Submitted to IEEE Transactions on Wireless Communications, October 2009.
Interference alignment (IA) is a multiplexing gain optimal transmission strategy for the interference channel with an arbitrary number of users. While the achieved sum rate with IA is much higher than previously thought possible, the improvement comes at the cost of requiring network channel state information at the transmitters. This can be achieved by explicit feedback, a flexible yet costly approach that incurs large overhead and limits throughput. We propose using analog feedback as an alternative to limited feedback or reciprocity based alignment. We show that the full multiplexing gain observed with perfect channel knowledge is preserved by analog feedback and the mean loss in sum rate is bounded by a constant when signal-to-noise ratio is comparable in both forward and feedback channels. When such feedback quality is not quite possible, a fraction of the degrees of freedom is achieved. We consider the overhead of training and feedback and use this framework to optimize the system's effective throughput. We present simulation results to demonstrate the performance of IA with analog feedback, verify our theoretical analysis, and extend our conclusions on optimal training and feedback length.
This preprint is available on arxiv.