Subchannel grouping and statistical waterfilling for vector block-fading channels

In this paper, capacity limits and power-control schemes are assessed for the block-fading (BF) additive white Gaussian noise (AWGN) channels under the assumption that the channel state information (CSI) is known at both the transmitter and the receiver sides. Two notions of channel capacity limits are considered in this paper, including the ergodic capacity for transmission without delay constraint, and the delay-limited capacity for transmission under a strict delay constraint. With the full availability of CSI, a BF channel can be decomposed into multiple space-time-frequency subchannels, and is, therefore, considered in this paper as a vector BF channel. We propose a class of multitarget (MT) power-control schemes, which we show to be computationally implementable and able to perform well in approaching the,capacity limits of vector BF channels. These power-control schemes are based on a subchannel grouping technique, which can be used to further divide subchannels in a vector BF channel into multiple grouped channels, each exhibiting different fading statistics. Each NIT scheme then assigns a particular set of power "targets" into these grouped channels, and maximizes the transmission rate via various statistical waterfilling solutions. Numerical results are presented to evaluate and compare the achievable transmission rates associated with these proposed schemes for some example vector BF channels.