Asynchronous Advantage Actor-Critic Algorithm Based Cooperative Caching Strategy for Fog Radio Access Networks

With the introduction of edge caching to fog radio access networks (F-RANs), users do not need to fetch the preferred contents from a far distant cloud center, which effectively reduces content downloading latency. Inspired by the adoption of deep reinforcement learning (DRL) in solving caching-related issues, this paper proposes a latency-oriented cooperative caching strategy for F-RANs based on the asynchronous advantage actor-critic (A3C) algorithm. Specifically, content popularity is first predicted in an online manner. Aims at minimizing the average download latency, the A3C algorithm is then employed to learn the optimal caching strategy based on the obtained content popularity results. Simulation results demonstrate that the proposed algorithm achieves a lower average download latency compared to other baseline policies.