Adaptive Centralized Clustering Framework for Software-Defined Ultra-Dense Wireless Networks

This paper develops a new centralized clustering framework to mitigate strong intercell interference in a software-defined ultra-dense network, where the number of clusters can be adapting to network conditions. A directed interference graph is designed to capture the dominant interference resulting from user mobility. An asymptotically optimal Max-K-Cut method is proposed to partition the graph, achieving a (1 - 1/K) approximation of the optimum in a polynomial time-complexity, where K is the number of clusters. As a result, K can be adaptively adjusted to leverage among the optimality loss, throughput, and complexity. Numerical results show that our adaptive centralized framework performs significantly better than other centralized or semidistributed clustering schemes in terms of throughput.