Wireless Fronthauls in Full-Duplex Cell-Free Massive MIMO Systems

Abstract

In this paper, we investigate a full-duplex (FD) cell-free massive multiple-input multiple-output (CF mMIMO) architecture with millimeter wave (mmWave) fronthaul, where uplink (UL) and downlink (DL) payload data and control signaling must be simultaneously supported. We first revisit the fronthaul requirements of representative wired low physical layer functional splits and show that the FD operation further aggravates the wired fronthaul bottleneck. To improve scalability beyond purely wired deployments, we propose a wireless fronthaul architecture in which the fronthaul links between access points (APs) and the central processing unit (CPU) operate over mmWave bands that are spectrally disjoint from the sub-6 GHz access links. Then, instead of forwarding antenna-domain baseband samples, we exploit low-dimensional sufficient statistics and develop a wireless fronthaul transmission framework. For the DL, Gram-regularized zero-forcing (Gram-RZF) and Gram-weighted minimum mean-square error (Gram-WMMSE) beamforming methods are designed using user-domain Gram matrices, thereby avoiding the transport of instantaneous channel state information. For the UL, the remaining two phases convey local UL signal estimates and slow-timescale second-order moments, enabling centralized large-scale fading decoding (LSFD) at the CPU. All UL information is delivered through subspace-domain wireless fronthaul transmission, together with a receive-subspace demultiplexing mechanism at the CPU for reliable packet recovery. Numerical results validate the proposed framework, demonstrating remarkable improvements over conventional half-duplex CF mMIMO.