Ground Vehicle Navigation through Traversability Analysis of Outdoor Environments

This thesis describes the research activity carried out on the navigation of unmanned ground vehicles in outdoor unstructured environments. These environments are very common in a wide variety of real-application scenarios, such as search and rescue for post-disaster response, monitoring of the environment and industrial facilities, precision farming, planetary exploration etc.. For this reason, robotics researchers have been investigating such kind of environments for decades. However, the problem of vehicle navigation in these scenarios is not fully solved. In fact, a general approach for ground vehicle motion planning, taking into account both the robot and the environment features, is still far to be defined. In the literature, this problem has been addressed through the so-called traversability analysis. It can be seen as an assessment of the difficulty for a specific ground vehicle, characterized by its own locomotion features, to cross a terrain area, which is in turn characterized by its own morphology and appearance. The works reported in this thesis are related to the problem of geometry-based traversability analysis. It consists in deriving maps of traversing costs from three-dimensional models of the environment. These costmaps are extremely useful in challenging environments, as they are used in robot motion planning to avoid unsafe paths for the vehicle itself. Still in the context of 3D reconstructions, a solution for the remote drive of mobile robotic platforms has been developed. The aim has been to enhance the operator interface with helpful information, including traversing costs, via graphical elements presented in a mixed reality context. Finally, coverage path planning for unmanned aerial vehicles has been investigated as well. It is a specific kind of path planning related to three-dimensional photogrammetric reconstruction. In particular, an approach to manage a flock of aerial vehicles has been developed, in order to parallelize the coverage mission. The coverage task is the first step needed for the environment reconstruction. Eventually the traversability analysis can be performed on the obtained reconstruction. Experimental and on-field trials have also been performed on a real mobile robotic platform, for the testing of the proposed approaches. The results achieved are reported and widely discussed.