Age of Information Analysis for Intermittent Updating Through a Gilbert Elliot Erasure Channel

Abstract

We consider information updating problems where generation of the update takes non-negligible time before making update transmissions in a sequential fashion through a Gilbert-Elliot erasure channel with causal channel state information (CSI), strictly causal channel feedback or no CSI. The generation time for each update is independent with a general distribution. The transmission (Tx) queue has a single data buffer to save the latest generated update. This model is inspired by sequential operations in intermittent computing based energy harvesting nodes. Once energy recharges, the node decides whether to generate a new update or to (re)transmit the update. We investigate window based and probabilistic retransmission schemes and obtain closed form average peak age of information (PAoI) expressions. We then provide numerical results that compare average PAoI performances with and without CSI and channel feedback, particularly with respect to threshold-based policies that allow young packets to be transmitted in each scenario, which are candidates for optimal policies.