Japanese Researchers Develop Sensor Skin

ScentCone writes "A University of Tokyo team has developed a flexible, laminated network of pressure and temperature sensors suitable for jobs such as robot fingers. Circuits as pressure sensors, and semiconductors as temperature sensors are not new, but the thin, networked laminate of the two is novel."

Comments

Re:Key invent

[Mycroft_VIII]

Mostly I agree, however I would like to point out that not all the components for a funtional finger need to be in the finger itself. Make tight fist while watching the fore-arm on the same arm.
    Same could go for some of the electronics related to sensor amplification and such.
    Though I do agree we are still a good ways from human equalivance in artificial muscles and skin sensors.

Mcyroft

it was there already

[jaromil]

at least something very similar to it, and patented by Sony: www.csl.sony.co.jp/person/rekimoto/smartskin BTW one of the developers of smartskin is the author of effectv.sf.net i'd rather add that slashdot news are getting lousier every day, please keep it up.

next up...

[hitmark]

direct neural interface. basicly a chip that allow the sensors to talk to someones neural system so that one feel the pressure as if it was ones own skin.

that in my view is realy the holy grail here. to get the feedback loop that is our interaction with the world going while using something thats not flesh.

Re:Key invent

[Mycroft_VIII]

Heh, yeah I've thought about this a bit. Two of my fantasy mad scientis inventions are an android and Power armor, which share some of these problems. The other is a time machine.
    It's a bit related to subsumption architechture in robotics, thought I didn't originaly think of it that way and intend to study more on that subject. It's just obvious to me that semi-independant semi-hiarchical controll/processing is the way to do it rather than try and do it all as one big cpu and software. Especially when you consider all the work the feet do keep you ballanced and moving smoothly while walking through 5 toes and the ankle pivot. Or all the work hand do in so many tasks.
    Teach the tiny components the details and the biger components then know just enough to order the next size down around and take feedback.
    To many layers could creat propogation delays and errors of imprecision, but then wires are faster than neurons and all sorts of minor errors that have to be constantly corrected 'in flight' occure in humans.
    If there is enough task specific learning capability and such at a low enough level the system could cope. Add in simple neural networks at say l2 or l3 and use GA's to find good fit baselines for them and the higher level heuristics have less to deal with specifics wise most of the time.
    Of course all these processors and various actuators will generate some heat (as might the power source), but if your emulating a human, a fluid cooled system (colored appropriately) could double as a mimic for a blood system and help provide 'body heat'. You could also pass cool air past some of these 'blood vessels' via a breathing mimicing system.
    Of course a mimic for the muscular system is a bit tricky. There are some pnumatic(sp?) muscle systems that somewhat behave like human muscles (rubber tubes in a mesh so that expaning the tubes shorten them) but I'm uncertain how thier strength and deformation properties compare to human muscles. Some other candidates are clearly not powerfull enough to do the job and if the same holds true for the phumatic muscles then more traditional motors and actuators may have to do some to most of the work with the other system being largely cosmetic in purpose. And of course the response time is important.
    And the problems with robotic vision are infamous, though progress is being made.
    It's all fun mental excersize though.

Mycroft