Toward Reinforcement-Learning-Based Intelligent Network Control in 6G Networks
Junling Li, Huaqing Wu, Xi Huang, Qisheng Huang, Jianwei Huang, Xuemin Shen
Abstract
Reinforcement learning (RL) is a critical enabler for optimizing performance, automating the deployment, and increasing the intelligence level of 6G networks. In this article, we first identify some advanced RL frameworks for diversified 6G service scenarios. We then envision RL-based intelligent network management for 6G from three different perspectives: cross-layer end-to-end network control for service-oriented software-defined networking (SOSDN), cross-network control for global coverage, and cross-service control for service customization. We also present the new challenges associated with RL-assisted network management in 6G networks and provide potential research directions. Finally, we use the smart grid as a typical 6G application scenario to demonstrate the critical role of RL-based methods in capacitating intelligent power system management.