Low Complexity Eigenmode Selection for MIMO Broadcast Systems with Block Diagonalization

We investigate new scheduling algorithms for MIMO broadcast systems to maximize the achievable sum rate of block diagonalization transceivers. In contrast to conventional schemes which select users directly, we investigate a more efficient type of scheduling algorithms, named as multiuser eigenmode transmission (MET). It is based on the selection of eigenmodes of all users. Utilizing an iterative procedure to calculate the Moore-Penrose generalized inverse of a matrix consisting of the already selected eigenmodes, we propose a low complexity multiuser eigenmode selection method based on the greedy algorithm. Then we show that the multiuser dominant eigenmode transmission can achieve the same sum rate scaling law as that of DPC. Therefore, to further reduce the complexity, we restrict our search space to the set of dominant eigenmodes and derive an alternative eigenmode selection algorithm with lower complexity at the cost of little performance loss, which we name as the low-complexity multiuser dominant eigenmode selection algorithm.