• aditya mehrotra.

chip updates: now maybe we can do walking? [updates]

Let's give this thing one last test, all we added at the end last time that we weren't able to test was shifting back to center. We want to try that now just to know that it works before we program in a whole walking sequence.

So that works pretty well :) - now theoretically it should be as simple as the following four lines of code which is why we spent so much time developing this function.

Let's test it out!!!! If I push the right joystick in any direction up/down the platform should take a few steps forward.

Ah yes, we tried and there's big error message. I think we indexed something incorrectly but we fixed it so hopefully this time we'll get an actual step sequence. Need to tune that parameter for terrain though.

Okay so even though the terrain parameter was set, the robot fell over flat on its face so I don't know what happened there here's the video, let's play with some parameters:

Alright so this thing has started acting up again doing those weird up/down dance moves it does when it's overheating but maybe we can review the footage to see what's really happening.

Let's look at the film frames in order of how the platform fell over.

Here's in the middle of it transitioning its weight towards the front. It's stable so far. It doesn't look like if you picked the back legs up now that it'd be able to handle not falling over backwards so we can't stop it here.

A little more forward and what happens is the front actually decreases in height since we told the "y" positions to decrease.

And now we kind of notice the real problem here which is that front right leg (because of slop) just barely slips backwards so traditionally leaning forward wouldn't cause a fall but since the legs are mismatched in the front...

The platform just kind of leans too much on its front legs and then...

Just goes in for a dive... it falls down face first. But the real question is how do we solve this and I think there are a few ways. First, we might be leaning too far on the front legs, the lower height at the front is one of the reasons its so easy for the platform to keel-over forwards. It also, just very well could be, the fact that the legs are trying to shift too far and we need to decrease the back shift because in frame 3 at least, it doesn't seem like the platform should be keeling over forward and lifting the back leg doesn't seem like an ordeal in that case.

Another possible solution is we start the rear legs farther back than the front. So when it stands up we use the regular algorithm to step the rear legs backwards and when it sits down make them step the rear legs forward and this give the platform a wider base which could make keeping the CoG in the triangle a lot easier. It could also be worth when the platform starts up, decreasing the z-distance now that we can move the feet anywhere. It might be simply worth putting the legs in a more ideal position for walking - maybe we'll try that. Like have the platform step inwards like this:

Just like step inwards when we're transitioning to walking mode, you know. And then when we're transitioning out, we step back out of that position. We don't want to do too much z because the problem is the leg could contact the body and bend things. Yes - the design is not ideal.

Well we'll get to this later.

#updates #omgrobots #chip

2 views0 comments
© copyright 2019 | aditya mehrotra