Self-Supervised Visual Exploration of Age of Information Process in IoT

The freshness of information in an Internet of Things (IoT) system can be measured using the Age of Information (AoI). AoI is a process that evolves over time and displays a distinct saw-tooth pattern, providing system operators with real-time performance information of IoT devices. However, the plethora of proposed AoI-related metrics offers limited contextualization of the timeliness of the collected information. The latter being crucial for the operator when re-dimensioning the system. This study is the first to demonstrate that improved contextualization through visual exploration of temporal relationships and patterns in the AoI process can reveal more subtle properties of IoT devices. To this end, we propose a new self-supervised framework that initially expands AoI measurements to 2D image representations using the Gramian Angular Summation Field (GASF) and subsequently groups images with similar properties through a novel AoI Deep Clustering approach. The proposed approach enables the ascertainment of transmission patterns and update frequencies of IoT devices by observing how AoI from a transmitting device evolves over time. We demonstrate performance on two datasets: the first dataset shows that it is possible to differentiate between random and deterministic transmissions from the source, and the second reveals that the system can detect how often will device transmit.