23 October 2025 in TechRxiv
AFAlfonso Fernández-Durán
Wi-Fi networks, which are based on the IEEE 802.11 standard, are among the most widely used wireless networks worldwide. In recent years, new standards have been introduced that offer a range of benefits to users. However, achieving full potential for these sophisticated standards hinges on the implementation of effective resource management strategies. This task is increasingly challenging due to the growing demand for high-performance services, particularly in massive networks, where the frequency channel allocation (CA) problem is of critical importance. Identifying the optimal channel configuration frequently represents a significant challenge. In this scenario, it is often the case that classical optimization approaches are often not sufficient. In such instances, machine learning (ML) has emerged as a promising solution. However, current solutions are insufficient for addressing real-time monitoring and bidirectional data and control flows, both of which are crucial for the management of the dynamic nature of modern networks. This study demonstrates that a reinforcement learning (RL)-based system, integrated with a digital twin (DT) methodology, can create a virtual representation of the network, allowing real-time adjustments without additional overhead. The MELODWIN system has been developed to optimize CA and enhance the performance of large-scale Wi-Fi networks. The validation results, obtained through simulations and in a real operational Wi-Fi network, provide clear evidence that MELODWIN improves performance in complex network environments. Compared to previous solutions, this approach integrates continuous monitoring and dynamic management capabilities, offering significant advantages. The proposed system establishes a foundation for future architectures that leverage RL and DT for wireless network management. With its ability to efficiently optimize performance in large-scale WiFi environments, MELODWIN holds promise for adaptation to other emerging wireless technologies.
1 January 2026 in IEEE Transactions on Aerospace and Electronic Systems