A Multi-Level Deep RL-Based Network Slicing and Resource Management for O-RAN-Based 6G Cell-Free Networks

With the deployment of the fifth generation (5G) of cellular networks, the focus of the information society has switched to the next era in which the limitations of 5G will be addressed, and the emerging services and applications will be satisfied. The sixth generation (6G) of wireless networks is envisioned to answer all demands of the next decade, which is only possible with advances in network design and management. This paper first presents a 6G-based network architecture that deploys emerging technologies, including Open-Radio Access Network (O-RAN) and Cell-Free massive Multiple-Input-Multiple-Output (CF mMIMO). Then, a hierarchical network slicing and resource management approach compatible with the presented architecture is defined. The proposed novel Reinforcement Learning (RL)-based scheme benefits from the openness of O-RAN to provide two levels of centralized multi-agent decision-making and decentralized single-agent execution for choosing proper service types by following the objective of maximizing the system capacity while guaranteeing the defined Quality of Service (QoS). To demonstrate the performance of the management method, Deep RL (DRL)-based algorithms for each level are proposed. Finally, the presented simulation results illustrate the effectiveness of the proposed solution in terms of peak data rate, user-experienced data rate, and latency.