Learning inverse kinematics

[evo85210]

I'm interested in learning how inverse kinematics works in PRC.
I've found lots of papers and tutorials online, but they're usually 2D kinematics.
Just wondering if anyone have any sources/pointers on where I can start if I want to learn about inverse kinematics in 3D

[Johannes @ Robots in Architecture]

Hello,

Our algorithm in KUKA|prc is now very customized and optimized to resemble the KUKA solver (which is not documented).
But basically the process is not that hard, you just need to split your robotic arm into two parts: One part that contains the first three axes - the one that you can basically solve in 2D, adding A1 afterwards, with very standard trigonometry - and then you treat A4 to A6 as a single spherical joint.

There are numerous implementations around, I'd look for code samples and then compare it with the more mathematical papers.

Sorry that I cannot just copy/paste our code, but it's quite long and much work went into it.

Best,
Johannes

[evo85210]

Thanks for the input, and of course, I'm not expecting you to post the PRC source code, you've put lots of work into it, and I appreciate that.
Your pointer about breaking it into two parts seems like a good starting point.

Just out of curiosity, how have you guys made PRC solve to similar results as the kuka solver?
I havent started anything yet, but id imagine there are cases where there's more than one possibility?

[Johannes @ Robots in Architecture]

Hello,

With much experimentation and trouble-shooting, and with the input of the community. It's still not "perfect", but then again the only close to perfect simulation approach is OfficeLite, which is the entire robot controller running within a virtual machine.
If you have got a license for it (may be included with EDU packages) then OfficeLite is very useful for troubleshooting, without always having to be close to a robot.

Best,
Johannes