A Deep Learning Assisted Approach for Minimizing the Age of Information in a WiFi Network

Motivated by the demands of time-sensitive applications, our paper studies methods for the age of information (AoI) minimization over a WiFi Network. Specifically, the AoI of a labeled device sending updates to the access point (AP) is of our concern. However, it is non-trivial to investigate the AoI optimization problem in such a scenario where an arbitrary number of background devices are also delivering updates to the AP; all of the network devices follow IEEE 802.11 based Medium access control (MAC) protocol to contend for the channel. Current works on AoI optimization over a WiFi network do not offer any practical methods for a single user to minimize its AoI readily and adaptively, limited to system AoI optimization with homogeneous traffic modeling or simplified MAC modeling. This paper develops a deep learning facilitated AoI optimization algorithm that provides direct guidance for easy implementation. Our novel method integrates service time analysis in 802.11 MAC and AoI queueing analysis with the deep learning enabled channel condition prediction. Specifically, we gather traffic rates of each node from the AP and train a channel condition estimation model. A labeled node can leverage the well-trained learning model to obtain accurate expected MAC service time and adaptively adjust its sending rate for minimal AoI. Simulation results show that our learning model can accurately estimate channels, and our algorithm guarantees fast adjustment of AoI-minimized sending rate.