出阵列，针对无人机的三种飞行轨迹（包括两种固定高度的水平锯齿形飞行和一种垂直爬升）捕获信道数据。这些轨迹旨在模拟空中用户设备操作，并通过对方位角和俯仰角的可控扫描来分析几何依赖的传播动态。我们研究了大尺度功率变化及其与几何特征（如俯仰角、方位角和三维距离）的相关性，随后通过分布拟合和莱斯K因子估计分析了衰落特性。此外，利用相关矩阵距离量化了时域非平稳性，并通过角度平稳性跨度展示了信道特性如何随无人机运动而变化。我们还结合K因子和均方根时延扩展分析了频谱效率，强调了二者对链路性能的共同影响。结果表明，俯仰角是接收功率的最强预测因子，各轨迹相关性均超过0.77；而Nakagami模型最符合小尺度衰落特性。K因子从低空约5dB增至高空超过15dB，表明视距主导性增强。非平稳性模式高度依赖轨迹和几何构型：水平飞行中方位角影响最大，垂直飞行中俯仰角影响显著。这些发现为6G非地面网络中无人机通信信道的建模与优化提供了重要参考。