Learning-Based Joint Antenna Selection and Precoding Design for Cell-Free MIMO Networks

TL;DR

The paper tackles downlink optimization in a cell-free MIMO network by jointly designing antenna selection and precoding. It introduces a distributed framework where each AP runs a CNN to select active antennas and a GNN to compute precoding using only local CSI, trained offline in a centralized manner with unsupervised loss for the GNN. An iterative-search-based dataset and a per-AP CNN enable scalable, fronthaul-efficient operation, achieving sum spectral efficiency close to centralized schemes while reducing computation time and eliminating cross-AP signaling. The approach demonstrates strong scalability, real-time feasibility, and substantial fronthaul savings, making it practical for large CF-MIMO deployments.

Abstract

This paper considers a downlink cell-free multiple-input multiple-output (MIMO) network in which multiple multi-antenna access points (APs) serve multiple users via coherent joint transmission. In order to reduce the energy consumption by radio frequency components, each AP selects a subset of antennas for downlink data transmission after estimating the channel state information (CSI). We aim to maximize the sum spectral efficiency by jointly optimizing the antenna selection and precoding design. To alleviate the fronthaul overhead and enable real-time network operation, we propose a distributed scalable machine learning algorithm. In particular, at each AP, we deploy a convolutional neural network (CNN) for antenna selection and a graph neural network (GNN) for precoding design. Different from conventional centralized solutions that require a large amount of CSI and signaling exchange among the APs, the proposed distributed machine learning algorithm takes only locally estimated CSI as input. With well-trained learning models, it is shown that the proposed algorithm significantly outperforms the distributed baseline schemes and achieves a sum spectral efficiency comparable to its centralized counterpart.

Figures (16)

• Figure 1: An illustration of the downlink cell-free MIMO network.
• Figure 2: An illustration of the TDD protocol.
• Figure 3: An illustration of Graph $i$. It is a fully connected undirected graph, where node $n_{i,k}$ represents user $k$, and $\textbf{v}_{i,k}$ is its feature vector.
• Figure 4: The structure of GNN $i$.
• Figure 5: An illustration of graph convolutional layer $\text{GNN}_{i}^{l}$.