What do you want your robot to do?

snipped-for-privacy@gmail.com wrote:
[...]


I think I did? Fri,Mar 18 2005 Subject: Re: machine learning ?
Of course in order to learn you need a sensory system of sufficient sophistication to provide something worth learning :)

Mmmm. A lot of *what* it does without the *how*.

I think we have a lot assumptions here?

Why would it ask that in particular? Later it might mistake a fridge magnet for a handle. A lot of things look the same in an image, it can depend on context. For example a circle could be a wheel, clock, ball and so on...
Vision is really hard. So despite what I wrote about "railway lines" I think we have to structure our visual world for the robot until it can be programmed to see better. In the case of the fridge put a high contrast "fridge" symbol on it. That will identify and orientate it relative to the robot.

The problem is how to write a "learning" program.

"Rectangle" can mean many things not just door.

Although I find the hardware boring I don't think that it isn't important, just expensive. I try to get the best but affordable hardware. I made sure I had a solid robot base from which to work. No noisy flimsy gears or motors.
The same goes for choosing the hardware for software. It is important that an idea is not stymied by a lack of computing speed or memory.
Practical advancements in robots is a partnership between the hardware and its associated circuitry and any higher level AI software. It depends on your interests, abilities, financial means and how you want to divide up your time.
John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
There was an article a few years ago about someone, I forget the details, but he ended up programming an industrial robot to tentatively touch its environment before committing to an action. This behaviour was modeled on his wife's way of interacting with the world, since she was blind. So instead of the robot slapping down an item in a storage spot regardless, it would first nudge the storage location to see if it was already occupied, then store the item if the slot was free. In this way it avoided making huge blunders, and without overengineering the environment or relying on vision.
It also was programmed with non-linear 'female brain' multi-tasking abilities. Very interesting concept.
wrote:

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Tim Polmear wrote:

This
world,
was
This is not unlike you poking around in the dark yourself, or probably the way most bugs and nocturnal mammals interact with their environments in daily life. Who says your robot needs a $10K vision system and IQ 170 to get along.
Along this line, I recently started reading "Legged Robots That Balance" by Marc Raibert, 1986. Among other things, he describes a quadruped walker built by Hirose in 1983 that could climb up and over obstacles. It used a touch sensor and a simple retract-and-lift-leg algorithm to sense and negotiate the obstacles. Plus a comparable routine for walking down the back sides of the obstacles. Early form of reactive control. You don't need complex sensors and planning algorithms to do everything, when simple sensors and feedback might suffice. You can't do "everything" this way but .... it forms a good basis for the other things.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Giving a robot a sense of touch anywhere approaching that available to a human being isn't that easy. Giving a robot a sense of vision that would be enough for a human to use means plugging a web cam into a USB port.
The storage robot most likely didn't have to feel itself around the factory or recognize anything accept that the space was clear.
Give your robot hand the same sense of touch that a human has and then tell me it is simple.
We can put our hands in our pocket and identify what is in there. We can feel just part of an object and identify it completely. We can manipulate multiple small objects in our hand with ease. The idea that our sense of touch is simple compared with vision isn't true.
Just as you can have a simple touch system you can have a simple visual system.
Vision is useful to find an object but to manipulate an object does require a sense of touch and force feedback. We can sense the texture, shape and weight of an object in an instant and know how to move our fingers and hand to do whatever we want with it.
A major computational task I would think?
It may be true that a simple reflex system will enable something like a cockroach scuttle across unpredictable terrain to the nearest dark spot. If this is all you want your robot to do?
If you want your robot to have an arm/hand that can manipulate objects the way we can that is a different ball game. Work is being done on all this but it is outside the financial status of the average hobby robot I would suggest?
Your basic robot needs its sonar/infrared obstacle detectors and whiskers/bumpers should they fail. But a web cam is cheap and offers a simple vision system with the possibility of a more complex visual system in the future.
The sense of touch that the fembots will become part of the future of robots. If I ever get around to building a robotic arm I would certainly want it to have a sensitive sense of touch.
As for female multitasking it does have a cost.
I like to explain these differences by the idea that men were the hunters and women nurturers :)
- John Casey
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
JGCASEY wrote:

This is basically true, but the sensory/cognition ratio is reversed for the two. It may be more difficult to construct a "skin" with millions of tiny sensors on it, but it's relatively easy to use the information from those sensors.
Conversely, a video camera can provide a highly detailed view of the robot's environment. But analyzing that view is anything but simple. In fact, science has barely cracked it.
In school, we learned the simplistic side of our body's senses: "eyes" for seeing, "ears" for hearing, and so forth. The other half of *sensation*, which is analysis, is given scant attention. The two can't be separated. Some 30 percent of the human brain is dedicated to vision processing. Given the complexity of the human visual cortex, the dorsal stream, and the various other lobes, there is some doubt we'd be able to master its subtleties before an affordable artificial skin could be developed.
This isn't to say that mimicking the human vision system is the best way to approach robotic vision, but it gives us a higher confidence that we know what the robot is seeing, and how it will react.
In the mean time, they are using simple motion analysis for things like red light cameras. One step at a time...
-- Gordon
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Gordon McComb wrote: JGCASEY wrote:

I have done enough visual processing of images to have some idea what *I* can and cannot do with a video camera. We are not talking "human vision" anymore than someone using ultrasonics is talking dolphin or bat "vision".
The only problem I have had, mainly because I am not a professional programmer, is getting the code to grab images from a video camera fast enough to be useful for a robot operating in real time.
[...]

Computer vision is also being used to read car number plates and road signs. Something that might be useful for a robot?
-- John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
JGCASEY wrote:

These are also limitations in the analysis of the data, which is what I was talking about. An ultrasonic sensor is actually far simpler than the echolocation in bats. But in any case, it's the little bat's brain that synthesizes this information to deduce whether there's a wall ahead of it, or a juicy praying mantis.
The human brain works mostly through recognition, not measurement, so most of the machine vision systems in use today are not remotely like how the brain works. We might design a robot eye to look for an orange ball, because orange is a specific color we can determine by looking at the RGB values of pixels. Balls have the same profile from any direction, so we can readily judge the distance from the size of the ball, though this assumes we know the diameter of the ball. Neat stuff, but as I said, far more crude than what a megapixel full color image might otherwise lend itself to.

Possibly, but this is more like optical character recognition (in fact, it is exactly like optical character recognition). It works because the system is looking for known two dimensional shapes, in fairly well established orientations, against high contrast backgrounds. In other words, these things are made to be highly visible, by man (or machine). It would be great if everything were so clearly dileneated.
-- Gordon
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Gordon McComb wrote:

I would say brains measure also, in order to recognize.

Certainly a modern video camera can have a lot more data than any machine vision system will use. But that's good. It makes it easier to decide what to throw away until all that is left is the bit you want. Such as your orange ball. With high resolution that ball will still be "round" at greater distances and not be reduced to say 4 square pixels.
Another example is stereo vision where the higher the resolution (particularly the horizontal resolution) the better.
The frog for example has fairly good eyes but not much of a brain. The higher the resolution the better it can "see" those flying bugs, even though I read its vision is essentially binary in nature.
Perhaps one last example for high resolution is texture. Different surfaces may have different textures and therefore that becomes a possible "feature" for recognition and delineation.

And just as we use reflectors, white (yellow/red) lines and high contrast signs to help the motorist there is no reason we can't do the same for our nearly blind robots.
-- John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
JGCASEY wrote:

Comparative not quantitative, look at "forced perspective" effects where something can be closer and appear bigger when the clues of its distance are obscured.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
mlw wrote:

Indeed it is all comparative and the same must be true for machine vision if it is to make sense of the 2D data collected by a camera with which it is to construct a visual "representation" of "what's out there".
Machine vision involves the same data as human vision. It can suffer the same illusions of size as a human visual system if it uses the same kinds of assumptions in estimating size or any other visual value such as shade, color or 3D shapes from 2D data.
The 2D data has many interpretations as to size and shape. Is it an oval or a circle seen at an angle? An absolute pixel value has many interpretations. Is that "gray" pixel value of 128 part of a white square or a black square? Will we classify that RGB value as yellow or brown?
Vision is an "ill posed problem". Machine vision would have to make use of assumptions as to how interpret those absolute measurements just as we do. Illusions unmask those assumptions.
-- John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Tue, Apr 26, 2005 at 05:13:37PM -0700, Gordon McComb wrote:

I'm not so sure about this - don't underestimate the processing required to analyze the the sense of touch. Touch can determine a huge amount of information about the environment including temperature, surface type, texture, complex shapes, etc. Just think about what it means to come to the conclusion that a surface is rough or smooth, #60 grit vs #220 grit vs paper smooth vs brushed metal vs polished. Our hands can sense all of these with very good accuracy. That's just texture, and only a highly limited sample. Think about holding a handful of sand, marbles, etc. The amount sensory input from skin, and the processing to analyze it, is quite large.
A fairly deep "understanding" of the surface can be determined as well - texture is a big factor, but also there is the amount of thermal conductivity and density which are taken into account as well. We can easily differentiate a steel ball bearing from a wooden ball not only by weight, but because the steel draws heat from our skin which we can feel because the object feels cold.
Additionally, we can assess pliability - is the surface hard or squishy? Or liquid? Is the surface hard, but has some give? How about combinations like wet sand?
When you were a kid, did you ever walk through mud barefoot where the mud squishes up between your toes? What a sensation!
These are all pretty easy for us, but I don't think it is because the processing is easy. It's easy in the sense that vision is easy for us. But for a computer, it's still a very difficult problem.

Even so, I think such an artificial skin would also present some serious complexities in analysis, as well as data bandwidth issues, just like we have with cameras input today. Maybe not to the level of vision, but I certainly don't think it can be said to be easy or even significantly easier.
It's relatively easy for a human to differentiate a puf of wind on our arm vs a spider crawling on our arm vs leaf brushing up against our arm from the tree we just walked past. If such an artifical skin that can provide the level of input of human skin was available right now, and assuming we had a high bandwidth interface to get the data into the computer for analysis, would the analysis of these signals be all that much easier to process than vision data to make these precise differentiations?
I guess this is a demonstration of where the power of the human brain architecture begins to show how weak our naive and single-threaded if-then-else type of logic is at analysing these raw data streams. Raw computing power aside, we haven't been able to really come close to what our biological built-in wet-ware processes are able to achieve in terms of analysing our environment, let alone sense it.

Certainly our ability to truly make sense of the world based on our crude sensory input devices and crude analyses is very limited.
-Brian
--
Brian Dean
BDMICRO - ATmega128 Based MAVRIC Controllers
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Brian Dean wrote:

I didn't really imply that replicating human touch would be easy, but "using the information" from these sensors is relatively straightforward. You can judge size of an object by determing how many cells are activated; shape by the arrangement of the activated cells; weight by the relative pressure on the cells; and "smooth" or "rough" by the variety of readings from adjacent cells, and so on.
There are ongoing experiments with fiber optic "fabrics" and organic transistors (grown on a sheet) that, while not yet duplicating the number of sensors per square inch of human skin, shows promise. But we're not yet even close to the point of perfecting "artificial skin," so much of this is academic.
By comparison, we now have cameras that provide a higher resolution than the human eye (though not the dynamic range). But what to do with that image...that's the problem!
Look here for an example of some ongoing haptic research. One mm spatial resolution isn't great, but I don't think you can say they are having any trouble analyzing the data that comes from the sensor:
http://www.kaist.edu/as_intro/as_nt_prm/as_pr_news/1177549_1836.html
-- Gordon
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Tue, Apr 26, 2005 at 09:44:09PM -0700, Gordon McComb wrote:

Agreed. However, using the information to the degree and effect that a human makes use the information is not at all straightforward and approaches the complexity of vision, IMHO. I'm refering to more subtle uses of touch that we make use of and rely on every day - not just simply to answer the question "is there something there". We use touch in far more sophisticated ways than that, and in combination with our other senses.

To be honest, what I see in this press release is the equivalent of snapping a photo with a camera, except with a grid of presure sensors instead of a CCD. I do not see any analysis of that "image", just the image itself. Am I missing something?
But this is a great advancement to be sure, and sure to take the mechanical sense of touch to the next level. But I really think this is quite a ways from duplicating what real human skin and associated nerves can sense, i.e., this appears to sense switch closure only, no temperature. Also, it's not clear if any gradients of pressure are reported or not, or just on/off data for this particular device.
But I do agree that, at a crude level, this "touch" image is more immediately useful than a visual image for computer analysis since there is less possible ambiguity in the information it represents.
-Brian
--
Brian Dean
BDMICRO - ATmega128 Based MAVRIC Controllers
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
There was an article on Catalyst (ABC Australia) about that sort of thing. The article is called Fembots. It is available at: http://www.abc.net.au/catalyst/stories/s973659.htm The robot was made by a Sydney based company called Kadence Photonics, (there is a link to their website in the article)
Regards
Andrew Wagstaff

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
The very article I was thinking of, thanks.
On Tue, 26 Apr 2005 23:02:19 GMT, "Andrew Wagstaff"

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
JGCASEY wrote:

Can you set your four year old child down in "any home" and have it find the refrigerator? Does a four year old child know what a "Beer" is? You need to start thinking about these robots as children and start understanding how they learn.
The blob vision systems you guys talk about sound just like how a baby sees things. The same goes for most voice recognition and motion control. We have established a starting point from which to work from not the end.
All this reprogramming and rewriting is the same thing that one must go through in raising a child. The child makes mistakes, gross mistakes at first and then we refine their abilities to the point where they can shoot basketball, feed themselves, and have semi-intelligent conversations.
I'm planning on working with the same robot for years possibly to build up its programming instead of building and rebuilding, writing and rewriting. I'm not saying that I won't make changes to the hardware along the way, just gradual changes.
BTW the flood fill mapping technique is exactly the same way I find my way around. It's probably the same method you use, just that you don't pay attention to it any more. When you were a kid you had to figure out that a wall was not a thing to walk on. I can't tell you how many times I've had to tell my kids to literally stop climbing the walls.
I think that the robot just needs to keep running and learning. Eventually "getting a beer" will be exactly "childs play".
Eljin
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
JGCASEY wrote:

The $500 is for the base platform.

I would find it hard to see any robot doing these tasks at the moment. Maybe the spiders.

That is actualy an interesting idea.

Have discussed this with my son. Having it dump food into the aquarium would be cool.

No one I know.

but with a better personality.

not likely.

It can speak, and I have an LCD screen for it.

Flashing lights are so 1960s.

Figured out, yes, long ago at Denning, not implemented on mine as of yet, and it is hardly perfect, obviously.

Wheel encoders are a very weak source for position information. I was atually wondering if the wireless router that I have could be used for some sort of triangulation, one transmit antenna, two reciever antenntas. It is theoretically possible, whether or not the information is available is a different matter.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Eljin wrote:

'Course, I was making a joke about the 60s, but I also agree that aesthetics play an important part of our enjoyment of most anything, robots included. Blinking lights *look* like robot stuff to most people. In addition to LEDs, don't forget side-emitting fiber optics and electroluminescent panels and wire. Sort of the "Tron Look."
Sound effects are also critical. Noises, chirps, relay clatter, and all that add a human dimension to robots. People prefer R2D2 over C3PO. There's a reason.
-- Gordon
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

people are obselete and ugly?
:-)
interesting observation. I never did care much for c3po.
I read Asimov's "The Norby Chronicles" and wanted a robot that would do THAT. I haven't figured out how to tune my bots' hyper-drive yet.
Rich
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
mlw wrote:

Take you word for it. Maybe they just don't admit to the satisfaction of a physical job well done like cleaning the sink or floor or chopping a pile of firewood.
I know one lady that uses her brain all week and one pleasure is to vege out by weeding the garden and doing other routine chores.

Of course it can. Just sing the following choreographic steps,
Just turn to the right, turn to the left jiggle those wheels and go beep beep beep.
move forward 3 steps move backward 3 steps
jiggle those wheels, flash those lights, spin to the left spin to the right
and so on....
Kids would love it if the moves were synchronized to accompanying music.

Don't underestimate the emotional appeal of a good light display and really cool sounds to the general population, particularly the young, even if they leave you cold.
Why do we like fireworks, why xmas lights, why does the Terminator have red glowing eyes? Why do movies use mood music. Think about what many people like when it comes to the appearance and sound of their car.
A domestic robot should look and sound great like all the other possessions we treasure.
Robots of the 2000's will need a "modern" look compared with those of the 1960's. But human emotions do not change, only their cultural expression. The Honda robot and its successors may not look like the steel robots of the 1960's but they still have that "human" appearance.
Today's cool look is tomorrows kitsch :)

Kind of expected you to expand a bit at this point :)
I would consider your basic sensors as part of the $500 robot base. Personally I think any self respecting robot needs a set of four ultrasonic and infrared obstacle detectors and, should they fail to prevent unwanted contact, a set of bumper switches.

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.