Cyclops

I did a quick lashup for doing some rough measurements with the Shart IR distance measurement instrument I bought.



Here you can see the instrument itself to the left and above the prototype USB controller board. I slapped on a few extra screw terminals to handle the three lines off of the instrument.




I was able to get it to spot a piece of 3 mm HDPE feedstock at 500 mm away.  This one has a range of about 1500 mm and a voltage range of 0-2.75v.  Given that it costs right at $12.50, that's not bad performance at all.  Sharp has another model which is focussed in to 300 mm.  That might a nice one to mount on a Reprap to scan an ongoing print job.

...like me. :-D

I've been mulling over ways to get my little Cyclops 3D scanner project going again with a minimum of effort. I decided to leverage what I already know how to do rather than get into all the drama of patching multiple scans that other open source 3D scanning systems like David do.

David uses a line projecting hand laser and a little PC web cam and a LOT of PC processing power to get you a 3D surface description of an object. It occurred to me that I already know how to make both encoders, gearmotors and steppers to work with PIC's. If I were to buy a little Sharp IR distance sensor like one of these.

I could shortcut much of the software sophistication that the Germans have gone through with David.  You point one of these little chips at your object and read off a voltage and then translate the voltage into a distance.  Mount one of these to swivel up and down and keep track of where it is at and put your object on a turntable with the same kind of tracking and you've got a very simple IR radar set that can easily develop a surface description of an object.

The only wild card was how much beam dispersal you got.  I talked with Matt at Acroname this morning and discovered that you can see the beam through an IR camera and that the dispersal that you can see off of the centreline is running no more than 0.8 degrees.  What that means is that most of the power in that spot is running in about 0.2-0.3 degrees off of centreline.

I bought a couple and look forward to seeing how they work. 

I can also leverage this tech into some robotic 3D imaging to let a mobile robot reconstruct its surroundings.  I have a project that has been on the back burner where that would be very handy.

The people that are doing SplineScan said that their VB6 code was not too efficient. I've been trying to move it over to VB.NET 2005 and it's pretty obvious that it isn't. I'd forgotten that it's been several years now that I've written anything in VB6.

Mind, they have what look like some good mesh fitting routines that I'll most likely use to good effect.

There has been a lot of interest in the RepRap community in the notion of creating an open source technology 3D scanner. The German Project David


and the British Project SplineScan

have already got well into the development of a cheap 3D scanner built up around webcams and a modified laser pointers. While David is freeware, SplineScan is proper open-source technology. David seems to get more traffic but both projects seem to produce very similar results.

Most of the emphasis around the RepRap community has been on the ability of such a scanner to create STL files of extant objects for replication in a RepRap (3D printer) machine. The issue of piracy and transmission of STL descriptions of patented or copyrighted objects has, of course, been much discussed.

I find myself not much interested in having a 3D scanner as such, not so much because I have any great aversion to copying the odd object that strikes my fancy and making a personal copy of it but rather because there is just not that much out there that I lust to have a copy of. Instead, I am interested in the notion of using such software to create a machine vision capability that I can leverage to create an open source pick and place machine...


...the point being that if you are to have a proper clanking replicator you have to have the ability to assemble the parts that our 3D printers create. Ordinary pick and place machines typically depend on being put to repetitive assembly tasks using standard parts delivered to specific locations. Where I want to go is for a pick and place machine that can do one-off tasks like, for example, gathering just-made parts from a 3D printer and putting them into a set of bins for later assembly. That sort of flexibility requires machine vision.

This sort of notion takes me off on a serious tangent from where the RepRap community is currently headed. It also informs the direction of development of my 3D printer project, Tommelise. I've not, for instance, been much interested in having the 3D printer capability to use support materials to confect complex objects, the reason being that if you do that you have to remove the support material before you can use the made object. I can't see my pick and place machine having that kind of flexibility just yet.

Developing a pick and place machine will also give me experience in building robot arms. I plan on leveraging this know-how, when I've developed it, into some work I want to do with telepresence 'bots.



If I can get one of these running cheaply that can be run over the web the social and economic impact of self-replicating 3D printers will be as nothing by comparison.

I noticed that SplineScan, while written currently in Python, was prototyped in Visual Basic 6.  The people at the SplineScan project were kind enough to post the VB6 code.  I intend to use that as a starting point for my own Project Cyclops.

I got into the RepRap project not so much because I wanted to develop RepRap machines but because I wanted to make things that only a RepRap machine would let me make.  I am within a few months now of having an operational bootstrap RepRap machine in Tommelise.  I expect that once it is operational Tommelise will be running pretty much 24x7 making things for my pick and place machine and telepresence 'bots.  I plan on improving Tommelise incrementally using the kaizen approach rather than going directly for a second generation machine.

Till then, I will be doing most of my Project Cyclops work within a VB.NET 2005 virtual environment.  That will let me develop my operational software for Cyclops rather, I hope, inexpensively.