This paper presents an obstacle avoidance method that is performed with an unmanned helicopter. The approach begins with a mapping step where information from sensor data about previously unknown dangers is extracted into an occupancy grid and eventually converted into a polygonal 3D world model. This continuously updating map is used by a path planner that generates and updates a 3D trajectory guiding the vehicle through safe passages around the detected objects. The algorithms are generic but optimized for unmanned aircraft and a stereo camera as the environmental sensor. Computation is fully executed on board so that a ground control station is only needed for supervision. With successful obstacle detection and avoidance flight tests, the paper shows the qualification of the presented method under real operational conditions.

Mapping the environment is necessary for navigation in unknown areas with autonomous vehicles. In this context, a method to process depth images for occupancy grid mapping is developed. Input data are images with pixel-based distance information and the corresponding camera poses. A measurement model, focusing on stereo-based depth images and their characteristics, is presented. Since an enormous amount of range data must be processed, improvements like image pyramids are used so that the image analysis is possible in real-time. Output is a grid-based image interpretation for sensor fusion, i.e. a world-centric occupancy probability array containing information stored in a single image. Different approaches to draw pixel information into a grid map are presented and discussed in terms of accuracy and performance. As a final result, 3D occupancy grids from aerial image sequences are presented.

This paper presents a mapping process that can be used for autonomous applications like obstacle avoidance and trajectory planning. The process is real-time capable, and works in full 3-D environments. The mapping starts with building an occupancy grid out of sensor data. Within this grid, single objects are recognized and their polygonal shapes are calculated. The model for object shape representation is rather rough and uses only right prisms and horizontal floor planes. This makes the shape calculation very fast. The extracted objects are not complex so that an external application will be fast as well. As a special application, the approach is tested on an aerial vehicle using a stereo camera system.

This paper presents an approach to create three-dimensional occupancy maps from an aerial vehicle with stereo vision. The main idea is to create an occupancy grid that moves along with the vehicle and extract features into a fixed global map. Vice versa, global features or a-priori knowledge can be inserted into the grid. The maps are calculated onboard to be used for autonomous behavior like path planning and obstacle avoidance. With the described method, maps are created and updated in real-time, and due to its flexibility, the vehicle is not restricted to a pre-defined area. The developed approach has been demonstrated in flights with a small unmanned helicopter.