Reinforcement Learning based Adaptive Resource Allocation Scheme for Multi-User Augmented Reality Service
KyungChae Lee, Chan‐Hyun Youn · 2022 13th International Conference on Information and Communication Technology Convergence (ICTC) · 2022
Nowadays, thanks to the rapidly evolving AR/VR industry and 5G high-speed communication technology, it is possible to provide high-quality AR services to general users. However, despite this trend, the remaining problem is that the use of existing AR devices is inconvenient for various reasons such as heat generation and lack of computational power. Overcoming this could be achieved by the scheme named computational offloading. Applying the computational offloading in the edge server environment as in the typical general user AR/VR service scenario, however, requires careful management, on both the server and the user. In this paper, we propose a novel RL based resource management agent MARCO, which can perform edge server resource scheduling in a way that increases the user service quality, fully utilizing the resources in the edge server in AR offloading scenario. From our experimental results, we concluded that with our carefully designed reward functions and the training pipeline, our MARCO can surely act as an effective resource scheduler by outperforming other baselines by a large margin.