Hi all,
We have a Kinova Mico, which is the smaller, two fingered version of the Jaco arm. We got the SolidWorks files from the manufacturer and using the SolidWorks to URDF exporter plugin, I generated the URDF files with various visual detail levels. Then, I obtained the .dae file through the collada_urdf package. I can display these models in OpenRAVE, manually move the joints without any problems, and even move the arm to desired joint values via the MoveManipulator function; however, when I try generating the full inverse kinematics model of the arm, I run into some serious problems. The ikfast methods cannot seem to find a solution. Attached, I have a sample output from the procedure of trying to autogenerate the ik model. Looking at the numbers printed at the very beginning, I can tell that there's clearly something wrong, but I cannot identify the root of the problem. I've tried several things like defining all joints as revolute joints with set limits, as continuous joints, and even building the URDF all over from scratch, making sure that the axes and frames are defined properly, etc., but none of my attempts were successful. The SolidWorks to URDF exporter output plus the .dae file (under the robots folder) for the simplified model can be downloaded from the following link: http://tekin.mericli.com/share/mico/mico_simplified.tar.gz I'd really appreciate it if someone could point to me what I could possibly be doing wrong and what I should try. Thanks a lot in advance! Best, Tekin |
Hi again,
I've been struggling with autogenerating the IK of the Mico model in OpenRAVE. Even though I've read every single forum entry that looks relevant, followed all tips and suggestions provided by Rosen and others, and played with the axes, the end effector orientation, decimal precision of the axes and anchor definitions, and pretty much everything else that could be played with, I haven't been able to get it working. Here I have a better looking model of the robot: http://tekin.mericli.com/share/mico/mico_with_rings.tar.gz The "robots" directory includes the URDF, .dae, and xml files. The mico_rounded_decimals.robot.xml file has reduced decimal precision for the axes and anchor definitions, whereas the mico.robot.xml file has the original values. When I try running openrave.py --database inversekinematics --robot=mico.robot.xml the ikfast says "found 3 consecutive non-intersecting axes links[0:5], vars=[j0, j1, j2]" and fails after some time. Specifying various --precision values has no effect on the result. While playing with the axes, I noticed that when I change the axis orientation of j1, which is the Arm_Joint, from this ![]() to this ![]() this time ikfast says "found 3 consecutive non-intersecting axes links[4:9], vars=[j2, j3, j4]" and proceeds with the calculations, finding a (of course wrong) solution in a couple of minutes. Interestingly, though, even this doesn't work for the mico_rounded_decimals.robot.xml file. I thought this observation might give you guys a clue about what I could possibly be doing wrong/right and guide me in a more informed manner. As a side note, I found a rough model of the Jaco model on the OpenRAVE forum (http://openrave-users-list.185357.n3.nabble.com/Compiler-errors-when-using-ikfast-td4026357.html), for which an IK solution can be generated. I've tried comparing and contrasting my Mico model with this Jaco model, and interestingly (and expectedly) their relative joint orientations are identical, yet ikfast works for one and doesn't work for the other. As another side note, I was able to get the IK for the robot working under arm_navigation and V-Rep, so I'm wondering what's going on with OpenRAVE. Your help will be much appreciated! Best, Tekin |
In reply to this post by Tekin Mericli
Could you find a solution?
I get this output when I run the same thing for a jaco arm using fivef's JacoROS INFO: ikfast 6d: [j0, j1, j2, j3, j4, j5] INFO: found 3 consecutive intersecting axes links[4:9], rotvars=[j2, j3, j4], translationvars=[j0, j1, j5] My doubt is: Jaco doesn't have any prismatic joints: why does the output say translationvars? Also Jaco's wrist is definitely not a wrist with intersecting axes. Why does it find three intersecting axes then? |
Administrator
|
something's wrong with the file... can you send the collada file you are testing (preferably with geometry)? 2014-07-05 17:00 GMT+09:00 ksatyaki <[hidden email]>: Could you find a solution? ------------------------------------------------------------------------------ Open source business process management suite built on Java and Eclipse Turn processes into business applications with Bonita BPM Community Edition Quickly connect people, data, and systems into organized workflows Winner of BOSSIE, CODIE, OW2 and Gartner awards http://p.sf.net/sfu/Bonitasoft _______________________________________________ Openrave-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/openrave-users |
Hi Rosen, Find attached a model that we use for the Mico arm (I'm working with Tekin). Let me know if this helps. ksatyaki, if you also have a model, feel free to share it.On Fri, Jul 11, 2014 at 1:54 AM, Rosen Diankov <[hidden email]> wrote:
------------------------------------------------------------------------------ _______________________________________________ Openrave-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/openrave-users |
Very sorry I didn't hit reply to all. I used the urdf from JacoROS package: https://github.com/ksatyaki/JacoROS ---> fork from fivef's JacoROS.
And from the urdf under jaco_description, I generated the .dae files... All those should be under: On 12 July 2014 00:58, Chris Dellin <[hidden email]> wrote:
------------------------------------------------------------------------------ _______________________________________________ Openrave-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/openrave-users |
Free forum by Nabble | Edit this page |