Channel estimation for double IRS assisted broadband single-user SISO communication

TL;DR

This paper investigates Bayesian channel estimation for a double IRS assisted broadband SISO system, deriving bounds and demonstrating the approximation of complex distributions with Rayleigh distribution as IRS elements increase.

Contribution

It introduces a Bayesian estimation framework for double IRS-assisted channels and uses numerical methods to evaluate the Bayesian CRLB without closed-form pdfs.

Findings

Bayesian CRLB can be approximated numerically for complex channel models.

The distribution of inner products can be approximated by Rayleigh distribution with more IRS elements.

MSE approaches the Bayesian CRLB in simulations.

Abstract

In this paper, two Intelligent reflecting surfaces (double IRS) assisted single-user single input single output (SISO) communication system is considered. The cascaded channels (mobile user (MU)$\rightarrow$IRS-1$\rightarrow$base station (BS), MU$\rightarrow$IRS-2$\rightarrow$BS and MU$\rightarrow$IRS-1$\rightarrow$IRS-2$\rightarrow$BS channels) are estimated under Bayesian setting. Here, the goal is to evaluate the performance of the estimator in case of MU$\rightarrow$IRS-1$\rightarrow$BS and MU$\rightarrow$IRS-2$\rightarrow$BS channel links using Bayesian Cramer-Rao lower bound (CRLB). Without the knowledge of closed form pdf of inner product of circularly symmetric complex Gaussian (CSCG) random vectors, we cannot obtain the fisher information. Hence, by numerical computation we obtain the Bayesian CRLB. In the simulation results, we show that we can approximate the pdf of the inner product of CSCG random vectors by a Rayleigh distribution by increasing the number of elements on the IRS, which is analogous to Central Limit Theorem (CLT). Also, the results convey that the mean squared error (MSE) almost matches with the Bayesian CRLB.