Skip to main navigation Skip to search Skip to main content

Agentic AI for Intent-Driven Network Orchestration: A Comparative Experimental Approach

Research output: Contribution to conference › Paper › peer-review

24 Downloads (Pure)

Abstract

Static, script-based network automation cannot keep pace with the complexity of modern IP-based networks. This paper proposes a hierarchical, stateful multiagent framework for intent-based network orchestration that translates operator intents into validated topologies and routing configurations, using specialised agents for interpretation, generation, validation, and deployment coordinated via a Model Context Protocol (MCP) server. To cut computational overhead, the framework substitutes a Small Language Model (SLM) fine-tuned with Low-Rank Adaptation (LoRA) for a monolithic LLM, and adds a Dual Modular Redundancy (DMR)-inspired dualagent check plus GNS3 pre-deployment verification. On 100 stratified intent instances (10–30 nodes), the SLM matches GPT-5-Mini’s correctness (McNemar p = 0.31) while cutting lifecycle time by 16%.

Workshop

WorkshopIEEE International Workshop on Computer Aided Modeling and Design of Communication Links and Networks, 2026
Country/TerritoryCyprus
CityNicosia
Period7/09/26 → 9/09/26
Internet address

Cite this