Transmission resource allocation in multi-antenna wireless communication systems with channel uncertainty

Abstract

In this thesis we investigate the design of transmission resource allocation in current and future wireless communication systems. We focus on systems with multiple antennas and characterize their performance from an information-theoretic viewpoint. The goal of this work is to provide practical transmission and resource allocation strategies taking into account imperfections in estimating the wireless channel, as well as the broadcast nature of the wireless channel. In the first part of the thesis, we consider training-based transmission schemes in which pilot symbols are inserted into data blocks to facilitate channel estimation. We consider one-way training-based systems with and without feedback, as well as two-way training-based systems. Two-way training enables both the transmitter and the receiver to obtain the channel state information (CSI) through reverse training and forward training, respectively. In all considered cases, we derive efficient strategies for transmit time and/or energy allocation among the pilot and data symbols. These strategies usually have analytical closed-form expressions and can achieve near optimal capacity performance evidenced by extensive numerical analysis. In one-way training-based systems without feedback, we consider both spatially independent and correlated channels. For spatially independent channels, we provide analytical bounds on the optimal training length and study the optimal antenna configuration that maximizes an ergodic capacity lower bound. For spatially correlated channels, we provide simple pilot and data transmission strategies that are robust under least-favorable channel correlation conditions. In one-way training-based systems with feedback, we study channel gain feedback (CGF), channel covariance feedback (CCF) and hybrid feedback. For spatially independent channels with CGF, we show that the solutions to the optimal training length and energy coincide with those for systems without feedback. For spatially correlated channels with CCF, we propose a simple transmission scheme, taking into account the fact that the optimal training length is at most as large as the number of transmit antennas. We then provide a solution to the optimal energy