Estimating the Quality of Transmission (QoT) of the optical signal from source to destination nodes is the cornerstone of design engineering and service provisioning in optical transport networks. Recent studies have turned to Machine Learning (ML) techniques to improve the accuracy of QoT estimation. In this paper, we survey the literature on this topic and classify the studies into categories based on their scope. Accordingly, we distinguish four categories of ML-based solutions: i) check lightpath feasibility, ii) estimate a lightpath's QoT, iii) enhance existing analytical models and iv) improve model generalization. We describe the proposed solutions in each category in terms of ML algorithms, inputs/outputs of the models, source of data and performance evaluation. Deploying a ML-based solution in the real field is not straightforward and presents several challenges. Therefore, we also discuss from an operator's perspective the potential of these solutions for real-field deployment.

The increase in computing capacity and the huge amount of available data have significantly accelerated the use of artificial intelligence and machine learning algorithms to solve complex real-life problems. In this context, optical networks can benefit from the power of machine learning algorithms to find innovative solutions to complex problems, in particular, by improving failure management. Failure management is an essential functionality of a network management system (NMS) to avoid degradation of signal quality or even the interruption of service. Existing failure management solutions are based on simple static rules, not suitable for modern optical networks characterized by a complex design and a large number of management parameters. In this paper, we use machine learning algorithms with data extracted from an NMS connected to an experimental setup to perform a rapid diagnosis and locate failures. We compare several machine learning models to identify, in the first step, correlated alarms that are related to the same type of failure—unexpected power attenuation—and to localize, in the second step, the position of this failure in the network. Using the four datasets generated by the experimental setup, results show that for both alarm classification and failure localization steps, the best performance is obtained by the light gradient boosting machine classifier with an F1-score greater than 90% and average precision and area under the curve scores close to 1.

Advance in optical transmission towards higher bit rate and denser spectral efficiency is challenged by nonlinear effects. The existing digital signal processing techniques to compensate nonlinear fiber transmission impairments suffer from heavy computations and require the knowledge of a large number of system parameters, which is impractical in field environment. Neural networks are among the investigated solutions in literature to cope with the complexity of those models. Taking advantage of the huge amount of data available in optical transport networks, WDM systems represent a fertile field to apply neural networks.This paper is a non-exhaustive survey on neural network applications for nonlinear impairments mitigation in optical fiber transmission systems. We distinguish two approaches. The first one is dependent of the NonLinear Schrödinger Equation (NLSE) while the second one is based on machine learning techniques. These two approaches achieve similar performance compared to the well-known nonlinear mitigation methods with reduced computational complexity.