Caching Transient Data for Internet of Things: A Deep Reinforcement Learning Approach

Connected devices in Internet-of-Things (IoT) continuously generate enormous amount of data, which is transient and would be requested by IoT application users, such as autonomous vehicles. Transmitting IoT data through wireless networks would lead to congestions and long delays, which can be tackled by caching IoT data at the network edge. However, it is challenging to jointly consider IoT data-transiency and dynamic context characteristics. In this paper, we advocate the use of deep reinforcement learning (DRL) to solve the problem of caching IoT data at the edge without knowing future IoT data popularity, user request pattern, and other context characteristics. By defining data freshness metrics, the aim of determining IoT data caching policy is to strike a balance between the communication cost and the loss of data freshness. Extensive simulation results corroborate that the proposed DRL-based IoT data caching policy outperforms other baseline policies.