2022-05-23-中国移动-6G_RAN自治的数字孪生网络方法白皮书（英）_52页_2mb

Summary of Digital Twin Network Approach for 6G Wireless Network Autonomy

Introduction and Vision

• Current 5G network operation and maintenance (OAM) automation is insufficient, with high energy consumption, complex interoperability, and low efficiency. 6G aims to achieve full autonomy (L5) through digital twin technology, enabling self-optimization, self-evolution, and self-growth.

Self-Intelligence Network

• Defines autonomy levels L0 to L5 (targeting L4/L5). L5 represents fully automated, zero-touch network services. Current 5G OAM relies on expert rules, leading to fragmented and inefficient automation, which DTN addresses.

Network Digital Twin Concepts

• Three Contents: Physical entity, digital twin content, digital plan content. Tracks network states and enables generating plans for autonomy.
• Five States: Initial, planning, service, twin, energy-saving to manage lifecycle stages dynamically.
• Double Closed Loops: Inner loop simulates and optimizes network decisions; outer loop evaluates and adjusts based on real-world feedback for end-to-end autonomy.

Technical Features and Architecture

• Four features: models (data/class), on-demand generation, parallel delivery, automatic simulation. DTN architecture includes data plane for real-time data services and intelligent plane for AI-driven control.
• Key technologies: Data acquisition and analysis, knowledge graphs, graph neural networks, pre-verification techniques to enhance robustness.

Whole Lifecycle Autonomy

• Supports continuous planning, virtual-real connection (e.g., base station self-start), and pro-active/preventive fault cure. Case studies demonstrate improvements in beam weight optimization, RAN slices, and multi-dimensional resource scheduling.

Case Studies

• Beam Weight Optimization: Uses digital twin for intelligent MIMO adjustment, improving network coverage efficiently.
• Intelligent RAN Slices: Enables adaptive slicing based on predicted traffic to reduce complexity and resource usage.
• Multi-Dimensional Resource Scheduling: Federates energy-saving strategies across resources, optimizing network efficiency.

Challenges and Future Outlook

• DTN faces issues like data privacy, model interoperability, and architectural scalability. Future research focuses on universal meta-models, reliable virtual scenario generation, and efficient data utilization to enhance generalization and migration capabilities.