Greedy user selection for zero-forcing and MMSE multiuser beamforming with channel estimation errors

In a multi-user MIMO downlink where the base station has only estimates of the channels of the users, the sum-rate of multi-user beamforming saturates at high SNR. However, this is not the case for single-user beamforming. We propose a low-complexity user scheduling algorithm that selects the number of active users based on a closed-form approximation of the average sum-rate, and in particular does not add multiple users in the single-user optimality range. In order to develop this algorithm we derive the expected value of the rate of zero-forcing beamforming and MMSE beamforming with estimated channels and modify the greedy user selection accordingly. The gain of the proposed method is shown in numerical simulations.