| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
All,
Seemingly dumb question. If I have a SLAM map feeding the Navigation Stack, and therefore localising itself then what purpose does the localisation provided by AMCL provide? Just to localise within the costmaps?
Cheers
Mark
a SLAM algorithm will localize the robot at the same time that it creates the map, possibly changing parts of the maps as it perform loop closures, also the origin of the map will change each time you start the SLAM algorithm.
AMCL will localize the robot at a previously created map, which guarantees that no changes will be done in the map and you will have always the same origin, as it has been previously recorded.
So if I have an existing map, where the environment of which might change then SLAM so that it can localise then update as necessary?
it depends on your application and on how much it changes. In my experience AMCL does a good job even if the environment changes significantly. With SLAM you may find yourself degenerating a good map due to errors in odometry too much dynamic objects, for example.
You won't need AMCL at all in such a use case. The SLAM node will be broadcasting the transform from the map frame to the odometric frame. You may have to configure the navigation stack's global costmap a bit differently though.
Asked: 2016-10-29 10:11:51 -0500
Seen: 4,562 times
Last updated: Nov 14 '16
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.
