“…The robot needs accurate data from sensors to build a model of the environment. This model (map) can be used to adapt the gait to the terrain profile (Walas and Belter, 2011), plan the motion (Zucker et al, 2010) and localize the robot (Yoon et al, 2013), and should thus be designed taking into account the real-time operation of the entire navigation system. Among many existing representations of the environment (raw point cloud, occupancy map, 2.5D elevation map, triangle mesh, octree (Hornung et al, 2013), multi-level surface (MLS) maps (Pfaff et al, 2007)), we work with 2.5D elevation maps.…”