Artificial intelligence (AI) has shown significant performance in optical network control and management. However, the reliability, complexity and deployment procedure of these AI-based applications need further investigation. To efficiently speed up the network automation and function extension, a digital-twin-based network control framework is proposed, which can intelligently synchronize with the practical system to support the upper-layer applications. To build a digital twin, high efficiency modeling, monitoring and self-learning mechanisms are the key building blocks. In this paper, we discuss our recent works on modeling, monitoring and self-learning methods for building a digital-twin for optical networks.
|