[ros-users] Problem with canonical_scan_matcher

Ivan Dryanovski ivan.dryanovski at gmail.com
Fri Jan 7 05:07:56 UTC 2011


Hi,

It is hard to tell from the video how much error you are getting from the
scan matcher - I can't make out what the scale of the axis is. However, keep
this in mind: both the canonical and polar scan matchers match scans
incrementally. The matching is only done on the last two scans, and not on
the entire history of scans. Therefore, even if you get a very small error
between any two scans, this error eventually builds up over time. This might
account for the error you are getting. Depending on the quality of your
laser, I would assume it would be normal to get some drift. Perhaps if you
tell me exactly how much drift you're getting (in m/s) I could tell you if
that's expected behavior.

You can try tweaking the parameters in order to minimize drift. Try setting
max_correspondence_dist and orientation_neighbourhood to something small -
this will hopefully remove more false correspondences. Also, try bumping
max_iterations higher, like 20 or 30, and see if that improves matters (at
the cost of computational time, of course).

Let me know if any of the above helps. If you describe your use case and
hardware in more detail, perhaps I would be able to give you more
suggestions for setting the parameters

Cheers,
Ivan



On Wed, Jan 5, 2011 at 5:05 AM, #TAN ZHI PING# <TANZ0066 at e.ntu.edu.sg>wrote:

>  Hi,
>
> I was trying out my pioneer robot using the canonical scan matcher but i
> realized that the pose2D (odometry) keep on changing despite the fact that
> the robot remained stationary throughout the whole event as shown in the
> link provided below.
>
> http://www.youtube.com/watch?v=cTQ0T2p32hc
>
> This video shown the x-axis and y-axis relationship between the odom from
> the p2os driver (blue) and the pose2D from the canonical scan matcher (red)
> using rxplot. As i mention earlier on, there is a constant offset in the
> pose2D causing the position feedback to move away from it's ordinary
> position.
>
> Is there any way which i used to solve this problem?
>
> With thanks,
> Zhi Ping
>
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20110107/c735a308/attachment-0003.html>


More information about the ros-users mailing list