Design of an Integrated Framework for EIT-Based Capacity and Channel Estimation, and Deep-Unfolded Beamforming in MIMO Systems

This paper establishes a unified framework for Multiple-Input Multiple-Output (MIMO) systems by bridging the gap between electromagnetic information theory (EIT) and practical discrete-space algorithms. First, we provide an ergodic capacity analysis based on prolate spheroidal wave functions (PSWFs) to address the capacity saturation phenomenon in extremely large-scale MIMO (XL-MIMO) and continuous-aperture MIMO (CAP-MIMO). Leveraging this EIT-based physical insight, we propose an enhanced PSWF-based channel estimation (PSWF-CE) method that outperforms traditional MMSE and compressed sensing approaches. To optimize the system’s operational efficiency, we further integrate a deep unfolding network into the Fast Fractional Programming (FastFP) algorithm for multi-cell beamforming. This mixed learning-and-optimization approach eliminates the need for large matrix inversions and complex parameter tuning. The proposed framework is rigorously evaluated through extensive simulations using MATLAB for physical layer modeling and NS-۳ for network-level performance analysis. Numerical results demonstrate that the proposed integrated approach significantly improves spectral efficiency and computational speed, providing a robust solution for next-generation wireless networks.