Minimizing Age of Information in Multi-Class Systems Using an Elapsed-Time-Based Hybrid Service Discipline

Minimizing the age of information (AoI) has recently become a key objective in Internet-of-Things technology. However, research on prioritized sources with distinct AoI constraints remains limited, despite its significant relevance in real-world applications. In addition, the existing service disciplines —preemption (PR) and non-preemption (NP)— result in a biased satisfaction across priority classes. Therefore, this work proposes a hybrid PR/NP service discipline based on the elapsed service time for a multi-class status update system, considering Poisson arrival process and Erlang service time distribution. Class-specific thresholds on the elapsed service time are set to limit further service preemptions. By applying the stochastic hybrid system approach, we analyze the AoI moments for arbitrary number of classes and derive key insights that underscore the benefits of our model over classical ones. Numerical evaluations of average AoI and AoI dispersion demonstrate that adjusting threshold parameters can achieve a more balanced performance across classes. Additionally, we explore threshold optimization to either enhance the performance of the most AoI-sensitive class or maximize overall network satisfaction while meeting quality-of-service constraints. For the latter, a near-optimal approach is proposed to reduce the computational complexity of determining the optimal solution as the number of priority classes increases.