Optimized Bandwidth Variable Transponder Configuration in Elastic Optical Networks using Reinforcement Learning

Abstract:
To maximize transmission capacities in Elastic Optical Networks, a large parameter space needs to be optimized. We show that by formulating this task as Markov Decision Processes and using multi agent reinforcement learning, near optimal parameters can be selected after training within seconds. In contrast to other heuristic optimization algorithms like Genetic Algorithms these agents are able to generalize to new link conditions (e.g. change in span count), removing the need to retrain agents for every possible link scenario.