Reinforcement Learning Based Routing, Modulation Level and Spectrum Assignment in Elastic Optical Networks

Routing, Modulation Level, and Spectrum Assignment (RMLSA) problems are critical to the success of Elastic Optical Networks (EONs). The literature reports several techniques for approaching RMLSA. However, the complexity of this problem necessitates the development of innovative approaches, which is the very essence of our work. Reinforcement Learning (RL) is an emerging alternative to dynamic RMLSA due to its ability to adapt to changes in the network's state through a learning process. Accordingly, this work proposes approaching the problem with RL techniques based on the Q-learning model. The experimental study conducts simulations over several test instances. The results confirm that the proposed approach shows promise in obtaining good results compared to the state-of-the-art heuristics