• aditya mehrotra.

chip updates: making a plan for transferring weight around [updates]

So here's the problem. We aren't using CMA correctly otherwise we would be able to transfer chip's weight forwards and backwards and pick up whatever leg we want to pick up. So we need to think about how best to do this. The reason CMA isn't working all that well is likely because of the SLOP in the gearboxes. This is an idea we won't give up on but will come back to, and in the meantime we're going to try other things. Such as the new stepping strategy of lifting the lower leg before we move the shoulder at all. I want to try that and see if the platform keels over or something.

This will be a very very slow walk kind of deal. Now another thing to note, is that the platform might actually do better if the y-position of the feet is higher than it currently is. The reason for this is right now, if the drop the y-pos of any leg we get into the region (around x=0.0) where the lower leg is at a very small angle to the ground resulting it being difficult to support. Let's try raising the standing position to 0.50 for all the legs.

There's the before and after. The first picture is very slightly lower than the original stance, the second picture is the new stance. As we can see the platform stands much taller in this configuration. It's about a ~5cm difference in height we're going to verify that now.

So with the new standing configuration the platform has about a 17.5in ground clearance. The old is about 16.5-16" ground clearance which is about 2.5cm and the rest would have gone to slop. We can actually make this even higher and then measure where the CoG would be. Let's make it about 0.52!

See this one is a little bit high if we look at the front leg positions. So what we're going to do is then, leave it at 0.50 or 17.5" of GC. Now let's figure out where the CoG in this configuration is so we know what to do.

And we're going to figure out the actual location of the feet after standing up. So to do so we need to pick a point to be the origin for measuring, and we'll mark that off and the x and y axis as well.

In this configuration the CoG is about 12in from the marked front and the feet positions look like this:

So re-drwaing this so a human can see it...

So this shows us the weight is defiantly on the back legs not in the center as we thought. And that's OK because now that we know this we can start using similar ideas to CoG motion analysis to, using this as a center configuration, to make the dog walk. For now, it seems we can easily lift the front-legs up and move them around which is definitely OK for now.

Now let's talk about clearance, the feet are 21.5 inches apart from front to back that means the center is at 21.5/2 = 10.75in from the front, which means the CoG only needs to move 1.25 in (3cm) forward before the platform becomes unbalanced and falls. This also means, for the back legs, we'd have to move the feet at least 3cm backwards before the CoG is within the parameters where we can lift up the back legs.

Most of the problem is being caused by the fact that we have an asymmetrical leg design. Had we gone with a symmetric one, the back legs might be easier to lift off the ground.

Knowing this we can test just lifting a front foot off the ground by simply rotating the knee joint.

Yea so I think it's a good thing to probably give the platform break. The jetson may have over-heated and shut-down. Let's give it an hour or so. Maybe even a day!

#updates #omgrobots #chip_in_for_good

3 views0 comments
© copyright 2019 | aditya mehrotra