Abstract
In this dissertation, two independent researches are presented. In the first one backgrounded in the single-user multiple-input-multiple-output (SU-MIMO) systems, we explore a low complexity MIMO precoder design based on the LDLH channel decomposition. While the second one backgrounded in the multiuser MIMO (MU-MIMO) systems, we explore the methods to improve the convergence property and the sum rate performance for an cooperative interference alignment (CIA) scheme called interference leakage minimization algorithm (ILMA). In SU-MIMO systems, transmit precoding is a common technique for canceling the intra-interference between two wireless terminals. Classical MIMO precoding systems are based on the singular value decomposition (SVD) to decouple the channel into parallel eigen modes. However, in addition to the high complexity, this way is also sensitive to the ill-conditioning of the channel. In this study, we suggest a channel decomposition method called LDLH decomposition for low complexity MIMO precoder design. We show that the LDLH decomposition yields better channel matrix condition number and hence better performance than that of SVD. We also show that the complexity of LDLH achieves one degree of magnitude lower than that of SVD-based schemes. In MU-MIMO systems, mutual interference has long been the most crucial factor affects the system performance. The cooperative interference alignment (CIA) is a promising method to effectively achieve interference mitigation. This study investigates methods of improvements for the interference leakage minimization algorithm (ILMA) which is recognized more feasible to real applications than other CIA schemes. Particularly, we look in depth at the effectiveness of initial conditions and the convergence property of the ILMA. For the former, we develop initial conditions of the ILMA from transmitter and receiver perspectives, respectively. For the latter, we devise a constraint relaxation formulation of the ILMA, named the rILMA. Mathematical proofs are given to show that the rILMA is convex and converges to a global minimum. Simulations verify that the proposed methods are effective.