PDA

View Full Version : IMU + LIDAR SLAM approach



hash79
08-02-2011, 12:52 PM
Has anyone seen this video?

http://www.gkmm.tu-darmstadt.de/rescue/?q=node/141

Using ROS with an IMU as their odometry and a LIDAR. The results look amazing! I was unable to find any more information. They say on their site they will be posting ROS drivers soon so I thought someone may have heard about this already.

I have an IMU I would like to stick in the XV-11 to see how well this approach works.

-Hash

SK.
08-02-2011, 03:55 PM
Hi,

being the developer of the scan matching/mapping approach used I can say a thing or two about Hector SLAM :)
Code and logfiles are already available and working at http://code.google.com/p/tu-darmstadt-ros-pkg/downloads/list . I intended to do some cleanup before a public announcement though, as there are still some rough edges.
With the default settings it will likely not work as well as with the Hokuyo LIDAR, as we leverage and rely on the high (40 Hz) scan rate to enable odometry free operation. Time permitting (which I can't guarantee right now) I'd like to look at some XV-11 bagfiles to play around with though.

I made a playlist of all videos we uploaded so far showing the approach in action:
http://www.youtube.com/playlist?list=PL0E462904E5D35E29

hash79
08-02-2011, 05:48 PM
Well very nice work, when I saw that video my jaw dropped!

From your description it sounds as if you are using the the map generated initially by the LIDAR and the distance measurements from the LIDAR to extrapolate odometry information. Is that a correct assumption? Is the IMU used to compensate for tilt and measure any rotation in the platform as it is moving around?

-Hash

SK.
08-03-2011, 02:01 AM
Yes, this is a very short summary of what is going on behind the scenes. Depending on the setup we use a EKF based INS for full 6DOF pose estimation that gets updated with pose updates from the 2D SLAM system and the IMU data (and potentially other sources), so there is cross coupling between sensors/measurement sources.