Sensor-based control for UAV navigation in GPS challenged environments

Abstract

Over the past years Unmanned Aerial Vehicles (UAVs) have become one of the fastest growing topics for robotics, both in academia and in consumer applications. Nowadays, UAVs are being used to serve a large amount of functions over a wide spread of fields such as military operations, defense, delivery, surveillance amongst others. However, the challenges that these vehicles present are of great complexity to the academic community. Control and navigation strategies are of crucial importance for the development of successful UAV platforms. The kind of sensors that are included on a UAV and their characteristics can greatly affect its performance. One of the main issues for UAV navigation is that it relies on GPS readings. However, GPS signals are not always available, especially in indoor areas. The loss of GPS connection during flight could have catastrophic consequences to the platform This project looks to implement a sensor-based UAV control system that can perform navigation tasks safely without GPS data. This project will be based mainly in the use of Monocular Vision for navigation. Finally, the main objective of this research is to develop safety measures in case of GPS data connection loss by implementing a system capable of maintaining stable hovering or performing controlled auto-landing based on vision pose estimation and sensor fusion.