A Decentralized Paradigm for Joint Clustering and Precoding in Cell-Free MIMO Networks

Leveraging its dynamic user-centric cluster topology, cell-free ultra-massive MIMO is poised to emerge as a key technology for facilitating ubiquitous connectivity in the 6G era. To reduce fronthaul load and computational complexity, this paper proposes a practical cell-free paradigm in which each cluster independently optimizes both clustering and precoding in a distributed manner, without aggregating complete CSI and data at the central controller. To address resource competition and inter-cluster interference inherent in the distributed architecture, a semi-coherent joint transmission (CJT) scheme is proposed. To cope with the resource competition, a cluster-cutting weight is introduced and updated heuristically, allowing for adaptive resource allocation among clusters. To alleviate interference, the effect of the current cluster on others is treated as a penalty term in the precoding design. Simulation results indicate that the proposed semi-CJT scheme outperforms in terms of both sum rate and throughput due to higher resource utilization rate and reduced computational complexity.