maybe were (read your) doing the cartesian robot wrong?

[oomlout.com]
total cost, about 150$ in pieces including the laser cut case. arms are also much more versatile. strap an extruder on it and you probably have a much cheaper more capable unit. why exactly are we sticking with the box design?

Edited 1 time(s). Last edit at 03/29/2009 06:55PM by ashadocat.

Do you think it will maintain 0.1mm accuracy at the end of the arm?

They haven't worked out how to write the software to do the inverse kinematics yet.

---

[www.hydraraptor.blogspot.com]

I see. I thought there was probably a glaringly obvious reason like that.

I think that the precision is more of a limitation than the lack of control software.

But I would not entirely discount the use of robot arms in future RepRaps; after the parts are extruded on the cartesian platform, it would be useful to have something that could pick them up and put them together.

Future project: study small robot arms that are useful for practical applications; create a design that can be built using reprapped structural parts.

... look here: [forums.reprap.org] (and especially the image in the first post on the second side)

Or the parallel scara robot concept in this paper: [www.harmonicdrive.de]

Some designs are very interesting in respect of simpler assembly and/or much lower count of parts ...

Viktor

The tripod robot looks like it might be workable, but more limited in the size of item it can build. Scara Robot looks like it could be precise enough, and usable in a setup with multiple switchable tools.

Parallel robots (the "tripod" one) have impressive motion capabilities, but also two major drawbacks. First, controlling them is an extremely complicated process, all the joints have to be operated in unison, and their motion does not map to the effector's position linearly. It's a tough piece of higher mathematics that requires something a lot more powerful than an ATmega (primarily due to its lack of an FPU, if not anything else) to process numerically, and that's only after you make a proper kinematic model, transform it and figure out how to deal with the damn thing giving you several different solutions for each move. I'm not saying that there's no one involved in RepRap with knowledge to do that, but I'm pretty sure it's just too much effort for the gain. Second, a much worse problem, is the high number of singularity points. A parallel robot is great at positioning the effector in space, but does very poorly when you ask it to do that using some arbitrary path between endpoints, because there's a risk of the robot making it into a singularity, locking up and losing stiffness. In operations such as pick and place, that's not a problem, the software will just route around those points, but you can't do that with additive manufacturing, as it's actually all about the path, not the endpoints. In fact, it might turn out that some objects can't be printed on such a robot, because they invariably contain some paths that cross a singularity point.

SCARA, on the other hand, could work, especially for printing odd-shaped objects that won't suffer from degraded quality of straight lines. It does, too, have some problems with moving along arbitraty paths, but simply restricting the workspace to a smaller, properly located area pretty much eliminates them for all practical purposes, while still providing a sufficient working volume. And, while it's still more complicated to control than a gantry robot, it's orders of magnitude simpler than the parallel robot.

Besides, the greatest strenght of the current design is simplicity. Almost everyone could understand it, figuring out the motion euqations for low speeds requires little more than high school math and some reasonable effort, and the resulting robot is accurate, durable and free of serious caveats such as singularity points.

Edited 1 time(s). Last edit at 04/03/2009 12:33PM by Enleth.