Autonomous Operations in Optical Networks

Machine Learning (ML) has already proven its benefits for network operation, being a sub-domain of artificial intelligence, it is highly suitable for complex system representation. In this paper, basic ML concepts are reviewed, as well as its integration into existing network control and management planes. Then, a use case focused on soft-failure detection is presented in detail covering optical spectrum analysis and ML algorithms; the technique relies on the widespread deployment of cost-effective optical spectrum analyzer (OSA). Finally, the retrieved optical parameters are analyzed using ML algorithms giving rise to illustrative results.