Low complexity channel estimation massive MIMO OFDM using deformable long-range attention convolutional graph networks

Abstract Enabling performance improvements provided by reconfigurable intelligent surface (RIS)-assisted wireless communications requires precise channel estimation. However, none of the current techniques account for the impact of synchronization issues such as carrier frequency offset (CFO). To improve the accuracy of channel estimation in large Multiple-input, multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems, especially under rapid subcarrier variation, by developing a novel deep learning-based framework that surpasses traditional interpolation techniques in reliability, scalability, and performance under complex communication scenarios. The proposed ODS-DLRACG-Net SGA integrates multiple orthonormal discrete S-transform for signal reconstruction with a Deformable Long Range Attention Convolutional Graph Network, optimized by the Snow Geese Algorithm, to model nonlinear mappings between pilot symbols and channel responses effectively. Simulation results demonstrate that the proposed model significantly outperforms conventional techniques in large MIMO settings, achieving a lower bit error rate, root mean squared error, mean squared error, and normalized mean squared error (RMSE, MSE, and NMSE), thereby confirming its high accuracy and robustness in dynamic and high-dimensional wireless communication environments. ODS-DLRACG-Net SGA offers a robust and scalable solution for channel estimation in massive MIMO systems, effectively addressing challenges of fast-varying channels and nonlinear relationships, thus enabling more efficient and reliable data transmission in future communication networks.