Adaptive Multilayer Contrastive Graph Neural Networks for Channel Estimation in Reconfigurable Intelligent Surface‐Aided MIMO Systems

ABSTRACT Channel state information (CSI) is necessary for wireless systems supported by reconfigurable intelligent surfaces (RIS) to regulate wireless channels and maximize bandwidth as well as energy effectiveness. In this paper, an Adaptive Multilayer Contrastive Graph Neural Networks for Channel Estimation in Reconfigurable Intelligent Surface‐Aided MIMO Schemes (RIS‐MIMO‐ CE ‐MCGNN) is proposed. Using reconfigurable intelligent surfaces, multiuser millimeter wave big MIMO systems with lower overhead may estimate frequency flat and frequency‐selective cascaded channels. Then uses a double‐structured sparsity property of the angular cascaded channel matrices and the common sparsity property among the different subcarriers since distinct angle cascaded channels that are visible to various users have completely shared nonzero rows together with user‐specific column supports. Support Detection using Adaptive Multilayer Contrastive Graph Neural Networks (AMCGNN). Then, the Channel estimation is performed using the Wader Hunt Optimization Algorithm (WHOA) technique. Finally, the system seeks to minimize the quantity of pilot overhead needed for precise channel estimate. The proposed technique attains 6.18%, 5.58% and 7.29% higher SNR and 7.88%, 5.88%, and 7.19% higher PSNR comparing with the existing methods: Channel estimation for reconfigurable intelligent surface‐dependent full‐duplex MIMO with hardware impairments ( CE ‐RIS‐MIMO‐HI), RIS‐supported Multiuser MIMO Systems Exploiting Extreme Learning Machine (RIS‐MU‐MIMO‐EELM), and RIS‐assisted mmWave‐MIMO channel estimation utilizing DL and compressive sensing (RIS‐MIMO‐ CE ‐CS), respectively.