Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Toward Autonomous O-RAN: A Multi-Scale Agentic AI Framework for Real-Time Network Control and Management

Hojjat Navidan, Mohammad Cheraghinia, Jaron Fontaine, Mohamed Seif, Eli De Poorter, H. Vincent Poor, Ingrid Moerman, Adnan Shahid

TL;DR

The paper addresses the growing complexity of O-RAN in 6G by proposing a multi-scale agentic AI framework that coordinates non-real-time, near-real-time, and real-time control loops. It introduces a three-tier architecture with LLM-based governance (Non-RT), SLM-based policy execution (Near-RT), and Wireless PHY Foundation Model (WPFM) inference (RT near the air interface), all connected via standardized O-RAN interfaces and telemetry. A proof-of-concept demonstrates governance, retraining, and intent-driven slice allocation in a live-like environment, validating stability and performance improvements over baselines. The work also discusses safety guardrails, model governance, data alignment, and standardization needs, outlining a practical path for incremental, standards-aware deployment of agentic O-RAN systems in 6G networks.

Abstract

Open Radio Access Networks (O-RAN) promise flexible 6G network access through disaggregated, software-driven components and open interfaces, but this programmability also increases operational complexity. Multiple control loops coexist across the service management layer and RAN Intelligent Controller (RIC), while independently developed control applications can interact in unintended ways. In parallel, recent advances in generative Artificial Intelligence (AI) are enabling a shift from isolated AI models toward agentic AI systems that can interpret goals, coordinate multiple models and control functions, and adapt their behavior over time. This article proposes a multi-scale agentic AI framework for O-RAN that organizes RAN intelligence as a coordinated hierarchy across the Non-Real-Time (Non-RT), Near-Real-Time (Near-RT), and Real-Time (RT) control loops: (i) A Large Language Model (LLM) agent in the Non-RT RIC translates operator intent into policies and governs model lifecycles. (ii) Small Language Model (SLM) agents in the Near-RT RIC execute low-latency optimization and can activate, tune, or disable existing control applications; and (iii) Wireless Physical-layer Foundation Model (WPFM) agents near the distributed unit provide fast inference close to the air interface. We describe how these agents cooperate through standardized O-RAN interfaces and telemetry. Using a proof-of-concept implementation built on open-source models, software, and datasets, we demonstrate the proposed agentic approach in two representative scenarios: robust operation under non-stationary conditions and intent-driven slice resource control.

Toward Autonomous O-RAN: A Multi-Scale Agentic AI Framework for Real-Time Network Control and Management

TL;DR

The paper addresses the growing complexity of O-RAN in 6G by proposing a multi-scale agentic AI framework that coordinates non-real-time, near-real-time, and real-time control loops. It introduces a three-tier architecture with LLM-based governance (Non-RT), SLM-based policy execution (Near-RT), and Wireless PHY Foundation Model (WPFM) inference (RT near the air interface), all connected via standardized O-RAN interfaces and telemetry. A proof-of-concept demonstrates governance, retraining, and intent-driven slice allocation in a live-like environment, validating stability and performance improvements over baselines. The work also discusses safety guardrails, model governance, data alignment, and standardization needs, outlining a practical path for incremental, standards-aware deployment of agentic O-RAN systems in 6G networks.

Abstract

Open Radio Access Networks (O-RAN) promise flexible 6G network access through disaggregated, software-driven components and open interfaces, but this programmability also increases operational complexity. Multiple control loops coexist across the service management layer and RAN Intelligent Controller (RIC), while independently developed control applications can interact in unintended ways. In parallel, recent advances in generative Artificial Intelligence (AI) are enabling a shift from isolated AI models toward agentic AI systems that can interpret goals, coordinate multiple models and control functions, and adapt their behavior over time. This article proposes a multi-scale agentic AI framework for O-RAN that organizes RAN intelligence as a coordinated hierarchy across the Non-Real-Time (Non-RT), Near-Real-Time (Near-RT), and Real-Time (RT) control loops: (i) A Large Language Model (LLM) agent in the Non-RT RIC translates operator intent into policies and governs model lifecycles. (ii) Small Language Model (SLM) agents in the Near-RT RIC execute low-latency optimization and can activate, tune, or disable existing control applications; and (iii) Wireless Physical-layer Foundation Model (WPFM) agents near the distributed unit provide fast inference close to the air interface. We describe how these agents cooperate through standardized O-RAN interfaces and telemetry. Using a proof-of-concept implementation built on open-source models, software, and datasets, we demonstrate the proposed agentic approach in two representative scenarios: robust operation under non-stationary conditions and intent-driven slice resource control.
Paper Structure (10 sections, 5 figures)

This paper contains 10 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Agentic AI O-RAN Framework
  3. Architectural Overview and Control Loops
  4. Agent Roles, Data Flows, and Telemetry
  5. Safety, Guardrails, and Conflict Resolution
  6. Proof of Concept
  7. Use Case I: Agentic WPFM Governance and Retraining
  8. Use Case II: Agentic Slice Resource Allocation
  9. Risks & Challenges
  10. Conclusion

Figures (5)

  • Figure 1: High-level architecture of the agentic AI framework.
  • Figure 2: The interaction cycle in the proposed agentic O-RAN architecture. Operator intent is translated by the LLM agent into A1 policies, executed by SLM agents via E2 control, and realized through WPFM dApps that act on the air interface.
  • Figure 3: WPFM governance for two scenarios: a) New class announcement for Technology Recognition (TR) in vehicular application. b) Interference Detection (ID) announcement for 5G-NR and LTE
  • Figure 4: Representative examples of agentic interaction illustrating how telemetry inputs and operator intent are processed through distinct reasoning stages to generate structured policies and control actions.
  • Figure 5: Comparison of static, heuristic, SLM-only, and agentic slice control strategies in a live 5G setup.