move_base and rtabmap

[aviadoz]

Hi,

I am doing some navigation tests in my lab. I am using rtabmap and move_base. I noticed that only when the "/rtabmap/localization_pose" publish data, move_base "heard" it and not take care about the data by /odom topic which I publish. For example, if the robot is in a place which there is a few frames that captured in the mapping stage (so it is more difficult to localized itself) the robot is rotate a lot until the "/rtabmap/localization" topic (I saw it in rviz) is lighting and the robot start driving. The same thing happen also when the robot arrives to its goal. In the odom msg the robot considered that it is arrive but move_base stop the robot only if rtabmap publish the pose. Is it okay?in the launch file move_base is subscribe to "odom" topic.

Thanks,
Aviad

[matlabbe]

Hi Aviad,

Normally "/rtabmap/localization_pose" is not used by move_base, only odom is used along with TF to get he pose of the robot. move_base would subscribe to rtabmap/grid_map topic for the global costmap.

Do you have a rqt_graph of the connections?

cheers,
Mathieu

[aviadoz]

Hi,

Thank you for your reply. I am adding the rqt node_graph.

rosgraph.dot

Thanks,
Aviad

[matlabbe]

Hi,

move_base seems correctly connected to rtabmap's map topic (through /rtabmap/grid_map topic):

however "/rtabmap/grid_map" doesn't seem to be connected to rtabmap?!? Is "n___rtabmap__rtabmap" the rtabmap node? normally, there should be a lot more topics around rtabmap node, here showing all of them:


what is your launch file?

[aviadoz]

Hi Mathieu,

I filtered rqt by /move_base node only.
I am uploading the whole graph.

Thanks
Aviadrosgraph.png

[matlabbe]

"/rtabmap/localization_pose" is not connected to move_base. If you are sending goals directly to "move_base_simple/goal", the behaviors (like rotating on place when the robot is stuck) you are seeing are coming from move_base, not rtabmap. The only thing that could influence move_base is if a relocalization is detected from rtabmap, the frame /map -> /odom TF will change, so move_base may have to replan a path. The /rtabmap/grid_map can also change, causing a replanning from move_base.