This paper introduces a novel three-stage graph neural network (GNN) designed to optimize reconfigurable intelligent surface (RIS)-assisted multi-waveguide pinching-antenna (PA) systems for downlink information transmission. The GNN learns optimal PA positions and RIS phase shifts based on user locations and channel conditions, respectively, and subsequently determines beamforming vectors. The proposed GNN is trained unsupervised and offers implementation strategies for integration with convex optimization, balancing inference time and solution optimality. Numerical results demonstrate the GNN's effectiveness, generalization capability, performance reliability, and real-time applicability. AI
IMPACT This research could lead to more efficient wireless communication systems through advanced AI-driven optimization.
RANK_REASON The item is an academic paper detailing a novel GNN-enabled optimization approach for antenna systems. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
