Robotic Arm

Thinking about building a robotic arm to play with.

Don’t have a specific use in mind yet but I would like something prosumerish that works with ROS/Gazebo etc and that can lift and manipulate something around a pound.

Hoping to stay in the $500 range.

I like the AR4 but its more than my toy budget right now. Any suggestions?

2 Likes

I think there is definitely room for a robot arm in the $500 range.

Sort of in the spirit of the MPCNC where the pro tools are very costly and a surprisingly good machine can be had for $500.

I respect the AR4 and the robot arms made by “Skyentific” on youtube, but I don’t think you get to something useful by taking a traditional concept and make it smaller/cheaper with plastic.

My idea is put counterweights everywhere. The elbow has a counterweight to offset the weight of the forearm at the elbow, and the shoulder has a counterweight to offset the entire weight of the arm. Then use block-and-tackle ‘tendons’ to amplify the forces from the motors to the joints. It will be relatively slow.

For extra credit, put the counterweights on a slide so they can move out and when you pick up some weight so the arm remains balanced.

For additional extra credit, use an optical feedback system to measure the precise position of everything and use that to maintain accuracy.

3 Likes

A lot more goes into those than I expected. Lucky I got hungry, or I might have gotten stuck in that rabbit hole all night. The hardware seems almost irrelevant to the drives and programming.

2 Likes

I think trying to design one would teach you a lot, very fast. The mechanics are not intuitive. But I bet there are rules of thumb, lile the should has to be 2x stronger than the elbow, which has to be 2x stronger than the wrist, etc. And I wonder if that varies by material. Something that was very strong for its weight, like titanium, could probably get away with a much smaller ratio than cast iron or plastic.

And at each joint, you will have very interesting mechanical problems. Things like backlash will be tricky to deal with because you will have a joint that could be at any orientation w.r.t. gravity.

You’ll learn a lot by doing something like this. Another way to look at it is, there is probably a lot of good info out there already.

1 Like

Yeah, I keep thinking you have to do it backwards. Start from the hand and work your way up…each joint getting progressively more rigid and stout. In one of those videos he mentioned two kinds of drives, I need to look at them, one being more accurate and one cheaper…sounds fun.

That is a great way to start. Try to make the second joint as small as possible while having a functional first joint, and try to get some kind of idea of how much bigger you need at each section. Then do some math to see if you can afford the shoulder.

Of course, I am a physicist by nature. The engineer in me says to just find someone else who has a design that works.

1 Like

It’s a lot to unpack for a non engineer like me… pretty good on software but on hardware I’m not familiar enough with the toolset to even know what’s possible.

I guess that’s part of the adventure! Thanks guys.

3 Likes

Jeremy Fielding did a series on building an arm.

There are a lot on thingiverse and printables. Are you thinking outside of those capabilities??

Really my goal is to get to the software side ASAP. I want to play with machine learning using real world hardware and use some existing libraries to handle the machine control.

My thinking hasn’t really progressed past that yet.

2 Likes

ROS (robot operating system) has a bunch of the groundwork for the software figured out already, including a simulation of robot arms and the inverse kinematics. I’ve worked on robots for a long time and having the ros infrastructure really helps. Just having things lole their pub/sub, data collection and replay, logging system, and visualization tools really helps.

I found this article that seems like a good start (I skimmed it, but didn’t read or follow it):

https://medium.com/geekculture/reinforcement-learning-path-planner-for-6dof-robot-in-ros2-518581dc72c7

3 Likes

So I just saw this and it looks amazing. Thought of this post right away!
https://www.printables.com/model/584932-parol6-3d-printed-desktop-robotic-arm

It is printed, but it looks like they have everything, inluding software on their github.

3 Likes

The PAROL6 looks great, no makes yet though?

EDIT: Do you have any good use-cases for a robotic arm?

Not sure yet. I have a few ideas but no idea if they are practical. Interested in computer vision and ai.

Mostly trying to stay ahead of my high schooler who’s really into robotics.

I had a computer vision project in my graduate college (shout out to the Colorado School of Mines) image processing class to provide feedback from a welding arm robot. There was an IR camera capable of recording images of the weld pool and I wrote an algorithm to measure the size of the weld pool. The pool was supposed to be feedback to a control algorithm to make good welds on various 3D surfaces, without it having a perfect model of the parts. After I left they gave the code to a post doc and he didn’t like my approach, so the code didn’t end up getting used.

It would take a lot of money and risk to build an actual welding robot arm. The camera may be highly specialized, so expensive.

But the problem is interesting. And I am sure someone would want to use a neural network and deep learning to solve it now. But it was simple image processing for me (threshold the image, remove speckled noise, Open to smooth it out, find the major axis with eigen vectors and return the length of the larger eigen value). None of that is “AI”, but marketing would call it that now.

Maybe these is a way to have a useful analog without actually having to weld? Like getting feedback from a marker to gauge how hard it is being pressed. Or the width of a flexible paint brush to make a good taper on a not-flat piece of paper.

2 Likes

It’s really hard to tell what to expect for BOM cost. Maybe $1000 to $2000 range?

I don’t think a single arm could assemble a copy of itself from parts, but maybe if you had a stationary gripper (like an electronic vise) then you might have a chance. Still extremely hard but with only one arm I am guessing you have no chance.

If the cost of parts is low then boom, there’s your robot army.

I must have missed something, I thought the o.p. just wanted to build a Robot Arm, not have it self reproduce???

1 Like

That’s what I love about this place. Someone is always taking it to the next level.

My wife wrecked the car yesterday (she’s fine) so my $500 budget just went to $0 for a while.

Probably start when the weather gets cold.

1 Like

Start printing pieces, then go big after that :wink:

Ok so when you are ready this is in your 500 dollar range! And they have kits like Ryan. Printed and hardware! Darn, but they charge 50 bucks for the cad files to print!!!

2 Likes