It creeps and scrawls.
June 12, 2008 1:17 PM Subscribe
My robot-that-can-draw-stuff idea has been patented and designed and prototyped to death hundreds of times, but never (to my knowledge) developed into a full scale product. Has anyone built something I can buy? Even a kit?
My idea is a drawing robot that works sort of like a mobile plotter for drawing with chalk, pencil, pen, cutting with a knife blade, painting with a brush, whatever. It uses omnidirectional wheels and stepper motors to move precisely in arbitrary directions without turning. You set it on a flat surface (or hold it against a wall) and trigger it with a button, at which point it draws whatever the positioning script on its a memory card says to draw, raising and lowering the pen/brush/whatever at the appropriate times during the drawing process. It's sort of a LOGO Turtle for the 21st century, but I'd like to use it to draw murals and make signage (in lieu of using a series of stencils or guides or something like that). The two keys are a) it's small and mobile - plotters and sign cutting machines tend to be enormous and b) you don't have to know how to draw. The 'invisible hand' does it for you based on the vector graphics you've fed it.
Now I've seen these guys doing something similar with a spray can (although theirs is tethered), and there's this guy who did it with a basic board and a homemade omniwheel, and of course the original series of LOGO Turtle robots, and probably 1,673 others out there (MAKE magazine has a stub) building similar prototypes. And that's wonderful!
BUT - has anyone developed this idea into something halfway practical (or even a toy!) instead of just a one-off art curiosity or hobby project? (You know...for the kids!) If it has, where can I buy it? If it hasn't, can I develop and sell it? Or is the idea considered to be in the public domain because it's been done to death over the last three decades?
Thanks!
My idea is a drawing robot that works sort of like a mobile plotter for drawing with chalk, pencil, pen, cutting with a knife blade, painting with a brush, whatever. It uses omnidirectional wheels and stepper motors to move precisely in arbitrary directions without turning. You set it on a flat surface (or hold it against a wall) and trigger it with a button, at which point it draws whatever the positioning script on its a memory card says to draw, raising and lowering the pen/brush/whatever at the appropriate times during the drawing process. It's sort of a LOGO Turtle for the 21st century, but I'd like to use it to draw murals and make signage (in lieu of using a series of stencils or guides or something like that). The two keys are a) it's small and mobile - plotters and sign cutting machines tend to be enormous and b) you don't have to know how to draw. The 'invisible hand' does it for you based on the vector graphics you've fed it.
Now I've seen these guys doing something similar with a spray can (although theirs is tethered), and there's this guy who did it with a basic board and a homemade omniwheel, and of course the original series of LOGO Turtle robots, and probably 1,673 others out there (MAKE magazine has a stub) building similar prototypes. And that's wonderful!
BUT - has anyone developed this idea into something halfway practical (or even a toy!) instead of just a one-off art curiosity or hobby project? (You know...for the kids!) If it has, where can I buy it? If it hasn't, can I develop and sell it? Or is the idea considered to be in the public domain because it's been done to death over the last three decades?
Thanks!
Best answer: Such products have been developed as toys, yes.
Back in 1990 I had a Lego kit that did this - a quick google reveals there are websites dedicated to scanned lego instructions(!) and that my toy was a "Technic Control Center (8094-1)" - here's a picture. I've heard it called an 'artbot'.
There's an updated version availiable as a matter of fact: Have a look for the "Lego Robotics Invention System 2" - see here for a picture.
The new and old designs look pretty different; the early one is a 2-wheel differential drive robot with the pen at the center, so you can turn on the spot to draw a corner without lifting the pen. The newer version has to accommodate a big 'control' brick so it moves the pens to the front and back, and has four wheels - this complicates the kinematics substantially if you want to draw certain shapes.
posted by Mike1024 at 2:03 PM on June 12, 2008 [1 favorite]
Back in 1990 I had a Lego kit that did this - a quick google reveals there are websites dedicated to scanned lego instructions(!) and that my toy was a "Technic Control Center (8094-1)" - here's a picture. I've heard it called an 'artbot'.
There's an updated version availiable as a matter of fact: Have a look for the "Lego Robotics Invention System 2" - see here for a picture.
The new and old designs look pretty different; the early one is a 2-wheel differential drive robot with the pen at the center, so you can turn on the spot to draw a corner without lifting the pen. The newer version has to accommodate a big 'control' brick so it moves the pens to the front and back, and has four wheels - this complicates the kinematics substantially if you want to draw certain shapes.
posted by Mike1024 at 2:03 PM on June 12, 2008 [1 favorite]
If it hasn't, can I develop and sell it? Or is the idea considered to be in the public domain because it's been done to death over the last three decades?
This is about technical approach, not marketing feasibility:
The problem you will hit, which is probably also why these things don't seem to be popular or easy to find, is that dead-reckoning without external reference points - the navigation method you describe - doesn't really work very well in the real world. Deceptively badly in fact.
In order to draw an image, the robot has to know where it is, and where it's facing. To do this, you need either an external reference point, or excruciatingly controlled conditions, or only need simple lines without much accuracy.
We can use our eyes, or move our hand relative to the position of our body, etc, but a robot that navigates by dead reckoning will rapidly accumulate navigational errors. If you get it to draw a simple square, the start point and the end point won't quite align, but you might not notice this as it has four sides that are pretty much at right-angles, so that's a square. But four lines isn't so interesting - try to do a picture, and the ever-enlarging error will become very noticeable - if you've ever tried to draw a complex picture blindfolded, without touching the paper (just stand there holding the pen against the paper and move the pen), you'll have seen the problem.
If an art piece requires exact conditions and produces simple things with some distortion, well, that's cool, that's art. If a consumer device does the same thing, that's a return and a refund. This is problem enough using contactless media like a spray can. Try it with variable-resistance like a cutting blade and things will get crazy even faster.
Solution:
Roboticists will likely advise you to forget the dead reckoning approach and build it using external reference points. perhaps something along the lines of "electronic road cones" - place three of them in a circle around your "canvas" area, and the robot uses laser position tracking (or whatever) relative to the cones. The cones must not move during the drawing, unless you intend to introduce distortion for the fun of it :-)
This approach has the advantage over dead-reckoning in that it can succeed, but is more expensive, which may explain why it's not a consumer product (that I know of (yet)) :-)
posted by -harlequin- at 2:13 PM on June 12, 2008
This is about technical approach, not marketing feasibility:
The problem you will hit, which is probably also why these things don't seem to be popular or easy to find, is that dead-reckoning without external reference points - the navigation method you describe - doesn't really work very well in the real world. Deceptively badly in fact.
In order to draw an image, the robot has to know where it is, and where it's facing. To do this, you need either an external reference point, or excruciatingly controlled conditions, or only need simple lines without much accuracy.
We can use our eyes, or move our hand relative to the position of our body, etc, but a robot that navigates by dead reckoning will rapidly accumulate navigational errors. If you get it to draw a simple square, the start point and the end point won't quite align, but you might not notice this as it has four sides that are pretty much at right-angles, so that's a square. But four lines isn't so interesting - try to do a picture, and the ever-enlarging error will become very noticeable - if you've ever tried to draw a complex picture blindfolded, without touching the paper (just stand there holding the pen against the paper and move the pen), you'll have seen the problem.
If an art piece requires exact conditions and produces simple things with some distortion, well, that's cool, that's art. If a consumer device does the same thing, that's a return and a refund. This is problem enough using contactless media like a spray can. Try it with variable-resistance like a cutting blade and things will get crazy even faster.
Solution:
Roboticists will likely advise you to forget the dead reckoning approach and build it using external reference points. perhaps something along the lines of "electronic road cones" - place three of them in a circle around your "canvas" area, and the robot uses laser position tracking (or whatever) relative to the cones. The cones must not move during the drawing, unless you intend to introduce distortion for the fun of it :-)
This approach has the advantage over dead-reckoning in that it can succeed, but is more expensive, which may explain why it's not a consumer product (that I know of (yet)) :-)
posted by -harlequin- at 2:13 PM on June 12, 2008
(Incidentally, dead-reckoning has a bit of a reputation in robotics circles - it's such a simple, obvious and intuitive solution that everyone initially latches onto it very strongly and refuses to hear that it doesn't work, and pours excessive time and money down the drain convinced that they're on the cusp of fixing the drift in their robot. It's a phase you have to go through, like puberty.)
posted by -harlequin- at 2:22 PM on June 12, 2008 [1 favorite]
posted by -harlequin- at 2:22 PM on June 12, 2008 [1 favorite]
As to the commercial feasibility of making and selling such a product, well, that's a complicated matter.
When you navigate by measuring wheel movement you're using what is sometimes called 'odometry'. It's popular on robots because it's cheap and works fairly well. The key disadvantage in your application is accuracy: you might only have an error of 1mm per meter, but if you had a 1m by 1m bit of paper and you wanted to infill part of it with your pen you could end up driving 10 meters easily, then you've got 10mm of error and suddenly you get very visible distortion. There are also bothers like wheels slipping, particularly if you're sign cutting or similar - and users would need a hard, flat floor (or large table) as plotting on carpet would be hard to achieve.
In other words, you might have difficulty achieving output quality matching fixed plotters/sign cutters, and you would need that output quality to compete in the professional market. You might find it difficult to make enough profit to cover your development costs.
Also your robot might take up more floor space as, when running, the bit of paper would have to be flat on the floor. Large-format plotters, on the other hand, can roll the paper (see here) to take up less than one sheet of paper's floor space. So your robot would be good for people who only need to plot once in a while, as it would take up less storage space, but people who kept the plotter permanently set up might prefer a normal plotter. Obviously, if you're selling to the 'occasional plotter' market you compete with high street and internet printing services.
In summary I think it's a neat idea, but you might be better to pursue it as a hobby rather than as a commercial product.
posted by Mike1024 at 2:35 PM on June 12, 2008
When you navigate by measuring wheel movement you're using what is sometimes called 'odometry'. It's popular on robots because it's cheap and works fairly well. The key disadvantage in your application is accuracy: you might only have an error of 1mm per meter, but if you had a 1m by 1m bit of paper and you wanted to infill part of it with your pen you could end up driving 10 meters easily, then you've got 10mm of error and suddenly you get very visible distortion. There are also bothers like wheels slipping, particularly if you're sign cutting or similar - and users would need a hard, flat floor (or large table) as plotting on carpet would be hard to achieve.
In other words, you might have difficulty achieving output quality matching fixed plotters/sign cutters, and you would need that output quality to compete in the professional market. You might find it difficult to make enough profit to cover your development costs.
Also your robot might take up more floor space as, when running, the bit of paper would have to be flat on the floor. Large-format plotters, on the other hand, can roll the paper (see here) to take up less than one sheet of paper's floor space. So your robot would be good for people who only need to plot once in a while, as it would take up less storage space, but people who kept the plotter permanently set up might prefer a normal plotter. Obviously, if you're selling to the 'occasional plotter' market you compete with high street and internet printing services.
In summary I think it's a neat idea, but you might be better to pursue it as a hobby rather than as a commercial product.
posted by Mike1024 at 2:35 PM on June 12, 2008
Eh, if you are just looking to have somebody build a prototype or build one to spec for you, or help you figure out if it IS marketable, Hanson Robotics can do it.
posted by Unicorn on the cob at 3:59 PM on June 12, 2008
posted by Unicorn on the cob at 3:59 PM on June 12, 2008
Response by poster: - dead-reckoning without external reference points
- you're using what is sometimes called 'odometry'
I'm not sure where either of those come into play with my design because I'm not using dead reckoning (estimating distance by time). That's why I specifically mentioned stepper motors. The "external reference point" would be determined by how many steps the thing has traveled relative to the starting point (or more practically, the "real" distance would be calculated by taking the wheel diameter, like 2", and step size, like 400 steps per revolution, into account). My homemade CNC machine can go millions of steps without losing accuracy, and that's just three stepper motors and three matching screws. A self-carrying machine shouldn't (shouldn't!) be any different. Of course there's things like wheel slippage and vibration problems, but risk of slippage seems like a fair price to pay for a portable plotter/cutter/whateverer. As for variable resistance - as long as the torque in the motor is great enough to overcome whatever is resisting it, the steps ought to have the same amount of accuracy. 10m north followed by 10m south ought to get me back to square one, every time, assuming there's no slippage or skipped steps or other (overcome-able) malfunctions. (Engineers reading this, does all this sound correct? Or just confused?) Typically they're fixed on a machine and used to drive other things, like leadscrews or a floppy drive platform, but I'm going to be using them as a locomotion system on two axes.
The thing definitely wouldn't be used for standard plotting applications like blueprints and large scale drawings - plotters are definitely more practical for that. It would be for more of an ad hoc application like a large home-made placard or very small run of signs or a mural, craft-type stuff. I think there's a market for that, at least, judging from all the craft magazines and TV shows and instructional DVDs and internet sites.
...(deep breath) thanks very much for the discussion, but I'm not here to vet the design, just to see if there's something out there now. The closest commercial product I've seen is the Lego kit, which seems to work pretty well but isn't "big vision" enough for what I want to do.
- Hanson Robotics can do it
Wow, they look extremely expensive. And that Evil-Dead-2-Movie-Poster-looking skull on their navigation menu? Yeah, that freaks me out pretty severely. (Or is it Lou Reed I'm thinking of?) And..um..Frubber? Shudder.
Thanks for your answers!
posted by ostranenie at 4:43 PM on June 12, 2008
- you're using what is sometimes called 'odometry'
I'm not sure where either of those come into play with my design because I'm not using dead reckoning (estimating distance by time). That's why I specifically mentioned stepper motors. The "external reference point" would be determined by how many steps the thing has traveled relative to the starting point (or more practically, the "real" distance would be calculated by taking the wheel diameter, like 2", and step size, like 400 steps per revolution, into account). My homemade CNC machine can go millions of steps without losing accuracy, and that's just three stepper motors and three matching screws. A self-carrying machine shouldn't (shouldn't!) be any different. Of course there's things like wheel slippage and vibration problems, but risk of slippage seems like a fair price to pay for a portable plotter/cutter/whateverer. As for variable resistance - as long as the torque in the motor is great enough to overcome whatever is resisting it, the steps ought to have the same amount of accuracy. 10m north followed by 10m south ought to get me back to square one, every time, assuming there's no slippage or skipped steps or other (overcome-able) malfunctions. (Engineers reading this, does all this sound correct? Or just confused?) Typically they're fixed on a machine and used to drive other things, like leadscrews or a floppy drive platform, but I'm going to be using them as a locomotion system on two axes.
The thing definitely wouldn't be used for standard plotting applications like blueprints and large scale drawings - plotters are definitely more practical for that. It would be for more of an ad hoc application like a large home-made placard or very small run of signs or a mural, craft-type stuff. I think there's a market for that, at least, judging from all the craft magazines and TV shows and instructional DVDs and internet sites.
...(deep breath) thanks very much for the discussion, but I'm not here to vet the design, just to see if there's something out there now. The closest commercial product I've seen is the Lego kit, which seems to work pretty well but isn't "big vision" enough for what I want to do.
- Hanson Robotics can do it
Wow, they look extremely expensive. And that Evil-Dead-2-Movie-Poster-looking skull on their navigation menu? Yeah, that freaks me out pretty severely. (Or is it Lou Reed I'm thinking of?) And..um..Frubber? Shudder.
Thanks for your answers!
posted by ostranenie at 4:43 PM on June 12, 2008
Best answer: I'm not sure where either of those come into play with my design because I'm not using dead reckoning (estimating distance by time). That's why I specifically mentioned stepper motors.
No, the stepper motor method is what I'm talking about when I say dead-reckoning (by which I mean calculating your position from records of your travels from a known start-point made via tracking your heading and distance traveled).
The "external reference point" would be determined by how many steps the thing has traveled relative to the starting point (or more practically, the "real" distance would be calculated by taking the wheel diameter, like 2", and step size, like 400 steps per revolution, into account). My homemade CNC machine can go millions of steps without losing accuracy, and that's just three stepper motors and three matching screws.
This is what I mean by how seductive dead-reckoning is. It all sounds great, but in the real world, you discover that your wheel diameter isn't quite mathematically perfect. No problem, you can improve that. Then the wheel traction isn't perfectly uniform either. Well you can work on that too. Then you figure out that the tires are picking up dust and hair, and you get to work on that, and every step of the way you think that once you solve this or that, it'll work. But it doesn't - you just end up winding the requirements higher and higher until it ceases to be practical. A mobile robot navigating in this way is not like the fixed armatures of a CNC machine, in several different critical ways.
Dead-reckoning can do some things, but for the practical device you describe, it's the wrong tree to bark up.
posted by -harlequin- at 5:17 PM on June 12, 2008
No, the stepper motor method is what I'm talking about when I say dead-reckoning (by which I mean calculating your position from records of your travels from a known start-point made via tracking your heading and distance traveled).
The "external reference point" would be determined by how many steps the thing has traveled relative to the starting point (or more practically, the "real" distance would be calculated by taking the wheel diameter, like 2", and step size, like 400 steps per revolution, into account). My homemade CNC machine can go millions of steps without losing accuracy, and that's just three stepper motors and three matching screws.
This is what I mean by how seductive dead-reckoning is. It all sounds great, but in the real world, you discover that your wheel diameter isn't quite mathematically perfect. No problem, you can improve that. Then the wheel traction isn't perfectly uniform either. Well you can work on that too. Then you figure out that the tires are picking up dust and hair, and you get to work on that, and every step of the way you think that once you solve this or that, it'll work. But it doesn't - you just end up winding the requirements higher and higher until it ceases to be practical. A mobile robot navigating in this way is not like the fixed armatures of a CNC machine, in several different critical ways.
Dead-reckoning can do some things, but for the practical device you describe, it's the wrong tree to bark up.
posted by -harlequin- at 5:17 PM on June 12, 2008
You seek the help of an open source community for the sake of marketing a product. Ick.
But as a builder of paintbots this post got me thinking about using a GPS for navigation. I'm going to try this when I get back to the States / my studio. In the mean time.. anyone here tried this yet? My Garmin hooks up to USB, is there a way I could get live data from it?
posted by miffed at 7:59 PM on June 12, 2008
But as a builder of paintbots this post got me thinking about using a GPS for navigation. I'm going to try this when I get back to the States / my studio. In the mean time.. anyone here tried this yet? My Garmin hooks up to USB, is there a way I could get live data from it?
posted by miffed at 7:59 PM on June 12, 2008
Response by poster: - from a known start-point made via tracking your heading and distance
I like to think this device would do something "in-between" dead reckoning and external referencing because there's no heading. It should never turn. Just up-down-left-right (or if you prefer, up-up-down-down-left-right-left-right-B-A-select-start.) It's two distinct axes because the wheels have wheels on them. I'm not expecting miracles - I'm sure I'll run into all sorts of interesting problems between the kooky wheels and the vibrating stepper motors and whatever else comes up. And I'll come up with all sorts of interesting solutions before I either give up on this science project (likely) or turn it into a vocation (much less likely). Appreciate the comments!
- You seek the help of an open source community for the sake of marketing a product
Did you read the whole post? I wanted to know if something like this existed already and if so, where I could buy one. If it's not been developed, maybe I could be the one to roll it out. If I wanted to market it seriously, the last place I would go for ideas is metafilter. It's more of a hobby/pipe dream/interesting side project. I'm not interested in striking it rich based on a questionably original idea, but more the practical application of that idea. Why did you think I linked to MAKE?
- But as a builder of paintbots
You mean the remote-control, wheeled, spray-paint kind?
- got me thinking about using a GPS for navigation
For the robot, you mean? I would imagine most commercial units are too inaccurate for anything more than a meter...but if you wanted to link a GPS unit to a robot you might be better off using the standard 4800bps serial connection the old units output and processing the string with an MCU. (Did my thread just get hijacked?)
posted by ostranenie at 8:50 PM on June 12, 2008
I like to think this device would do something "in-between" dead reckoning and external referencing because there's no heading. It should never turn. Just up-down-left-right (or if you prefer, up-up-down-down-left-right-left-right-B-A-select-start.) It's two distinct axes because the wheels have wheels on them. I'm not expecting miracles - I'm sure I'll run into all sorts of interesting problems between the kooky wheels and the vibrating stepper motors and whatever else comes up. And I'll come up with all sorts of interesting solutions before I either give up on this science project (likely) or turn it into a vocation (much less likely). Appreciate the comments!
- You seek the help of an open source community for the sake of marketing a product
Did you read the whole post? I wanted to know if something like this existed already and if so, where I could buy one. If it's not been developed, maybe I could be the one to roll it out. If I wanted to market it seriously, the last place I would go for ideas is metafilter. It's more of a hobby/pipe dream/interesting side project. I'm not interested in striking it rich based on a questionably original idea, but more the practical application of that idea. Why did you think I linked to MAKE?
- But as a builder of paintbots
You mean the remote-control, wheeled, spray-paint kind?
- got me thinking about using a GPS for navigation
For the robot, you mean? I would imagine most commercial units are too inaccurate for anything more than a meter...but if you wanted to link a GPS unit to a robot you might be better off using the standard 4800bps serial connection the old units output and processing the string with an MCU. (Did my thread just get hijacked?)
posted by ostranenie at 8:50 PM on June 12, 2008
Programming a robot similar to what you describe was given to the 3rd-year Software Engineering students at my university recently. 2 wheels with stepper motors, skid-steer, central pen, all controlled by a PIC... and you have to make it do drawings from HPGL, including curves.
posted by polyglot at 10:41 PM on June 12, 2008 [1 favorite]
posted by polyglot at 10:41 PM on June 12, 2008 [1 favorite]
But as a builder of paintbots this post got me thinking about using a GPS for navigation. [...] My Garmin hooks up to USB, is there a way I could get live data from it?
The $50 GPSes you see on ebay usually connect as a serial port (some with a built in USB to serial converter) then use a protocol called NMEA 0183. Some suppliers also offer their own protocols which are faster/contain more information, but these protocols depend on the supplier. See here and here for a start I guess.
Unfortunately you are likely to be disappointed by the precision affordable GPS units can offer. If you've got $10,000 to spend you can get a GPS set up accurate to 2cm in open sky; there is then a sort of gap in the market and you reach $100 GPSes and you're dealing with errors of a few meters.
GPS is interesting to play with, but with most hobbyists' budgets won't get better than a few meters of accuracy, I'm afraid!
posted by Mike1024 at 1:10 AM on June 13, 2008
The $50 GPSes you see on ebay usually connect as a serial port (some with a built in USB to serial converter) then use a protocol called NMEA 0183. Some suppliers also offer their own protocols which are faster/contain more information, but these protocols depend on the supplier. See here and here for a start I guess.
Unfortunately you are likely to be disappointed by the precision affordable GPS units can offer. If you've got $10,000 to spend you can get a GPS set up accurate to 2cm in open sky; there is then a sort of gap in the market and you reach $100 GPSes and you're dealing with errors of a few meters.
GPS is interesting to play with, but with most hobbyists' budgets won't get better than a few meters of accuracy, I'm afraid!
posted by Mike1024 at 1:10 AM on June 13, 2008
This thread is closed to new comments.
If the idea is in the public domain, you can develop and sell it.
Everything you wrote above is now in the public domain.
Best of luck with your project.
posted by JimN2TAW at 1:58 PM on June 12, 2008