Intelligent Reflecting Surface Aided Multicasting With Random Passive Beamforming

Qin Tao, Shuowen Zhang, Caijun Zhong, Rui ZhangIEEE Wireless Communications Letters 2021 · DOI

TL;DR

This paper introduces a novel "random passive beamforming" scheme for Intelligent Reflecting Surfaces (IRS) in multicast systems. By randomly changing IRS reflection patterns over short time slots instead of relying on complex channel state information (CSI), this approach significantly reduces training overhead and outperforms traditional CSI-based methods, especially in large-scale systems.

What Problem Does This Paper Address?

Intelligent Reflecting Surfaces (IRS) are a promising technology for enhancing wireless communication by intelligently reflecting signals. However, optimizing IRS performance typically requires accurate channel state information (CSI) for all links involved. Acquiring this CSI is a major challenge, as the training overhead grows linearly with the number of IRS reflecting elements (N), making it impractical for the large N envisioned for IRS. This paper addresses the need for an IRS-based multicast system that can achieve good performance without the burden of CSI acquisition.

Methods

The core innovation is a random passive beamforming scheme. Instead of using CSI to precisely control each IRS element's phase, the IRS randomly adjusts its reflection pattern over multiple "reflecting slots" within a single channel coherence interval. The system divides each coherence interval into $Q$ slots, and in each slot, the IRS applies an independent, random reflection pattern. This eliminates the need for CSI acquisition for the IRS itself.

The authors derive a closed-form approximation for the outage probability of this random scheme, particularly in the high signal-to-noise ratio (SNR) regime. They also establish a lower bound for the outage probability of traditional CSI-based beamforming for comparison. The analysis focuses on finding the optimal number of reflecting slots ($Q$) to balance the trade-off between reduced training overhead and improved channel diversity.

Datasets

Not discussed in the paper. The evaluation is based on simulations with synthetic channel models.

Key Insights and Ideas

  • CSI-Free IRS is Feasible and Effective: The paper demonstrates that a practical IRS system can operate effectively without requiring full CSI, a significant hurdle for current CSI-dependent designs.
  • Randomness as a Solution: Introducing randomness in IRS reflection patterns, over multiple short slots, can effectively mimic the benefits of beamforming by averaging out channel fades, even without explicit CSI.
  • Optimal Q Trade-off: The number of random reflection slots ($Q$) is a critical parameter. A small $Q$ is beneficial in high-outage (low-SNR) scenarios, while a larger $Q$ is preferred in low-outage (high-SNR) scenarios. This highlights the need for dynamic adjustment of $Q$ based on system conditions.
  • Scalability Advantage: The proposed random scheme's training overhead scales with $Q$, which is typically much smaller than $N$, making it highly advantageous for systems with a large number of IRS elements.

Key Results

  • The proposed random passive beamforming scheme significantly outperforms traditional CSI-based passive beamforming when the number of IRS elements ($N$) and/or the number of users ($K$) are large, especially when considering training overhead.
  • The derived closed-form approximation for outage probability is shown to be accurate, validating the theoretical analysis.
  • Numerical results show that the optimal $Q$ increases with $N$ and transmit power ($P$), but decreases with the rate target ($\tau$).
  • In large-scale systems (e.g., large $N$ and $K$), the CSI-based scheme's performance degrades due to training overhead, while the random scheme maintains or improves its performance.

Limitations and Future Work

  • High-SNR Approximation: The primary theoretical analysis focuses on the high-SNR regime. While numerical results show good agreement even at lower SNRs, a full analysis for all SNR regimes would be beneficial.
  • Optimal Q Determination: While the paper shows how to find the optimal $Q$, the practical implementation of dynamically adjusting $Q$ based on real-time channel conditions or system parameters is a direction for future work.
  • Feedback Overhead Omission: The paper assumes feedback overhead for CSI is omitted for simplicity in the CSI-based scheme comparison. A more comprehensive comparison might consider this.
  • User Distribution: The paper assumes users are clustered locally. Investigating performance with more diverse user distributions could be an extension.

Summary updated

BibTeX
@article{tao2021intelligent,
  author = {Qin Tao and Shuowen Zhang and Caijun Zhong and Rui Zhang},
  title = {{Intelligent Reflecting Surface Aided Multicasting With Random Passive Beamforming}},
  journal = {IEEE Wireless Communications Letters},
  year = {2021},
  doi = {10.1109/lwc.2020.3021473},
}
Figure