Development of an autonomous navigation system in highly dynamic environment

[ES] La navegación autónoma de vehículos aéreos no tripulados en entornos complejos ha sido un tema de investigación activo en los últimos años. En particular, en entornos desconocidos y repletos de obstáculos que carecen de GPS, como bajo la cubierta forestal. Este proyecto pretende combinar técnicas de detección, cartografía y generación de trayectorias sin colisiones para facilitar la navegación autónoma en 3D de un UAV en un bosque, siendo consciente de cualquier obstáculo, como árboles, arbustos, ramas y vegetación del sotobosque. El objetivo de esta pila de navegación autónoma está orientado a la investigación forestal y a aplicaciones operativas en casos en los que las difíciles condiciones plantean un problema de accesibilidad tanto para las personas como para los vehículos terrestres, tales como la cartografía forestal para la gestión forestal sostenible o tareas de manipulación, entre otras.

[EN] Autonomous navigation of UAVs in complex environments has been an active research topic in recent years. In particular, GPS-denied unknown obstacle-rich environments such as under the forest canopy. This project aims to combine techniques for sensing, mapping, and collision-free trajectory generation to facilitate the 3D autonomous navigation of a UAV in a forest, being aware of any obstacles, such as trees, bushes, branches, and understory vegetation. The aim of this autonomous navigation stack is oriented to forest research and operational applications in cases where difficult conditions pose a problem of accessibility for both people and ground vehicles, such as forest mapping for sustainable forest management or manipulation tasks, among others.

[EN] his thesis explores the development and application of an autonomous navigation sys tem in dynamic environments. The primary focus is leveraging contact information to over come challenges such as GNSS­denied environments, sensor degradation, and other common failures, enabling collision recovery. The thesis begins by addressing the theoretical and technical challenges of autonomous navigation in obstacle­dense environments and the existing approaches to contact detec tion techniques. After comprehensively evaluating various contact detection alternatives, we propose a novel collision platform and contact detection algorithm integrated into an existing drone frame. These enhancements enable contact­based autonomous naviga tion capabilities in challenging environments. The designed collision system was evaluated independently and as part of an autonomous contact­based mission. The findings demonstrate the system’s remarkable ability to ac curately predict contact events, even integrated within an autonomous mission. This showcases the system’s versatility, performing well in static conditions and during real navigation in flight. Conclusively, this research highlights the potential of contact information to enhance au tonomous navigation significantly. The study suggests avenues for future research, par ticularly in improving collision platform robustness for more challenging environments and integrating additional sensor modalities to further enhance reliability.