pic: GBX-162, Realistic Swerve

GBX-162 is a realistic swerve drive. In the past, many of my designs have fallen short in one way or another or had failure points, but this is pretty much how I would make a swerve drive if I had to do one in-season. It weighs 6.4lbs with a CIM or 5.8lbs with a MiniCIM (without the versaframe shown in the photo).
Geared for 17.4fps free speed, it’s fairly fast without sacrificing acceleration. To change the ratio to something slower I would need to switch to a larger bearing to suit a larger gear, but it wouldn’t be a terrible change to make.
Of course 1072 will never use this design unless the students come up with something on their own! :stuck_out_tongue: But I figured it would be a good idea to make a swerve drive that I’d feel confident in fielding in CADathons at least.

CAD files are located here:
Look in GBX-162 -> Release for Pack&Go and STEP files.

You always have the most gorgeous renders.

One thing I noticed about your mounting: Have you considered 1533-style slots instead of the holes? It nerfs some of the pocketing, but it does give you the ability to locate it along any length of VersaFrame without having to poke new holes in the tubing.

Do you have the parts list/CAD for this?

It’s a small thing, but thank you for listing free speed rather than “top speed” with some unknown/arbitrary “speed loss constant” applied to it.

Nice and compact! Is that a ball bearing on the end of the CIM shaft and in the center of the large rotational gear? I would omit that part for fear of overconstraining the motor shaft. The shaft is strong enough to support the pinion without that bearing.

What is being used for the large bearing for the module rotation?

I’m asking this because I can’t clearly see it, but how are you planning on mounting encoders to track distance and rotation of your module? Beyond that it looks great, good work!

I love the packaging and small form factor of this swerve! Having attempted to design swerve on my own multiple times, I understand how difficult it is to get something nicely packaged. I was wondering if those screws on the bottom were all the support that the fork gets in terms of radial loads? Do you have another bearing, or is all the load through the bolts on the bottom? Also, what do you have in terms of taking the thrust load?

You know, my renders used to look only decent until I saw Nick Coussen do a render on a dark background, so now I use the same background for all my renders. :stuck_out_tongue:
I like the idea of the slots, but I don’t particularly see the need honestly. I don’t like using Versaframe over just making my own tubing, after spending hours this year getting VF just right. Given that we have a CNC and a manual mill, it’s actually faster for us just to cut tubes to length on the manual rather than use VF.

I’m just generating the STEP and Pack&Go files now, and will update the description with the link as soon as that’s done.

I always get mildly peeved when I see “adjusted” speed listed. Adjusted for what?

There’s a balance to be struck here with the overconstraining. On one hand I agree with you and don’t want to risk the shaft binding. But on the other hand, if I don’t have it, tolerances could totally throw the spacing of the CIM pinion off relative to its meshing gear. Tricky, tricky stuff. I’ll probably remove to make room for an AMT CIMcoder on the CIM shaft.
I’m using a 6814 bearing, 70mm inner by 90mm outer, 10mm thick. Very cheap alternatives to the Silverthin bearings a lot of designs use.

Whoops, I knew I forgot something! The plan was to use Muchskull’s AMT CIMcoder for the drive, I’ll add that in before releasing the CAD. Rotation is settled with a Versaplanetary Integrated Encoder.

The setup here is similar to the one 1323 runs, so I’m not too worried about most of the lower module specifics. The screw is a dead axle for the wheel, and the forks are attached to the turning gear by means of 4 10-32 screws tapped into the forks with a couple of tiny pins on one fork for locating the bevel gear. Because of the small wheel it’s quite easy to make it compact and strong.
If you mean the thrust loads from the bevel gear, those are taken just on regular ball bearings. Other swerves have run the same setup without issues. There is space for a roller thrust bearing, but I wouldn’t add it unless I needed it.

Is it realistic because you’re never gonna use it?

Is it realistic if you make it and it doesn’t work?

You know Anand just as well as I do - no way he’d make a swerve in season.

Link to CAD models (also in description): https://workbench.grabcad.com/workbench/projects/gc8mrFALmKICo4R9VHLXRcuk-mX7BOz2JCA6ISgFtmNUQy#/space/gcSTdRRbtyPVJLJ_9o0kq0T2DyGoJ3GMIQSQ0z1UVX5D8O

I added Munchskull’s CIMcoder and removed the bearing in the middle gear.

Bingo. Not worth the time it takes IMO. COTS swerves, however, are on the table. The Swerve&Steer still has numerous issues, but here’s to hoping for improvements!


Care to detail what’s wrong with Swerve&Steer?

Apart from tolerance/machining issues in 2015 when they came out, there’s no way to add a drive encoder, it’s heavy, and it’s super tall. They require you to use an MA3 for rotation, the top speed is fairly low, and the use of a PG71 for steering means that it’s slow to rotate. A copy of 1323’s swerve, or even a Team221 module, would perform better. Of course the latter is not available anymore.

We will def have some form of a swerve module for sale. After 3 iterations we think it’s ready for sale!


Having more robots moving around the field like 1323 would be a game changer. Awesome news R.C.

Your season has been a very good advertisement so far