pic: GBX-150, CIM-mounted swerve

GBX-150 is my 50th design for a swerve drive. Few of my swerve drives make it past the sketch phase, but there are 50 unique ideas for swerve drives in corners of my computer.

-14 to 20fps free speeds determined by selection of pinion and mating gear (18 or 20 tooth)
-149rpm module rotation speed using cycloidal VP stage
-2.2" tan-nitrile treaded wheel
-5.1lbs (I tried so hard to get it sub-5lb)
-Ultra-compact (probably my most compact ever)
-Mounts to face of CIM, and CIM is mounted via a large clamping arm with a 6-32 screw (although 8-32 and/or set screws are doable)
-S4 for turning encoder, with no speed encoder. Hall effect zeroing sensor can be added if desired
-5 CNC’d plates, 2 turned shafts, and a VP shaft drilled out for 1/4" S4 encoder. Everything else is COTS.

I’m pretty proud of this swerve drive, but feel free to kick me down a notch. :stuck_out_tongue: I’m sure there are aspects of the design I’ve overlooked (the least of which being the use of a clamp on the CIM to hold the module to the chassis). That being said, the main weak point of this design is the clamping attachment, and the use of the black paint on the CIM as a shoulder coupled with possible extra set screws should make it able to resist whatever the game throws at it.

I’d switch the turning encoder out for one of the versa planetary integrated ones but other than that, this looks cool. Good work!

Independent of the swerve aspect, the idea of mounting a wheel based only on the two 10-32 mounts on the face of the CIM is a bit scary. Mounting that arm with a 6-32 and putting a swerve mechanism between there and the wheel have a bunch of alarms going in the back of my neck.

I’ve seen plans for a VP-compatible cycloid drive, but I don’t recall that it’s been built, much less made a COTS item. Did you include that in your machining requirements? Is there any advantage of the cycloid drive (vs planetary) in this application other than the obvious weight/space savings?

The main forces on the CIM are thrust loads against the face rather than shear forces, and she shear forces are limited by wheel friction. I’m more worried about the attachment arm. Looking back, I think I can redo the design to use no friction locking at all and use traditional plates, so I might try that next.

VP cycloid should be included in the machining bill, you’re right. I like the height savings it provides, mainly.

The thrust load isn’t the concern, especially as it’s compression. However, if the CoF of the wheel on carpet is greater than one*, the shear forces are potentially **greater **than the thrust load. What really concerns me are shock loads.

  • Except for Lunacy swerves, this seems to be the rule rather than an exception.

Looks very cool. I’m glad to see you are still iterating on your swerve designs.

What I can’t gather from this picture is how the primary bearing setup would work. Could we get a cross section view, or an explanation for that?

That’s a good point; I hadn’t thought of it that way. Looking back on it, the whole clamping mechanism is a cool gimmick, but not something I actually need… I could also reduce the weight somewhat with other constructions I think.

Thank you!
The bearing setup is definitely an oddball one. Here’s a link to the 2015 Offseason folder with the CAD, if that helps:

Essentially, a plate is screwed into the CIM and firmly attaches to the inner race of the 6711 bearing. The outer race of the bearing is pressed into the 64t turning gear and captured with the flanged buttonhead screws seen in the picture.

I like the concept, as someone who has utilized the large steel sleeve of the CIM structurally before.

I will warn against taking any loads with the face of the CIM, this is a rather thin wall cast component, which I have seen break in normal mounting configurations.

See if you can find a way to clamp higher up on just the steel sleeve, for both parts, the module rotation and the main frame mount. It’ll be interesting to accomplish that and still properly pilot the CIM for the gear mesh.

Also I have next to zero confidence in that wheel/tread setup surviving for a match, once you get past a certain point there is simply not enough contact to transfer the forces required without extreme wear, one of the reasons I stopped at the ~3" diameter territory with my designs.

R30 (as of 2016) allows modifying a motor housing for the purposes of mounting, which certainly opens the door for structural mounting of a CIM.

It may be possible to entirely replace the output shaft plate of the CIM with your main anchoring plate. Only restrictions I’m seeing from the rulebook would be ensuring that your design is not lighter than the original, and that the electrical and mechanical operation of the motor have not been modified.

+1 :slight_smile:

That CIM sleeve is a beefcake. See linked cartoon-CAD of our 2016 winch. (We ended up using an igus bearing, not the PTFE called out in the cartoon. Metric 64mm i.d. fits nicely over the CIM sleeve paint. See picture of the innards also.)

My theory as a motor designer: if you have more than the minimum steel required for magnetic flux, the excess should perform another useful function to justify its weight, cost, claimed space, etc.

Ok, so does anyone have a spec for the wall thickness for a CIM? Mini CIM? Bag CIM? I’d love to see more and more designs delving into this somewhat uncharted territory of using the CIM body as a mounting point.

I think last time I measured it was ~0.090" thick, it’s not going anywhere.

Thanks for the advice. I thought the face was machined, but if it breaks that easily it looks like I’ll have to rethink a lot of it. I’m currently working on an iteration of this idea that should be stronger while pushing the weight under 5lbs (finally).
Does the tread just come off such small wheels, or is there another problem? I was thinking of just using a colson if what you’re saying is the case.

I recall seeing old threads about a team machining the paint off the CIM to make it shiny, but I want to avoid machining more COTS components than absolutely necessary. That being said, replacing the front plate of the CIM would be an elegant solution.

Ok, thanks. I’ve only opened a CIM once, and I wasn’t paying attention to the sleeve as much as the internals.

A colson would work better as the bond to the polyolefin core is very strong, but it would still wear out faster than you’d think.

Replacing the ends of motors has been on my “i’d like to do this…” list, but I never have as I’d likely lose the rules argument, you are getting a weight/performance advantage when you integrate the motor into the structure like that. And I read “modified to facilitate mounting” as “mess with it some if need be, but don’t replace it”.

I had a concept once that would replace the dead axle of a window motor with a powered shaft through to get an on-axis steering motor, but never made it past that as I’m pretty sure I’d lose the legality argument.


Nice design. You even have the weight of the robot pinning the steering bearing into the wheel module and the mounting plate. Others have hung the weight of the robot off the screws on the edge of the bearing which is a no-no. I don’t like relying on screws that aren’t seated fully like the ones you have on the main steering bearing but since you aren’t holding the weight of the robot off them it should be fine. I would use stripper bolts or machined tabs but with a little loc-tite the screws shouldn’t be an issue. Plus they are very close to the main plate and if one came loose it could lockup your steering for a match or burn out a steering motor.

The answer to your problem of mounting the CIM to the main plate is easy.
Flange the motor/bearing adapter (plate between the CIM and steering bearing), and make that flange larger than the hole in the main plate, assemble the whole steering module and slide the whole thing up into the plate until the flange contacts the main plate. Then you can clamp on the minor OD of the adapter instead of the CIM or bolt the flange to the main plate (my preference). If I am seeing everything correctly you should be fine. It will add a 1/4" to the overall height but that shouldn’t be an issue.

One thing we’ve found is when the steering gear is smaller in diameter like this, the backlash in the steering gear to steering gearbox and within the Versa Gearbox, you get a decent amount of steering backlash. We are going to belt steering over gears due to this. Modifying a steering gear is simpler and easier, but for autonomous routines a low backlash steering helps a lot.

Awesome new evolution of the type of co-axial setup Aren started a couple of years back where the bevel gear is on the wheel axle. Though it looks like you’re using a tiny 1/2" x 2" wheel, that could also be part of the compact look. Our 2015 1"x3.25" wheels looked laughably small.

Thank you! I’ve never really liked using the screws partially to capture things, but other solutions require so much more machining or cost I don’t like to use them. Fortunately most of the time they don’t support any load anyway.
I actually came up with the idea for using a bevel-beside-wheel before it came up on CD, but my execution was nowhere near the finesse that Aren’s (or 1323’s) had.

I see what you mean about mounting the CIM; I realized that after I already did the render. The method you’re describing is a little different from what I had in mind, however. I was just going to replace the clamp with a plate that goes beneath the CIM to mount to.

I really dislike using things like large timing pulleys for this (I’ve done it a couple times), but I completely see your point about why to do it. Usually for autonomous I’ve had experience just using a NavX for guidance, and I don’t think you need complicated swerve maneuvers for auton anyway, but I can think of several autonomous problems that are harder to solve here.

Now I’m getting paranoid about the small wheel. :stuck_out_tongue: I remember that nitrile tread not wearing out for around 25 matches + practice + offseason back in 2014, but the AM treads wears out a lot faster iirc. I wouldn’t think that tread wear increases exponentially over a lower width + diameter (in this case 4x faster assuming it’s a linear relationship), but please correct me if I’m wrong. ATM I can live with 4x faster tread wear.

I revamped some of it using some of the suggestions, and finally got it under 5lbs! (4.99lbs, to be exact) I’ve been trying to do that for two years, and this is definitely less sketchy than some of my other attempts.
The top support clamps onto the CIM with set screws, but the bottom uses a plate to support the CIM and directly interfaces with the bearing. It’s a much stronger connection. I also upgraded to a 30t gear over a 24t gear for the turning gear to reduce the backlash a bit.
That being said, I’m still not sure I would use this for most competitions, or at least not without a couple extra pieces. The most sketchy part is that I had to resort to using a press fit to hold the large bearing in the gear. The thrust loads are all opposite the press fit, so theoretically this is 100% ok, but we all know how theoryland is deceiving.
The most dangerous parts of this are the small bevel gear shaft bearings, and the press fit. Everything else I feel fairly comfortable with.

Very cool! And thanks for sharing the native SolidWorks files. I’m very interested in working across teams and iterating on each others designs.

I look forward to seeing video if/when you build & test it!

I always love seeing your designs.; they are the reason I taught myself CAD so Icoukd aspire to make stuff like you do. A sub 5lb swerve module is an amazing accomplishment! I have a few questions:

  1. How is the small bevel cluster shaft being held in (on the top and bottom)? I see there’s something under the small bevel gear but I don’t see what it is.

  2. Unless it’s an interference fit, I could imagine a scenario where the robot rides up on a field obstacle or another robot and the gear gets pulled off the large bearing. In your first version of 150 you have screw holding the gear and bearing together. Did you remove them for weight savings or was there another reason?

  3. Is there an encoder on the CIM that I can’t see? I though that was usually a requirement for swerve.