Animatronic Dragon Project

The first step was to do copius ammounts of research on dragons. I looked at various sources online, renderings, graphic art, and about 5 or 6 "B movies". The next part of the process included deciding on the various movements the dragon would be capable of. The list below is what I decided would be a "full featured" dragon.

• Full Eye Movement (Left, Right, Up, Down)
• Eyelids - Perhaps two layers (like cats?)
• Articulating Eyebrows
• Nose Flare
• Lip Flares
• Jaw Open/Close

And of course he needs to breath smoke through his nose. If I could afford the insurance to have him shoot flames, I would...

The second step was to create a full size clay model of the dragon's skull. It was easier to make the skull on top of an existing plaster bust. This model will serve as the base for everything to come, it will be cast and turned into a fiberglass shell. The fiberglass shell (see step 8) will then house the microprocessor and all of the electronics that are responsible for the movements. I'll then use this as a base to build the dragon's skin.

This part was the most time consuming element of the early stages. The skull had to be proportionately accurate: all the electronics and motors responsible for the dragon's 14 unique movements have to fit inside this space.

Once the clay model was finished, I cleaned up all the surfaces to prepare for the casting process. At this point I took a pair of calipers and a tape measure and collected measurements to create a scale 3d model in CAD software. This will be used to help arrange all of the servos and hydraulics in the most space-efficient manner.

Here I have the prepared version of the skull, ready to get siliconed. I will be creating the mold in halves so this is the dragon's right side.

Once the skull was ready to accept the silicone, it was mixed and poured. Here I used 2 pounds of GI1000 with an ultrafast catalyst.

The brush was used to coerce the silicone back up onto the higher parts where gravity was fighting against us. Fortunately the ultrafast caused a pretty rapid cure, so I only had to babysit the silicone for about 10 minutes.

As the silicone cured, it was wrapped with plaster bandages and allowed to fully cure.

Once the right side cured, he was flipped over and siliconed from the other side, this time it took 3 pounds of GI1000 to adequately cover the larger surface of the left side with the barrier wall removed. Notches were cut into the exposed sides of the silicone to allow for reliable alignment. Using a tiny bit of vaseline to coat the other silicone surface, you may make it easier to prevent sticking and split the halves; using too much vaseline, however, prevent the silicone from curing.

Image to the left shows the dragon head fully siliconed and bandaged.

After the plaster had set the clay wall was removed and the mold pieces were split. The clay model was removed and placed aside in case I need it again. The plaster should be easy to remove while peeling the halves apart. The layer of vaseline prevented the halves from curing together, the notches that were cut allow the two halves to align properly for molding. Once they were split, I cleaned the silicone exposed to the interior of the mold.

After the silicone was cleaned, I began working on fiberglassing the skull. Through some great miscalculation on my part, I ordered the slowest-setting hardener possible. What should have been a thirty minute process became some horrendous two and a half hour ordeal. Fortunately I was working in the spray booth, and had the ability to put the room into "cure mode" and crank the heat to about 120 degrees. This took a bit of an edge off the duration of the project.

Fun fact: It's important to be in a well ventilated area and wear a respirator while working with fiberglass resin.

What-should-be-obvious fact: Using the proper harderner makes this process run a lot more smoothly.

Once the fiberglass halves cured they were cut to match, sanded down, and fiberglassed together. About this time, the correct hardener had arrived in the shop, this step was sooo much faster! Hooray.

The fiberglass skull was cleaned up and checked against the main model for accuracy. After a few minor modifications, I was ready to start cutting him apart.

At this point I started cutting the moving pieces out of the fiberglass shell. The jaw was seperated and holes were cut for the eyes and eyebrows. I also cut an access hole in the head to get to things inside. Meanwhile a steel frame was being prototyped in the machine shop.

The steel jaw was placed in the mandible and I started laying out the 7 heavy duty servos that exist in his head- only to find that I didn't leave room for the eyes. Some servos will just have to live outside the head. Holes were punched in the metal jaw and corresponding holes were placed in the fiberglass for mounting the metal jaw to the skull.

The metal jaws were attached together and placed in the fiberglass skull using machine screws. Here, I made sure the jaw opened and closed to my liking. The head would need to be mounted and have an articulating jaw before I could skulpt the teeth (to make sure teeth weren't bumping into each other).

The teeth were sculpted right on top of the skull in the top jaw first. I used your everyday off the shelf FIMO oven-bake clay. I took care to make sure the teeth would close over the bottom jaw and sculpted them right on the exterior of my dragon's "skull". After making sure the mouth would close all the way I placed the top portion of skull in the oven.

With the top teeth, the jaw was reconnected through the convenient holes in the gums. The bottom teeth were then sculpted in the negative space created by the top teeth. Here are both side by side, the left showing the painted teeth after baking; the right being the lower jaw prior to baking and painting.

The teeth were baked at 375 degrees for about 30 minutes in the oven. Since the skull was made of fiberglass it took the heat fine- it was here that I was glad I didn't use vacuuform or thermalplastics to shape the skull. Unfortunately the fan on the oven caused the teeth to settle inward; a few minutes with a dremel and some dental work solved that problem.

I had some concerns about breaking the FIMO so I added a few layers of fiberglass resin over the teeth to give them strength and a cool slobbery look. Alas, some teeth are still breaking from time to time, more resin and a fresh paint job will come after the mechanicals are all done.

Pictured to the right is the concept for the eyeballs, thanks to photoshop and google images. I sent this over to Jim at Tech-Optics in Newport Beach, California. While the artists out in Newport went to work, I created a temporary pair out of 1.5" diameter wooden balls. Using a drill press, a hole was punched through the center, and one side was notched in the size of the iris. This allowed me to work on the mechanics and I was also abled to exert what would otherwise be an eyeball shattering amount of force on the metal fittings.

The various eyeball mechanisms I tried to implement failed, one after another, so I finally decided to go my own way using my own design. The eye table was cut from UHMW plastic on a ShopBot CNC Machine. The eye will pivot right and left on an internal skewer, moved by the servo behind it. The entire assembly pictured to the right will pivot up and down at the eye's centerpoint to gain up and down movements. Eyelids will be attached on the exterior of the holding unit to prevent them from pivoting with the eye.

There were a few problems with the mechanism above. The material was too flexible, the servo spline didn't mount inline with the eyeball, and the shaft ended up protruding above the eye. I decided to start from scratch, which resulted in the picture to the right. The platform is designed for a Hitec HS-322HD servo, placing the center of the spline inline with the center of the eye, hopefully eliminating some weird side torque. I've also added two hinge points distal to the eyeball and gave the eyelid mechanism it's own mounts. This should eliminate the issues I had with the up/down movement interfering with eyelids opening and closing. The whole platform will still move, but a servo fixed to the base of the skull will control the open/close of the eyelids, this should allow for the eye platform to move independant of the eyelids. This was then machined out of aircraft grade aluminum on a CNC mill, hopefully killing the issue of flexion.

Without the eyes mounted, there wasn't much to be done inside the head. The arrangement of servos to control the eye depend entirely on how much space would be required by the mechanism itself. In the downtime while waiting for materials to arrive (aircraft grade aluminum takes a while apparently) I decided to work on the neck mechanism to support the head. I knew we'd need to move some servos outside the head, and I knew there would need to be something to hold the head up; I figured a good neck to support the head could also support some servos. The time to design the neck was fast approaching. Here's the whole of the neck mechanism. The black box at the bottom is a computer power supply I found online for $15. I'll go into more detail on the power supply in a future section. Near that is a terminal strip where the connections to the 'brain' will be made.

After going through a few iterations of neck designs (and realizing I had some downtime waiting for parts) I decided to give the head a single of axis of movement at the neck. You can sort of see the design in the drawing above, but here's a closer look. There's a fixed gear attached to the steel neck support. That gear is then connected to a swivel plate. The other (spinning) side of the swivel plate will be connected to the head. Also fixed to the head is a high torque servo gearbox with a second gear. This gear will engage the fixed gear on the neck and turn the head around it.

Pictured to the right is the beginnings of the neck mechanism. I used a four inch aluminum gear spaced about a half an inch from the turntable and the neck bracket. The bolts protruding from the top of the gear are the fasteners that attach to the neck bracket pictured in step 23. I found that the spacing causes the turntable to collide with the servo gearbox if you use less than a three inch gear (like the two-inch one I ordered). I had a three and a half inch delrin gear sitting around and rather than ordering a new aluminum gear I decided to just use the delrin gear on the servo.

Heres a crappy shot of the welded neck bracket I took in my living room (I left my camera at home). A coat of primer and black spraypaint and this guy will be ready to be mounted. There will be a 6" swivel plate mounted to the bottom of the neck to make working on him easier. The swivel plate is then mounted to a piece of wood.

Just to make sure all my calculations were correct, I made a simple motor mount out of some pine I had sitting around. Pictured is the mockup of the motor mount and attachments. The gears interface really well and theres plenty of strength so far. Everything worked well, the pine just isnt sturdy enough to do what we need so this piece will be machined out of aluminum as well.

After assembling all the pieces, the head was mounted to the neck mechanism and fired up with a battery and servo tester. I'm not sure how old the battery was, but I'm starting to worry about weight of the head once its loaded. To the right is a video that shows the progression, models, and the final product.

The cost of a 6VDC power supply off the shelf is several hundred dollars. The cost to engineer one or two after purchasing components and printing circuit boards is also close to $200 plus the countless hours choosing parts, designing and assembling the pcb. The cheapest solution was to purchase a 5VDC powersupply. An ATX computer power supply can be as cheap as $5 used at your local thrift store. The trick is to find one that doesn't have a lot of circuitry dedicated to protecting from spikes in power consumption- ie to find the best crappy power supply. I purchased a 580w black power supply thats capable of providing 45amps at 5vdc for about $20 new online.

ATX power supplies tend to shutdown if there's not a sufficient load (ie no computer attached). They are also switched on by the motherboard. There are plenty of online tutorials on how to convert an ATX power supply to something else, so I'll be brief on the details. The basics to modifying a power supply are sinking the power-on wire to ground, connecting a 3.3v wire to the sensing wire, and placing some sort of inductive load on the power supply. For the inductive load I used a 10 watt 10 ohm resistor. Pictured to the right is the power-on wire soldered to a ground wire.

When done with the internal modifications I grouped all the wires of various voltages together, rerouted them through braided sleeving, and attached spade terminals for the terminal block.