Optical registration for Handibot?

I did some experimenting with computer vision based registration for the Handibot and thought I'd share. I have more images than I think I can share on this forum, so it's probably best to read the post on my blog. Happy to chat about it here or there.

VERY interesting!! nice work! There are a couple of updates to FabMo being worked on that may make that process even easier for you.

One thing that could help a lot is the ability to skew, shift or rotate a cut in real time. We've been thinking about that more as a way to correct for uneven surfaces when depth really matters (circuit board engraving)--but it could work just as well in the XY plane to allow your tiles to be cut without such exacting placement being required. Gonna pester the FabMo guys about this even harder now!

Also--Ted and I are working on an accessory for cardboard cutting--kind of a tangent reciprocating blade that uses the Z axis to create the sawing motion. We built one here that we just use for cutting hand-holds into our shipping boxes...but the problem we wanted to solve was how hard it was to move the handibot around on a sheet. You want the handibot to be heavy so that it doesn't shift during cutting...but you want it to be easy to move. What Ted's come up with is a new handle for the handibot that has a trigger which can lower a set of rollers--and lift the handibot up by about a 1/4" you can then roll it around to where you want it then release the trigger to drop the handibot down onto the material and lift the rollers.

Those rollers will be fantastic to have. I have been trying to figure out something like that for myself. I've had a lot of health issues in the last year and I'm gradually building my strength back up, but I've also been having worse carpal tunnel symptoms. So I've not been able to use the handibot nearly as much as I'd planned, simply because I have to be at ideal conditions to be able to shift it around as many times as I might need in a round of practice cuts and then final cuts. Any idea of price range on the rollers? Or if plans/parts lists will be made available?

CL,

Sorry to hear that you've been having a rough time...

I think that we could get those rollers out by the end of the year--they were designed as part of a larger setup (cardboard cutting tool with a pressure foot and a new app....etc...) but no reason they couldn't be sold separately.

Very interesting read, Hackaday worthy.

Thanks guys. Brian, wow, both the Fabmo changes and roller base could really make this viable. Is there anywhere I can learn more (or help?)?

I'll call out that while I've done most of my iteration using a Linux laptop with a USB camera, the ideal end state for Handicam is that it runs on a phone or tablet. The only required accessory would be a mount. The image matching operations are CPU-intense, but on a multi-core device, they should be fine -- will just drain batteries quickly.

This may or may not be of help...but there is a hidden app in fabmo that should eventually allow a user to connect a camera to their tool. The code is here: [github.com] it's only visible in fabmo when you look at apps from the configurations menu.

I'll find out where we are on software position correction and update here.

As mentioned above--the rollers are all but done...we'd been holding them back in order to release them with other accessories where they would feel most relevant...

Kyle, very cool! If we could use auto path generation/regeneration to solve skew (pitch and yaw) as well as slight positioning errors my CNC world would be much happier.

Brian: Are the rollers (wheels?) meant for both standard bases and LST? Do they swivel to allow movement in all directions?

As an aside...

I like Kyle have a manual positioning jig but have often thought of building a sliding tray assembly. I do a lot of 3D carving, and while I try to stay within the 6x8 zone, it can be very limiting. Many will suggest I should have something like a Shopbot Desktop machine but since I do this as a hobby and my workspace is limited, a tool with a more extensive work area is currently not viable, besides carving is not the only thing I do with the Handibot.

The idea I kicked around for a while was to build a gantry system (I am not an engineer!). 1-1/2" t-slotted rails mounted along the edge of my cutting table with 1-1/2" t-slotted cross-members. The cross-members would ride along the edge rails using sleeve-bearing carriages, and then the Handibot would be mounted on a tray that moved along the cross-members again utilizing sleeve-bearing carriages. Positioning could be done using drop pins every 6 and 8 inches (X and Y) minus any overlap desired. You would need brakes on the carriages as well to stop vibration issues. I would need to be able to accurately drill the drop pin holes in the rails and then align with the opposite side. That is kind of where the whole thing falls apart.

So my manual placement system remains in use, and I tend to use a 1/4" overlap to try to lessen tiling artifacts. The 6x8 work area limitation is what drives my interest in the LST; it at least resolves the issue in one direction.

A couple of quick observations / ideas:

1) Under Fabmo > Configuration > General, there's a configurable Offset and Rotation that could surely be used for small values. In practice, a user could output grid of 5x7" tiles instead of 6x8", then if they can maneuver to within +/-1/2" of the target, Handicam can communicate the correct offsets to Fabmo.

2) Looking at the ShopBot commands generated by VCarve for one of my projects, I'm surprised to see that every instruction is a J3 (Jump 3d) or M3 (Move 3d). So even though my model has arc-like curves, the output is a expressed as a bunch of short point-to-point moves instead of arcs or other commands. Transposing or rotating points is easy. It doesn't look like it would be necessary to support transforms for a large command set.

Interesting.