A Review of Graph Neural Networks for In-Network Computing of Real-Time Metaverse Tasks
- Version
- Download 16
- File Size 326.70 KB
- File Count 1
- Create Date 8 December 2025
- Last Updated 8 December 2025
A Review of Graph Neural Networks for In-Network Computing of Real-Time Metaverse Tasks
Likhith Rao K1, Pradeep Nayak2, Maithri3, Manya4 ,Mithun H5
1 Student, Information Science & Engineering, AIET, Karnataka, India
2 Assistant Professor, Information Science & Engineering, AIET, Karnataka, India
3 Student, Information Science & Engineering, AIET, Karnataka, India
4 Student, Information Science & Engineering, AIET, Karnataka, India
5 Student, Information Science & Engineering, AIET, Karnataka, India
ABSTRACT
The realization of the “Metaverse”—a persistent, parallel 3D virtual world—depends on the network’s ability to deliver immersive experiences without motion sickness. This requires meeting the strict motion-to-photon latency constraint, typically under 20 milliseconds. The reviewed paper by Rashid et al. argues that current network architectures are insufficient for this requirement. Cloud servers incur large propagation delays, while Mobile Edge Com-puting (MEC) servers face heavy queuing delays during large-scale virtual events such as Metaverse concerts.
To address this, the authors propose the use of Computing in the Network (COIN), where programmable switches participate in computation. Rendering tasks are modeled as fork-joint structures and partially processed within the network. The key innovation is replacing slow optimization solvers with a fast Graph Neural Network (GNN) for real-time task placement.
Keyword : - Metaverse, Graph Neural Network, In-Network Computing, Real Time Task Placement, Mobile Edge Computing
Download