This paper proposes a safe 3D path planner for an urban operation of VTOL-type UAV. One of the difficulties of navigating an UAV in an urban environment involves its degraded self-localization accuracy due to GPS signal occlusion. In order to keep UAV navigability in such a GPS-denied area, alternative navigation systems (mostly using visual sensing) have been proposed in many literatures. This paper suggests incorporating different means of self-localization and their position-dependent availabilities in path planning task. Localization uncertainty corridor is defined by a space swept by the localization error ellipsoid evolving along the planned path. The proposed path planner applies classical A~*/Theta~* algorithms and sampling-based RRT~* algorithm to a 4-dimensional search space (3D Euclidean space + localization modes) to find a path which minimizes a volume of the uncertainty corridor. In addition, for each path segment, collision check between the uncertainty corridor and obstacles is performed to confirm UAV traversability. The proposed 3D safe path planner is firstly tested through simulations with obstacle configurations defined in the UAV obstacle field navigation benchmark. Then, some of the planned paths are executed by the ONERA ReSSAC UAV platform and resulting trajectories are evaluated with the benchmark metrics.
展开▼
