Electromagnetic Theory-Guided Finite-Aperture Shared-Modal Cascaded Channel Estimation for RIS-Assisted MIMO

Abstract

Reconfigurable intelligent surface (RIS)-assisted cascaded-channel estimation incurs a high training burden when each phase-state channel is estimated independently. Within a quasi-static interval, we express each such channel as a known finite-aperture modal excitation of propagation-dependent coefficients shared across states. Electromagnetic-theory-guided probes are projected into a feasible phase-only codebook, and column-normalized sparse recovery estimates the shared coefficients for observed-state reconstruction and unseen-state synthesis. In fixed-deployment simulations, the method achieves competitive held-out-state normalized mean-square error (NMSE) across the tested SNR and training-overhead ranges; phase-quantization, modal-order, and geometry sweeps quantify sensitivity.