ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
1

Gmapping overwrites map

asked 2012-10-31 03:36:52 -0500

Mendelson gravatar image

updated 2012-11-01 07:34:46 -0500

I'm using a simulation with p2os package and I built my own map (a simple map, with some walls) in order to check if everything was going well. When I start moving the robot, the map is done correctly at the beginning, but when I drive the robot out of the main room, the generated map gets wrong. We obtain something like different parts of the map in wrong places, overwriting one each other. Does anybody know why it is happening?

(obs: I'm building a 2D map with a laser scanner)

Original map:

Original

Original map:

Map

Thanks!

edit retag flag offensive close merge delete

Comments

1

The image link doesn't work. From the description it might just be the case that gmapping failed.

dornhege gravatar image dornhege  ( 2012-10-31 03:55:40 -0500 )edit

thanks, I have fixed it. I'm new at ROS so I don't know if this map is considered complex.

Mendelson gravatar image Mendelson  ( 2012-11-01 07:36:08 -0500 )edit

3 Answers

Sort by ยป oldest newest most voted
4

answered 2012-11-01 16:10:25 -0500

weiin gravatar image

From your images, it looks like something wrong with the rotation odometry. GMapping can usually close loops for small rotational errors, but once that gets large, the map will mess up badly.

Have you run through these yet?

http://www.ros.org/wiki/navigation/Tutorials/RobotSetup

http://www.ros.org/wiki/navigation/Tutorials/Navigation%20Tuning%20Guide

edit flag offensive delete link more
3

answered 2012-11-02 04:16:10 -0500

This scenario looks like you should be able to get good results with hector_mapping (at least if your LIDAR ist not severly range-limited). You should be able to use it as a drop-in replacement for gmapping as described here. Can now also be installed via packages:

sudo apt-get install ros-fuerte-hector-slam
edit flag offensive delete link more

Comments

I have to use gmapping in this project, sorry.

Mendelson gravatar image Mendelson  ( 2012-11-06 04:14:16 -0500 )edit
2

answered 2012-11-01 01:13:23 -0500

Victor_ocv2 gravatar image

updated 2012-11-01 01:13:39 -0500

If you hand us your map it would be easier to answer but I guess that the problem is simple yet hard to correct:

Your environment may be a bit complex and have one or several loops which is a common problem in SLAM algorithms. You can see here that it is a problem that is currently being addressed by many researchers http://scholar.google.com/scholar?q=slam+loop+closing

I suggest that instead of using one 180degree laser you use two and that you should lower down your robot speed.

Hope this helps you.

edit flag offensive delete link more

Comments

From the structure in the images it doesn't seem too complex given that the laser range is sufficient to cover the area.

dornhege gravatar image dornhege  ( 2012-11-02 05:02:09 -0500 )edit

Question Tools

1 follower

Stats

Asked: 2012-10-31 03:36:52 -0500

Seen: 1,458 times

Last updated: Nov 02 '12