Intelligent Reflecting Surfaces for Wireless Communications: Low-Complexity Techniques and Performance Analysis

Abstract

Intelligent reflecting surface (IRS), also referred to as reconfigurable intelligent surface (RIS), is an emerging technology aimed at dynamically controlling the propagation environment of next-generation wireless systems. An IRS comprises a large array of passive phase shifters whose configurations can be adaptively adjusted to co-phase the reflected signals at the receiver, thereby mitigating the channel impairments such as fading and shadowing. Despite its promising benefits, the practical deployment of IRS entails several signal processing challenges that require careful investigation. This thesis addresses three fundamental problems in this context, with an emphasis on low-complexity solutions and rigorous performance analysis of IRS-assisted wireless systems.

The first part of this thesis focuses on developing a low-complexity approach to harness the full potential of IRS-aided systems. Conventionally, optimizing the IRS phase configuration involves a three-stage procedure: (i) estimating the channels of all the links, (ii) computing the optimal phase shifts, and (iii) transporting the phase angles from the base station (BS) to the IRS, resulting in a three-fold overhead. To circumvent this, we develop an alternative scheme wherein, in every time slot, the IRS phases are randomly sampled, and a user equipment (UE) is scheduled opportunistically for data transmission. The central idea is that, with a sufficiently large number of UEs, a randomly chosen IRS configuration is likely to be near-optimal (i.e., beamforming-aligned) for at least one UE. Scheduling such a UE, for e.g., using a proportional-fair (PF) scheduler, for data transmission allows us to obtain most of the IRS gains without explicitly optimizing the IRS, thereby eliminating the associated overheads. Moreover, it is easy to find the UE that experiences the highest SNR (or PF metric) for a given randomly chosen phase configuration using one of several splitting or timer-based methods. We perform an in-depth analysis of this scheme under various channel models and frequency bands. For example, we show that the number of UEs required to attain a spectral efficiency (SE) close to the beamforming SE grows exponentially with the rank of the IRS spatial covariance matrix. We also propose alternative strategies that improve this scaling. Furthermore, we extend the scheme to frequency-selective channels using orthogonal frequency-division multiplexing (OFDM) and characterize the resulting performance tradeoffs.

The second part of this thesis explores the impact of IRS deployment in a multi-operator wireless system, where each operator serves its UEs over non-overlapping frequency bands. Given that IRSs are inherently passive and lack active RF components such as band-pass filters, they cannot selectively reflect signals confined to specific frequency bands. This raises a key question: if an IRS is deployed and controlled by one operator, how does it affect the performance of out-of-band (OOB) operators operating in adjacent frequency bands? To address this, we first analyze a two-operator scenario where one operator controls a single IRS. Our findings reveal that, although the IRS introduces random phase shifts to the OOB signals, the IRS elements collectively enhance the OOB UE performance, albeit with lower gains than those experienced by the controlling operator. Specifically, in sub-6 GHz bands, the IRS induces rich scattering for the OOB UEs, while in mmWave bands, it increases the likelihood of virtual line-of-sight (LoS) paths. We further show that deploying multiple IRSs amplifies these benefits for OOB users due to spatial diversity. Further, for ease of implementing a multiple-IRS setup, we propose a low-complexity channel estimation technique tailored for distributed IRS deployments, which notably maintains constant pilot overhead when the number of IRSs does not exceed a threshold. Lastly, we generalize our analysis to multi-operator networks where each operator controls its own IRS, and investigate the limits of inter-operator cooperation gains in realistic wireless environments.

The final part of this thesis investigates the performance of wideband beamforming with IRS, where the IRS configuration is designed to beamform wideband signals. A fundamental challenge in this case arises due to the frequency-flat nature of phase shifters, which limits their ability to focus energy uniformly across a wide frequency band. This results in a spatial-wideband effect at the UE, leading to the beam-squint or beam-split (B-SP) effect in the frequency domain, causing a degradation of the array gain and the achievable throughput. To address this issue, we propose two low-complexity mitigation strategies. First, we introduce a distributed IRS architecture in which multiple smaller IRSs are strategically deployed. This architecture reduces the severity of the spatial-wideband effect by parallelizing the spatial delay spread and exploiting the angle diversity. However, it may introduce temporal delay spread (TDS) unless signals from different sub-IRSs arrive within the same sampling bin. In this view, we formulate and solve an optimization problem to determine the optimal placement of IRSs that minimizes TDS over a set of all possible UE locations. In the second approach, we leverage the law of energy conservation to show that, under the B-SP effect, the IRS must inherently form different beams across different frequency components. Exploiting this insight, we propose an opportunistic OFDMA framework that assigns different subbands to different UEs, enabling the system to harness full array gain across the entire bandwidth through multi-user diversity. Lastly, we observe that in sub-6 GHz bands, frequency selectivity is dominated by rich multipath effects rather than the B-SP phenomenon. Accordingly, we develop a joint IRS phase optimization strategy using a majorization-minimization framework tailored for OFDM systems that maximizes the overall sum-rate at a given UE.

In summary, this thesis explores three core aspects of IRS-aided wireless systems: opportunistic user scheduling, performance of IRS-aided multi-operator systems, and wideband beamforming strategies. It demonstrates how even a randomly configured IRS can extract most of its benefits and improve the performance of wireless communication systems. The major takeaways are as follows:

1) We develop novel opportunistic scheduling schemes that employ randomly selected IRS phase configurations and leverage multi-user diversity. These schemes deliver near-optimal IRS gains without explicit optimization of the IRS phase configurations or incurring the associated signaling overheads.

2) We analyze the impact of the IRS in multi-operator environments, particularly on the OOB users. A key finding is that IRSs, despite being controlled by one operator, can enhance the performance of OOB operators by improving their effective channels.

3) Finally, we develop low-complexity techniques to enable wideband beamforming using phase-shifter-based IRSs. In particular, we show that a distributed IRS architecture inherently mitigates the B-SP effect, while an opportunistic OFDMA strategy exploits it to enhance system throughput under a randomly configured IRS phase.

By benchmarking the proposed techniques in this thesis against other existing methods, we show that our solutions can achieve superior performance while maintaining low time and computational complexities.