Traversable ground detection based on geometric-featured voxel map

The process of finding traversable ground is important to autonomous vehicles. To achieve this, 3D maps that contain the information about vehicle´s surroundings are generally used. One of the simple methods to represent 3D maps is to use 3D point clouds collected by range sensors. But, it requires lots of memory and computation time to deal with point clouds. To solve these problems, we propose a novel traversable ground detection method using a 3D mapping algorithm. The proposed mapping algorithm includes plentiful geometric information based on voxels, which is called Geometric-Featured Voxel (GFV) maps. In our experiments, point clouds collected in urban environments are tested to evaluate performance of the proposed algorithm.