Global Motion Planning for Unmanned Aerial Vehicle Automation

Global motion planning is a critical problem to be explored in the context of autonomous UAVs. It involves two main objectives: finding an optimal path between a source and destination, and effectively avoiding obstacles during autonomous missions. This involves both path planning and trajectory planning, although some research proposals solely concentrate on defining control policies for autonomous obstacle avoidance. This paper presents a technique for global motion planning to address path planning and trajectory planning sequentially. The mission path is planned after defining no-fly zones using an extended version of QGroundControl developed specifically for this purpose. The RRT* algorithm from the OMPL is then employed to generate a path that effectively avoids the defined obstacles. The flight path is continually updated whenever an obstacle is detected to ensure UAV safety during autonomous missions. This update process incorporates the initial mission path, the UAV's current GPS coordinates, and the output from the obstacle detector. It relies on machine learning models for obstacle detection and collision prediction, and involves generating waypoints to avoid obstacles while following the initial flight path. The proposed technique has been tested in the Gazebo simulator and appears to be suitable for autonomous missions in outdoor environments.