3D Indoor Mapping for Micro-UAVs Using Hybrid Range Finders and Multi-Volume Occupancy Grids

Autonomous micro-UAV navigation requires techniques that allow for the accurate mapping of unstructured 3D environments such as stairwells, tunnels, and caves. As a step towards that goal, this paper presents a system to build three-dimensional maps of rectilinear environments, where the horizontal cross-section of the world is invariant at different heights. The assumption that the environment is structured makes our approach suitable for mapping of indoor spaces. One of the largest challenges in 3D mapping is correctly estimating the full 6-DOF pose of the vehicle. We present an approach that estimates the pose by fusing the information of an altimeter, an IMU, and a horizontally-mounted laser range-finder. A key step in the estimation is an orthogonal projection of the laser scan data, allowing for accurate scan matching in 2D. We also propose a novel map data structure called a Multi-Volume Occupancy Grid. MVOGs explicitly store information about both obstacles and free space. This allows us to correct previous potentially erroneous sensor readings by incrementally fusing in new sensor information. In turn, this enables extracting more reliable probabilistic information about the occupancy of 3D space. Observations are grouped together into continuous vertical volumes, which makes this new data structure considerably more space-efficient than point cloud or voxel-grid representations.