Low-Complexity MU-MIMO Nonlinear Precoding Using Degree-2 Sparse Vector Perturbation
Yi Ma,Abderraouf Yamani,Na Yi,Rahim Tafazolli
DOI: https://doi.org/10.1109/jsac.2016.2525378
IF: 16.4
2016-01-01
IEEE Journal on Selected Areas in Communications
Abstract:Multiuser multiple-input multiple-output (MUMIMO) nonlinear precoding techniques face the problem of poor computational scalability to the size of the network. In this paper, the fundamental problem of MU-MIMO scalability is tackled through a novel signal-processing approach, which is called degree-2 vector perturbation (D2VP). Unlike the conventional VP approaches that aim at minimizing the transmit-to-receive energy ratio through searching over an N-dimensional Euclidean space, D2VP shares the same target through an iterative-optimization procedure. Each iteration performs vector perturbation over two optimally selected subspaces. By this means, the computational complexity is managed to be in the cubic order of the size of MU-MIMO, and this mainly comes from the inverse of the channel matrix. In terms of the performance, it is shown that D2VP offers comparable bit-error-rate to the sphere encoding approach for the case of small MU-MIMO. For the case of medium and large MU-MIMO when the sphere encoding does not apply due to unimplementable complexity, D2VP outperforms the lattice-reduction VP by around 5-10 dB in Eb/No and 10-50 dB in normalized computational complexity.