| Popis: |
In this paper, we introduce a model-free online algorithm that is driven by dynamic regret to minimize the network traffic in edge computing in a non-stationary environment that is caused by user mobility and their varying demands for edge services. Our approach is based on Monte Carlo sampling of past experiences that are used to calculate the regret and adjust the exploration and exploration tendencies; moreover, the proposed approach does not assume any mobility patterns for the user, nor does it require any hyperparameter tuning. Preliminary results show that the proposed approach adapts to the ever-changing environmental conditions and converges towards the new optimal service deployment. |
