Cost Efficient Intelligent Sensing Big Data Caching in ICN-IoT Networks.

The move towards sixth-generation (6G) is expecting to support not only more device connections but also the soaring multimedia big data traffic demands, which spawns the promising approach of in-network caching of information centric networks (ICN) in Internet of Things (IoT). By prefetching popular contents during off-peak traffic periods, caching nodes can make them available to users during peak periods, effectively improving the quality of experience. In this paper, considering the limited cache capacity, unknown popularity distribution as well as non-stationary user demands, we explore this problem by optimizing content caching with the objective of minimizing long-term transmission cost. The content caching process is modeled as a cooperative Markov decision process (MDP), aiming to maximize caching reward. To handle this optimization problem, we propose a deep Q network-based content caching (DQN-CC) algorithm to obtain the approximate optimal solution in an online fashion, thus the agent at the controller is able to adaptively learn and track the underlying dynamics. To update the cache of each node, a replacement rule based on the marginal gain is used, offering faster update and reduced complexity. Extensive simulations verify the superiority of the introduced design.