Radio Map-Empowered Fast CSI Tracking for RIS-Enabled Low-Altitude UAV Communications

This paper addresses the challenges of extending cellular network coverage into low-altitude airspace by deploying reconfigurable intelligent surfaces (RISs) near base stations (BSs) to serve unmanned aerial vehicles (UAVs). Existing approaches typically assume the channel is temporally and spatially correlated and the UAV location is known. Without requiring the linearity or continuity of the channel state information (CSI) transition process. we propose to recover the CSI evolution by leveraging a channel covariance-embeded radio map. At each time slot, one pilot is used to observe the high dimensional RISUAV channel. By exploiting both the radio map and the mobility model, we develop a strategy to use sequential one-pilot CSI measurements to jointly estimate the UAV location and CSI. Additionally, the phase shift vector of RIS is adaptively designed with an entropy-minimizing approach, enhancing observation quality under stochastic channel conditions. To initialize phase shifts effectively, a Random-and-Tracking strategy is introduced, combining early-stage random sensing with refined CSI tracking via eigenvector-based sensing. Numerical results, using ray-traced channel data over real city maps, demonstrate that our proposed framework achieves over 94% of the capacity attained with perfect channel knowledge at 15 dB signal-to-noise ratio (SNR), with a UAV speed of $50 ~\text{km} / \mathrm{h}$.