Opportunistic Fog Computing for Next-Generation Radio Access Networks
One or more files will be made publicly available from 2022-09-20.
MetadataShow full metadata
The next-generation radio access networks (RAN) will not only enhance peak data rate and encourage ultra-low-latency applications, but also play a crucial role in connecting a broad range of edge devices resulting from the Internet-of-Thing (IoT) paradigm. This will lead to an exponential growth of data traffic and a need for ubiquitous processing. To support this, researchers have proposed the concept of cloud radio access networks (C-RAN) where client data received by base stations (BSs) are transmitted over fibre links to a commodity cloud platform for processing. However, the current C- RAN may not be a suitable candidate when it comes to dealing with issues of i) limited backhaul capacity; ii) excessive load concentration on the centralised base band unit (BBU) pool; and iii) challenges of meeting the delay-sensitive requirements of 5G. A promising alternative to C-RAN is fog-based radio access networks (F-RAN), which is based on the philosophy of harnessing the distributed resources of collaborative edge devices to deliver localised RAN services to end users. Still, the current F-RAN is mainly utilising dedicated processing hardware and does not leverage on the available large number of distributed edge devices. Upon undertaking an extensive literature review, we realised that there exists a considerable research gap in tackling the following issues: i) under-utilisation of resourceful end-user devices and constrained backhaul capacity; ii) optimal resource assignment for computationally intensive tasks; iii) limited flexibility and scalability of current solutions for computation offloading; and iv) secure management of distributed resources. The research undertaken in this thesis aims to provide insights and potential solutions to the aforementioned issues. Simulation and experimental evaluations, along with prototype implementation are also presented. The key research contributions of this thesis are reported in the following four chapters: Chapter IV addresses research gap (i) by proposing an opportunistic fog radio access network (OF-RAN) which can contribute a scalable solution to the key challenges faced by current RANs and address the issue of service load balancing in this type of RAN. To address research gap (ii), Chapter V investigates and resolves the task-node assignment problem in the proposed OF-RAN by utilising a multi-objective optimization approach. Chapter VI addresses research gap (iii) by analytically modelling and evaluating the performance of our proposed OF-RAN against existing RANs in order to gain insights into how OF-RAN can complement the existing architectures. Chapter VII addresses research gap (iv) by proposing the concept of blockchain-enabled OF-RAN which builds on the inherent security of blockchain decentralization and the collaborative processing of OF-RAN. The concept is investigated by both simulated and real experiments for a federated deep learning application that harnesses the edge devices in OF-RAN for object detection.