Autonomous Precision Landing for UAVs on a Mars-like environment in ROS/Gazebo

Space exploration and UAVs have been undergoing exponential growth in recent years, and with them the related technologies deployed. More than a hundred space launches are planned for 2023, and it is expected that UAVs will be increasingly associated with these types of missions. The process leading to their success is very complex and expensive, both in terms of economics and schedule, and for this reason, simulators are often employed within the implementation process. As a result of these, it is possible to test algorithms without causing damage to the physical hardware, extend safety and operability, and definitely have a less environmental impact. The work in this thesis aims to create a simulation environment to reproduce a possible Mars mission performed through the use of two robots: a rover and a UAV. Specifically, this work will mainly focus on implementing and simulating a precision landing algorithm of the UAV on the rover, exploiting technologies that could potentially be deployed on the red planet. This has been achieved first in simulation using the Software In The Loop (SITL) framework, and later, the same configuration has been implemented and tested on real hardware, to validate the development precision landing approach in reality. A comparison of current technologies mainly adopted for this purpose has been made, after which the combination of UWB and AprilTag was chosen as the solution. The graphical user interface has been created in Gazebo, while the Robot Operating System (ROS) has been used to develop the software architecture. Gazebo and ROS are used to generate a simulation environment with a rover (including an AprilTag on it) and drone on the Martian ground, as well as to estimate the relative position between drone and rover using a Kalman filter, and the last part responsible for the robots control algorithms. The precision landing algorithm was tested and analyzed in three different situations: using the UWB only, using the UWB and the AprilTag as inputs to the Kalman filter for more precise position estimation, and, finally, using the UWB and the AprilTag as inputs in the control part. All the three configurations share the same control strategy consisting of two phases. First, a Proportional (P) controller is used when the UAV is far from the target. Then, within a certain distance, a more performing PID (Proportional-Integrative-Derivative) controller is used to get the drone closer to the rover. In the last configuration, during the approach to the ground touch phase, the drone will no longer use data from the UWB, but only data from the AprilTag that will be clearly visible from the camera. In order to perform experimental tests, the drone has been adapted to install the RaspiCam used for the precision landing. The CAD of a support for that camera has been created and, then, 3D printed using an FDM technique so that it would be as efficient as possible in terms of weight, strength, supports, and interface. Subsequently, the RaspiCam has been calibrated, and it has been tested first that the AprilTag detection was working properly, and then that the presented algorithm was efficacious by performing tests on a drone in flight.