Optimization of Integer-Forcing Precoding for Multi-User MIMO Downlink
share
Integer-forcing (IF) precoding is an alternative to linear precoding for multi-user (MU) multiple-input-multiple-output (MIMO) channels. IF precoding can be viewed as a generalization of lattice-reduction-aided (LRA) precoding where symbol-level detection is replaced by codeword-level decoding, allowing achievable rate expressions to be derived explicitly. A particular form of IF precoding that is asymptotically optimal for high SNR is the regularized diagonally-scaled IF (RDIF) scheme, which is specified by a diagonal scaling matrix D and an integer-valued effective channel matrix A. Optimal values for these parameters have been found (analytically) only for the case of K = 2 users. In this paper, a low-complexity method is proposed to find good parameters for the RDIF scheme for K > 2 users. The proposed method computes matrix D by solving a relaxed optimization problem and then matrix A by the application of the LLL algorithm—which, in this particular case, is shown to converge with a fixed number of iterations, resulting in an overall complexity of O(K^3). Simulation results show that the proposed method achieves 1a higher sum rate than choosing D = I (as in LRA precoding) and significantly outperforms traditional linear precoding in all simulated scenarios1
READ FULL TEXT
