Rotary Hexabot Platforms

[chuckknigh]

OK, guys...this one is new to me. It's a ROTARY hexabot platform.

http://www.hexel.com/rotobot.htm

It's addressed a lot of the problems for home builders, like variable length legs. Its legs are a fixed length, and the base length is adjusted. Brilliant...it's inspirational.

BUT, being the cheapskate that I am, I WANT to build one of these, but on the cheap! Any ideas? It appears that each of the legs has freedom to move, only until it hits its neighbor...there *is* interference between them. Could it be as simple as a set of nested "lazy susans?" In their graphic, the center looks to be nothing more than stacked ball bearings with attached "arms."

Any thoughts as to the ball joints? I wouldn't mind one of these units, but I'd want something small...maybe a socket-wrench u-joint could work, instead?

And, is it just my imagination, or are the strut connections to the ball joint at the top of the platform, co-planar? i.e. just a hinge? That would simplify the construction tremendously.

So, what do y'all think? Worth it to build one of these, or not?

-- Chuck Knight

P.S. I know EMC has support for hexabots...would it control something like this?

[sdwilson]

very ineresting design, i think your right about the nested lazy susan idea, i wonder about the pivot points and the overall rigidity, and about how it is locked into position.
are those set screws by the lower pivots, at the end of the arms.

[chuckknigh]

Well, it's a commercially available design, so overall rigidity can't be too bad. And, since errors are averaged in hexabot design, rather than additive like in X-Y-Z designs, the system should be able to hold to higher tolerances, without a vastly increased tolerance for each individual part. A homebuilder's dream. The downside is that there are no fewer than 6 steppers included in this design, and it's overkill for most folks.

My primary concern is not in building it, but in controlling it. Since posting my message, I've run through several dozen variations of this machine, in my mind. (Doesn't everybody do this?) Making it won't be a major problem...it'll be complicated, but not impossible, even with just a home woodworking shop.

My concern is that, as the struts move apart, they move in an arc...that means non-linear equations. They're not that hard to program, but does any easily available software support them? (And, does anyone have the math skills to figure them out? I've got a background that includes differential and integral calculus, but I'm rusty.)

It also means that resolution will be variable...increasingly coarse resolution as the base length shrinks. I think that's right...but that aspect is common to all hexapod designs, I think.

The genius of this unit is that it takes an already very flexible 5 DOF machine, and integrates a 6th DOF into its fundamental design. By rotating all the struts around the path, simultaneously, you get a rotational axis with minimal increase in complexity, and one which still averages the errors.

So, is anyone interested in helping me figure out the dynamics of this gizmo? ;-)

-- Chuck Knight

[sdwilson]

it almost seems as though 4 axises(<--- is that right) would cover it in a straight line (well 3 and a rotory, bear with me here it will make it more complicted but probably easier to program)
because really its a 3 pivot idea that goes to 6(for the up and down) so if there was a way to move it up and down(maybe a screw) that pushed directy at the upper pivot point (the 3 )
and was nested in the middle of the corrosponding lower pair of legs, i don't know i explained that right but it seems like a good idea, makes a 4 axis a 6 axis (other axis on the rotary)

[sdwilson]

and your right about the arc,.5" will give you a foot at the end of travel, so being able adjust it thru a screw mechanism will give you greater control

[chuckknigh]

I think I see what you're saying, but I'm not sure it will work. This platform not only tilts, but *moves* across its working envelope in terms of height, tilt, pitch, yaw, etc...

Take a close look...the hexapod design is a variable geometry octahedron. All sides are triangles, which makes it stable. Adjusting the geometry of those triangles makes it move. Go to the videos section of the Hexel web site (under gallery, I think) and you'll see how these things can move all oer the place.

Now, if you removed the triangles, and just put in a single vertical strut, then it would no longer be stable. Same principle as geodesic domes. It would also not move, except up and down, and tilt.

Check out my avatar...my picture. Notice the model I'm holding in my hand. I'm a dome nut..."natural geometry and geodesic structures", is one of my hobbies. Inherently stable structures are impossible to build, without triangulation...in the model I'm holding, the panelized approach creates "virtual triangles" since the corners are connected by the panels themselves.

Back to CNC...I think this design will only work if it is built pretty much as-is. We might be able to simplify or modify some of the details, but the overall geometry is going to have to remain the same. The control has got to be done in a layer of software...be it from within EMC, or in some custom software that would have to be written.

I wonder if Hexel would make their control routines available to us? The home workshop is definitely not their target market...

-- Chuck Knight

[jimbo]

that design is ingenious and looks very doable to build but math is another story as you said. regardless, it would still be interesting to build.

with this design, the top half of the machine is trivial with fixed legs and table. I dont quite think a standard hinge will work for the joints but that is just a gut feeling. something like ball joints from suspension or steering i think would be better.

what if you used an automotive flywheel for the base with the six motors walking around it, granted the teeth are coarse but it is flat. also I am wondering if the arms at the base have to form a triangle at the center, if not the arms could be shortened slightly and mount to a ring that rides a single bearing in the center.

[ToyMaker]

This site: http://freeandeasy.sourceforge.net/rotproto.php
may give some hints to the basic math required to get started on the software.

robotic regards,

Tom
= = = = =
Do not pray for tasks equal to your powers; pray for powers equal to your tasks.
- - Phillips Brooks, bishop and orator (1835-1893)

[bb99]

You might be able to get away with three motors. I see two potential ways:

1. Each corner of the table has two legs. If you were to fix one of these legs to the base, then you would only need to move the other leg. This however would involve more complex math for some simple movements. For example, if you wanted to simply tilt one edge of the platform down, you would have to take into account a slight twisting of the platform since only one of the corners supporting legs is moving, the corners position ins 3D space will travel proportionately in the same direction as the moving support leg. This would necessitate a counter rotation of the entire table to counteract.

2. Have the two corner legs attached to the same motor. This would mean that the two legs travel the same distance in opposite directions for any movement. A real upside for the do-it-yourself’er is that any single movement will be restricted to less complex set. That is to say, if you wish to move corner A, then the two supporting struts will travel an equal distance and corner A’s position in space will be easier to calculate.

--bb99

[chuckknigh]

Originally posted by jimbo with this design, the top half of the machine is trivial with fixed legs and table. I dont quite think a standard hinge will work for the joints but that is just a gut feeling. something like ball joints from suspension or steering i think would be better.

I don't know that I'd call it trivial, but it's definitely home buildable. With nice modular parts, like ball joints, it could almost be a "tinkertoy" approach. My thing is, I'm a cheapskate, and I'm not very familiar with automotive parts. But, I seem to remember that the smaller automotive ball joints have a very limited range of motion. That's my concern.

I seem to remember some toy, years ago, that utilized ball joints...I doubt it was very accurate or long lived, but as I remember, they just "snapped" together. Toobers and Zots, or something like that? Oh well...it's been a while since I played with such toys.

what if you used an automotive flywheel for the base with the six motors walking around it, granted the teeth are coarse but it is flat. also I am wondering if the arms at the base have to form a triangle at the center, if not the arms could be shortened slightly and mount to a ring that rides a single bearing in the center.

I wonder what kind of strength the motors would have to have? I have several ~50-60 oz-in steppers just lying around...

I love the idea of the automotive flywheel -- the teeth gave me a good idea. Essentially a rack and pinion system for motion...they're supposed to be quite good, very fast, and easy to adjust for backlash. The teeth are already precision ground on the flywheel, and the price would definitely be right. I *know* they're cheap, on the used market.

As for the center shaft, I wonder if it'd need to have bearings, or if a bronze bushing would be sufficient? We're talking about relatively little and slow motion, after all...lots of starting and stopping, though, so it'd probably be a good idea to use bearings.

Now I'm wondering if the central shaft is even necessary. Assuming a downward thrust, the model I posted above would hold together without the central shaft. I'm guessing that it would help hold it in alignment, but a reasonably "tight" construction around the outside edge would do the same thing. And, a pinion driven around the outside, wouldn't need any connection at the center.

Power distribution might be helped, though -- it might be possible to make a 360 degree electrical conection through the bearing? Otherwise, the rotation could cause all sorts of interesting twists and turns in the wiring.

I think this is a GOOD START!

-- Chuck Knight

[chuckknigh]

Originally posted by bb99 You might be able to get away with three motors. I see two potential ways: 1. Each corner of the table has two legs. If you were to fix one of these legs to the base, then you would only need to move the other leg. This however would involve more complex math for some simple movements. For example, if you wanted to simply tilt one edge of the platform down, you would have to take into account a slight twisting of the platform since only one of the corners supporting legs is moving, the corners position ins 3D space will travel proportionately in the same direction as the moving support leg. This would necessitate a counter rotation of the entire table to counteract.

I don't know that this is true. You can get reverse threaded screws -- a turnbuckle is a perfect example of this. One rotation, and even motion out both sides. This would eliminate the twisting.

Down side is that it would also remove the rotational aspect of the construct -- we'd be back down to 5 DOF, but that's more than enough for most purposes.

Thinking about it, though, I'm not sure that would work. It'd raise and lower the corner of the platform, but I think it would eliminate any side to side movements that were available. I think that two legs *have* to be independently adjustable.

Check out the motion in these videos...
http://www.hexel.com/video.htm

-- Chuck Knight

[BeerFizz]

You could always build the base as a square and not a circle. The legs would then move along the sides of the square. Two legs per side, with the max height when they are in the middle and min when they are at the corners. Eight legs total (an OctoPod ). Moving in a straight line might also simplify the math.