@caesarm

Hey v3.

This round was all about getting the design closer to the final product, mainly the robot base and the swivel base assembly. A lot more is actually in the CAD now: screws, heat-set inserts, bushings, bearings, and servos. I also designed the parts that mount the main arm assembly to the swivel deck, and fixed a handful of my own design setbacks and assembly issues that only showed up once real hardware was in the model.

The base, fully detailed
From top to bottom, the stack is:

• the rotating deck with the mounts for the arm
• the top cover with the pocket for the J1 servo
• a hub plate
• the thrust roller bearing
• the deep groove ball bearing
• the 2:1 herringbone gear
• the housing, with a 608 bearing at the bottom for the gear shaft

Every screw that goes into plastic lands in a brass heat-set insert instead of printed threads. PETG/ PLA threads strip after a few cycles, and this guy will probably get taken apart a lot during build prototyping, not to mention strength concerns. Modeling the actual screws also forced me to check head clearance, screw lengths, and whether an allen key or t-handle can even reach them once everything’s stacked. It was boring but worth it.

The swivel deck is now the real interface between the base and the arm. The two J2 servos will sit on either side of the deck, and every arm pivot gets a bronze bushing as a cheap, replaceable wear surface. The pins and bushings carry the linkage loads, so the servos only have to supply torque instead of holding the arm up by their splines. I thought of direct driving just now, but it’s currently just a thought.

Some setbacks:

• The coupling with the curved support linkage. After realizing some ROM by messing around a lot with the assembly, realized the front counterweight option I planned for had to go. To add, I moved the mounting towards the inside of the base instead of on the outside, which should give me some more movement liberty in the long term.
• The actual shape of the robot base. I rounded it out and focused on symmetry so I would have an easier time assembling. Converting to a circular shape and not just the simplest form also opened the gate for more equalized load bearing.

Reality check !!
I don’t have any of the hardware yet, so everything is designed off online CAD models. From experience, that’ll set me back once the real parts show up.. nothing is ever perfect without real references.

• Downloaded models are usually nominal or simplified.
• Heat-set insert hole sizes change by brand.
• Bearing fits depend on the printer.
• Some more boring stuff to do with tolerances and measurements

So I’m trying to keep the critical interfaces easy to tweak, and once everything arrives I’ll print small fit tests before committing to full parts.

Next up I’m planning to..

• Give the arms the same treatment: pins, bushings, spacers, fasteners, the works; making it super functional.. and look super wicked.
• Double-check shoulder screw lengths against the bushing stacks.
• Update the BOM with the real fastener and insert counts.

The tool-head standard and the 1:1/direct drive question are still on the list too, but those have to wait for hardware.. gabagoo.

This is v2.

What’s new
I designed the full base assembly: outer shell, swivel bearing interface, and rotating deck. The waist joint (J1) driven by a 2:1 herringbone gear reduction off a STS3215. After research, I chose to use these motors with a 345:1 @12V on each, up to four motors total: one for the yaw, 2 powering the shoulder linkage (J2), and another for an eventual 4th DOF (J4) on the toolhead. This tiny motor already has a built in 12-bit encoder, which eliminates the need for an additional electronic part, aka additional point of failure.

Nailed down the project’s actual design goal (see below — this matters more than it sounds)
Started thinking seriously about tool-head modularity for the business end of the arm
The base / waist joint

I decided on herringbone gears since they self-cancels axial thrust; a straight spur or single-helical gear at this scale wants to walk sideways under load, herringbone doesn’t (that much). Plus, I’m 3D printing it, so machining manufacturing isn’t a concern (yet). I also use a deep groove ball bearing on the outside for smooth rotation and a thrust roller bearing on top to reduce my axial loads. Initially, I was planning to use opposing tapered roller bearing, but this approach is cheaper (and lighter..).

For the 2:1 ratio, my main motivator was to get the least strain (and heat) on the only tiny motor carrying the weight of the entire robot (excluding payload). Using torque calculations, I realized that the 2:1 might be unnecessary and slow, so I already envision converting to a 1:1 or direct drive mechanism after seeing the arm in action. In the next iteration, I’ll tackle the modularity of the bearing base to be able to use an assortment of different sized bearings and sprockets, making it easier for anybody wanting to give the build a shot. My only concern is that this joint could be a silent generator of slop, which can hopefully be mitigated by the servo encoder.

While most hobby arm projects chase 6 DOF because that’s what “real” robots have, I’m deliberately not doing that. It’s kind of overdone/ overrated, but it still amazingly effective. This robot has a different purpose.

The actual goal: a cheap, consistently precise 3-DOF arm waist, shoulder, elbow, amazing to use by itself but increasingly more capable with added attachments/ accessories.. a bit like an Ender 3..lol. That’s the basis of the palletizing topology this guy was inspired by in the first place; those robots don’t need a wrist full of DOF to do useful, precise, and repeatable work, and one robot can be adjusted/ modified to perform a plethora of tasks. Precision, repeatability, and real world practicality is the actual metric I’m chasing, not DOF count. In my eyes, I don’t want to build this guy and be done with him, I want to get and learn the most of the robot, especially since he’ll be mostly plastic.

For modularity, the tool end is the extensibility point

Instead of building DOF into the arm, the plan is to build capability into swappable tool heads at the end effector. Same 3-DOF base arm, different job depending on what’s connected on.
Next up I’m planning to..
Streamline the tool-head mounting standard design (connection pattern + electrical/pneumatic passthrough).
Print the base and validate the herringbone mesh + swivel fit in (preferably in) PETG
Get all 3 servos on the bus talking together and do a first coordinated move. There’s also some more stuff, but I can’t fit it all here lol.. until next time.

This is v1 of the overall robot arm build, which is a 3 DOF antagonistic drive pick and placer, inspired by factory palletizing robots. It’s the first physical version of a linkage-driven arm I’ve been working on in SolidWorks. I already got a few issues with this version, just from looking at it on screen, but I’m going to 3D print it to see how it actually behaves in the real world before making changes.

Design Approach
I used existing automation robot references for primitive geometry; referenced from Fanuc, KUKA, and Panasonic industrial/palletizing robots. These companies have decades of field-proven arm geometry behind them, so rather than guessing at joint offsets and link ratios from scratch, I used their designs as a sanity check for what actually works in practice. I’m already researching their electronic work using different control options like CANBUS and EtherCAT to implement (at least one version) into this robot.

For the largest (by volume) components (anything with organic-looking cutouts or holes) I ran SolidWorks’ topology optimization instead of hand-sketching lightening patterns. To make sure my studies were actually improving my design instead of making it worse, I cross referenced different parts across the industry (excavators, cranes, and other heavy equipment, since an excavator is what inspired me to create this robot) and used my knowledge in steel mechanics to see if a part would hold from a glance.

I designed with FDM 3D printing’s strengths and weaknesses specifically in mind rather than just printing “machined part” geometry and expecting it not to break. At the same time, I kept track of which components are candidates to eventually remake in aluminum once the design is proven out, so the plastic-specific DFM choices don’t paint me into a corner for a metal version down the line.

What’s Next

Print V1 and put it through real motion /hand-stress testing to confirm (or rule out) the issues I’m already suspecting. Then, I’ll revise the design based on what breaks, binds, or flexes more than it should have. After, I’ll start on the electronics side, although I’m going in with limited parts on hand, so that’ll be its own challenge.