High-Level Service Type Analysis and MORL-Based Network Slice Configuration for Cell-Free-Based 6G Networks

Network slicing has garnered significant attention within the telecommunications community since the introduction of 5G. However, achieving dynamic and intelligent network slice configuration to accommodate diverse service types remains a critical challenge in advanced network orchestration. With the advent of 6G, which is characterized by its highly dynamic and robust nature, there is an urgent need for an intelligent and slice-compatible assignment approach to meet the evolving demands of next-generation networks. In this context, this work introduces an end-to-end network slicing framework that spans from the user to the Centralized Unit, within a system model incorporating an Open Radio Access Network and Cell-Free massive Multiple-Input Multiple-Output architecture. Our contribution begins with a detailed review of the anticipated 6G Key Performance Indicators and their implications for network slicing. We then propose a novel approach that leverages Multi-Objective Reinforcement Learning (MORL) to enable a single intelligent agent to address multiple service requirements through a unified training phase. By replacing multiple specialized agents with a single MORL agent, our approach significantly improves the scalability, reduces the complexity, and enhances the practicality of network slicing orchestration—while maintaining optimal system performance. Numerical results validate the effectiveness of the proposed MORL-based solution. The trained agent not only ensures the Quality of Service for diverse user service requests but also successfully manages the coexistence of conflicting service types. This includes accommodating the stringent requirements of Extremely Reliable and Low-Latency Communications alongside Further-Enhanced Mobile Broadband services within the same network environment.