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

Researchers propose a multi-scale agentic AI framework for Open Radio Access Networks (O-RAN) that uses hierarchical AI agents—from Large Language Models to wireless foundation models—to autonomously manage 6G network control across different timescales. The framework addresses operational complexity in disaggregated networks by enabling coordinated AI decision-making across standardized interfaces, demonstrated through proof-of-concept scenarios.

This research addresses a fundamental challenge in next-generation telecommunications: managing increasingly complex, software-defined networks without proportional increases in operational overhead. O-RAN's disaggregated architecture enables flexibility but creates coordination challenges when multiple independent control applications interact unpredictably. The proposed solution leverages recent advances in agentic AI—systems that interpret goals and autonomously coordinate multiple functions—rather than relying on isolated models or manual intervention.

The framework's three-tier architecture reflects practical network constraints. Non-Real-Time LLM agents handle strategic policy translation from operator intent, Near-Real-Time Small Language Models execute low-latency optimizations with millisecond response requirements, and distributed wireless foundation models provide inference at the edge near radio units. This hierarchical delegation mirrors how human network operators work, but replaces manual decision-making with AI coordination through standardized O-RAN interfaces.

For the telecommunications industry, autonomous network management reduces operational expenses and human bottlenecks while improving responsiveness to changing conditions. The proof-of-concept demonstrations—handling non-stationary environments and intent-driven resource allocation—suggest practical applicability beyond theoretical frameworks. However, deployment faces regulatory scrutiny around autonomous critical infrastructure control and requires extensive validation before carriers adopt such systems at scale.

The significance extends beyond telecom engineering. Successful implementation would validate agentic AI architectures for managing other complex, real-time distributed systems. The open-source approach using accessible models and datasets lowers barriers for research adoption, potentially accelerating O-RAN ecosystem development and 6G standardization timelines.

• →Multi-scale agentic AI framework organizes RAN control across Non-RT, Near-RT, and Real-Time layers using LLMs, SLMs, and wireless foundation models
• →Hierarchical agent coordination through standardized O-RAN interfaces enables autonomous network management without proportional operational complexity
• →Proof-of-concept implementation demonstrates viability for non-stationary network conditions and intent-driven resource control scenarios
• →Open-source foundation accelerates research adoption and potential 6G standardization, reducing barriers for carrier ecosystem development
• →Autonomous critical infrastructure control raises regulatory and validation challenges that must be addressed before production deployment