Intelligent Spectrum Learning for Wireless Networks With Reconfigurable Intelligent Surfaces

Reconfigurable intelligent surface (RIS) has become a promising technology for enhancing the reliability of wireless communications, since an RIS is capable reflecting desired signals through appropriate phase shifts. However, intended that impinge upon are often mixed with interfering signals, which usually dynamic and unknown. In particular, received signal-to-interference-plus-noise ratio (SINR) may be degraded by reflected from RISs originate non-intended users. To tackle this issue, we introduce concept spectrum learning (ISL), uses appropriately trained convolutional neural network (CNN) at controller to help infer directly incident signals. By capitalizing on ISL, distributed control algorithm proposed maximize SINR dynamically configuring active/inactive binary status elements. Simulation results validate performance improvement offered deep demonstrate superiority ISL-aided approach.