AI in 6G Networks: Service and System Aspects

3GPP has been working to integrate AI into networks since the early stages of 5G. Technical specification groups such as RAN (Radio Access Network), SA (Service and System Aspects), and CT (Core Network and Terminals) are each standardizing various AI-related features to enhance network performance and improve user experience.

In particular, SA2 is focused on AI-driven automation of the 5G core network. To this end, it introduced a dedicated function called NWDAF (Network Data Analytics Function), which supports 23 analytics services as of Release 19. Through this, it provides analytics data for optimizing QoS decisions, analyzing UE mobility, and responding to network signaling congestion.

Federated Learning (FL) technology has also been introduced to protect data privacy and improve operational efficiency.

• Horizontal Federated Learning (HFL): Standardized in Release 18, supporting model training and inference across multiple NWDAF instances.
• Vertical Federated Learning (VFL): Standardized in Release 19, enabling more efficient AI operations between NWDAF and AF (Application Function).

Going forward, in the 6G era, the plan is to design AI-native and AI-friendly wireless communication systems based on cutting-edge AI technology, maximizing network performance and supporting a wide range of new use cases.