Partially Observable Minimum-Age Scheduling: The Greedy Policy

This paper studies the minimum-age scheduling problem in a wireless sensor network where an access point (AP) monitors state of object via set sensors. The freshness sensed state, measured by age-of-information (AoI), varies at different sensors and is not directly observable to AP. AP has decide which query/sample order get most updated information (i.e., with minimum AoI). In this paper, we formulate as multi-armed bandit partially arms explore greedy policy minimize expected AoI sampled over infinite horizon. To analyze performance policy, 1) put forth relaxed that decouples sampling processes arms, 2) process each arm Markov decision (POMDP), 3) derive average under sum from individual arms. Numerical simulation results validate excellent approximation terms