Interference management for Device-to-Device communications in 5G networks

Abstract

While offloading has great potentials to relieve increasingly congested cellular networks, its benefits come at a cost, namely uncoordinated interference resulting from other Device-to-Device (D2D) pairs, which reduces wireless network capacity. In this dissertation, we propose to align the interference occurring from nearby D2D pairs using Topological Interference Management (TIM) instead of depending on the instantaneous channel state information (CSI), which is a task that hinders the practicality of D2D technology. We recast the TIM problem as a low-rank matrix completion problem (LRMC), which is usually NP hard due to the rank's non-convex and discontinuous nature. For this, we have progressively developed three efficient and novel approximation solutions to overcome the TIM matrix special structure with hard constraints (having all ones on the diagonal entries). While dealing with large networks in practice, we proposed to divide the whole network into clusters, and apply TIM on each of the resulting sub-groups in a parallel way, so that the scalability issue gets resolved. To this end, we carefully designed a clustering framework that works in favor of the LRMC-based TIM scheme, by relying on graph theory. Moreover, to make TIM more realistic, we also considered practical scenarios through accounting for path losses and mobility. Simulations have shown that the proposed methods minimize the matrix rank better than the existing works, while maintaining a polynomial complexity. By successfully approximating the rank completion problem, the system degrees-of-freedom (DoF) of a partially connected network of D2D-enabled devices (even of large dimensions) with no CSI increases, the occurred interference is managed, and hence the network throughput increases. To improve the DoF values even further, we also have utilized the successive interference cancellation (SIC) built-in capability in the 5G handsets for decoding, while maintaining in the overall an LRMC-based TIM approach.