This is a robot arm similar to the Barrett WAM (which is differential based). However my robot arm has differentials at all arm joints. Thus the arm can be configured (depending on how you drive the motors) as a 4, 5, 6, or 7 DOF robot arm. The base rotates which is one of axes followed by 3 pairs of differentials.
Where my arm is different is in its cost, at appx. 500$ cost + electronics, gripper and camera.
- 1 meter reach for actual usable capacity
- 1 KG payload (hopefully at full reach)
Hopefully you can use such a robot arm for many uses:
- Pick and place (PNP) with vision system
- 3D printer
- CNC or 3D printer machine loading/unloading/processing
- Do the dishes
- Fold your clothes!
After having built the first two iterations, I realized I needed a better base design. I wasn't happy with the accuracy or stiffness of the 16ga steel folded at 90degrees so I have decided to move the motors eternal to the arm structure, which ends up making it quite a bit more compact. It is also a bit similar to the design now in use by Boston Dynamics "mini cheetah" / electric mini-dog arm.
Here the motors are still NEMA34s geared ~4:1 or 5:1 to the 20mm axle.
In order to counteract the base torque, I have added a counter-weight (with associated cutouts) which also acts as the thrust washer mount/pivot. The counter-weight is massive - 1/2" steel plate laser-cut, weighing around 3 lbs!
I have kept the original adjustable base tensioner:
and now also added them in the arms:
Here you also see one of the modifications I did for assembly - the motors are installed on the diagonal (small hole + slot). The larger hole is for being able to assemble the opposite side motor.
Thoughts on the redesign?
I have ordered the parts and hope to start assembling this weekend (iteration 3). It's a long process and I hope to get the motors moving soon.
In the meantime, I will try to get the existing arm moving a bit. I have bought three Makerbase MKS V1.3 32-bit smoothie clone boards as testbeds (so I don't fry my legit Smoothieboard 5XC), I hope to connect them up and start testing soon.
Laser cutting definately have advantages. One is that steel makes an excellent spring material.
Why not kill two birds with one stone?
I've now integrated adjustable leaf springs into my laser cut tube directly.
The springs are the three "lines' present, which are not directly mirrored but are set 180degrees from each other on each side. By changing the thickness of the printed "counter-balance", I can "engage" one, two or three of the springs. In addition, I can change the preload by redesining the printed part, or by adding a u-bolt around the entire tube (similar to truck suspension).
This is one of the possible designs I have been considering for the stackable "pusher". As you can see it is bolted to the 8mm axle pivot (which is now a solid aluminium bar instead of a printed part with bearings).
In the first picture you also see some modifications to the old tube design (which was squared off and which was snagging on the rotator/pivot). The end is now all curved to allow the pivot to go around. In addition I added holes for IGUS igubal self-aligning bearing installation.
On the base I have added a very large (huge) counter-weight to balance the arm. See next post for more details on that and the rest of the substantial redesign, all of which originated from building the original arm.
After assembling the first, then the second arm, I am really getting tired of tapping all those threads manually in the plastic. I am thinking of just using heat-inset tapped inserts but it would be another thing to buy for everybody. Also bolting the motors and joints take too much time (especially with nylock nuts). Any suggestions for making things easier?
Here are pictures of the two together.
Already so many parts need some redesigns, I'll be updating the STLs on Github hopefully before the weekend.
What feature do you want to see in the robots? Any suggestions for the next post?
The original goal of the robot arm was to have a high-accuracy 4-DOF palletizing robot.
However over the last few months I have been working on something far more adaptable - a 1meter reach collaborative robot arm with 7 DOF. The arm should be just as accurate when combined with visual servoing with a camera on a stick above the robot and one on the wrist.
The main design architecture of the arm is a NEMA17 stepper on the base to rotate it, and differential drives on each pair of joint DOFs after that - NEMA34 on the base, NEMA23 on the first arm and extra-large size servos (~30 kmcm) on the last arm.
The end-effector connector is an ISO-standard mount (same as on the Universal Robot UR3/UR5/UR10 robots).
On the github project link you can find the files in STL (3d printed parts) and SLDPRT for the metal parts (Solidworks design but I'll try redesigning them in Solvespace). The issue tracker there is also live.
See pictures below for the first arm build (which has already been modified to save weight)
In order to get a better idea of the mechanisms involved, I ordered and assembled a MeArm. I must say, for the price it's unbelievable! Here are a couple of photos.
There are a few improvements I'd make. One, I wouldn't try supporting the entire mass of the arm a tiny 9g servo. I would try to support it on a bearing plate, or thrust washer or something similar. It can even be a piece of plastic. Two, actually there is no two! It works well!
In order to start designing mine, I have begun to use the above 2D-lasercut parts as a "baseline" for my own arm. I used to use Solidworks but now I'm trying to keep it all open-source including the design. Thus, OpenSCAD. I have posted the beginning of my SCAD code here: