Slashdot Mirror
No Servant, Japan's Build-a-Robot Delivers Joy
isabotage3 writes to tell us that a new Japanese build-a-robot product may offer up a bit more participatory joy than models past. Even though it took this novice reporter over eight hours to assemble he still seemed to think that the end result was worth it. With a quick interface that allows everything from basic movement mapping to complex dance and aerial maneuvers, this robot seems to offer the user an experience far removed from the ASIMOs and AIBOs of years past. From the article: "You don't have to be a scientist, or even very smart, to play with Manoi AT01. But there's a catch: A lot of work is required to get it going. The $1,260 machine, which can walk, wave its arms and do other simple moves, comes in a kit that requires assembly — a sprawling, mind-boggling concoction of matchbox-size motors, plastic Lego-like parts, twisted wiring, 200 tiny screws and a 100-page manual."
...or does that "catch" sound like a damn good reason to buy it?
[command INSERTWITTYQUIP failed: insufficient wit]
OK, so a good after dinner project for any decent engineer here.
What's the big deal? RTFM, put it together, and play.
I can't be the only one who did 10k-piece puzzles as a kid, and those only had a picture.
- Adam L. Beberg - The Cosm Project - http://www.mithral.com/
The article mentions that this is a Manoi robot. A spot of online searching leads me to this MPEG video.
Cheers,
Ian
I thought "Build-a-Robot Delivers Joy" was a euphamism for "custom built sex android."
If you're interested in these things, you should also compare the featureset (hardware AND programmability) of other Japanese robotics kits. Two come to mind: the Kondo KHR-1 / KHR-2 series, and the Robonova series. There are others, but these two seem to have solid support and continual development. The Robonova has a nice feedback system, allowing you to hand-pose the robot and "snapshot" that pose, and then you string together all the poses into various actions. It's almost like you're programming it via stop-motion modeling.
One area I think that most are weak is that of vision support. I'd like to work with recognizing various symbol targets (even barcodes or stripes) and get specific command feedback. Also, this scale of robot is just now getting familiar with gyro inputs, but it's not like it's suddenly able to walk up inclines and catch itself falling. They seem best able to work in a very simple flat-smooth environment only.
[
Now, that would be cool.
How we know is more important than what we know.
Would have been helpful if the submitter included a link to a PICTURE of the robot, which the article didn't either...
I found it here. And in English.
Tequila: It's not just for breakfast anymore!
Ummm, not to be a racist or anything, but wouldn't reading a 100-page instruction book translated from Japanese be considered torture in some countries?
1. Please to put little engine 3X later than subassembly YY.
2. Set us up the arm, but not to be rotated wisely.
3. Enjoy super happy fun robot!
The assembly is about typical for a Kyosho product. Try building one of their better 4WD R/C cars, with a working suspension, transmission and differentials to assemble. Very similar experience.
The actuators for this robot are apparently still output-only R/C PWM-type servos. The competitive product Robonova, though, has position and current feedback from the servos to the control computer, which moves it out of the dumb preprogrammed category into something that has potential for real autonomy.
The sensor suite on these things is still below par. These things really need a 6DOF inertial navigation system for balance, which means three rate gyros (about $22 each) and three accelerometers (about $6 total). They need force sensing in the feet and hands. With that, a camera, and a WiFi link to an external computer, you have almost ASIMO-level hardware functionality. I'll bet we see all that in a year. It's the obvious next step.
Then the problem is to develop software for robust legged locomotion. There's been work on that, but most of it is with expensive one-off machines. Once that moves to commercial robot hardware in the $1K range, progress will be rapid.
There's a heck of a lot of things that are hard about that problem. I do, however, think we have the technology to do it. The Scale-Invariant Feature Transform algorithm is now 2 years old. There are open source implementations and many demonstrations of it being used effectively. This algorithm makes recognising parts something you can do in realtime. All the dexterity required to fiddle about with those parts and put them together has been solved a number of times, but mostly by academics who don't commercialize their research, so you'd probably have to solve that again.
How we know is more important than what we know.
http://video.google.ca/videosearch?q=manoi
pretty cool robot. still not springy. once they will be able to use inertia in robots, instead of ignoring it, thats when robots will be life-like.
"You don't have to be a scientist, or even very smart, to play with Manoi AT01."
I think I'll wait for the Womanoi TA36-24-36.
If you can read this sig, you're too close.
Well somebody has to take the lead in this endevaour. I don't see a GUNDAM coming from the United States anytime soon. And I want a GUNDAM!
No. See, in academia, you solve a little reduced set of the problem. To commercialize your research you have to actually expand the work to include real world problems and until you do that, academic research is about a useful as poetry.
How we know is more important than what we know.