An asymptotically sum-rate optimal precoding scheme for MIMO Gaussian broadcast channel

In this paper, we study the downlink precoding schemes for MIMO Gaussian broadcast channels (MIMO GBC). A novel low-complexity zero-forcing dirty-paper-coding (DPC) scheme, named as Successive ZF-DPC, and a multiuser scheduling algorithm are proposed to asymptotically approach the sum capacity of MIMO GBC. After the ordered subset of active users determined by the scheduling algorithm, the Successive ZF-DPC successively puts the precoding matrix of each user in the null space of the composite matrix stacked by channel matrices of preceding users in the subset such that each user is not subject to the interference caused by the following users in the subset. Then, DPC can be used to eliminate the interference caused by the preceding users, which ultimately nullifies the mutual interference among users. Our analysis and simulations show that the Successive ZF-DPC is asymptotically sum-rate optimal in both spatially uncorrelated and correlated channels even if each user has multiple antennas. Furthermore, it is shown that the multiuser scheduling algorithm can make efficient use of the multiuser diversity to increase the achievable maximal sum rate and accelerate the approach to the sum capacity of the MIMO GBC.