AI/ML Life Cycle Management for Interoperable AI Native RAN

TL;DR

This paper reviews the development of a standardized AI/ML life-cycle management framework for 5G RAN, addressing challenges like model drift and vendor lock-in, and paving the way for AI-native transceivers in 6G.

Contribution

It provides a comprehensive review of the evolving 3GPP standards for AI/ML life-cycle management in RAN, highlighting architecture, mechanisms, and open challenges.

Findings

Introduction of a five-block LCM architecture

Standardized interfaces for model transfer and monitoring

Identification of open challenges in resource efficiency and environment drift detection

Abstract

Artificial intelligence (AI) and machine learning (ML) models are rapidly permeating the 5G Radio Access Network (RAN), powering beam management, channel state information (CSI) feedback, positioning, and mobility prediction. However, without a standardized life-cycle management (LCM) framework, challenges, such as model drift, vendor lock-in, and limited transparency, hinder large-scale adoption. 3GPP Releases 16-20 progressively evolve AI/ML from experimental features to managed, interoperable network functions. Beginning with the Network Data Analytics Function (NWDAF) in Rel-16, subsequent releases introduced standardized interfaces for model transfer, execution, performance monitoring, and closed-loop control, culminating in Rel-20's two-sided CSI-compression Work Item and vendor-agnostic LCM profile. This article reviews the resulting five-block LCM architecture, KPI-driven monitoring mechanisms, and inter-vendor collaboration schemes, while identifying open challenges in resource-efficient monitoring, environment drift detection, intelligent decision-making, and flexible model training. These developments lay the foundation for AI-native transceivers as a key enabler for 6G.