Transforming Robots: Smart Blocks

Andy Smith writes: "The BBC reports that researchers from Dartmough College in New Hampshire aim to create robots made from 'smart building blocks.' The idea is that the robots can then transform into other objects. According to the story, the reseachers eventually 'hope to use thousands of microscopic units to make infinitely flexible machines, fit for any task'. The article goes into a lot more detail about how the units will work, and the research that is currently being done."

Comments

But what we really want to know is....

[ChrisGoodwin]

Will the Autobots wage their battle to destroy the evil forces of the Decepticons?

--

Will this ever be useful?

[Chyron]

I wonder... it seems like a very nifty idea, but I don't think these smart morphing robot appliances will be a practicality, not for a long, long time, perhaps not ever. Even if the robots can be made to intelligently morph themselves into an appropriate form for the job at hand, without "real" AI, they'll be incapable of any useful work. At best, they'll need constant supervision (do I really want my lawnmower to blithely slash through power lines, my garden hose, the flower bed, or my neighbour's garden?) And AI of the sort necessary for all but the simplest jobs is still very far off. Especially the sort of generic, non-specialized intelligence that would be required for this type of multi-purpose appliance.
--

Re:Will this ever be useful?

[Kintanon]

I wonder... it seems like a very nifty idea, but I don't think these smart morphing robot appliances will be a practicality, not for a long, long time, perhaps not ever. Even if the robots can be made to intelligently morph themselves into an appropriate form for the job at hand, without "real" AI, they'll be incapable of any useful work. At best, they'll need constant supervision (do I really want my lawnmower to blithely slash through power lines, my garden hose, the flower bed, or my neighbour's garden?) And AI of the sort necessary for all but the simplest jobs is still very far off. Especially the sort of generic, non-specialized intelligence that would be required for this type of multi-purpose appliance.


I think the true commercial application for this technology beyond the benefits to military and scientific equipment will be the Multi-Shovel or some such thing. You have a little grey box, it has on it a panel with some selection mechanism. You can choose 'Rake', 'Shovel', 'Hoe', 'Hedge Clippers', 'Post Hole Digger', etc... And then you watch as the little thing constructs from pre defined plans the selected item.
That way you just spend 1500$ for this one item and it can act as a few dozen different tools. You could get updates to the plans to have it become new stuff, and you could increase or decrease the mass of the cube to have it make smaller or larger items by adding or removing blocks of the little bots.
I know this isn't doable now, but I think this technology might head in that direction for home use once it makes it down that far.

Kintanon

Re:Modular Robots?

[NevDull]

.NET bots? Don't MCSE training courses already make these?

-Nev

T2

[twisty]

This very sort of component would be just what is needed to build the Liquid Metal Guy from Terminator 2... on the right scale. If you think about it, the movable components would have to be on the same scale as organic cells more or less. They'd also need some kind of peer-to-peer network to coordinate their composition as well as distribute memory. (A tall order, but this is a much-needed step.)

The Philosophy of the design has a lot of merit too. I remember a music composition program called Bars and Pipes on my old Amiga. It made numerous effects possible, and easy, by letting you drop the gizmos of your choice into the pipeline. It's also a reason I'm migrating development from Windows over to Linux, because the latter has a much better grasp of piping one universal tool's output into the input of another.

Yawn...

[myosin]

This is rather like the robot snake, and indeed similar 'robots' can transform (This one by Xerox). Not exactly orignial, and the miniturisation seems to be little more than hype at the moment.

-----
"Almost isn't good enough - but it's almost good enough."

do we see a trend here?

[giberti]

This is not the first slash story about a product like this. There are some great video's of this sort of robot in action, running sin wave's for movement. The originial /. story here. And the CONRO original site which has some interesting videos if you have the bandwith to download them here.

It's "Dartmouth".

[Alik]

Dartmough? Feh. They mean Dartmouth College, my very own alma mater. I graduated in the same class as the guy who built the little block they show in the picture.

Yes, the idea of miniaturization is kind of hype. OTOH, Daniela trained under Dartmouth's MEMS guy back at Cornell, so she has some contacts in the field. He's already built (years ago, actually) prototypes of these "smart manipulating surfaces". They look like just a flat chip, but when powered, they'll spin things around, act as conveyor belts, and generally create 2D "force fields".

Shrinking the things isn't the issue. Even if they're an inch cubed, they could still be useful, especially if we borrowed from Lego the idea of having a few "special bricks". The problem is control. Can you imagine having to specify your body one cell at a time? These things are going to need to be able to work out where they should be with minimal cues from the central brain. She does have some work in the field (algorithms to move around furniture with a team of robots, all of whom have limited sensing and communications power; also, the stuff I worked on with transportable agents), but there's a long way to go.

Her own page on the subject is here.

Re:It's "Dartmouth".

[Alik]

Shouldn't you be graduated by now?

You're right; we can simulate the behavior of the cells in order to work out control algorithms for them. They've actually done some of that; Daniela has some animations of some of the older modular designs carrying out simple functions like climbing stairs. (I think they're on that webpage link.)

However, as I have learned in my attempts at robotics, simulation often doesn't mean jack shit. It's still basically impossible to simulate all the nastiness of the physical world, so in the end, you've got to put them out there and see what breaks. (As an example: when I took CS88 from her, we built Legobots. The IR sensor I made was actually sensitive to the color of the brick in which it was embedded; it only worked in blue bricks. Probably a function of the absorption/reflection of the dye, but who the hell's going to code that into a simulator?)

Moreover, in the end, if you want to use them, you've gotta make them. Therefore, it makes more sense (IMHO) to deal with the real-world issues up front rather than gluing them into idealized code at the last minute.

WRT to the surfaces... each one *can* do multiple things. Think of them as a set of "motion pixels" with almost-arbitrary pushing abilitity. The same chip can be a conveyor, an agitator, an aligner, or several other things, depending on what the control code tells the pixels to do.

BTW, thinking of Dartmouth in the news... did you realize that our very own Marty Vona, architect of the most recent modular robot unit, is also the man who hacked the Billy Bass? Every time I start to feel competent, I just need to look at what he's up to.

Asimov

[MouseR]

This is straight out of Asimov's Robot City.

It's a moderatly good read from an Asimov original idea, presented by him but written by a pletora of other sci-fi writers such as Stephen Leigh and William F. Wu.

Karma karma karma karma karmeleon: it comes and goes, it comes and goes.

Humaniform robots would be better

[DeadVulcan]

Morphability is cool and all, but I think that it would be far more useful to build a robot in human form. Such a robot could use existing tools, essentially making it modular.

A morphing robot would (ultimately) outstrip any modular robot in usefulness, but I think the research and fabrication costs of such a robot are gigantic in comparison to simply building a human-shaped robot. A humaniform robot (to borrow Asimov's term) can instantly use of all our human tools, making itself useful very quickly.

--

Nanobots!

[jaavaaguru]

The nanobots have escaped from Red Dwarf!

Morphing vs Modular

[rde]

Morphing robots would be cool; however, later on in the article we find that she actually means 'modular'. While cool in and of itself, this is a bit of a disappointment. Instead of turning into Robert Patrick and trying to kill whiny kids, all it'll do is plug a shears into one of its appendages and mow the lawn. How downletting is that?

Anime U

[Paladin128]

In the newly formed Anime U, found in Neo-hyper-mega-Tokyo IX, Will add this to there curriculum for a BS in Ultrapower Combat Engineering. The new class will be called Morphing Metcha Design. It's pre-requisites will be:

• Bypassing Newton's 1st and 3rd Laws II
• Changing gravitational constants via martial arts.
• Advanced Transformable Mecha
• Improbable Hair

"Evil beware: I'm armed to the teeth and packing a hampster!"

Xerox parc already has done this...

[catch23]

I saw a live demo of Mark Yim's modular reconfigurable robots when I was an intern there. When I was there, they still seemed to have some trouble getting the robot to walk or crawl. It worked great as a wheel or a snake though.

Not Fractal Robotics again...

[AlexH]

There is a guy here in the UK called J. Michael (I forget his first name) who has been trying to sell something he calls "fractal robotics" for years. The whole thing is a big joke in the robotics industry as he's yet to demonstrate any actual working hardware despite often claiming that he has received large amounts of investment from respected companies and that fractal robotics is "the next big thing" which will revolutionise agriculture, space exporation, construction, etc. For a while at least, he used to regularly troll comp.robotics.misc and uk.misc with lengthy articles about how NASA is digging it's own grave by refusing to invest in his company. He even got published in a hobbyist electronics magazine several years ago. He has a web site where you can see obviously fake video clips of a "prototype fractal robot" in operation and read lots of propaganda.

AI is more important than "morphing"

[MotoMannequin]

I cannot morph, yet I am suitable to many varied tasks. An application to this type of technology may be to produce a robot that can repair itself when it is damaged, but comments from the article make me doubt they are on the right track:

She suspects that a process of top-down planning that "cascades" the process of form-changing will be needed to make the system change shape quickly.

They will not be able to make a robust system by trying to control these "building blocks" from the top down. The cells in our body are not being told by a controller that they are an arm, or a kidney, the information is stored in the DNA. Yet our bodies do have arms and kidneys.

Interesting research into complex systems has shown that robust systems are not controlled top-down, but are the emergent properties of lost of small agents that are reacting with each other based on a simple set of rules.

This type of research is the holy grail for scientists in this field, but we are still stumbling on much simpler problems right now.

Moto Mannequin

"With all appliances, and means to boot!" - William Shakespeare