| Multiple-input multiple-output (MIMO) can be used to achieve substantial throughput and reliability via spatial multiplexing and diversity. It can also be used to allow multi-user (MU) scenarios, thereby further boosting throughputs. However, the interference caused by multiple users sharing the spatial channel might greatly degrade potential gains promised by MU-MIMO. In this dissertation, we mainly investigate two MU-MIMO systems, namely downlink channels and interference channels, and propose transmission strategies to benefit MU systems by efficiently utilizing available resources.;First, we investigate downlink MU-MIMO channels and consider the impairments from imperfect channel state information (CSI). In MU-MIMO systems, perfect CSI is essential to eliminate the interference; otherwise, residual interference due to CSI errors can severely degrade the performance. Closed-form expressions are derived for achievable rates of such systems with imperfect CSI, caused by quantization and delay. We compare the performance of single-user (SU) and MU transmission and provide guidelines to determine which mode should be activated for a particular environment.;For MIMO links with limited feedback, vector quantization (VQ) is efficient but incurs high complexity. Motivated by the theory of sparse coding, we propose sparse coding quantization (SCQ) to reduce the computational complexity of VQ, by feeding back a linear combination of multiple codewords. Through both analysis and simulation, we show that SCQ can achieve the same sum rate as VQ at a reasonable cost in feedback overhead. Furthermore, compared with scalar quantization (SQ), SCQ provides higher quantization effectiveness. By evaluating these techniques, we show that SCQ is a good option for balancing performance, overhead and complexity. For MIMO links with delayed feedback, the relative performance of MU and SU is also affected by the system configurations. Despite the vulnerability of MU to the delay in time-varying channels, MU can retain its theoretical advantage over SU in scenarios with alternative channel delay models. Moreover, the degradation due to delayed CSI can be significantly mitigated by utilizing high-order prediction at the transmitter.;Besides sum rate or spectral efficiency (SE) performance, energy efficiency (EE) is another critical metric. We consider joint power allocation and antenna selection to maximize the EE of MU and SU systems. Extending the SU-MU comparison to the EE performance, we illustrate that SU-MIMO is desirable when the transmit power is low, while MU-MIMO is favored in the case of high transmit power.;Another realistic feature of MU-MIMO systems is receive antenna heterogeneity, meaning that users are equipped with different numbers of antennas. To exploit this reality, a user selection algorithm is proposed for downlink systems. The scheduling algorithm maximizes the system throughput using user grouping and spatial stream allocation. The proposed algorithm outperforms the classic homogeneous approach, where the heterogeneity is not taken into account.;In addition, we consider MU-MIMO interference channels and discuss interference management approaches for these channels. Existing algorithms, such as the minimum mean-square error (MMSE) and the minimum leakage interference alignment (ML-IA), achieve performance gains after several iterations, which incur a large amount of time overhead. As an extension of ML-IA, the algorithm we propose, maximizing of the signal with interference alignment (MS-IA), combines one-shot interference cancellation and signal strength maximization. By applying, in addition, a novel, partitioning strategy, we show that MS-IA surpasses MMSE, in one-shot performance.;Finally, we explore the benefit of MIMO to orthogonal frequency division multiplexing (OFDM) systems in a large delay spread environment, where inter-channel interference (ICI) and inter-symbol interference (ISI) may exist and degrade the performance. An ICI/ISI-aware beamforming algorithm is proposed to mitigate these interferences as well as enhance desired signals. We show this technique can significantly eliminate the ICI and ISI, achieving beamforming gains for OFDM systems. |
