Machine Learning-Based Channel Analysis for User Concentric Optical Switching Networks

Abstract

Optical switching networks (OSN) rely on both hardware components and signals for performing efficient switching operations achieving service level requirements of the end-user. In this manuscript, an incremental learning-based wavelength assignment is introduced to minimize asynchronous channel selection and dispersion-free switch-over in OSN. This method accounts on the wavelength dispersion characteristics of the light path for establishing and reconnecting communications between end-to-end devices. The learning process segregates favorable and conflict channels based on blocking probability and channel capacity assigned to the optical users. A conventional wavelength division multiplexing technique is employed for improving data transmission rates. This method is appropriate for evading blocking rates in optical networks to improve throughput rate and to leverage OSN performance. Machine learning (ML) paradigm is designed for OSN leverages the communication rate influenced by channel conflicts and utilization capacity. Therefore, the number of transmissions likely requires optimal switch-over with the knowledge of channel constraints to improve the performance of OSN.