ArchivesSyndicate This Blog |
IR RangingWednesday, February 18. 2009In which your narrator dodges the whole issue of computer vision... As I mentioned some time ago, Tommelise 2.0 has gone into production mode as a CNC milling machine ... until I get bored and build an extruder. Mind, my plane is to use Tommelise 2.0 to make the parts for a much cheaper Tommlise 3.0. In that one, the expensive Haydon linear steppers will be replaced with reprapped ones driven by cheap tin can steppers. One of my original motives for getting into Reprap was to get the technical background to fulfill a lifelong ambition, acquired at University, to print buildings. Printing large objects presumes that the product is much larger than the tools that make it. My inspiration stemmed from the recently deceased Professor Wolf Hilbertz, whose design studio I took as an undergraduate. Wolf conceived of the built environment as being both automatically constructed by extrusion machines. Similar machines would reclaim the materials of buildings no longer needed and use it in constructing others. His seminal thinking on the subject was published in Progressive Architecture in 1970. Wolf conceived of quite large machines building whole cities. My ambitions at the time were much more modest. My idea was for very personal spaces being built by a small swarm of machines no larger than shoe boxes. Interestingly, Dr. Bowyer entertained the same notion some years ago and went so far as to get his ideas off of paper and into reality after a fashion. He was kind enough to link me to his own work in this area recently. One of the big issues when designing extruding 'bots is figuring out where one is. That's nearly so difficult is figuring out where other 'bots in your swarm are so as to avoid crashing into them. About a year ago I happened across a Sharp IR range finding assembly that could, depending on the model, calculate distances out to 5 meters. For starts, I bought a couple of more nearsighted ones that can spot things as far away as 1.5 meters. It struck me that if you put one of these on an alti-azimuth mount run by a MCU not all that different from what we use with our Reprap machines, you ought to be able to map one's surrounds directly in spherical coordinates with very little computing power. I was at the University of Texas in the 1960's when Professor Agarwal and his graduate student Steve Underwood began trying to reconstruct 3D environments using binocular vision. It was, computationally, horribly costly then. The years since haven't reduced it from being a very wicked problem. How much more simpleminded to simply do rangefinding! All that had to wait, though, because my first priority was getting Tommelise 2.0 going. I reasoned that with a CNC capability constructing a workable alti-azimuth positioning system would be trivial. As you can see, it was indeed. I designed the parts for a simple system over the weekend and milled them out in a matter of a few hours. 
It is hard to describe the bliss one feels when one knocks out a stepper mounting plate in Art of Illusion using shop drawings of the stepper and after milling it discovering that all of the bolt holes are precisely aligned. I've leveraged the prototype board that I designed to test the I2C slave chip's ability to run a stepper motor via a 754410 into handling two NEMA 17 stepper motors that I happened to have on hand.
Trackbacks
Trackback specific URI for this entry
No Trackbacks
Comments
Display comments as
(Linear | Threaded)
Regarding your interest in automatic building construction:
http://openfarmtech.org/index.php?title=Global_Village_Construction_Set
They are creating and testing open source equipment for creating replicable sustainable villages/communities (with fullblown fablabs) from local materials. See the end of the page for the CEB press, which creates structural elements (walls) from dirt. It still needs labour, but is a practical step towards automation. See also the CNC torch cutting table on the wiki! They also have ideas about metal extrusion from scrap etc.
Please consider becoming a donating member (a True Fan) with 10$/month
Funny, I was going to point that out too. These guys at OpenFarmTech are doing a great job. I'm a proud supporter. Personal fabrication is central to their sustainable, replicable village.
Also check out 'contour crafting' (just google it). It's not as biomimetic, it's large scale 3D printing instead of meso-scale.
I guess I just get a little impatient with these spinoffs of the "Five Acres and Independence" folks. "http://www.amazon.com/Five-Acres-Independence-Handbook-Management/dp/0486209741 I first ran into this sort of thing in the flesh, so to speak with Max's Pot, which began at the University of Texas Architecture when I was there in 1970. Max's pot is still there and has rather metastasized in the interim. http://www.cmpbs.org/cmpbs.html It's been my observation that the sort of people who get involved with such movements are long on enthusiasm and seriously short on technical background. I'll remember to my dying day the stink of some of the original Max's Pot building efforts using building blocks made of sulfur, a byproduct of oil production in Texas. (shudder)
I'm not sure about their technical background, what's important is that they are really producing results. They only need money and a few more developers. And they are really doing just what you (and Reprap) are after: local automated fabrication for the masses, and in the same peer to peer spirit that drives the reprap community. Did you doubt Adrians ideas of exponential propagation of the repraps when all there were was an experimental extruder? Can you at least appreciate the open source tractor they have prototyped, which costs about one tenth of of the closest commercial competitor? The same for a sawmill for local timber from local forests. All the equipment you need to create susteinable villages from scratch for some 10 000$, including labour. Whit that (plus a small group of teachers) you could jumpstart an African country after a war by teaching the poor to make the tools for themselves on others around them. That would cost a fraction of the amounts of money that is dumped in the developing countries.
I won't bother you anymore, but see the video presentations on their frontpage at least.
It wasn't my intent to be disrespectful. I am sure that they are passionate about what they are doing. I haven't, however, been very attracted to that sort of thing.
Interesting, didn't know that was your motivation. I'm interested in contourcrafting but on a smaller scale.
Those sharp sensors are really beautiful tools, used a bunch of them six years ago in a video installation to detect people walking true a projection. 6 sensors in a row formed a grid. The sensors where rated up to 3 meter but actually you got more range. Problem was that it was an urban area and bicycle reflectors (was in holland) always interfered with the sensors, even from 10 meters if they where in line of sight.
I remember comming accross this, http://craft.usc.edu/Description.html, last year and being quite impressed with the ability to print whole concrete walls. Printing at least the shell of a building looks like it's not that far away.
I wonder if a ceramic / concrete + trowel extruder would be possible on a reprap? Your not going to be able to print buildings, but there are all sorts of possibilities.
I find myself more than a little annoyed that Behrokh Khoshnevis pretends that what he is doing is original. Ed Allen first mooted the concept way, way back in 1970. The problem with Khoshnevis' implementation is that he still apparently does not recognize that a building shell is probably the cheapest part of a building. The subsystems are what kills the budget. Unless your extrusion system can make headway in doing subsystems from scratch, there is little to be gained from building such a system.
Hey, looking good! I made an X-Y mount out of a couple servos and a lot of polymorph plastic when I first got an Arduino, but I haven't done anything other than stick a laser pointer on it and slave it to my mouse.
I'm curious as to how the spot resolution of those Sharp sensors will affect your output - if the detection area is fairly large, will you end up with a fuzzy volume, or a sort of minimum clear area?
I'd love to do closed loop Z control even on a bog standard RepRap, just to try and level the bed before a build - one of those sensors might be just the ticket.
Looking forward to seeing some output from your sensor!
Wade
The D:S ratio of this kind of sensor, as best as I've been able to determine, is somewhere between 20-50:1. That means that you're going to get a fuzzy volume that is very distance dependent. If, however, you put the system on wheels, and move it around in the environment you are trying to may you will be able to sharpen it up considerably.
I think most of the entrants in the DARPA autonomous vehicle competitions use laser range finders for their mapping. To speed things up and simplify the design, the rangefinder itself is fixed and a mirror spins around.
LOL! Oh yeah, those are nice. I was at a conference a few months ago and met the guy who makes and sells those systems to the DARPA competitors. Sadly they cost about $75000. They're lovely things, though. I actually designed and built an IR mapping system that used the spinning mirror trick years ago when I worked for the South African Council for Scientific and Industrial Research. The problem with that approach when you're using IR, as opposed to visible radiation, is that you have to have a front surface specular reflective surface. To get that you've got to sputter aluminum onto a selected piece of float glass to make that happen. I don't have access to that kind of equipment right now like I did then.
Try your local amateur astronomer club, they will generally know someone who can relatively cheaply sputter aluminium onto glass for mirrors.
The spinning mirror system I built in South Africa cost some $12,000 in 1983 dollars. Mechanically, I don't think that too much has happened since then to improve on that level of expense. The IR sensor I'm using has a 40 msec sample acquisition time so the speed achieved with spinning mirrors isn't much help. You need to have your IR sensor stabilised at a temperature either much higher or much lower than what you are measuring as well as a chopper to get really high acquisition speeds, sadly.
|
