Data-driven Modelling Of Optical Communication System Devices

With the continuous development of mobile internet and the rapid spread of digital information,communication data traffic is experiencing unprecedented explosive growth.In the era of 5G/6G,coherent optical communication systems are playing an increasingly important role in meeting the growing communication demands.However,coherent optical communication systems face the challenge of signal quality degradation in high-speed data transmission,mainly due to the working characteristics of optoelectronic devices and inherent noise,which can cause signals to be easily affected by device impact and experience nonlinear distortion during transmission.Therefore,modeling and researching the problem of signal nonlinear distortion in the transmission process of coherent optical communication systems has significant research significance for future system design and signal compensation.However,current nonlinear modeling research for damaged coherent optical communication devices has reached a bottleneck.Traditional device modeling methods use a divide-and-conquer approach,where the modeling subject is a single optical communication device or a single damage model,which is difficult to simulate the nonlinear characteristics of the device and form a comprehensive systemic optical communication modeling framework.At the same time,the impact of noise has been frequently ignored in previous research,making the adaptability and robustness of the model in complex environments questionable.Therefore,this article proposes a new modeling algorithm and improves the existing problems in previous research.The main contributions of this article are as follows:(1)Traditional device modeling mainly focuses on individual optical communication devices or single types of damage models,which cannot form a comprehensive and systematic optical communication modeling framework.Therefore,this paper proposes an end-to-end framework that can model multiple devices and multiple types of damage,with the aim of forming a comprehensive and complete device modeling system.In addition,this paper takes the actual environment as a basis,and verifies the performance of the proposed modeling algorithm on a commercial platform.(2)In the existing research on coherent optical communication system device modeling algorithms,it is difficult to effectively decouple complex coupling damage patterns and handle noise interference.To address this issue,this paper proposes a modeling algorithm based on a hybrid variational autoencoder regression.The algorithm uses a variational autoencoder model for signal modeling to improve accuracy and reliability,and to remove noise and redundant information in the input data.The design of the regression model is used to learn latent feature vectors to achieve the device modeling task for coherent optical communication systems.(3)In existing modeling algorithms for coherent optical communication systems,there is a lack of research on learning implicit features in data.Therefore,this paper proposes a new deep metric learning method called hierarchical triplet loss.The hierarchical triplet loss function studies the features of different damage factors by exploring the hierarchical structure of anchor point space.The hierarchical triplet loss function is derived from the nonlinear mapping of latent space,which constrains the relative distances between hierarchical categories and obtains differentiated latent representations.The algorithm combines different neural network modules such as attention mechanism and deep residual network to process different damages,and each module has clear functional divisions,providing better interpretability and achieving adaptive learning of damage patterns.