Optimizing Age-of-Information in Adversarial and Stochastic Environments

Abstract

We study a multi-user downlink scheduling problem for optimizing the freshness of information available to users roaming across multiple cells. We consider both adversarial and stochastic settings and design scheduling policies that optimize two distinct information freshness metrics, namely the average age-of-information and the peak age-of-information. We show that a natural greedy scheduling policy is competitive with the optimal offline policy in the adversarial setting. We also derive fundamental lower bounds to the competitive ratio achievable by any online policy. In the stochastic environment, we show that a Max-Weight scheduling policy that takes into account the channel statistics achieves an approximation factor of 2 for minimizing the average age of information in two extreme mobility scenarios. We conclude the paper by establishing a large deviation optimality result achieved by the greedy policy for minimizing the peak age of information for static users situated at a single cell.