Deep Learning Based Optical Performance Monitoring for Digital Coherent Optical Receivers

Abstract

Optical performance monitoring techniques are required to ensure reliable transmission in all types of optical systems. Optical performance monitoring techniques facilitate the estimation of link-degrading impairments such as optical signal-to-noise ratio degradations and nonlinear intrusions that are difficult to assess using conventional measurement methods. The development of new optical performance monitoring techniques will aid in the deployment of new links and monitoring of deployed networks. The objectives of this research are (a) to develop machine learning techniques that can estimate optical performance monitoring metrics in optical communication when deploying a new optical link and assess the condition of established links; (b) to assess the performance of the associated machine learning techniques; (c) to understand the factors that limit performance estimation; and (d) to identify optimal proxies for applying machine learning in digital coherent optical receivers.