Experimental demonstration of local AI-Agents for lifecycle management and control automation of optical networks

This paper presents an innovative approach to automating the full lifecycle management of optical networks using locally fine-tuned large language models (LLMs) and digital twin technologies. We experimentally demonstrate the integration of generative AI and digital twins to create powerful AI-Agents capable of handling the design, deployment, maintenance, and upgrade phases in the lifecycle of optical networks. By deploying and fine-tuning LLMs locally, our framework eliminates the need for public cloud services, thereby ensuring data privacy and security. The experimental setup includes a commercial-product-based testbed with eight optical multiplex sections in the C-band, showcasing the effectiveness of the AI-Agents in various automation tasks, such as API-calling for service establishment and periodic power equalization, as well as log analysis for troubleshooting. The results highlight significant improvements in operational accuracy and efficiency, underscoring the feasibility of this approach in real-world scenarios. This work represents a significant advancement toward intent-based networking, showcasing the transformative potential of AI in automating and optimizing optical network operations.