Somebody should be watching Daniela Rus and her pals at MIT, because what they are doing is so crazy, so potentially important, people need to know about them. Not because they're dangerous, but because what they're doing might be changing the world and nobody should change the world without the world noticing.
In a nutshell, they are giving machines the power to redesign themselves.
They're not teaching a robot to copy itself. No, it's bigger than that. They are building robots that will soon change shape on their own, to better accomplish some task. These machines can choose their own future. (Guided, for the moment, by human inventors.)
Step by step (and I've already blogged about their magical bag-o-sand toolkit) they are creating the basic building blocks of a reconfigurable robot, and to my amazement — all of a sudden, robot bits have learned to leap through the air!
The concept of a reconfigurable robot is simple: Make a cube that can move to sit on another cube, then another, so that it can build different shapes. Sometimes like this ...
Sometimes like that ...
And eventually, the bits can turn themselves into a chair, a support beam, a tool, or a Terminator robot that saves little boys from being killed so they can save the future world.
Snap In, Snap Off, Move
The hard part is getting the bits to do this on their own. First off, they have to learn three basics: How to snap into place, unsnap, and move.
John Romanishin, who's been building robots and competing in robot derbies since he was a baby geek, figured out how to slip an itty bitty flywheel into the cube. It looks like a miniature desk fan that rotates up to 20,000 revolutions per minute — it comes with a battery, and when the flywheel is on, the cube has momentum: it can climb, roll across the ground, crawl up another cube, even leap through the air. "Which is pretty exciting," John says, "because it allows robots to jump on top of each other and go places that they couldn't go if they were only moving directly on the structure."
It has no moving parts on its outside. The motion comes from within.
When the fans turn off, a series of magnets on all six cube surfaces allow it to snap into place wherever it is, onto the cube it's next to, which creates an enormous number of possibilities ... Just take a look.
Daniela Rus, who has a knack for finding very whizzy whiz-kids, even for MIT where the whiz population is pretty high, says her grad students John Romanishin and Kyle Gilpin are still telling the cubes what to do, sending instructions via wifi — but one day, they won't have to.
“ [They] decide how, when and where to move.
- Kyle Gilpin
Says Kyle:
In the future, we envision putting the algorithms on the modules themselves so they completely autonomously, in a distributed fashion, decide how, when and where to move, so we want to be able to take a large group of cubes and tell them, "form this shape" and give those instructions at a very high level, and then have the cubes decide on their own how to go about accomplishing that task.
I'm fantasizing now ... but here's how my dreams go. First I'm creating a self-assembling hammer, then a self-assembling car, then a self-assembling hut when I'm off hiking in the mountains, and then ... one day that cannot come too soon ... a self-assembling Robert Krulwich that can "go to work" while I spend the whole day kayaking in Long Island Sound.
As long as these thingies are tethered to human needs, I think what Professor Rus is doing is something I'd like to see more of, so I say (and I think I am speaking here for many of us) Keep goin', Daniela! We are all pulling for you ...
Here's the MIT press release that describes what Rus' guys are doing. Thanks to Erik Olsen for sending this my way.