Reconfigurable Intelligent Surfaces: How to Enhance Multiuser Channel Prediction?

Reconfigurable intelligent surface (RIS) is envisioned as the key enabler of a smart wireless environment, offering a nearly energy-neutral solution to control the dynamic propagation of electromagnetic signals. On the other hand, channel prediction can bring promising benefits in various aspects of wireless systems, e.g., reducing channel state information feedback overhead. Prediction of a highly dynamic channel is cumbersome. This letter focuses on enhancing the performance of multiuser channel prediction by introducing time correlation in the composite channel while exploiting quasi-RIS. We show that the intelligently configured RIS improves channel prediction performance while saving sufficient RIS elements for other applications and simplifies machine learning architecture.