Ask Your Question
2

Navigation Slam gmapping multiple laserscans

asked 2015-01-14 13:03:55 -0600

fabs gravatar image

Hello, i'm trying to create a map with two asus xtion on my turtlebot2. They are mounted back to back in the middle of the turtlebot. One to the front and one to the back. After bringing up the turtlebot and the gmapping_demo i can see both laser scans in RVIZ, so this seems to work. I can also see the messages published on /scan from both cameras. The problem is now, slam_gmapping only creates the map basing on one camera as i can see in RVIZ. Is there a possibility to create a map basing on multiple laserscans? So, i think the problem is the gmapping node only uses one laserscan from one camera. I've already found this discussion on this topic: http://answers.ros.org/question/9294/... Maybe there are some new insights.

edit retag flag offensive close merge delete

3 Answers

Sort by » oldest newest most voted
3

answered 2015-01-15 15:06:00 -0600

fabs gravatar image

updated 2015-01-16 11:09:28 -0600

Today i found the following very interesting project from the iralab. They developed a laserscan toolbox for merging laserscans. In the example they got another assembly of the scanners, but maybe it can be modified. I think the problem point is the destination_frame. I think i have to create a "virtual" camera with 60° in the front and 60° in the back. Correct me if i am wrong.

  1. https://github.com/ros-perception/sla...
  2. http://arxiv.org/pdf/1411.1086.pdf
edit flag offensive delete link more

Comments

This is the right solution :). I've got not enough points to mark it as the correct answer.

fabs gravatar imagefabs ( 2015-01-21 10:04:21 -0600 )edit
0

answered 2015-01-14 13:30:00 -0600

Hello,

This question is interesting! I was thinking in doing the same but I don't have the time for that. To answer your question for gmappimg, you must write a node that subscribe to both scans and publish one new larger scan. my advice is not to use 180 deg separation between the sensors. It is better to use the the field of view of your sensor as the separation. Then your node must deal some how with the interlaced pixels. This is expected to improve the resultant map.

edit flag offensive delete link more

Comments

hi, dude, have you solved this problem? thanks.

asimay_y gravatar imageasimay_y ( 2016-02-02 20:34:11 -0600 )edit
0

answered 2015-01-15 02:50:31 -0600

See also this previous Q/A (you'll likely find some others, too). I know of no existing ready-to-use solution so far. You can of course hack something by fusing two scans into one, but depending on LIDAR locations this could introduce distortions the data. The other option is diving into gmapping's code and adding true multi lidar support.

edit flag offensive delete link more

Comments

1

hi, dude, have you solved this problem? thanks.

asimay_y gravatar imageasimay_y ( 2016-02-02 20:34:45 -0600 )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

3 followers

Stats

Asked: 2015-01-14 13:03:55 -0600

Seen: 1,711 times

Last updated: Jan 16 '15