PaperlayerView exact revision on arXiv

Source-linked AI summary

Blind Interference Suppression for IRS-Aided Robust Wireless Communications

Tao Wang, Xiaohui Zhang, Hehe Ban, Yiwei Guo, Ming Yi

arXiv:2609.38859v1eess.SP

TL;DR

Accurate CSI for IRS-assisted interference suppression is challenging to obtain under strong interference, particularly from non-cooperative sources. The paper combines CSM and PPI to configure the IRS from received signal power without prior CSI; simulations report interference reduction to the noise level, improved SINR, and better performance than benchmark schemes.

  • Problem

    Existing approaches generally rely on CSI, but accurate acquisition entails pilot overhead and delay, and non-cooperative interference makes pilot-based interference-channel estimation infeasible.

    Source: §I · p. 1

  • Method

    The proposed scheme combines CSM-based phase alignment with PPI and configures the IRS using received signal power without prior CSI.

    Source: §I · p. 2, A. Overall Procedure · p. 5, C. PPI Algorithm · p. 6

  • Results

    The scheme reduces interference power to the noise level, improves SINR, and outperforms benchmark schemes under typical interference conditions.

    Source: §I · p. 2, §VII · p. 11

  • Takeaways & Limitations

    The scheme supports discrete and imperfect reflection coefficients, multi-UE scenarios, and low computational complexity.

    Source: §I · p. 2, §VII · p. 12

Abstract

from arXiv · show

The application of intelligent reflecting surfaces (IRSs) to suppress interference in wireless communication systems has recently attracted significant research attention. Most existing approaches rely on complete or partial channel state information (CSI) to configure the IRS. However, acquiring accurate CSI in IRS-assisted systems involves considerable pilot overhead and introduces non-negligible delays. This issue is further exacerbated under strong interference conditions, where interfering sources are typically non-cooperative, making CSI acquisition even more challenging. As a result, existing CSI-dependent interference suppression methods become difficult to deploy in practice. To address these limitations, we propose a novel blind interference suppression strategy that combines a proportional phase-inversion (PPI) algorithm with the conditional sample mean (CSM) method. The proposed approach determines the IRS configuration using only the received signal power, without requiring any prior CSI. We conduct a comprehensive performance evaluation by deriving the theoretical performance of the proposed scheme, which is subsequently verified through numerical simulations. Furthermore, simulation results across various parameter settings demonstrate that the proposed blind interference suppression scheme reduces the interference power to the level of noise, thereby achieving a marked signal-to-interference-plus-noise ratio (SINR) improvement and outperforms existing benchmark schemes.

I. INTRODUCTION

The paper motivates blind IRS interference suppression by the difficulty of acquiring CSI and proposes a received-power-based approach combining CSM and PPI. Its system model considers an IRS near the UE and seeks to reduce interference without CSI; simulations report suppression to the noise level and improved SINR.

I. INTRODUCTION

  • Prior studies generally assume perfect legitimate-channel knowledge, and some also assume partial interference-channel knowledge, while CSI acquisition faces practical challenges under aggressive interference.Non-cooperative interference makes pilot-based interference-channel estimation infeasible; estimating legitimate channels adds overhead and delay, and strong interference can make CSI obsolete before acquisition completes.Source: §I · pp. 1–2
  • Earlier blind suppression schemes assume ideal continuous phase and amplitude control, whereas discrete phase and constant-modulus constraints can prevent convergence or degrade performance.The cited ZO method may fail to converge, and adaptive beamforming may suffer significant degradation or lack a feasible solution under these practical constraints.Source: §I · p. 2
  • Combining CSM with PPI configures the IRS using only received signal power, without prior CSI or channel-distribution knowledge, and supports discrete, imperfect reflection coefficients and multi-user scenarios.The scheme is also reported to have low computational complexity.Source: §I · p. 2
  • The proposed scheme reduces interference to the noise power level, improves SINR, and outperforms benchmark schemes under typical interference conditions.The paper also derives theoretical performance and validates it through numerical simulations under various parameter settings.Source: §I · p. 2

II. SYSTEM MODEL

  • The system model places an IRS near the UE in the presence of an interferer and seeks an IRS phase vector that minimizes interference power without CSI.The UE controls the IRS; interference is defined as disturbances from uncontrollable sources, rather than controllable multi-user or inter-cell interference.Source: §II · p. 3

III. DEPLOYMENT SETTINGS FOR IRS-ENABLED INTERFERENCE SUPPRESSION

The deployment analysis relates the minimum number of IRS elements to IRS–UE distance and carrier wavelength. It gives example element counts across distances and frequency bands, while noting that very near deployments may introduce model mismatch and long distances may make deployment costly.

III. DEPLOYMENT SETTINGS FOR IRS-ENABLED INTERFERENCE SUPPRESSION

  • The element-count relationship is derived under assumptions including an IRS near the UE, an interferer much farther away, identically distributed path losses, and a line-of-sight-dominated IRS–UE link.The preceding analysis assumes perfect CSI and continuous IRS phase shifts; the following scheme addresses operation without CSI and with discrete phase shifts.Source: §III · p. 4
  • At 1 m, the minimum element count is approximately 7, 70, and 700 for the 0.3 GHz, 3 GHz, and 30 GHz bands, respectively.The minimum count is directly proportional to IRS–UE distance and inversely proportional to carrier wavelength.Source: §III · p. 4
  • At 0.1 m, the counts are approximately 0.7, 7, and 70 for 0.3 GHz, 3 GHz, and 30 GHz; at 10 m, they are approximately 70, 700, and 7000.The paper notes that 0.1 m may induce near-field model mismatch, while 10 m imposes prohibitive hardware and operational costs.Source: §III · p. 4

IV. PROPOSED BLIND INTERFERENCE SUPPRESSION SCHEME

The scheme configures IRS phases to make reflected interference combine destructively with the direct interference channel. It uses silent-mode RSS measurements to initialize phases with CSM, then applies PPI for suppression.

A. Overall Procedure

  • The UE triggers a periodic four-step cycle: silent-mode activation, random-phase RSS measurement, CSM phase initialization, and PPI interference suppression.The AP enters silent mode when measured interference exceeds a predefined threshold; the cycle repeats to adapt to dynamic interference conditions.Source: A. Overall Procedure · p. 5

B. CSM Algorithm

  • For each IRS element, CSM selects the discrete phase whose conditional sample mean of measured interference power is largest.The conditional mean is computed from trials in which that element used the given phase.Source: B. CSM Algorithm · p. 5
  • With θCSM, reflected interference components fall within ±ω/2 of the direct interference channel’s phase.This phase alignment provides the initialization for subsequent phase inversion.Source: B. CSM Algorithm · p. 5

C. PPI Algorithm

PPI searches for an IRS configuration that minimizes measured interference power by inverting selected element phases and deactivating the rest. The paper also describes the scheme’s extension to multi-UE settings and its robustness to imperfect reflection coefficients.

C. PPI Algorithm

  • PPI tests activated-element proportions and random element combinations, inverts selected phases by π, deactivates the rest, and chooses the configuration with minimum measured interference power.For each proportion S, trials randomly select round(SN) elements; the final configuration minimizes the measured interference-plus-noise power.Source: C. PPI Algorithm · pp. 5–6
  • Deactivating unselected elements reduces power consumption while maintaining interference suppression performance.The PPI search is over both the activated-element proportion and the specific element combination.Source: C. PPI Algorithm · p. 5
  • The combined CSM–PPI scheme configures the IRS using received signal power alone, without requiring CSI.CSM provides the initial configuration and PPI performs the final suppression search.Source: C. PPI Algorithm · pp. 5–6
  • In multi-UE settings, each UE can run the scheme with a dedicated IRS; non-serving IRS interference is described as relatively weak and not substantially degrading performance.The authors state that simulation results in Fig. 12 validate effectiveness in this setting.Source: C. PPI Algorithm · p. 6
  • The scheme is described as robust to imperfect IRS reflection coefficients because CSM and PPI require no prior model of coefficient characteristics.The authors state that simulation results in Fig. 13 validate this robustness.Source: C. PPI Algorithm · p. 6

V. PERFORMANCE ANALYSIS

The analysis characterizes interference and legitimate-signal power changes, then derives conditions for effective interference suppression based on the number of IRS elements. Its threshold examples indicate the deployment scale needed under common scenarios.

V. PERFORMANCE ANALYSIS

  • The analysis defines IPCR as the optimized-to-original interference-power ratio and LSPCR as the corresponding ratio for legitimate signal power.It assumes i.i.d. reflected interference channels and a uniform distribution of selected phase offsets over a specified interval.Source: §V · p. 6

A. Interference Power Change Ratio (IPCR)

  • The IPCR analysis minimizes a quadratic expected-power function over activation proportions from 0 to 1 and considers three cases for the optimum.Other channel distributions can be analyzed with minor modifications.Source: A. Interference Power Change Ratio… · p. 6
  • When deployed IRS elements are insufficient, activating all of them still cannot fully cancel the direct interference component.In this case, all N elements are activated.Source: A. Interference Power Change Ratio… · p. 7
  • Under common scenarios, effective suppression requires operation in the saturation regime, with threshold element counts ranging from several tens to several hundred.The threshold depends on the direct-to-reflected interference amplitude ratio and phase resolution.Source: A. Interference Power Change Ratio… · p. 7, Table II · p. 7
  • The analytical IPCR results assume an optimal activation proportion, whereas the proposed PPI algorithm also searches combinations of activated elements, whose additional gains are difficult to characterize analytically.The overall SINR improvements are evaluated numerically in Section VI-B.Source: A. Interference Power Change Ratio… · p. 7

B. Legitimate Signal Power Change Ratio (LSPCR)

The IRS configuration is based only on the interference channel, so optimization leaves the reflected legitimate-channel distributions unchanged. The subsection represents LSPCR and shows phase-alignment performance across sample sizes and phase levels.

B. Legitimate Signal Power Change Ratio (LSPCR)

  • Because IRS configuration depends on the interference channel, the reflected legitimate-channel distributions remain unchanged under the optimized configuration.The expected legitimate signal power is then given by the subsection’s stated expression.Source: B. Legitimate Signal Power Change… · p. 7
  • Fig. 5 presents CSM phase-alignment performance for N = 200 across K = 2, 4 and T = 200, 2000, 20000, 200000, with the red box marking the ideal phase-difference distribution.The caption identifies the plotted conditions and ideal reference.Source: Figure 5 · p. 8
  • The subsection represents LSPCR mathematically.The provided passage introduces the representation but does not state its resulting values.Source: B. Legitimate Signal Power Change… · p. 8

C. Complexity Analysis

The complexity analysis accounts for RSS measurements and computation in both CSM and PPI. It concludes that total computational complexity scales linearly with the number of IRS elements.

C. Complexity Analysis

  • The CSM algorithm requires T RSS measurements and O(NT) computation, while PPI requires M/s measurements and O(NM/s) computation.These are the two components of the proposed scheme’s computational complexity.Source: C. Complexity Analysis · p. 8
  • The total computational complexity scales linearly with N, and the total number of RSS measurements is T + M/s.The passage characterizes the measurement count as typically a few thousand in practice.Source: C. Complexity Analysis · p. 8
  • Because each RSS measurement takes one symbol duration, execution time corresponds to transmitting T + M/s symbols and typically ranges from milliseconds to seconds.The stated duration is a typical range, not a fixed runtime.Source: C. Complexity Analysis · p. 8

VI. NUMERICAL RESULTS

The numerical section compares simulations with theoretical predictions and examines CSM alignment under finite sampling. The reported curves closely match theory, while element activation saturates as the IRS deployment grows.

VI. NUMERICAL RESULTS

  • Monte-Carlo simulations use P1 = P2 = 30 dBm, noise power σ2 = −94 dBm, and |h0|2 = |g0|2 = −100 dB, yielding an original SNR of 24 dB without IRS.The evaluation sets T = 10N, s = 0.01, and M = 10; benchmarks include the upper bound, random phase, and RMS.Source: §VI · p. 8

A. Verification of Theoretical Analysis

  • As CSM sample size T increases, phase differences approach the ideal distribution, while finite-sample misalignment is worse for larger K.The passage notes imperfect alignment at T = 2000 and says alignment inaccuracies can degrade achievable SINR.Source: A. Verification of Theoretical Analysis · p. 9
  • IPCR falls below −20 dB, and simulations in Figs. 6–8 align closely with theoretical predictions, with minor discrepancies attributed to CSM imperfections.The simulation comparison covers activated-element count, IPCR, and LSPCR.Source: A. Verification of Theoretical Analysis · p. 9, Figure 7 · p. 9, Figure 8 · p. 9
  • The number of activated IRS elements initially rises with N and then saturates; the paper identifies the saturation regime as necessary for effective interference suppression.The figure plots activated-element count against total element count for K = 2 and K = 4.Source: A. Verification of Theoretical Analysis · p. 9, Figure 6 · p. 9
  • The legitimate signal power is only weakly enhanced because IRS configuration is independent of the legitimate channel.This result is reported alongside the theoretical verification curves.Source: A. Verification of Theoretical Analysis · p. 10

B. SINR Improvement

The proposed scheme improves SINR by suppressing interference, with performance evaluated across interference power, sample size, multi-UE deployment, and IRS reflection imperfections. The conclusion highlights operation without CSI and support for practical IRS implementations.

B. SINR Improvement

  • SINR gains exceed 20 dB before saturation, with the saturated result about 2 dB below the upper bound and interference suppressed to the noise level.Under strong interference, the proposed method also outperforms the CPP method, which uses perfect legitimate CSI but performs suboptimally in that regime.Source: B. SINR Improvement · p. 10
  • At P2 = 60 dBm, SINR rises from −30 dB to about 15 dB (a 45 dB gain); the proposed method shows the slowest SINR decline as interference increases.As interference diminishes, CPP outperforms the proposed method.Source: B. SINR Improvement · p. 10
  • At T = 1000, the proposed scheme reaches near-optimal performance, with SINR rising quickly with sample size before saturation.The RMS scheme improves more slowly with T; CPP and random phase are unaffected by T.Source: Figure 11 · p. 10, B. SINR Improvement · p. 11

C. Performance in Multi-UE Scenarios

D. Impact of Imperfect IRS Reflection Coefficients

  • SINR remains stable as the IRS reflection imperfection level ϵ varies from 0.01 to 0.2.The model uses β = 0.8 insertion loss and represents quantization errors, mutual coupling, and thermal drift as additive error to the ideal coefficient; coefficients with magnitude above one are renormalized.Source: D. Impact of Imperfect IRS Reflection… · p. 11, Figure 13 · p. 11

VII. CONCLUSION

  • The conclusion states that the scheme requires no CSI, supports discrete and imperfect reflection coefficients, scales to multi-UE scenarios, and has low computational complexity.The paper reports that numerical results show interference reduction to the noise level and improved SINR under typical interference conditions.Source: §VII · pp. 11–12

APPENDIX

The appendix derives the mean of W, invoking linearity of expectation and independence and considering uniformly distributed phases.

APPENDIX

  • The mean derivation uses linearity of expectation and independence of the variables.The passage introduces this step before presenting the mean of W.Source: §1 · p. 12
  • The appendix states the mean of W after evaluating the terms involving |gn|ejϕn.The supplied derivation also identifies the mean as a separate step.Source: §1 · p. 12
  • The phase variable ϕn is taken to be uniform on [π − ω/2, π + ω/2] in the mean derivation.The passage then proceeds to an expression for the mean.Source: §1 · p. 12

2. Variance of W

The appendix derives the variance of W using independence and stated moments, then gives its expected power as a quadratic expression in S.

2. Variance of W

  • The variance calculation treats |gn|ejϕn, n ∈ N, and V as independent.The passage identifies these independence assumptions before proceeding with the variance.Source: §2 · p. 12
  • For the stated moments, the derivation gives E[(|gn|ejϕn)2] = E[|gn|2] = ρ2 and separately gives E[|gn|ejϕn]2 = ρ2c2.Source: §2 · p. 12

3. Expected Power of W

  • The appendix gives the expected power as E[|W|2] = AS2 + BS + C.It presents total variance before substituting into the expected-power expression.Source: §3 · p. 12, §2 · p. 12
Loading 2609.38859v1…