Interference alignment with cyclic unidirectional cooperation

Recent results on interference alignment (IA) in a K-user n × n MIMO interference channel, have shown that the ratio of total degrees of freedom d̂ to the single user degrees of freedom varies from d̂/n <; 2 in the case of arbitrary MIMO channels (i.e., channels with no structure) to K/2 in the case of channels with diagonal structure (i.e., channels with time or frequency varying coefficients). These results confirm that the information theoretic upper bound on degrees of freedom with no channel extension can only be achieved for K ≤ 3 with an overwhelming overhead of the global channel knowledge or an extensive number of back-and-forth iterations between the transmitting and the receiving nodes [3]. In this paper, we show that this limit on the number of users can be extended to K ≤ 5 as a result of introducing a partial cooperation into the interference channel. This partial cooperation is such that the message of the kth transmitter or base station (BS) is known to the [k - 1)th transmitter (mod K) and each receiver or user is simultaneously served by two transmitters. We introduce a one-shot algorithm that utilizes the reciprocity of wireless network to achieve interference alignment with no iterations between the transmitting and the receiving nodes when d̂/n ≤ 2. Monte-Carlo simulation results show that the proposed partial cooperation used with the one-shot algorithm can considerably enhance the sum rate in an interference channel system at practical SNR levels as a result of improved spatial multiplexing gain.