Modelling and mapping of terrain characteristics for situational awareness
One of the problems that an autonomous machine needs to solve, to be able to safety navigate a terrain, is to define where it can drive through. In the literature, terms like traversability, trafficability, and obstacle maps are used. In this report, we discussed first a typical pipeline of generating traversability costs maps. A subset of on-board sensor data (e.g. lidar) and information from third party (e.g. digital elevation maps) can be used to build a world model. We also discussed several different world models, such as, elevation maps and its variations. These models are then converted to traversability cost maps. However, we showed that this is a rather limited view of the traversability, because there are not clear guidelines how to perform the latter transformation.
The report will be present the level of technology and research and it focuses on on-board sensors (both proprioceptive and exteroceptive), also giving touch on open data, third party data and multi-modal data fusion.