This is an old revision of the document!
Contact: Vladyslav Usenko.
In this work we present a real-time approach for local trajectory replanning for MAVs. Current trajectory generation methods for quadrocopters achieve high success rates in cluttered environments, but assume static environments and prior knowledge of the map. In our work we utilize the results of such planners and extend them with local replanning algorithms that can handle unmodeled (possibly dynamic) obstacles while keeping MAV close to the global trajectory. To make our approach real-time capable we maintain information about the environment around MAV in a occupancy grid stored in 3D circular buffer that moves together with a drone, and represent the trajectories using uniform B-splines that ensures that trajectory is sufficiently smooth, at the same time allowing for efficient optimization.
For source code with simulation examples visit the Github repository (coming soon).