Multiuser precoding in wireless communication systems: parameter and resource optimization via large system analysis

M ULTI-user Multi-Input Multi-Output (MU-MIMO) technologies have become an important feature of the physical layer in modern wireless communication systems, such as in wireless LAN 802.11 and 4G networks including LTE-Advanced and mobile-WiMax, that need throughput in the order of hundreds of megabits per second or more. MU-MIMO has become an important research area, particularly for downlink or broadcast transmissions. In that scenario, one transmitter or base station (BS) with multiple antennas sends different and independent messages to each user simultaneously. The resulting multi-user interference (MUI) can limit the maximum achievable sum rate. Designing a good precoder for the users’ data symbols at the transmitter can reduceMUI. In this thesis, we focus on finding the optimal design parameters that maximize the system performance, i.e., SINR or sum rate, in multiple-antenna broadcast channels (BC) employing a particular precoding strategy, called Regularization Channel Inversion. We investigate the system performance maximization under various scenarios. In single-cell BC, different forms of channel state information (CSI) at the transmitter such as perfect CSI in independent and identically distributed (i.i.d.) and spatially correlated as well as partial CSI with uncertainties are considered. We present the optimal strategy for the regularization parameter of the RCI precoder to adapt in those scenarios. For some cases we also study how we choose the cell-loading, defined as the ratio between the number of users and the number of transmit antennas, to maximize the system performance. Considering clustered or grouped users where each group has different path-losses, we explore the optimal power allocation across the groups that maximizes the sum rate per antenna. In two-cell BC, we investigate CSI feedback optimization for different levels of base station cooperation. iii The analysis in this work is conducted in the large system limit where the number of single-antenna users and transmit antennas tend to infinity with their ratio being fixed. The mathematical tools are based on spectral theory of random matrices that explore the eigenvalues of large dimensional random matrices. Even though the analytical results are in the asymptotic regime, we show their validity for the finite size system design through numerical simulations.