Machine learning applications for advanced optical networks

• Grant number: 22/07488-8
• Support Opportunities: Regular Research Grants
• Duration: November 01, 2022 - October 31, 2024
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Convênio/Acordo: CONFAP - National Council of State Research Support Foundations
• Principal Investigator: Darli Augusto de Arruda Mello
• Grantee: Darli Augusto de Arruda Mello
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated researchers: Daniel Augusto Ribeiro Chaves ; Gustavo Sousa Pavani ; Helder May Nunes da Silva Oliveira ; Helio Waldman ; Joaquim Ferreira Martins Filho ; Leonardo Didier Coelho ; Nelson Luis Saldanha da Fonseca ; Raul Camelo de Andrade Almeida Júnior

Abstract

The new digital services that have become popular in recent years, such as video streaming, cloud computing, and group video calling, have generated a significant increase in the demand for data transmission. Transport networks, which enable channel flows at high transmission rates between different regions of a city or between cities, are especially impacted. To meet this demand, it is necessary to expand the infrastructure based on optical fibers, the only technology capable of meeting transmission rates above 1Tbps over long distances, as well as making better use of existing infrastructures. There are great challenges related to the complexity of problems involving the different strategies and technologies to achieve these high transmission rates. It is considered to use other transmission bands in the optical domain, avoiding the waste of band in legacy systems, and the deployment of Space-Division Multiplexing (SDM). There are several possibilities for expanding network capacity, respecting the required level of resilience and transmission quality to ensure high rates, either by using alternative transmission bands or by implementing advanced optical fibers that support parallel spatial channels. In this scenario, machine learning models emerge as a viable alternative to deal with the growing complexity of optical networks and communication systems. In fact, they appear as important tools for the most diverse tasks, such as device modeling, parameter estimation, resource allocation and degradation mitigation. This project aims to bring together researchers from São Paulo and Pernambuco, who already work independently in the area, in a synergistic environment of collaboration. It is expected, therefore, to incorporate new machine learning techniques to optical communication networks, with special attention to elastic SDM and multiband networks, resulting in more efficient, sustainable systems with greater transmission capacity. (AU)