free ranging buildbots

I've been contemplating freeranging devices with precise orientation and location sensing, that can be set free as a team to build an object of larger size. The team members should preferably form component parts of an ecosystem, as too much specialisation reduces group flexibility. The spacial sensing technology is now commonplace already. Consider graphics tablets, silicon compasses, and GPS.

An approach like this allows for faster build speeds, and easier construction of shapes with undercuts. Textural flexibility is an inherant feature of a system build with components multiscaled for fractal flexibility.

It's very difficult for multiple wandering robots to work together at machinists tolerances. More a multi-million dollar institutional project than a multi-hundred dollar home project.

I wish you luck.

There's been some breakthroughs with Swarm Robotics, but its still pretty difficult. Reminds me of the Ghost in the Shell quote... "Over specialize and you breed in weakness."

What you're proposing reminds me of this site I found years ago
[www2.parc.com]

So thinking of a robot arm of 3 DOF, then as work progresses it needs 5 DOF for a short while then returns to a nominal 3 DOF

While I think that anyone deciding on tackling rearranging robots for building stuff should be left for someone planning on doing a PhD in Engineering and Robotics, its really quite complicated stuff.

Though i'd be more impressed if a swarm poly (1 DOF per module) robot built up a robot module from the 3D printer to build more modules would not be too impossible.

... search for 'foglets' or 'utility fog' - that's the ultimate swarm-technology

But beside the fun ... with many crosslinked modules in a chain or reconfigurable 2D-array you can make some really sophisticated repositioning of tools and workpieces ...

Viktor

Tolerances for large projects are rarely as close as you find for CNC apps. A long-time rule of thumb in house construction, for example, is that if your stud wall positioning is okay as long as they are within 4-6 inches of where they are supposed to be as long as the errors are consistent. Builders know that people can simply not see differences that small in room dimensions. Mind, ceiling heights are another matter. There, you might have an inch to play with.

---

-------------------------------------------------------

Hell, there are no rules here - we're trying to accomplish something.

Opportunity is missed by most people because it is dressed in overalls and looks like work.

Thomas A. Edison

... but it's a big difference, when you're trying to fit a wall unit

Viktor

You might be interested in James McLurkin project in 1994

The Ants: A Community of Microrobots
http://www.ai.mit.edu/projects/ants/

It was a very interesting project with social behaviour like clustering and playing tag.

Don't sound to difficult. I can picture a few ways of doing this.

What's the purpose though? Is the purpose to have small things, which are low cost material wise, to build large things? If so, what do you imagine these things to building? Or is the purpose that you simply like to have a challenge?

Tony

---

Creating the society of the future
[conceivia.com]

Have you looked here?-
[www.hexapodrobot.com]

Greetings all,

I think this is an interesting notion, although getting a team of buildbots geometrically coordinated (well enough to make useful parts instead of scrap) is certainly a challenge. A team of robots is one approach to making objects bigger than any one robot/machine; I find this an interesting idea.

One idea that could make it easier to coordinate a team of robots is to provide them some sensor infrastructure for positioning (like GPS, but at a more local scale.) This could either be a series of active beacons at known locations, or something passive, such as a series of lines on the walls and ceiling at known spacings.

Of course, sensing the position of the 'bots is only part of the challenge. If one is using any subtractive machining technique (like the hexapod router), one must either have robots that can hold their position (against machining forces), or else be limited to cutting really soft materials (like the foam used in the hexapod router video.) So, lockable joints and/or magnetic/suction-cup feet might also be needed.

---

Larry Pfeffer,

My blog about building repstrap Cerberus:
[repstrap-cerberus.blogspot.com]

Subtractive processes don't have to require contact forces to work. Think about plasma cutters, waterjet cutters, oxygen-acetylene torches, etc... I really want to build a cnc lathe that uses a plasma cutter as the tool head because then I can throw out all the vibration and tool head stiffness requirements. Or so I hope.

Another thing to consider, the more energy dissipated at the tool tip the less force required to push it along.

I had a similar idea to this a while ago, what killed it for me was building a robot that precise, even for the big stuff, isn't doable with the tools I have on hand. So, I'll work on RepRap, etc... all to give me the tools to let me build something that works.

The haxapod robot is really impressive, I was looking at this and drooling again a week ago. (I often go back to look some more and drool)

On some of the haxapod builders other pages he has been cnc carving with it.

It has got to be easier to extrude, than mill, less vibration and reactive forces.

aka47

---

Necessity hopefully becomes the absentee parent of successfully invented children.

I have occasionally though about how to design a swarm robot system. One of the things you need (especially for fabricating) is positioning.

This does not have to be an absolute real world positioning, but is local to the job being undertaken.

A long time a go (back in the old DOS days), I used to play a game called PC robots. In this you had to program virtual robots to destroy each other. One of the robots I worked on was capable of using the position of other robots (allies) to co-ordinate an attack on a target. I recently realised the algorithms use for this would be similar to co-coordinating a swarm robot system.

If you can work out the angle and rage between several robots (minimum 4 for 3D work), you can use trigonometry to calculate positions relative to them.

As an example, if you have 3 robots (2D work) and had the range and angle between all 3, then you can construct a triangle between them. Then, using these as base points you can work out the position of any other robot by checking it's position and angle to each of the base point robots (and you can even do this with just knowing the range from the base points)

This will give you a position relative to the base point robots, one that is unique in space. With 4 base point robots, you can work out a position in 3D space (what is needed for fabrication)

The limit to the accuracy of this positioning is down to the accuracy of the sensors used to detect range and angle, the better these are the more accurate you can be.

You can also increase accuracy by using more base points and getting an average position from them (and would also help for failures/interference/batteries going flat/etc).

So for this you would, not move the base point robots for the entire job, but other robots in the swarm would move around and perform the fabrication.

Each robot would also need to be able to communicate between the others (positioning information and task assignments), and this could be handled using any number of wireless technologies (but multi-channelled would be best with each robot having a unique ID - GUID or IP address).

Absolute position and orientation is unimportant (except for maybe the vertical orientation - which way is up) so long as it remains constant through the build.

Hmm

I like the sound of that. Zigbee or wifi would actual do you for the radio comms.

A shared channel would mean that each robot only needed to tx it's position once for all the others to get it.

Ie a broadcast, zigbee and wifi though also allow point to point over the same medium necessary for other things.

Your 3d will be limited to articulation of your robot plus being able to stand on what ever they have built so far.

Building that can be used to climb on needs factoring into the build.

aka47

---

Necessity hopefully becomes the absentee parent of successfully invented children.

Yes the build height would be limited to what the robots could achieve. But if they could be made to co-operate and connect together to increase their height (a bit like molecubes could) then this would solve the height issue but limit them to what they could achieve combined (so still a limit, but not limited to an individual robot).

The other option is to have them construct a scaffold that they could use to climb up to the height and position they needed (the ability to recycle the scaffold would be good for this too).

These two option could be used in conjunction to construct arbitrarily (to the structural limit of the materials and volume of materials you have) large objects.

Constructing the scaffold in real time would be processor intensive, and so this would probably be done at design time for the object so that the scaffold was included in the design file sent to the swarm. But, using each robot as a sub-processor in a multi processor system could allow the swarm to perform the complex processing needed to do real time scaffold design.