Fallacy or Fact - The Cost of Swerve

This would be a great title for a white paper, but lets just start with a discussion :stuck_out_tongue_winking_eye:

I’d argue that the true cost of swerve may be expensive, but for teams that are at least a couple years old, and have access to a proper machine shop and/or a metal working sponsor. The cost is actually quite subsidized to the point where it is just as affordable as mecanum wheels.

For example if you look at all the non machinable parts of the 221 revolution. (Bearings, Wheel, Gears, Thrust Washer, etc) You get a cost of 60 dollars per module which is only $10 more than the inferior mecanum wheels and all centrally purchasable from VEX. Arguably everything else your team most likely has already or can machine it.

Now I understand on the BoM cost is expensive, but coming from a 9 year old team who has horded pretty much everything that has come into our shop. The actual cost to produce a swerve module is much less.

I’m on one of those older teams that doesn’t have NC machining and has never built a swerve.

I’m afraid the cost of swerve is more than dollars. You really have to develop it before the season. It will take much time–your whole summer. If you waited until build season, you wouldn’t have time to build the rest of the robot. Also, to commit to swerve on a robot before you’ve ever done one would be too risky.

Getting the CAD right is just the first step. The money-price–comparing it to mecanum–isn’t really fair–a good swerve is so much more capable.

As the mechanical engineer on our team, I’m sure I could help the students to design a swerve, but to get the programming to work is another question.
Just to understand it to begin programming is mind numbing. 27 modes of driving?.

And I’ve already posted that I think we’re expecting large floor obstacles next year. Not ideal for swerve.

Don’t forget to add the cost of extra controllers and sensors. 4x extra controllers would be ~$200-400 and 4x extra encoders would be ~$100-200. You might be able to get away with not having encoders on the drive part of the swerve.

I would agree that it could be a lot cheaper. However I think it does make sense to go with a COTS swerve module at least in the beginning to find out how everything works. Having a thrust bearing on the steering would never have struck me (maybe that’s just because I’m a programmer). Plus a lot of the teams that are 9 years old can afford to spend a little extra on something like a COTS swerve module. We might do something in the offseason, as well.

And yeah, you forgot absolute encoders which are like 30-40$ per module (there’s the programmer coming out).

Also hello, my team number is 1 more than yours :D. Maybe we could talk since we are geographically pretty close.

Yea i’m surprised a 3081-3082 feud and or partnership has not undertaken at 10k lakes. We also are always geographically close in the pits.

Like I said I would argue that a team that has been around long enough that has focused on control systems well enough to attempt a non COTS swerve has encoders to use just as they have the motors and other COTS that go into the swerve.

p.s. Im a mechanical guy, and I completely spaced out about thrust bearings in swerve mods until I picked apart some CAD drawings

p.p.s Encoders are not included in COTS swerve mods. From what I understand most COTS swerve mods are just the mechanical assemblies.

COTS is good for the first year, especially the Revolution Pro; other COTS modules aren’t as good IMO. Plus, it’s easy to go from offseason to competition with a COTS module, so it’s a direct transfer of code.
That being said, any really good swerve IMO will be custom, and that takes a lot of effort. 192 made a brilliant 3D printed swerve, but they used a high-end printer. All of the good custom modules I’ve seen are made on a CNC.

AMT encoders are my favorites. $20 virtually indestructible incremental encoders with index is hard to beat. Good for module rotation (with a magnet + hall effect for zeroing) and speed sensing if you want it. IIRC 1323 doesn’t run speed sensors on their swerve, which I think is fairly ok.

EDIT: The biggest cost is indeed time, which is why a better COTS swerve can be nice. If a team goes in with the right planning and code, however, they can get a COTS swerve finished almost as fast as a WCD.

The main cost of swerve isn’t the financial cost of building one. It’s the time and energy investment into making it work effectively, and the time and energy that was not invested in other parts of the robot as a result.

The dollar cost of swerve has never been a deterrent for me… It’s all about the cost in:

  • Risk of failure
  • SWAP (Size, Weight and Power) resources
  • Time (before the season, during build season, and during competition season / at events)
  • Machining resources

Chris hit the nail on the head:

The AM swerve does have one, but it’s before the gear reduction so it’s not very effective but usable. I guess your point does still stand.

I know it might be hard for high school students or some teachers to understand, but my own time per day is worth well more than a few hundred dollars (to me, and to others).

Not to mention the added cost of complexity. More things can go wrong and do, and often times when your drive-train isn’t working (for whatever reason) you can’t do anything in the match.

My team fabricated their own mecanum wheels in 2006, and we’ve used them again in 2010 and 2012. I can only recall 1 match that we’ve sat part of a it due to the drive train (in 2006, which was the battery coming loose, letting pneumatic pressure out of the robot which caused our wheel modules to fold up into the robot - really nothing to do with the wheels).

We used swerve in 2011, and we sat still because of drive-train issues for easily 4-5 matches.

yikes. This makes it more scary.

It’s an interesting technical and programming challenge that, if executed (and driven) correctly, can make a robot stand out. If you want to have a chance to do anything at all, you will have to make a robot in the offseason that works really well.

Oh hey its me.
I’ve been messing around with swerve drive designs in my free time inspired by 2767 and 2468’s VersaSwerve, but it’s nothing 5254 will ever run, for example. We value the bulletproof and simple traction drivetrain over something more complex that might give us an advantage, but take time away from our scoring mechanisms. And as it turned out, this year we needed plenty of time on our scoring mechanisms until we were happy with them (we’re still iterating :rolleyes: )

I do think it would be interesting to see how much of the cost of swerve is an up-front cost, and how much your expenses would decline as you use it in consecutive years.

Yeah. Swerve looks and sounds cool in theory, but without extensive practice the driver ends up spending more time trying to figure out which way the robot is going rather than just driving.

KING TeC tried it on one robot but dropped it for more standard drive train on future ones.

If you have field-centric swerve drive it is remarkably intuitive to drive, probably more so than tank is. It is also really fun, so there’s that.

Shh don’t tell my team the real reason I want swerve…

The biggest drawback of doing a swerve is purely time. Ideally we just move to a cots solution vs making our own modules next year. I believe we spent 2 weeks or so machining and making all 9 modules, maybe it was 3 weeks…

Link to Swerve Drive

You think that’s scary? In my first year of FRC, 2220 ran a custom swerve (having built an off-season swerve too), and literally didn’t move on the field until our last match on Friday at our second regional. The cost of swerve can be any chance at competitive success if you don’t execute well.

Again, Chris is really right here-- you can get the machining donated, you can probably get the parts donated or covered by sponsors, but that isn’t where the real cost is at.

So I’ve only seen one robot with it, but I wonder if an Ackerman steering (car steering) set up could provide me with alot of the benefit of swerve as a way to move into it slowly.

Lots of people drive R/C cars and they can be amazing.

First I would just design one swerve module and leave the wheel off, put it in the middle of the robot with two car-type tie rods to steer wheels at the corners on kingpins. Driver could start driving, get used to it. I could do this for the back wheels next; rear steering.

Then when I’m happy with that, I can switch out the front wheels with kingpins to swerve modules and we’d already have the main driving mode ( I guess) down. Finally, switch the rear for full swerve.

It would be a step by step mechanically, but maybe not help software as much.

Useful idea?