Home Creators Posts Import Register Favorites Logout

Content

Unlike my previous shop which was packed to the brim with project scars and the skeletons of projects past, the new shop feels cold, sterile, and just plain boring. The new shop has some huge blank walls and I decided they were a canvas just begging to be painted. Ideally by a robot. How cool would it be to have the windows XP background in full 50 foot glory?

I've been working on the design and fabrication of a robot that could do this for the past several weeks. I'm nearing the end and as I type this the robot is diligently r̶u̶i̶n̶i̶n̶g̶  painting the wall. 

When I started to really get into the nuts and bolts of this project it turned out to be more of a challenge than I anticipated. Let me walk you through a few of the bigger issues I've been working on.

1. Painting arbitrary colors

This was the crux of the project and has dictated the design of pretty much everything. Ideally I would have a robot that could mix up any arbitrary paint color then apply it to the wall. Unfortunately this is completely infeasible. Imagine if I need to paint a single square inch with a specific green color. Due to practical accuracy considerations, the robot would have to mix up much more green than I would need, apply it, then somehow dispose of that green paint and clean itself for the next color. Extremely wasteful and difficult.

Fortunately for me this problem is faced by practically all machine generated color printing. There's a really good solution that only requires four colors to approximately generate any color. The process is called halftoning and all I have to do is apply hundreds of thousands of variably sized dots to the wall. This certainly feels like exchanging a hard problem for a hard problem. But at least the new hard problem feels solvable...

2. Reliable paint delivery for days

It's one thing to say "yeah I'll apply a million dots to the wall," it's quite another to actually do it. If like me you do it one dot at a time it will take upwards of a week to do. Making a (prototype) robot that can run reliably for this amount of time is a challenge. The software can crap out in endless ways and when you put that many cycles on a mechanical system it tends to fall apart. In the process of making this I've had a fair share of those issues, but the real doozy has been the paint delivery system. 

Somehow you need a system that can make controllably sized dots day in and day out. This is a problem that I considered the best approach for at length. Bingo dabbers, brushes, sprayers, self inking stamps, giant ballpoint pens, and many other applicators. I couldn't find a system that felt great on paper, but all things considered I felt that a paint spray gun had the best odds of holding up and being controllable. So this is what I designed and built the system around. 

So far the sprayer has worked out pretty well, though I'd estimate that 85% of integration hell has centered around paint application related issues...

3. Moving the robot anywhere on the wall. AKA Gigantic 50 foot wide gantry

The paint sprayer can make a dot, but to paint the whole wall I have to be able to move it anywhere on the wall (obviously). I also have to be able to precisely control how far the sprayer is from the wall since this has a major impact on the size and density of what it paints. Also, since I have to paint so many dots and I'd like it to finish this century, I need a system that can move around super fast. I considered an arrangement like a Maslow style CNC router  but practical considerations with speed and running over the paint led me away from it.

I ultimately settled on a very strange but neat arrangement that I've always wanted to build. I have a very big slow gantry that moves around a small fast gantry. 

This gives me the best of both worlds. I can move the robot anywhere on a very large area with the super heavy, slow, cheap gantry and then use the small gantry that is light and fast to apply a ton of spots quickly. There's some neat details that allow me to make this accurate, but I'll save those for the video. 

You may be wondering why I wouldn't put a small gantry on a maslow style cable system. This certainly would have been easier, but there were some fundamental issues related to balance and reachable wall area. 

In any case I'm in the late stages of tuning and laying down a bunch of paint on the wall. During this process I've printed a series of increasingly less bad portraits of wifemadehere....

 I should be making the video with all the juicy details in the coming week and a half or so. I think the results are going to be pretty cool and will hopefully be a big step towards making the shop feel awesome. 

Let me know if you have any questions and as always thank you so much for supporting these projects!

- Shane 

Files

Comments

David H

“Giant mural robot” is something I’ve thought should exist for a long time, so I’m glad you’re making it!

Mike O'Dell

A video went by yesterday of machine that does precisely this. I believe it leverages dot matrix technology from the big printers that do large format printing like billboards. it does vertical strips that look to be a few inches wide and the machine rolls horizontally on rails that fo down on the floor. They look like the ones used by a camera dolly to create tracking shots. the thing makes a single pass down the wall and prints beautifully. While the 4-color model is theoretically sufficient, in practice, there are at least two saturations of each color used and sometimes even 4 shades of gray (white to black).

Dumbtex

Do you have a GitHub I’m rlly curious

stuffmadehere

Very interesting. I had considered using some kind of industrial print head but I avoided it because: a) I wanted the challenge of dealing with the printing since that is the hard part b) totally unknown lead times / technical risk / etc If I do a V2 I would definitely be going for some kind of highly parallel, highly precise super fast droplet generator.

Benjamin Clerc

Can’t wait to see the video ! Also, « When I started to really get into the nuts and bolts of this project it turned out to be more of a challenge than I anticipated. » has to be your motto 😂

Rick W

Possible mod: a line drawing bot. A steady stream of paint and a steady velocity should make a massive Pac-Man labyrinth doable if you put down a base coat that complements the line color.

Finn Merlett

I love the idea of this project but two things jumped out at me as possible improvements. 1) Instead of having the spray head move in and out, just have it at a fixed distance with different length timed sprays. This should make the colours more continuous as it would reduce white space between the dots, whilst still allowing you to vary the colour intensity. 2) Paint a small section for colour balance calibration - do some kind of test pattern and print some kind of test pattern on good quality large size paper and pin it to the wall. Then you could take a picture and calculate the colour difference to that which is expected. A couple of repeats later and you could have vast improvements to the colour accuracy.