I agree its a pretty weak article but its a blog, not a paper, so I would not expect it to rise to the level of rigor for something intended for one of the IEEE journals.
That being said, if you've ever heard of 'charity burnout' there is something similar called 'robotics burnout.' You discover it, you engage in excitement, but nothing happens, you can only be excited (or charitable it seems) so long before you need to take a break. But lets start at the beginning shall we?
So in 1985 the Homebrew Computer Club was waning, it had been the center of attention between 75 and 85 but with people like IBM in the game and Apple going much more 'corporate' it was more of a users group than the folks who were changing the world.
One of the special interest groups from that meeting was the Robotics SIG. A guy named Dick Prather who was active in that SIG decided that even if the computer group was dying, robotics was just getting started and so he did the SIG equivalent of a setsid(2) call and made the Homebrew Robotics Club an independent organization. It has met continuously since then (yes 26 years).
One of the things about 'robots' that most people don't get, is that fundamentally a robot is any machine that has some level of program-ability that does one or more tasks while adapting to its environment. Your dishwasher is a good robot, it washes the dishes for you, or the pots and pans, or the stem ware, it uses a variety of sensors to decide if the dishes are clean yet and it dries them afterwards. If you didn't have it you would be getting a sore back moving dishes from the counter-top to the sink, to the drying rack.
Folks have argued that to be a robot it would have to do it like humans do, but that is an angels-on-pins sort of argument. Generally robots are the expression of automation, and they have (as a market) been growing where ever it makes economic sense. As processors get cheaper and more powerful more and more things make economic sense.
Of course for things that are really expensive or really dangerous its really easy to justify the cost. So for things like disarming bombs, or hunting people down in a country you are not technically at war with and killing them are both easily justified costs if you can automate them.
ISRobotics, the guys who make the Roomba, make most of their money selling robots for things like mine clearing and recon, and yes shooting people. Founded in 1990 they were making $1M/yr in 1996 with 16 employees[1], that wasn't particularly sustainable but in 2001, with 9/11 they demonstrated the value of their packbots. And now at $401M/yr they are doing quite well except that over 60% of their revenue is "G&I" which is code for "Government and Industrial".
Now that isn't all bad, its just that instead of comparing the robotics revolution to the 'PC' revolution you have to compare it to the 'computer' revolution, which is to say that a whole lot of investment and development is focused on corporate and government use until the the industry can supply enough automation to slake their thirst and to give some time to 'regular' folks.
So in the 80's we had companies like Androbot and Heathkit approaching mobile robotics as gimmicks/educational, in the 90's we had a lot of toys and the start of robot (really armored R/C vehicles with minimal automation) combat. And now in the second decade of the 21st century we're seeing prototype self driving cars, a number of walking designs, and energy systems that can sustain things for more than a couple of minutes.
Bottom line, things are going along, and getting better, and in many ways getting better faster now than they have in the past, but there is so much eco-system that has to develop around building mobile robotics that can do useful work that the rate of change feels much more evolutionary than revolutionary.
When we look at robotics startups today, places like Willow Garage and Anybot you can get a good feel for how cool things could be and how far we are from seeing the kind of uptake like we did with personal computers.
Generally robots are the expression of automation, and they have (as a market) been growing where ever it makes economic sense. As processors get cheaper and more powerful more and more things make economic sense.
Yeah that's kind of it in a nutshell. The PC revolution was easy to see because computer hardware has been performing the same fundamental task for decades: computation. Memory, storage, processing, and I/O; that's it. All computers are basically the same in that sense. But that one fundamental task is so useful that virtually everyone benefits.
Because of this, an entire industry has focused on improving those 4 basic components of computing, making them ever cheaper and more capable. And as the capability increases, the utility increases as well.
It's not the same with robots. Robots are more likely to evolve gradually over time. There won't ever be one single thing you can point to and claim "that was the robot revolution" and it's going to be a lot slower because there won't be a Moore's Law for robotics.
This is a weak article. First, it ignores the existence of commodity robots. Something like 4 million Roombas have been sold. They are easy to use commodity robots that sell in volume from Best Buy.
Second, many people, going back to the early 20th century, have imagined buying cheap general purpose robots, just as they have imagined buying jet packs and holidaying on space stations. Just because I imagine it doesn't make it feasible. The reason there aren't many autonomous-robot startups is that no one knows how to do the AI. It may make sense to you that you ought to be able to put together robots like you put together a spreadsheet, but that's not how it turned out to be. The article admits this, describing day to day tasks as "incredibly complex". Yet then asks us to imagine the cheap, capable robot.
Fourth, Microsoft has Robotics Studio, a package that competes with ROS in many respects. It is exactly a "robot operating system" on top of Windows. Researchers pretty much ignore Robot Studio, but why does this journalist do so when he says that "what we need Microsoft [...] to do is build an operating system....". It has been trying for years. Thanks for the advice, though.
A low-quality piece for Spectrum, picked up from a blog.
I'm thinking it has something to do with the fact that they have been the next revolution since Karel Čapek gave us the term in 1920. If you haven't read R.U.R. you should. It feels like it belongs in last weeks Wired.
After reading R.U.R., it's amazing how many themes of the subsequent 90 years' worth of dystopic literature it predicts! Robot rebellion, androids, human fertility crisis - it's all there, and more.
It's also kind of ironic that the same piece of literature that introduced the concept of the robot also introduced the perennial concept of the coming robot rebellion.
It's not actually that ironic. Robots tend to be symbols for an oppressed underclass - their masters have (or think they have) a high degree of control over their every thought and action, they're often tasked with menial or dangerous labor, and frequently considered 'other' and 'less' than their squishy human overlords. And what's every slave driver's worst fear?
As literary devices, robots don't have much to do besides rebel. If they just did what we told them to do, there wouldn't be much story to tell. (Note that that's part of why _I, Robot_ was so mind blowing - a lot of it was about things going wrong when the robots did exactly what we told them.)
I think it had more to do with Čapek subscribing to Marxist theories about historical inevitability and such. His robots are in fact a mental abstraction, a caricature of early 20th century proletariat that was supposedly bound to revolt, sooner or later. Čapek is taking this to extremes, purporting to show that even a race that was bred especially for labor would inevitably revolt, if only they were afforded the modicum of intelligence that was needed for that labor.
Of course, we now know that Marx's inflexible theories were wrong, and proletarian revolution is by no means inevitable. We are, however, still stuck with a half-baked idea of the inevitable robot revolution, which tinges the whole field of robotics research, at least as seen from the outside, with a subconscious bitterness of fear.
If you were thinking multipurpose C3P0 style robots, then the revolution has been 5 years in the future for the last 30 years.
However single-purpose robots have become part of the fabric of most industries. Heck, most people don't even think of them as robots. So the revolution happened, it is now the new normal.
I had a similar thought the other day while using an automated paper towel dispenser. "Isn't this primitive robot?" Yet, I'd never really thought to apply the term "robot." It'll be interesting to watch the evolution from dumb/single-purpose robots to smarter/multi-purpose ones.
If that's a primitive robot, we've had primitive robots since the first push-button elevator or automatic door, many decades ago.
It's like the defining-down of AI. Proponents promised strong AI Real Soon Now for decades; now they say that we're just being unfair if we don't count email filters as "AI".
There's a (probably apocryphal?) story of a professor asking his students to design a robot to wash dishes. Most came up with something along the lines of Rosie from the Jetsons; only one offered the 'right' answer which was something along the lines of "why waste time designing when I can buy one from the appliance section at Sears?"
It's too bad there apparently aren't any robots ready to put into service in Fukushima. They might have robotic arms or cranes for doing things like shuffling fuel rods, but what can help with doing improvised welding or wiring in a hostile environment?
In recent months NHK reported on experiments in using simple robots that could help the elderly in Japan with a few chores, but they were a flop mainly because the people involved rejected them instead insisting on human companionship. There's a shortage of working-age people, and a growing percentage of retirement age people there. Many of the younger people stay in the larger cities making the shift even more dramatic in some areas. In the 60's many predicted flat-panel on the wall television sets to be in use by the 80's. It looks like general purpose robots will take longer than many would have hoped too.
Perhaps something remotely controlled by a human with VR gloves etc is viable for the emergency repairs? Things would have to be well shielded to avoid having stray neutrons altering data bits.
Believe it or not... It is fairly trivial to consume _all_ computational resources on the robot. A few examples of concurrently running tasks: AMCL SLAM (mapping); assembling incoming laser scans and transforming them to various frames; multiple occupancy grids (obstacle detection / planning) at various resolutions; 3D point-cloud perception for object recognition, person tracking, or 3D mapping; motion-planning controllers for mobility (base) or manipulation (arms) running at 100+ Hz; computer vision algorithms; the list goes on...
It is possible to offload some of these tasks (especially high-level planning or recognition tasks) to people via MechTurk or the cloud (see "Cloud Robotics"). However, there will always be some computation that needs to occur with real-time constraints -- which likely means on-robot.
The _big deal_ in the last 2 years is ROS. ROS has changed everything, and is causing robotics to accelerate! Now, if you need one of those previously mentioned tools... they're open source, well documented, and ready to use.
Robots need a lot more computing power than a desktop PC, maybe even more than a current supercomputer, because the computer off-loads the hard parts of interacting with the world to people. In a sense, you could consider programmers and users robot peripherals that let less powerful computers get more done.
That is actually something I have been thinking about recently. People spend hours upon hours playing games that resemble, to some degree, real jobs. (Think FarmVille, CafeWorld, etc.)
If you could craft the game in such a way to provide the control to the robots for things that are not easily automated, people's entertainment would provide the labour required to control the operation for free. A capitalistic dream.
Sorting recycling comes to mind. The manual labor version of it involves standing over a slow moving belt of stinky trash and picking out the valuable recyclables.
Seems like a Fanuc with a suction gripper and a hires web cam is all it might take to get a little "trashville" going. Give the housewives who collect the most alu while the kids are away at school day passes to local spas as prizes and you've got a win.
Very astute. I think you hit it dead on. The robot revolution will be about outsourcing menial tasks to cheaper humans, not super-duper AI. It's really more of the network revolution continued into servos, not a "robot revolution".
Isn't the world in an awful state when we value other humans' contributions on a lower level than robots. Not saying it hasn't been in the past, but one would have thought the wondrous fruits of Capitalism would have got us past that stage by now...
The Mturker's value for his time is set by him, not "us". If he doesn't derive enough value from doing the HITs to give a pile of servos on the other side of the world its smarts, then he doesn't have to.
The reality is that a human can still do in a few seconds what it takes a rack full of S3 instances an hour to do. Yes, the human's contribution costs less in absolute terms and is readily available so is valued less.
Where the value to the turker and the value to the employer meet in the middle, you get the wondrous fruits of capitalism (small 'c', not a religion).
Extracting information and optimizing actions based on it via purchasing/timing decisions. Requires much less capital commitment and potentially much higher ROI.
FTA: "Buzzing you in when you get locked out, signing for a package, taking that frozen chicken out of the freezer while you’re at work, feeding your pet, and of course the veritable classic of robo-problems: getting you a beer." - imho these are terrible examples of what a robot could do... there are much simpler solutions for most of these problems than buying this: http://www.willowgarage.com/pages/pr2/overview
Looking for general purpose robots misses the evolution in home robotics already in progress. I've used a iRobot Roomba for vacuuming our home for about two years now, it works great. I expect future versions will be able to do more tasks.
Consider the progression of the iPod as a music player to a general purpose touch computer (running iOS apps of all kinds) that is is today -- this is perhaps a good picture of how home robotics may progress from single purpose to multi-purpose household machines.
It's the term robot. For most people that means human like machines. Making human like machines is a terrible waste of energy compared to just making the form of a machine match it's tasks.
If you instead say machine automation rather than robot, well then you see that we probably are in that revolution.
I've been working on a theory on this one.
In the most abstract sense, computers are an extension of the human self. Cave Paintings, Papyrus scrolls, Guitars, Printing Presses, Pens, Televisions, Computers are all an evolution of this. You can think of a mobile phone as a inanimate (non-sentient) robot that lacks the ability to self-ambulate. There is a distinction between sentience which has not yet been achieved, and partial autonomy. There simply is not a mass market use case that makes small scale robots like the finch popular enough. Instead, I believe that mobile devices will continue adding new and interesting features. NFC will be the big thing next. After that, some new pathways will be: printing from mobiles, and possibly some of them will have movable parts that developers can program.
Eventually, the mass market appeal of mobile devices will combine with the usefulness of a physical manifestion of our ideas into reality. You need to make a cheap robot USEFUL.
Ultimately, it is the person that controls the robot that will get the most use out of it. Just like a master craftsperson can use his or her tools far more effectively than any random person.
So...in one idea, the reason robots are not really here is that there is no mass market appeal to justify the hundreds or thousands of dollars that they cost.
I have been thinking about this a lot lately. All the parts are laying around, and I just need to start putting them together..literally, I have steppers and servos, and some DSP test boards. I am getting de ja vu all over again. When I was finishing up school, i commented that if I had a little more sense i would just quit, and make web pages. Here it is 15 years later, and about time to close up the pottery shop. Zafka Robotics does have a nice ring to it.
iRobot is amazingly successful not just in home robots but also in the military and public sectors. I love my Roomba and I'll be purchasing the Scooba 230 too.
There are very real technical limitations that make the kinds of robots available limited. Tele-operation is ok for the military, but interfaces aren't good enough yet to, say, enable a human size robot to open a door quickly. Automated driving is up and coming, but still a few years away from being a real product.
This is a combination of perception and manipulation problems. There is only so much that can be done by a robot moving from point a to point b, and we're seeing products that leverage that. Roomba, toys, packbot, grand challenge bots - none of them really see, touch, and grab things. This isn't a matter of some discoveries in a university that are waiting to be productized. No one has found the right answer.
I left a job in robotics to work on more interesting products, but I hope to return when the tech is real.
I almost got a Roomba but I went with a Dyson instead because all reviews said you still need to vacuum. I'll probably get the next generation, which should be out in a year or so.
Moore's Law is a bit flat at the beginning. For example, only 1% of the human DNA was sequenced at the first half of the project and the other 99% during the second half, if I remember the story correctly. We need a few more 18 month periods before robots start to become more useful in a non-controlled environment. My money is on 2015-2018.
It's all about price points. The crux of the problem is glanced in the last paragraph of the article "Imagine a robot that you could buy at Best Buy for somewhere between $2k and $4k." Right now, that just isn't feasible. For autonomous mobility you need components that qualify the following criteria: fast, quiet, safe, and precise. Unfortunately, such components just aren't affordable yet.
The revolution frenzy won't hit until your upper middle class income family can afford it.
It seems to me the problem is that there is so much variation in the underlying robotics hardware that it will be difficult to come up with a 'one-size-fits-all' OS.
the hardware, as usually, is abstractable into drivers.
Take for example hand movement - you may have the model of the robot's hand including actuators' input/output abstractions, length/weight of the arms, etc... It is common for all the robots with arms and very mathematically-mechanically complicated task to calculate the dynamic of the hand from current position, at current speed to another position with another speed, including recoil on the rest of the body, ... Mathematically it is a complicated smooth manifold in the high dimension space and various analyses and optimizations on such objects have been a very fruitful source of many Ph.D.s and articles :) Though on practice any "good" suboptimal solution would do. The 3D orientation and environment sensing is also abstractable - here again the calculation and effective algorithms (which are really hardware independent - for example stereo analysis is dependent basically on the resolution and the speed of the input sensors and teh CPU power and effective algorithm - sounds familiar?) is the heavy part. The [speech] command processing - has been also abstracted into input devices and processing core algorithms.
<fanciful-speculation>
That's why I wouldn't be surprised to see Apple, with its superior hardware/software integration, on the forefront of the consumer robotics revolution.
</fanciful-speculation>
One big reason robots aren't very good is because actuators are not very good. There are no good artificial muscles. Motors are great to power wheels and propellers but not much else. On the sensor side, robots can see, hear and even touch quite effectively. I think the software is slowly making progress, as is the # of computations we can do per watt. But lack of actuator progress severely hampers the future of robotics IMO.
Hydraulics and pneumatics work pretty good, but they are really tedious to work with, especially for people used to electronics. Their biggest problems are cost and seals, especially seal durability. And of course the cost in energy to keep them powered - when I get around to it, I expect to need to use a IC powered version out in the garage. But then energy budget is a problem with all mobile robots.
I figured robots haven't gained momentum at anywhere the same rate as personal or distributed computing since robust sensing and interaction of the physical world is still mostly an unsolved problem. Robots are useful in carefully physically constrained areas like industrial assembly lines and can be finicky toys that are too limited to do much anything useful, and there are things like Roombas that are somewhere in between the two.
A solution to the robust physical interaction thing sounds like something that would be pretty much AI-complete and would imply a lot bigger changes than housecleaning robots. This would just get you the science fiction humanoid robot that can walk into a house and start cleaning up and renovating the walls, but given the amount of flexible smarts and learning ability a general solution for being able to pull that off involves, it could probably start doing your job for you as well.
I'd guess there is also a less scary technology branch, where the current limited scope robots become somewhat more feasible with better manufacturing, materials, computing and communication technology, taking use of things like human telepresence and heavy duty data mining that have become easier due to just plain Moore's Law instead of groundbreaking AI research. These kind of robots are going to require environments tailored to their limitations, like the existing assembly line robots, as well as doing the R&D on figuring out just what kind of robot and environment combinations are going to do useful stuff, so it's not going to be the sort of lightning-fast development we saw with personal computers. More like the gradual development from 19th century homes to late 20th century homes full of electrical household appliances.
In fact, electrical household appliance probably is the metaphor for thinking about these sorts of robots, not a metal man walking about and enslaving humankind. That might explain why people aren't vastly enthused about them, even though there's probably a market for them.
Robots are the next revolution, but not because of the examples listed in the articles. I don't get that excited to hear about robots getting my packages, cooking my dinner, talking to my (future)wife, etc...I think the real revolution is with robots that can provide oxygen to people in burning buildings, locate/extract people under debris after a natural disaster, and find my lost car keys.
As for the robot in every home, I think a cheap, easy-to-program home automation processor that connects wirelessly to custom hardware is more appealing. That way I could set my bird-feeder, alarm system, and central-heating to "vacation-mode" and leave with peace-of-mind.
However, if someone could create a general way to power robots with intelligence, i.e. plug a robot into the intelligence grid and there it goes, they'd be billionaires.
Comparing the software revolution with a robot revolution is just unfair. The hardware part that fueled the software revolution was multi-purpose, so there was a fixed initial cost while the cost to develop the subsequent (amateur) programs was negligible.
The develop-test-debug cycle is also much faster in the software industry.
But I do dream of a world where Lego Mindstorms are widely available and used as any other toy.
Seems to me that until we have robots that collect energy and resources and build other robots - or alternatively, some kind of programmably reconfigurable substratum like a low-tech T-1000 (Terminator 2) - the capital costs are hard to overcome.
Exactly! that is why I am so interested in Self-reconfiguring modular robotics, there are several companies that are starting to offer modules to the general public, although all but one are for manually reconfigured system. But I see a lot of movement in both business and research over the last years. I think that it is likely that you will be able to buy reasonably priced Self-reconfiguring systems in 5-15 years. and in the meantime you can play with cublets :-) http://www.modrobotics.com/blog/?p=187 more on SRCMR http://en.wikipedia.org/wiki/Self-reconfiguring_modular_robo...
what about the robots in people's pockets. iPhones and Androids can do so many things and have so many sensors. These are robots, just not in the shape we imagined they will take. And they are not yet in the autonomous decision level we imagined they will be.
Disclaimer: I'm a masters student in robotics and don't use iPhone or Android (yet)
Really interesting idea. While I know nothing about robotics, this makes me wish I did.
Does anyone know the startup costs of building a robot vs. making a computer in the 70s? My guess is the cost and complexity of building a robot far exceeds that of building your own computer (even from just a # of required parts perspective). Then again, a walking robot like the one Honda has dumped millions into may just be the wrong way of thinking about it. Perhaps a rumba-esque creature is a better indicator of what the first PR (personal robot) will look like.
That being said, if you've ever heard of 'charity burnout' there is something similar called 'robotics burnout.' You discover it, you engage in excitement, but nothing happens, you can only be excited (or charitable it seems) so long before you need to take a break. But lets start at the beginning shall we?
So in 1985 the Homebrew Computer Club was waning, it had been the center of attention between 75 and 85 but with people like IBM in the game and Apple going much more 'corporate' it was more of a users group than the folks who were changing the world.
One of the special interest groups from that meeting was the Robotics SIG. A guy named Dick Prather who was active in that SIG decided that even if the computer group was dying, robotics was just getting started and so he did the SIG equivalent of a setsid(2) call and made the Homebrew Robotics Club an independent organization. It has met continuously since then (yes 26 years).
One of the things about 'robots' that most people don't get, is that fundamentally a robot is any machine that has some level of program-ability that does one or more tasks while adapting to its environment. Your dishwasher is a good robot, it washes the dishes for you, or the pots and pans, or the stem ware, it uses a variety of sensors to decide if the dishes are clean yet and it dries them afterwards. If you didn't have it you would be getting a sore back moving dishes from the counter-top to the sink, to the drying rack.
Folks have argued that to be a robot it would have to do it like humans do, but that is an angels-on-pins sort of argument. Generally robots are the expression of automation, and they have (as a market) been growing where ever it makes economic sense. As processors get cheaper and more powerful more and more things make economic sense.
Of course for things that are really expensive or really dangerous its really easy to justify the cost. So for things like disarming bombs, or hunting people down in a country you are not technically at war with and killing them are both easily justified costs if you can automate them.
ISRobotics, the guys who make the Roomba, make most of their money selling robots for things like mine clearing and recon, and yes shooting people. Founded in 1990 they were making $1M/yr in 1996 with 16 employees[1], that wasn't particularly sustainable but in 2001, with 9/11 they demonstrated the value of their packbots. And now at $401M/yr they are doing quite well except that over 60% of their revenue is "G&I" which is code for "Government and Industrial".
Now that isn't all bad, its just that instead of comparing the robotics revolution to the 'PC' revolution you have to compare it to the 'computer' revolution, which is to say that a whole lot of investment and development is focused on corporate and government use until the the industry can supply enough automation to slake their thirst and to give some time to 'regular' folks.
So in the 80's we had companies like Androbot and Heathkit approaching mobile robotics as gimmicks/educational, in the 90's we had a lot of toys and the start of robot (really armored R/C vehicles with minimal automation) combat. And now in the second decade of the 21st century we're seeing prototype self driving cars, a number of walking designs, and energy systems that can sustain things for more than a couple of minutes.
Bottom line, things are going along, and getting better, and in many ways getting better faster now than they have in the past, but there is so much eco-system that has to develop around building mobile robotics that can do useful work that the rate of change feels much more evolutionary than revolutionary.
When we look at robotics startups today, places like Willow Garage and Anybot you can get a good feel for how cool things could be and how far we are from seeing the kind of uptake like we did with personal computers.
[1] http://money.cnn.com/magazines/fortune/fortune_archive/1996/...
[2] http://investor.irobot.com/phoenix.zhtml?c=193096&p=irol...