Inverse Kinematics

I'd like to create an inverse kinematics chain for a 6-axis robotic arm, and use it to generate the G-Code necessary to operate the robotic arm. For starters, does anyone have experience with inverse kinematics (just lines and joints) in Grasshopper?

Replies to This Discussion

Permalink Reply by pyrit on September 5, 2010 at 2:36pm

hey !

would be very interested in this topic too.

Permalink Reply by Daniel Piker on September 5, 2010 at 7:38pm

I did look into doing some more serious constraint solving as part of Kangaroo, but it is tricky stuff. Maybe if I had some funding and a few months to study and work on it I could get somewhere...

There has been some interesting discussion related to LEDAS and Solvespace in the past, but I don't know if there have been any further developments with that (?)

In the meantime I decided to have a go with Galapagos.

(It is trying to match the position of the yellow object to the blue one.)
I think it's an interesting test, but not really practical.
(It's very slow, sometimes gets stuck, the solutions are only approximate, and it's only for static solutions not actual toolpaths).
Perhaps it's possible to get it to converge quicker - I haven't really got the hang of tweaking the different Galapagos settings yet. Here's the file if anyone's interested.
robot_.3dm
robot_.ghx

Permalink Reply by Daniel Piker on September 6, 2010 at 11:14am

Okay, so I did some real IK !
I realised that in the special case where the robot has a spherical wrist (the 3 axes intersect in a common point), as many of them do, then you can do something called kinematic decoupling, which makes it simpler to solve.

It can do paths (and it's fast too)

Permalink Reply by Jacek Jaskólski on September 6, 2010 at 11:19am

wow!

Permalink Reply by Avihay Bar on September 6, 2010 at 12:03pm

WOW!

Permalink Reply by Evert Amador on September 6, 2010 at 1:54pm

Hi Daniel,

Really good stuff!

Is this done with Kangaroo?

Cheers

Evert

Permalink Reply by Daniel Hambleton on September 9, 2010 at 11:21am

Nice!

I was trying to get through this paper a while ago on the topic. It's pretty intense, but might be useful....

Permalink Reply by Tuan N. Tran on September 9, 2010 at 1:13pm

schweetness! can i play too? :)

Permalink Reply by Aaron M. Ryan on November 1, 2010 at 1:56pm

daniel,

Can you share your solutions?

aaron

Permalink Reply by nirulfen on September 6, 2010 at 10:58pm

absolutly amazing!!!
next step is maybe to build it
I hope to see the all project directly working within Rhino.

Permalink Reply by Daniel Piker on September 7, 2010 at 10:39am

Thanks all for the response :)

Evert - right now this solver is separate, but I am thinking of making it part of Kangaroo.
(It does work with Kangaroo, so you can control physical simulations with the arm, and you can control the effector position/orientation with physical simulation, you just can't make systems mixing arbitrary geometric constraints with physical forces.)

I'd be interested to hear more about how people use robots and IK now in other programs, and how they might want something like this to work in Grasshopper.

and Nirulfen - thanks for the link, that's a nice project.

Permalink Reply by Matt Trimble on October 18, 2010 at 4:15pm

Thanks for your work on this, Daniel! Your IK solver seems to work quite well. My interest is in creating a text file that would contain position and angle data for each of the robot's axes for the duration of its movement (eg, based on a toolpath). I'd love to discuss this with you further. You can reach me directly at matt@radlabinc.com, or at my office in Boston at 617.440.3588.

I look forward to hearing from you.

Matt