MIT's Infinite Size CNC Router | Hacker News Reader

MIT's Infinite Size CNC Router

alecrivers.com

144 pointsMCompeau38 comments

This would be an amazing boon to hobbyist woodworkers (and i hope that's what they mean by "consumer version"). A large amount of time routing is spent making jigs or patterns templates or transferring measurements or ... Even simple things like dado grooves require clamps and a straightedge.

Given the right price, almost every woodworker on the planet would likely buy this.

They mention the precision is 0.009" average error, which is good enough for woodworking, but not say, metalworking, or even any kind of precision plastic parts.

They also mention "digital automation" in general, which this wouldn't replace. Cabinet shops/etc use CNC machines not just because they are accurate, repeatable, etc, but because they do not require humans to baby them.

This is very cool and has a lot of potential, however a few comments. I have done a lot of hand routing and also have a 3-Axis CNC Router. Both hand routing and CNC each have their own Setup Costs. (1) I can see this device being very useful for artists and designers that are doing a lot of one-off shapes and designs (perhaps for Set Design, pattern makers, templates, etc). The setup costs to prepare CNC files for these types of Jobs account for the majority of the costs. (2) However I would guess that the majority of field (vs. shop) routing is done to "shape" the edge of wood (i.e. with a ogee or roundover, etc). Short of having an on site upside down router table, this is typically done (a) using a plunge bit with a top mounted guide for the Router guard to follow or (b) using a router bit with a small Guide Wheel that follows a top or bottom mounted template/guide edge. The router bits with Guide Wheels can not easily be used for plunge routing and are a pain to use in multi-pass scenarios. I don't know if they can get the precision, however I would think that if this team could offset their sensors to follow the middle of a piece of trim allowing someone to shape the edge of a piece of wood "FreeHand" (after applying some guide tape) and get away from Guide wheels, Clamps, Guides, etc. That would be a MAJOR use case. (3) After watching the Video, I would urge users to wear dusk masks when routing.

While I think this idea is awesome, it solves a common problem in a unique and creative way, I sadly cannot watch the video: "Unfortunately, this EMI-music-content is not available in Germany because GEMA has not granted the respective music publishing rights."

Oh my, one more video I cannot watch just because I am german and one more day of my life ruined by them :(

The video demonstrates the use of the router with no additional explanation.

Observations from the video:

1) On the screen is a picture of the object, with a blue path representing the actual path of the router and a circle representing the fault tolerance of the system. Apparently, keep the path inside of the circle and you win.

2) The computer can move the router independently of its mount, within the tolerance represented by the onscreen circle.

3) It looks really easy to operate.

Especially since music is not essential to this clip. There is no "music" at all, just easy listening in the background. ;)

Hint: Use a (free) youtube proxy ...

One word: ProxTube

I have no idea how kosher this is on Hacker News, but I have mirrored the 480p version of this video for your viewing pleasure:

http://nacr.us/help/mitRouter.mp4

By far the best solution I have found to this problem (I am german too) - preface the video link with 'ss'

Like this: http://www.ssyoutube.com/watch?v=-UmL7xZZSUk (remove the 'feature=player_embedded&' crap after copying the link to your url bar)

The great thing about this is: no extra software needed (though there is an extension and an bookmarklet) and it gives you links to all video resolutions. Proxtube didn't work for me when I tried it and most of the time I like to download the videos with a download manager, which speeds things up on my slow connection (though the Flash Player seems to have gotten faster on my netbook, both download and playback-wise.

Programming it to make straight cuts without setting up a fence alone would make this a buy for me, shapes be damned.

Prediction: The router will be cheap, the special plywood with the calibration stickers will be very expensive.

Isn't there a positioning system that it can use to track the router with accuracy? I'm not aware of any, but I'm sure it must exist.

Wouldn't it be more likely for the reverse to hold true? The razor-and-blades business model doesn't exactly apply here given the low level of technical sophistication required to print the calibration stickers. Constructing inkjet cartridges and machining modern razor blades are far more complex feats of engineering.

As the paper describes, the stickers can be applied "in any pattern so long as some tape is visible from every position that the device will move to". Thus, there is no need to align the stickers in a special pattern beforehand.

Anyone should be able to print similar calibration stickers on their own dime.

I imagine eventually they could do away completely with calibration stickers and just use regular surface tracking with any old plywood.

I wonder how a optical mouse sensor would react to the vibrations of the router.

But then you might want to use it with lets say glossy plain-coloured melamine sheets - most surface trackers have a lot of problems with this, so the calibration stickers are a cheap and reliable solution.

The only issue I can see is kick-back from the router bit, especially if you use larger diameter bits - this can throw the router significantly, and unless the control system compensates quickly yet gently, you may well lose tracking, leading to undefined results.

With optical mouse type tracking wouldn't you lose your position relative to other cuts if you lifted the router off the workpiece?

I imagine you can integrate information from a QR-code for global positioning and an optical mouse sensor for fine tuning

A neat application, indeed - but there is definitely something to find flaw with: it is purported to provide the ability to route arbitrarily large pieces at a very low cost. However, a human operator is not like a gantry-style CNC router, they cannot simply "fly" over the work area. Typically, must operate the router in a "push and pull" fashion to maintain control. A human would become weary very quickly if you put them in a harness over a 50-ft piece of material and forced them to push and pull with their arms hanging below them. Finding motors with enough torque, but not so much to put the human in danger (you wouldn't want a software flaw to push the router in a dangerous direction without the operator being able to override it with their own physical strength) will limit what bits and materials the process can be used on. Granted, I'm fairly certain few are expecting to CNC route metallic materials by hand. =)

Overall a cool new application - I find the display on the screen one of the best features, and, in fact, I'd kill to have something like that -without- the motor correction. One could, with some work, replace the motors with encoders attached to spinning shafts, which were limited in which direction they could spin, and get rid of the tape - simply showing the user that they are following the pattern or not.

This is awesome! But in this case, instead of introducing the human element, wouldn't it be (relatively) easy to make the entire apparatus self-driving and programmable? As in, mount this thing on a Roomba-like robot and give it a coarse driving path?

Getting a grip on the surface good enough to give you the control you need in the presence of sawdust is a lot harder than it looks at first glance. You might be able to make it work by putting a big magnet on the underside of the surface to attract the router, though.

It would have to an extremely powerful magnet though, since it can't be right below the router.

It could if had its own set of wheels+servos. But it would still have to be powerful enough to attract the router through the surface.

I think that the magnet idea could work if you added 3 electro-magnets to the router chassis and made another chassis (to ride around underneath) with 3 corresponding ferro magnetic pads.

You could just have removable weights on it. Suction feet is the other obvious one.

It would be neat if it could follow a line drawn on the board, eliminating need for a digital version of the design and the calibration stickers.

I don't want to be pesimist, but I give it 6 months to a open source alternative to be released.

Why do you need to put barcode markers on the surface? What's wrong with optical mice?

My guess is for a sort of absolute positioning as opposed to the relative positioning of an optical mouse. A mouse produces deltas relative to your previous position, while the bar codes are always in the same place. The problem with relative positioning is that even a small inaccuracy will accumulate over time, resulting in edges being off in the end.

because optical mice aren't very accurate on a scale larger than one motion of a few inches

Genius - even though the router is made by Dewalt, I hope it's patent protected and licensed by MIT so we can start seeing this in Home Depot / Lowes without one company owning the rights.

Of course, my question is answered by this: "A consumer version of the position-correcting router is being developed, to be released by a forthcoming startup." - No doubt the same people in this video, technology licensed from MIT. Company to be acquired by Dewalt within 24 mos. after commercialization.

Not really within the remit of Dewalt, whose branding is "big manly construction equipment", and this is more of a fine-woodworking/hobbyist product. (The pros will still use their big CNC table routers, which don't require human intervention, and can cut much much faster) Maybe Dremel?

Dewalt is basically the prosumer brand of Black & Decker (which owns Dewalt)

what about projecting the sticker patterns onto the wood rather than real stickers?

The wood will move slightly when it's cut, so that won't work.

I am quite sure this can be countered - the wood could be clamped and the router could be tracked and the picture adjusted.

Further, the amount of movement would not make your routes completely useless/wildly inaccurate.

And I am quite sure it can not. Just walking near the table will move it more than the stated 0.009" accuracy.

And supposing you could counter it - your method is wildly more complicated than a sticker. Why not just use a sticker?

What you just described is getting halfway to a CNC machine.

Another benefit of this system would be portability. You can set it up and use it almost anywhere, like at a job site.

Having a system with clamps, overhead tracking, etc. is going in the wrong direction. It's not like the stickers are some huge inconvenience or expensive consumable.

Reply on news.ycombinator.com