Intelligent optical performance monitor using multi-task learning based artificial neural network.

Area of Science:

• Optical communications
• Artificial intelligence
• Signal processing

Background:

• Accurate optical performance monitoring is crucial for advanced optical networks.
• Simultaneous monitoring of optical signal-to-noise ratio (OSNR) and modulation format identification (MFI) presents a significant challenge.
• Existing methods often require separate systems or exhibit lower accuracy.

Purpose of the Study:

• To design an intelligent optical performance monitor (OPM) capable of simultaneous OSNR monitoring and MFI.
• To leverage multi-task learning based artificial neural networks (MTL-ANN) for enhanced performance.
• To reduce the complexity of OPM devices for multi-parameter estimation.

Main Methods:

• Utilizing amplitude histograms (AHs) of optical signals as input features for the MTL-ANN.
• Developing and training a single MTL-ANN model for both OSNR monitoring and MFI.
• Validating the approach through simulations and experiments with NRZ-OOK, PAM4, and PAM8 signals.

Main Results:

• The MTL-ANN achieved 100% accuracy in MFI for all tested modulation formats.
• OSNR monitoring demonstrated high accuracy, with a mean-square error (MSE) of 0.12 dB in regression and 100% accuracy in classification.
• The MTL-ANN approach showed superior accuracy and stability compared to single-task learning based ANNs (STL-ANNs).

Conclusions:

• A single MTL-ANN can effectively perform simultaneous OSNR monitoring and MFI.
• The proposed intelligent OPM offers a simplified and highly accurate solution for optical network monitoring.
• This technology paves the way for reduced-complexity OPM devices in future heterogeneous optical networks.