Group Decoders for Correlated Massive MIMO Systems: The Use of Random Matrix Theory

Future MIMO terminals are expected to be equipped with a higher number of antennas. A possible intermediary solution between using the optimal but complex SIC decoder and the simple but poor performing MMSE decoder, is to consider different group decoders, namely, group SIC (GSIC) and group parallel decoding (GPD). The performances of such decoders strongly depend on the grouping strategy. In this paper, we introduce a unified framework to handle different group decoders. We use tools from random matrix theory to present a tight approximation of the average sum rate achieved in the case of adopting any kind of these decoders using only statistical CSI. For large number of data streams and fast changing channels, finding the optimal grouping for each channel realization is very complex. We formulate an optimization problem in which we use our developed approximations to find a static grouping which is shown to lead to a performance near to the optimal SIC and much better than the MMSE, especially for high transmit correlation. We also show how the performance depends on different system parameters, such as correlation strength and number of groups.