Rate-balancing in massive MIMO using statistical precoding

In a typical massive MIMO system, the limited coherence time of the wireless channels leads to interference during the uplink training phase, since pilot sequences have to be reused between different users. This interference ultimately limits the achievable rate when basic linear beamforming is used. We show how to push this limit by exploiting the statistical properties of the channel. Specifically, we use a lower bound on the achievable rate to formulate a rate balancing problem which is independent of the instantaneous channel state information but only depends on the second order information. The rate balancing problem is then solved by methods known from classical MIMO systems leading to beamforming vectors and power allocations which significantly outperform the standard matched filter or zero-forcing approaches.