ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
0

Random Orientation Vector for Hand Pose

asked 2013-02-26 06:20:44 -0500

MartinW gravatar image

Hello all,

I was wondering if there was a way to create a random orientation. In my project I get the xyz coordinates of an object I want my robot to pick up, however I can't get the orientation correct enough that it works with IK services. So I was thinking of making a random orientation vector with my coordinate and just loop over until I find a valid solution. Is this a good way to approach this problem? Or should I just try find a few orientations manually in Warehouse Planner and saved these values in some array and just loop over those when I try to pick up the object.

Kind Regards, Martin

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted
0

answered 2013-02-26 06:55:06 -0500

dornhege gravatar image

Does your robot's workspace cover full 6-Dof? In that case you should be able to find an IK solution.

If it isn't, yes, randomization will work and be simple to implement. However, there might be better solutions. If you use a numeric solver for IK, this could give you the nearest pose, even if the IK was not solved exactly.

Alternatively you could also get the grasps for you object first (if possible) and find IK solutions only for the valid grasps.

edit flag offensive delete link more

Comments

Hey dornhege, yea my robot has 6-Dof and two fingers for grasping. Right now I am using the IK services provided by the warehouse planner, but it can't seem to find any solutions. However these are precise poses I am looking for, is there a way to ask for a position xyz with any orientation?

MartinW gravatar image MartinW  ( 2013-02-26 07:35:48 -0500 )edit

Are those poses reachable? Then maybe, you just need a better IK. Technically you can query for xyz, but I'm not sure, if that's possible with the default implementations.

dornhege gravatar image dornhege  ( 2013-02-26 23:13:20 -0500 )edit

Can a better IK be generated by using a denser search within the sampling modes of the Planning Description Configuration Wizard? I just use the default IK services

MartinW gravatar image MartinW  ( 2013-02-27 07:17:23 -0500 )edit

Disclaimer: I don't know how the wizard works! But my guess is that the sampling is only used to determine necessary collision pairs for the configuration and has nothing to do with the IK. You should be able to test the IK manually from rviz though. That should give you some insights.

dornhege gravatar image dornhege  ( 2013-02-28 00:08:27 -0500 )edit
1

Depending how that goes, you can look into: http://openrave.org/docs/latest_stable/openravepy/ikfast/

dornhege gravatar image dornhege  ( 2013-02-28 00:10:32 -0500 )edit

Question Tools

Stats

Asked: 2013-02-26 06:20:44 -0500

Seen: 258 times

Last updated: Feb 26 '13