Advanced Learning Schemes for Metaverse Applications in B5G/6G Networks

Sherief Hashima, Mostafa M. Fouda, Kohei Hatano, Eiji Takimoto

Abstract: The metaverse, a virtual world where various users can jointly/independently interact with digital objects seamlessly, is expected to be a key application area in beyond fifth generation (B5G)/ six-generation (6G) networks. However, the metaverse design poses several challenges, such as efficient usage of limited resources and the provision of high-quality user experiences. In this paper, we recommend advanced machine learning techniques such as multi-armed bandits (MABs), federated learning (FL), meta-learning, etc., to address these challenges. We illustrate the evaluability of these techniques with two specific use cases: resource allocation for virtual reality (VR) applications and personalized content delivery in the metaverse.