Minimizing Age of Information by Deep Reinforcement Learning for Next Generation Multiple Access

In the next generation of multiple access, with the explosive growth of multimedia communication services and the emergence of new application scenarios, the timeliness of information has become the main goal for the future. And age of information (AoI) has been proposed to indicate the freshness of information. It is crucial to explore a channel access policy that optimizes AoI under feasible mechanisms in order to better improve the timeliness of the system. However, such channel access policy and corresponding mechanism are basically unknown. In recent years, the rapid advancement of machine learning, especially deep reinforcement learning (DRL), has provided promising solutions to address this challenge. Therefore, this article considers the problem of multiple access and aims to minimize the average AoI of the entire network. To achieve this idea, we first represent it as a decentralized Markov decision process (Dec-MDP). Then we propose a deep Q-network (DQN) based algorithm to help distributed nodes make access decisions independently, such that the freshness of network information could be improved. The corresponding simulations were finally conducted, and the results indicate that the DQN based policy can reduce global average AoI to a certain extent.