Joint Spectrum and Power Allocation for Cognitive NOMA Communications With RF Energy Harvesting

To address the challenges of energy and spectrum scarcity in Internet of Things (IoT) systems, radio frequency (RF) energy harvesting (EH), non-orthogonal multiple access (NOMA), and cognitive radio (CR) have emerged as effective solutions. In this paper, we investigate a cognitive IoT network where IoT devices operate as secondary users (SUs), accessing temporarily unoccupied licensed frequency bands of primary users (PUs) while harvesting energy from transmission signals emitted by both the primary base station and other secondary users. To enhance spectrum efficiency, NOMA is employed for secondary uplink transmissions, with NOMA grouping implemented to mitigate intra-group interference. We formulate an optimization problem to maximize the long-term average throughput of IoT devices by jointly optimizing spectrum and power allocation. To solve this problem, we decompose it into two sub-problems: spectrum allocation, modeled as a many-to-one matching game where SUs and sub-channels iteratively refine their matching strategies, and power allocation, optimized using a deep deterministic policy gradient (DDPG) algorithm. By iteratively solving the two sub-problems, we obtain an efficient resource allocation strategy. Simulation results demonstrate that the proposed algorithm achieves fast convergence and consistently outperforms baseline schemes in terms of average throughput across various scenarios.