• Computer Science > Networking and Internet Architecture [Submitted on 20 Feb 2026] Title:Graph-Neural Multi-Agent Coordination for Distributed Access-Point Selection in Cell-Free Massive MIMO View PDF HTML (experimental)Abstract:Cell-free massive MIMO (CFmMIMO) systems require scalable and reliable distributed coordination mechanisms to operate under stringent communication and latency constraints. • A central challenge is the Access Point Selection (APS) problem, which seeks to determine the subset of serving Access Points (APs) for each User Equipment (UE) that can satisfy UEs’ Spectral Efficiency (SE) requirements while minimizing network power consumption. • We introduce APS-GNN, a scalable distributed multi-agent learning framework that decomposes APS into agents operating at the granularity of individual AP-UE connections. • Agents coordinate via local observation exchange over a novel Graph Neural Network (GNN) architecture and share parameters to reuse their knowledge and experience. • APS-GNN adopts a constrained reinforcement learning approach to provide agents with explicit observability of APS’ conflicting objectives, treating SE satisfaction as a cost and power reduction as a reward. • Both signals are defined locally, facilitating effective credit assignment and scalable coordination in large networks.

Article Summaries:

  • Researchers have developed APS‑GNN, a distributed multi‑agent learning framework that uses a graph neural network (GNN) to coordinate access‑point selection (APS) in cell‑free massive MIMO systems. Each agent controls an AP‑UE link, exchanging local observations over the GNN and sharing parameters to reuse knowledge. The framework employs constrained reinforcement learning, treating spectral‑efficiency satisfaction as a cost and power reduction as a reward, with policy initialization via supervised imitation of a heuristic baseline. In realistic simulations, APS‑GNN meets target spectral‑efficiency while activating 50‑70 % fewer access points than heuristic or centralized MARL baselines, and achieves one to two orders of magnitude lower inference latency, demonstrating its scalability and practicality.

Sources: