Multi-user multi-input multi-output (MU-MIMO) techniques, which are capable of transmitting radio signals through the same radio resource by spatial multiplexing, can improve capacity. However, this requires the separation of spatially multiplexed signals at the base station. Recently, signal separation techniques using iterative algorithms such as Gaussian belief propagation (GaBP) which is capable of separating signals with high accuracy have been attracting significant attention. In open radio access networks (O-RAN) which are becoming used in actual systems, the base station is configured with a radio unit (RU) which receives radio signals, and a distributed unit (DU) which separates signals. Reducing the data sent from the RU to the DU can lead to cost savings for the base station.

In this paper, we propose two receive beamforming methods aimed at reducing the amount of data sent from the RU to the DU without diminishing the quality of separation by GaBP. GaBP requires unitary matrix-based receive beamforming, considering the characteristics of GaBP. One proposed method uses a left singular matrix derived from singular-value-decomposition, and the other one uses a Q matrix derived from the QR-decomposition of the channel matrix. Singular-value-decomposition-based beamforming or QR-decomposition-based beamforming can reduce the amount of data sent from RU to DU without sacrificing signal power.

Furthermore, receive beamforming methods based on singular-value-decomposition or QR-decomposition can reduce the correlation of received signals that decrease the separation accuracy of GaBP. This not only reduces the amount of data, but also contributes to enhancing the performance of GaBP. Specifically, singular-value-decomposition-based beamforming decorrelates the received signals, and simulation results demonstrate a significant improvement in the separation performance of GaBP. Additionally, evaluations of the computational cost required for receive beamforming and separation processing show significant reductions compared to methods without receive beamforming.

The combination of conventional signal separation and receive beamforming techniques, in conjunction with a receive beamforming method that considers the characteristics of GaBP, shows promise in reducing costs for base stations. This contribution is highly appreciated and makes the paper worthy of this distinguished paper award.