Average AoI Minimization in WP-IoT Networks: Optimal Scheduling for NOMA Transmission

Age of information (AoI) serves as a critical metric to measure the freshness of information. This paper investigates AoI minimization via sensor scheduling in a non-orthogonal multiple access (NOMA)-based wireless-powered Internet of Things (WP-IoT) network. Specifically, a base station (BS) transmits radio frequency (RF) signals continuously over the downlink to power multiple IoT sensors (IoT-Ss). In each time slot, two IoT-Ss are scheduled to transmit status update packets to the BS via NOMA over the uplink. Since the packets are usually short, we formulate an average AoI (AAoI)-minimization optimization-scheduling problem in the finite blocklength (FBL) regime. To obtain the optimal scheduling policy, we convert the sequential decision-making problem to a Markov Decision Process (MDP) and develop a deep reinforcement learning (DRL) solution. We also derive an AAoI performance lower bound for the system under consideration. Simulation results verify that NOMA transmission achieves significant performance improvement over orthogonal multiple access (OMA) transmission. Moreover, our obtained scheduling policy for NOMA transmission can achieve a much smaller AAoI than the benchmark scheduling policies and approaches its performance lower bound.