Synergistic Interplay of Large Language Model and Digital Twin for Autonomous Optical Networks: Field Demonstrations

TL;DR

A digital twin (DT)-enhanced LLM scheme is proposed to facilitate autonomous optical networks by leveraging monitoring data and advanced models, and furnishing LLMs with dynamic-updated information for reliable decision-making.

Abstract

The development of large language models (LLM) has revolutionized various fields and is anticipated to drive the advancement of autonomous systems. In the context of autonomous optical networks, creating a high-level cognitive agent in the control layer remains a challenge. However, LLM is primarily developed for natural language processing tasks, rendering them less effective in predicting the physical dynamics of optical communications. Moreover, optical networks demand rigorous stability, where direct deployment of strategies generated from LLM poses safety concerns. In this paper, a digital twin (DT)-enhanced LLM scheme is proposed to facilitate autonomous optical networks. By leveraging monitoring data and advanced models, the DT of optical networks can accurately characterize their physical dynamics, furnishing LLMs with dynamic-updated information for reliable decision-making. Prior to deployment, the generated strategies from LLM can be pre-verified in the DT platform, which also provides feedback to the LLM for further refinement of strategies. The synergistic interplay between DT and LLM for autonomous optical networks is demonstrated through three scenarios: performance optimization under dynamic loadings in an experimental C+L-band long-haul transmission link, protection switching for device upgrading in a field-deployed six-node mesh network, and performance recovery after fiber cuts in a field-deployed C+L-band transmission link.

Figures (6)

• Figure 1: Schematic of the synergistic interplay between DT of optical networks (ONet) and LLMs.
• Figure 2: Schematic of (a) hybrid data-driven and physics-informed DeepONet for fiber multi-channel power evolution modeling in DT. The topology of a long-haul C+L-band experimental link, a field-deployed six-node mesh network, and a field-deployed C+L-band link is illustrated in (b) with established DT GSNR accuracy calculated by MSE with positive for aggressive and negative for conservative. Mean absolute error is marked for GSNR accuracy w/ and w/o DT refined physical parameters.
• Figure 3: Schematic of the interplay between DT, including functions of power evolution prediction, parameter refinement, NLI calculation and ASE noise accumulation, and LLM, equipped with domain knowledge and plugins and tools. Verified strategies are pushed from the DT to ONet.
• Figure 4: Demonstration of autonomous optical performance optimization with dynamic loadings in System 1. Circled numbers represent steps.
• Figure 5: Demonstration of autonomous protection switching for device replacement in System 2. Circled numbers represent steps.