To design an optimized logical topology design of the optical network requires to use a suitable signaling scheme. In this case, the signaling scheme must have the ability to built for each incoming traffic the set of the needed path that guarantees the QoS (Quality of service) according to traffic requirements. Moreover, the established path must reduce the use of the available resources in terms of the lightpath, Fiber delay line, and wavelength conversion. For this reason, the signaling and switching scheme must cooperate for utilizing the required path. Finally, the used path must try to improve the QoS and minimize resource usage according to designed topology using machine learning.
There is already a rich literature on using ML to optimize routing decisions.
There was a presentation at IETF-100 on deep reinforcement learning.
There is a a research team at the Hebrew University that studied "Learning to Route".
Huawei Hong Kong has presented its "MIND" architecture.
Google is using ML for energy reduction (an important goal you didn't mention).
Several researchers have presented results on on-line (competitive) learning for routing and resource placement.
Y(J)S
I strongly encourage you to read this one:
Article Artificial Intelligence (AI) Methods in Optical Networks: A ...
I would also refer you to my group working on AI for optical networks:
https://www.researchgate.net/project/Deep-Learning-for-Optical-Networks
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts