Pilot Contamination-Aware Graph Attention Network for Power Control in CFmMIMO

TL;DR

The paper tackles real-time downlink power control in CFmMIMO under pilot contamination and varying UE activity by introducing a pilot contamination aware graph attention network that learns a mapping $\mathbf{M}=f(\boldsymbol{\Phi},\mathbf{B};\boldsymbol{\Theta})$ in a self-supervised manner. It builds a heterogeneous graph with AP-type and UE-type edges, embedding pilot contamination information through $\boldsymbol{\Phi}$, and employs a multi-layer attention mechanism to update AP-UE pair representations before projecting to the feasible power-control set. The method preprocesses large-scale fading, handles varying numbers of UEs with a padding strategy, and uses a differentiable objective to optimize max-min SE without labeled targets. Results show the proposed GAT achieves performance comparable to the accelerated projected gradient baseline while delivering substantial runtime speedups, and it demonstrates robustness to channel estimation errors and pilot contamination, making it viable for real-time CFmMIMO power control.

Abstract

Optimization-based power control algorithms are predominantly iterative with high computational complexity, making them impractical for real-time applications in cell-free massive multiple-input multiple-output (CFmMIMO) systems. Learning-based methods have emerged as a promising alternative, and among them, graph neural networks (GNNs) have demonstrated their excellent performance in solving power control problems. However, all existing GNN-based approaches assume ideal orthogonality among pilot sequences for user equipments (UEs), which is unrealistic given that the number of UEs exceeds the available orthogonal pilot sequences in CFmMIMO schemes. Moreover, most learning-based methods assume a fixed number of UEs, whereas the number of active UEs varies over time in practice. Additionally, supervised training necessitates costly computational resources for computing the target power control solutions for a large volume of training samples. To address these issues, we propose a graph attention network for downlink power control in CFmMIMO systems that operates in a self-supervised manner while effectively handling pilot contamination and adapting to a dynamic number of UEs. Experimental results show its effectiveness, even in comparison to the optimal accelerated projected gradient method as a baseline.

Figures (5)

• Figure 1: Illustration of graph characteristics.
• Figure 2: Result comparison between $3$ methods across $5$ scenarios.
• Figure 3: Performance comparison for the proposed method trained with a varying number of UEs but tested with fixed UEs.
• Figure 4: Performance comparison for different estimation errors in Scen. 2
• Figure 5: Ablation study on pilot information awareness.