DRL-Based AoI-Driven Scheduling for EH-Enabled CR-NOMA IoT Networks With Queue Constraints

This letter investigates age of information (AoI) minimization in uplink cognitive radio non-orthogonal multiple access (CR-NOMA) networks with energy-constrained secondary users (SUs) powered by non-linear energy harvesting (EH). We develop a deep reinforcement learning (DRL) framework that explicitly accounts for the coupled dynamics of stochastic traffic arrivals, finite queue and battery capacities, interference protection for the primary user (PU), and non-linear EH behavior. The proposed scheduling approach learns adaptive transmission policies that balance AoI minimization, reliability, and energy sustainability. Simulation results demonstrate that our framework closely approaches exhaustive search, while achieving near-optimal performance with substantially lower computational complexity.