Autonomous Constellation Shaping for Coherent Optical Fiber Communications

光纤通信系统在整个互联网中广泛使用，允许用户无缝地发送和接收跨越数十公里到数千公里距离的信息。企业、教育、医疗和娱乐行业对基于云的存储和服务、社交媒体服务、视频流和机器对机器应用的需求呈爆炸式增长，这推动了对连接性的需求，而这种需求只能通过光纤传输系统的持续创新来满足。高速数模转换器、模数转换器和数字信号处理的最新进展为光纤传输带来了新的范例;基于场调制、相干检测和数字信号处理的系统在性能和功能方面提供了巨大的优势。拟议的研究解决了光网络面临的最紧迫问题：需要增加在单根光纤上传输的总比特率，即，传输能力。该研究将集中在星座整形，因为它可以（i）抑制由于光纤中的信道内和信道间效应而由传播信号产生的性能限制非线性干扰（NLI）的产生，以及（ii）通过在给定接收光信噪比（OSNR）的情况下设置可实现的频谱效率方面提供相当大的灵活性来增加存在NLI时的传输容量。整形的目的是最大限度地提高可实现的，每信道的频谱效率，因此传输容量，对于一个给定的系统configuration.The研究将在弹性光网络（EONs）的背景下进行动态，异构的信道计划，支持多种数据速率，调制格式和信道带宽。与具有静态、均匀信道规划的传统光网络相比，这大大增加了设计最合适的星座整形的挑战。在EON中，相邻信道以及因此NLI可以沿着感兴趣的信道所穿过的路径沿着从一个跨度到另一个跨度而变化，并且随着网络自适应的发生，跨度上的相邻信道可以随时间而改变。因此，将研究基于机器学习技术（例如人工神经网络）的光学性能监测，作为从接收信号学习OSNR和NLI的性质的手段，从而允许自主地优化星座成形。研究的主要成果是实验验证了为自主优化星座成形而开发的方法，演示了由于设计成形的功效而增加的传输容量，以及熟练掌握前沿技术的高素质人员。

Optical fiber communication systems are used extensively throughout the Internet, allowing users to seamlessly send and receive information traversing distances from tens of kilometers to thousands of kilometers. The explosive increase in demand for cloud-based storage and services, social media services, video streaming, and machine-to-machine applications across the business, educational, healthcare and entertainment sectors is fueling a need for connectivity that can only be met by continued innovation in optical fiber transmission systems. Recent advances in high-speed digital-to-analog converters, analog-to-digital converters, and digital signal processing have led to a new paradigm in optical fiber transmission; systems based on field modulation, coherent detection, and digital signal processing provide substantial benefits in terms of performance and functionality.The proposed research addresses the most pressing problem facing optical networks: the need to increase the aggregate bit rate transmitted over a single optical fiber, i.e., the transmission capacity. The research will focus on constellation shaping as it can (i) suppress the generation of performance-limiting nonlinear interference (NLI) generated by propagating signals due to intra- and inter-channel effects in an optical fiber and (ii) increase the transmission capacity in the presence of NLI by providing substantial flexibility in setting the achievable spectral efficiency for a given received optical signal-to-noise ratio (OSNR). The shaping is aimed at maximizing the achievable, per-channel spectral efficiency, and hence the transmission capacity, for a given system configuration.The research will be pursued in the context of elastic optical networks (EONs) with dynamic, heterogeneous channel plans supporting multiple data rates, modulation formats, and channel bandwidths. This substantially increases the challenge of devising the most suitable constellation shaping compared to conventional optical networks with static, homogeneous channel plans. In EONs, the neighbouring channels, and hence the NLI, can vary from span to span along the path traversed by a channel of interest, and the neighbouring channels on a span can change with time as network adaptation occurs. Consequently, optical performance monitoring based on machine learning techniques, such as artificial neural networks, will be investigated as a means of learning the OSNR and properties of the NLI from the received signal, thereby allowing the constellation shaping to be autonomously optimized. The primary outcomes of the research are experimental validations of the methodologies developed for autonomously optimizing the constellation shaping, demonstrations of increased transmission capacity due to the efficacies of the devised shapings, and highly qualified personnel skilled in leading edge technologies.