Overview of Quality of Transmission Estimation in Optical Networks

Accurate Quality of Transmission (QoT) estimation is essential for optical networks, yet conventional methods rely on large design margins that reduce efficiency and increase cost. This paper surveys the evolution of QoT estimation, beginning with performance indicators such as BER, Q-factor, OSNR, GSNR, ESNR, and EVM, and then reviewing analytical, simulationbased, and machine learning (ML) techniques. Regression and classification models are examined alongside ensemble methods such as XGBoost, which combine predictive accuracy with interpretability. The survey also discusses emerging large language model (LLM) frameworks, whose reasoning and generalization capabilities extend QoT estimation toward real-time automation. By integrating performance metrics, ML estimators, and LLM-driven intelligence, QoT modeling is positioned to enable marginless design, efficient RSA, amplifier optimization, failure management, and adaptive network reconfiguration.