Joint Downlink-Uplink Channel Estimation for Non-Reciprocal RIS-Assisted Communications

TL;DR

This paper introduces a novel channel estimation method for non-reciprocal RIS-assisted MIMO systems, using a three-phase protocol and Tucker decomposition, reducing user terminal complexity while maintaining high accuracy.

Contribution

It proposes a new closed-loop three-phase protocol and a Tucker decomposition-based algorithm specifically designed for non-reciprocal RIS-assisted MIMO systems, shifting processing burden to the base station.

Findings

The proposed method achieves satisfactory channel estimation accuracy.

It outperforms benchmark FDD LS-based and tensor-based techniques.

The approach reduces user terminal hardware complexity.

Abstract

Reconfigurable intelligent surface (RIS) is a recent low-cost and energy-efficient technology with potential applicability for future wireless communications. Performance gains achieved by employing RIS directly depend on accurate channel estimation (CE). It is common in the literature to assume channel reciprocity due to the facilities provided by this assumption, such as no channel feedback, beamforming simplification, and latency reduction. However, in practice, due to hardware limitations at the RIS and transceivers, the channel non-reciprocity may occur naturally, so such behavior needs to be considered. In this paper, we focus on the CE problem in a non-reciprocal RIS-assisted multiple-input multiple-output (MIMO) wireless communication system. Making use of a novel closed-loop three-phase protocol for non-reciprocal CE estimation, we propose a two-stage fourth-order Tucker decomposition-based CE algorithm. In contrast to classical time-division duplexing (TDD) and frequency-division duplexing (FDD) approaches the proposed method concentrates all the processing burden for CE on the base station (BS) side, thereby freeing hardware-limited user terminal (UT) from this task. Our simulation results show that the proposed method has satisfactory performance in terms of CE accuracy compared to benchmark FDD LS-based and tensor-based techniques.