[Paper Review] Optimizing a Binary Intelligent Reflecting Surface for OFDM Communications under Mutual Coupling

Emil Björnson|arXiv (Cornell University)|Jun 8, 2021

Advanced Wireless Communication TechnologiesEngineering10 references37 citations

TL;DR

The paper studies a wideband OFDM system aided by a binary-state IRS with mutual coupling and two unbalanced states per element, and proposes channel estimation and configuration methods validated on the SP Cup 2021 dataset.

ABSTRACT

An intelligent reflecting surface (IRS) can greatly improve the channel quality over a frequency-flat channel, if it is configured to reflect the incident signal as a beam towards the receiver. However, the fundamental limitations of the IRS technology become apparent over practical frequency-selective channels, where the same configuration must be used over the entire bandwidth. In this paper, we consider a wideband orthogonal frequency-division multiplexing (OFDM) system that is supported by a fairly realistic IRS setup with two unbalanced states per element and also mutual coupling. We describe the simulation setup considered in the IEEE Signal Processing Cup 2021, propose a low-complexity solution for channel estimation and IRS configuration, and evaluate it on that setup.

Motivation & Objective

• Motivate and model IRS-aided wideband OFDM communications with practical two-state, unbalanced, and mutually coupled IRS elements.
• Develop low-complexity channel estimation and IRS configuration methods suitable for binary IRS configurations.
• Demonstrate the proposed methods using the IEEE Signal Processing Cup 2021 dataset and code.

Proposed method

• Model the IRS as a two-state, mutually coupled array affecting an OFDM wideband channel.
• Derive a pseudo-baseband OFDM model and express the sum rate over subcarriers as R = (B/(K+M-1)) sum log2(1+P|h_theta[nu]|^2/(BN0)).
• Propose LS-based channel estimation exploiting an N×N Hadamard pilot matrix to estimate hd and V under coupling.
• Introduce dimension reduction by expressing V with a row-wise structure V = (1_{NV} ⊗ I_{NH})^T V_row and estimate reduced parameters.
• Develop a binary-constraint IRS configuration method using a power-method-inspired iterative projection to maximize the quadratic form c^H B c with c ∈ {±1}^{N+N_V}.
• Address noise estimation and mismatch in the pilot-to-Ω mapping to robustly estimate channels and configure the IRS.

Experimental results

Research questions

• RQ1 Can a practical two-state (±) binary IRS with mutual coupling be effectively estimated and configured for OFDM in wideband channels?
• RQ2 How does mutual coupling and amplitude unbalance affect IRS reflection and system performance in OFDM?
• RQ3 What low-complexity estimation and configuration algorithms can approach optimal rate with binary IRS constraints?
• RQ4 How can pilot design and dimension reduction enable reliable estimation of hd and V under coupling?
• RQ5 What performance gains can be achieved with the proposed methods on realistic SP Cup datasets?

Key findings

• The proposed LS-based channel estimation can identify hd and V under coupling using Hadamard pilot blocks.
• Dimension reduction via V = (1_NV ⊗ I_NH) V_row enables workable estimation with limited pilots.
• The binary-IRS configuration method with a power-method-like iteration achieves higher rates than selecting from pilot configurations or using a uniform surface, especially in NLOS scenarios.
• Across 50 UEs, the optimized IRS yields a sum rate approximately 3.3 times larger than a uniform surface in the reported dataset.
• The dataset-driven results demonstrate substantial performance gains even with two-state, unbalanced, and coupled IRS elements in wideband OFDM.