Capacity Distribution for Interference Alignment With CSI Errors and Its Applications

Interference alignment (IA) is known to achieve the degree-of-freedom (DoF) capacity of the interference channel, if full channel state information (CSI) is available at the transmitters perfectly. Challenges, however, arise when CSI is not perfect, and the achievable capacity of IA is not well understood. In this paper, we study the achievable performance of the interference channel using perfect IA techniques based on imperfect CSI. In particular, we obtain the statistical distribution of the maximum achievable rate per stream of the channel. Utilizing our analytical results, we derive new nonasymptotic performance metrics that are then used to 1) optimize the number of streams per user for maximizing the network sum-rate and 2) assess the performance of IA in the time-varying block fading channel. Numerical results are provided to reveal the accuracy of our analytical results.