Enhanced resource allocation in elastic optical network using deep learning and optimization process

Abstract

The elastic optical network offers several advantages in bandwidth allocation compared to traditional fixed-grid optical networks. These advantages stem from their ability to flexibly and efficiently allocate resources, meeting modern communication networks’ dynamic and diverse demands. It is crucial to handle dynamic traffic loads and proactively manage the resources in an elastic optical network with a productive technique. Deep learning is an effective tool for complex data analysis and real-time decision-making. We address a model that integrates two deep neural networks: generative adversarial network (GAN) for data augmentation; and echo state network (ESN) for network’s requirement prediction. Furthermore, an optimization process is carried out for efficient spectrum allocation. The GAN provides a considerable and reliable quantity of data necessary to train the ESN model that could provide the desired output. The ESN model is further enhanced by optimizing the essential parameters, enabling it to learn diverse traffic patterns and anticipate unusual situations. By using a GAN-ESN approach, there is a substantial benefit in reducing latency, saving energy, and optimizing bandwidth allocation. The simulation results confirm that the proposed scheme can significantly improve the performance of resource management and achieve a high degree of fairness(95%accuracy) in the evaluation metrics.