Linear Interference Alignment in Full-Duplex MIMO Networks With Imperfect CSI

In this paper, we consider a system where full-duplex (FD) base-stations (BSs) communicate with half-duplex (HD) downlink (DL) and uplink (UL) users in a multiuser multi-cell network, where all nodes are equipped with multiple antennas. The introduction of FD BSs offers potential to increase spectral efficiency, however, it also causes a surge in the number of interference links compared with the HD network counterpart. Here, we apply linear interference alignment (IA) to manage interference in this network under imperfect channel state information (CSI). First, we characterize the performance losses incurred with respect to the achievable sum rate and degrees of freedom (DoFs). Results show that the general trend in performance loss is mainly determined by how the error scales with the signal-to-noise ratio (SNR). In particular, full UL and DL DoF can be achieved even under imperfect CSI when the channel error is at least inversely proportional to SNR. Moreover, in such cases the sum rate loss is always finite, and either goes to zero or is upper bounded by a derived value. Second, we design two linear IA algorithms applicable to the system under consideration. These are based on minimizing the mean square error (MMSE) and maximizing the signal-to-interference-plus-noise ratio, and consider statistical knowledge of the CSI error for added robustness. The proposed algorithms follow specific design principles that distribute the different interference components amongst the various beamformers and result in unitary receivers and precoders. In addition, we show that under certain conditions both designs result in identical beamforming solutions, even though the MMSE algorithm has lower computational complexity. Finally, we also derive the proper condition for IA feasibility in the multi-cell system under consideration.