Massive MIMO Statistical CSI Correlation: Measurements and Analyses

With knowledge of statistical channel state information (CSI) at the base station (BS), one can reduce the pilot overhead and improve the channel estimation performance in massive multiple-input multiple-output (MIMO) systems by exploiting the non-orthogonal pilot sequences. However, due to the limited uplink (UL) training resources, it is challenging to acquire the high-accuracy statistical CSI in real-time. To tackle this problem, an emerging technique is to construct a channel knowledge map (CKM) based on the statistical CSI. In order to construct such a CKM, it is necessary to analyze the spatial and temporal correlations of statistical CSI. Specifically, this paper analyzes the spatial and temporal correlations of beam domain channel power matrices (BDCPMs), obtained from channel measurements in a massive MIMO testbed. Furthermore, to verify the effectiveness of the stored aged BDCPMs, we also simulate channel estimation performance using BDCPMs measured at different dates. Both measured and simulated results demonstrate the effectiveness of the CKM based on statistical CSI.