[ros-users] Simple Navigation

Gonçalo Cabrita goncabrita at gmail.com
Fri Feb 18 17:01:26 UTC 2011


Hi guys,

Thanks for your feedback.

I'll give it a try on Monday and post my findings! I'll definitely also use
the bumpers, the Roomba does love to bump on things. And if you think about
it using the IR array to avoid obstacles means that it is not a matter if
the Roomba will hit something or not, it's just a matter of when!

Gonçalo Cabrita
ISR University of Coimbra
Portugal

On Fri, Feb 18, 2011 at 4:43 PM, Brian Gerkey <gerkey at willowgarage.com>wrote:

> 2011/2/18 Gonçalo Cabrita <goncabrita at gmail.com>:
> > Here at the lab we use mostly the Roombas. Thanks to Eitan we
> > now successfully run our Roombas with the navigation stack using a Hokuyo
> > laser for obstacle detection.
> > However the Roombas come equipped with many cool sensors. I was wondering
> if
> > it would be possible to use the Roomba as is, to run a simplified version
> of
> > the navigation stack. My idea was to publish the IR data as a point cloud
> > and feed the information into move_base running solely on a local dynamic
> > map, so no map server or amcl. The performance wouldn't be great for
> sure,
> > but maybe the Roomba would be able to move around without crashing.
> > What are your thoughts on that? Is it even worth a try or is it just
> idiotic
> > ( the IR sensors have a range of about 20cm :P )?
>
> It's interesting enough to give a try to. You could also use the bump
> sensor information to produce obstacles.  They're coarse, but you can
> adjust the resolution of the costmap accordingly.  I would expect the
> robot to run into things, but it's a Roomba, designed to do that
> right?
>
>        brian.
> _______________________________________________
> 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/20110218/caa056d1/attachment-0003.html>


More information about the ros-users mailing list