Parallel manipulators

[Gonçalo Cabrita]

Hi guys,

I've been searching the move it website and malling list but I could not find an answer to my question.

Is it possible to use move it with parallel manipulators?

We are currently working on a de-mining project and we developed our own arm to move the sensors. It is a 2 DOF arm and the last link, which actually contains the sensors is meant to always be parallel to the ground, so it has 2 parent links. This is not supported by URDF, and so I think it does not work with move it. Am I right to assume this? Is there a way to make this work or a plan to address this in the near future?

I'm attaching a screenshot of our URDF, the model was hacked to work with ROS, so the last link only has one parent in the model, the top tube.

Best regards,

Gonçalo Cabrita

[Dave Hershberger]

Hi Gonçalo,

Seems to me like you could use a "mimic joint" to make the model look right:

http://wiki.ros.org/urdf/XML/joint in section 3 under "<mimic>".

That's what we use to make PR2 parallel-jaw grippers move correctly.

You wouldn't actually model the closed kinematic chain, since URDF doesn't support that.  You would just make joints at the near and far ends of the boom mimic each other (likely with a -1 multiplier) to keep the end parallel to the ground.  Then you could make another joint so the upper boom mimics the lower boom, but don't actually attach the upper boom to the end link, since that would close the kinematic chain.

Dave

[Sachin Chitta]

Hi Gonçalo,

We have done four-bar linkages with MoveIt! before by setting up the URDF the way Dave suggested - although they were for robots with more than 2 DOF.  Are you looking at using planning or IK? With only 2 DOF, you could write a small custom IK plugin that should work incredibly fast for your arm but you might not even need that.

Very cool robot!! Could you send us a movie for the MoveIt! blog once you get it working?

Best Regards,
Sachin

[Gonçalo Cabrita]

Hi Sachin

The arm just needs to sweep from side to side at a constant speed, however, since this is going to be used offered I need to maintain the arm at a constant distance from the ground. I was thinking about writing a custom node to read my 3D point cloud and do that. Then I remembered about obstacles and I thought about using Move It. Do you think that Move It will work well on this setup or would it be better to try something simpler?

I also have another question regarding Move It. I'm currently using the ROS navigation stack to move the mobile robot, and I was thinking about Move It to move the arm. Is there a way to bring everything together? Can both software packages work in order to plan trajectories and avoid obstacles or is mobile robot and robotic arm planning and navigation completely independent?

Finally, it would be my pleasure to send you some pictures and footage of the robot looking for some mines :)

Best regards,

Gonçalo Cabrita

<Mail Attachment.png>

[Sachin Chitta]

Hi Gonçalo,

Building your infrastructure around MoveIt! would be the right way to go here. MoveIt! was designed to be modular so you should be able to do something "simpler" and keep it lightweight. Your motion itself might be fairly simple since you are using essentially 2 DOF but you can reuse the perception components (self-filtering, octomap, etc.). You will still have to write your own custom stuff for putting together the perception and your motion - I would use MoveIt!'s infrastructure for that.

Regarding bringing together MoveIt! and navigation, this had been in our plan a while back but in the absence of resources we won't be able to get to it anytime soon. You can still design for your use-case by keeping both separate and changing the height of your sweeps based on upcoming terrain using MoveIt!. Keep the mailing list posted on issues as you go along.

BTW, this is a very cool project :-)

Best Regards,

Sachin

[Gonçalo Cabrita]

Hi guys,

So I did as you guys suggested, and I added the mimic tag to my URDF and everything seems to be ok.

I started by running the setup assistant, following the online tutorial. Once finished I continued to the rviz tutorial to check if everything was working, however, after checking if everything is according to the tutorial, I cannot see the interactive markers to move the goal and start states. I tried this with the pr2 example and everything works fine. Could this be related to something I might have done wrong during the setup assistant? How should I continue?

Best regards,

Gonçalo Cabrita

[Sachin Chitta]

Hi Goncalo,

This is hard to debug without more information. The typical reasons why the interactive markers may not come up:

(1) You have the wrong fixed frame in Rviz - try changing it to the planning frame (the root frame)

(2) you have not setup the group the right way

(3) you have not setup the end-effector the right way

Without taking a look at your URDF/SRDF, its hard to say.

Sachin

[Gonçalo Cabrita]

Hi Sachin,

I've gone through the three topics and it seems everything is according to the tutorials, however, since I'm using a parallel manipulator, and my end effector is not a gripper I'm not sure if this will have any ill effects on the setup. I was trying to avoid bothering you by sending you the files, but if you could take a quick look at the setup I would be very thankful :) This is the first time I'm setting up move it I'm a bit lost!

I'm attaching the files generated by the setup assistant.

Best regards,

Gonçalo Cabrita