Deep Learning-Augmented SHS Model for Accurate AoI Analysis in Heterogeneous Unsaturated CSMA Networks

Accurate analysis of the age of information (AoI) is crucial for timeliness-critical systems that often rely on carrier sense multiple access (CSMA) networks. Existing models that minimize the tagged node's AoI in CSMA networks, where nodes contend for channel access, perform well with near-saturated background traffic but struggle in heterogeneous unsaturated networks. To address this, we propose a deep learning (DL)-augmented stochastic hybrid systems (SHS) model for fast and precise AoI analysis. Our approach aggregates background nodes into a single virtual saturated group while reflecting their heterogeneous and unsaturated characteristics via the channel access rate. By leveraging DL, the access point (AP) augments the SHS model with locally monitored traffic, enabling the tagged node to achieve precise AoI analysis. This integration significantly improves SHS precision, leading to accurate AoI estimation in heterogeneous and unsaturated settings. Validation through 802.11-based simulations in the ns-3 simulator demonstrates our model's robustness and efficiency in practical CSMA network scenarios. Additionally, we introduce a joint evaluation metric to balance AoI and sampling cost, ensuring an optimal trade-off between information freshness and resource consumption.