Theory - Beyond RepRap

What happens to manufacturing scalability when you take the extruder and put it on a swing-arm attached to a RC or computer-controlled car?

Well,
- you can have multiple machines working on the same product.
- this can include several specialist machines
- the product scale isn't bound to the dimensions of a box
- it becomes possible to manufacture an entire machine's set of parts

My concept would require:

- a motorized car (stripped down RC car?) with accurate DC motors controlled by an external source and transmitted by wire (computer, person, circuit board..)

- a feed system to move material to the car. (dependent on material) My preference was something like a fishing line with material being moved in solid form.

- extruders/nailguns/welders/cutters/clampers/laser pointers and other such devices can be assigned to cars arbitrarily. A standardized mount point would be sensible.

My line of thinking is to be able to take an truck out to a patch of land with some material in the back, drop a few of these robots, and have them start assembling a house, boat, plane, or more of themselves ad nauseum.

Definitely up for more ideas and/or criticism.

that'd be awesome;
but I'm afraid this still is science fiction.

I cannot imagine how to synchronize multiple rc cars or thelike to build ONE single piece.
Not without having any precise reference;
and if it's not a solid frame... I guess you'll spend more than just a few to lay it out. The base must be perfectly flat or at least scanned to know where the bumps are, the outer frame should be not only perfectly straight but perfectly angled too and so on.

Also all cars must contain a big power supply batterie to fullfill the task,
or they will tie each other up with the powercord.

But yes.. that'd be very nice

For now: How about one of those giant cnc cartesians they use to mill car bodies out of foam?
That'd be at least possible (although it might take more than a year to print a car with just few extruders )

'sid

Cars? What about this one used to mill masters for boat hulls and interiors

Edited 1 time(s). Last edit at 08/25/2009 12:16PM by unfold.

This guy did something similar - milling foam with a cutting tool equipped hexapod robot:

[www.hexapodrobot.com]

Once the robot starts moving over large distances, it'll be tricky to keep the precision, but it could be done with some feedback.

Wade

Wade, isn't that a beauty..? Amazing indeed.
but it can't draw or mill anything that's bigger than his "armlength" unfortunately.
IIRC Matt was here quite a while ago maybe he has some "off his site" notes on this topic

unfold, true they are bigger, but since I do drive a car more often than a boat, that's what came to my mind first

'sid

And now comes the guy that shows an airplane getting milled

Cool that Matt was hanging around here, and the tests are really nice. Indeed it does not overcome the major problem and that's the positioning when they walk around.

As to giant cartesian bots: there's a company that works on a z-corp style printer for large scale stone objects: [www.d-shape.com]

... AFAIK some development in 'cooperative robots' was done with measuring the exact position and orientation of the specific bots by triangulation with coloured lights in the corners of the working area or simpler with ultrasonic beacons placed around.

For robot-soccer the bots are colour-tagged on top and the position/orientation is measured and calculated with cameras above the arena.

Another method is a 'coded' surface and the bots strooling around with two cameras scanning the actual position-code ...

Viktor

Hmm... so it's generally perceived to be a sensory problem? There's probably some things that could be done to resolve it using systems other than a camera and physical markers. Lidar, ultrasound mapping, or other such isn't so bulky and unwieldy that it's impractical, perhaps too expensive for us hobbyists?

Is it possible to use a ruler, or make one on the fly? Would it be possible to leave laser markers, or work as a group where one or more act as a reference point?

My iphone has GPS and compass. I know it's definitely not accurate enough, but a sufficiently accurate localized system which gave you direction and position would make it possible, yeah?

I think that actually the key would be to have the machine like the hexapod built with a small infa-red camera built into it. Then you build 3 or 4 infa-red LED's. In front of each you place a piece of card with a 1mm hole built into it(to increase the fine tuning vs just an LED which is a much bigger target). To make it easier, connect each LED to a different out port on an arduino. Code each one to flash at a specific speed. (some flash 20 times a second, some 15 etc etc).

Use code like this [www.youtube.com] and the machine can now work out the exact position of each of the 4 dots(each dot blinks a certain amount and the computer can thus tell them apart). given the dots are an exact distance apart (that we have calibratoed once) with some very quick and simply geometry (the kind of math a computer can work out very very fast without using too many clocks) the hexapod could work out it's exact known position.

Thus we could drop the machine in a room like this and it could start working, it primarily would keep track of it's position using internal feedback with things like knowing how many steps to move etc etc. But say once every minute(to cut down on it's need to process all this work for positional geometry) it could map it's assumed position in it's internal view of the world (ie what XYZ co-ordinates it thinks that it's at based on it's past minute of moving) vs the position that the LED array gives it. Maybe this would also take care of the calibration effect with a simple feedback loop which takes care of the stepper etc calibration.

Edited 1 time(s). Last edit at 09/12/2009 01:05PM by letsburn00.

Some kind of laser pointer interferometer might work?
[amasci.com]

You don't need absolute positioning for a system like this. What you need is relative positioning.

It is similar to what VDX is talking about, but the robots themselves co-ordinate to create a reference frame for the job

What you could do is place 4 robots (4 minimum for 3D printing) and have them transmit a signal (like the flashing LEDs that letsburn00 was talking about) that other robots can use for triangulation. These 4 robots can't move once the become the "signal" robots, so these might be dedicated robots, or could be assigned from some method within the swarm.

You will need the distance and relative angle between each of the "signal" robots. This information is transmitted to the worker robots and they can calculate their position relative to these "signal" robots.

Once the reference frame is worked out it becomes trivial (for a computer) to calculate its position from them (you only need the distance to each of the signal robots to work out where it is, but you need distance and angle for the signal robots to work out the co-ordinate system).

I also agree that some method of scanning the work area is needed because lumps and bumps or other obstacles in the area would disrupt the build. Perhaps, once the reference (signal) points were set, other robots could move around the area and map out that portion of it.

Each robot would be assigned an area to map; calculating their position and height at all times to map the topography of the area. They could also register collisions with obstacles and mark out places where an area is blocked. They would share this information with all the members of the swarm and they would collectively build up a map of the work area.

Once the map is built up (and they could keep testing and updating it as the work progresses) each robot can know where it is, the terrain in the work area and where each other robot is and what they are doing, and can thus make decisions about how to construct the job.

link:
[dev.forums.reprap.org]

---

My CNC/extruder creation: [grael-minifactory.blogspot.com]