Federated Meta-RL for Network Slicing Aware VNFs Orchestration in 6G Core Networks

The emergence of sixth-generation (6G) wireless networks demands innovative solutions to address the complexities of network slicing and the placement of Virtualized Network Functions (VNFs) in core networks (CNs). This study introduces a Fed-MRL (federated meta-reinforcement learning) framework for intelligent and adaptive VNF orchestration in 6G CNs. Using a Deep Q-Network (DQN)-based meta-reinforcement learning (MRL) approach, Fed-MRL dynamically optimizes VNF placement to meet diverse Quality of Service (QoS) requirements while adapting to real-time network conditions and traffic patterns. To enhance scalability and security, Fed-MRL integrates a distributed federated learning (FL) mechanism, enabling collaboration across decentralized CN elements without compromising data privacy. Our solution incorporates network slice awareness into the VNF placement process, ensuring optimal resource allocation, minimized latency, and improved network performance across multiple slices with distinct service demands. An optimization framework is developed to minimize service latencies and achieve slice-specific QoS objectives through intelligent, slice-aware VNF orchestration. The proposed Fed-MRL approach establishes a foundation for next-generation 6G CNs by enabling efficient resource management, enriched user experiences, and seamless deployment of adaptive services.