Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large language models for optical network O&M: Agent-embedded workflow for automation

Shengnan Li, Yidi Wang, Fubin Wang, Yujia Yang, Yao Zhang, Yuchen Song, Xiaotian Jiang, Yue Pang, Min Zhang, Danshi Wang

Abstract

With the continuous expansion of optical networks and the increasing diversity of services, existing operation and maintenance (O&M) approaches are increasingly challenged to meet the rising demands for intelligence and efficiency. Large language models (LLMs), endowed with advanced semantic understanding and contextual analysis capabilities, are emerging as a promising enabler for intelligent optical network O&M. Recent studies have demonstrated the feasibility of applying LLMs to optical network management, marking an important step toward intelligent automation. However, systematic investigations into how LLMs can be effectively integrated into existing O&M workflows remain limited. This paper addresses this gap by drawing inspiration from best practices in real-world O&M workflows and systematically identifying scenarios that are well suited for LLM integration. We highlight that agent-based design is key to improving the executability of tasks, and we propose a multi-Agent collaborative O&M architecture that integrates LLM capabilities with existing O&M tools. The proposed architecture leverages core LLM-related technologies including prompt engineering and tool invocation, to build Agent solutions targeting key tasks such as optical channel management, performance optimization, and fault management. This work presents a conceptual framework for embedding LLM-based Agents into optical network O&M workflows, forming agentized processes that demonstrate the feasibility of LLM-assisted task execution and lay the groundwork for future autonomous O&M systems featuring closed-loop perception, decision-making, and action.

Large language models for optical network O&M: Agent-embedded workflow for automation

Abstract

With the continuous expansion of optical networks and the increasing diversity of services, existing operation and maintenance (O&M) approaches are increasingly challenged to meet the rising demands for intelligence and efficiency. Large language models (LLMs), endowed with advanced semantic understanding and contextual analysis capabilities, are emerging as a promising enabler for intelligent optical network O&M. Recent studies have demonstrated the feasibility of applying LLMs to optical network management, marking an important step toward intelligent automation. However, systematic investigations into how LLMs can be effectively integrated into existing O&M workflows remain limited. This paper addresses this gap by drawing inspiration from best practices in real-world O&M workflows and systematically identifying scenarios that are well suited for LLM integration. We highlight that agent-based design is key to improving the executability of tasks, and we propose a multi-Agent collaborative O&M architecture that integrates LLM capabilities with existing O&M tools. The proposed architecture leverages core LLM-related technologies including prompt engineering and tool invocation, to build Agent solutions targeting key tasks such as optical channel management, performance optimization, and fault management. This work presents a conceptual framework for embedding LLM-based Agents into optical network O&M workflows, forming agentized processes that demonstrate the feasibility of LLM-assisted task execution and lay the groundwork for future autonomous O&M systems featuring closed-loop perception, decision-making, and action.
Paper Structure (16 sections, 11 figures)

This paper contains 16 sections, 11 figures.

Table of Contents

  1. Introduction
  2. Overview of Current Optical Network O&M
  3. Three-Layer Architecture of Optical Networks
  4. O&M Scenarios in Optical Networks
  5. LLMs for Optical Network O&M Automation
  6. Key Technologies for LLM Deployment in Optical Networks
  7. LLM-Based AI-Agent for Optical Network Operation and Maintenance
  8. Multi-Agent Architecture for Optical Network O&M
  9. Agent for Optical Channel Management
  10. Agent for Performance Optimization
  11. Agent for Optical Network Fault Management
  12. Challenges
  13. Real-time Data Access
  14. Digital Twin Modeling Capabilities
  15. LLM Reliability
  16. ...and 1 more sections

Figures (11)

  • Figure 1: Three-layer architecture of an optical network. (a) The physical layer comprises optical network elements and optical fibers, where each network element includes both hardware components and a local management system. (b) The control layer consists of the network management system (NMS), SDN controller, and databases. (c) The application layer encompasses the digital twin of the optical network and O&M applications, including optical channel management, performance optimization, and fault management.
  • Figure 2: Evolution of optical network O&M workflows from manual operations to fully agentic workflows.
  • Figure 3: Key enabling technologies for LLM-based solutions. The framework includes six core techniques—Prompt Engineering, RAG, Planning/Workflow, Tool Integration, and Memory—that collectively support the construction of robust and adaptable AI Agents in domain-specific scenarios.
  • Figure 4: Execution workflow of an LLM-based multi-Agent system. The Supervisor Agent interprets the user intent and coordinates subtasks among multiple specialized Sub-Agents, each utilizing tools such as prompt templates, RAG, and external APIs. The process iteratively refines the response until the final result is delivered.
  • Figure 5: Multi-Agent collaborative architecture for optical network operations and maintenance. The interaction layer includes bidirectional communication between human operators and the Supervisor Agent, as well as between the Supervisor Agent and task-specific Sub-Agents. The Sub-Agent layer includes three sub-Agents, each responsible for a key O&M domain: optical channel management, performance optimization, and fault management. The function layer defines core capabilities required to fulfill each Sub-Agent’s responsibilities.
  • ...and 6 more figures