Looks great, excited to try it! I wish the license was more permissive (mit, etc). ACT training code would be great, this is the same thing I’m working on doing, but haven’t got that far yet. Thank you for sharing this!!

Yeah pwm + sensor seems like the right move here

Any chance you got to it yet?

So you assume calibration is wrong if there is a crash and attempt to fix it?

can you please change license to be more permissive? would be happy to contribute if so. MIT License for example?

any chance you have got around to doing this yet?

I just manually manipulated it back into place. Not sure why this happened, do the limits not work with the cartesian and tool coordinate methods of manipulation?

It was, and I had tightened it super tight, but still slipped. I put some bearing compound on it and that fixed it so far.

the rods are slipping for me when using the J5 motor (i have put the set screws). Do you have this problem too?

drill it out with a 3mm bit. I had the same issue.

I’m at that point! :)

Is your firmware open source? Maybe I can contribute when I get to that point. Seems like we’re doing similar things

Omg, thank you! Saved me a ton of work!!

Was thinking of modifying this design to meet the same proportions as AR4: https://github.com/AlexanderKoch-Koch/low_cost_robot As long as the proportions and joints are the same, it should work fine. I have to get better at cad to learn how to modify it though! Will post when I get it done.

An arm (usually built with cheaper materials / motors) that can be used to remotely control the follower arm (for example AR4). This is used to teleop or to train AI models for the arm. The leader is only used to read values for the joints from.