Wireless Communication over Time-Varying Channels With Limited Feedback

The number of deployed wireless communication systems has grown rapidly in the last years. Their popularity is mainly due to the effortlessness with which the systems can be deployed. Further, the new generation of wireless systems, e.g., 802.11n, starts to close the performance gap to their wired counterparts. The performance of these systems can be improved if channel state information (CSI) is available at the transmitter. Unfortunately, having perfect CSI at the transmitter all the time is difficult to realize. CSI can be acquired through training, but becomes outdated very quickly afterwards due to the inevitable time-variability of the wireless channel. However, if the coherence time of the channel is sufficiently large, then the CSI becomes time correlated. One of the main topics of this thesis is to exploit this time correlation to improve the performance of the wireless system. Another problem is that the CSI can often just be acquired at the receiver and must be fed back to the transmitter over a data-rate limited feedback link. Thus, the CSI has to be quantized before it can be fed back. The application of tools from the large vector quantization framework to CSI quantization is the other main topic of the thesis. Throughout the thesis we investigate different scenarios where CSI knowledge at the transmitter plays a paramount role. In the first part of the thesis we consider a single-link MIMO system. We start by investigating how the unitary precoder can be extrapolated based on the knowledge of previous precoders using geodesic interpolation. Next, we apply tools from vector quantization to quantize the precoders by exploiting the time correlation of the channel. In the next part we consider the MISO broadcast channel. There, we propose the use of predictive vector quantization to exploit the time correlation to improve the CSI quantization. Further, we introduce a novel channel model, and generate the corresponding CSI quantization codebooks. We investigate in the last part the feedback of different metrics to enable round-robin scheduling, and near round-robin scheduling, assuming a limited feedback link. We present different metrics, and also propose an algorithm to design the corresponding codebooks