Efficient Safe Trajectory Planning for an Omnidirectional Drone

The computation of a safe path between two target points for an omnidirectional drone is considered. The drone is equipped with eight fixed unidirectional thrusters, enabling full pose control. Given a priori knowledge of the environment map, a preliminary path is generated using a variant of RRT^*. A corresponding trajectory is then fitted to this path and executed by the drone. To enhance safety and efficiency, the velocity along the path is adaptively assigned, balancing caution near critical obstacles with minimizing travel time in open spaces. The adaptive velocity limits are proportional to the distance between the drone's convex hull at various poses along the path and the surrounding environment. Distance computation is optimized by focusing on obstacles in the direction of motion, representing obstacle facets with axis-aligned voxels, and pruning distant obstacles before calculating the exact distance between the convex hull and the environment map. The proposed approach is validated through simulation studies, demonstrating effective navigation through narrow gaps at odd angles while ensuring minimal travel time and maintaining required safety margins.