Resource Allocation for Next Generation Wireless Networks
Type of DegreePhD Dissertation
Electrical and Computer Engineering
MetadataShow full item record
The rapid development of various emerging applications, such as machine learning, virtual reality, and internet-of-things (IoT) brought massive data traffic. Next generation (6G) wireless networks are expected to support extremely high data rates and a wide variety of applications. To fulfill the requirement of 6G, enabling technologies such as intelligent reflecting surface (IRS) and machine learning have been proposed. In this dissertation, we explore the critical resource allocation problems in 6G wireless networks with the two enabling technologies. The first part of the dissertation focuses on wireless resource allocation in IRS-assisted networks. Two examples are explored. In the first example, we studied a fairness-aware resource allocation in a rate splitting (RS) network assisted by IRS. Joint active beamforming at the BS and passive beamforming at the IRS design is proposed so that the minimum user rate can be maximized. In the second example, an IRS-assisted federated learning system was studied. An efficient resource allocation algorithm based on optimization theory is proposed to jointly configure the communication and computing parameters to minimize the system energy consumption. The second part of this dissertation investigates resource allocation problems for wireless networks with machine learning via two examples. In the first example, the energy efficiency of a device-to-device (D2D) network is investigated. A deep learning approach is proposed to allocate the power resources to maximize the sum rate of all devices. In the second example, we studied a downlink resource block allocation problem in a radio access network (RAN) via deep reinforcement learning (DRL). Our results suggest that IRS has great potential in improving the system network performance. Optimization methods still play vital roles in resource allocation for next-generation wireless networks. Meanwhile, machine learning approaches would be indispensable tools to address some challenges that the optimization methods could not handle. At the end of this dissertation, we present a workflow to address the general wireless resource allocation problems. Future research directions are also given.