Statistical CSI Aided MIMO-NOMA Transmission in Correlated Rician Fading Channels

In this article, we investigate the ergodic sum-rate maximization problem for a downlink multiple-input multiple-output non-orthogonal multiple-access system with correlated Rician fading channels. Specifically, by exploiting the statistical channel state information and using large-system approximation theory, the ergodic sum-rate is approximated by a deterministic expression. Then the optimization problem of designing transmit covariances for the strong and weak user is formulated under the constraints of total transmit power budget for base station and minimum rate requirement for the weak user. Since the variables to be optimized are highly correlated with each other, we resort to the alternate optimization method by breaking the problem into two subproblems. For the weak user, the optimal covariance can be acquired by water-filling strategy, while the subproblem for strong user is non-convex. To overcome this, we further adopt the Taylor series approximation on the objective function and the reformulated problem can be efficiently solved with generalized singular value decomposition. After acquiring the covariances, the power allocation is also optimized by bisection search. Finally, the numerical simulation validates the accuracy of derived approximate expression and also the effectiveness of proposed transmission scheme, which significantly outperforms other schemes.