A Path Forward: 6G Resource Allocation from a Deep Q-Learning Perspective

The 6G paradigm presents a myriad of challenges, as it promises complex features such as managing diverse traffic profiles under a unified infrastructure. While many studies propose deep-Q learning (DQN) approaches for resource management in Network Slicing (NS) schemes, these algorithms often face a core issue: they are not easily reproducible in real-world environments due to their high dimensionality. In this study, we analyze a distributed DQN-based radio resource allocation methodology, designed to efficiently meet specific Service Level Agreements (SLAs). Our contribution includes making the code publicly available for further research and evaluation. We then assess its performance through a comparison with a Baseline DQN approach, highlighting the strengths and limitations of both models.