Botty v2 — Bailing Out
Failing is only the first step to success.
This is a story about failing.
Well, not so much a story so much as it is a retelling of events in process.
Anybody can tell you about their successes, and given half a chance they’ll do so at agonizing length. It’s much rarer that you hear about failure, in part because everyone is terrified of failure even though the very act is the grounding process by which we learn. You don’t learn from getting it right; the only thing you learn is that you have learned. Education comes from failure, often repeated failure, sometimes embarrassingly repeated failure.
I’ve been pursuing “the art” of 3D design for over a year now, cramming in a good chunk of the last couple of centuries of mechanical design as I go along. As an ex-software guy, that’s a pretty big jump. You can blame the increasing prevalence of 3D printers and the decreasing cost of outsourcing 3D prints for that interest. Along with a lifetime fascination with tabletop wargaming, a hobby tailor-made for an obsession with little plastic figures.
I’ve written multiple articles about things that I’ve pursued and things that I’ve learned along the way. Some of them in agonizing detail. Others with just a light gloss.
I sent out about a week ago to design a robot. Not just any robot, but something ridiculously poseable. Something sort of inspired by my girlfriend’s ball joint manikin art doll, but more mechanical and — dare I say — slightly more insane.
When I get a wild hair to chase down some sort of design, I don’t sit down and go digging through “how someone else did it”, even though that’s the sensible course of action. I want to come to these things as a novice, even naïve, because the process of banging against the wall as hard as I can often turns out more knowledge than simply trying to replicate what someone else is done. It also leads to a lot more failures. As I’ve said before, I believe these things are linked.
Besides, if I had done even basic due diligence and research on the idea to start with I would have run into things like this:
I won’t lie, if I had run into that before I started I would have been far less interested in beating my head against that particular wall. How do you compete with that?
I encourage everyone to take a moment and check out the assembly videos on that link, in particular how it uses elastic bungee as a synthetic muscle fiber, keeping things tight and allowing high degrees of flexibility.
I would’ve never come up with that.
Which probably tells you that I probably shouldn’t be trying to figure out how to design poseable figures, given that I have absolutely no experience with them, right?
If you let that sort of thing stand in your way, you’ll never learn anything new.
So instead, I started with an idea based around a much less ambitious idea for a robot that could be assembled in any pose but which would require joints to be glued. Which is fine — everyone has to start somewhere. I had the basic, underlying idea that spherical joints — literal balls — would serve well enough to build some robots poseable enough to be used on wargaming terrain.
That led to wandering around over in this space:
Pretty nice, as far as it goes.
I was particularly proud of the design of the reverse knee and hip joints, reasoning that offset spheres would allow enough clearance for different poses while providing enough surface area for glue to hold the thing together. You’ll also note that it is digitigrade, in particular a chicken-walker. I’ve always loved digitigrade leg designs, so it came naturally wanting to design one of my own.
(You can’t see it in most of the pictures, but each of the toes has a spherical connection to the footpad as well. You can never go too far in the pursuit of posability.)
Problem: once I got to this point I needed to come up with a usable idea for a torso. You wouldn’t think that would be an issue, but while the foot and leg design fell out of my head really quickly, the torso just wouldn’t come together.
The first cut was really weird:
See, this is what happens when you give someone who is primarily inclined to model things with parametric extruded shapes and volumes access to a decent surface modeler. You get complete crap.
Fusion 360 has a one of the best surface modeling systems available outside of really expensive high-end modelers. Their technology is called T-splines for reasons which are obvious when you can see the representation of the model; the surface is split up into areas which are essentially rectangular in typology and framed by a mesh where three lines come together in the shape of a T. This requires a whole lot less line description compared to mesh modelers like you find in Blender, which describe the surface as a mesh of quadrangles, which then get reduced to triangles. T-splines are almost an intuitive system.
If you have even a little bit of artistic talent in your body.
I’m getting by on sheer bloody mindedness and a modicum of engineering ability.
Every time I started working on the torso, it ended up looking like a chimpanzee head with legs coming out of its ears.
Now, go ahead, look back at that video. Tell me it doesn’t look like a chimpanzee head with legs coming out of its ears. I’ll wait. I know the truth.
I knocked around on that thing for two days straight trying to get something that works better with the idea to come out.
No dice.
Eventually, after an embarrassingly long time, it occurred to me that I could literally go back to some of the original inspirations. I had already decided that whatever torso I went with would be balanced over the feet in that neutral pose so that it would have at least one guaranteed, straightforward useful standing pose. F360 has an excellent Center of Gravity system where you can tell it what bodies to analyze and it will update a point that represents the COG no matter what changes you make to it.
There were a number of digitigrade inspirations, but for an old-school geek like myself nothing beats the granddaddy of them all, the Tyrannosaurus rex. That actually started going in a better direction.
(If you can’t forgive me for using the environmental rendering system to put a robotic T Rex in the middle of an open field, you’re not the kind of person I want reading this thing. Go away)
On the positive side, I came up with a really nice off-cylindrical support axle to join the shoulders of the digitigrade legs. I am completely proud of that. And an oversized tadpole which leveraged the modern paleontological ideas on how large bipedal dinosaurs probably carried their tales rigidly out behind them for balance worked out pretty well.
The problem was that I ended up painting myself into a conceptual corner. It wasn’t a giant chimpanzee head, that’s true — but it was a design that really couldn’t go anywhere. It was stuck being exactly what it was.
That’s a terrible situation to be in.
In a struggle to try and get out of the intellectual hole I decided to do what artists and engineers have always done over the centuries…
I gave up, scrapped the whole thing, and decided to start from scratch.
That’s a hard thing to do. I really hate to give up on an idea. I have a dogged, obsessive drive to finish whatever I start. That can be a real problem when you’re trying to do something you’ve never done before.
I’m trying to get better.
So — version 2.
As before, I decided to start at the bottom, with the feet. But this time I wanted to pull together a little more detailed, more artistic design.
You’ll notice something very important: I decided that part of my problem was that the simple, technically unmoving ball joints were just not sufficient to my needs and I needed to design something from the ground up which actually functioned like a ball joint. So I designed one — with a plastic lock pin, a plastic cap designed to be glued on the end of that lock pin, and a couple of pins intended to be glued into segments to allow them to rotate.
(Does anyone see the problem here? It took me a day to realize what I’d screwed up.)
Whatever else is going on here, that foot is a thing of beauty. Faceted, highly structured, really cool looking…
If only it solved the real problem.
I had lost track of the original intent, to create a digitigrade simple robot which hinted more than it expressed and played to imagination of the viewer. I got caught up in making really cool toes — and a ball joint.
But the original problem was that I wanted to make something. I wanted to learn something. And what I was doing wasn’t a learning something. In fact, what I was doing couldn’t teach me anything except how to manipulate the F360 interface with more practice. Practice is good, and practice is a thing that we need to seek, but that wasn’t what I intended to achieve. I need a clear metric of success to feel comfortable.
This foot design could never be right because it could never be wrong. It’s the sort of thing that you can elaborate on for hours on end and never get “done”.
That was kind of depressing to realize, but it came with a moment of clarity. There was something in this design that absolutely had a metric of success, which could clearly be a testable idea.
The ball joint.
I spent over a week pursuing an idea which ultimately was wrongheaded from the ground up, but what came out was an interesting, even fascinating problem to be solved.
Tomorrow: we start solving.
