Novel spectral decision-making model using artificial intelligence for multi-user access in decentralized cognitive radio networks
Diego Armando Giral-Ramírez, César Augusto Hernández Suárez, Enrique Rodríguez-Colina · Results in Engineering · 2025
ABSTRACT Next-generation wireless networks require fast information transfer with new spectrum access policies. Therefore, Cognitive Radio (CR) is a technology that proposes solutions to spectrum utilization and congestion problems generated by the fixed allocation of resources. This paper proposes an intelligent spectrum decision-making model for Decentralized Cognitive Radio Networks (DCRN) focused on multi-user access and spectral mobility. The model comprises six modules: spectrum characterization, decentralized network, deep learning, multi-user access, non-predictive decision-making, and evaluation metrics. The novelty of this work is the integration of these modules, applied to a simulation environment with data measured from the 5 GHz Wi-Fi band. An intelligent decision-making strategy is used by employing deep learning to extract features and machine learning to classify spectral data converted into Red, Green, and Blue (RGB) images, identifying three traffic levels: high, medium, and low. Furthermore, three non-predictive strategies are assessed to improve the decision-making process in a decentralized multi-user environment: Combinative Distance-based Assessment (CODAS), Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), and Multi-Criteria Optimization and Compromise (VIKOR). The number of total handoffs and the number of failed handoffs is used as metrics to evaluate the performance of spectral mobility. The results show that the proposed model improves spectral mobility, and the system enhances performance in decentralized environments. The deep learning-based decision-making process for DCRN with multi-user access improved user communication, reducing the spectral handoff rate, then this work results as a novel idea for its application in DCRN in practical scenarios. Despite numerous studies that have addressed different challenges and provided specific advantages and disadvantages in cognitive radio, there is still a need for adaptive solutions in this field. In this context, incorporating an adaptation module into the existing model is suggested, allowing recommendations and adjustments based on the requirements of the applications. This approach would improve the flexibility and effectiveness of the system under various conditions.