Abstract:
The report provides an overview of the motivation behind using reinforcement learning in network survivability and routing, modulation and spectrum allocation. The reinforcement learning algorithms that were explored throughout the study, namely Multi-armed bandit algorithms, Monte Carlo Methods, Q-Learning, and Deep Q-Networks, have found various applications in Q-Networks. This study aims to assess the application of these reinforcement learning frameworks to Routing, Modulation and Spectrum Allocation in Elastic Optical Networks. After considerable literature review, a deep Q-Learning based application of routing, modulation and spectrum allocation has been decided as the baseline for the research work.