Revision history [back]

Hi There, It is a good point to clarify. We use a map to estimate a drivable path in an environment. We can use either 2D or 3D map depending on the application what robot is designed for. If we use 3D motion planning (6 DOF), we definitely need a 3D map to avoid obstacles and reach the goal in a 3D space. However, in most wheel robot applications, a robot is driving on a 2D plane and a 2D map is sufficient to achieve navigation task. For those applications, laser scanners based mapping methods such as Gmapping can be used. In addition, if someone uses an RGB camera like Kinect for mapping, Gmaping may not give an accurate result since, RGBD sensors are low FoV and short-range sensors. But it gives an accurate result with RGBD mapping (3D). Therefore, people use an extracted 2D slice from a 3D map for 2D navigation with path planning algorithms such as A*. On the other hand, a 3D map provides details information on the environment. Also, it provides registered RGB images at each pose and they can be used for place recognition, and for other decision-making tasks based on the Visual information. Therefore, RGB sensors give more capabilities for 2D navigation.