Achievable rate improvement through channel prediction for interference alignment

Interference alignment (IA) is a promising interference management technique to efficiently eliminate multiuser interference in a K-user interference network. However, channel state information (CSI) at the transmitter is indispensable for the interference alignment precoding matrices design. In spite of the fact that the CSI can be obtained through feedback from receivers to transmitters in a frequency-division duplex (FDD) system or through the reciprocity in a time-division duplex (TDD) system, it may be imperfect for the reason such as estimation inaccuracy, feedback delay and time-varying of channel. In this paper, the impact of imperfect CSI on the achievable sum rate of interference alignment network is considered and a channel prediction technique based on Kalman filtering is proposed to verify the theoretical analysis. Through analysing, we find that the sum rate performance of an interference aligned network is affected by an integrated parameter, i.e., the product of user number, transmit power and channel error variance. Simulation results reveal that the performance of interference alignment is more sensitive to the uncertainty of CSI at high signal-to-noise ratio (SNR) regime than that at low SNR regime. It is also verified that the performance can be improved through channel prediction, comparing with interference alignment based on the delayed feedback CSI.