Uplink Multi-User Massive MIMO Using Gaussian BP in Highly Correlated Actual Environments

This paper focuses on multi-user detection in uplink Massive multiple-input multiple-output (MIMO) systems. As a low computational complexity signal detection algorithm, matched filter (MF)-based Gaussian belief propagation (MF-GaBP) using adaptively scaled belief (ASB) has been proposed. In addition, to suppress the negative impact of spatial correlation, an MMSE detector is introduced as the prior stage processing of MF-GaBP. Through computer simulations, this algorithm exhibits high performance even in high- spatially loaded Massive MIMO scenarios. However, in actual radio environments, there is a possibility to observe an extremely high channel correlation among some antenna elements or some users due to co-located polarization antenna elements or closely deployed mobile stations (MS). The highly correlated channel results in the performance degradation of MF-GaBP using ASB because of the low reliability of initial beliefs. This paper aims to verify the performances of MF- GaBP in the actual radio environments. The predetermined parameters in MF- GaBP are also adjusted according to actual environments for improving the convergence property of the iterative detection. To obtain propagation channel data for performance evaluation in the actual environments, an indoor experimental trial using Massive MMO equipment with 32 antenna elements was carried out under a condition that 16 MSs are closely deployed. Computer simulations using the obtained propagation channel data validate that simply applying the MMSE detector before MF-GaBP with ASB suppress an error floor in the bit error rate (BER) performance even in highly correlated channel.