Downlink multi-user MIMO scheduling with performance guarantees

We consider a long-standing open problem pertaining to scheduling over a wideband multi-user downlink. In this problem a base-station (BS) must assign multiple subbands to its served users such that a weighted sum rate metric is maximized subject to sum power and cardinality constraints. On each subband multiple users can simultaneously be scheduled. Such scheduling is complicated by the fact that the rate achieved by a user on any subband assigned to it depends not only on its own channel condition, but on the set of other users co-scheduled on that subband as well. The latter dependence is via the transmission scheme adopted by the BS in order to simultaneously serve multiple users on the same subband. This problem has received wide attention for over a decade and while numerous heuristics have been designed, there is no known algorithm that offers provable constant-factor worst-case guarantee. In this paper we obtain an important result which demonstrates that when the transmitter employs capacity-optimal dirty paper coding, constant-factor approximation guarantee can be achieved via simple algorithms. Indeed, we show that for a wideband scheduling problem in which a permissible set of user groups is specified as input, a simple deterministic algorithm yields a constant-factor approximation guarantee. Further, for the generalized case where any user group subject to a cardinality constraint is permissible, a greedy algorithm yields a constant-factor guarantee over certain practically relevant regimes.