ICANEWS

Engineering & Technology

AI-Native Open RAN: From xApps/rApps to Autonomous Network Agents - A Roadmap

This paper presents a comprehensive review of Artificial Intelligence (AI)-enabled Open Radio Access Network (O-RAN) systems, offering a unified perspective on

Overview

This paper presents a comprehensive review of Artificial Intelligence (AI)-enabled Open Radio Access Network (O-RAN) systems, offering a unified perspective on the evolution of intelligence within wireless networks. It systematically examines the O-RAN architecture, specifically focusing on the role of intelligence operating within both near-real-time and non-real-time RAN Intelligent Controller (RIC) frameworks. The authors also develop a taxonomy categorizing various AI approaches applied to O-RAN.

Research Context

Open Radio Access Networks (O-RAN) are characterized by openness, virtualization, disaggregation, and programmable intelligence, primarily facilitated through the RAN Intelligent Controller (RIC). This paradigm has emerged as a transformative development for future wireless systems. The existence of standardized interfaces and near-real-time control loops has created opportunities for integrating AI into the management and optimization processes of radio access networks.

Over several years, a diverse range of AI techniques have been proposed to address key challenges inherent in O-RAN. These challenges include radio resource management, network slicing, traffic prediction, mobility management, interference mitigation, and spectrum sharing. Despite considerable progress in these areas, existing solutions often exhibit specific limitations. They are frequently task-specific, necessitate extensive retraining, and show limited generalization capabilities across different deployment environments and varying network conditions.

Approach

The research approach involves a two-fold structure. Firstly, the paper examines the architectural framework of O-RAN, specifically analyzing how intelligence functions within the distinct near-real-time and non-real-time RIC frameworks. Secondly, it develops a taxonomy designed to classify various AI approaches applicable to O-RAN. This taxonomy encompasses a range of methodologies, including:

  • Machine learning
  • Deep reinforcement learning (DRL)
  • Digital-twin-assisted optimization
  • Emerging foundation-model-based architectures

Key Findings

  • Open Radio Access Networks (O-RAN) leverage openness, virtualization, disaggregation, and programmable intelligence via the RAN Intelligent Controller (RIC) for future wireless systems.
  • Standardized interfaces and near-real-time control loops in O-RAN facilitate AI integration into radio access network management and optimization.
  • AI techniques have been proposed for O-RAN challenges like radio resource management, network slicing, traffic prediction, mobility management, interference mitigation, and spectrum sharing.
  • Existing AI solutions for O-RAN often remain task-specific, require extensive retraining, and show limited generalization across deployment environments and network conditions.
  • A taxonomy of AI approaches for O-RAN includes machine learning, deep reinforcement learning (DRL), digital-twin-assisted optimization, and foundation-model-based architectures.