Dynamic power implementation with reinforcement learning for excursion-free amplified optical systems
Résumé
In optical networks, wavelength dependent gain in optical amplifiers, resulting in power excursion at their output, is still a challenge to address. Current proposed solutions are based on specific wavelength assignment, which can go against the wavelength continuity constraint in transparent optical networks. In this paper, we propose a dynamic power predistortion implementation using reinforcement learning to counter this restriction. We show that a dynamic adjustment of input optical power per each channel permits to reduce the mean power excursion at the output of the optical line system.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|