Load aware self optimisation based capacity scaling techniques for 5G cloud radio access networks

Abstract

Expanding network capacity is the main pillar to realising the fifth-generation (5G) vision as well as the use cases of Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communications (URLLC) and Massive Machine-Type Communications (mMTC). Capacity extension approaches involve bandwidth extension by using the Millimetre wave (mmW) band and enhancing the existing bandwidth efficiency through network densification as well as Multiple Input Multiple Output (MIMO). However, both approaches increase system complexity and cost. The Cloud Radio Access Network (C-RAN) architecture is a promising solution for the network densification cost issue in addition to introducing a high level of cooperation between the cells. Furthermore, a Self-Organising Network (SON) is another solution aimed at bringing intelligence and autonomous proactive adaptability, thereby reducing the cost and complexity of cellular networks. For this reason, the main focus of this thesis is measuring the merits of a SON in C-RAN, whereby various energy-efficient capacity scaling systems are proposed for the upcoming 5G and beyond. First, a self-organising C-RAN is introduced, which dynamically adapts to varying network capacity demands. The number of active Base Band Units (BBUs) and Remote Radio Heads (RRHs) is scaled based on the Cell Differentiation and Integration (CDI) technique, according to the load demand, which is aimed at efficient resource utilisation without sacrificing the overall Quality of Service (QoS). A CDI algorithm is proposed in which semi-static CDI and dynamic BBU-RRH mapping for load balancing are jointly performed. Network load balance is formulated as an optimisation problem with constraints. Discrete Particle Swarm Optimisation (DPSO) is developed as an Evolutionary Algorithm (EA) to solve the optimisation problem and compared to the Exhaustive Search (ES) Algorithm. The CDI-enabled C-RAN shows significant throughput improvement compared to a fixed C-RAN. Specifically, an average throughput increase of 45.53% and average blocked users reduction of 23.149% is achieved. Additionally, a power model is proposed to estimate the overall power consumption of C-RAN. Approximately 16% power reduction is calculated in a CDI-enabled C-RAN when compared to a fixed C-RAN. Second, since 5G in its preliminary deployment is going to coexist with 4G infrastructure, the CDI performance is tested for interworking between the 5G wireless access technology New Radio (NR) and Long Term Evolution-Advanced (LTE-A). There is an increase of 282.9% and 121.7% in throughput, whilst a reduction of 97.9% and 96.4% in the number of blocked users is observed in the NR CDI-CRAN network compared to fixed C-RAN and CDI-CRAN, respectively. Third, a scheme for reducing interference and increasing the rate coverage of the conventional macrocell by exploiting the mmw band spectrum is proposed. Since the cell edge area has a lower coverage data rate due to high interference and less received power, it is served by an mmw band. Small mmw RRHs are overlaid onto a conventional macrocell at the cell edges in a Heterogeneous Cloud Radio Access Network (HC-RAN) architecture. The RRH clustering is formulated as a bin packing optimisation problem. This scheme has significantly higher coverage and rate compared to conventional HCRAN. A gain of 20% and 40% is achieved in coverage and rate when compared to systems applying the Soft Frequency Reuse (SFR) technique and conventional HC-RAN.