tom_hampton wrote: ↑
Sun Mar 28, 2021 11:57 pm
HA! When I was in college....waaaaaay tooo long ago (how did that become measured in decades?) I was co-lead of a multi-disciplined (6 each: ME, EE, SwE) project to design a 6-legged robot sponsored by NASA. It looked a lot like that (except with 6-legs). Our lesson's learned at the end of the report was...a VERY similar list. Mechanical power is hard....electrical power is harder...and getting 18 college students to all pull their weight is impossible.
I had similar when I did a masters over a few years outside of work as a post-grad... the group project was a nightmare. Once the others see that you're motivated and want to get a good grade to help your career, they stop contributing because they have other areas of their study/life to focus on. That really sucked for me.
And yep... powering this sucker is hard. I've blown two servos so far, and only had two spare, so pretty nervous about breaking any more (getting them imported, the postage is three times the price of the actual servos!)
m.cavallerin wrote: ↑
Thu Apr 01, 2021 9:13 pm
Wow, amazing! Could you upload somewhere (You tube?) a video while PiRantula is moving?
Lol, nope! Currently blown out two of the servos! I need to refit them and take a massive amount of care not to blow any more as I've no backups left!
Basically, lifting 5kg is fine. Just... if one leg is actually pushing down more than the other legs... you can't see that it is... until magic smoke appears and your servo is destroyed
I've just had some current sensors arrive though, so I'm thinking to hook them in (though bit unsure if I've enough analogue pins exposed through the servi hats). Thinking that I can have scenarios... standing, limit servos to 2A each leg... walking, limit to 3A per leg. To try keep it from burning out again. Idea being to limit and get an even spread from the feedback.
freedman wrote: ↑
Fri Apr 02, 2021 6:06 am
Have you tried spring loading the legs so the servos don't need to work as hard to lift.
I remember something about spiders; I think they move their legs up and down by filling/releasing them with air, the legs are like air bags. They can stand in one spot and use zero energy.
That unit you made is cool and scary.
That's a really interesting thought!
It would totally make sense to figure out a standing pose... and have it spring loaded so that's the equilibrium. Then movement is "just" the delta from the natural position. It'd add some compliancy too, allowing it to have some margin for error on moving.
I think I'll have to have a bit more of a think on that. There might be enough size on this thing's legs to add some springs... though I've zero understanding of that side of things. That said, even if it only partly asssists... that'd be less kg that the motors are pushing against... so all good... mmmm. Plenty to think about
And yep... it scares me turning this mofo on!
Thanks for the link and thought.
I'm very very much an amateur with this whole modelling and electronics thing. It's definitely not my job, nowhere near. 29 pages of discussions about how the MBDyn library works and learning of it... I think is pretty far out of my understanding and available time to do so currently. I think that kind of research would take me 3 months to get even a glimmer of understanding... whereas I'm currently thinking of focusing on ROS next (actually trying to make a slower version of this rover
from thingiverse as a platform to try learning a bit more electronics, coding, SLAM, vision, depth-scanning, linux, python and ROS)... quite a lot to learn!
The movement of the digger arm is quite interesting though. That's kinda what I've ended up going for... after working out that it's all about the radius from the center of the servo... if you can bring that radius in... so bring the legs in... then stand from that position, it's a winner! I did these awful calculations to determine that it really needs to bring those toes in to be able to stand...
gbroques wrote: ↑
Fri Apr 02, 2021 9:40 pm
What was the design process like?
Why model it in FreeCAD?
For example, did you create the entire model in FreeCAD first to flush out what you wanted to do physically?
Or was it an iterative approach where you designed some in FreeCAD, built a little physically, design more in FreeCAD, repeat?
I really enjoyed designing it. It was a learning exercise as well as something fun. Learning about ASM3, and starting to give a bit more thought to my filesystem for re-use of my sub-components in future projects was pretty frustrating... but I'm glad I got through it. It's great to only have to draw a model once and get it right, then pull it in as an assembly and base your geometry off it, then you know it'll match first time.
FreeCAD... because I started learning F360, and just as I was getting to have some fun with it, they added the new licencing. And even though I'm just a hobby guy... I don't want to be paywalled when there's a free product just as good for my purposes. Plus, using F360 today is assuming they don't up the price or imposing more pain on hobby people. FreeCAD takes me out of that completely, no ifs or buts. Safe.
Process-wise... my main concern was making sure all parts would fit on an Ender 3 printer.
Second to that was strength... I didn't want it snapping. My first robot was an OpenCat clone, similar to this
(original now 404's). First thing it did on standing was rip itself in half. That was a sad climax. I've rebuilt the lil' sucker, but the motors are tiny and it was too delicate. So I gave PiRantula a 10mm wall, and 5mm base. Should be one heck of a tough cookie.
Third thought was that I want it to be able to take a payload. So the top panels have enough screw holes into the walls so I can add new ones with extra modules to the top. I'm thinking a Pi4 and something for SLAM. Maybe a LiDAR or this depth camera
which I bought a couple of weeks ago.
I was a bit nervous about the shock of slamming legs down, so there's a toe made of TPU, and a 1mm TPU shock absorber where each leg joins the body. Not a lot, but could prevent a repeat OpenCat incident. The toe being insert-able also gives me an option on making them bigger in future if need be.
So yeah, just drew a set of shapes, split them into assemblies so I could space them against each other (just to learn about ASM3 really there), and then tried to figure how to jam the electronics into the design.
Next iteration looks to be including current sensors, so no changes to the model needed just yet. Just... more... wires. Oh god.