Neural Network Based IRSs-UEs Association and IRSs Optimal Placement in Multi IRSs Aided Wireless System.

Area of Science:

• Wireless Communications
• Signal Processing
• Artificial Intelligence

Background:

• Beyond 5G networks face significant blockage issues at high frequencies, necessitating advanced solutions.
• Intelligent Reflecting Surfaces (IRSs) offer virtual line-of-sight (LOS) links to enhance spectral efficiency (SE) for user equipment (UEs).
• Optimal IRS-UE association and multi-IRS deployment are critical for maximizing system performance but pose computational challenges.

Purpose of the Study:

• To propose an efficient neural network (NN) based scheme for IRS-UE association in multi-IRS aided MIMO systems.
• To develop a criterion for optimal multi-IRS deployment to enhance network performance.
• To address the complexity and inefficiency of conventional association and deployment methods.

Main Methods:

• A neural network (NN) is trained using estimated angles of arrival (AoAs) to associate each UE with its optimal IRS.
• Passive beamforming is performed within each IRS after UE association.
• A deployment criterion is proposed that maximizes the average sum UE signal-to-interference-plus-noise ratio (SINR).

Main Results:

• The NN-based scheme achieves performance comparable to exhaustive and iterative search methods with significantly reduced complexity and delay.
• The proposed association method operates without real-time feedback signaling, improving power efficiency.
• The optimal deployment strategy outperforms reference methods, leading to better system and per-UE spectral efficiency.

Conclusions:

• The NN-based IRS-UE association scheme effectively enhances spectral efficiency in beyond 5G networks.
• Optimal multi-IRS placement is crucial for maximizing network performance, and the proposed criterion achieves this.
• This research offers a computationally efficient and practical solution for IRS deployment and management in future wireless systems.