Performance Analysis of IA Techniques in the MIMO IBC With Imperfect CSI

In this work we consider the multiple-input multiple-output (MIMO) interference broadcast channel (IBC) and analyse the performance of interference alignment (IA) under imperfect channel state information (CSI), where the variance of the CSI error depends on the signal-to-noise ratio (SNR). First, we derive an upper bound on asymptotic mean loss in sum rate compared to the perfect CSI case and then we quantify the achievable degrees of freedom (DoF) with imperfect CSI. Both sum rate loss and achievable DoF are found to be dependent on the number of cells in the system and the amount of users per cell, in addition to the CSI error parameters themselves. Results show that when error variance is inversely proportional to SNR, full DoF are achievable and the asymptotic sum rate loss is bounded by a derived value. Additionally if the CSI imperfection does not disappear for asymptotically high SNR, then the full DoF gain promised by IA cannot be achieved; we quantify this loss in relation to the CSI mismatch itself. The analytically derived bounds are validated via system simulation, with the cellular counterparts of the maximum signal-to-interference-plus-noise ratio (Max-SINR) and the minimum weighted leakage interference (Min-WLI) algorithms being the IA techniques of choice. Secondly, inspired by the CSI mismatch model used to derive the bounds, we present a novel Max-SINR algorithm with stochastic CSI error knowledge (Max-SINR-SCEK) for the MIMO IBC. Simulations show that the proposed algorithm improves performance over the standard one under imperfect CSI conditions, without any additional computational costs.