Correlation Matching Pursuit for MIMO Vector Perturbation

In this paper, a correlation matching pursuit (CMP) procedure is proposed to handle the vector perturbation (VP) problem for nonlinear precoding (NLP) in downlink multiuser multi-antenna (i.e., MU-MIMO) systems. Basically, CMP consists of two sub-procedures, i.e., correlation matching and correlation pursuit. The former takes the charge of direct calculation of the perturbation integers through a set of established convex optimization problems; the latter is responsible for selecting the perturbation vector for updating the precoded vector. Iterative execution of both procedures renders CMP being modelled as a searching tree which consists of multiple nodes and paths. The sequence correlation between the precoded vector and perturbation vectors is revealed crucial to the performance optimality and thus used as metric of the path search. Given that the single-path search is suboptimal, we propose multi-path schemes to enable exploitation on the path diversity and thus further improve the performance. Complexity analysis and computer simulations demonstrate that CMP-based NLP algorithms serve as low-cost VP solutions with significantly lowered processing latency and meanwhile, comparable performance compared to prior arts.