What About Peak Age? Average vs. Peak AoI Minimization in Finite-Horizon Scheduling

Current communication standards typically emphasize latency, reliability, and throughput as the main performance metrics. However, a promising line of research is adopting age of information (AoI) as a more direct measure of data freshness, which is key for real-time applications and ambient sensing. In particular, industrial or mission-critical scenarios would likely require an AoI minimization over a finite horizon, e.g., corresponding to the duration of an intended observation window, under the constraint of a limited number of updates being exchanged. From the standpoint of a standard to be implemented in future communications and networking platforms, the choice would further be conflicted between minimizing the average or peak value of AoI. Anticipating this conundrum, we explore an optimization approach based on dynamic programming recursion, where we consider both average and peak AoI as possible objectives; we also evaluate one metric when the other is optimized. We further consider both independent and correlated errors. In all these cases, we are able to show that minimizing either AoI-related metric over a relatively short horizon often converges to similar threshold-based criteria. From a practical standpoint, this supports the insertion of either AoI-related metric in future standards, making further debates over average versus peak AoI minimization amount to just semantic distinctions with negligible practical impact.