Research On Nonlinear Equalization Technology Based On Machine Learning In Coherent Optical Communication

In the context of the rapid increase of internet traffic and services,optical fiber communication systems with ultra-high speed,ultra-large capacity,ultra-long distance,and ultra-strong protection are inevitable development trend.Coherent optical communication systems can significantly increase optical fiber transmission capacity,and Quantum Noise Stream Cipher(QNSC)combines physical layer encryption technology to achieve high-speed and secure optical transmission.In order to further improve the transmission capacity of the above-mentioned systems,this paper mainly studies the nonlinear equalization technology in the coherent optical communication system and the QNSC system.Existing nonlinear equalization algorithms usually require complex signal processing and accurate prior link information of the transmission system.The nonlinear equalization scheme based on machine learning can achieve end-to-end fully blind nonlinear equalization without prior information and complex signal processing.This paper focuses on the nonlinear equalization strategy based on machine learning algorithm in the traditional coherent optical communication system and the QNSC system.The innovative research results obtained are summarized as follows:Firstly,there are many shortcomings of nonlinear equalization algorithms in traditional coherent optical communication systems.For example,traditional nonlinear equalization algorithms require prior link knowledge and complex signal processing,and some nonlinear equalization strategies based on supervised machine learning algorithms require training cost and signal cost.Therefore,this article studies a nonlinear equalization scheme based on fuzzy logic C-means clustering(FLC)algorithm.However,the results of FLC algorithm are easily affected by the initial clustering center points which are randomly selected.Therefore,the FLC algorithm easily converge to the local optimal solution.To solve this problem,the clustering algorithm based on heuristic algorithm is introduced such as the FLC algorithm based on the annealing evolution(FLC-AE)algorithm.The nonlinear equalization performance of these algorithms is verified in a 16QAM 120km non-realy coherent optical transmission system.Results show that FLC and FLC-AE algorithm can effectively compensate the nonlinear impairments in fiber.Compared with FLC algorithm,the FLC-AE algorithm is more stable and needs less time complexity.Secondly,in order to improve the transmission capacity of the QNSC transmission system effectively,while ensuing the security of the QNSC system,this article studies how to introduce a nonlinear equalization algorithm based on machine learning into the QNSC system.In addition,as the order of the modulation format increases,the performance of the clustering algorithm decrease and the time complexity increase.For example,the FLC algorithm and FLC-AE algorithm cannot normally compensate nonlinear impairments in 64QAM transmission system.To solve this problem,FLC algorithm based on subtractive clustering(FLC-SC)algorithm is introduced.The nonlinear equalization performance is demonstrated in 16QAM experimental system,16QAM and 64QAM simulation system.Results show that FLC-SC algorithm can compensate the nonlinear impairments in fiber effectively.Compared with FLC algorithm and FLC-AE algorithm,the FLC-SC algorithm is more stable and have less time complexity.Thirdly,to improve the nonlinear equalization performance of the nonlinear equalization algorithm,this paper attempts to use limiting technology in the transmitter,and the clustering algorithm is used in the receiver.The scheme is verified in the traditional coherent optical communication system and QNSC system.Results show that the combination of the two technologies can achieve better nonlinear equalization performance than only using clustering algorithm.