Terrain traversability in rescue environments

3D Terrain understanding and structure estimation is a crucial issue for robots navigating rescue scenarios. Large scale 3D point clouds, even if crisp and yielding a detailed representation of the scene, provide no information about what is ground, and what is top, what can be surmounted and what can be not, what can be crossed, and what is too deep to be traversed. In this work, we propose a new preliminary method for point cloud structuring, leading to the definition of a traversability map labeled with a cost that specifies how far is the considered region from a traversable one. The representation comes with a real-time algorithm that can be used for the safe navigation of a specific robot, according to its own limitations or constraints. Here, by robot constraints, we mean the length, height, weight of the robot, together with its kinematics constraints (in terms of ground mobility). We present results of the method with experiments taken on different scenarios, furthermore we illustrate the pros and contras of relying only on points cloud data set, without resorting to a surface reconstruction.