Print Your Own Robot: Part 4

I have lots of updates to share. First, there’s a new video of the latest tentacle prototype in action below the fold. Second, We’ve made some excellent progress on the method of manufacture, reliability, and repeatability of the designs we’re producing. It’s almost at the point where we need to figure out a good real-world experiment to test our ideas against. I’m currently torn between creating a grasper and something that walks. Jim and I have been working out ways of getting a self contained power source inside of a soft robot, and it seems like we might be able to use a canister of compressed gas to do everything from timing movements to articulating valves. I’m working out ways of integrating peristaltic pumps and timing mechanisms that will be simple to prototype… which is a pretty tall order. However, I think a combo of laser cut bits and creative molding can have this one solved.

This past week, up at Viridis3d, I filmed a mini documentary on the process we’re using and how their 3d printing technology integrates with the whole project. I should have that one edited up and published by the end of next week.

The current process we’re using deviates from my initial ideas pretty substantially, but I think it will end up for the best. We’ve abandoned the concept of soluble cores printed straight off of the machines. Although these have a lot of potential, they keep giving us problems and slowing down the iteration time substantially. The biggest frustration is that the silicone always soaks into them at least a little, forming a boundary layer between the solid silicone and the soluble powder. It doesn’t stick well to either, but also doesn’t come out of the skin without substandial abuse. It’s a lot like what you get when you dump flour into boiling soup. You get this sticky hydrolyzed layer that’s not soup and not starch surrounding lumps of perfectly dry powder. We’ve fiddled with the idea of coating the cores in wax, but the resulting cores take a lot of effort to get out cleanly as well. When heated it ends up turning into a paste that has to be massaged out of the tentacle (not quick and elegant prototyping by any measure).

What we’ve come up with is pretty promising. We’re stacking all the internal parts into one-piece molds, casting them in silicone, and then using the silicone mold to make waxes that get assembled into cores. It’s a method very similar to the one described by Michal Zalewski in the Guerrilla Guide to CNC Machining (which I can’t recommend enough for inspiring clever fabrication techniques). We can print a mold out of MakerDust, infiltrate it with resin, pour silicone, and have it producing cores in just a couple of days, and most of that time is spent doing other things waiting for materials to cure.

Although this new method throws a bit of a wrench in my “make a silicone volume with an arbitrary hollow volume inside” plan, when you consider the constraints of getting this thing to articulate (paths to supply all the parts of the hollow volumes with air, features to suspend the parts that will form these hollow volumes in place while things are getting molded, and the resolution of powder printed parts) the method of creating one piece molds that generate wax cores isn’t too unreasonable. It does mean that it can’t simply replicate internal geometries with a genus greater than 0 (which eliminates spongy shapes like coral and the inflatable tissues found in a certain part of the human male anatomy), but with planning it definitely can capture some holes. Besides, if I’m desperate for a silicone sponge, I can always print a core directly in wax.

The next version of the tentacle design that’s in progress is intended to solve the problem of the meat on one side of the central wall preventing the inflating volume on the other side from doing its job. Essentially, when you have the axes of motion opposing each other by ninety degrees, they spend all their time either doing work or getting in the way. The new design looks a bit like a three leaf clover when seen from above. I figure setting the inflating volumes 120 degrees away from one another will eliminate the most material opposing the axis of motion, while enabling a pretty good amount of control and flex.

I’ve got a pretty detailed photo series on the process up on Flickr and will be releasing the major CAD iterations on Thingiverse soon. In the meantime, keep checking back here for updates.