[ros-users] Collisions Outside the Bounding Box (planning_environment)

Gil Jones gjones at willowgarage.com
Tue Aug 10 00:03:09 UTC 2010


David,

Apparently ODE, which is the collision checking algorithm we employ, does
not support cylinder to cylinder collisions - you can see what kind of
collisions will be detected on this page:

http://opende.sourceforge.net/wiki/index.php/Manual_%28Collision_Detection%29

Changing one or both pipes to be boxes should mean that collisions between
the objects will be detected - that level of intersection should result in
detected collisions.

-Gil

--
E. Gil Jones (gjones at willowgarage.com)
Research Engineer
Willow Garage, Inc.
68 Willow Road
Menlo Park, CA 94025
650.475.9772


On Mon, Aug 9, 2010 at 4:30 PM, David Lu!! <davidvlu at gmail.com> wrote:

> Also, in the screenshot attached to the ticket, the cylinders in the middle
> do not collide with meshes at all (according to the algorithm). (I confirmed
> this by only checking collisions between the cylinders and the meshes)
>
> Do the colliding surfaces have to be closer to each other to be considered
> collisions?
>
> Thanks,
> David!!
>
>
> On Mon, Aug 9, 2010 at 3:26 PM, David Lu!! <davidvlu at gmail.com> wrote:
>
>> After narrowing down my model, I've found at least one bug that was
>> causing me problems.
>> https://code.ros.org/trac/wg-ros-pkg/ticket/4629
>>
>>  <https://code.ros.org/trac/wg-ros-pkg/ticket/4629>Short Version:
>> Collision detection ignores the scale of meshes
>>
>>
>> On Fri, Aug 6, 2010 at 1:52 PM, Sachin Chitta <sachinc at willowgarage.com>wrote:
>>
>>> On Fri, Aug 6, 2010 at 1:26 PM, David Lu!! <davidvlu at gmail.com> wrote:
>>> >> (1) This is difficult to debug without more information. One thing you
>>> >> could do is display the pose of the robot corresponding to the last
>>> >> collision in rviz. This tutorial -
>>> >>
>>> http://www.ros.org/wiki/motion_planning_environment/Tutorials/Tutorial%20B
>>> >> - helps you do that.
>>> >>
>>> > I've been running off
>>> > of
>>> http://www.ros.org/wiki/motion_planning_environment/Tutorials/Tutorial%20A
>>> ,
>>> > except with my own robot and no laser data, so I know exactly what pose
>>> the
>>> > collisions are happening in.
>>>
>>> True, but the pose that the planning_environment is checking
>>> collisions for might not represent the one you are sending in. Doing
>>> this will show you where the collision environment thinks the robot is
>>> - if there's a potential bug here we might catch it this way.
>>>
>>> Can you send me a snapshot in rviz with collision display enabled. It
>>> looks like (from the marker message), PIECE1 is colliding with PIECE2
>>> in BASE_FRAME. Where is BASE_FRAME defined with respect to PIECE1 and
>>> PIECE2 - maybe send me a urdf as well?
>>>
>>> Sachin
>>>
>>>
>>>
>>>
>>>
>>> >
>>> >>
>>> >> (2) Also, using the rviz "select" button, you can click on the markers
>>> >> and they will give you more information on what is colliding.
>>> >>
>>> > Good to know.
>>> >
>>> >>
>>> >> (3) Are you moving the base of the robot around? We are working
>>> >> towards a release that handles motion of the robot base properly - the
>>> >> current collision environment is not yet fully tested for motion of
>>> >> the base.
>>> >>
>>> > No. I'm only running GetStateValidity with one static pose.
>>> > _______________________________________________
>>> > ros-users mailing list
>>> > ros-users at code.ros.org
>>> > https://code.ros.org/mailman/listinfo/ros-users
>>> >
>>> >
>>>
>>>
>>>
>>> --
>>> Sachin Chitta
>>> Research Scientist
>>> Willow Garage
>>> _______________________________________________
>>> ros-users mailing list
>>> ros-users at code.ros.org
>>> https://code.ros.org/mailman/listinfo/ros-users
>>>
>>
>>
>
> _______________________________________________
> 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/20100809/6e2d72cd/attachment-0003.html>


More information about the ros-users mailing list