Deep Q-Learning-Based Cooperative Caching Strategy for Fog Radio Access Networks

Abstract

To reduce the burden on fronthaul link as well as transmission delay, this paper proposes a cooperative edge caching strategy based on the deep Q-learning (DQN) algorithm considering the cooperative caching behavior between fog access points (F-APs) for Fog Radio Access Network (F-RAN). Specifically, to obtain the desired content popularity, we first predict the user preference probability with the topic model. Furthermore, considering the coupled multi-variable nature of the optimizing problem, a deep reinforcement learning (DRL) based content caching strategy is adopted to acquire the optimal content placement policy by combining the content popularity prediction results and content popularity. Finally, numerical simulation results prove the proposed scheme can reduce the average download delay compared with the existing algorithms.