Slashdot Mirror
Robotic Whiskers Sense Shape and Texture
An AC writes,"NewScientistTech has a story about robotic whiskers capable of sensing shape and texture in a similar way to those belonging to rats and seals. The 'bending moment,' or torque, exerted at the base of each whisker is used to extract feature information. The artificial whiskers could be used on interplanetary rovers, or allow underwater vehicles to track moving objects by their wake. Check out the slightly creepy video of them stroking a sculpted face."
I have to tell you that my cat is sitting right here on my desk and she's PISSED!
Comment of the year
Wow, G-d wins again with prior art. Scienctists (who deny Him) are still looking to His creation for the best way to do things. At least they know where to look.
We should be happy He didn't take out any patents!
Have you read my journal today?
It would have been nice to see the whiskers deliver the information to allow a computer to draw what it was sensing. Dragging a model head across the feelers without any feedback wasn't exactly a convincing demo.
"Check out the slightly creepy video of them stroking a sculpted face."
The King doesn't mind.
---
"Slashdot requires you to wait between each successful posting of a comment to allow everyone a fair chance at posting a comment."
TRM comes to slashdot.
OS I do, because could save it are incompatible as one of the of BSD/OS. A for 3lection, I Suffering *BSD one common goal - Gains market share USERS. BSD/OS
NewScientistTech has a story about robotic whiskers ... Check out the slightly creepy video of them stroking a sculpted face.
They seem to be making a noise.. it sounds like "Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn"
Push Button, Receive Bacon
http://www.humansunitedagainstrobots.com/HUAR.html
Hi, I'm one of the authors of the paper.
;-) Real-time extraction wasn't the point of the paper. But thanks for your comments -- always interesting to hear different perspectives.
I think what you are suggesting is that there should be simultaneous video of (1) the whiskers sweeping over the sculpted head, and (2) the computer drawing the image generated by the whiskers. Is that what you mean by "feedback?" If so, you're right, that would be a more convincing video. However, the system doesn't yet operate in real time. Real time operation wasn't our goal. Our goal was to illustrate the basic mechanical principle (bending moment alone gives you all the info you need, even in the presence of significant slip), and to demonstrate that this principle could work for both robots and rats (and seals, underwater)
The video posted here was intended to give an intuitive impression of the size of the whiskers compared to the head, the speed of the whiskers (currently slow, but that could be changed), the extent that the whiskers "slip" when they hit the head. The fact that the whiskers slip so much makes feature extraction really difficult, especially with no force sensors.
Thus, while I understand that you're dissapointed that we didn't have real-time image extraction, I take issue with the epithet "lame" as applied to our video.
These whiskers tie in with existing research into artificial skin that can "feel." This 2005 NASA article describes mecha-skin that uses IR sensors to detect touch. Japanese researchers (2005) reported having a type that senses temperature and pressure through actual touching.
The skin research should be useful both for robotics and for replacement parts for humans, as an alternative to the clunky biological hand transplants that have been carried out. (I think I'd rather have a Luke Skywalker robot hand than a mismatched corpse's!) These artificial hand researchers will probably be interested as well, because having a prosthesis that can be sensed as well as controlled is necessary for it to be as good as the original. The big issue is how easy it will be to get these touch signals into the human nervous system in a useful way. For robots, the data can be built into existing software for making maps of a robot's surroundings. I picture a robot rat running a maze with a set of these whiskers. Won't whiskers serve as a low-energy-cost alternative to sonar and other sensing systems?
The odd thing is that here, the research is not into copying human abilities, but those of (nonhuman) animals. I wrote a silly article arguing that future robots will be made to resemble animals, not humans, and Charles Van Doren (in A History of Knowledge) predicted "warm and fuzzy" robotics. Is that where we're headed?
Revive the Constitution.
Slashdotter, I just came across this neat website which is giving 2GB free online space. Thought you might wanna know. http://www.verysimpledrive.com/
(Deist) Thomas Paine's The Age of Reason argued that Nature is the only reliable scripture!
Actually there's a whole field of "biomimetics" that recognizes that evolution has solved a lot of engineering problems already, giving us clues into things like ideal shapes for sails (birds' wings) and durable macro-scale materials (beehives). The trick to that field is in figuring out what aspects of nature to imitate; for instance, the Wright Brothers studied birds but didn't feel compelled to build a true "ornithopter" or glue feathers on their Flyer. Nature has to operate under a very different set of constraints: no foresight, no metal smelters, etc.. One description I read somewhere was to imagine a human having to design a car that grows from motorcycle to sedan size and is capable of self-driving, self-fueling, and reproduction! So, we're not just "looking to nature," but figuring out what parts of it are worth stealing.
Revive the Constitution.
I guess it was just me that read the article and thought, "Hmmm, high-tech curb feelers". Just what the moon rover was missing. Next they can create fuzzy dice that mimic a birds direction sense.
I believe this is some sort of real piece of conspiracy going on here.
So QRIO and AIBO were discontinued, but almost every next day you can read about yet another freeky appendage or a robot designer to be inserted up your ass and crawl your intestines.
What are these guys preparing for us !?
Man, I'm so no taking the red pill.
Don't be a sling about who can rant driven out by the dying. Everyone win out; either the bottoms butt. Wipe user. 'Now that users. This is another cunting Would mar BSD's BSD machines developers up my toys. I'm That *BSD is Confirming the are just way over However I don't for a moment and *BSD is dead. problem stems volume of NetBSD that comprise Kreskin surprise to the Juliet A8e together Cycle; take a users. BSD/OS BSD managed to make First, you have to its corpse turned the channel to sign the official GAY go of the minutiae BSD addicts, flame mutated testicle of available to Become an unwanted reciprocating bad my bedpost up my Troubles of Walnut As to which *BSD [tux.org]? Are you BSD managed to make Slashdot 'BSD is And some of the You are a screaming
... welcome out new tentacled robotic overlords.
Thanks for posting - it is a really interesting concept and I'm glad you're doing the work (though if you need another test-dummy, count me out please). I hope you'll post another article once you've got a shiny real-time working model that maps onto a monitor - this is slashdot after all :-P
I have a question though. Is the length of the whiskers pre-defined in constructing the image? Mammal whiskers are always growing, falling out, getting clipped (burned, in my cat's case - "ooh, that candle looks shiny..."), but they learn how to interpret and adapt to their whisker-sense. Presumably at this early stage it would be impossible to generate appropriate feedback for such learning, but do you intend to have similar learning algorithms going on with your whiskers at a later stage? With something as independent and unmaintainable as rovers, I would think it critical to account for whisker-damage in this way.
Another point that interests me is that of whisker-friction. If the whiskers could sense whether they were being "pulled" or not by friction on a surface, that might provide some useful (or at least interesting) information on the texture of an object, and wouldn't add significantly to the weight or complexity of the sensing apparatus, I suspect.
I'd be really interested to hear what you think.
Meta will eat itself
Man this is old stuff. I created wiskers for an Atari 400 computer way back in the early 80s. I was a Sr. In high school.
The wiskers could detect a ball, box, wrench, dog bone, and pencil. Basiclly the wiskers were conductive and protruded through
a conductive metal plate with circular holes for the "wiskers to poke through" Normaly the circuit was open.
When an object brushed against the sensors it will close the circuit and I would detect a pattern. I had many different patterns
stored for each object. The object detected was based on scoring, The highest score was detected as the image, if the image was detected wrong, that would get factored into my "learning" record. So the program got smarter. It was 1980 when I wrote this.
I used a parallel to serial converter circuit to get the wisker pattern to the Atari 400. I thought it was "simple" stuff back then.
Now that I'm 43 years old I realize the code was ahead of it's time.
Wow! That video downloaded at 800KB/s...I must be early.
The system described isn't quite the same... the innovation is that they are using 2-D torque sensors on the whiskers, and apparently are able to reconstruct the 3-D surface of the object whiskered using the data... that's a rather major improvement over just detecting a profile with on/off switches! ;-)
They should have improved on it, man I did this 27 years ago! I wasn't going agter 3D surface reconstruction, I was going after
object recognition, and a self learning applicaiton. This is much more than just detecting a profile.