ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Navigation Stack with gmapping

asked 2011-03-15 11:46:53 -0500

jdt141 gravatar image

Has anyone had success getting the navigation stack to work with the gmapping package? If I understand it correctly, it appears that you could stop using the map_server and amcl, replace those with gmapping, and you could then perform autonomous mobility in an environment while you SLAM. Is that understanding correct? Do the local_costmap and global_costmap support this idea, and if so, what are the key parameters to configure this to happen?

edit retag flag offensive close merge delete

3 Answers

Sort by » oldest newest most voted

answered 2011-03-16 02:19:25 -0500

jdt141 gravatar image

I was able to get this to work by not running both amcl and map_server, and running gmapping. The only parameter that I had to change (from the defaults in the navigation tutorial) were to change the global_costmap to NOT have a static map (static_map: false) and to set the rolling_window to true (rolling_window: true) in the global_costmap_params.yaml file.

edit flag offensive delete link more


I tried to do as you said. I tried to set the goal point in rviz in unknown space and the robot wouldn't go there. Do I have to change some parameter in base_planer like global_base_planner and local_base_planner?
ParNurZeal gravatar image ParNurZeal  ( 2011-12-20 22:19:28 -0500 )edit
I think this is a wrong guideline. Even though it can work, the navigation stack will never receive the updated map. The static map should be set to true and the rolling window should be set to false.
ParNurZeal gravatar image ParNurZeal  ( 2012-01-13 13:12:24 -0500 )edit

answered 2011-03-22 08:20:23 -0500

eitan gravatar image

updated 2011-12-21 07:40:12 -0500

Eric Perko gravatar image

You shouldn't have to change the global_costmap parameters at all to run the navigation stack with gmapping, replacing AMCL and the map_server with gmapping should be sufficient from CTurtle onwards. For an example of the navigation stack running with gmapping in stage, you might want to check out the navigation_stage package here: . One thing to mention is that the navigation_stage package is configured for a Diamonback setup, though it'll probably work on CTurtle as well.

edit flag offensive delete link more


I know it's a little late, but: please delete the trailing dot from the navigation_stage link, so it works. Cheers.
tom gravatar image tom  ( 2011-12-15 19:06:41 -0500 )edit
Fixed the navigation_stage link.
Eric Perko gravatar image Eric Perko  ( 2011-12-21 07:40:29 -0500 )edit

answered 2011-04-05 04:17:09 -0500

You can find a fully functional demo in There you can find two gmapping-demo launch files for a real or a simulated (with gazebo) erratic robot.

edit flag offensive delete link more


I saw that in erratic launch file. Even in slam demo launch file, they still use static map. I thought if we wanna do slam, we should not use static map in global cost map. Is the launch file correct?
ParNurZeal gravatar image ParNurZeal  ( 2011-12-20 23:27:42 -0500 )edit
@Pablo Iñigo Blasco : The gazebo model is very good - Thank you !
Arkapravo gravatar image Arkapravo  ( 2011-12-24 04:40:07 -0500 )edit

@ParNurZeal: you misinterpret "static" here. It is a parameter to the costmap that means the map comes from the outside and has not to be build from the laserscanners. i.e. static is correct here for the planner, because it get's a finished map from gmapping and does not have to build its own one.

Achim gravatar image Achim  ( 2012-08-30 22:33:32 -0500 )edit

Question Tools



Asked: 2011-03-15 11:46:53 -0500

Seen: 5,236 times

Last updated: Dec 21 '11