MIMO communications systems provide much greater capacity potential than their SISO counterpart. However, to fully unleash their capacity potential, careful design of transceiver architecture is necessary. Most traditional MIMO precoding techniques focus on single transmission sessions. In this dissertation, we integrate the MIMO precoder design with hybrid ARQ for performance improvement and error control. In addition, it is unrealistic to assume perfect channel state information (CSI) at the transmitter side for precoding. Hence, we focus our precoder design on scenarios where only partial CSI is available at the transmitter. This dissertation investigates the MIMO precoding design for H-ARQ system with only partial channel state information.
First, channel is assumed to be relative stationary, and the mean of the MIMO channel matrix can be estimated at the receiver and sent back to the transmitter for precoder design for subsequent transmissions. When a particular data packet fails, the transmitter would transmit the data packet again by using a new precoder designed by the imperfect current CSI information. The receiver should utilize both previously received signals and newly arrived signal from the retransmission for joint detection of the data packet. Our precoder aims at maximizing ergodic capacity, and demonstrates significant performance gain over conventional precoding schemes.
Next, we investigate precoder design based on another channel statistics, namely the covariance matrix. Due to the topology setting of the transmitting/receiving antennas, MIMO channels sometime exhibit stationary correlation which is irrelevant to any particular scattering. We formulate the precoder design problem as a convex programming problem, and utilize KKT conditions to find an optimal power allocation.
Although channel statistics form one type of partial information, a different kind of partial CSI is also available for time varying channels, when only a finite number of bits are sent as feedback from receiver to update the CSI. This model is more appropriate when the communication devices or major reflecting objects around them are moving relatively fast, and, thus, the channel exhibit strong non-stationary characteristics. Another motivation for such modeling is the fact that feedback channels are usually of limited capacity, and the overhead controlling information transmitted should be as low as possible. We show that for the MIMO-ARQ systems, an effective precoder may consist of two parts: a conventionally designed precoder which neglects ARQ mechanism, and a carefully chosen permutation matrix, whose main task is to assign data symbols to different subchannels. We propose an efficient method of finding the optimal permutation matrix, and show the conditions under which the permutation matrix selection method is optimal. We also derive an explicit analytical formula for the BER/outage performance for 2 × 2 system.
Most existing MIMO precoders essentially transmit signals along the directions of the right singular vectors of the channel matrix. However, their BER performance is rarely investigated. We further develop a generic approach to determine the diversity order for various MIMO precoder schemes via a concise analytical formula. The results obtained can be viewed as an extension of the fundamental tradeoff between diversity and multiplexing obtained by Dr. Zheng and Dr. Tse.
Finally, we design an MIMO-ARQ transceiver scheme that aims to maximize the system throughput. By retransmitting only partial information that is not well received in the previous transmission(s), we may significantly boost the system through-put. The scheme is based on only finite feedback bits and does not require any transmitter side channel information. It is practical even for highly mobile wireless communications.
