Agentic AI Empowered Intent-Based Networking for 6G

Genze Jiang; Kezhi Wang; Xiaomin Chen; Yizhou Huang

cs.AIJan 10, 2026

Agentic AI Empowered Intent-Based Networking for 6G

Authors: Genze Jiang, Kezhi Wang, Xiaomin Chen, Yizhou Huang

Organizations: Department of Computer Science, Brunel University London, UK · Department of Computer Science, University of Reading, UK

Abstract

The transition towards sixth-generation (6G) wireless networks necessitates autonomous orchestration mechanisms capable of translating high-level operational intents into executable network configurations. Existing approaches to Intent-Based Networking (IBN) rely upon either rule-based systems that struggle with linguistic variation or end-to-end neural models that lack interpretability and fail to enforce operational constraints. This paper presents a hierarchical multi-agent framework where Large Language Model (LLM) based agents autonomously decompose natural language intents, consult domain-specific specialists, and synthesise technically feasible network slice configurations through iterative reasoning-acting (ReAct) cycles. The proposed architecture employs an orchestrator agent coordinating two specialist agents, namely Radio Access Network (RAN) and Core Network agents, via ReAct-style reasoning, grounded in structured network state representations. Experimental evaluation across diverse benchmark scenarios demonstrates that the proposed system outperforms monolithic agents, rule-based systems, and direct LLM prompting, with architectural principles applicable to Open RAN (O-RAN) deployments. The results also demonstrate that whilst contemporary LLMs possess substantial general telecommunications knowledge, structured prompt engineering encoding domain-specific decision thresholds and operational policies is essential for reliable network automation performance.

Figures & tables

Fig. 1: Network system model illustration showing RAN sectors characterised by load

L_{r}

, active users

U_{r}

, and spectrum availability

B_{r}

, alongside core nodes characterised by compute load

\kappa_{c}

and tier

\theta_{c}

. The latency matrix

\mathbf{T}

captures end-to-end latency

\tau(r,c)

between each sector-node pair.

Fig. 2: Hierarchical multi-agent architecture for IBN orchestration.

Fig. 3: Sequential coordination protocol for multi-domain network slice provisioning.

Sector ID	Active	Load	Spectrum Available (MHz)			Context
	Users	(%)	mmWave	Mid	Low
stadium_central	45,000	88	800	20	5	High-density urban venue
city_plaza	1,200	35	800	80	30	Commercial district
industrial_park_a	500	60	400	100	20	Factory environment
suburban_residential	3,000	45	0	60	40	Fixed wireless access
rural_highway	150	15	0	20	50	Wide-area coverage

TABLE I: RAN Sector Topology Characteristics

	Latency to RAN Sector (ms)
Node ID	stadium_	city_	industrial_	suburban_	rural_	Compute	Description
	central	plaza	park_a	residential	highway	Load (%)
mec_stadium_1	2	12	15	25	45	65	Stadium co-location (edge)
mec_industrial_1	15	12	3	25	40	30	Factory proximity (edge)
metro_agg_hub	8	5	10	15	25	50	Metro aggregation (metro)
regional_dc_north	35	30	35	40	50	40	Regional data centre (core)

TABLE II: Core Network UPF Node Characteristics

Scenario ID	Category	Semantic	Utility	Tokens	Time	Iterations
		Accuracy	Score		(s)
esports_stadium	URLLC	1.0	0.91	13363	3.2	3
industrial_automation	URLLC	1.0	0.94	13573	4.0	3
connected_ambulance_rural	URLLC	1.0	0.54	13617	3.5	3
ar_maintenance_industrial	URLLC	0.5	0.94	12989	3.1	3
stadium_4k_streaming	eMBB	0.5	0.71	12587	3.3	3

TABLE III: System Performance Across Benchmark Scenarios

Iteration 1: Intent Parsing and RAN Consultation

Orchestrator classifies traffic as URLLC based on keywords “ultra-low latency” (<5ms explicit threshold) and “safety-critical operations”, identifying dual constraints of latency and reliability for industrial control application.

ACTION: CALL_AGENT | agent_name=ran_specialist | request=Can industrial_park_a support ultra-low latency and high reliability for automated robotic assembly with mmWave, mid-band, or low-band spectrum? What spectrum allocation is recommended?

Iteration 2: RAN Response and Core Consultation

RAN specialist evaluates spectrum options considering reliability requirements in industrial environment (potential interference, penetration challenges). Recommends mid-band despite mmWave availability, justifying: “Mid-band offers better balance between capacity and penetration, crucial for maintaining high reliability in industrial setting with potential interference.” Issues load monitoring warning (sector at 60% utilisation).

Orchestrator synthesises RAN output and consults Core specialist, formulating URLLC-aware query emphasising stringent latency requirement:

ACTION: CALL_AGENT | agent_name=core_specialist | request=Given mid-band allocation at industrial_park_a for automated robotic assembly line requiring ultra-low latency (<5ms), which UPF node can provide necessary latency and what considerations for high reliability and current 60% sector load?

TABLE IV: Condensed Reasoning Trace for URLLC Industrial Automation Scenario

System	Semantic	Utility	Tokens	Time
Configuration	Accuracy	Score		(s)
Proposed Multi-Agent	0.67	0.75	13281	3.8
Monolithic Agent	0.62	0.74	1894	0.8
Rule-Based System	0.46	0.75	N/A	0.1
Direct LLM	0.58	0.66	1157	0.6

TABLE V: Comparative Performance Against Baseline Approaches

Fig. 4: Comparative performance across baseline approaches.

Configuration	Semantic	Utility	$\Delta$ Sem	$\Delta$ Util
	Accuracy	Score
Full System	0.67	0.75	0.00	0.00
Remove Prompts	0.58	0.73	-0.09	-0.02
Remove Specialists	0.62	0.74	-0.05	-0.01
Remove ReAct	0.54	0.55	-0.13	-0.20

TABLE VI: Ablation Study: Component Contribution Analysis

Fig. 5: Ablation study results across architectural components.

Intent-Based Networking

Intent-Based Networking translates high-level goals into validated network policies and uses feedback to manage services, resources, and performance.

See all 23 papers

6G Needs Agents: Toward Agentic AI-Native Networks for Autonomous Intelligence
May 2, 2026Mohamed Amine Ferrag, Abderrahmane Lakas, Merouane Debbah6G Networks LLM Agent Orchestration
Department of Computer and Network Engineering, United Arab Emirates University, UAE · Research Institute for Digital Future, Khalifa University, UAE
Agentic Wireless Communication for 6G: Intent-Aware and Continuously Evolving Physical-Layer Intelligence
Feb 19, 2026Zhaoyang Li, Xingzhi Jin, Junyu Pan +2Wireless Communications 6G Networks
College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou, China.
Performance Comparison of IBN orchestration using LLM and SLMs
Feb 27, 2026Wai Lwin Phone, Brahim El Boudani, Tasos Dagiuklas +1Intent-Based Networking Multi-Agent Orchestration
Dept. of Computer Science & Digital Technologies, London South Bank University, London, UK · Digital Catapult, London, UK

May 2, 2026J/K move · Enter open · S save

6G Needs Agents: Toward Agentic AI-Native Networks for Autonomous Intelligence

Mohamed Amine Ferrag, Abderrahmane Lakas, Merouane Debbah

Department of Computer and Network Engineering, United Arab Emirates University, UAE · Research Institute for Digital Future, Khalifa University, UAE

6G Networks LLM Agent Orchestration Wireless Communications Multi-Agent LLMs

Open paper AskPDF arXiv

Sixth-generation (6G) networks are increasingly envisioned as AI-native infrastructures integrating communication, sensing, and computing into a unified fabric. However, existing approaches remain largely optimization-centric, relying on closed-loop control with limited reasoning capability. In this paper, we argue for a paradigm shift toward Agentic AI-Native 6G, in which Large Language Model (LLM)-based agents operate as bounded, policy-governed reasoning entities within a semantic control plane layered above deterministic 3GPP infrastructure. We propose a four-layer architecture that integrates deterministic network infrastructure, semantic abstraction of intent and context, hierarchical reasoning, and a distributed multi-agent fabric spanning device, edge, and core domains. To assess feasibility, we develop a proof-of-concept agentic reasoning and orchestration framework and conduct an extensive empirical study using a domain-specific 6G benchmark under realistic deployment constraints. Our results reveal a fundamental tradeoff between reasoning capability and system efficiency, showing that no single model simultaneously satisfies latency, throughput, and accuracy requirements. Instead, heterogeneous deployment of LLM agents across the device--edge--core continuum is necessary to balance these constraints. We further demonstrate that quantization introduces non-uniform effects across models, reinforcing the need for system-level optimization rather than model-level compression alone. These findings establish agentic intelligence as a viable architectural direction for 6G and highlight key challenges in achieving scalable, trustworthy, and self-reasoning networks. All experimental results and evaluation scripts are publicly available to support reproducibility.

Abstract

Intent-Based Networking

Explore similar work

6G Needs Agents: Toward Agentic AI-Native Networks for Autonomous Intelligence

Agentic Wireless Communication for 6G: Intent-Aware and Continuously Evolving Physical-Layer Intelligence

Performance Comparison of IBN orchestration using LLM and SLMs