Svetnikov's Scrapyard

[Kieve]

Between Career missions and baiting the Kraken with tasty ship-shaped kibble, I try other experiments. Sometimes, they even work!

More often than not however, they end in explosive failure or hilarious Kraken attacks. Or both.

Here you may see the results of my ferret-like attention span.

Remember kids, R.U.D.'s are B-A-D!

*That's "Rapid Unplanned Disassembly"

---

Modular Cities:

As if a new career run and building a supermassive aircraft carrier for Laythe weren't enough, I got inspired by JAD's wonderful Minmus Acres project, and have added city-building to my list of Career-Mode to-do's.

Javascript is disabled. View full album

These are some prototypes I'd worked on in Sandbox - each "block" is a structural baseplate, fitted with B9 landing gear (with the motors and steering enabled), and docking ports mounted 2x per side. The blocks are controlled independently by OKTO cores, and can be linked together in any configuration.

However, even with the sturdy landing gear, weight does play a part. The massive City Tower module is heavy enough to push the wheels down, and cause docking trouble. In my test above, I was only able to secure it with one port.

Also as a rule: never retract the landing gear, or your city will become lodged in the terrain (particularly the heavy blocks).

The intent is to use EL and Karbonite to build them in-situ, and with the B9 gear they can be driven to more scenic locations for link-up, if the terrain is navigable.

Still working on the idea, so input is always welcome. I've been considering lately what it would take to adapt the Civilian Population mod to work on a smaller scale using stock-alike parts (such as the Stockalike Station Parts pictured above). Ideally the end result is a fully functional expandable Kerbal City, for any world with the resources to build one.

Edited March 26, 2015 by Kieve

[Kieve]

Walker Experiments: (The first of many)

This time I am squarely pinning the blame on Fengist for what follows.

I decided to try my hand at IR walkers. Why? Because:

1. They're awesome. I mean, do I even need to explain this?

2. I have a habit of setting stupidly impossible Kraken-bait tasks for myself. Adding robotic flailing deathtraps to the list just feels natural, at this point.

So let's begin with the Hex.

Now right away, I'm sure you can already tell this is going to end badly. We'll get to that.

But why start with 6 legs? Level of complexity. Bipeds may be the coolest, but they're also by far the most difficult. Balance, center of mass, momentum, it all gets a lot trickier when you try to do it on just two legs. I mean, look at all the falling down Kerbals do! Plus I have a Hexbug sitting on my desk, and that helped provide some reference.

However, Hexbugs aren't known to attempt a backflip, and explode.

Even Jeb seems mildly concerned. On the emotive meter of most Kerbals, that would be "Terror-induced heart failure."

Yes, that's right, I created something that makes even the great Jebediah Kerman worried.

At this point I don't even know what's exploding. It's just kind of a big flailing fireball of failure. Let's move on.

---

Meet Hex's 4-legged counterpart.

Jeb: "We're doing this again?" */gulp*

Yes Jeb. Commence the flailing!

So, this one actually managed to move around without excessive combustible failure.

Granted, not in any particular, desired direction, but it did move!

Okay, more like shuffled. Randomly.

It was like watching a chihuahua have a siezure.

Then I noticed why: The leg joints were locked up, for some reason. Not certain where I went wrong in the assembly process, but those washers with the red stripe on top?

Yeah, those should turn. And they don't. So the leg just gets stuck vertical, and doesn't pivot as it should. I'll have to go back and spray some WD-40 on it I guess.

Let's go smaller.

---

Disclaimer time: I modified the tweakscale file for IR.

My interest in robotics has tended towards the miniature scale, rather than behemoth-sized (I love Battletech, personally, but I'm not planning on recreating a 75t Timber Wolf / Mad Cat in KSP).

Anything useful that comes out of these experiments is liable to be a compact probe rover, designed to crawl over difficult terrain where wheels would just slide. So, pretty much what NASA would use them for.

But getting back to my previous statement, IR's settings for tweakscale are... weird. They don't align with your set of 1.25m/2.5m/3.75m scales, and the smallest they go is .75m. What scale is that, even?

I dug into the .cfg for it and shuffled some numbers around, added a few settings, and now the docking washers line up just fine at fuel-tank scales, plus a "Small" .5m and "Micro" .25m setting, for really tiny work.

It hasn't helped with this tiny elephant however. It behaves even worse than the four-legged Hex, just wallowing around and doing a whole lot of nothing.

I built something vaguely similar to this with my old Lego Mindstorms kit, and actually got it working. I wish I'd kept it together because now I can't remember what I did to make it functional...

Anyway, let's have our last failure for the night.

---

This is actually just a test of a test.

I wanted to work out some foot-ankle-leg dynamics, and slapped this together. The washers in the lower-rear of the leg are powered, and meant to tilt the foot as the walk-cycle progresses.

And it worked, too. I mean, it was bolted down tight with the docking clamps, so all it really could do was wave the shin-piece back and forth.

That seems a shame. So I decoupled it.

Now this may look like a failure to you all, but it actually gave me some design insights.

Y'see, the way the two halves of the calf/shin framework were moving, they looked very much like legs already in stride.

Of course, it was laying on its side going absolutely nowhere at the time, so that's not particularly helpful in this case. But it's made me reconsider my notion of where to place the powered cams, and how the entire leg operates in relation to the walker's body. Legs are the heavy bits, requiring the most structure and support, as well as doing all the actual work. Building the motive cam into them, rather than the relatively lightweight body, may help attain a stability that my attempts have thus far lacked.

[Fengist]

If it's any consolation, you're going through the same exact painful process I did. First, thinking you can just build one and second, realizing it might be a good idea to put it on a test stand.

Right now, one of the biggest problems I see with your designs is something I'll get to in the not too distant future in my post, the feet. The angle of attack for your toes is probably too steep. Take a look at both my designs and those of Parallax again. You'll notice our 'toes' are more horizontal than vertical. Any time a landing strut angle is anywhere between vertical and 45 degrees it tends to grab the ground (braking action). If it's on a forward toe and grabs the ground, you instantly pitch forward. Too much angle (too horizontal) and you'll end up sploding the strut when you put the foot down, or even more fun, sploding the runway and going nowhere in the process.

In order to get away with a steep angle strut, your leg motion has to be really good and the foot has to 'rotate' through so that the heel strikes first and then the toes. Take a look at the 'advanced' design in my tutorial. That's an example of what I mean by rotating through. If the toe hits first, you will come to a screeching halt. I've even flipped Octo end over end by not having the angle right. The reason Octo has hinges on the ankles is so that I can adjust the angle of the feet depending on my speed and direction of travel to make sure the heel hits first and the toe stops it from falling forward without becoming a brake. The heel angle is just as important because at some point, you're going to want to go backward and it'll have to act as a toe.

Your 4 legged Hex, if it's just shuffling around, change the toes and I think you'll find you get propulsion.

The last piece of advice, I didn't see any reaction wheels, or not a lot. They're pretty essential for even multi-leg walkers. Since designing a ball-and-socket hip join in KSP is a nightmare to even think about, turning your walker will have to be done by either reaction wheels or some other force like RCS. Even Octo has 4 rather large reaction wheels to help it steer when moving.

Hope this helps.

[Kieve]

Walker Experiments:Cheezebot

I was on a voice chat with my girlfriend while working on this, and remarked that he resembled a cheeseburger (this was before adding the legs).

I'd originally planned to do a single B9 light for an eye, but after noting certain similarities, I went with something a little cuter.

Practically nothing here except legs and reaction wheels.

This was my experiment that followed the last set of walker images. I honestly didn't expect my attempts at a biped to function at all (not that I expected much better from the previous multi-legged attempts either).

To my astonishment however, Cheezebot actually walks pretty decently - if only in reverse. He has a "falling down" problem though.

But with the feet mounted on powered hinges, he's got little problem picking himself back up again.

---

So what did I learn? Well, for starters, I may as well just dive in with biped bots. Yes, they're more complicated, trickier to balance, and more difficult to control, but there is one big thing in their favor: I understand them better.

I played with robotics, and LEGOS in particular, a lot right through highschool, and spent my senior year designing a cam-operated biped with Mindstorms tech. As far as walk-cycles and leg motion are concerned, I have a big head-start in that area. I may as well follow through.

Also, I managed to completely flub one of the key design elements I was intending to crib from - in the previous walker leg, the "motive" washer is between the Knee and Ankle sections, and provides a fair amount of torque without running the risk of "flipping" a joint (when the cam shafts flex out of alignment enough that the leg rotates past its intended range of motion). You can see that in the last pic of the previous post, in fact - the far section has flipped and is on the wrong side of the test foot. Cheeze has this issue pretty bad, and a large part of that is because the powered motive washer is located directly at the knee.

As for the "walking backwards" thing, I've a sneaking suspicion that a lot of it has to do with the phasing in the walk cycle. Right now, one washer and one cam linkage powers both the upper and lower leg, and the result is a force that pushes to the rear. Weight balance is also a factor - the legs are heavy, and most of the weight falls behind the foot rather than in front of it. As Fengist pointed out in his biped tutorial:

Once upon a time, I read somewhere that bipedal walking was the art of falling down and not getting hurt. Each step you take you essentially fall forward. Before you land on your face, your other leg moves forward and stops you from falling, it then, makes you fall forward again. When designing a walker in KSP, that's exactly what you have to achieve. To fall forward, stop, then fall forward again.

Cheeze is currently balanced such that he's always falling backwards.

Time to head back to the drawing board...

[Fengist]

I love that look! I never thought of putting the gyros on the legs... brilliant. Or, for that matter, the rotors on the knees.

One thing your design is going to prevent though is more speed. One of the reasons I use the pistons on the lower leg is by extending them, I get a much, much longer stride. Octo normally runs at around 10-12 m/s. By extending the lower pistons, I can achieve 18+. The same was true with the Mun Runner. If you re-watch that video, you'll see it at one point making 20+m/s with upper and lower pistons extended, and that was on Kerbin. I was hitting 22-24m/s on the Mun.

You may want to look over the IR parts. There is one piston you can bury inside the leg so that it only sticks out a short bit.

Dunno if you've tried but the larger landing struts may give you more extension on your heel and keep you from falling backward. Both work well enough, but I've blown up a lot of the one's you're using just stomping around on the runway.