Iterative water-filling algorithms for power efficiency optimization in the MIMO downlink

In this paper we present algorithms to minimize transmit power consumption in the multi-antenna downlink under quality of service (QoS) constraints in terms of (sum) rate. This fundamental problem arises frequently in dynamic resource allocation. It occurs, for example, if a certain throughput has to be guaranteed due to stability issues or QoS reasons while interference to adjacent cells should be as small as possible. This perspective is in contrast to the traditional viewpoint of weighted rate sum maximization or throughput maximization, where the power budget is fixed but no rate can be guaranteed. Our algorithms exploit the specific structure of an equivalent optimization problem and are based on iterative water-filling techniques. They are very fast even for demanding scenarios (such as 100 users with multiple antennas each) and provably converge to the global optimum.