Our neighboring planet Mars has been a primary target for exploration since the beginning of spaceflight and is today the celestial body with the highest number of exploration missions. However, due to large signal delays and demanding environmental conditions, missions such as NASA’s Curiosity rover have only been able to explore the local surroundings within a few kilometers of their landing sites. Moreover, scientifically highly interesting regions, such as the Valles Marineris, have not yet been explored in situ by rovers or probes in the search for e.g. water and traces of life. To reach such environments, greater autonomy and robustness in robotic exploration systems are required. Concepts such as the DLR’s Valles Marineris Explorer (VaMEx) propose scout Unmanned Aerial Vehicles (UAVs) to assist autonomous rovers in identifying regions of interest and traversable terrain. The Ingenuity UAV demonstrated the first proof of concept, operating with the Perseverance rover, while future concepts aim for ranges of 40 km or more for scout UAVs. This thesis presents a possible navigation system suitable for such a future Mars long-range scout UAV, capable to robustly fusing different sensor inputs while following a commercial-off-the-shelf (COTS) design approach inspired by the success of Ingenuity. The main effort is based on an implementation of a loosely coupled Error State Extended Kalman Filter (ES-EKF) algorithm that adheres to the COTS philosophy and targets a small computational footprint by relying on non-GNSS sensor inputs. Extensive tests in simulation as well as analogue missions are presented. The navigation system design proposal is accompanied by a novel method for deriving sensor reliability requirements in very early phases of system development by inverting classical Fault Tree Analysis (FTA). Developing sensor fusion algorithms often requires time- and resource-intensive parameter tuning. Several works have already proposed solutions using dedicated Bayesian Optimization (BO)-based tuning algorithms. However, such solutions are often tailored to specific tuning problems and use cases. Therefore, as a second main work of this thesis, a BO parameter tuning approach, based on a state of the art framework (following the COTS approach), is introduced. Different settings, objective functions and parameter selections are derived and evaluated to provide a simple-to-use yet versatile tool for parameter tuning. The thesis concludes with a focus on the usability and validation of simulation environments for developing new robotic navigation solutions, followed by a conclusion and outlook for future tasks and challenges for long-range distance navigation with UAVs on Mars and other GNSS-denied environments.     «

Our neighboring planet Mars has been a primary target for exploration since the beginning of spaceflight and is today the celestial body with the highest number of exploration missions. However, due to large signal delays and demanding environmental conditions, missions such as NASA’s Curiosity rover have only been able to explore the local surroundings within a few kilometers of their landing sites. Moreover, scientifically highly interesting regions, such as the Valles Marineris, have not yet...     »