Narrowband-MIMO and MIMO-OFDM systems: Exploiting channel statistics for efficient use of feedback.

摘要

Wireless communication systems employing multiple antennas at both the transmitter and receiver, called multiple-input multiple-output (MIMO) systems, can provide higher data rates and increased reliability compared with single antenna systems. The performance gain can be further enhanced by enabling the transmitter to adapt the transmit signal to the current channel state. Often in practice, the receiver selects a precoding matrix in a precoding codebook, which is agreed a priori by the transmitter and receiver, based on the current channel state. Then, the transmitter obtains the index of the selected precoding matrix and multiplies the corresponding precoding matrix with a source symbol vector before transmission.;Many existing precoding codebook design algorithms and feedback-aided transmission schemes possess limitation such as a lack of consideration of channel statistics and reliance on numerical optimization techniques. These limitations lead to inefficiencies in the use of the scarce feedback resource, or restrict its use in practice. Hence, we develop precoding codebook design algorithms and feedback strategies that are practically feasible and account for channel statistics. In order to improve efficiency, our two approaches are to match a precoding codebook to the channel statistics and to eliminate feedback of redundant information that can be obtained via interpolation or estimation. The major contributions of this dissertation are the development of algorithms for (i) precoding codebook design, (ii) precoding matrix interpolation, and (iii) "overhead free" estimation of time-varying channel statistics.;First, we consider the spatial correlation aspect of narrowband MIMO channels. Focusing on the design of a transmit direction matrix codebook, we develop a completely systematic algorithm for designing a codebook adapted to the spatial channel statistics.;Second, we consider the spectral correlation aspect of wideband MIMO channels. The channel states of adjacent subchannels in the frequency domain can be highly correlated. Hence, the transmitter may use an interpolation technique to reduce the amount of feedback. We develop a new precoding matrix interpolation algorithm that works gracefully with spatial and spectral power allocation.;Finally, we consider the time-varying nature of the channel statistics. We develop an algorithm for adapting a precoding codebook based on the past feedback indices.