Ask Your Question

Robot location jump in RVIZ window

asked 2016-09-13 03:01:11 -0500

samping gravatar image

Hi everyone,

I have a robot positioning problem in my simulation.

In the simulation my robot will suddenly jump place near the it`s original location. Some time it will recover by itself.

And some time it just stay there, which lead to problem of collision of something even worse.

I cannot figure out which part I have done wrong.

I will put the video of my error in below, and I can provide other detail as request.

you might skip to 40second to see what happen in my rviz window

edit retag flag offensive close merge delete


Where does the position information for the RVIZ comes from? From a localization method? Could you provide more information on what you're doing in the simulation?

vmatos gravatar image vmatos  ( 2016-09-13 03:38:31 -0500 )edit

I am currently running a simulation using teb_local_planner.

The following url is the paste bin for the .yaml file

samping gravatar image samping  ( 2016-09-13 04:04:08 -0500 )edit

Ok, What are the red arrows? Where does that information come from?

vmatos gravatar image vmatos  ( 2016-09-13 04:36:31 -0500 )edit

there is some issue with localisation... provide more information on the launch file

non-conventional_stuff gravatar image non-conventional_stuff  ( 2016-09-13 04:46:09 -0500 )edit

The red arrows is the data from amcl.

Yes sure, the following url is the paste bin for my launch code, and there is a command after this comment

samping gravatar image samping  ( 2016-09-13 23:46:53 -0500 )edit

roslaunch teb_tut robot_carlike_in_stage1.launch map_file:="/home/samping/Desktop/teb_tut/maps/tutorial.yaml" world_file:="/home/samping/Desktop/teb_tut/stage/" initial_pose_x:=5.0 initial_pose_y:=0.0 initial_pose_a:=0.0

samping gravatar image samping  ( 2016-09-13 23:47:02 -0500 )edit

2 Answers

Sort by » oldest newest most voted

answered 2016-09-14 06:35:02 -0500

Procópio gravatar image

I guess you are using amcl to localize the robot. If that is the case, what is happening is that you have very little or redundant information about the map you are in, which results in many candidate poses for acml.

Imagine your robot is in an empty square room, facing a corner. How can you tell which corner the robot is facing?

To solve you problem, either create a map with more distinct features or play with amcl parameters, as reducing the number of particles or setting the parameters recovery_alpha_slow and recovery_alpha_fast to 0.

edit flag offensive delete link more


Could you tell me how to create a map with more distinct feature. I am confuse about how to add more feature.My map are basically a .png file convert from my cad drawing I have try to set recovery_alpha_slow and recovery_alpha_fast 0.0, and unfortunately it does not work in this case.

samping gravatar image samping  ( 2016-09-15 20:46:24 -0500 )edit

I have increase my resample_interval:1 to 100 in amcl_params.yaml, and It does improve my positioning problem, but the jumping problem is still happen during the simulation.

samping gravatar image samping  ( 2016-09-15 21:27:15 -0500 )edit

to create a better map for localization, try to add more features, as walls, objects, etc...

Procópio gravatar image Procópio  ( 2016-09-26 05:43:48 -0500 )edit

answered 2016-09-14 03:33:42 -0500

vmatos gravatar image

I would suggest reading up on SLAM, particularly particle filter SLAM. It seems the robot can't find itself due to the similarity of map regions. What's the laser range? You can also have a discrepancy in unit size between odometry/grid map/laser data.

edit flag offensive delete link more


I do not think I am using slam for my package

samping gravatar image samping  ( 2016-09-15 20:47:18 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower


Asked: 2016-09-13 03:01:11 -0500

Seen: 500 times

Last updated: Sep 14 '16