Proportional fair space-time scheduling for wireless communications

We extend the proportional fair (PF) scheduling algorithm to systems with multiple antennas. There are K client users (each with a single antenna) and one base station (with n(R) antennas). We focus on the reverse link of the system, and assume a slow-fading channel where clients are moving with pedestrian speed. Qualcomm's original PF scheduling algorithm satisfies the PF criteria only when the communication is constrained to one user at a time with no power waterfilling. However, the original PF algorithm does not generalize easily when we have nR receive antennas at the base station. In this paper, we shall formulate the PF scheduling design as a convex optimization problem. One challenge is in the optimal power allocation over the multiantenna multiaccess capacity region, which is still an open problem. For practical consideration, we consider multiuser minimum mean-square error processing at the base station. To obtain first-order insight, we propose an asymptotically optimal PF scheduling solution. Using the proposed PF solution for a multiantenna base station, the system capacity is enhanced by exploiting the multiuser selection diversity, as well as the distributed multiple-input multiple-output configuration. It is found that the PF scheduler achieves a good balance between fairness and system capacity gain.