Semi-Decentralized Interference Aware Scheduling in D2D-Enabled Cellular Networks

We investigate the problem of semi-decentralized, interference aware scheduling in a cellular network with Device-to-Device (D2D) links. Our goal is an algorithm that allows optimal resource allocation and power control in spite of mutual interference from and to D2D links, as well as the creation of a reporting structure that is independent of the exact optimization goal and D2D modes. To achieve this, we formulate an interference aware scheduling problem with general rate-utility and generalized reuse constraints. By applying a Generalized Benders Decomposition, the problem is decomposed into two coupled sub-problems, a "primal" power control problem and a "master" scheduling problem, that can be dealt with independently. By this, we decouple the tedious power optimization from the scheduling part. We evaluate performance of the proposed structure analytically and find it to allow optimal scheduling independent of the used utility and D2D modes, provided that the master scheduling problem is solved optimally. Finally, we propose a solution for the master problem of weighted sum-rate maximization. Our simulations indicate that using the proposed scheme, D2D reuse can increase by an order of magnitude compared to the often targeted one-fold reuse, reaching a maximum of 31-fold frequency reuse in our set-ups. It is further capable of improving sum-rate performance by around 35% over existing works, while keeping signaling overhead and optimization delay at the same order of magnitude.