An Obstacle Avoidance Concept for Small Unmanned Rotorcraft in Urban Environments Using Stereo Vision

Abstract

Small Uninhabited Aerial Vehicles to be operated at low altitude in obstacle prone environments require an effective collision avoidance capability. An obstacle avoidance concept, consisting of a stereo camera sensor, real-time on-board image processing, and adaptive mission and trajectory planning is presented. When a mission is to be flown into an area with obstacles, the onboard mission planner first computes a flight trajectory that is based on a known obstacle data base. This is accomplished by using a Probabilistic Road Map (PRM) approach, which has been adopted for the use in helicopters. When flying the mission a stereo camera checks the flight path ahead by analyzing the depth images. An image processing algorithm reduces the complex high data rate images to a 3-dimensional obstacle map that is built by polygon shapes. This enables the system to constantly renew and update the existing world model by adding new obstacles as well as deleting obstacles from the data base if they have disappeared. Since the resulting world model has only a moderate complexity, close to real-time onboard mission planning and replanning using the aforementioned planning algorithms becomes possible. In case of an imminent collision with obstacles, the trajectory control mode is interrupted by a real-time collision avoidance behavior. A fully autonomous miniature helicopter has been used as a testbed to verify the the obstacle recognition part of the concept, while the planning and updating methods have been tested in simulation.