The Octo-Bouncer | Hacker News Reader

The Octo-Bouncer

electrondust.com

334 pointsspchampion228 comments

Five ball juggling robot: https://www.youtube.com/watch?v=9asDO_1A27U

Three ball juggling / devil stick robot: https://www.youtube.com/watch?v=pKJEbs64Y2o

Partner juggling: https://www.youtube.com/watch?v=83eGcht7IiI

High speed rock paper scissors champ - https://youtu.be/3nxjjztQKtY

Ishikawa Komuro Labs had a bunch of crazy high speed robotic vision and actuation system ~10 years ago. https://youtu.be/-KxjVlaLBmk?t=65 if you watch the slow motion there's so much backlash in the system it's hard to believe it works at all, but somehow they factor all of that in. Crazy stuff.

That said, for a little side project this is very cool. The simulation idea is dope and the end result is fun to watch.

Yeh.. the scissors paper rock 100% winning rate.. robot is cheating.. every time its waiting till human has made a choice, easy to see in slow mo

Alternative is bunny carrot gun..

That's what the robot is supposed to do - it explains that on the slide at the beginning of the video.

Build one these but using something like micro mirrors [1], esp in microgravity this could be used for mass transfer and sorting of fine powders. The powders could be parked in clumps in an electroacoustic grid potential (ultrasonic phased array).

[1] https://en.wikipedia.org/wiki/Digital_micromirror_device

Sounds similar to some work NASA did with to "walk" lunar / Martian dust simulants off of glass using electrostatics, intended for use in colony habitat windows: https://www.nasa.gov/content/scientists-developing-ways-to-m...

this is a amazing! I'm working on a flight controller and vertical stability system for a high power model rocket and it's so freaking hard. You have to be an expert fabricator, electrical engineer, controls engineer, and software engineer. You also have to have the drive and dedication to get it done. It makes the webdev I get paid to do seem so trivial.

This article is really cool structurally, beyond the content itself. What I mean is that it's formatted somewhat like a programming tutorial (build X with Y in Z minutes) except that it's perfectly readable to people like me with very little hardware background. Most of the specific implementation details are abstracted out into links. I think that this kind of writing can be very effective for technical people trying to reach a broader but still technical audience.

Regarding the content itself, I think it's really cool that the author made a sound-absorbing box for the CNC machine, and I'm surprised that it is safe to totally enclose the machine like that. After spending last summer working at a 3D printing company that makes metal printers, I wonder how a 3D printer would compare to the CNC mill in terms of time to create that many small parts, I imagine it would only take one print-wash-sinter cycle, under 48 hours, about 1/4 of the article's reported machine time. That said the CNC mill used was very small.

The 3D printers you were using probably cost in the $100k-250k range. I think you can get tiny bench-top mills like that for something like $200.

And a CNC mill at a fraction of that $100k price point would probably be much more than 4x faster (due to spindle power and liquid cooling of the parts).

edit: Also since it looks like 2d work you can also use a fiber laser to cut it out which would probably be faster and less messy than a CNC but more expensive of a machine.

Mechanically this seems like something that could also be built out of lego (technic). Instead working with metal it seems like everything needs to be custom machined.

Is there a reason there aren't similar kits of metal components with regularly drilled holes and so on that can just be put together in whatever way one chooses?

It exists! https://en.m.wikipedia.org/wiki/Meccano

https://www.servocity.com/actobotics

There are some kits, like Actobotics. Generic structural parts are inherently heavier and less stiff than custom-designed ones, so they're not usually suitable for high speed projects like this one.

This is amazing! But based on the title I thought it would juggle 8 balls.

"I basically ran my Benbox CNC 1310 (almost) nonstop for multiple weeks."

Wow. That's some serious dedication. What an excellent multi-disciplinary project, congratulations!

This needs to be an exhibit at the San Fran Exploratorium.

https://www.exploratorium.edu/

Why four servos instead of just three? You could still have a square surface with three, just mount to the glass underneath and maybe have a frame on the glass to provide the triangular mount points. That way each servo is truly independent, right now you only have 3 distinct degrees of freedom anyways so you've essentially made an "overdetermined system".

Intuitively I would assume three servos would require more precision, practically, even though theoretically it's no problem.

Well, no, it wouldn't take any extra precision. If anything, the current setup takes more precision as you have to keep all servos in sync or it'll be pushing one servo against the others. Right now you could unbolt one of the 4 servos that are on it and it would still work just fine. You wouldn't need to reprogram anything, just having 3 working servos and one completely disconnected would still keep it in the exact same position.

That is awesome work and an awesome article. If that was a commercially available product for sale I would buy it, because, fun!

Watching what it can do, it feels like this is the perfect machine to run one of those marble mazes.

This is incredibly impressive.

My only domain knowledge for this thing is building a 3d printer from scratch, but the design and micro-manufacture of this thing, combined with the software that allows it to do what it does, is jaw dropping. Great work!

Video of the machine and Windows app: https://youtu.be/lYyAMDYzJQM

Why "Octo"?

There are only 4 motors, so it's probably named after the octagon-shaped opening in the frame.

10/10 great work

id be curious to know more about the planetary reducers. did you just need more precision? or torque? seems like servo motors would solve the second issue.

Reply on news.ycombinator.com