Clustering for Interference Alignment in Multiuser Interference Network

Abstract

Interference alignment (IA) has been shown to be a promising technique for achieving the optimal capacity scaling of a multiuser interference channel at asymptotically high-signal-to-noise ratio (SNR). However, in practical communication systems, mitigating interference from all interferers via IA is not necessary since some users' interference have negligible effect due to large path-loss. Moreover, the feasibility constraint and the heavy signaling overhead hinder applying IA on interference from all interferers. Clustered IA puts users in disjoint clusters where IA is applied to users within each cluster. It provides a mechanism for mitigating the signaling overhead and maximizing the achievable rate. However, how to properly form IA clusters has not been well studied. We consider the application of clustered IA in a multiuser interference network with asymmetric channel attenuation at finite SNR. We model the interference network as a connected graph, transforming the clustering problem into a graph partitioning problem. By exploiting the variation on the interference levels from multiple interferers, efficient clustering algorithms are proposed such that clusters formed can capture strong interference as intracluster interference, leaving relatively weak interference as intercluster interference. Then, the intercluster interference can be coarsely modeled as noise. We also consider the precoder/equalizer design in a clustered system and show the importance of incorporating the aggregated intercluster interference in the design. Simulation results show that proper clustering combined with generalized IA precoder/equalizer design leads to significant gains on the achievable sum rate.