| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
Hi,
My environment is very dynamic and I know that RTABmap uses ICP to localize using the laserscans under Reg/Strategy = 1. So is it possible to always update the map where laserscans/camera scan the area and reuse the updated map at the next turn? So in that case, ICP will still be accurate under a dynamic environment.
So is it possible to always update the map where laserscans/camera scan the area and reuse the updated map at the next turn?
Yes, if --delete_db_on_start argument is not set, rtabmap re-uses the previous map and will update it. By default when restarting, the robot assumes that it is starting from the last position of the previous session. If the robot moved when offline, it will localize automatically only if the camera is used. With or without camera, we can still set the localization manually by sending a pose on /rtabmap/initialpose topic (similar to AMCL).
Under dynamic environment, it depends if most of the scan's rays hit static obstacles. I suggest to build rtabmap with libpointmatcher support to have more robust ICP in dynamic environments, particularly by tuning Icp/PMOutlierRatio parameter.
cheers,
Mathieu
It depends on which sensor you are using, because ray tracing is not enabled by default for camera input, just for 2D lidar. To turn on ray tracing (to clear obstacles), set Grid/RayTracing=true . After creating a map, if you restart in localization mode only, new obstacles won't be added in the static map. If you restart in SLAM mode, new obstacles would be added to the map.
Asked: 2019-12-03 16:44:40 -0600
Seen: 1,131 times
Last updated: Dec 04 '19
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
