Cognisant
cackling in the trenches
- Local time
- Today 11:48 AM
- Joined
- Dec 12, 2009
- Messages
- 11,393
Just something I've been working on for a while.
To emulate muscles I intend to use two servos applying counteracting forces to an axis, each via its own torsion spring, so if both are active at the same time it's like tensing up a pair of counteracting muscles which makes the limb rigid and immobile, whereas having them both inactive at the same time is like the mentioned muscles relaxing, leaving the limb floppy, I imagine the torsion springs may also afford the servos some shock force protection, though I'll probably use further suspension for that.
This setup should provide a wide range of force and rigidness, though I'm concerned about how quickly this can occur, the muscles of the human body are astonishingly fast, I guess I could accelerate the actuation by sacrificing power, maybe use a miniaturised dual cone CVT although that's getting ridiculously complex.
Torsion springs on wikipedia.
What a serious robotics servo looks like, and cost
To control all this I'll be using the "wired intelligence" most commonly seen in BEAM robotics (using 555 timers and such) this involves having analogue circuitry respond mechanistically to input, for example realigning a misaligned limb, indeed this is how servomechanisms work, the potentiometer is the input and if the servo arm or hub is misaligned more or less power will be applied to correct it. Because there's arguably no processing going on (rather a singular process) this analog control circuitry can respond incredibly quickly, very close to the theoretical limit in fact, which is why our bodies work in much the same way.
BEAM robotics on wikipedia.
The AI methodology I'm working with is a form of artificial neural net, it's design is simple yet incredibly counterintuitive, literally took me years to figure out, no doubt you've noticed the effect it's philosophical implications have had on my mind, my obsession with robots...
Hence why I'm building a body for it now, the AI requires extensive training and the complexity involved in training it in a virtual environment increases exponentially as the concepts it needs to understand become increasingly abstract, so it's far easier to teach it in real space as allegorically one would raise a child.
The body itself will have a very basic level of intelligence, enough so that it can successfully perform the instructions it receives from a stationary (cluster) mainframe, the body will be powered by a direct connection to the power grid, so if the AI wants to go somewhere it simply switches between bodies while the bodies themselves remain tethered in their respective locations, eventually a fibre optic connection will become part of everyday household power points, I see this happening first in Japan where such technology to care for the elderly is anticipated and desperately sought, and also in various businesses like Starbucks, 7Eleven, etc.
The internal frame will be comprised primarily of water-jet (or laser) cut sheet metal (or carbon fibre, though I'm worried it may be too brittle) held together by screws and right angle brackets, so it won't be a pretty terminator style endoskeleton, but it'll do the job.
Externally I've considered various options, latex horrifies me, a plushie look would be cute but flammable, flat plating would be fine but ugly, the all round best option is a doll like plastic shell, the molds I'll create could be reused for casting metal or ceramic versions as well.
Ball-jointed doll construction.
Potentially NSFW, it's a nude doll, nothing overly risqué but still hard to explain.
Gyroscopes, potentiometers, accelerometers, and various pressure sensors will be mounted on the frame and on the inside of the shells, so it won't have the complex tactile sensory arrays we do but it'll be able to detect pressure, vibration and impacts; furthermore I'm considering using small microphones so it can "feel" by listening, so if the fingertips have small ridges and it brushes it's fingers over a surface it'll be able to feel/hear the surface texture.
I'm still pondering how big it should be, the larger it is the easier it'll be to build and find parts, but high performance servos are very expensive, then again the smaller it is the faster it's reaction time will need to have due to lacking inertial stability, although at a smaller size I can use lighter materials, increasing the power to weight ratio.
To emulate muscles I intend to use two servos applying counteracting forces to an axis, each via its own torsion spring, so if both are active at the same time it's like tensing up a pair of counteracting muscles which makes the limb rigid and immobile, whereas having them both inactive at the same time is like the mentioned muscles relaxing, leaving the limb floppy, I imagine the torsion springs may also afford the servos some shock force protection, though I'll probably use further suspension for that.
This setup should provide a wide range of force and rigidness, though I'm concerned about how quickly this can occur, the muscles of the human body are astonishingly fast, I guess I could accelerate the actuation by sacrificing power, maybe use a miniaturised dual cone CVT although that's getting ridiculously complex.
Torsion springs on wikipedia.
What a serious robotics servo looks like, and cost
To control all this I'll be using the "wired intelligence" most commonly seen in BEAM robotics (using 555 timers and such) this involves having analogue circuitry respond mechanistically to input, for example realigning a misaligned limb, indeed this is how servomechanisms work, the potentiometer is the input and if the servo arm or hub is misaligned more or less power will be applied to correct it. Because there's arguably no processing going on (rather a singular process) this analog control circuitry can respond incredibly quickly, very close to the theoretical limit in fact, which is why our bodies work in much the same way.
BEAM robotics on wikipedia.
The AI methodology I'm working with is a form of artificial neural net, it's design is simple yet incredibly counterintuitive, literally took me years to figure out, no doubt you've noticed the effect it's philosophical implications have had on my mind, my obsession with robots...
Hence why I'm building a body for it now, the AI requires extensive training and the complexity involved in training it in a virtual environment increases exponentially as the concepts it needs to understand become increasingly abstract, so it's far easier to teach it in real space as allegorically one would raise a child.
The body itself will have a very basic level of intelligence, enough so that it can successfully perform the instructions it receives from a stationary (cluster) mainframe, the body will be powered by a direct connection to the power grid, so if the AI wants to go somewhere it simply switches between bodies while the bodies themselves remain tethered in their respective locations, eventually a fibre optic connection will become part of everyday household power points, I see this happening first in Japan where such technology to care for the elderly is anticipated and desperately sought, and also in various businesses like Starbucks, 7Eleven, etc.
The internal frame will be comprised primarily of water-jet (or laser) cut sheet metal (or carbon fibre, though I'm worried it may be too brittle) held together by screws and right angle brackets, so it won't be a pretty terminator style endoskeleton, but it'll do the job.
Externally I've considered various options, latex horrifies me, a plushie look would be cute but flammable, flat plating would be fine but ugly, the all round best option is a doll like plastic shell, the molds I'll create could be reused for casting metal or ceramic versions as well.
Ball-jointed doll construction.
Potentially NSFW, it's a nude doll, nothing overly risqué but still hard to explain.
Go on make your jokes, it just happens that doll making has already solved many problems in articulated design.
If the image didn't load:
http://www.souldoll.com/shop/step1.php?pagenow=3&number=1065
Scroll down.
If the image didn't load:
http://www.souldoll.com/shop/step1.php?pagenow=3&number=1065
Scroll down.
I'm still pondering how big it should be, the larger it is the easier it'll be to build and find parts, but high performance servos are very expensive, then again the smaller it is the faster it's reaction time will need to have due to lacking inertial stability, although at a smaller size I can use lighter materials, increasing the power to weight ratio.
