Télécom Paris (enst) Institut Eurécom These Rubén De Francisco Martín Optimisation De La Performance Des Systèmes Mimo Avec Connaissance Partielle Du Canal Specialization: Communication and Electronics Performance Optimization of Mimo Systems with Partial Channel State Information

Multiple-input multiple-output (MIMO) wireless communication systems have the potential to offer high data rates as well as link reliability. Over the past years, MIMO systems have been regarded as firm candidates for future mobile communication standards, both in single user and multiuser modes of operation. The feasibility of these systems in future standards depends in great measure on the ability to provide high rates with a reduced amount of channel state information at the transmitter (CSIT), due to limited resource availability on the feedback link. This thesis addresses the problem of optimizing MIMO wireless communication systems with partial CSIT. On the one hand, we focus on how to design methods for obtaining CSIT, identifying what types of CSIT are relevant. On the other hand, we propose techniques to exploit the available sources of CSIT to optimize the system performance. In the first part, point-to-point MIMO channels are considered for the purpose of error rate minimization. Linear precoding techniques are proposed to enhance the performance of space-time coded (STC) MIMO systems, based on statistical information on the MIMO channel. As we show, the performance of such systems can be significantly improved by appropriately combining mean and covariance information. In the second part of this thesis, we focus on sum-rate performance optimization in MIMO broadcast channels with limited feedback. Lowcomplexity cross-layer approaches are proposed for systems with joint linear beamforming and multiuser scheduling, optimizing the following parts in the MIMO communications system: linear beamforming techniques, scheduling algorithms, feedback strategies and feedback quantization techniques. Different feedback models are considered. In a simple scenario where feedback consists of only channel quantization indices, we study the benefits of generating quantization codebooks adapted to the channel statistics, exploiting spatial and temporal correlations. A different feedback model con-