Partial CSI feedback design for interference alignment in MIMO cellular networks

Interference alignment (IA) can achieve the optimal capacity scaling with respect to SNR but most existing IA designs require full channel state information (CSI) at the transmitters. In this paper, we consider IA processing with partial CSI feedback in MIMO cellular networks and we use the feedback dimension to quantify the first order CSI feedback cost. Conventional IA cannot be used because only partial CSI knowledge can be used to design the IA pre-coders. Therefore, we establish a new set of feasibility conditions for IA under the proposed partial CSI feedback scheme. Based on these results, we formulate the problem of CSI feedback dimension minimization subject to the constraints of IA feasibility. We further propose an asymptotically optimal solution and derive closed-form trade-off results between the CSI feedback cost and IA performance in MIMO cellular networks.