Such a cool build, 2-4mm accuracy seems too off, if this is more of a parts quality issue maybe you could add a redundant array of sensors to get positional data besides what the stepper motors provide and use that data with a PID algorithms to compensate it.
Very cool demo. With the steppers its a bit less important but are you going to explore dynamic modelling of the system? I suppose with knowledge of the stepper state + closed loop feedback you could predicatively throttle the motion based on the fastest joint rate to ensure stepper torque always exceeds the required system torque
Hey. Thanks for sharing your awesome work with all of us! Quick question. What software are you using to create your Forward/IK models and your motion planning? I am just starting building of an Annin Robotics AR4 Mk2, so I hope to join you in this field very soon! ;)
Great work! For what it's worth, I've got three of those X17-30A Nema 17 wave drivers, and while I was initially over the moon about them in theory, I discovered that they have enormous axial slop when any weight/leverage is applied. Like a solid degree or two of "wobble" you can squeeze out of them with a little leverage. Their bearings are simply not up to the task, IMO. I have three genuine Harmonic brand drives of similar size (CSF-14-30-2XH-J) I bough off e-bay and those things are tanks. While I had to print adapters to mount nema 17's to them they are flawless.