On the performance of covariance shaping in massive MIMO systems

The low-rank behavior of channel covariance matrices in massive multiple-input multiple-output (MIMO) settings can be exploited for pilot decontamination and statistical precoding. In this regard, many well-known algorithms are designed building on the assumption of signal subspace separation among user equipments (UEs), which is hardly satisfied in practice. To cope with this issue, our prior work introduced the concept of covariance shaping to enforce statistical orthogonality by exploiting the multiple antennas at the UE-side in the simple case of two UEs. This paper focuses on: i) extending the covariance shaping approach to a more general multi-user setting, and ii) evaluating its performance in several realistic scenarios, providing useful insights into its behavior and practical implementation.