Minimizing the Age of Information of Cognitive Radio-Based IoT Systems Under A Collision Constraint (2001.02482v1)

Abstract: This paper considers a cognitive radio-based IoT monitoring system, consisting of an IoT device that aims to update its measurement to a destination using cognitive radio technique. Specifically, the IoT device as a secondary user (SIoT), seeks and exploits the spectrum opportunities of the licensed band vacated by its primary user (PU) to deliver status updates without causing visible effects to the licensed operation. In this context, the SIoT should carefully make use of the licensed band and schedule when to transmit to maintain the timeliness of the status update. We adopt a recent metric, Age of Information (AoI), to characterize the timeliness of the status update of the SIoT. We aim to minimize the long-term average AoI of the SIoT while satisfying the collision constraint imposed by the PU by formulating a constrained Markov decision process (CMDP) problem. We first prove the existence of optimal stationary policy of the CMDP problem. The optimal stationary policy (termed age-optimal policy) is shown to be a randomized simple policy that randomizes between two deterministic policies with a fixed probability. We prove that the two deterministic policies have a threshold structure and further derive the closed-form expression of average AoI and collision probability for the deterministic threshold-structured policy by conducting Markov Chain analysis. The analytical expression offers an efficient way to calculate the threshold and randomization probability to form the age-optimal policy. For comparison, we also consider the throughput maximization policy (termed throughput-optimal policy) and analyze the average AoI performance under the throughput-optimal policy in the considered system. Numerical simulations show the superiority of the derived age-optimal policy over the throughput-optimal policy. We also unveil the impacts of various system parameters on the corresponding optimal policy and the resultant average AoI.