Joint Beamforming and Scheduling for a MIMO Downlink with Random Arrivals

We study a number of downlink signaling strategies under perfect and non-perfect channel state information at the transmitter (CSIT) for the case of random packet arrivals. Under this setting, the relevant fairness criterion is the stabilization of all buffer queues which guarantees a bounded average delay for all users. In the case of perfect CSIT, different linear beamforming schemes are compared with the stability optimal policy based on dirty-paper coding (DPC). It is found that simple linear beamforming schemes with greedy user selection achieve near DPC performance in terms of the average delay. In the case of non-perfect CSIT due to a delayed feedback, we propose a novel greedy beamforming (heuristic) scheme together with "analog feedback" of CSI, where each user sends back its estimated channel vector without quantizing and coding. The proposed scheme outperforms the previously proposed schemes such as opportunistic beamforming and requires a similar feedback overhead