Robust weighted K-means clustering algorithm for a probabilistic-shaped 64QAM coherent optical communication system

Area of Science:

• Optical Communications
• Signal Processing
• Machine Learning

Background:

• Probabilistic-shaped 64 quadrature amplitude modulation (PS-64QAM) is crucial for high-capacity optical networks.
• Accurate clustering of constellation points is essential for reliable signal detection.
• Existing K-means algorithms may struggle with complex signal impairments.

Purpose of the Study:

• To introduce a novel weighted K-means algorithm for PS-64QAM signals.
• To improve the accuracy of decision point localization in optical communication systems.
• To enhance the robustness and efficiency of clustering algorithms.

Main Methods:

• A weighted K-means scheme utilizing a Maxwell-Boltzmann distribution weighting factor.
• Implementation and testing within a 120-Gb/s PS-64QAM coherent optical communication system.
• Comparative analysis against standard K-means in back-to-back and transmission scenarios.

Main Results:

• The proposed algorithm achieved superior bit error rate (BER) performance compared to K-means.
• Demonstrated significant optical signal-to-noise ratio (OSNR) gains (0.6dB-1.8dB) in back-to-back tests.
• Showcased improved robustness in locating decision points across a wider power range during transmission and faster convergence.

Conclusions:

• The weighted K-means algorithm offers enhanced accuracy and robustness for PS-64QAM signal clustering.
• It provides significant performance benefits, including OSNR gains and wider operational power margins.
• The method is particularly effective in mitigating impairments from fiber Kerr nonlinearity, reducing computational complexity.