Last week I headed up to Viridis3d for some more hacking. We got some beautiful results using some vaccuum casting with the trefoil design, parts printed for both the internals and outer shell of the quadruped, and schemes for tempting new mechanisms. All in all it’s been really exciting seeing the progress. Also, I have some updates on controlling the trefoil tentacle with an arduino powered set of air solenoids.

One of the confounding factors in getting this flavor of robot moving predictably has been how difficult it is to control wall thickness and bubble inclusions when casting the final silicone pieces. Like almost every mold, you do the best you can, try to create a nice, sterile, well ordered universe, and hope. Although Dragon Skin has performed really well as a durable, flexible silicone, it has the nasty habit of trapping bubbles in inconvenient spots when curing. Many silicones have a thin, pancake syrup consistency when mixed, but Dragon Skin is much more like honey or molasses, meaning it’s really easy to trap bubbles in the mix while stirring and have them set in place when everything’s curing. A good solution for the problem is pulling them back out with a vacuum chamber.

Thankfully, my friend Kristen Stubbs has a really swanky vacuum chamber over in her space at Artisan’s Asylum. What I did was assemble the trefoil mold, fill it with the required volume of Dragon Skin, vacuum degas the whole assembly, insert the mold core, and repeat until all the bubbles were cleared. This yielded a really good tentacle that was almost completely free of bubbles. The downside was the process took a really long time and the bubbles never really fell like I was expecting. I was hoping that it would blow up, and then vent all of the bubbles in the mix, but instead it stayed inflated even as I pulled a really hard vacuum and I just had to wait it out until the bubbles popped. I might hunt around for another silicone if this process keeps giving me agony.

So, with a very nice looking tentacle in hand, it was time to start experimenting with robotic air control. I believe I’ve found a system that works in a pretty simple and straightforward way. It still needs some work when it comes to the programming end, but I think the mechanics are well sorted. The idea is to pulse air into the tentacle using a solenoid valve, and have a constant bleed on the line so that flex will entirely be controlled by how long the valve stays on. It’s sort of a low frequency PWM. I’d like to get this working using a visual interface in Processing but, given how little I program, progress has been slow. I’ve got a thread on Adafruit with what I’ve come up with. In the meanwhile, you might like to check a rough video of the trefoil inflating.

I initially thought that keeping the system air tight was going to be a problem, and it definitely is something we’re going to have to engineer in detail as designs get more complex, but since I’ve found a simple system for plumbing things and sticking them together without too many air leaks. My problem with many soft robot designs is they rely on adhesives exactly where adhesives perform their poorest. Many robots rely on a sheet glued to a cast silicone maze, turning the open troughs and channels into sealed bladders. Although this method is elegant in several respects, it relies on an incredibly thin layer of adhesive to hold up to greater forces than the surrounding materials. This is a guaranteed failure point and one reason I’ve been pushing so hard to develop robots with as few seams as possible.

My current method for getting a good seal on the base of the trefoil is to laser cut a wood sheet with press-fit holes that line up the air tubes with the bladders. I push the air tubes through the sheet with about an inch sticking out, slather everything with Sil-Poxy, and press on the cast tentacle, making sure to massage out any air pockets I can. The advantage to this system is that I can hook luer fittings (which are remarkably airtight and convenient) directly into the tubes. The large lateral surface area of adhesive securing the tentacle down to an inflexible board means that it can take lots of pressure and distortion without directly stretching the adhesive. I think similar methods will keep things airtight over a long working life.

I’ll save the huge advancement made on the quadruped design for its very own post, but I’ll leave you with one of the molds designed to produce a wax of the internal air structure. It’s pretty cool, and I’m predicting that if the first design doesn’t perform as expected, tweaking the CAD and producing a completely new version would only take a week or so of full on work. I’m very excited.