Skynet is waaaay in the future

[Argonor]

Just to show what cutting edge robot tech is about these days:

www.scapetechnologies.com

(this is the company I work for, now)

[flooglestreet]

Now, one of those for dumpster diving would be

[Westfalia Chris]

Well, at least we´ve progressed from the "Griff ins Klo" stage to the "Griff in die Kiste" stage*. So I guess world domination isn´t too far away.

*Apologies to non-German-speaking users. Can´t really translate that idiom.

[Argonor]

Now, one of those for dumpster diving would be

Problem is, you'd have to know exactly what you're looking for, having made a CAD drawing of it, trained the robot to grip it - AND make sure that enough of the desired object is actually visible to the robot camera(s) to ensure a recognition..... 

[Gluteus Maximus]

I don't know whether to be relieved or disappointed that robots are still so primitive 

No offence to you or your company, of course 

[Hauptgefreiter]

I don't know whether to be relieved or disappointed that robots are still so primitive 

not all of them are THAT primitive 
Look here:
http://www.youtube.com/v/KZPnssjzsCQ&hl=en&fs=1&autoplay=1

[Bako]

I don't know whether to be relieved or disappointed that robots are still so primitive 

What about automated floor cleaners and those lawn-trimmers, and the robot (actually two working in tangent) that can act like a waiter, in addition to loading and unloading the dishwasher?

Or the study group with small robots that created their own "personality" (admittedly not what we'd call one though)? A few became 'friends', others disliked others, and one continuously tried to kill itself.

And some of those prototype UAVs the military is working on...

And no, I'm not crazy. The Discovery Channel and Tech TV exist for a reason beyond pictures of ancient jungle ruins and the latest from NASA .

[Argonor]

Floor cleaning and grass trimming robots actually run on far less complicated programs than a random bin-picker. Everything in those is hard-coded, whereas a bin-picker actually has to 'think' in reacting to the contents of the bin.

The vision technology needed for automated recognition of randomly placed shapes has only been slowly evolved over the last decade, whereas the need for automated bin-picking has been present ever since the beginning of industrial production (even though only the last 20 years or so, that need has been recognized - before that, endless repetition of the same motions was just something workers had to live with...)

[Gluteus Maximus]

Fair enough, robot tech is very impressive considering how hard it is to replicate animals' behaviour, but my point was just that we don't see anything remotely near to the stuff we get in our favourite SciFi films and tv programmes. Would anyone trust their lives to a AI-controlled aircraft?

When I see a Terminator or Robocop in real life, I'll be impressed. And scared sh**less 

[Argonor]

Would anyone trust their lives to a AI-controlled aircraft?

Nope, because the AI would have to rely solely on the instruments, unlike a human pilot who can react to anomaly and ignore a faulty result (the thing measuring the height being wrongly calibrated, for instance).

Sci-fi movies seem to assume, that a machine can mimic the function of the human eyesight - which may be possible somewhen in the future, but right now vision systems rely on cameras taking pictures, which are then processed to see if the program can recognize any features from the shapes stored in its database.

Actually, my company is refining a method using only 1 or 2 cameras, whereas the competitors rely on 12-16 cameras to get recognition 

Still, if the bin is not placed correctly, the system might not be able to find all the parts in the bin.

So a robot running around, recognizing faces, reacting to the environment almost like a human, is still a thing of the distant future... which may, or may not be a good thing....

[Sky Captain]

So a robot running around, recognizing faces, reacting to the environment almost like a human, is still a thing of the distant future... which may, or may not be a good thing....

It is true that a robot is far away from behaving like a human, but science is taking big steps in lower level behaviour like reflexes.
Of course most of it is still far away from mass production and just working in the lab right now.

Here is an example of a very interesting human like robot hand.

I'd guess it recognizes the sushi mostly by color, but when touching it it can sense the pressure and grab it without squeezing it. In addition with soft and flexible robot arms I think it can be a valuable help in everyday live in not to distant future.

Here is an other example of a robot runing on four legs like an animal.

Even if surely not recognizing the entire envirenment it can sense enough to pass difficult and unregular terrain. And when watching the part on the ice you realize how good relexes are to keep standing when slipping.

They are still basics, but science is much further in building an animal like robot with some kind of basic "feelings" like a living body.  Feelings in this case more simple feelings like presure, heat or pain etc rather than abstract ones like love or hate.

Last but not least a video about Exoskelettons.

Still a little clumsy but it may be the first step to some kind of space marine. 
No, but I think when making it stable enough and finding a way to produce it kind of cheap it will be great for certain civilian tasks like construction works.

[Dr. The Viking]

Just keep developing those neural networks. One day SkyNet will be among us!

[Bako]

And humaniform robots...

I admit, my first few examples were pretty poor. But the last one was more of what I was striving for. Their only programing was very limited, with only the proper pathways for activating their locomotors and something to do with their sensors. Just about everything they did they had to learn on their own. The exact details were a bit beyond me, but that's what allowed them to develop their own simple personalities. Though in reality they were little hunks of metal with 3 wheels slightly larger than a balled fist. Maybe not a Robbie or R. Daneel, but baby-steps must be taken before long strides can happen.

One day though robots and whatnot will exist regardless of how well they fit the lazy (ahem)s needs.

That robotic hand is pretty cool. Would beat the standard prosthetics of Today by a long mile if paired with a further augmented form of the technology behind "mind-powered" wheel chairs.

[Sky Captain]

I admit, my first few examples were pretty poor. But the last one was more of what I was striving for. Their only programing was very limited, with only the proper pathways for activating their locomotors and something to do with their sensors. Just about everything they did they had to learn on their own. The exact details were a bit beyond me, but that's what allowed them to develop their own simple personalities. Though in reality they were little hunks of metal with 3 wheels slightly larger than a balled fist. Maybe not a Robbie or R. Daneel, but baby-steps must be taken before long strides can happen.

Robots can learn using neural networks. Its just like evolution: try and error. They remember what did well and always make little changes to get new ways.
Here is a cool video about very simple little robots who can learn to act together forming a bigger robot. They learned to build a snake and to build a spider, so more comlix behaviour lerning is possible. 

[Argonor]

I think the very signifcant difference between some of the above projects (with the exo-skeleton as an exeption), and the work we do in Scape, is, that we know exactly where we are aiming.

We are making solutions for industrial production lines that are going to be implemented more or less here and now (at least when measuring in industrial terms), whereas I hear a lot of vague ideas of what might be the use for the different technologies being developed.

There's a huge difference between 'scientific' and 'engineering' approaches, and we need to take the engineering approach to satisfy specific commercial needs, whereas scientists often start out in one direction, and then changes that along the way, and see what they end up with

I'm not saying that scientific research has no place in robotics, of course it has, but when I hear that many hundreds of scientists are involved in a relatively small project like the multibot, which has no set purpose other than to see if it is possible to make it, I cannot help thinking, that maybe EU money could be spent better on projects aimed at more specific ends.

We are a small company with 7-8 programmers/developers, and we are involved with some of the very big players in the automotive and aerospace industries (I'm not allowed to reveal them, sorry).

I'm not a developer, I mainly work with administration and marketing, but the 3 months I've been here, I've got a much deeper understanding of the problems faced in the automation area.

People think of a bin-picker as performing a simple task, because they see it performing the same task again and again - but actually it doesn't.

A robot painting car parts is performing the same task again and again, because the car is always the same, and always placed in the same spot, in the same direction, etc. so the procedure can be hardcoded - no decision-making involved here.

A random bin-picker has to find a part, and figure out how to best grip it. At the same time it has to be aware of the sides of the bin/box and the other parts, to avoid collision. If the part has to be placed in a fixture or similar for further processing, it involves orientation control - the robot must be able to tell what is up and down, and maybe left and right (and sometimes even how it is positioned in a 6-dimensional space - I cannot even begin to understand what the developers are talking about, here  ), and the figure out how to turn it, before dropping it in the designated place.

And I did like my old job better - being a systems consultant/first line supporter in a company that sold and bound books for libraries and schools, but as the production moved to Copenhagen, I had to find something closer to home.... Compared to my old job, this is rather boring - and there are no cheap/free books