Reimagining RAN Automation in 6G: An Agentic AI Framework with Hierarchical Online Decision Transformer

2026-04-04

AI Summary

This work addresses the limitations of current 6G radio access networks (RANs)—notably insufficient automation, difficulty in interpreting human intent, and inadequate response to anomalies—by proposing an agent-based AI framework. The framework integrates a hierarchical online decision-making Transformer, retrieval-augmented generation (RAG), and generative AI for time-series forecasting to enable multi-agent collaboration in resource allocation, service orchestration, and self-healing control. It further introduces a novel two-tier human intent verification mechanism. Experimental results demonstrate improved system throughput alongside reduced latency and energy consumption; the system accurately identifies degraded user intent with 88.5% precision and autonomously recovers 90% of baseline performance through its self-healing capability. This study represents the first integration of generative AI with hierarchical decision Transformers for RAN automation.

Abstract

In this paper, we propose an Agentic Artificial Intelligence (AI) framework for wireless networks. The framework coordinates a pool of AI agents guided by Natural Language (NL) inputs from a human operator. At its core, the super agent is powered by a Hierarchical Online Decision Transformer (H-ODT). It orchestrates three categories of agents: (i) inter-slice, intra-slice resource allocation agents, (ii) network application orchestration agents, and (iii) self-healing agents. The orchestration takes place with the help of an Agentic Retrieval-Augmented Generation (RAG) module that integrates knowledge from heterogeneous sources. In this proposed methodology, the super agent directly interfaces with operators and generates sequential policies to activate relevant agents. The proposed framework is evaluated against three state-of-the-art baselines, showing improved throughput, reduced network delay, and higher energy efficiency at both slice-level and system-wide performance metrics. Also, the proposed Agentic framework introduces a bi-level human operator intent validation methodology, both at the slice-level and Key Performance Indicator (KPI)-level using generative AI-based time series predictors. We could rule out performance-degrading operator intents with an accuracy of 88.5%. Lastly, while being interrupted by any performance-degrading events, the self-healing capability of Agentic AI in our framework automatically recovers 90% of its previous performance, avoiding quality-of-service drifts when there is no human involvement.

Problem

• RAN Automation
• 6G
• Agentic AI
• Human Intent Validation
• Self-healing Networks

Innovation

• Agentic AI
• Hierarchical Online Decision Transformer
• RAN Automation
• Retrieval-Augmented Generation
• Self-healing Networks