Thesis Diseño de un simulador PON para la optimización de DBA mediante aprendizaje por refuerzo
Loading...
Date
2025-12
Authors
Journal Title
Journal ISSN
Volume Title
Program
Ingeniería Civil Telemática
Departament
Campus
Campus Casa Central Valparaíso
Abstract
Actualmente existe un aumento masivo en la demanda de ancho de banda, impulsado por servicios exigentes como el streaming en alta resolución y los juegos en la nube. Las redes PON son la infraestructura estándar para soportar este tráfico, pero sus mecanismos de control (DBA) deben evolucionar para cumplir con requisitos de calidad cada vez más estrictos, integrando desde soluciones heurísticas clásicas hasta nuevas técnicas de Inteligencia Artificial. El problema principal es que investigadores y empresas carecen de herramientas de simulación flexibles para validar estas nuevas estrategias. La mayoría de los simuladores actuales son cerrados o difíciles de integrar con tecnologías modernas, lo que impide probar y comparar ágilmente distintos algoritmos de gestión de red. Para cubrir esta necesidad, este trabajo presenta el diseño y validación de PonLab, un simulador de redes PON de código abierto escrito en Python. Esta herramienta funciona como un banco de pruebas modular, diseñado para que académicos y desarrolladores puedan implementar sus propios algoritmos, ya sean heurísticos, SDN o basados en Aprendizaje por Refuerzo. Su arquitectura híbrida permite reducir la carga computacional sin perder precisión, ofreciendo un entorno unificado para corroborar el comportamiento de la red bajo distintos enfoques.
Currently, there is a massive surge in bandwidth demand, driven by demanding services such as high-resolution streaming and cloud gaming. Passive Optical Networks (PON) serve as the standard infrastructure to support this traffic; however, their control mechanisms (DBA) must evolve to meet increasingly strict quality requirements, integrating everything from classic heuristic solutions to new Artificial Intelligence techniques. The primary challenge is that researchers and industry professionals lack flexible simulation tools to validate these new strategies. Most current simulators are closedsource or difficult to integrate with modern technologies, hindering the agile testing and comparison of different network management algorithms. To address this need, this work presents the design and validation of PonLab, an open-source PON simulator written in Python. This tool serves as a modular testbed, designed to enable academics and developers to implement their own algorithms, whether heuristic, SDN, or Reinforcement Learning-ased. Its hybrid architecture reduces computational overhead without compromising precision, offering a unified environment to corroborate network behavior under different approaches.
Currently, there is a massive surge in bandwidth demand, driven by demanding services such as high-resolution streaming and cloud gaming. Passive Optical Networks (PON) serve as the standard infrastructure to support this traffic; however, their control mechanisms (DBA) must evolve to meet increasingly strict quality requirements, integrating everything from classic heuristic solutions to new Artificial Intelligence techniques. The primary challenge is that researchers and industry professionals lack flexible simulation tools to validate these new strategies. Most current simulators are closedsource or difficult to integrate with modern technologies, hindering the agile testing and comparison of different network management algorithms. To address this need, this work presents the design and validation of PonLab, an open-source PON simulator written in Python. This tool serves as a modular testbed, designed to enable academics and developers to implement their own algorithms, whether heuristic, SDN, or Reinforcement Learning-ased. Its hybrid architecture reduces computational overhead without compromising precision, offering a unified environment to corroborate network behavior under different approaches.
Description
Keywords
Redes ópticas pasivas (PON), Asignación dinámica de ancho de banda (DBA), Simulación de redes, Código abierto, Aprendizaje por refuerzo (RL), Python
