Why do high-tier teams not use Swerve Drive

I have seem many posts, articles, and people, who say that swerve drive is the ultimate drive train as far as benefit to the robot. It has speed, maneuverability, and defense capabilities. It this is so then why have the top teams not use it, but instead used tank?

Although it is true most teams do go with a tank drive even at the championship level, there are many notable teams that used serve this year. The two most notable ones in my opinion were 33 and 2767.

1 Like

A tank (or West Coast Drive) is usually easier mechanically and programmatically, allowing the team to focus on improving manipulator/scoring mechanism design. Swerve drives are also often (not always) heavier, taking precious pounds away from the rest of the robot.

Add 1323, 1533, 2910, and 3707 among others.

1 Like

Swerve drive is very complex and prone to breakage. If your drivetrain doesn’t work, you will most likely lose your matches. Why risk a higher chance of breakage by using swerve when tank drive is good enough?

The other “problem” with swerve is that it isn’t always necessary. For some games, like this year’s game, the extra costs (both time and money) of swerve drive don’t provide enough benefit over other drive trains.

It could come down to weight, time spent on programming, motors, and personal preference.

Swerve is a beast when done well (16 and 2767 come to mind). However attempting swerve for the first time is quite challenging and could potentially delay your scoring mechanisms. If you need to strafe, mecanum and H-drive are much easier to build and program. They will also give around the same level of performance as an average swerve drive.

As for weight, the AndyMark swerve module weighs 9.4lbs. That’s ~38lbs for just your gearboxes/wheels. Compared to a standard CIMple box with 4 CIMS that weighs ~18lbs, it could be hard to justify trading ~20lbs for swerve. If you want to make your own lighter swerve drive, you lose a lot of engineering resources that could be spent into other mechanisms.

Another big concern is your drivers. Do they have experience with swerve? You could have the best swerve in the world, but if your drivers aren’t experienced with it then it won’t do you much use on the field.

The overall biggest issue with swerve is the learning curve required. Teams need to decide to go with swerve and stick with it for a few years to make it worthwhile. Most higher-level teams wouldn’t want to make that kind of commitment. There’s a lot of risk trying out a new system.

1 Like

Reliability, motor usage, and weight, are three significant factors I think.

Most high-tier teams could probably build a very reliable swerve, but the risk of losing a random match due to an unforeseen problem (stalling motors, broken gears, etc) could very well place them in 2nd seed or lose them a playoffs match at the level of play that they must maintain.

Modern swerves generally utilize 8 motors for direction and speed, half of the slots available in a PDP. This removes the motors you have available for other systems, and further increases weight.

With weight as a finite resource that also negatively impacts driving characteristics, it is beneficial to both leave weight for other systems and keep weight as minimal as possible. Sheet metal and lightened West-Coast drives are capable of weighing under 30 lbs, however the lightest swerve I’ve personally seen, 33 Killer Bees’ from this year (which was 3D printed from nylon and only used 6 motors), I believe still weighed greater than 30lbs with the supporting infrastructure in between the modules (but don’t quote me on that).

1323 and 2767 showed how amazing swerve can be. 254, 148, and many others showed how swerve isn’t necessary.

3 Likes

I’ll leave this here:

1 Like

I have seen similar things.

Whenever you say something is the “best”, you are always making some assumptions. The majority of the assumptions for swerve are mentioned by others in this thread.

When deciding if it’s “best for your team”, simply take into account all the assumptions you can find. Is it valid to make those assumptions in your case? If so, yes, it’s best. If not, you’ve got some more brainstorming to do.

High tier teams that use swerve drive (and there are many) do so because they are already good at it, and they know that part of being high tier is doing things you are good at until you are great at it. These things include fielding a reliable robot, knowing how to move it quickly and efficiently, getting lots of practice with it, and having lots of time to develop scoring mechanisms to go on top of the drive base.

High tier teams that don’t use swerve have precisely the same criteria.

16, 2767, 1323, etc. all keep doing swerve because they have gone from good to great while developing swerve.

254, 1678, 1619, etc. all keep doing WCD because they have gone from good to great while developing WCD.

Teams like 33, 2451, and 118 have been chimeras and don’t fit my model.

HOT (Team 67) gave a seminar at MSC this year about robot design. They pointed out the 3 most important things of a robot is the drivetrain, the drivetrain and the drivetrain. If your robot cant move, you cant score points (ignore some robots 2015). Swerve tends to be more complicated, making it more prone to failure, and meaning more design time. Dedicating resources to a very complicated and resource requiring drivetrain is hard to justify for a lot of people when a WCD can work just fine, and be competitive at the same time.

Serve also takes a lot of time for practice. 2767 makes driving swerve look easy but I cant even imagine how many hours their driver has on that robot practicing.

We have now completed our 4th competition season using swerve. We have seen drastic improvements in each of the areas you cite:

  1. Reliability - this year’s design has been remarkably reliable. We did not have any swerve breakdowns. We did build 2 spare modules so that we were ready in the event of a breakdown, but we did not need them. I credit this to a couple of improvements we made this year. We switched to Colson wheels which have had much better tread life than the wheels we used in the past. We switched to mini CIMs for the drive motors instead of the 775s that we used last year that caused most of our issues. We paid attention to the workmanship details like grease in the gearboxes and solder on the encoders.

  2. Motor count - OK, not much you can do there.

  3. Weight - our module this year was 6.8 lbs, or 27.2 lbs for a set of 4. This is certainly competitive with most other drives.

https://www.chiefdelphi.com/media/photos/46428?

In one word: Resources

Swerve is a lot of parts.
Swerve is a lot of money.
Swerve is a lot of motors.
Swerve is a lot of weight.
Swerve is a lot of packaging space.
Swerve is a lot of development.
Swerve is a lot of code.
Swerve is a lot of driver training.
Swerve is a lot of …resources.
Swerve is often a lot of compromises.

Teams that have gotten through all that work tend to then make minor incremental improvements on their system.
Most teams that are unsuccessful with Swerve are able to do the hardware portion by either themselves, or buying the parts. Most fail at the controls and thus driver practice portions.

All this being said, that is why 2767 is made with magic. Swerve, and an robust elevators and roller claw and climber and ramp systems for something on the order of 106 lbs… They truly made their robot from magic.

They also went through a lot of years with their swerve drive where they were “under the radar” because their other systems were less effective. While they may have just shown up on most picklists the last couple of years, they have had a steady climb with years of experience and development. 111, 16, Aren Hill… same story. Lots of development, and lots of experience lead to some amazing systems.

1 Like

We can bang out a rock-solid WCD in our sleep, then put 100% of our effort into the scoring functions, software, practice, etc.

8 Likes

If you think those “high tier” teams haven’t dabbled with it…

1114 - Swerved in 03. I know they’ve played with it since
67 - swerved in 05, I happen to have a very good source saying they are interested in experimenting with it again
111 - One of the OG kings of swerve…
118 - uh, 05-08 called, they want their week 3 release videos back
148 - 2008 was a thing…
33 - has now run swerve multiple times (05, 09, 17)

I’m sure 254 has toyed with it, if there’s ever a game that really calls for it maybe they’ll run it.

We used it for a very specific situation. Hated the entire experience. If we rebuilt that robot with the same design-goals we would not do swerve on it again. Remember: no one hates Tumbleweed as much as we hate Tumbleweed.

Would 148 use swerve in the future? It would need to be a game with VERY specialized tasks which require it.

No one should EVER use 148 as a reason to build a swerve drive.
148 says: Do NOT build a swerve drive.

8 Likes

My point was, you tried it… The thing I’ve noticed about the teams consistently winning is they are further along the design curve than the rest of us.

BUT, being as I have you here and you’re uniquely qualified to answer - why WONT you do it again? What was so terrible about it?

Jared hit it right on the nose.

Also - more failure modes, more complexity/weight, and typically it is more of a distraction than benefit.

1 Like

Then follow up - how would you have achieved the design goals of Tumbleweed now?

Then follow up to that - do you feel the design goals of Tumbleweed were the correct design goals given the game constraints?

Apologies for questions, but trying to get a bit more content in here than “going sideways is worthless lol” and maybe get some real engineering design…

Parts - yes and no. You can certainly make swerve a lot more complicated than it needs to be. When I compare our latest design to a tank drive with a shifting gearbox, and encoders it is not really that many more parts. But if you compared it to a basic KOP tank drive, then yes, there are more parts.

Money - If you have access to a waterjet cutting facility, then the amount of money you need to invest in the COTs parts to build a swerve drive is not that bad, especially when you compare it to a shifting tank drive.

Motors - OK, you need to allocate 8 motors to swerve which is typically more than you would use on a tank drive.

Weight - marginally so. We are at 6.8 lb per module this year without anything exotic. I’ve seen some pretty beefy gearboxes and chains being used for tank drives.

Packaging space - this depends on how you look at it. Our setup this year is a vertical module. It is tall, but the footprint is small which allows it to be packaged in the corner of the robot leaving the entire middle of the robot open. We used a similar tall configuration last year. In 2016, we used a configuration where the motors were mounted next to the module which increased the footprint a bit but with low height. So, you can package the modules differently depending on the needs of the rest of the robot. You do need to plan on using the corners of the robot for the modules, but there is a decent amount of packaging flexibility if you need it.

Development - there is no denying that you need to develop swerve before you try to use it in competition. We always tell people that when they express an interest in swerve. But once you have it developed, it is relatively easy to adapt and evolve the design year after year. We were able to modify our design this year to use Colson wheels and adapt to versaframe with minimal effort. I think we had the swerve parts sent to be waterjet cut within a few days of kickoff. We are already looking at modifying our current modules for lunacy wheels. :ahh:

Code - Well, sort of. We are able to use the same basic code year after year and just change the x-y positions of the 4 wheels. Again, once you have developed it, it is not that hard to implement each year. The vector math inside the code is not all that complicated. If you think about the problem and solve it on paper before putting it in code, you can write a pretty efficient code.

Driver training - not with field orient! If you use field orient, then you don’t have to learn to think about steering backward when the robot is coming toward you versus when it is moving away from you. It does require a different type of thinking about driving. You don’t put yourself in the frame of reference of the robot like you are trying to drive a vehicle. Instead you put yourself in the frame of reference of the game. Most drivers find this easier to visualize. It also opens up a lot more possibilities of movements. This is where the real driver learning comes in - learning how and when to strafe, juke, spin or translate. When you take it out of field orient, then you need to orient your mind to the robot. But the controls are still intuitive (one joystick for translation with up = forward and the other joystick for rotation with left = turn left).

Resources - again, once you have gone through the development process, the resources required to implement swerve each year is not extensive.

Compromises - well, everything is a compromise. If you admit that swerve allows for a wider range of movements than tank drive, then a tank drive is a compromise where you are giving up those other degrees of freedom in exchange for a simpler drive. Like Woodie said - we have now seen thousands of different solutions to the same problem. Each one has a different set of compromises.