Dynamic Spectrum Access Using Cognitive Radio With Deep Reinforcement Learning and Non-Orthogonal Multiple Access

This work proposes a dynamic spectrum access (DSA) system that uses Cognitive Radio (CR), Deep Reinforcement Learning (DRL), and Non-Orthogonal Multiple Access (NOMA) to make better use of the wireless spectrum. Efficiency, throughput, and collision rates are all improved by the approach, which allows secondary users (SUs) to access channels opportunistically. DRL techniques like actor-critic and Double Deep Q-Network (DDQN) allow for real-time adaptive learning, while NOMA may multiplex users to boost spectrum efficiency. Important parameters that are assessed in the model are interference, channel occupancy and bit decoding rates. The outcomes of the simulation indicate that the throughput decreases with relatively small changes during episodes, but the mentioned system is flexible. Future developments are aimed at collision avoidance and throughput stability in highly dynamic environments.