Dynamic Spectrum Fusion: An Adaptive Learning Approach for Hybrid NOMA/OMA in Evolving Wireless Networks

This study introduces an advanced adaptive learning algorithm to optimize the choice between Non-Orthogonal Multiple Access (NOMA) and Orthogonal Multiple Access (OMA) in wireless networks. Our approach utilizes a dual-node simulation environment to evaluate the performance of a novel hybrid NOMA-OMA scheme, especially under varying conditions of user mobility. A key introduction of our research is the implementation of a versatile, data-driven scoring system, designed to maximize system performance across diverse applications. To improve the decision-making process, we incorporated a range of machine learning techniques; Random Forest, Gradient Boosting Machine (GBM), and Multilayer Perceptron. This comprehensive approach, blending deep learning and traditional algorithms, is pivotal in determining the most effective Multiple Access (MA) scheme in various spatial scenarios. The integration of reinforcement learning techniques enables our system to dynamically adapt to changing conditions, facilitating real-time decision-making. The result is a dynamic and adaptive Hybrid NOMA/OMA scheme, demonstrating a significant improvement over traditional systems.