Complexity reduction of pico cell clustering for interference alignment in heterogeneous networks

Interference Alignment (IA) in heterogeneous networks (HetNets) is a promising technique that improves the spectral efficiency significantly. We showed in [1] that transmit antennas at pico BSs could be utilized more efficiently by clustering pico cells in IA in HetNet where the clustering formation was optimized so as to minimize the rate loss caused by inter-cluster interference. In [1], the optimum clustering formation was selected by comparing all possible formations, that is, the value of the objective function for all possible formations was calculated. Therefore, we required the enormous complexity to construct pico cell clusters. In this paper, we propose a novel algorithm for clustering pico cells that reduces the complexity of the clustering process. In particular, we define rate-loss matrix that represents the rate loss caused by inter-pico interference, and translate the optimization problem to the construction of rate-loss matrix. Clearly, the proposed algorithm is sub-optimum in terms of achievable rate compared to all-search algorithm used in [1]. However, the simulation results show that the difference of achievable rate between the proposed algorithm and all-search algorithm is negligible and becomes smaller as the cluster size decreases. We evaluate the complexity of proposed algorithm quantitatively comparing to all-search algorithm, and show that our algorithm reduces the complexity of clustering process significantly while achieving almost same performance.