Network Management and Decision Making for 5G Heterogeneous Networks

Abstract

Heterogeneous networks (HetNets) will form an integral part of future cellular communications. With the proper management of network resources and decisions, the coexistence of small cells with macro base stations will improve coverage, data rate and quality of service for users. This thesis investigates critical issues that will arise in HetNets.

The first half of this thesis studies major consequences of the disparity between HetNet tier transmit powers, namely that of interference and load balancing. To reduce the effects of harmful interference to small cell users arising from powerful macro transmissions, we first design a precoding matrix in the form of a generalized inverse, which, unlike conventional precoding methods, allows the base station to target a user specifically to reduce its own interference to that user. Even with a transmit power constraint, the affected user can achieve significant improvement in its interference reduction at the slightly compromise of existing macro users.

Next, we study load balancing by showing the benefits of a dynamic biasing function for cell range expansion over a static bias value. Our findings indicate that a dynamic bias is a more intuitive way to prevent small cell overloading, and that associating closest users first is a preferred association order.

We conclude our study into load balancing by proposing a new notion of network balance. We describe how network balance is different to user fairness, and subsequently define a new metric called the network balance index which measures the deviation of the actual base station load distribution with the expected load distribution. We show using an algorithm that the network balance index is more useful than fairness in improving sum rate for clustered networks.

The second half of this thesis explores more advanced user-centric issues for HetNets. Chapter 5 proposes a user association scheme that achieves high fairness, and considers user association behaviour with network dynamics. In order to reduce the computation needed to re-associate a large network, we study the probabilities that each user will have to switch associations when a user or base station enters or leaves. In the process, we find that a shrinking network has more effect on user association than a growing one.

Finally, Chapter 6 extends the conventional idea of HetNets to include device-to-device (D2D) communications. We propose a D2D decision making framework that more suitably selects D2D modes for potential D2D pairs by using a two-stage criteria that leads to fewer incorrect D2D mode selections. Once a suitable D2D mode is selected, we demonstrate how to determine optimal or near-optimal power and resource parameters for each mode in order to maximize sum rate. We present a geometric approach to solving the co-channel power control problem, and closed form expressions where possible for orthogonal frequency allocation. Our comprehensive study validates the potential for D2D integration in future cellular communications.

The proposed techniques and insights gained from this thesis aims to illustrate how networks can be better managed and improve their decision making processes in order to successfully serve future users. Show more