Use this link to cite:
http://hdl.handle.net/2183/39526 Vuelo Inteligente: Aplicación de Machine Learning en Drones Cuadricópteros
Loading...
Identifiers
Publication date
Authors
Rey Barreiros, Abel
Advisors
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: Este trabajo fin de grado desarrolla una plataforma de vuelo autónoma y modular que permite configurar misiones y probar algoritmos de Machine Learning en un entorno real. Compuesta por una Pixhawk 2.4.8 y una Raspberry Pi 4B, la plataforma transforma el dron en un laboratorio portátil, superando las limitaciones de las simulaciones teóricas. La integración de la Google Coral es esencial, ya que acelera la inferencia de los algoritmos ejecutados en la Raspberry Pi. Aunque la inferencia en tiempo real durante el vuelo no se ha implementado en esta versión, el sistema está preparado para hacerlo en el futuro con conectividad 4G LTE. El control de vuelo, gestionado por DroneKit, permite la ejecución de misiones, incluyendo el retorno automático al punto de lanzamiento. Además, se ha implementado un sistema de detección de objetos basado en YOLO optimizado para la Google Coral, facilitando la segmentación y anotación en tiempo real. El código del proyecto, liberado bajo la licencia Mozilla Public License 2.0 (MPL-2.0), está disponible en el repositorio público [1]. Este trabajo no solo facilita la experimentación práctica con algoritmos en entornos reales, sino que también establece una base para futuras aplicaciones que requieran procesamiento avanzado en tiempo real.
[Abstract]: This Final Degree Project develops an autonomous and modular flight platform that allows configuring missions and testing Machine Learning algorithms in a real environment. Comprising a Pixhawk 2.4.8 and a Raspberry Pi 4B, the platform transforms the drone into a portable laboratory, overcoming the limitations of theoretical simulations. The integration of the Google Coral is essential, as it accelerates the inference of algorithms running on the Raspberry Pi. Although real-time inference during flight has not been implemented in this version, the system is prepared to do so in the future with 4G LTE connectivity. The flight control, managed by DroneKit, enables mission execution, including automatic return to launch. Additionally, an object detection system based on YOLO optimized for Google Coral has been implemented, facilitating real-time segmentation and annotation. The project’s code, released under the Mozilla Public License 2.0 (MPL-2.0), is available in the public repository [1]. This work not only facilitates practical experimentation with algorithms in real environments but also lays the foundation for future applications requiring advanced real-time processing.
[Abstract]: This Final Degree Project develops an autonomous and modular flight platform that allows configuring missions and testing Machine Learning algorithms in a real environment. Comprising a Pixhawk 2.4.8 and a Raspberry Pi 4B, the platform transforms the drone into a portable laboratory, overcoming the limitations of theoretical simulations. The integration of the Google Coral is essential, as it accelerates the inference of algorithms running on the Raspberry Pi. Although real-time inference during flight has not been implemented in this version, the system is prepared to do so in the future with 4G LTE connectivity. The flight control, managed by DroneKit, enables mission execution, including automatic return to launch. Additionally, an object detection system based on YOLO optimized for Google Coral has been implemented, facilitating real-time segmentation and annotation. The project’s code, released under the Mozilla Public License 2.0 (MPL-2.0), is available in the public repository [1]. This work not only facilitates practical experimentation with algorithms in real environments but also lays the foundation for future applications requiring advanced real-time processing.
Description
Editor version
Rights
Atribución 3.0 España







