TLDR: A research paper by Nan Li et al. outlines the design and standardization principles for AI-Native Radio Access Networks (RAN) in 6G. Unlike 5G, 6G will integrate AI from the outset to manage complexity and support ubiquitous AI applications. The paper proposes three essential capabilities: AI-driven RAN processing, reliable AI lifecycle management, and AI-as-a-Service provisioning. It introduces a new ‘AI Node’ architecture and validates key features through large-scale field trials by China Mobile, demonstrating significant improvements in latency, root cause analysis, and energy saving. The work emphasizes the importance of Day 1 standardization and multi-SDO collaboration for successful 6G deployment.
The next generation of mobile networks, 6G, is poised to fundamentally change how we interact with technology, and at its core will be Artificial Intelligence (AI). Unlike 5G, where AI was often an afterthought, 6G is being designed with AI deeply embedded from the very beginning. This approach, known as AI-Native Radio Access Network (RAN), aims to tackle the increasing complexity of future networks and support a wide array of AI-powered applications.
This vision for 6G is explored in a recent paper titled Towards AI-Native RAN: An Operator’s Perspective of 6G Day 1 Standardization, authored by Nan Li, Qi Sun, Lehan Wang, Xiaofei Xu, Jinri Huang, Chunhui Liu, Jing Gao, Yuhong Huang, and Chih-Lin I. The paper draws on extensive experience in mobile network operations and standardization, offering a practical roadmap for integrating AI into 6G RAN from its initial deployment.
The Core of AI-Native RAN
The concept of AI-Native RAN revolves around three essential capabilities:
- AI-driven RAN Processing, Optimization, and Automation: This involves using AI to intelligently control and optimize various aspects of the radio access network, from the air interface to overall resource management. AI can learn in complex scenarios, make real-time decisions in vast environments, and predict patterns for proactive adjustments.
- Efficient, Controllable, and Reliable AI Lifecycle Management (LCM): This capability ensures that AI models are effectively managed throughout their lifespan. It covers everything from collecting the necessary data for AI, managing the models themselves, and enabling collaborative computing across the network. This includes continuous monitoring and adaptation of AI models to ensure stability and reliability in dynamic network conditions.
- AI-as-a-Service (AIaaS) Provisioning: This transforms the RAN from just a communication pipe into a unified platform that supports both traditional communication and advanced AI applications. It means the network can offer services like guaranteed connection quality for AI applications, provide valuable network data to AI services, and even expose its computing resources (like GPUs) for external AI applications.
A New Architecture for 6G
To realize this AI-Native vision, a new functional architecture is proposed, featuring a novel component called the “AI Node.” This AI Node, equipped with significant AI computing resources, works in conjunction with the 6G Base Stations (6gNBs). The AI Node handles computationally intensive tasks like AI model training and complex inferences, while the 6gNBs manage real-time communication and simpler AI tasks. This hierarchical design allows for efficient use of AI resources, supports collaborative AI across the network, and enables a more programmable and adaptable RAN.
Validating the Vision: China Mobile’s Field Trials
To validate these concepts, a large-scale field trial was conducted by China Mobile across 31 cities in China, involving over 5000 5G-Advanced base stations. The trials demonstrated significant improvements in key areas:
- Differentiated Service Assurance: AI-enabled solutions significantly reduced air interface latency for applications like short video streaming (by 25.6%) and QR code scanning (by 21.9%), even under challenging network conditions. This also led to a 3-6% increase in overall network traffic, as improved quality encouraged higher resolution streaming.
- Root Cause Analysis for Poor User Experience: By leveraging multi-dimensional data and AI models, the system improved the accuracy of identifying the root causes of poor user experience by 20% and classification recall by 30% compared to traditional methods. This allows for more proactive network optimization.
- Network Energy Saving: AI-enabled solutions achieved substantial energy savings, with one advanced solution reaching 34.16% energy reduction compared to a baseline without energy-saving mechanisms. This highlights AI’s potential for building more sustainable mobile networks.
Also Read:
- AI-Powered Resource Allocation for Reliable Wireless Control Networks
- Asia’s AI-Native Startups Spearhead Transformative Growth Across Key Industries
The Path Forward for Standardization
The success of AI-Native RAN in 6G heavily relies on effective standardization from Day 1. This includes defining new data collection mechanisms, robust AI model management procedures, and protocols for collaborative AI computing. It also emphasizes the need for forward compatibility, ensuring that hardware and software can evolve with future AI advancements. Furthermore, the paper stresses the importance of collaboration among various standardization bodies like 3GPP, ITU, and O-RAN Alliance to ensure a unified and consistent approach to 6G AI-Native RAN development.
The insights from these trials and the proposed framework provide a strong foundation for the standardization and commercialization of AI-Native RAN, paving the way for more intelligent, efficient, and sustainable communication networks in the 6G era.
