cdsteinkuehler wrote:UPDATE: I just got the inverse kinematics working with LinuxCNC. It's way too much fun to do coordinated jogs in X and Y and watch both arm motors move! Video to come in a bit...
False alarm. I did pretty much get this going, but there seem to be some issues with either my math or the LinuxCNC support for inverse only kinematics. So I took an unexpected detour and had to code up forward kinematics. I gave up on a closed form solution, my math mojo just isn't up to the task given the available time. Instead, I'm using "simple" wally kinematics, where I assume the arm position doesn't affect the motor drive (as if the motor was mounted on the shoulder arm). Then it's just intersect the two circles to get a predicted XY, use the predicted XY to calculate the arm positions, calculate a the new effective drive command based on the current arm position, lather-rinse-repeat until the error from one step to the next is below your threshold. So far it's taking about 25-30 iterations to get below 0.05mm error.
Now I've got to pack, so Wally with full kinematics looks like a show-floor activity. Nothing like working in demo mode!
