• aditya mehrotra.

chip updates: combining known motor data from previous testing w/ some basic math [updates]

So this morning we wanted to make sure that 10:1 gearboxes would actually be enough for what we were trying to do with the new chip robot. Note that we only took apart one gearbox so far so going back to the 100:1 would not be terribly difficult. Here's some math in the PDF below:

Note May 31, 2021
Download PDF • 864KB

The basic testing from last time showed us at the 80A current limit we can expect 1.6-2 Nm from the motor. And w/ a 10:1 gearbox that's only 20Nm. That's really not enough to make this system work reliably.

So I think the conclusion here is to keep the 100:1 gearboxes but do better forms of both packaging and design from mechanical standpoint. Re-design and fix some fundamental issues.

Also look at: https://www.ghostrobotics.io which shows their new quadruped has a leg design very similar to chip's where the rotations go shoulder, hinge, knee instead of the swing joint first. Since we are using 100:1 gearboxes, if we can get a decent impedance controller working with those we'll use that otherwise we can try going the boston dynamics SEA route.

So things we need to fix:

  • positioning of the motors and mechanical robustness

  • upper leg length should equal lower leg length

  • impact mitigation

  • power system, legs should be able to move faster

  • motion profiling and smooth control on the legs not just PID control

The leg lengths honestly might be the biggest atrocity, so there's a version of chip where we fix the slop in the lower joint, make the leg lengths the same, and just make it mechanically better and be done with it. I think this is the strategy we are going to try with a stronger body and lighter legs and everything. Going to try to make everything lighter.

We'll start designing this tomorrow!

0 views0 comments
© copyright 2019 | aditya mehrotra