Scholarlink

<p>This paper proposes an Interference-Aware Complex Phasor Message Passing Network (CPMPN) for joint beamforming optimization in multi-user IRS-assisted systems. The proposed framework introduces a complex-valued graph representation that preserves the amplitude and phase characteristics of wireless channels and employs an interference-aware message passing mechanism to capture constructive and destructive signal interactions. A feasibility-preserving phase output layer guarantees unit-modulus IRS reflection coefficients, while a joint prediction head simultaneously generates IRS phase shifts and base station beamforming vectors. The proposed method is evaluated using the DeepMIMO O1 ray-tracing dataset under realistic millimeter-wave propagation conditions. Experimental results show that CPMPN achieves an achievable sum rate of 25.3 bps/Hz at 20 dB SNR, outperforming Alternating Optimization (23.4 bps/Hz), Real-Valued GNN (21.9 bps/Hz), Random IRS Phase (14.8 bps/Hz), and No-IRS transmission (13.6 bps/Hz). Moreover, the proposed framework approaches the semidefinite relaxation (SDR) upper bound of 27.1 bps/Hz with an optimality gap of only 3.3%, while maintaining substantially lower computational complexity through single-pass inference. These findings demonstrate that physics-aware complex graph learning provides an effective and scalable solution for real-time beamforming optimization in IRS-assisted 6G wireless systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Interference-aware complex phasor message passing network for joint beamforming optimization in multi-user IRS-assisted 6G systems

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月4日

查看全文
  • Mustafa Ihsan Mustafa,
  • Osman Nuri Ucan

摘要

This paper proposes an Interference-Aware Complex Phasor Message Passing Network (CPMPN) for joint beamforming optimization in multi-user IRS-assisted systems. The proposed framework introduces a complex-valued graph representation that preserves the amplitude and phase characteristics of wireless channels and employs an interference-aware message passing mechanism to capture constructive and destructive signal interactions. A feasibility-preserving phase output layer guarantees unit-modulus IRS reflection coefficients, while a joint prediction head simultaneously generates IRS phase shifts and base station beamforming vectors. The proposed method is evaluated using the DeepMIMO O1 ray-tracing dataset under realistic millimeter-wave propagation conditions. Experimental results show that CPMPN achieves an achievable sum rate of 25.3 bps/Hz at 20 dB SNR, outperforming Alternating Optimization (23.4 bps/Hz), Real-Valued GNN (21.9 bps/Hz), Random IRS Phase (14.8 bps/Hz), and No-IRS transmission (13.6 bps/Hz). Moreover, the proposed framework approaches the semidefinite relaxation (SDR) upper bound of 27.1 bps/Hz with an optimality gap of only 3.3%, while maintaining substantially lower computational complexity through single-pass inference. These findings demonstrate that physics-aware complex graph learning provides an effective and scalable solution for real-time beamforming optimization in IRS-assisted 6G wireless systems.