Minimizing AoI of Non-Uniform Multi-Source Real-Time Data Updates: Model Generalization, Analysis and Performance Evaluation.

This work studies the non-uniform multi-source data update problem for real-time monitoring systems, where a set of heterogeneous data sources transmit their updates to a Base Station (BS) through wire or wireless channel(s). The performance metric called Age of Information (AoI) - which measures the time elapsed since the last data update of each source received by the BS - is commonly used to quantify the freshness of the data updates. However, most existing work on minimizing AoI of multi-source data updates assume that all sources have a uniform size of data updates which unnecessarily reduces their applicability. This work explores a more general model where individual sources can have non-uniform sizes of data updates, and provides thorough analysis to optimize both peak and average AoI of the target system. Based on these analysis, an optimal scheme to minimize the peak AoI is first developed by guaranteeing the delivery frequency of each source proportional to the function determined by its data size. A $(2+\delta)$ -approximation algorithm based on random sampling (RS) and a heuristic called Ratio-driven Maximum Age First (RMAF) are further proposed to minimize the average AoI. Our extensive experiments validate the bound of RS, and show that RMAF can achieve close performance to the lower bound of the minimum time-average AoI and outperforms the state-of-the-art schemes.