First to Blog

This is a response, more of a high five, to Zach Hoeken’s post up on MAKE: “First to File? Nah, First to Blog!” Basically his post was a series of ideas that have been hanging around in his notebooks, possibly eligible for patents, that he would rather see out there and made in the world than locked away between the pages of a personal sketchpad forever or exploited to the chagrin of mankind by some unruly technological entity, wrapped up in complex patent labyrinths, and never put to a more just use than in sole product from a sole company (see 3d Systems vs the Form 1, Patent Busting3d printing patent challenges, etc). Even worse is the possibility of an idea getting patented and never implemented, only used as a club to hit innovators over the wallet (see Intellectual Ventures). I’m in favor of this. Truth be told I’m pretty aggressively anti patent, which is why all of my recent robotics projects have been released into the open source. Although I realize there’s a difficult road ahead, finding ways to keep funding innovation and novel IP in the world patent abolitionists have been gunning for, I believe open access to information and the network effects it generates far far and away outweigh the small innovation boost you get from researchers confident they’ll be the only people able to profit from the particular idea they’re developing.

If you want a stark, dramatic demonstration on how having access to all the data makes each bit of data incredibly more valuable, and how open review routinely saves the world, take a look at this case where a single internet comment on a large meta-analysis of medical literature clued the Cochraine Collaboration into a huge pile of missing, unpublished research data that was clouding their results. Their subsequent examination has left the entire field of pharmaceutical research and regulation on shaky ground. Until recently, prior art meant information that the patent office could easily find that would invalidate a patent application. Blog posts and online tutorials haven’t held much weight in patent applications up until recently. See, before the web, the best way to disseminate information directly to the patent office, saying “Hey, I’ve made something but don’t want to patent it. Please stop anyone from trying to patent one of these.” with a Statutory Invention Registration. It sped up patent office searches and made sure your work lived in the public domain. As of last year, however, the America Invents Act has eliminated that particular device. It seems like officially dispersing your work now means filing for a provisional patent, and then allowing it to expire, cementing your concept in the public domain. Although publications online might fall within Section 102 of Article 35 in the US code regarding patents’ definition of prior art I’m not familiar with any cases where blog posts have blocked patents. This article suggests they may as long as there’s an indelible time stamp (such as an Archive.org cached copy of the info) testifying to the information’s date of publication.

Soluble cores for the first conical tentacle. These were printed at Jim Bredt’s lab.

So, why blog about it? It’s possible blogging won’t stop other researchers from patenting similar ideas, preventing people from expanding the technology in an open environment. It’s possible that my reluctance to close down IP guts my chances of getting funding to extend these ideas as more powerful more impactful designs. However, I believe that technologies are functions of ecosystems. Soft robots are nowhere to be found in any industrial, medical, or commercial application. Their development has been slow despite lots of funding, public interest, and hours spent at the bench researching the topic. I believe this is because the environment of solutions (infosec might call it an attack surface) is incredibly sparsely populated. In traditional robotics if you’re hunting for an idea for a simple cost effective chassis to prototype your bot on you’re in luck.  Wondering whether your boat should be made of aluminum or steel? Bingo. Searching for information on laser cut part assembling considering kerf for interference fits? Done. Where do you look for the guide to best practices in soft robots? Where are the descriptions of which structures give the most flex force per psi in an inflating chamber? Where are the documentations of failures and iterations that let you pick up a person’s methods and procedures to replicate their success? They’re pretty thin on the ground.

I’m writing about this (and I believe in the open source movement) because I know that more data, more eyes on the problem, more experiments, and more exposure means that efficient solutions will start proliferating. It means that successful designs will be replicated, improved, and the technology will move forward as a whole. Just look at the projects using layers and layers of technology developed to make laser cutting faster and objects assembled from laser cut bits more functional. Ten years ago they were nonexistent. So, in that spirit, here are the most promising ideas that I’m developing into the next round of soft robots:

Twisting Tentacle

If I take any of my tentacle prototypes and twist the whole assembly in CAD the resulting tentacles should have an inherent twist when inflated. This will probably result in an octopus-style grip generated by the form folding in on itself like a rubber band after you’ve twisted it up and it starts rolling over into a spiral.

Internal Power Source

I’m developing a system for using compressed gases (specifically nitrous or liquid CO2 cartridges) as a power source. Based off some quick back-of-the-envelope calculations (figured from how many paintballs you can fire with 12g of liquid CO2) it looks like liquid gas would give me about 100x the power (J) of an alkaline battery. I’m thinking of making robots that have a small battery on board and some basic electronic valves to handle articulation and using an air cartridge to do all of the mechanical articulation. It’s even possible I could save power on the valve operation by having them use gas from the cartridge to do the majority of the mechanical work of flipping open or closed.

Pressure Activated Suction Cups

One of the coolest aspects of making robots with soft materials is developing analogues of cool mechanisms out there in the world of biology. Almost all creatures on earth are made up of soft squishy bits. They’re doing millions of complex clever things with their meat from creating rings of it to meter their food as they break it down internally, to growing and shrinking differently colored bubbles of it for camouflage.

I’ve been developing some designs for octopus suckers that are cast into a tentacle and activated by air pressure. The plan is to have solid plugs of silicone surrounded by a hollow ring with a divot in the surface of the silicone where it meets the plug. Hopefully when pressure is applied the ring will inflate, and create a suction cup.

Pneumatic Logic

I’m interested in keeping the complexity of manufacture down for these devices and having as many processes taken care of in a single casting as possible. Each time a part needs to be inserted or suspended in the mold, a glue seam needs to be made, or I have to run tubing from one chamber to another it increases the chances of messing up the whole prototype and is another point of failure. I’d prefer to have some simple deterministic mechanisms, like a valve that dumps air into the next chamber after the last one has filled to a certain point, to be cast right into the piece rather than piloted by an external control unit. With such pneumatic logic (like is found in spiders) I’d be able to get complex motion out of a small robot without piping air control lines or fitting in hard electronic valves and wires.

Jamming Tentacle

If I design a tentacle with a hollow internal chamber, separate from the inflating ones the control the distortion, I could fill it with a jamming material like coffee grounds or styrofoam beads. The jamming material could form a spine that I’d be able to activate by sucking the air out of that chamber. This might be useful for getting a gripper wrapped around an object and freezing it there with the jamming spine, allowing a soft robot to get around a tight corner or grip a complex shape and then rigidize to manipulate it.

Durometer Gradients

An essential feature of anatomy is smooth transitions. Creatures survive by distributing force from running in such a way that it doesn’t rattle their brains out of their skulls or eject all their hard won food right out of their bowels. This takes a lot of elegant fastenings moving force from one area and spreading it out over a strong framework like a skeleton or cartilaginous spine. If you’ve ever cut up a fresh squid for calamari you’ll know that they’re pretty soft creatures. You can poke your finger through one if you’re feeling particularly sadistic. How, then, does it hold on to such a sharp beak capable of eating fish whose flesh and bones are many times harder than the squid itself? The answer is a beautiful gradient of chitin extending out from the tip. The layers at the very end are super hard, and each subsequent one is slightly softer forming a gentle seam with the body and distributing all the force the beak endures.

I’m trying to get the same kind of elegant transitions in my own devices using gradients of printed powder: meshes that start out dense and get sparse as they travel up through the soft articulating part of the robot.

Stretch and Bend Sensors

Eventually piloting these bots will take some pretty sophisticated control systems. Ideally I’d have a closed loop both metering the power I’m putting into the system and having sensors to tell me where the robot’s various bits are in space. I’ll be trying some tests with bend and stretch sensors embedded right in the robots to get feedback on how they’re moving, and start piloting them remotely.

Embedded Bones

After a recent accident I’ve found that I can embed hard pieces into my silicone bots by partially infiltrating the printed parts with liquid resin and then casting silicone over them. This might mean increasing the force these bots can deliver. By restricting the directions in which they can bend I can convert more distortion into force and possibly get these things to do some useful work.