The robots can adjust their caterpillar tracks, to ensure they are all pulling in the right direction. "Each robot has a traction sensor inside that detects all the external forces on it," explains Dorigo. A robot uses its sensor to identify any conflicting forces, and then changes direction accordingly.
So, once its carrying your cargo along the path and begins to slide down a slope all the tracks will turn in unison to help carry it down the hill to its doom. They won't think anything is wrong because everyone will be pulling in the same direction.
So, once its carrying your cargo along the path and begins to slide down a slope all the tracks will turn in unison to help carry it down the hill to its doom. They won't think anything is wrong because everyone will be pulling in the same direction.
I'm assuming maybe this was more of a humorous comment, but I'll take the bait. It would be trivial (I would think) to add a pitch sensor of some sort, then do a little bit of simply physics/trig to adjust the force calculations.
So, once its carrying your cargo along the path and begins to slide down a slope all the tracks will turn in unison to help carry it down the hill to its doom. They won't think anything is wrong because everyone will be pulling in the same direction.
The robots were developed by Marco Dorigo at the Free University of Brussels, Belgium, along with colleagues at the Institute of Cognitive Science and Technology in Italy and the Autonomous Systems Laboratory and Dalle Molle Institute for the Study of Artificial Intelligence, both in Switzerland.
"It could be carried by an African swallow."
"Oh yeah, an African swallow, maybe, but not a European swallow. That's my point. "
"But then the African swallow's not migratory..."
The Invincible [wikipedia.org] by Lem describes a planet populated by a swarm of self-replicating robots, that are very small and simple individually, but as a group display extremely complex behaviours, have swarm memory.
So all we need to do is to show these robots how to self-replicate, I am sure most people on/. know enough about this practice that they should be able to explain this to a robot. And then we are all set. (did I mention that the swarm of these robots killed off everything else on the planet? But it wil
The object was apparently to demonstrate something or other regarding cooperation strategies between robots with limited communication abilities and limited knowledge of the surroundings.
What, precisely, was gained by doing this with actual physical robots, rather than a computer simulation?
What, precisely, was gained by doing this with actual physical robots, rather than a computer simulation?
Gee, maybe things like accounting for things you never thought or had the ability to simulate? What makes you think that a computer can model every single thing (frictional forces, heat and stress on motors, etc) as well as actual reality?
No, but surely it can model the _cooperative_ aspects.
I wouldn't trust a computer to predict whether a robot hand is capable of cracking an egg and peeling off the shell without damaging the membrane underneath.
But I would trust a computer to model the effect of having robot A shine a blue light, robot B shine a red light, have robot A programmed to move toward a red light at 1 mph, and have robot B programmed to move away from a blue light at 2 mph. And I would trust it to model the effect of a twenty such robots.
Interaction effects result in "emergent" behavior because it's not readily apparent from the behavior of individuals. That is to say, sense-and-respond cycles are not easy to model unless you start out with a lot of data. A good example is ants--single ants wandering around demonstrate chaotic behavior in time and in space, whereas large numbers of ants demonstrate very ordered behavior (purposeful movement, all taking rests at the same time, etc.). We can model this because we've seen it, but before we ever saw it, it would probably have been outside of our abilities to predict that it would happen.
In terms of complexity hierarchy, it doesn't make sense to make a model that is just an aggregation of different objects. You don't talk about the group behavior entirely in terms of the objects making up the group, because the objects don't demonstrate group behavior--the group does--so in some sense "half a herd of robots" doesn't make any sense. From the perspective from which the group behavior is evident, the group is a unitary individual.
But we're not talking about just shining lights here, we're talking about moving an object. That involves modeling the frictional forces, possibly balance, etc. Modeling parts of the co-operative aspects before hand would probbably help in the programming design of the robots, but if you want to know if it's actually going to work (how quickly does each robot need to respond to change X, etc), it's probbably easier to just build the damn things.
They are just too annoying to program that it's probably easier to build a robot especially when dealing with unideal situations. An ultrasonic sensor can have so many different modes of failure (ie Specular reflection. Sheets. Cross Talk. Etc. Etc. Etc.) that it's better just to use an actual robot.
What, precisely, was gained by doing this with actual physical robots, rather than a computer simulation?
"In theory, there is no difference between theory and practice. But, in practice, there is." -- Computer Scientist Jan L. A. van de Snepscheut (or Yogi Berra, depending on who you believe)
between robots with limited communication abilities and limited knowledge of the surroundings.
You answered your own question right there. It can be extremely difficult to simulate the 'unknowns' present in a real environment. So yes, you can simulate comms degredation and limited sensor range, but what about "unknown unknowns"? Things that fall into that category might be if your comms are short range modulated IR, what is the effect of reflections? Or if it is RF, similarly, what about environmental inte
All these Japanese companies are dead set on the whole humanoid robot concept. While the AI systems are clearly a joke, why focus so much energy on bipedal movement? It is clearly not the easiest mode of transportation. Human walking is essentially controlled falling. Oh great a bunch of things that kind of resemble humans can lift something heavy all together. Why not just build a smart forklift to do the same job autonomously. It just doesn't make any sense.
- They already ARE making lots of functional type robots. Toyotas factories have these all over the place. - The world as it is is designed for bipedals. If you can model a humans' movement, you can operate much of the human world potentially (climb ladders, etc) - There is a potential market for the humanoid concept. I think the market is validated in some ways by the amount of coverage they are getting for these things.
Well, when you can only send one type of robot to mars, it is best to find the most adaptible one. Having one style of robot that is capable of lifting heavy objects, climbing a rope, driving a vehicle, fixing other robots, etc. is a powerful argument. It lessens the number of separate robots that need to be sent, it allows them to function for a longer time (self repair) using a single common stock of replacement parts. Most significant, they are capable of being modified purely in software to accomplis
While the AI systems are clearly a joke, why focus so much energy on bipedal movement?
You don't need strong AI to perform bipedal movement. The world of researchers does not have to choose between researching bidpeal movement and strong AI capable of interacting with other AIs; this is not a video game - you can research more than one thing in a nation, or even the world.
Oh great a bunch of things that kind of resemble humans can lift something heavy all together. Why not just build a smart forklift to do t
The Japanese have created humanoid robots. They will create human-like robots. These robots will be made to look exactly like female humans that don't actually exist. Once that's done, it's just a matter of time before we will all welcome our Japanese Pr0n Overlords.
Can a smart forklift go up stairs? Or unload directly from the back of a truck, take a box up some stairs, through the front door, and into the living room?
I've often wondered this myself, why the preoccupation.
But I'm going to put that off for a more important question. Since factory robots have essentially just replaced peoples' jobs, turning former manufacturing laborers into Walmart workers as they are replaced with machines and Chinese laborers, what happens when there's a robot capable of every job we have on earth. Essentially would everyone just be poor? Modern conveniences are supposed to make our workload easier, but when robots and computers do
Watch the power of thousands of/.ers! Individually, they can't crash a web server, but together - with limited communication and intelligence - they can...
There have been a few previous efforts in this direction.
Somebody, I think at UCLA, did some nice work in this area around 1990. They had a pair of small forklift-type machines which worked together to lift larger objects. One would get on each end of a couch, for example, and with very limited intercommunication but good force sensing, they'd move the couch together.
That seemed a very practical idea, but it wasn't followed up at the time. There are many industrial and construction applications where
Looking at the video, try to spot the project management behavior that shows up towards the end once four of the bots figure out how to drag the object over. One of them just stops doing anything and stands out away from the group as if trying to think of ways to empower the resources to realize their action items.
It is this bot that must be destroyed before the future of robotics is harmed.
I think that nature is a wonderful place from which researchers can gain inspiration. But when you stop and think about it robots and swarm intelligence just doesn't work. Robots are have different properties then insects and as a result, insect behavior doesn't transfer well to robots.
It's all about energy. Both insects and robots have to be designed to optimize the amount of work done with their limited energy source. There are three basic tasks that consume energy.
- thinking - communication - acting (ie, moving)
Insects:
Thinking is the most expensive task for an insect. Brains are expensive. They use a lot of energy and require a physical body to support the increased energy usage. As a result, insects don't think - they act. There actions are hard coded so as to minimize cost.
Communication is difficult for insects to implement. Pheromones are relatively inexpensive, but impose serious limitations. Visual communication is possible (look at bees) but isn't used much. It also requires good eyesight and a neural network to decipher the images (both of which consume energy.) Audio communication is also possible but requires significant resources (ie, a brain) in order to be effective. In real life it is only used for the most basic forms of communication (look at crickets.) What I'm basically saying is that communication is expensive - as a result, insects found ways to work with minimal communication.
Acting requires energy, but it is the most efficient of the three tasks. If you take into consideration that insects already require a body to acquire food and reproduce, the added cost of using that body to perform an action is minimal. Acting requires no additional parts, it only consumes a small additional amount of energy.
Robots
Acting is the most expensive task for a robot. To act, a robot requires a body. This adds weight, motors, complexity. Batteries suck, have a limited lifespan and are difficult to recharge. Nothing reduces an actor's lifespan quicker then acting. While new technology can improve the lifespan, it won't improve fast enough.
Communication is expensive for a robot, but much cheaper then acting. Wireless communication allows for sophisticated communication between robots while using only a minimal amount of energy. New technology will improve the efficiency of communication more then it will the efficiency of acting.
Thinking is cheap for a robot. New CPUs allow for complex programs while only consuming microamps.
So this is what you have: (hight energy usage to low energy usage)
They are the exact opposite. Does it really make sense to have robots mimic insects? It's crazy. A more efficient way for robots to perform a group task is to have them cooperate explicitly. Elect a leader, create a plan of action, distribute that plan, then act together while minimizing the amount of energy required.
So would it be legal to upload a copy of these videos to YouTube or Google Video in order to take the load off the NewScientist server, and make the content accessible?
You'd have to assume it's copyrighted content (since everything is copyrighted unless otherwise stated), and therefore not allowed. But how is a cached copy of a video on Google Video any different than a cached copy of a web page on Google Cache or MirrorDot? The purpose would be the same.
hmmmmm might need modification (Score:5, Insightful)
So, once its carrying your cargo along the path and begins to slide down a slope all the tracks will turn in unison to help carry it down the hill to its doom. They won't think anything is wrong because everyone will be pulling in the same direction.
Apart from this minor detail i think w00t!
Re: (Score:2)
Re: (Score:3, Informative)
I'm assuming maybe this was more of a humorous comment, but I'll take the bait. It would be trivial (I would think) to add a pitch sensor of some sort, then do a little bit of simply physics/trig to adjust the force calculations.
Re:hmmmmm might need modification (Score:5, Funny)
Yeah, totally unlike humans!
Er, wait...
Parent
Re: (Score:2, Funny)
The title certainly does. When I first read the header/title, I thought it said "Robot Swarm Shits Heavy Objects."
It's a simple question of weight rations. (Score:5, Funny)
Re: (Score:2, Funny)
Re:It's a simple question of weight rations. (Score:5, Funny)
"Wait a minute! Supposing two swallows carried it together?"
Parent
Re: (Score:3, Funny)
Re: (Score:3, Funny)
Re: (Score:2)
"It could be carried by an African swallow." ."
"Oh yeah, an African swallow, maybe, but not a European swallow. That's my point. "
"But then the African swallow's not migratory..
I welcome our new army of robot-swarm overlords (Score:2, Funny)
Re: (Score:3, Funny)
Imagine a Beowulf cluster of these!
Hey - it had to be said.
Meanwhile... (Score:2)
But can they.... (Score:5, Funny)
a HA! (Score:2, Funny)
The Invincible (Score:2)
So all we need to do is to show these robots how to self-replicate, I am sure most people on
Why wasn't this a simulation? (Score:5, Insightful)
What, precisely, was gained by doing this with actual physical robots, rather than a computer simulation?
Re: (Score:2)
Re:Why wasn't this a simulation? (Score:5, Insightful)
What, precisely, was gained by doing this with actual physical robots, rather than a computer simulation?
Gee, maybe things like accounting for things you never thought or had the ability to simulate? What makes you think that a computer can model every single thing (frictional forces, heat and stress on motors, etc) as well as actual reality?
Parent
Re:Why wasn't this a simulation? (Score:4, Interesting)
I wouldn't trust a computer to predict whether a robot hand is capable of cracking an egg and peeling off the shell without damaging the membrane underneath.
But I would trust a computer to model the effect of having robot A shine a blue light, robot B shine a red light, have robot A programmed to move toward a red light at 1 mph, and have robot B programmed to move away from a blue light at 2 mph. And I would trust it to model the effect of a twenty such robots.
Parent
Re:Why wasn't this a simulation? (Score:4, Insightful)
In terms of complexity hierarchy, it doesn't make sense to make a model that is just an aggregation of different objects. You don't talk about the group behavior entirely in terms of the objects making up the group, because the objects don't demonstrate group behavior--the group does--so in some sense "half a herd of robots" doesn't make any sense. From the perspective from which the group behavior is evident, the group is a unitary individual.
Clear as mud?
Parent
Re: (Score:2)
Re: (Score:2)
Simulations are horribly annoying to program (Score:2)
Re: (Score:2)
In theory, you're right: there's no difference. But in practice...
Re: (Score:3, Insightful)
"In theory, there is no difference between theory and practice. But, in practice, there is." -- Computer Scientist Jan L. A. van de Snepscheut (or Yogi Berra, depending on who you believe)
Re: (Score:2)
You answered your own question right there. It can be extremely difficult to simulate the 'unknowns' present in a real environment. So yes, you can simulate comms degredation and limited sensor range, but what about "unknown unknowns"? Things that fall into that category might be if your comms are short range modulated IR, what is the effect of reflections? Or if it is RF, similarly, what about environmental inte
Wasting time w/Humanoids? (Score:4, Insightful)
Re:Wasting time w/Humanoids? (Score:5, Informative)
A few misconceptions to clear up:
Parent
Re: (Score:2, Interesting)
- They already ARE making lots of functional type robots. Toyotas factories have these all over the place.
- The world as it is is designed for bipedals. If you can model a humans' movement, you can operate much of the human world potentially (climb ladders, etc)
- There is a potential market for the humanoid concept. I think the market is validated in some ways by the amount of coverage they are getting for these things.
Re: (Score:2)
Re: (Score:2)
You don't need strong AI to perform bipedal movement. The world of researchers does not have to choose between researching bidpeal movement and strong AI capable of interacting with other AIs; this is not a video game - you can research more than one thing in a nation, or even the world.
Oh great a bunch of things that kind of resemble humans can lift something heavy all together. Why not just build a smart forklift to do t
Re: (Score:3, Funny)
The Japanese have created humanoid robots. They will create human-like robots. These robots will be made to look exactly like female humans that don't actually exist. Once that's done, it's just a matter of time before we will all welcome our Japanese Pr0n Overlords.
Re: (Score:2)
Re: (Score:2)
slashdotted (Score:5, Funny)
Cooperating robots (Score:2)
There have been a few previous efforts in this direction. Somebody, I think at UCLA, did some nice work in this area around 1990. They had a pair of small forklift-type machines which worked together to lift larger objects. One would get on each end of a couch, for example, and with very limited intercommunication but good force sensing, they'd move the couch together.
That seemed a very practical idea, but it wasn't followed up at the time. There are many industrial and construction applications where
Management (Score:5, Funny)
Looking at the video, try to spot the project management behavior that shows up towards the end once four of the bots figure out how to drag the object over. One of them just stops doing anything and stands out away from the group as if trying to think of ways to empower the resources to realize their action items.
It is this bot that must be destroyed before the future of robotics is harmed.
Re: (Score:2)
Minority Report (Score:2)
This page was generated (Score:2, Funny)
SWARM intelligence is just silly (Score:3, Interesting)
It's all about energy. Both insects and robots have to be designed to optimize the amount of work done with their limited energy source. There are three basic tasks that consume energy.
- thinking
- communication
- acting (ie, moving)
Insects:
Thinking is the most expensive task for an insect. Brains are expensive. They use a lot of energy and require a physical body to support the increased energy usage. As a result, insects don't think - they act. There actions are hard coded so as to minimize cost.
Communication is difficult for insects to implement. Pheromones are relatively inexpensive, but impose serious limitations. Visual communication is possible (look at bees) but isn't used much. It also requires good eyesight and a neural network to decipher the images (both of which consume energy.) Audio communication is also possible but requires significant resources (ie, a brain) in order to be effective. In real life it is only used for the most basic forms of communication (look at crickets.) What I'm basically saying is that communication is expensive - as a result, insects found ways to work with minimal communication.
Acting requires energy, but it is the most efficient of the three tasks. If you take into consideration that insects already require a body to acquire food and reproduce, the added cost of using that body to perform an action is minimal. Acting requires no additional parts, it only consumes a small additional amount of energy.
Robots
Acting is the most expensive task for a robot. To act, a robot requires a body. This adds weight, motors, complexity. Batteries suck, have a limited lifespan and are difficult to recharge. Nothing reduces an actor's lifespan quicker then acting. While new technology can improve the lifespan, it won't improve fast enough.
Communication is expensive for a robot, but much cheaper then acting. Wireless communication allows for sophisticated communication between robots while using only a minimal amount of energy. New technology will improve the efficiency of communication more then it will the efficiency of acting.
Thinking is cheap for a robot. New CPUs allow for complex programs while only consuming microamps.
So this is what you have: (hight energy usage to low energy usage)
insects - thinking, communicating, acting
robots - acting, communicating, thinking
They are the exact opposite. Does it really make sense to have robots mimic insects? It's crazy. A more efficient way for robots to perform a group task is to have them cooperate explicitly. Elect a leader, create a plan of action, distribute that plan, then act together while minimizing the amount of energy required.
Willy
Re: (Score:2)
Re:Hah. (Score:5, Funny)
Well, your office is clearly blocking material that's NSFW.
Parent
Cache? (Score:3, Interesting)
So would it be legal to upload a copy of these videos to YouTube or Google Video in order to take the load off the NewScientist server, and make the content accessible?
You'd have to assume it's copyrighted content (since everything is copyrighted unless otherwise stated), and therefore not allowed. But how is a cached copy of a video on Google Video any different than a cached copy of a web page on Google Cache or MirrorDot? The purpose would be the same.
I did get a copy of one of the videos before the
Re: (Score:2)
Re: (Score:2)
(google gets a lot of variants of this)
Re:Peep hole (Score:4, Funny)
Parent
abuse of moderation (Score:2)