Q-Learning Based Social Community-Aware Energy Efficient Cooperative Caching in 5G Networks

To satisfy the vast demand of data traffic caused by various mobile devices, caching contents at Base Stations (BSs) and User Equipment (UE) has become a promising solution. A lot of recent works have proved that an approach of caching at edge devices is efficient to reduce latency and alleviate backhaul load so that the probability to meet users' Quality of Service (QoS) is much higher. However, due to the selfishness of mobile users, it is hard to convince users to store contents for other users. In this regard, we consider the social connection between users in order to induce users to store contents for the social community. So, to implement the socially aware Device to Device (D2D) caching, we construct two types of graphs: i) physical D2D graph which is constructed based on users communication over D2D links and ii) logical social graph which is consist of a group of people who share the common interest. Then, we formulate socially aware D2D caching problem into Markov Decision Process (MDP). Additionally, we consider the distance and the social ties between users, and mobile devices' capacity to optimally store contents under the constraint that energy consumption of cellular link is much higher than D2D link. Finally, we solve the D2D caching problem by using Q-learning where the goal is to minimize the total energy consumption over Macrocell Base Station (MBS) and mobile devices.