HN Explorer

Ask HN: Desktop Robotic Arm?

29 points by lokl ↗ HN
I'm in the market for a desktop/tabletop 6DoF robotic arm to experiment with, any recommendations? Not certain of uses, but: I might develop custom grippers, repeatability of 0.25 mm or less desired, option to extend workspace with linear rail would be a plus. Thanks!

21 comments

[ 97.1 ms ] story [ 1859 ms ] thread
helph67
There's plenty of Pi projects, here's one article https://robotbuildingkit.com/the-6-best-raspberry-pi-robot-a...
lokl
Thank you, DIY options are attractive because of price, and it's impressive there are even 6DoF projects, but many seem to lack the repeatability I am after (or their repeatability isn't clearly stated). If you have personal experience with one, I would be interested to hear about it.
Iv
Depending on how much you are comfortable tinkering with control electronics and encoders, know that getting a good repeatability can just be a matter of tinkering with the control eletcronics:

https://www.youtube.com/watch?v=SioCwvR_PYY

How does this channel does not have more views is a real mystery to me.

lokl
Likewise, your comment should have more upvotes. This is a very interesting approach.
lukeinator42
A cheap option is one of these sainsmart 6 axis robot arms: https://www.amazon.ca/gp/product/B00UMOSQCI.

I haven't documented anything yet, but I got the arm working with ROS/MoveIt! for kinematics. The code is here: https://github.com/lukeinator42/sainsmart_arm_ros. It uses cheap servos though so I don't think the accuracy is near 0.25mm.

Here's a video that shows the arm in action: https://www.youtube.com/watch?v=Yi_QydaN-34

I'm currently also building a Roboteurs RBX1 Remix arm that uses stepper motors, but it's not complete yet.

lokl
Thank you and good luck with the RBX1 Remix, looks interesting.
omneity
My personal experience with two different robotic arms (< $500) and some additional research is the lack of high level software and libraries, which is quite crippling as it’s very complex to control.

If you’re not used to control theory and inverse dynamics it’s quite tricky to control it (especially at 0.25mm) and you’ll never get to actually use it to perform what you want.

You might have better luck with low-end industrial robots but the cheapest are $2500 and more.

(I’m similarly in the market if a product can prove me wrong, especially interested in something with some computer vision integration)

lokl
Thank you, this seems like a very important point. Which arms did you use (if you're comfortable sharing)?
omneity
Just random brands. I think it was something like Yehboom and Sunfounder or something.
ianai
So I’m very far out of my depth here. But it seems like we humans learn how to use our limbs through something like trial and error. Ie in other areas, babies vocalize a lot of odd sounds that vaguely, eventually sound like words. It “looks like” what random variation on inputs and weighting based on outputs might “look like” third person. Is that way far off the realm of possible with motors/etc?
omneity
It’s a very interesting thought. I personally am not a robotics expert, but it seems to me there are two obstacles: loading the arm with enough sensors and encoders to detect outside feedback reliably, and the speed of learning iteration (limited in the real world by physical laws etc).

Maybe look into this? https://github.com/cyberbotics/webots

high_byte
idk about anything specific or anything recommended here. but thought I'd share my thought: I've seen one on kickstarter about 2 years ago. not sure the name but it seemed very practical.