Khare Prasunika, Costa Nelson, Ruiz Marc, Napoli Antonio, Comellas Jaume, Pedro Joao, Velasco Luis
Advanced Broadband Communications Center (CCABA), Universitat Politècnica de Catalunya (UPC), 08034 Barcelona, Spain.
Infinera Unipessoal Lda., 2790-078 Carnaxide, Portugal.
Sensors (Basel). 2025 Mar 20;25(6):1948. doi: 10.3390/s25061948.
C+L+S multiband (MB) optical transmission has the potential to increase the capacity of optical transport networks, and thus, it is a possible solution to cope with the traffic increase expected in the years to come. However, the introduction of MB optical technology needs to come together with the needed tools that support network planning and operation. In particular, quality of transmission (QoT) estimation is needed for provisioning optical MB connections. In this paper, we concentrate on modelling MB optical transmission for provide fast and accurate QoT estimation and propose machine learning (ML) approaches based on neural networks, which can be easily integrated into an optical layer digital twin (DT) solution. We start by considering approaches that can be used for accurate signal propagation modelling. Even though solutions such as the split-step Fourier method (SSFM) for solving the nonlinear Schrödinger equation (NLSE) have limited application for QoT estimation during provisioning because of their very high complexity and time consumption, they could be used to generate datasets for ML model creation. However, even that can be hard to carry out on a fully loaded MB system with hundreds of channels. In addition, in MB optical transmission, interchannel stimulated Raman scattering (ISRS) becomes a major effect, which adds more complexity. In view of that, the fourth-order Runge-Kutta in the interaction picture (RK4IP) method, complemented with an adaptive step size algorithm to further reduce the computation time, is evaluated as an alternative to reduce time complexity. We show that RK4IP provided an accuracy comparable to that of the SSFM with reduced computation time, which enables its application for MB optical transmission simulation. Once datasets were generated using the adaptive step size RK4IP method, two ML modelling approaches were considered to be integrated in the OCATA DT, where models predict optical signal propagation in the time domain. Being able to predict the optical signal in the time domain, as it will be received after propagation, opens opportunities for automating network operation, including connection provisioning and failure management. In this paper, we focus on comparing the proposed ML modelling approaches in terms of the models' general and QoT estimation accuracy.
C+L+S多波段（MB）光传输有潜力增加光传输网络的容量，因此，它是应对未来几年预期流量增长的一种可能解决方案。然而，MB光技术的引入需要与支持网络规划和运营的必要工具相结合。特别是，配置光MB连接需要进行传输质量（QoT）估计。在本文中，我们专注于对MB光传输进行建模，以提供快速准确的QoT估计，并提出基于神经网络的机器学习（ML）方法，该方法可轻松集成到光层数字孪生（DT）解决方案中。我们首先考虑可用于精确信号传播建模的方法。尽管诸如用于求解非线性薛定谔方程（NLSE）的分步傅里叶方法（SSFM）等解决方案由于其极高的复杂性和时间消耗，在配置期间用于QoT估计的应用有限，但它们可用于生成用于创建ML模型的数据集。然而，即使在具有数百个通道的满载MB系统上也可能难以实现。此外，在MB光传输中，通道间受激拉曼散射（ISRS）成为主要影响因素，这增加了更多复杂性。鉴于此，评估了相互作用绘景中的四阶龙格-库塔（RK4IP）方法，并辅以自适应步长算法以进一步减少计算时间，作为降低时间复杂度的替代方法。我们表明，RK4IP在减少计算时间的情况下提供了与SSFM相当的精度，这使其能够应用于MB光传输模拟。一旦使用自适应步长RK4IP方法生成了数据集，就考虑将两种ML建模方法集成到OCATA DT中，其中模型在时域中预测光信号传播。能够预测传播后接收到的时域光信号，为网络运营自动化创造了机会，包括连接配置和故障管理。在本文中，我们专注于在所提出的ML建模方法的模型通用性和QoT估计精度方面进行比较。
