Description
Date depot: 3 mai 2021
Titre: Automated Control of Future Transport Networks aided by Data Analytics & Machine Learning Methods
Directeur de thèse:
Adlen KSENTINI (Eurecom)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé: The research ambition is to propose and design a framework for a Machine Learning-assisted autonomous and dynamic optical network controller supporting real-time operations enabling to evaluate and prove the applications of the concepts of « Zero-Touched Optical Networking Control ».
The proposed thesis will start studying the identification and the evaluation of applying and extending Machine Learning techniques to the operations of optical networks with the objectives to make optical networks more adaptive (reconfigurable), agile (fast) and reliable (highly available) to deliver transport connectivity services on-demand (a.k.a., transport slices). The focus will be on closed-loop control with intelligent monitoring to increase end-to-end reconfiguration and survivability of open optical networks from simple workflows based on pre-defined policies to Reinforcement Learning based algorithms enabling SDN Controller to integrate autonomously new policies to improve its future control decisions.
The second year will define novel Software Defined Networking Controller architecture, Machine Learning based algorithms as well as the models that can integrate advanced monitoring and data analytics capabilities to improve optical networking control policies. The operational life cycle of the candidate Machine Learning methods, together with update strategies will be considered too. Further evaluations of the proposed Open Optical SDN Control platforms will continue being performed.
The third year will prove the concepts of Open Optical SDN Control platform based on Open-Source frameworks for device, topology and service models enhanced with machine-learning capabilities by demonstrating the benefits through different use-cased notably optical network service provisioning and maintenance with a special attention to proactive failure detection of optical network services to increase reliability and optical infrastructure utilization.
Doctorant.e: Errea Moreno Javier