Double Deep Q-Network Based Joint Edge Caching and Content Recommendation With Inconsistent File Sizes in Fog-RANs

In this paper, the joint edge caching and content recommendation problem for the case of inconsistent file sizes is investigated for fog radio access networks (F-RANs). Firstly, we transform the joint caching and recommendation policy into a ‘single’ caching policy. Then, a time-varying personalized user request model is proposed to describe the fluctuant demands of users. To maximize the long-term net profit of each fog access point (F-AP), we formulate the caching optimization problem and resort to a reinforcement learning (RL) framework. To address the impact of file size inconsistency, a ‘pre-split’ mechanism with a dynamic upper limit is adopted to meet the constraint of storage capacity, and a ‘lazy’ updating mechanism is introduced into the training process. Finally, a double deep Q-network (DDQN) based distributed edge caching algorithm is proposed with content recommendation. Simulation results show that compared with the existing methods, the average net profit of our proposed algorithm can be increased up to 29.7
${\%}$
while content recommendation can not only increase caching efficiency but also accelerate convergence.