Content Placement with Unknown Popularity in Fog Radio Access Networks

Abstract

The fog radio access network (F-RAN) is a promising paradigm to support the future explosive data growth by leveraging edge caching and edge computing. With the help of the new designed fog access points (F-APs), F-RANs can take the full advantage of local caching capabilities to relieve the load of fronthaul and reduce the transmission delay. However, the caching contents placement is a challenging issue due to the uncertainty and dynamics of the user content requests. Motivated by the recent development of artificial intelligence, a machine learning based F-AP contents placement method is designed in this paper. Wherein, the unsupervised learning is utilized to classify the popularity of requested contents first, then the content placement problem is solved by the deep reinforcement learning. The core idea of the presented method is that the network controller first intelligently predicts the popularity of requested contents through the unsupervised learning based algorithm, then it utilizes the deep reinforcement learning algorithm to fill the limited F-AP cache spaces with various contents meanwhile minimizing the fronthaul cost. Through numerical simulations, the convergence of the intelligent proposal is demonstrated, and the superiority of the proposal is verified by comparing with other baselines.