A Deep Q-Learning based Architecture for 2 and 4 Users to Optimize Power Allocation and Signal Detection in NOMA Cognitive Radio Networks

Non-orthogonal Multiple Access (NOMA) is a crucial technique in Cognitive Radio Networks (CRNs) that improves frequency band use efficiency. However, NOMA may encounter difficulties due to its complicated methods for canceling out interference and identifying channels, particularly when conditions change often. To address these issues, this work proposes a novel strategy that combines Power Allocation (PA) and Deep Learning (DL) to improve signal identification and interference manage-ment in NOMA systems with two or four users. We employ a Deep Neural Network (DNN) to detect signals and Deep Q-learning (DQL) to intelligently allocate power among users. This strategy improves our ability to recognize several users at once, as well as make decoding mixed signals easier and more efficient. Our technique performs far better in coping with the complicated and shifting situations seen in real-world communications. The findings reveal that Deep Q-Learning minimizes the complexity of interference cancellation and significantly improves the system's overall performance, addressing frequent concerns encountered in NOMA in CRNs.