MIMO Transceiver Design for Multi-Antenna Communications over Fading Channels

In wireless communications, the use of multiple antennas for both transmission and reception is associated with performance gains of fundamental nature. One such gain stems from the spatial-multiplexing capabilities of wireless multiple-input multiple-output (MIMO) channels: Many propagation environments admit several data streams to be conveyed in parallel over a single point-to-point link, setting the stage for significantly increased data rates. The optimal maximum likelihood (ML) receiver pays a high price for spatial multiplexing in the form of a heavy computational burden. Another receiver candidate is the decision feedback (DF) equalizer, reaping the benefits of spatial multiplexing with far more efficient receive processing. Its combination of independent data-stream decoding with successive interference cancellation (SIC) is sufficient to achieve several information-theoretic performance limits, both for point-to-point and multiple-access channels. Nevertheless, in practical systems there is often a clear performance degradation associated with DF processing compared to ML. In this context, linear precoding is a transmit pre-processing technique that can enhance performance by adapting the transmission to available channel-state information (CSI). This thesis deals with MIMO transceiver design for DF equalization: the joint optimization of transmit precoding and receive equalization. A crucial aspect is the assumptions made on the kind of CSI available. The thesis considers a practical case with long-term, statistical CSI at the transmitter and perfect short-term CSI at the receiver. The thesis presents an optimization framework for MIMO transceiver design, contributing in several respects. Firstly, a number of relevant performance measures are presented, and novel expressions are provided for spatially correlated MIMO channels. Secondly, it is shown that optimization with respect to such performance measures can be cast as convex optimization problems, enabling efficient solutions in general. The thesis also provides an in-depth analysis of specific optimization problems, enabling very efficient solutions; novel connections are established between MIMO transceiver design, convex-hull algorithms, and submodular optimization. Thirdly, an extension to the multi-user uplink is provided. The thesis considers not only the joint optimization of the users’ precoders, but also joint optimization with the decoding order employed at the receiver. The thesis shows how to address the hard design problem in a computationally efficient manner using alternating optimization between precoders and the decoding order, which is observed to converge fast with close to optimal performance.QC 20130327