An optimization problem on the joint content caching and mode selection for slice instances in fog radio access networks (F-RANs) is researched in this paper, characterizing the unknown content popularity distribution and time-varying channel assumptions. In particular, hotspot and vehicle-to-infrastructure scenarios are considered and corresponding network slice instances are orchestrated in F-RANs. Considering different users' demands and limited resources, there exists a significant high complexity in solving the original optimization problem with traditional optimization approaches. Motivated by the advantages of deep reinforcement learning in solving sophisticated network optimizations, a deep reinforcement learning based algorithm is proposed, wherein the cloud server takes intelligent actions to maximize the hit ratio and sum transmit rate. The performances of the proposed algorithm are demonstrated to be significantly improved.
A Deep Reinforcement Learning Based Content Caching and Mode Selection for Slice Instances in Fog Radio Access Networks
01.09.2019
267764 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Cooperative edge caching via multi agent reinforcement learning in fog radio access networks
BASE | 2022
|IEEE | 2023
|Online Reinforcement Learning of X-Haul Content Delivery Mode in Fog Radio Access Networks
BASE | 2019
|