Deep Reinforcement Learning Based Coded Caching Scheme in Fog Radio Access Networks

Abstract

Fog radio access networks (F-RANs) have been presented as promising architectures for the future wireless system to provide high spectral and energy efficiency. With the help of the new designed fog access points (F-APs), F-RANs can take the full advantage of local caching capabilities, which relieves the load of fronthaul and reduces transmission delay. However, the cache resource optimization is a challenging task due to the uncertainty and dynamics of user file requests. Considering the high utilization of cache space and file diversity by coded caching, a deep reinforcement learning (DRL) based algorithm is developed for coded caching enabled F-RANs. The core idea of the proposal is that the network controller intelligently allocates the limited cache spaces of F-APs to different coded files based on the historical requests of the user. While the successful transmission probability of user requests is maximized during the learning process. Through numerical simulations, the convergence of the DRL based caching scheme is demonstrated, and the superiority of the proposal is verified by comparing with other baselines.