Home
Creators
Search Recent Missing Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

New video: CNC Chainsaw Robot (Patreon)

Published:
2021-04-04 09:05:54
Imported:
2026-06

Content

I'm so happy to finally have this one working. I was expecting this project to be on the easier side, but I think it ended up taking more hours than any other project I've done for the channel. Pretty interesting though. If you want to get more details on the project, I'll be doing a livestream in a couple of days where I plan to talk more generally about the project and try to answer questions. It will be streamed here: https://www.youtube.com/watch?v=pwZ28XT7DYY

As always thanks for the support and I hope you enjoy this project!

- Shane

Files

Making a Robot Chainsaw. Carving dogs.

Check out https://Kiwico.com/StuffMadeHere for 50% off your first month of any crate! The robot arm I used: https://hubs.ly/H0K42Hg0 Please consider supporting these projects through my patreon: http://patreon.com/stuffmadehere When I was offered the opportunity to beta test an upcoming robotic arm from Tormach, I knew exactly what I wanted to do with it. CNC chainsaws baybeeeeee. Aside from being fundamentally awesome, it is a very interesting software problem. Since I didn't have to build the arm I was able to focus the project and the video on the software challenges of making something like this work. Thank goodness I didn't have to build the arm because the software almost killed me. I can imagine all kinds of next steps for this creation like machining textures with the chainsaw or cutting topiaries. I hope you like it! Sources: * Low Polygon Labrador Model: https://www.thingiverse.com/thing:446787 * RRT GIF: https://en.wikipedia.org/wiki/Rapidly-exploring_random_tree#/media/File:Rapidly-exploring_Random_Tree_(RRT)_500x373.gif Other great things (your purchase helps support this channel): * I use a glowforge pro for rapid prototyping and it's great. You can $500 off a Glowforge pro here: http://glowforge.com/stuffmadehere

Comments

blizzy

At 15:05, I'm wondering if using a different programming language could speed things up. Is your source available? (and input files, of course)

craig T kaplon

I have run many chainsaw mills and watched individuals carve these sculptures in person. The ability of the chainsaw to perform over the end of the bar is very difficult and you will find these sculptors sharpen and frequently change chains. From a chainsaw perspective, the log itself would almost need an additional axis (inclination) to accommodate large removal so that the tip could be solely for end precision shaping. Or you could use extremely dry and soft wood.. Be cautious if you go this route as a under lubricated bar and chain will get to dangerous temps indoors with a dry soft lumber..

Mike O'Dell

I too was wondering about python, but given that the python libs for matrix math, etc, are pretty tensely tuned already, and this is a Known Hard Problem. I expect the largest opportunities for improvement are algorithmic and hackorythmic. This assumes replacing the chain saw with something with more cutting finesse is not acceptable. If cutting foam is an acceptable endpoint, replacing the chainsaw with something like a giant Dremel MotoTool to cheaply (and poorly) approximate the vertical spindle in your CNC machine. That could make a huge difference - plus, i believe there is toolpathing code available to build on (or borrow from). Finally, how did you come to assemble such an insanely well-equipped shop? Your episodes are extreme metal shop porn. Is there a listing of the kit, ideally with a bit of commentary about likes, dislikes, etc. A guided tour with commentary would make a great episode. We got a deep dive when the new lathe. There is a fine line between deep admiration and agonizing jealousy. Normally I can tell where I am on that scale. In the case of your shop and your chops, not so much. (grin)

Andrew B

Insane video. So impressed that it actually worked in the end - I thought this might have been too ambitious, but it just makes it that much better. Would love to see that lock to LPL next!

Dakota Schneider

Hey Shane. Loved the video, as always. Figured I'd drop a few thoughts here for the inevitable V2: For pathing, did you consider cutting the convex hull directly just by slicing through the planes of the faces? You may need a longer bar for this, or to take two cuts on some faces (one from each side). This would allow a fairly effective bulk removal (for most shapes) and make the subsequent detail work easier. Depending on the depth of remaining cut, I think you could do a lot of the work directly with the apple slicer algorithm. For the remaining deep pockets, you could perform bulk removal using your current algorithm until the remaining depth of cut was within range for apple slicer. With enough bulk removal, you could perform a fairly fine surface pass with the tip of the chainsaw, appropriately modified of course. I think that you could get quite good results this way. Looking forward to the next unpickable lock!

NotMyRealName

I don't know why, exactly, but the googly eyes on the elbow as the synthetic voice said the word dog had me laughing harder than I have in quite a long time. Thanks for that

Kevin Kays

Hi Shane - First, some obligatory praise - your builds are cool, but the stories you tell through your builds are especially great. You do a great job of humbly and humorously communicating your problem solving process, and I really respect how straightforward and easy-to-understand your explanations are. I know it must take quite a bit of time and thought, and I think you've found a great level of compromise between jargon-filled info-dumps and hand waving. As a software engineer, I can't help but notice from your quick screen-cap-scrolls through your source code that there are a few quick tips (and a maybe a few longer tips) that would make your code a lot cleaner and easier to debug - this one first and foremost: https://testing.googleblog.com/2017/06/code-health-reduce-nesting-reduce.html It's kind of my career/hobby to pick apart code and help people solve large problems with software - if you ever want a code review or some more "readability" tips, I'm happy to volunteer my time. Good luck with v2!

Yiding Jia

I know this channel is about doing somewhat ridiculous things, but I can’t help but think that what you’ve done here could be applied to a nichrome wire foam cutter to make foam sculptures. For the rest of us without an industrial robot arm and don’t want to worry about dismemberment.

Casey Handmer

Awesome stuff as usual. Please make a patreon tier where there is github access and PRs. As your shop build out asymptotes to completeness, your challenges will become more software driven and assembling a circle of software sages may help take your tasks to the next level. Re algorithms, humans carving with chainsaws do a bit of the Zeno's Paradox. More or less standing still they cut the profile (a 2D projection of the shape) very roughly. Then they walk around the log to a new angle and cut another profile, this time a bit more carefully. Wash rinse repeat until the not-necessarily-convex hull is complete. At this point the material remaining to be removed is less than an inch in thickness and is abraded away with the tip of the bar. By doing successive passes, curved sections can be cut as the cut width itself can be significantly wider than the bar. For the next step, build a basic sim of the physical actions of the chainsaw and train a neural net to govern its operations. You might also have to upgrade its ability to avoid running into itself.

stuffmadehere

Thanks for sharing your experience - I appreciate the tips! I think next time I'm just going to use a different tool than the chainsaw. I did it for the challenge but I can get way better results with a burr / mill / etc.

stuffmadehere

I think you're generally right on the computation - all the heavy lifting was done by numpy / scipi which are using C / fortran etc on the backend. The one code / language opportunity for major improvement would be to parallelize with gpgpu programming - this algo is very parallelizable. I plan to put a mill or burr in it next. I did the chainsaw not because it was the best, but because it is hard / interesting software that I can't get off the shelf. The shop is a combination of ALL my money from the last 15 years + some deals with brands and my youtube channel. Also I actually did a livestream tour of my shop for patreon! Check it out here: https://youtu.be/5oaEEx695f4

stuffmadehere

It's funny reading your comment because I know I'm guilty :) When I'm writing this code I KNOW that it's pretty bad but I still do it that way. It's always a struggle between doing things clean and knowing that the code is going to majorly change as I work through the problems. I basically treat a lot of the code as throw away prototype code and either do a massive refactor towards the end or when I do a V2 of the project. That said even throw away code can be better (and I have been burned by this several times) so I appreciate the tips!