Predictive Autonomic Transmission for Low-Cost Low-Margin Optical Networks

Low-cost low-margin optical networks play an essential role in the growth and upgrade of metro network infrastructures required to support the challenging requirements of 5G (and beyond) services. The use of low-resolution Analog-to-Digital Converters (ADCs) in conjunction with soft-decision Forward Error Correction (FEC) techniques is a promising technology for reducing typical high expenses of high data-rate optical transmission equipment. However, the resulting transmission systems are more sensitive and fluctuations of the state of polarization (SOP) caused by fiber stressing events might have a strong impact in terms of Quality-of-Transmission (QoT). To guarantee robust in-field operation, soft-decision FEC computation can be intensified to guarantee target post-FEC Bit Error Rate (BER) performance at the expense of increasing power consumption at the receiver. In this paper, we tackle at reducing this drawback while keeping the maximum QoT robustness by means of Machine Learning (ML) enabling predictive autonomic transmission capabilities. Specifically, the Autonomic Transmission Agent (ATA) based on ML is proposed for soft-decision FEC tuning based on short-term predictors using monitoring data including SOP and pre-FEC BER. By anticipating when it is actually needed to increase the computational effort dedicated to FEC process, minimum power consumption can be achieved. In addition, long-term prediction is proposed to suggest reconfiguration actions in the transmitter side. A set of experimental measurements is used to train and validate the proposed ATA system. Numerical analysis allows concluding that ATA based on ANN predictors achieve the maximum QoT robustness with a power consumption reduction up to 80% compared with static FEC configuration.