Signal processing methods for large-scale multi-antenna systems

Abstract

The data traffic demand in wireless communication systems has been growing at a fast pace with the widespread use of cellular systems and the emergence of the Internet of Things. To meet the large traffic requirements, the fifth-generation (5G) of cellular technology envisions transceiver systems operating at the millimeter wave spectrum with large-scale antenna arrays. However, this 5G system design faces many engineering challenges. Signal processing methods and architectures employed in classic multi-antenna systems are inadequate in the large-scale scenario. Standard signal processing techniques become computationally expensive and classical radio-frequency front-end architectures exhibit low energy efficiency. This thesis presents lowcomplexity and energy-efficient solutions to the design of large-scale multi-antenna systems. First, we propose multilinear filters to tackle the complexity issue in large-scale receive processing. We show that the proposed multilinear filtering methods drastically reduces the computational complexity with a slight performance deterioration compared to the classical linear approach. Concerning the energy efficiency of transceiver architectures, we investigate hybrid analog/digital (A/D) massive multiple-input multiple-output (MIMO) systems with low-resolution data converters. We present efficient precoding schemes for hybrid A/D systems with fully- and partially-connected phase-shifting networks. We also introduce low-complexity double-sided massive MIMO transceiver schemes, where both the base station and the user equipment employ large-scale antenna arrays. In particular, we leverage the multi-layer filtering strategy to reduce the computational complexity and channel state information requirements of the transceiver design. Finally, we consider the problem of channel estimation under synchronization impairments. More specifically, we develop tensor-based algorithms for channel estimation in the presence of carrier frequency offset (CFO) and phase noise (PN). We start with the frequency-flat case by assuming CFO-corrupted measurements. Then, we turn our attention to the frequency-selective case by including both CFO and PN.