User scheduling based on angles with block diagonalization in MU‐MIMO broadcast

Summary This paper investigates user selection scheme in the multiuser multiple-input multiple-output (MU-MIMO) broadcast (BC) scene with block diagonalization precoding. Block diagonalization is a suboptimal but practical linear precoding method, which can eliminate the multiuser interference by turning the MU-MIMO BC channel into parallel MIMO channels. With this precoding method, we propose the best user from the user subset to maximize the total throughput in the MU-MIMO BC system. The angles between subspaces used in this paper are induced from n-inner product, an extension from norm space to the n-dimensional space, which characterizes the orthogonality between subspaces. One of the algorithms achieves good performance by comparing the capacity greedily, the other one attains high capacity by reducing the cardinality of the user subset to improve the orthogonality between the user channels, which could be seen as a complexity reduction algorithm with respect to the former one. Indeed, they are all based on the angles between subspaces. Analysis shows that both of the proposed algorithms have lower complexity and better performance than the classical algorithms. The numerical results also confirm our analysis.