AI/ML in 3GPP 5G Advanced -- Services and Architecture

TL;DR

This paper surveys the AI/ML developments in 3GPP Release 19 within the SA group, distinguishing AI for network (AI-driven optimization) from Network for AI (network-enabled AI applications). It details CN enhancements via NWDAF, location services, VFL, policy/QoS improvements, and abnormal behaviour mitigation, as well as AI/ML media services, application-layer analytics, and OAM management enhancements. It also discusses Release-19's D2D AI/ML capability, split inference, and KPI-driven requirements, and outlines AI/ML beyond Release-19 with Release-20 study items and sustainability considerations. The study provides concrete architectural, data, and governance mechanisms enabling edge-cloud collaboration, privacy-preserving learning, and end-to-end AI-enabled 5G networks, with implications for future 6G.

Abstract

The 3rd Generation Partnership Project (3GPP), the standards body for mobile networks, is in the final phase of Release 19 standardization and is beginning Release 20. Artificial Intelligence/ Machine Learning (AI/ML) has brought about a paradigm shift in technology and it is being adopted across industries and verticals. 3GPP has been integrating AI/ML into the 5G advanced system since Release 18. This paper focuses on the AI/ML related technological advancements and features introduced in Release 19 within the Service and System Aspects (SA) Technical specifications group of 3GPP. The advancements relate to two paradigms: (i) enhancements that AI/ML brought to the 5G advanced system (AI for network), e.g. resource optimization, and (ii) enhancements that were made to the 5G system to support AI/ML applications (Network for AI), e.g. image recognition.

Figures (5)

• Figure 1: Figure shows an autonomous car connected to 5G network. The car uses hierarchical AI/ML models. In-vehicle ML model manages real-time tasks like emergency braking and obstacle avoidance. For more complex and computation-intensive tasks, ML models are available at the edge and cloud nodes. Edge (cloud, resp.) models are used for moderate latency (non real-time, resp.) tasks like route optimization (optimization of battery management system, resp.). This hierarchical deployment ensures efficient task allocation based on latency and computational requirements. The ML model in the car may use data from its sensors, the model at the edge may use data from the cars in the same neighbourhood, and the model in the cloud can utilize global data. The data path in edge and that in cloud is shown with red and green dotted line, respectively.
• Figure 2: Figure shows different elements in the 5G System and corresponding AI/ML enhancements. Blue circle shows network components that are enhanced, while summary of enhancement is given in the green box next to the network component.
• Figure 3: An overview of AI/ML enhancements in 5G system in Release 19 (SA group).
• Figure 4: AI/ML enhancements to 5G CN.
• Figure 5: AI/ML enhancements to media services and application layer in Release 19. Green boxes indicate AI/ML enhancements.