Machine Learning as a Crucial Enabler for the Future Fiber-Optic Communication Systems

• Grant number: 24/01543-2
• Support Opportunities: Scholarships abroad - Research
• Start date until: July 15, 2026
• End date until: July 14, 2027
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Helder May Nunes da Silva Oliveira
• Grantee: Helder May Nunes da Silva Oliveira
• Host Investigator: Alejandra Beghelli
• Host Institution: Centro de Engenharia, Modelagem e Ciências Sociais Aplicadas (CECS). Universidade Federal do ABC (UFABC). Ministério da Educação (Brasil). Santo André , SP, Brazil
• Institution abroad: University College London (UCL), England

Abstract

The continued growth of Internet traffic is driving telecom operators to provide robust, cost-effective, and performance-adaptive connectivity solutions to meet the Quality of Service requirements of applications. As a result, the risk of a capacity bottleneck is growing, which means that Internet technologies will not be able to satisfy the ever-growing demand for capacity. Optical networks are crucial for transmitting most internet traffic, as they offer the possibility of transporting data at high bit rates over distances ranging from national to global dimensions. For the projected growth to be feasible, the cost per bit of fiber transmission systems must be reduced, which precludes the installation of more fiber cables, as this is costly and disruptive. Therefore, increasing the capacity that optical networks can transmit is crucial to mitigate the risk of a capacity bottleneck. Since optical networks generate a massive amount of data, it is natural to consider applying data-driven techniques to achieve this. Machine learning (ML) is a promising way to unlock additional capacity in optical networks that has been extensively explored in the literature. In this context, the crucial problem addressed by this research project is the application of Deep Reinforcement Learning (DRL), to enhance the performance and adaptability of Multi-Mode Fiber (MMF) networks. This project continues the research efforts we started in the previous FAPESP project #20/05054-5 and is closely linked to project #22/07488-8.