• aditya mehrotra.

chip updates: getting the IK working and the system back to normal [updates]

So we turned the system on this morning, wrote the Inverse Kinematics node (we won't describe in detail because it's the same as the last system. Nothing has changed there there's now just a node in-between master (which published XYZ) and the write node (which takes in commands).

The above video is the result of the normal stand-sit trajectory. It's pretty clear the hinge and the shoulder PID needs more tuning. What we'll do is increase both KD and KI. Okay the oscillation got bigger but it came back to zero better. What if we increase KD even more?

Okay... KD is now 0.5... let's see. It still over shoots KI = 0.0001?

Okay we've tuned it to the point the motor commands sound terrible. We've increased the KD of the knee just for kicks. Now we've tuned a bit more and here's the result:

As we can see there's still some NOISE from the motors because the command isn't as stable as we'd want it to be. However, this should be okay for now because of how stable the system response is. I don't want to mess it up. Now let's see if the same sit/stand trajectory/points will work on this new-and-improved software stack or if we will need to like update that to something else. Time to put chip on the floor.

***This is also with the bad leg we have to fix***

Mmkay well clearly that didn't work. Let's try making the back farther back hm? YES! I want to try it from the beginning though, zeroing-sit-stand-sit.

So those were some final two videos of it sitting and standing. What this means is we're now basically back to where we were before Wednesday? Tuesday? Whenever we started re-writing everything. So there's only one thing left before we are back where we were Tuesday (functionality wise) is a forwards kinematic node that continuously publishes current XYZ values to the network. We're going to write that as well really quickly.

#updates #chip #omgrobots

0 views0 comments
© copyright 2019 | aditya mehrotra