A Framework of Topological Interference Management and Clustering for D2D Networks

Abstract

In this paper, we develop a joint clustering and topological interference management (TIM) framework for a device-to-device (D2D) network. This scheme divides the whole network into multiple groups, each served on a different frequency, and the interference within each group is managed by TIM, based only on the connectivity pattern and not on the instantaneous channel state information (CSI). To this end, we model TIM as a low-rank-matrix-completion problem (LRMC) problem and solve it using a novel and low-complex scheme based on semidefinite programming (SDP). As for the clustering part, we develop a clustering algorithm that is suited for the LRMC approach to solve TIM while building on the SDP relaxation of the maximum- $k$ -cut algorithm, and extending it to account for each cluster’s capacity. This clustering problem turns out to be a capacitated maximum- $k$ -cut problem, for which we derive a relatively tight upper bound, that helps in determining the performance guarantee of many clustering algorithms. Simulation results show that the joint clustering-TIM can help, in some cases, improve the system degrees-of-freedom (DoF), especially in large D2D networks. Our proposed scheme also reduces the computation time of the LRMC-based TIM approach.