<scp>RIS</scp> ‐Aided <scp>mmWave MIMO</scp> Channel Estimation Using Rotation‐Invariant Coordinate Convolutional Neural Network and Compressive Sensing

ABSTRACT Channel State Information (CSI) is necessary for wireless systems supported by reconfigurable intelligent surfaces to regulate wireless channels and increase bandwidth along energy efficiency. In this paper, RIS‐Aided mmWave MIMO Channel Estimation with Rotation‐Invariant Coordinate Convolutional Neural Network and Compressive Sensing (RIS‐MIMO‐CE‐RICCNN) is proposed. Reconfigurable Intelligent Surface (RIS) is utilized to estimate frequency‐flat and frequency‐selective cascaded channels with minimal overhead in multi‐user millimeter wave big multiple inputs multiple outputs (MIMO) systems. The unique angle cascaded channels visible to different users have totally shared non‐zero rows including user‐specific column supports; it makes use of both the double‐structured sparsity property of angular cascaded channel (ACC) matrices and the common sparsity property among the various subcarriers. Rotation‐Invariant Coordinate Convolutional Neural Networks (RICCNN) is used to accurately detect channel supports. Then, channel estimation is done by Harbor Seal Whiskers Optimization Algorithm (HSWOA). The performance metrics like Signal to Noise Ratio (SNR), Normalized mean square error (NMSE), Bit error rate (BER), Peak‐signal to noise ratio (PSNR), Computational complexity, Loss function, BER versus SNR and energy consumption are evaluated. The performance of the RIS‐MIMO‐CE‐RICCNN technique is evaluated against existing methods. The RIS‐MIMO‐CE‐RICCNN achieves 16.28%, 30.78% and 25.29% lower SNR, 15.08%, 20.58%, and 15.25% lower NMSE, 28.96%, 30.21%, and 23.89% lower BER, and 26.28%, 31.26%, 19.66% lower PSNR when compared to the existing models: RIS‐assisted mmWave MIMO channel estimation utilizing deep learning with compressive sensing (RIS‐MIMO‐CE‐CS), channel estimation for reconfigurable intelligent surface enabled multiple user mmWave MIMO systems (RIS‐MIMO‐CE‐MU), Beam Pattern along Reflection Pattern Design of Channel Estimation in RIS‐Assisted mmWave MIMO Schemes (BPRPD‐CE‐RIS‐MIMO)respectively.