Fifth-generation (5G) networks offer ultra-high speeds, ultra-reliable low-latency (URLLC), and massive connectivity, crucial for IoT, autonomous vehicles, and industrial automation. Yet, practical deployments often fall short due to static resource allocation and limited adaptability.
This thesis proposes AI-driven dynamic resource allocation through a custom Network Data Analytics Function (NWDAF), predicting a real-time user needs to proactively manage resources. It also focuses on seamless mobility via Multiple Access Edge Computing (MEC) service migration using Deep Reinforcement Learning and secures Open RAN environments against Denial-of-Service (DoS) attacks.
Experiments conducted on real-world testbeds reveal significant gains in latency, throughput, security, and resource efficiency, offering clear guidance for the design of the forthcoming 6G networks.