A Greedy Algorithm of Data-Dependent User Selection for Fast Fading Gaussian Vector Broadcast Channels

Abstract: User selection (US) with Zero-forcing beamforming is considered in fast fading Gaussian vector broadcast channels with perfect channel state information (CSI) at the transmitter. A novel criterion for US is proposed, which depends on both CSI and the data symbols, while conventional criteria only depend on CSI. Since the optimization of US based on the proposed criterion is infeasible, a greedy algorithm of data-dependent US is proposed to perform the optimization approximately. An overhead issue arises in fast fading channels: On every update of US, the transmitter might inform each user whether he/she has been selected, using a certain fraction of resources. This overhead results in a significant rate loss for fast fading channels. In order to circumvent this overhead issue, iterative detection and decoding schemes are proposed on the basis of belief propagation. The proposed iterative schemes require no information about whether each user has been selected. The proposed US scheme is compared to a data-independent US scheme. The complexity of the two schemes is comparable to each other for fast fading channels. Numerical simulations show that the proposed scheme can outperform the data-independent scheme for fast fading channels in terms of energy efficiency, bit error rate, and achievable sum rate.