Channel Estimation for Reconfigurable Intelligent Surface Aided MISO Communications: From LMMSE to Deep Learning Solutions

We consider multi-antenna wireless systems aided by reconfigurable intelligent surfaces (RIS). RIS presents a new physical layer technology for improving coverage and energy efficiency intelligently controlling the propagation environment. In practice however, achieving anticipated gains of requires accurate channel estimation. Recent attempts to solve this problem have considered least-squares (LS) approach, which is simple but also sub-optimal. The optimal estimator, based on minimum mean-squared-error (MMSE) criterion, challenging obtain non-linear due non-Gaussianity effective seen at receiver. Here we present approaches approximate MMSE estimator. As first analytically develop best linear LMMSE, together with corresponding majorization-minimization-based algorithm designed optimize phase shift matrix during training phase. This estimator shown yield improved accuracy over LS approach exploiting second-order statistical properties noise. To further improve performance better globally-optimal propose data-driven solutions deep learning. Specifically, posing estimation as an image denoising problem, two convolutional neural network (CNN)-based methods perform solution. Our numerical results show that these CNN-based estimators give superior compared approaches. They low computational complexity requirements, thereby motivating their potential use in future RIS-aided communication systems.