Deep Reinforcement Learning-based Dynamic Lightpath Provisioning for Elastic Optical Networks
Keywords:Deep reinforcement learning, elastic optical network, routing and spectrum assignment, network control algorithm
In this paper, we develop a deep reinforcement learning-based routing, modulation format, and spectrum assignment (RMSA) algorithm for elastic optical networks that enable provisioning dynamically lightpath services. In order to enhance the network performance, the developed RMSA exploits deep reinforcement learning (DRL) mechanism for selecting efficient route and spectral resource by learning experiences of dynamic lightpath provisioning. Numerical simulations have been utilized to estimate the performance of the elastic optical networks applied to the proposed DRL-based RMSA solution. The obtained results demonstrate that our proposed network solution outperforms the conventional shortest path algorithm significantly and offers a notable performance enhancement in terms of blocking probability and accepted traffic volume.