[ros-users] canonical way to use multiple laser scanners

[Aaron Solochek]

Done: http://answers.ros.org/question/262/using-multiple-laser-scanners

However, I have some more general questions that don't seem suited for inclusion
in that post.  Maybe they should be separate questions, I'm not sure, but I'm
hesitant to start posting questions there that go into multiple issues.

One method of merging the LaserScans I've considered is allowing the lidar nodes
to both publish on the same topic.  Is it safe to assume that any mapping node I
might use is agnostic to the number of scanners publishing?  I'm worried about
there potentially being some inefficiencies due to half the incoming scans
coming from a different frame (e.g. unnecessary overhead of having to refetch a
transform rather caching it -- I don't know if anything does that though,) or
even just the problem of applying a kalman filter or similar to the data.

What I'm doing currently is:

scanning node A->filtering/processing node A-> laser scan merging node

scanning node B->filtering/processing node B-> laser scan merging node ->
mapping node (which takes a PointCloud, since LaserScans can't be merged cleanly)

My laser scan merging node is basically a laser_assembler node, modified to
subscribe to two different topics, and when it has received a scan on both
topics to merge them and publish a PointCloud.  I'd like to make this node more
flexible.  Ideally the number of laser scanners it could merge would be settable
at run time, but I'm not sure the best way to do that.

Either way though, nodes like slam_gmapping don't take PointClouds, which
presents me with another problem.  I could convert the PointCloud back into a
LaserScan, as if it were a 360 degree scanner located at the origin of the
base_link, but obviously that would require that the scanners are all coplanar,
and it also throws away data.  I haven't looked at the slam_gmapping code yet,
but I'm think I might need a locally modified version of it that takes
PointClouds.

Thoughts/suggestions?

-Aaron


On 03/04/2011 01:39 PM, Tully Foote wrote:
> Aaron,
> 
> Could you ask this on answers.ros.org <http://answers.ros.org>?
> 
> Thanks,
> Tully
> 
> On Fri, Mar 4, 2011 at 9:10 AM, Aaron Solochek <aarons-ros at aberrant.org
> <mailto:aarons-ros at aberrant.org>> wrote:
> 
>     I have a robot with 2 hokuyo scanners, one on the front, and one on the rear.
> 
>     What is the cleanest way to merge them into one point cloud, so that the merged
>     scan can be used for mapping?
> 
>     I was hoping to use one of the laser_assemblers, but they seem designed only for
>     assembling scans from a single nodding scanner.
> 
>     I'm hoping I'm missing something obvious here, because this doesn't seem like it
>     would be all that strange a setup.
> 
>     Thank you.
> 
>     -Aaron
>     _______________________________________________
>     ros-users mailing list
>     ros-users at code.ros.org <mailto:ros-users at code.ros.org>
>     https://code.ros.org/mailman/listinfo/ros-users
> 
> 
> 
> 
> -- 
> Tully Foote
> Systems Engineer
> Willow Garage, Inc.
> tfoote at willowgarage.com <mailto:tfoote at willowgarage.com>
> (650) 475-2827
> !DSPAM:4d71321720581132719651!
> 
> 
> 
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
> 
> 
> !DSPAM:4d71321720581132719651!