Channel Estimation for IRS Aided MIMO System with Neural Network Solution

Intelligent Reflective Surface (IRS) is a promising technique for Beyond 5G and 6G wireless communications. IRS is comprised of plenty of passive reflecting elements that an external controller is able to control by software. To develop the great potential of IRS-aided wireless communication system, we should get a full understanding of Channel State Information (CSI). However, it’s an exceedingly challengeable task for channel estimation in IRS-aided system. In this paper, we propose a residual neural network to achieve cascaded channel estimation for an IRS-aided multiple-input multiple-output (MIMO) communication system. Numerical results prove the effectiveness and high robustness of our Deep Learning (DL) method.

I. Introduction

RECENTLY, Intelligent Reflective Surface (IRS) (also called Large Intelligent Surface or Software-Defined Surface) [1] - [2] has been considered as a possible future for Beyond 5G and 6G mobile communication with substantial benefits like low power consumption, easy deployment, low cost and low latency. IRS is a 2-D electromagnetic artificial surface comprised of a sizable number of inexpensive passive reflective components. Since each component can function independently and be controlled by means of external signals, we can easily manage the phase and magnitude of the incident signal in real time. By controlling the propagation environment, IRS is able to increase the effectiveness of the spectrum and the efficiency of energy.