Two More Swerve Designs you Can Make In-House

Hi CD. It’s been a year and a half since I posted a design for a Swerve Module that teams with modest manufacturing resources can make in house. Although my team hasn’t had a chance to build one yet (our quarantine only recently ended), two other teams built variants of the design in 2020/2021. That’s extremely gratifying, but also provides some good “Operating Experience” to feed back into design improvements!

With that in mind, let me present two more Swerve Designs which you can build in-house (if you have access to a router and a 3D printer and mayyybe a lathe, but maybe not).

1) CX21-Swerve7-A001

OK, the name is pretty uninspired, but it’s arose after mucking around with at least 6 other design ideas. I’m happy to hear better naming suggestions! The key points of this design are:

  • Motors are flipped, for an overall height of only 6.5". I think this is possible the shortest module design I’ve seen for 4" wheels, and it means the whole module fits below the bumpers even if the ground clearance is only 1.5".

  • Uses a “single main plate”, for reduced machining time and costs.

  • Main plate is designed to fit a “WCD” style frame, but it could be adjusted to mount in the corner of a simpler square frame.

  • Drive is 2 stage (instead of 3) for better efficiency. Ratio can be adjusted to your liking, since the pulleys are 3D printed. Uses a double wheel setup like the Thrifty Bot module as a result. The forks are made from 2D cuts in 4" square box tube.

  • Needs a 1/2" hex bore hub with a 1.625" hole PCD to support the larger miter gear, which no one makes right now, so that’s a bit limiting

  • designed for 4" diameter x 0.875" wide Vex Colson wheels.

  • Steering is 2 stage, using one of my favourite inexpensive gearboxes - a double belt reduction with a 3D printed compound pulley (mounted on a dead axle / shoulder screw). This saves the cost of extra spur gears and/or planetaries. Steering ratio is about 15.6:1

  • Absolute encoder mount is a variant of my previous design, with thicker walls in the 3DP gears (thanks to feedback from team 359)

  • Module weight with motors, wheels and hardware (but not controllers) is 5.8lb according to SolidWorks, which is a bit heavier than I wanted, but close to COTS modules.

Sizing drawing: cx21-swerve7-a001.pdf (111.4 KB)

`

2) CX21-Swerve9-A001

Again, ignore the naming. Swerve8 was a dead end, but I’m happy enough with the 9th try :slight_smile: . The key points of this design are:

  • More traditional motor-on-top mounting makes the module 1.5" taller than Swerve7, but also about 1.5" less long. At 8" tall, it’s still shorter than most COTS options.

  • Moves the “Azimuth plate” down to just above the small miter gear. The miter gear shaft has also been pushed a bit further from the steering axis than my previous design (1.25" instead of 1"). Doing both these things lets me accomplish a lot of goals:

    • Lets me support the miter gear shaft with flanged bearings in the azimuth plate. There are no bearings in the forks. The shaft then becomes a simple cut-and-tapped bit of thunderhex. The upper bearing is supported in the 3D printed steering pulley.
    • This makes the forks much simpler - in fact they can be almost entirely made in a single Op in a router. Only the tapped mounting holes have to be added, and I imagine a 3D printed jig could make that into a hand tool job.
    • Lets me fit a 1.5" Colson Wheel beside the bevel gears. To mount the gears I imagine that the WCP gear could be silver-soldered to a 1.25" OD x 1.125" ID tube, and then they would be pressed into the Colson Wheel. If that doesn’t work, this design is also compatible with both SDS and WCP billet wheels
  • Drive Ratio is somewhere between 9.5:1 and 5.3:1 (or 10 ft/s to 17 ft/s free speed), depending on motor pinion choice (between 12 and 17 teeth). Steering ratio is the same as Swerve7. Ratio calcs can be found here.

  • Module weight is a little lighter at about 5.6 lb

layout drawing: cx21-swerve9-a001.pdf (188.8 KB)

3) Overall Fit and More Details

Here’s how the two modules might look mounted in a WCD style chassis (both types of modules are shown, but I don’t think you’d ever need to mix and match them on one frame.

How much do these puppies cost? I was hoping that by having so many parts routered or 3D printed, the cost of these modules would come down under $300 ea. Unfortunately, when I took the time to write out and price the Bill of Materials, I found that just the COTS parts alone cost over $350 :frowning: That’s still cheaper than the WCP or SDS COTS modules, but still more expensive than hoped. I clearly have work still to do to reduce costs.

Here are the CAD files in STEP and original SolidWorks. Feel free to download and remix to your heart’s content - but please post to CD if you do so we can all learn and improve!
https://drive.google.com/drive/folders/15k6W7pmGoGp6dXrbeTZCc_KWBiqROgzY?usp=sharing

23 Likes

Love the work. I would suggest that you switch from 3mm pitch belts to 5mm belts. When we did our testing for Swerve X all teams noticed issues with stripping the teeth on the belts . Our solution was going to 5mm pitch x 9mm wide.

You can probably also run the math and see if 3mm belts are within the load ratings.

16 Likes

I like these designs. What I’ve found is that the major way to save money on these is to move to metric hardware and source as many parts are possible from overseas (including using a 68xx bearing instead of an X-contact and a AS5600 encoder over a CANcoder). With COTS options being like they are these days, it’s much harder to justify a custom module unless you have specific dimensional requirements.

2 Likes

The designs have 3mm pitch GT2 belts on the first stage, and 5mm pitch HTD belts on the second stage. Does that change your recommendation? I’m hoping that’s still ok, because HTD would require a bigger motor pulley.

1 Like

You’re not kidding! I tried to make these modules as simple as possible, and still ended to with over 50 lines in the BOM.

The biggest cost drivers are in order:

  1. Motors (can’t do anything about it)
  2. Cancoder (not many cheaper options yet)
  3. Miter gear pair (not likely to get cheaper)
  4. X-contact bearing (could change to 80x100mm bearing, but mounting becomes harder)
  5. Wheel

Those five alone take you to $250 before you machine anything!

The COTS options are great, and come with the added benefit of being able to reuse them year over year without breaking the rules. We’re considering making our own mostly for the challenge (especially since they wouldn’t be much cheaper). And I keep designing them for the same reason :slight_smile:

2 Likes

Ryan skipped these in Thrifty Swerve and has EDIT: MACHINED DELRIN ON AL greased up and running directly on each other. It works much better in real life than my armchair engineering would have said looking at the paper design.
Now whether you consider anodizing plates an achievable in-team process step is a different question :stuck_out_tongue_winking_eye:

4 Likes

This is similar to the very successful (world championship winner worthy) design of Stryke Force Swerve that also includes a suspension. There is an aluminum top hat that runs inside of a larger plastic hub.

103 made a prototype of this setup and was impressed with how well it worked. With the generally short lifespan of robots I see future COTS designs utilizing similar tech, eventually with injection-molded plastic parts.

2 Likes

Actually Delrin running in anodized aluminum. I clarify this because folks will freak (including myself) if they think that aluminum is running on aluminum.

Ours is aluminum running on Onyx.

6 Likes

Sadly no :/. We had the gt2 3mm belts fail on the first stage of reduction. Some of the failures occurred on lighter than a fully loaded robot too (70 lb robot).

So in this exact application I would be totally hesitant to use that belt on the drive motor.

9 Likes

I think you can cross off or modify your list here though for some cost.

Motors - lot of teams run 4 motors anyways and would now do to 8.
Wheels can be crossed off since most bots have wheels to begin with.

1 Like

I don’t think knocking off the motors and wheels from the BOM is the right solution here. Just because “most” teams have these parts already, doesn’t mean every team does. There are also a number of teams that purchase new wheels and motors each year.

For the purposes of comparing swerve designs/solutions, I think having everything listed is the right way to go. Just about every swerve design out right now (especially when focused on price as the bottom line) lists everything on their BOM. Removing these key items from the BOM only lowers the price artificially IMO (if I can remove motors/wheels from this BOM, I can do it on every other one. Therfore the difference in price between modules doesn’t change).

From here, if a team decides they don’t need or already have certain components, they can adjust the BOM themselves.

3 Likes

15 Likes

Delrin is a magical material

3 Likes

Sorry, I don’t mean YOU should take the parts off the BOM (wrote my last response too fast)

I meant more for teams who are considering this.
A serve drive IS going to cost more than say a 6-wheel tank, but there’s a cost there that is already being paid for by most teams. OP said each serve will cost around $350, so $1400.
Aside from kit-bot teams, teams should figure out how much they’re already spending on gearboxes, ~4 motors/controllers, wheels, and other stuff.

Neos + controllers (OP’s items)
$125 x 4
Toughbox minis (cheaper end for gearboxes)
$86 x 2
4" highgrip wheels
$8 x 6

There’s $720 without bearings, chain/belts, and pullies/sprockets. I think our last build was about $900 for the drivetrain. So for another $500 we could have tried swerve.

4 Likes

Bingo - swerve is expensive. But so is a skid steer. More so if it shifts or runs 6 motors or other options.

The big downside to swerve for many teams is that there’s a threshold of minimum functionality that’s much higher than skid steer. A poorly done skid steer probably still works. A poorly done swerve less so. So there’s more risk.

So wish people would compare apples to apples for costs more often though.

8 Likes

Gotcha! In that case, I 100% agree. This is definitely something we would consider should we ever decide to go down the swerve rabbit hole. We make a WCD currently, so the price difference between swerve and our current setup is not $0 for tank and $1400 for swerve. It’s much closer as you point out.

1 Like

I agree with your suggestion. It’s important to compare apples with apples. $350 per module is also a good ways cheaper than the COTS modules, which go for ~$600 all in. But keep in mind, my numbers assume you don’t pay for any material or time to build in-house, which also isn’t very realistic.

6 Likes

This is good to know. We are using the 5mm pitch 9mm wide belts from WCP for our swerve that Brendan helped us design. Had no issues on a fully max weight robot during our off-season events.

3 Likes

Don’t also forget to include the cost of spares. With swerve, you pretty much want to have at least 1 (preferably 2) “extra” swerve modules fully assembled and ready to mount on the bot, including motors.

Of course, with say WCD in theory you should likewise have at least 1 full spare gearbox and other spare parts ready to go. In practice, this hasn’t been quite as big an issue as the simpler WCD or kit chassis is much less complex and in my experience less prone to damage/failure.

2 Likes

One reason to justify, is that when we try to make one and fail, we can really learn from our mistakes! I have no regrets when we learn from failure.