Optical Performance monitoring using Q-learning optimized least square support vector machine in optical network

Abstract

Optical performance monitoring (OPM) is one of the important technologies to ensure the normal transmission of optical network data. Although there are many techniques have been applied to OPM, it is still necessary to improve its monitoring performance and implement multidimensional monitoring. A novel method is proposed for signal-to-noise (OSNR) and chromatic dispersion (CD) estimation by using least square support vector machine (LSSVM) optimized based on Q-learning in optical communication system with the signal modulation format of QPSK, DP-16PSK and DP-16QAM. The signal features extracted from the asynchronous amplitude histogram (AAH) is used as the input to the model. Experiment simulation have been carried out in the value of OSNR range from 21dB to 32dB and the CD coefficient range from 8-17ps/km/nm. The results show that the proposed method achieves high estimation accuracy with the estimation error of OSNR and CD coefficient are 0.2064dB and 0.1597ps/km/nm, respectively. Compared with the traditional support vector machine (SVM), Q-LSSVM improved the performance of OSNR and CD coefficients by 16.3% and 27.9%, respectively. The proposed method is helpful to the multi-dimensional and accurate optical performance monitoring in the future optical networks.