The development of ROS-based offboard algorithms for autonomous UAVs intended for Mars exploration

Unmanned Aerial Vehicles (UAVs) have received increased interest in the field of mobile robotics. Recently, research in the field of UAVs is also studying their use in space applications and planetary exploration. An example is Ingenuity, the first drone on Mars developed by NASA. This is because a UAV offers a wider area of operation with respect to an unmanned ground vehicle (UGV), and it provides the mission with a higher resolution regarding images if compared to a satellite or an orbiter. Moreover, Software in the loop (SITL) simulation of such vehicles is preferred during the development phase since it enables to evaluate the algorithms implemented without making use of a real vehicle and without the risk of damaging the hardware. This thesis proposes a set of algorithms related to UAV-based planetary exploration. Such software implementations range from the more straightforward functionalities, related to the navigation and off-board control of the drone to more complex ones which also incorporate on-board sensors, such as the available ventral camera, to provide imagery data of an object by encircling it through a circular trajectory. Additionally, an area coverage algorithm is used to map an area of the terrain by following a grid sweep path, while providing images of the underlying terrain. Each of these algorithms is composed of a path generation functionality and subsequent, through the usage of a PID controller, trajectory tracking capability. All the proposed algorithms have been developed with Python. The drone off-board control software implementation has been developed through the Robotics operating system (ROS), more specifically the ROS2 version, and the PX4 autopilot. They are designed to work with an external Linux computer along with the firmware available on board the drone, without needing to resort to the UAV computational resources. Each of the proposed navigational features are organized according to a request/reply model available through the interface offered by ROS. Besides, the UAV has been equipped with a precision landing algorithm making it capable of re-entering on top of a dedicated platform, in this thesis case a UGV. Moreover, the drone is supplied with a ROS2 node implementing a Kalman filter algorithm that produces the relative position and velocity estimates between the drone and the UGV landing platform, from sensor readings. The available sensors are the ones already implemented via the PX4 autopilot flight stack, furthermore ultrawide-band antennas and a top-view camera for AprilTag detection are added to the simulation environment to increase the estimate precision. Every sensor combination available on the estimation algorithm is tested in the simulation environment, to check their effectiveness and interaction. Finally, the simulation environment has been developed through ROS and the Gazebo simulator, it provides with the condition simulating the drone itself, a rover, and a reproduction of the Martian ground. The testing phase of the algorithms has been done by following the Software-in-the-Loop (SITL) approach. The drone model available through the PX4 autopilot has been used as the starting point and modified according to the estimation algorithm needs by adding the necessary sensors via Gazebo plugins.