Hierarchical Planning with 3D Local Multiresolution Obstacle Avoidance for Micro Aerial Vehicles

Abstract:
Micro aerial vehicles (MAVs), such as multicopters, are particular well suited for the inspection of human-built structures, e. g., for maintenance or disaster management. Today, the operation of MAVs in the close vicinity of these structures requires a human operator to remotely control the vehicle. For fully autonomous operation, a detailed model of the environment is essential. Building such a model by means of autonomous exploration is time consuming and delays the execution of the main mission. In many real-world applications, a coarse model of the environment already exists and can be used for high-level planning. Nevertheless, detailed obstacle maps, needed for safe navigation, are often not available. We employ the coarse information for global mission and path planning and refine the path on the fly, whenever the vehicle can acquire information with its onboard sensors. To allow for fast replanning during the flight, we present a 3D local multiresolution path planning approach making online grid-based planning for our MAV platform tractable.