ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
1 | initial version |
It will be easier to answer to question if you have a database to share on some tests you tried. The 1 FPS is the update of the map, not the odometry. It is recommended to run odometry at 30 FPS or more. The 1 FPS is because unless you move very fast, we don't need to keep all 30 HZ frames (which would have a lot of overlap anyway). With Kinect for Azure, I've seen better visual odometry using IR image + Depth image, which has a larger field of view than the RGB camera. It is also possible to do only ICP odometry too, which gives quite accurate results like in this post. The point cloud can be regenerated at the the resolution you want as rtabmap keeps all raw data. For detecting corners and make a nice 2D floorplan from the point cloud, rtabmap cannot do that (only occupancy grids but it seems not what you want to do).