Use GMapping nodelet #158
Use GMapping nodelet #158
Conversation
Modified GMapping launch files to use the new nodelet version of gmapping instead of the older node. This allows the GMapping nodelet (which reads /scan) to be run in the same nodelet manager as depthimage_to_laserscan (which outputs /scan), allowing for more efficient data processing.
|
The above-referenced PR in the slam_gmapping package has been merged, so this can now be tested and merged if approved. |
|
@tfoote -- any feedback on this? Given it's a launch file change only, I wonder why the kinetic build is failing. |
|
The kinetic build is failing due to some CMake warnings, which were not introduced by this PR. I will try to test this sometime in the next couple days. |
| @@ -3,7 +3,7 @@ | |||
| <arg name="base_frame" default="base_footprint"/> | |||
| <arg name="odom_frame" default="odom"/> | |||
|
|
|||
| <node pkg="gmapping" type="slam_gmapping" name="slam_gmapping" output="screen"> | |||
| <node pkg="nodelet" type="nodelet" name="slam_gmapping" args="load SlamGMappingNodelet camera/camera_nodelet_manager" output="screen"> | |||
There was a problem hiding this comment.
People use the turtlebot gmapping with Lidars, in which case they would not have the camera nodelet manager.
I am not entirely sure that it is a good idea to make loading the nodelet the default.
I think it would be better to make a separate laser gmapping launch, that does not use the nodelet, before merging,
|
@tfoote |
Modified GMapping launch files to use the new nodelet
version of gmapping instead of the older node. This
allows the GMapping nodelet (which reads /scan) to be
run in the same nodelet manager as depthimage_to_laserscan
(which outputs /scan), allowing for more efficient
data processing.
Note that this was done in conjunction with a forthcoming change to the gmapping package (pull request: ros-perception/slam_gmapping#41), and serves as an example of the two nodelets running together under the same nodelet manager. Initial testing showed a ~11% reduction in overall memory usage and ~20% CPU usage running the slam_gmapping nodelet and depthimage_to_laserscan nodelet together, as compared to running slam_gmapping in a separate node. This PR is thus dependent on the above-linked PR, and needs that to land before it can be merged in.