Abstract
Large MIMO base stations remain among wireless systems designers’ best tools for increasing wireless through-put while serving many clients. Still, current system designs, sacrifice throughput with simple linear MIMO detection algorithms. Higher performance detection techniques are known, but remain off the table because of their related complexity and latency requirements. In this PhD thesis, novel signal processing approaches are presented, that have the potential to reclaim this wasted MIMO channel capacity while meeting challenging latency requirements by employing parallel processing and efficient tree pruning techniques. The core of this work builds a novel framework for massively parallel signal processing for large MIMO systems applicable to both uplink and downlink. The proposed approaches are asymptoticly optimal, adapt to the processing capabilities of the base station and the current MIMO channel realization, support powerful a posteriori probability (APP) decoding and have latency requirements similar to simple successive interference cancellation. The proposed massively parallel precoder and the detector are validated in over-the-air experiments. In order to make the approaches practical, novel solutions for fast rate adaptation (for both uplink and downlink) are proposed, necessary to translate the improved detection/precoding capabilities into actual throughput. In addition, this thesis introduces the novel principle of Antipodal detection and decoding, that enables the complexity efficient demultiplexing of tens of interfering streams even in the most challenging transmission scenarios (e.g., when the number of transmitted streams equals the number of the base stations antennas). For the first time, the proposed Antipodal detector leverages the unexploited relation between the detection reliability and the detection complexity of sphere decoding to extract implicit reliability information (at no extra processing cost) and to substantially reduce the detection complexity itself. In particular, the detector polarizes its output into highly reliable bits and erasures. While a traditional belief-propagation decoder can handle such an outcome, the Antipodal decoder proposed is tailored to the properties of the Antipodal detector output.