Toward a Dynamic Tasks Offloading and Resource Allocation for the Metaverse in In-network Computing

The Metaverse is anticipated to provide an immersive experience to a large group of simultaneous users. Optimal computing resource allocation is critical to meet the massive users’ demand. However, existing studies pay little attention to the Metaverse tasks decoupling in joint dynamic resource allocation problems. This paper considers the metaverse tasks decoupling, offloading and resource allocation problem for massive deployment. We model the joint tasks offloading problem as the mixed non-linear programming (MNLP) to minimize network energy consumption and task completion delay. We employ fuzzy c-means (FCM) clustering to group the users into groups. In each zone, a computing resource is available to serve several resource-constraint users. Each user can perform the tasks locally or offload to nearby computing nodes (FIN) or to a more distant, rich edge computing resource (EIN) in the operator’s access network, such as MEC. The experimental results suggest that a joint dynamic solution is critical for enabling massive metaverse task offloading.