Interference Alignment: A one-sided approach (1103.5738v1)

Abstract: Interference Alignment (IA) is the process of designing signals in such a way that they cast overlapping shadows at their unintended receivers, while remaining distinguishable at the intended ones. Our goal in this paper is to come up with an algorithm for IA that runs at the transmitters only (and is transparent to the receivers), that doesn't require channel reciprocity, and that alleviates the need to alternate between the forward and reverse network as is the case in Distributed IA (Gomadam, Cadambe, Jafar 08'), thereby inducing significant overhead in certain environments where the channel changes frequently. Most importantly, our effort is focused on ensuring that this one-sided approach does not degrade the performance of the system w.r.t. Distributed IA (since it cannot improve it). As a first step, we model the interference in each receiver's desired signal as a function of the transmitters' beamforming vectors. We then propose a simple steepest descent (SD) algorithm and use it to minimize the interference in each receiver's desired signal space. We mathematically establish equivalences between our approach and the Distributed IA algorithm (Gomadam, Cadambe, Jafar 08') and show that our algorithm also converges to an alignment solution (when the solution is feasible).