Conceptual
Browse Instruct Curate

Multi-Armed Bandit Resource Management for AR in 5G Edge Computing

A proof-of-concept scheme that meets a roundtrip delay budget for augmented-reality applications on 5G mobile edge computing by splitting the budget across three hops and using a per-hop Multi-Armed Bandit to adapt physical resource blocks at the base station and GPU frequency at the edge, exploiting the monotonic resource-to-QoS relationship to avoid pre-trained neural networks and their sim2real gap.