Transfer Learning-Based EdgeAI Resource Allocation in NGEPON and B5G X-Haul Network for Immersive Media

This study introduces an EdgeAI-based framework for the dynamic bandwidth allocation and traffic priority prediction in 25G-Next-Generation Ethernet Passive Optical Network (NGEPON) and Beyond 5G (B5G) X-haul networks, targeting the requirements of immersive media services such as Virtual Reality (VR) and Augmented Reality (AR). The proposed framework employs a Software-Defined Networking (SDN) Cloud-Edge Service Orchestrator with an LSTM-transfer learning-based mechanism (EdgeAI_DBA) to estimate network traffic and optimize content caching at the network edge while dynamically prioritizing bandwidth allocation based on real-time traffic demands for the Fixed wireless NGEPON & B5G X-haul access network. Simulation results under realistic network conditions indicate that EdgeAI_DBA significantly improves packet delay, drop probability, and system throughput. These results emphasize that our EdgeAI_DBA mechanism can deliver high-capacity, low-latency services critical for future immersive media delivery.