.png){kind=icon}
.png){kind=icon}
Return
Downlink two-stage precoding and cluster CSI feedback for FDD-based cell-free systems
.png){kind=icon}
PRE
AI
DOI：10.1016/j.phycom.2026.103078.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Cell-free system can improve cell-edge user experience in cellular networks. Given the prevalence of frequency division duplexing (FDD) in 5G networks, FDD-based cell-free system is poised to be crucial in future wireless networks. Accurate channel state information (CSI) feedback and efficient precoding algorithms are key to ensuring user experience rates in FDD systems. Existing CSI feedback algorithms lack precision, and precoding algorithms underperform in interference cancellation. In this paper, we propose a cluster-level CSI feedback scheme that leverages the sparsity of the distributed channel. This scheme constructs co-located user clusters in the angle and power domains and employs a round-robin feedback strategy within each cluster to significantly reduce feedback overhead. Furthermore, a two-stage precoding scheme is introduced to tackle interference challenges in multi-user scenarios. In the first stage of CSI acquisition, the precoding matrix is designed using inter-cluster statistical channel information to suppress dominant inter-cluster interference. In the second stage of scheduling, an optimized signal-to-leakage-plus-noise ratio (SLNR) algorithm is employed to dynamically mitigate real-time intra-cluster multi-user interference. Simulation results show that the cluster-level CSI feedback not only reduces the feedback overhead by 42% at the cost of moderate computational complexity at the base station, but also, when compared to Angle-MMSE precoding, the proposed two-stage precoding scheme achieves a 10.5% sum rate improvement. Furthermore, as user mobility increases, the cluster-level CSI and two-stage precoding scheme demonstrates superior robustness.
Keywords:
FDD-Based cell-free massive MIMO
Co-located user cluster
CSI Feedback
Two-stage precoding