Paralle Kinematics

[chuckknigh]

OK...big words, simple meaning. X and Y aren't the only way to define points.

There is a very cool painting robot, called Hektor, which uses 2 steppers, some cable, and a spray can hanging on a wall, to produce graffiti art.

http://www.hektor.ch

I think this type of approach shows a lot of creativity and even potential. Imagine a raster version, for printing low-resolution wall murals! All sorts of possibilities, with this one.

Other than custom software, though, does anyone know how to control this sort of thing?

I'd have preferred to put this over in the woodworking section, with all the activity...but I didn't think the moderator would allow that.

-- Chuck Knight

[High Seas]

Chuck - You've re-piqued my intrest in Hector. A while back I thought he was wire driven, wire on spool - now I see it is toothed belts! Guess the software does 2d trig to find its position, if you know (x,y) and (x', y') the starting points, then count the length of each of the belts by counting teeth through the gears -- you'd come to the point the spray paint is at? Just then say "pfft" and its done!
It might be one way? (well it would be one of 2 possible locations - but Hector don't fly!)
Jim

[posix]

I'm just thinking about how it does it (the xy relationship that is) and have noticed something that made me smile. it's your normal xy table rotated 45 deg clockwise (and hung on a wall, of course) with the added twist that one axis would have to compensate for the other. hexapod with only 2 "legs" perhaps?
ok I don't know about the actuator for the spray can but I think a much better result can be had with a salvaged inkjet printer head which I'm sure you allready have somewhere hidden in your salavaged treasures. you'd just have to figure out how to control it. since I believe you have lots of little steppers why not give it a shot?

[aplatt]

Nah, this is easier than you guys are making it out. No trig necessary. The length of the rope on each side is the distance from the can to the pulley. If you consider the left pulley to be at (0,1) and the right pulley to be at (1,1), a position (x,y) has a left rope length of sqrt((x^2)+(1-y)^2), and a right rope length of sqrt((1-x)^2+(1-y)^2). (x and y being between 0 and 1 in this example.) I'm particularly impressed with the software for this, myself. Optimizing those toolpaths seems no trivial task.

[ViperTX]

Pathogrean Theorem....

Welcome aplatt!

I just glanced at Hektor......looks like he is using the vector produced by the graphics program....I wonder if he used Bresenham's approach....hmmmm....

I suspect this can't be too fast...as it relies on the weight of the can and it's mount to dampen the motion....

Pretty amazing.....

[aplatt]

Pythagoras is indeed the fellow. And to get rate of change of those two lengths you can just take the first derivative with respect to x and y of each length.

I took a look at the "scriptographer" system they're using for software. It still looks quite complicated to me. If you know any sources describing in general how to go from vector art to toolpaths I'd enjoy 'em. I'm not familiar with the "Bresenham" of whom you speak.

The site has flash movies of the can in action (often time-lapse, but you get a good sense of the speed of things). It isn't fast, but I'd say it was faster than I expected. They stick a bunch of extra little loopy bits into the toolpath to minimize swinging and it seems to work remarkably well. (The can never stops abruptly or changes directions sharply without describing almost a full circle.)

[lerman]

Chuck,

EMC can handle non-orthogonal and other special purpose kinematics. My understanding is that people are doing hexapods (Stewart platforms) with it.

Ken

[cwiliam]

http://www.shopbottools.com/bill's_corner.htm#Hacking%20ShopBot
checkout this one same principle as Hecktor .