How to improve amcl performance?

Hi all,

I made a simulated environment and a robot in Gazebo and I'm trying to localize the robot on a map with amcl.

So, what I did so far was to:

launch the simulation environment
run the map_server with the available map
run amcl
open rviz to see the result

and this is what happens:

As soon as the robot turns the pose estimate is very bad. Can please someone give me some hint on how to get this working?

Thanks.

Asked by schizzz8 on 2018-06-22 09:20:07 UTC

Comments

This looks to me like potentially 2 things: 1, that your URDF with your laser might not be correctly configured, so that when you turn, it offset by a bad amount. 2, simulation environments have issues with AMCL due to lack of unique features from its "perfectness" if you give us your urdf/configs..

Asked by stevemacenski on 2018-06-23 00:40:21 UTC

... we might be able to help narrow down more

Asked by stevemacenski on 2018-06-23 00:40:28 UTC

sure! all the config files are in this package (https://github.com/schizzz8/lucrezio_simulation_environments) to test it run: roslaunch lucrezio_simulation_environments empty_world_with_apartment_and_robot.launch

Asked by schizzz8 on 2018-06-23 03:37:31 UTC

@stevemacenski any update on this?

Asked by schizzz8 on 2020-05-28 06:14:42 UTC