[MagnusWootton]

So this robot, is the one im going to build,   this is just an open s-cad render procedural generation cad script and it knocked my socks off how good it turned out!

Its a hydraulic one,  im not sure if my pipes and actuators are big enough,  but I'll be doing a test actuator first, (just the leg with a single actuator) this one is nearly done, just a bit more work to go before I go try to build it.

If I had access to a 2 metre 3d printer I could print it out pretty easy, (the body is only 20 centremeters but the legs are just too long) but I dont, so I've actually got a technique organized which is going to use a large big metre 2d print instead, which just mostly is just the pipes and actuators and then  build the shell off it from there,  in a top half and bottom half that just mirrors together,  should be pretty similar,   I just mainly need something that holds water, cause thats what hydraulic robots do,  just push water from the body to the arm.

So you can see the leg sorta looks like a snake, thats because on the 2d print its just flat straight out with the leg, but I actually bend and twist it into position afterwards with a heat gun and some extra braces im going to make to get it exactly the right angle, and thats how I get all the axi I need off a flat bar,    I'm not sure if it works that well for a humanoid, but since a spider is a pretty scary creature the extra twistiness to the legs I dont think makes it look less spidery, it sorta works for it.

It has a cool hinge configuration,  cause a) it can retract its leg closish to its body,  and b) its roll and yaw hinges interchange for each other as the pitch hinge changes angles.  so its always got access to yaw and roll,  but it swaps what hinge it uses for them depending on the prior pitch angle.


So a bit more work to go,  but I think is going to be the one!  7 years or so after I first wanted to go do this with lots of missing pieces in the jigsaw puzzle.

[MagnusWootton]

I did some walking tests,  and it needed a reconfigure,  I've already got it lookin' nice! and I dont want to ruin the look of it,  but the walking was pretty bad (but still showing a little potential)  So this is the reconfigure, and ill have a walking vid tomorrow.

I'm going to be getting my print done,  but maybe I need 2 weeks before I do it.  this is the first time ive built anything this professional!  Im really excited.

All this extra effort has gone into something important,  I need retractability of the legs so it can make it through tighter door-ways,  if it cant fit through the door it cant protect me inside, and thats usually where you invited the vampire in,  so it needs to be in-doors.

[MagnusWootton]

Its not right yet, theres some improvements, but the new configuration works alot better,

the limbs are fully retractable, and it can walk in its retracted state, so it makes it through doors easy even tho it has a crab like posture.

It took me a while to get the leg ordering right in the movement,  its not a gallop exactly,  its like a gallop crossed with a walk,  so its not pearling around 1234,  its actually 1324, slightly different, to get the two front legs 180 out of phase, instead of 90.  and the other thing was to get the sides swapped between front and back, so its not landing the same side at the same time.   then it looked about right.

but my unit isnt going to be using IK to do its motions,  this is just an animation test in open s-cad.

[MagnusWootton]

so, heres the fixed one.       it wasnt quite what i was expecting, it seems fairly monstrous!!! (like a little too freaky for my liking)  but I dont know what to do about it,    The project is still a go, It's good enough better not complain.  having that full retractable leg makes it walk a little funny, but its not too bad,   ill have to see what the computer comes up with,  these walking animations arent used in the finished thing!

[MagnusWootton]

Here he is,  with the kinect and jetson nanos and arduino placeholders there.



also in the body, is alot of rechargable batteries, one per actuator,   its one battery and hi-voltage generator PER actuator!  I need to make sure I have the amp requirement! (even tho its electrostatic i do have to hit the microamps or it doesnt work!  Actually I dont even know for certain but best be safe and sure.   and then theres a big circuit board full of high voltage switches I'm going to drench in timber oil to get a nice hi-voltage seal on. (if a spark plug gets oil on it, it actually cant make it through!)  and thats all I need,  but with only 22cm cubed to put it in, its going to be a tight fit to get it all in the little box.

[WriterOfMinds]

Your bellows actuators look to be grouped in pairs - is each pair associated with one joint or two? How many degrees of freedom does each leg have total?

[MagnusWootton]

I dont know if this is going to work,   I havent even built the test actuator yet. 

There is two actuators one for pushing left and one for pushing right, it goes 45 degrees each polarity for a total of 90 degrees.    so its only low flexability, I think I can get the robot to walk with only low angular freedom, if I'm flexing plastic to turn it I thought I'd better keep it down for this first one,  more chance it's going to work.

The animation I made doesn't express the full flexability of the system, it looks a little short stroked but thats cause I didnt do the animation properly,   it has more flexability than whats shown.   but it is just 90 degrees total for each leg,  I thought if I keep it down its got more chance in working.

So I dont know if its got enough power to just push its way back through the other "vein" back into the other actuator,  if it doesn't,  I might have make it pull as well as push but then the internal resoviour is alot harder to make push and pull, so I'm trying to get away with it.

[MagnusWootton]

so here i've done up the head a bit more to match the legs. (and the stuff is inside, fitting.) - sortof, you can see it poking out the sides and bottom, im just gonna go put some lime green caps on it and call it done, for now.




you can see the kinect mounted vertically in the middle of the front of the head.
The problem with that is i have the horizontal res as the vertical, so it means it sees more up and down than left and right,   I guess that'll make him have better feet detection for the pose detector, so maybe it might not be that bad.

So its just too many pipes for the double actuator system, its 12, and they all have to be in a line across the bending plastic to keep the volume the same throughout the angle change  and they are just too skinny,  they are only about 2mm pipes and the resistance is all the way to the actuator AND back... but I can get it to 6 if I make it single actuator, and the path is only down the leg once. so Ive reduced it to one and am going to implement suction for the reverse actuation.   Ive worked out how im going to get the suck energy, its a bit of a bitch of a build but  i should be able to do it.

So now I just have to organize everything nice and snug inside its head, I think I have enough space,  I gave it a fairly eggy head so I could fit it all in.


<edit>


here he is with the caps.  he looks like hes got decoy eyes! q:)

</e>

[MagnusWootton]

so I've got him in the physics sim, ready to breathe life into him.
There was a million bugs to tackle to get this to happen cause im pretty much coding it all cross wired, but just managed to get it going in about 5 days from the last post.
this is just random motor movements, sorta looks like hes struggling to get off the floor but its just your imagination.
Ive got to fix the physics first before I get the motor search going,  his legs are going through the ground - needs a few fixes then off I go.   I have a really good rtx 3080 here I can use, (My little bro is staying over atm and I get to use his sick video card.)

[MagnusWootton]

Some modifications to the spring system,   I put alot more momentum into it now,     if u just leave it freewheeling it develops oscillations that eventually make it rip itself apart,   you need to add a form of breaking to the springs to make it hold together with adding as little viscocity as possible.  (thats the tricky bit.) this is 0.5 viscoscity so its not that bad, I havent added too much and its still holding together,  but if its less viscoscity then it loses control and splits apart.

Its still penetrating the ground with the "physical body"  (being the model ive attached to the spring model) so I need to fix that yet.    when the springs hit the ground they tend to bend against it and the "physical body" ends up going out of synch,  I think if I make it more taught against the ground by putting the ground verts in precedence in the spring update order it might fix it.





<edit>

I managed to stop it going into the ground, and the sticks are more stuck to the body better.
But now it springs off the ground with the slightest tap of its leg.     but once ive fixed this its probably good enough!




<edit2>

actually come to think of it,  I better work on this some more,  it got better plus more bugs.   so I might just go improve the aesthetics a while then Ill get back to the physics a bit later.     soon as the physics done,  its onto the gait search system!




<edit3>

fixed the render!!  looks just like the openscad render!!!
just showing off my global illumination hack,  I do it as a post process,  lifts all the shadows, and Ive added a terrain.

[MagnusWootton]

Just learnt ALOT about local raytracing.    check this out.



This thing doesnt work with ordinary diffuse lighting,  the whole thing is actually specular reflection,  and its really horsepower intensive,   but I can run a bigger window multisampled tomorrow on my lil' bros rtx3080.