Deep Reinforcement Learning for Cooperative Coded Caching Strategy in Fog Radio Access Network

Fog radio access networks (F-RANs) are seen as potential architectures to support large-volumn data services. Fog access points (F-APs) act as content caching helpers in F-RANs, which relieves the transmission pressure of backhaul and decreases the transmission delay. Nevertheless, due to the random data request and stochastic location of users, how to take full advantage of the limited cache resources and achieve higher probability of successful transmission in F-RANs is still a tough task. In this paper, we propose an algorithm based Deep Reinforcement Learning (DRL) named DRL based cooperative coded caching (DB3C) method to search optimal content coded caching strategy with random linear network coding (RLNC). The main idea of DB3C is to search the optimal caching strategy in every request scenario by the controller in High Power Node (HPN) intelligently in terms of the deep Q network (DQN) model. Considering the Quality of Service (QoS) requirement and limited cache spaces in F-APs, the probability of successful transmission is regard as the vital factor to assess the performance of the DB3C method simultaneously. Simulation results show that the probability of successful transmission in proposed method performed much better compared with other baselines.