Due to its significant advantage in spectral efficiency, multiple-antenna communication technology will undoubtedly be a major component in future wireless system implementations. However, the full exploitation of this technology also requires perfect feedback of channel state information (CSI) to the transmitter-- something that is not practically feasible. This motivates the study of limited feedback systems, where CSI feedback is rate limited. This thesis focuses on the optimal design of limited feedback systems for three types of communication channels: the relay channel, the single-user point-to-point channel, and the multiuser broadcast channel. For the relay channel, we prove the efficiency of the Grassmannian codebooks as the source and relay beamforming codebooks, and propose a method for CSI exchange between the relay and the destination when global CSI is not available at destination. For the single-user point-to-point channel, we study the joint power control and beamforming problem and address the channel magnitude and direction quantization codebook design problem. It is shown that uniform quantization of the channel magnitude (in dB scale) is asymptotically optimal regardless of the channel distribution. The analysis further derives the optimal split of feedback bandwidth between the magnitude and direction quantization codebooks. For the multiuser broadcast channel, we first prove the sufficiency of a product magnitude-direction quantization codebook for managing the multiuser interference. We then derive the optimal split of feedback bandwidth across the users and their magnitude and direction codebooks. The optimization results reveal an inherent structural difference between the single-user and multiuser quantization codebooks: a multiuser codebook should have a finer direction quantization resolution as compared to a single-user codebook. It is further shown that the users expecting higher rates and requiring more reliable communication should provide a finer quantization of their CSI. Finally, we determine the minimum required total feedback rate based on users' quality-of-service constraints and derive the scaling of the system performance with the total feedback rate.

Due to its significant advantage in spectral efficiency, multiple-antenna communication technology will undoubtedly be a major component in future wireless system implementations. However, the full exploitation of this technology also requires perfect feedback of channel state information (CSI) to the transmitter-- something that is not practically feasible. This motivates the study of limited feedback systems, where CSI feedback is rate limited. This thesis focuses on the optimal design of limited feedback systems for three types of communication channels: the relay channel, the single-user point-to-point channel, and the multiuser broadcast channel. For the relay channel, we prove the efficiency of the Grassmannian codebooks as the source and relay beamforming codebooks, and propose a method for CSI exchange between the relay and the destination when global CSI is not available at destination. For the single-user point-to-point channel, we study the joint power control and beamforming problem and address the channel magnitude and direction quantization codebook design problem. It is shown that uniform quantization of the channel magnitude (in dB scale) is asymptotically optimal regardless of the channel distribution. The analysis further derives the optimal split of feedback bandwidth between the magnitude and direction quantization codebooks. For the multiuser broadcast channel, we first prove the sufficiency of a product magnitude-direction quantization codebook for managing the multiuser interference. We then derive the optimal split of feedback bandwidth across the users and their magnitude and direction codebooks. The optimization results reveal an inherent structural difference between the single-user and multiuser quantization codebooks: a multiuser codebook should have a finer direction quantization resolution as compared to a single-user codebook. It is further shown that the users expecting higher rates and requiring more reliable communication should provide a finer quantization of their CSI. Finally, we determine the minimum required total feedback rate based on users' quality-of-service constraints and derive the scaling of the system performance with the total feedback rate.