Hacker News new | ask | show | jobs
by bsder 1328 days ago
As much as I love seeing this kind of hardware stuff, the problem with "pick and place" is the vision integration, not the motor control. That's where everybody gets stuck.

It's not a "pick and place" until you've got the cameras and alignment working.
1 comments
theacodes 1328 days ago
You're not wrong, but it's also not the focus of the article. There's a lot of complexity in a pick and place system, for example I didn't even touch on the entirely separate board that's controlling the tool head. If I tried to capture all that at once, I'd be writing an entire textbook, not an article.

If you're interested in how the LumenPnP handles vision, you can check out the OpenPNP project which handles all of the computer vision involved.

https://openpnp.org/

link
bsder 1328 days ago
Those are the same non-functional machines I have seen for the past decade in some cases. All of them got to vision and got stuck.

I truly hope your machine gets past that point, because it would be really good to have a genuine, open-source pick and place machine. With the current advances in ML/AI and running on graphics cards, it should be within the realm of a hobbyist nowadays.

I wish you good luck and hope to see your stuff on one of the crowdfunding sites one day.

link
theacodes 1328 days ago
I think you're working on some outdated assumptions. As mentioned in the article, the LumenPNP project is successful. Myself and many others have been successfully assembling boards with it since earlier this year. Lumen isn't my project, it's just the machine that the board I designed is intended to work with.

https://github.com/opulo-inc/lumenpnp

link