Task Offloading and Resource Allocation for 5G and Beyond Networks

Abstract

In this thesis, we give a detailed analysis of the challenges encountered in the 5G and beyond networks in the field of resource allocation. To start with, we provide an introduction to the development of communication technologies and highlight the distinct features of 5G and beyond networks. This serves the purpose of providing the readers with a more general view of the cause behind the challenges encountered in current networks and understanding the importance of addressing these challenges.

To solve these challenges, we proposed three schemes in this thesis, with one focusing on the resource allocation in 5G radio access network (RAN) slicing and two on task offloading and resource allocation in space-air-ground integrated networks (SAGIN). Specifically, these schemes are proposed to satisfy the following requirements under different environment settings: low complexity, online decision-making ability, strong scalability, strong generalization ability, and ease of implementation. To achieve these goals, all schemes are formulated into a block structure where each block accomplishes a dedicated task. The design of each block depends on the function to be accomplished, and both optimization-based approaches and DRL-based approaches are utilized to formulate a block. Each block can be easily implemented into other schemes to achieve the same function as in our work, which greatly improves the generalization ability of our proposed schemes. Additionally, this block structure enables the scheme to have each block designed separately. Several blocks can also be trained in parallel to reduce the complexity of the scheme.