Abstract
Recent advancements in large language models (LLMs) have enabled the development of AI agents capable of complex problem-solving and autonomous operations. This work demonstrates the first, to our knowledge, field trial of an LLM-powered AI agent for autonomous optical networks (AONs). The AI agent addresses challenges in managing complex optical network lifecycles. Three operational modes—LLM-native, LLM-centric, and rule-centric—are proposed to enable the management of multiple typical events during the optical network lifecycle. Demonstrated on a 440 km field trial testbed, the AI agent shows advanced capabilities such as wavelength add/drop, soft/hard failure management, and optical power optimization. Comparative analysis highlights LLMs’ potential in advancing AON operations, showcasing their role in achieving intelligent, efficient, and autonomous optical network management.