Now, the programming, design, and production of a functional robot takes several years, is extremely expensive, and involves multiple disciplines, such as hardware and software design, advanced programming technique, and machine learning and vision. The research team's goal in the new project is to automate the process of producing functional 3D objects. Interestingly, the other major goal is to let ordinary people design and build fully functioning robots from everyday materials like sheets of paper, not an entirely new idea in 3D printing.
A robotic gripper 3D printed with easily accessible materials could be used by people with limited mobility. (Source: Jason Dorfman, CSAIL/MIT)
"Our vision is to develop an end-to-end process; specifically, a compiler for building physical machines that starts with a high level of specification of function, and delivers a programmable machine for that function using simple printing processes," said Rus in a press release. "We believe that [this research] has the potential to transform manufacturing and to democratize access to robots."
Researchers' topics of interest are focused on a number of areas, including developing an application programming interface (API) for function specification and design, writing algorithms to control assembly and operation of a device, creating an easy-to-use programming language environment, and designing new, programmable materials for automatically fabricating robots.
In addition to Rus, other members of the CSAIL team include Martin Demaine, Wojciech Matusik, Martin Rinard, and Sangbae Kim of MIT's department of mechanical engineering. The NSF project team also includes Harvard University's associate professor Rob Wood.
Very cool initiative, but I have to wonder about the complexity of creating a 3D printer that is capable of allowing the average consumer to actually produce something that is so complex is terms of functional behaviors, not just physical form. It's one thing for a 3D printer to effortlessly crank out a screw or a bolt or some other physical piece of hardware that can fix a household appliance, but doesn't perform any movement. It's quite another to 3D print an entire robot that has motion to unscrew a jar or open a door.
No doubt it's possible in a research lab; I'm just wondering about the realities of commercialization on a grander scale.
Beth, I agree with you on the commercialization aspects of this project. Many of us would support NSF funding for research activities but the focus on "developing a desktop technology that lets the average person design, customize, and print a specialized robot in a few hours" is curious. Certainly the average person isn't going to produce results in terms of research funding, but I assume the project has primarily been funded on its merits as significant research.
I agree with your skepticism to some extent, Al and Beth. By looking at the background data, it appears that the researchers have used the words "design and customize" to really mean "customize" on a couple of different levels. What intrigued me about this, aside from the robot angle, is that it's quite in line with other developments Beth has written about regarding the use of blueprints by consumers to 3D print household items. This just takes that a couple steps farther with slightly more complex machines.
I totally agree with you, Ann, and it is in keeping with some of the other stuff we've written about. The thing is these robots are real robots thus have to actually move and perform tasks. That's where the reliance on 3D printing is questionable, in my book. We wrote about this initiative My Robot Nation, where 3D printer companies were trying to encourage lay people to design and print their own robots, but these were toy robots. Very doable. This NSF thing--maybe not so much.
I think this is very cool. Of course, at best, this will develop a collection of basic devices that can be linked together with quasi-rigid peices that the user lays out. But if the basic devices are at the right "scale" this could be a big step forward.
Think of the Lego Mindstorms system. It enables kids to build autonomous devices using a half a dozen special purpose Lego bricks, and their Lego sets. This project would use 3D printing to free the user from the limitations of the Legos. The other component would be simplifying the programming. Even Mindstorms programming is pretty rigorous. Developing the next level of abstraction in programming would be worth the $10 Million by itself.
I agree, this is cool, but I'm a little skeptical too... not that it couldn't be done, but I don't think we're there yet... or at least not a the level I think is implied by the teams.
Are the pictures actual working prototypes or theoretical mock-ups/proposals? The pictures can be telling... the 'frame' may be paper, the PCB may be paper or 3D printed, but it's pretty obvious that the components (op amps? transisters?) are not... so it's like a paper breadboard. I also noticed the wires tethered to the back, suggesting that power and/or control must not reside within the robot.
I've been interested in 3D printing, and even the upcoming revolution in 'printable' technology, such as printable solar panels, printable circuitry, etc. using special inks/toners with standard hardware.
I'm also curious how they'd deal with joints and control of motion... I could see motion powered via piezoelectrics if the material can be conductive and deposited via ink (paper) or 3D printer (maybe sintering as opposed to the plastic printers).
For what application would these be used? As 3D printers improve and become more versatile, I can see application with more rigid structure, but I have a little trouble with the paper aspect, unless we're mainly talking circuitry. These look like origami art mixed with parts from an old radio... The Lego Mindstorm system is cool, although I have yet to get my own system to tinker with (it's really for my son :)), but part of the limitation of the Legos is it's advantage... a reconfigurable rigid modular system.
Although it's neat to think of using paper as a construction medium, it strikes me more as a novelty in practicality... Just thinking of possible applications... self-delivering mail (or notes, across class), toys to chase or pique your pets' interested (until they try attacking/chewing it), paper Roomba, process serving (subpoenas/summons, etc.), negotiations in dangerous/hostage situations, swarm of cellulose assassins (new meaning to the term "death by a thousand papercuts"), or (my favorite) make paper robots out of the NSF grant money and watch it walk away.
As far as programming, programming can be pretty simple... I have little toys 'robots' that run off simple neural networks... although like 'hard wired' into the circuitry, a software/programming variation isn't that complex... I'm sure I could make simple neural net programs in assembler if I wanted, and it's one step away from machine code.
While I'm all for basic research and (useful) applied research, I'm guess I'm not sold on sending $10M in taxpayer money for this type of research project. Just because a bureaucrat gives out a grant doesn't mean that it's worthwhile, and there are numerous examples of that I don't won't to go into right now. In my opinion, I don't have a problem with this research at this time if it is paid for by somebody else... the universities, industry (maybe the paper and/or 3D printer industries), private investors, and/or venture capitalists, but given our current government fiscal incompetence, I'm against this kind of grant... I don't see the return to the government (or their boss, the taxpayers and citizens) for the investment, no matter how often one of them uses the term "democratize".
Neural nets aren't programmable. At least not in the sense of the programmer positively determining what they do. At best, a neural net can be trained to respond a specific way to specific stimuli. When it then gets dissimilar stimuli, it is anyone's guess what it is going to do. I wouldn't recommend them for this kind of thing at all.
Save them for applications where their flexibility is a bonus and their lack of determinism is not a hazard or a detriment.
On the subject of paper, I felt the same way about the pictures, but I have also seen a true 3D printer that uses paper and glue for its build matrix. So I withheld comment on that point. The possibility is still there.
JCG, I agree the photos of paper "robots" do look like mockups, since as naperlou pointed out, the prototypes shown clearly don't have working joints. The article also states they are prototypes, not functional robots. This is a five-year project, so clearly there aren't any results yet. It's also interesting that the research will examine "new, programmable materials."
SparkyWatt, thanks, I hadn't thought of the Lego analogy, but I think that's a great one. Other, equally "far-fetched" methods are already being used for mass-producing small functional robots, such as this one that produces a robot insect from a single sheet:
Beth, the My Robot Nation initiative you wrote about was to create models, so it's admittedly much simpler. But the NSF-funded project is allotting time and money, and particularly some of the best brains in robotics, to a lot of research clearly lacking in the My Robot Nation initiative. That research will focus on several topic areas, including "new, programmable materials." If 3D printing can be used to make parts for aircraft I don't see why it can't be used to print parts for functional robots.
Ann, from the pictures I can see that this is a primitive device. Where are the joints? This will not have the mobility that is discussed.
What I really don't understand is the use of the term "democratize". To do what these researchers talk about you could certianly use a wheeled vehicle to better effect. What is the NSF doing funding this? If you could make really useful robots from just a specification language and a 3-D printer then people would be doing it.
naperlou, the research has only just begun, so you're right that no one is doing this yet. The photos show prototypes, no doubt the ones shown to the NSF. The fact of who is involved also piqued my interest: the concentrated brainpower here is quite high, and many of the people involved have already done some pretty amazing things in robotics.
I don't know if this would be classified as applied or basic research, but I like the fact that it has a five-year goal of compiling printable, programmable machines. The fact that it has a $10 million NSF grant must mean that someone thinks it's realistic.
Chuck, the price tag, the five-year span and the assembled brainpower says to me that if these guys can't do it on this project, it probably can't be done--at least not yet. The NSF doesn't hand out that kind of money every day. The NSF is also highly involved in trying to keep the US competitive in STEM.
Seems to me that makerspaces (hackerspaces) are already moving in this direction w/o the government or millions of dollars for grants. Anyone check out Jeremy Blum on YouTube and the makerspace movement?
The information on the web page is very interesting. However, I wonder about the company no real company information all I remember seeing are several generic email addresses. If anyone has first hand working knowledge or relationship with this company please post.
I like the idea that the average person would have the ability to select, print and program a robot. Exactly how this will be accomplished and made available will be interesting. Some people have a difficult time programming universal remote controls. A professor once told his programming class creating an easy to use program means a lot of work on the programmer and I think this comment applies to this concept. To develop something that is simple will take a lot of upfront work and planning plus a reasonable costing 3D printer. Nevertheless, I like the concept.
As cool as the concept sounds, I doubt that the "average" person will be embracing it any time soon. Let's face it, technology has to be "packaged up" and simplified before the average person will use it. Witness the remote controlled television. I can't tell you the number of times my kids went searching for the "lost" remote to turn the TV off, while I actually got up out of my chair, walked over to the box and pushed the "off" button.
Scott, I think you're right, and the expectation is that consumers will be buying some kind of package that they then customize. The problem with using the terms "design" and "program" is that they mean something quite different to a consumer than they do to an engineer.
Good point about the differences in understanding between the engineer and the average consumer in terms of terminology and functionality It's that difference that can often make or break a product. I'm reminded that for years we sold a sensor product with an embedded functionality that made the interface more flexible. We sold a few. We decided to package up that flexible interface in a separate "black box" that lived in a control cabinet and in no time we were selling thousands every year. Counterintuitive? Yes - from the engineering point of view. Perfect for the user.
Scott, thanks for your input on this subject. It keeps coming up during our discussions of robot packages, and I've also run across the different usage of these terms in a couple of other areas, such as machine vision.
Just think of how much waste we could produce if the general public were able to attempt to produce an individual robot! We have folks with no technical understanding and no concept of cause and effect, and now those could spend a bit of effort and consume resources in creating something robotic. Aside from the mounds of wreckage, consider the implications of a robot produced by somebody who has no grasp of inertia or kinematics. Just think about that!
In addition, consider those "bright young kids" who could be crating assorted robots, learning about the process of creating functional robots, without ever understanding a whole lot of basic engineering and physics fundamentals. IT looks to me like a handy process for producing "unintended consequences", from where I stand.
William, I agree that how this is packaged for consumers is key. I've made the same type of criticisms about giving consumers the ability to 3D print other things, such as toaster parts. However, apparently the idea is that they could only customize existing blueprints, not actually "design" their own robots in an engineering sense.
Instead of sifting through huge amounts of technical data looking for answers to assembly problems, engineers can now benefit from 3M's new initiative -- 3M Assembly Solutions. The company has organized its wealth of adhesive and tape solutions into six typical application areas, making it easier to find the best products to solve their real-world assembly and bonding problems.
Many of the materials in this slideshow are resins or elastomers, plus reinforced materials, styrenics, and PLA masterbatches. Applications range from automotive and aerospace to industrial, consumer electronics and wearables, consumer goods, medical and healthcare, as well as sporting goods, and materials for protecting food and beverages.
Engineers trying to keep track of the ever-ballooning number of materials and machines for additive manufacturing and 3D printing now have some relief: a free searchable database with more than 350 machines and 450 different materials.
At JEC Europe Dow Automotive introduced a new ultra-fast, under-60-second molding cycle time for its commercial-grade VORAFORCE 5300 epoxy resin matrix for carbon composites. It's aimed at high-volume automotive manufacturing.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.