This is the third post in my blog series about building a real life moving castle. I’ve interwoven this text with videos – they’re there to deepen the subject. The text is written as a standalone, so you can skip them if you like.
What is the difference between a real life moving castle and an ordinary tiny house? I strongly feel it’s the house walking. There are other things to Howl’s castle of course, but the sequence where mr Turnip shows up with it in tow, ground shaking from its steps sets the tone. Studio Ghibli animated the castle as a character of its own, and like real life builders of walking machines they quickly ran into problems. Miyazaki wanted ten legs for it, and some of the concept art shows six legs. In the end that many legs proved to be too many moving parts for the computer graphics (the entire castle was created in gc), and they ended up using four. A bummer for anyone planning to build a real life version, because a four legged walking machine is tricky to get right.
There is no shortage on walking machines in the world. Most impressive is probably the dragline excavators used in mining – big and heavy machines dragging themselves across the surface (instead of using a “real” walk cycle). On the other side of the spectra are the lightweight strandbeests made by Theo Jansen. Built mostly from plastic tubes they dance across the beach where mr Jansen releases them. Somewhere in the middle is small toys like mechanical spiders. What all of them have in common is passive balance. That is, if you stop them they still stand, because they have enough of support points on the ground. The minimum of legs you need for that is six.
Four legged machines stands in a kind of middle ground between passive and active balance. It’s tempting to think of them as using passive balance, but it’s more true to say they can use passive balance. The largest walking robot in the world (as of 2023), Tradinno the dragon of Furth am Wald, is moving that way, always keeping three leg on the ground while moving one. This is awesome character design, by the way, with the dragon slowly moving forward like a living threat.
If we want a fast machine on four legs it has to make use of active balance. That is, most of the time it needs to balance on two legs. During a walk cycle a quadruped topples from one pair of legs to the other, either diagonally over the body or from side to side. On the surface; easy. In practice it takes active balance, ie a continouos adjustment of movement, to make sure the toppling doesn’t turn into falling, and many a robots has fallen on this test.
Robotics have worked for a long time on the problem with walking. The star position is taken by bipedal robots like the fictional C3PO and Mathilda Junkbottom – we’re dreamed about building our likes for a very long time. And the road to actually getting a robot to walk has been long and filled with failures. The first humanoid soccer leages at RoboCup were rife with unintentional physical humor, with clunky robots jerkily and slooowly walking up to a ball – only to tilt over once they lifted a leg to kick it. It’s no coincidense that Asimo was a world sensation when the robot first walked and ran on stage.
When it came to four legged robots the US military funded research through DARPA around 2010. They wanted a vehicle that could carry equipment in terrain where wheels weren’t an option. That’s what payed for the construction of Spots ancestor Big Dog by Boston Dynamics. In the clip below the robot demonstrates both the advantages of legs instead of wheels in terrain, as well as active balance. It’s not perfect balance wise – more like a drunkard on ice. The project was discontinued, and the military part of robot building seems to have been transfered to Ghost Robotics. (Do you too hear Isac Asimov rotate at warp speed in his grave?)
Ideally a walking castle should not jump around since a house interior rarely is shake proof. In addition active balance takes energy. Muscles, or in the case of robots pistons and motors, needs to be moved for the balance to work. I haven’t done the maths yet (frankly, I need to be paid for that) but for a big, heavy vehicle like a castle walking that loss would concern me. The more passive balance I can add to the walk, the better.
This brings us to the six legged walk. With six legs all in the walk three of them will always be at the ground – like a three legged stool. Three legged stools are easier to topple over than four legged ones, but they have one distinct advantage; they are never rickety. All items with four legs run the risk of having one leg too short, and then you get that irritating micro rocking that can drive anyone insane. If the item has a flat top level that area may be askew, but it’ll be stable. (This is the reason why a lot of lab stands are three legged by the way.) Gaining that stability means the legs can be locked in place and make use of passive balance while the three others move.
So, six legs pointing almost straight down from the body. This is a far cry from Howl’s castle that keep its legs straight out from the body in distinct angles. That’s the power of magic and animation – it works on film. In real life this would put the stress of the weight of the castle on twelve tiny points; the four “shoulders” and eight “knees” (each leg has a knee hidden under the lamellae ‘skirt’ close to the body). Hypothetically the knees are supported by the bottom legs, but to walk these parts have to be pendulum moved, meaning that support is frequently diminished.
Is that so bad? Yes, due to the many moving parts joints are the weakest part of a limb. Depending of what you chose – ball and socket joint, pivot joint or hinge joint – you have things that can dislocate or simply break because it’s a tiny peg joining two pieces together. There are some impressive materials out there, but for a castle of Howl’s size we’re looking for a wishalloy if the joints should last. The more realistic alternative is to make legs with as few joints as possible, and also make use of the laws of mechanics – ie. constructing legs not built as levers.
Surfing around YouTube to find illustrations for this post made the algorithm hickup the Hacksmith Industries Spider Mech (for once a useful sugestion). This is an awesome project and I want one to go groceryshopping! But for the sake of this post I have to say that they’re kind of combining the worst of two worlds. They’re using six legs and get that passive balance, yet since each leg has four joints – two knees, one shoulder and one wrist – the entire thing keeps rocking putting extra strain on mechanics and sensors. The six legs means the walk pattern is complicated, and each one must have it’s own programming. Since the legs points out from the body instead of straight down the weight of the platform and the legs is hanging on the joints and pneumatic tubes. They even managed to get the pneumatics to explode. The project was ended when one leg simply tore itself off the platform, and after a repair it tore off again, splitting the platform in the process. They look pretty crestfallen in the end, but they did reach their goal; they made the spider mech walk.
The Spider Mech has a similar walking pattern to Howl’s moving castle. But it didn’t have anything on top – that set up with a chair and a joystick is featherweight compared to the huge living unit of the castle. Its interior hides one combined kitchen/living room/magician’s office, one bathroom, and two sleeping rooms (Sophie sleeps under the set of stairs at the bottom floor). That would make it heavier than a 34foot tiny house, which weighs over six metric tons. It’s hard to find a better illustration on how important it is to not just calculate for the legs alone, but also the weight to be carried and the mechanical stress that will appear.
(If you want to know more about the process behind making Howl’s Moving Castle I recommend “The Art Of Howl’s Moving Castle” from Studio Ghibli Library. Still available in print.)