Strategic Uses of Swerve Drive

[Jared Russell]

Originally Posted by Lostmage333

This is a discussion that my team has had every season for the last 3 seasons. The result has always been to build the swerve drive.

Really, it depends on the game. 2007 was Team 469's first time building a swerve drive. Why? We determined that we had the resources, and thought it would be somewhat helpful. It was a big challenge to undertake. Our first swerve drive design had numerous problems. Therefore, we designed and machined a 2nd iteration. Was it better? Yes. Was it still flawed? Very. We used 2 chains per module. Tensioning was a pain, and issues with chains jumping were big throughout the season. However, in the end, it was still successful.

Pros: Omnidirectional drive
Cons: Chains jumping, Wheels not holding up well to load (remember the AM wheels from 07?), Motor heavy (4xCIM + 2xFP + 2xGlobe), Difficult to program.

In '08, we again had the discussion about doing swerve. Well, in the end, we decided to do it again. Again, we decided to go for the full swerve (independent power/steering). It was quite resource heavy. However, we learned from the chain fiasco from the previous year. Using gears and a more "simplistic" design, made essentially of 3 CNC'd pieces, most of the mechanical/breakdown issues went away.

This last year, we again decided to go with swerve because mobility was deemed to be extremely important. As the base driver, I can comfortably say that the swerve drive provided huge benefits this year, from both an offensive and defensive standpoint. Might it be just as useful next year? I don't know.

However, swerve shouldn't be one of those things that you do automatically. Its not absolutely necessary to win. If executed well, it can provide a huge advantage, but at what cost? A simple 4/6 Wheel Drive system would use less motors (no steering), and almost definitely weigh less. The conventional drive systems eliminate the need for drive calibration and stuff. Some great teams have never built a swerve drive, such as 1114. Others, such as 217, have built it once. From what I hear, 217's experience with swerve was so bad that it's likely that they will never do it again.

Well I hope that provides some answers to your questions.

Just a minor correction - 1114 used swerve in 2004, and 217 used swerve in both 2002 and 2003

[Chris is me]

This was kinda my thinking with it, that it would be a complicated weight and time commitment for not necessarily much benefit, but I'd never built one so I'd prefer to find out what people who'd done it before think. So thanks! (By the way, keep responding so I can hear more perspectives on it).

My team's been designing a swerve this summer, so these discussions come up a lot, and it'd be nice to have discussion on CD for reference from teams that know what they're doing.

[AndyB]

The drawbacks of swerve that I've experienced are:

• Harder to control from a driver's perspective (when compared to a regular tank steer)
• Generally heavier and require an extra motor for steering
• More complex to design and build.
• More expensive.
• Don't do ramps too well.
• Robot turning issues if you drive all modules together

If you are going to do a swerve drive, build one in the fall first. Get your practice in early. The hardest part in this process could possibly be to opt against a swerve after the game is announced. Don't do it just because you feel obligated.

[Josh Goodman]

A couple of things as far as implimentation go. As far as actually building the mechanism, dont try it out during build season. A local team decided to do that this year, and they spent most of their build season on the Mill building and rebuilding parts. Test building/programming a swerve makes a great summer/preseason activity. If you don't try it out first, you will run into many problems and, imo, not as good robot due to lack of time in build season.

So my suggestion is this. Get a group together and figure out how the things work. Build it, program it, get a full fledged working prototype (use the KoP materials even). Then when you see the game, evaluate your necessity/pros and cons. If the game screams out Crab Drive...DO IT! You now have the experience and capability. If you think "Well....we don't really need it, but it would be cool to have", it's all up to your team, but I would say save the weight for the other mechanisms. As many people have said before, there are tons of drivetrain styles out there with less weight, time and cost.

You want something in between? Try Mechanum Wheels.....they're pretty awesome too.

[Ken Patton]

I think Josh's advice to you is pretty good.

My opinion (and while we have not done swerve in any of our 13 years so far, we did do active steering on the drive in 2 of those years) is that it is very game-specific. Some games - and the mechanisms that go with 'em - might lend themselves to a swerve design.

One example of that was Wildstang's 2003 bot which could get on the top of the ramp and then move sideways to block their opponent. Another example was Wildstang 2005 which could do a "drive-by" pickup of tetrahedrons from the side of the field.

The point is that both of those functions were highly game- or field-dependent. And the people who swerved could think of ways to use it to their advantage. Maybe even simplify the gathering/scoring mechanisms too.

The other side of the coin are the teams that don't put the complexity in the drive, go with a simpler drive, but figure out how to re-orient the robot to acquire objects and then score them. It might take more practice for them to get great at it, but then they might get more practice if they are up and running sooner. Again it'll be game dependent.

I admire the guts and talent - and the drivers! - of those teams that do swerve systems...

Ken

[JVN]

Originally Posted by Ken Patton

Another example was Wildstang 2005 which could do a "drive-by" pickup of tetrahedrons from the side of the field.

I find it hilarious that one of your examples of a "mechanism that lends itself to a swerve drive" was on one of the only Wildstang robots that didn't have one...

Maybe Raul didn't get the memo?

Rather than believe that to be possible, my theory is that Raul just wanted to make his programmers parallel park on a vision tetra in autonomous mode to pick it up, then parallel park on a goal to score it.

-John

[Ken Patton]

Originally Posted by JVN

one of the only Wildstang robots that didn't have one...

ohhh, man, do I feel like a dummy... I assumed it was a swerve. Bad idea. Sorry.

IF that bot had had a swerve, it woulda been awesome. I mean, more awesome. Yeah, thats it.

Thanks for setting me straight, John

Ken

[EricH]

Not quite a swerve, but 330 once decided to experiment with mecanum wheels for omni-directional motion. At that point, we'd had a couple years of practice with VEX-scale wheels that we'd had built after seeing mecanums used to move airline cargo containers around, but none full-sized.

Well, we built our set of mecanums and designed the 2005 robot to accomodate a 6WD or a mecanum drive, obtained a couple more kitbot trannies, and set up a mecanum drive on the kitbot and a 6WD on the competition robot. We did some testing, like putting a tetra on a pole and attaching to the kitbot (note: it did a nice circle while going sideways). But what killed that drive was the Defense test: two goals, regulation distance apart, and our 4WD 2003 robot. Objective: get mecanum-bot through the gap past 2003's defense. Result? No success. The 6WD stayed in place, and we did pretty well with it. We haven't had another omni-directional robot, until 2009, where the drivebase rotated under the robot that was held in one orientation by the trailer. (Results: not exactly what was hoped for.)

It's really game-dependant, and team-dependant. A team with swerve experience may choose a swerve when a non-swerve may be better, because they figure they can get it to work--then they face a team that has a non-swerve that beats the wheels off of them. It may also go the other way, but that doesn't happen often.

Chris, the guy who said that "if a team knows how to do a swerve, they shouldn't opt out of it" doesn't quite understand that while a swerve is the best combination of pushing, speed, and maneuverability that is currently available, it is at best a compromise, and many times compromise won't work quite like you think it will. If a team knows how to do a swerve, they know how to do a swerve should they decide that a swerve is necessary.

[Akash Rastogi]

Also, since Eric mentioned mecanums, mecanums are sometimes great alternatives (depending on the game) for teams who don't want to opt for a swerve.don't have the best resources available. For example, since 08 didn't really require "defense" per say, we tried out mecanums for the first time. I personally really liked the features that you get for little effort compared to swerve. But then again, there are the benefits of swerve that you get along with the complexity/cost.

Swerve > Mecanum > Omni wheels/ Kiwi drive

Would this statement be fairly accurate? You get at least some footing with mecanums over omni wheels, and then swerves trump mecanums in defense and maneuverability. If teams want a decent middle ground, would you say that mecanum (with proper suspension) is the way to go?

[Chris is me]

After doing holonomic in 2008, I don't think I / my team would really be willing to play with a mecanum drive or any traction compromising drive in FRC, barring another Overdrive type game where defense is (mostly) limited and tight navigation can be helpful. From what I've seen the traction benefits to a mecanum drive don't result in pushing power or effective defense, especially if not being pushed head-on. At least mecanum drives wouldn't have the "drift" around corners that made our drive unstable, but basically the team consensus (and the reason to build a swerve in the first place) is that we thought it was the only viable way to build a drive if omnidirectional movement is required. Perhaps I'm wrong though...

[Jared Russell]

We used a holonomic omniwheel based drive in both 2007 and 2008, and I cannot foresee us doing it again unless defense is explicitly forbidden ala 2008 (even though we managed to win a regional in both of those years). Simply put, the gain in mobility did not offset the loss in traction. Determined defense from a skid steer bot can still shut you down.

As far as omni vs mechanum, the dynamics are identical. The only benefits of mechanum are that it is often easier to mount the wheels and motors parallel to the frame. Also, it is somewhat easier to make/buy mechanum wheels with larger floor contact patches and better tread material (though I have seen omniwheel with roughtop tread rollers before).

I should also emphasize that using omni/mechanum drives effectively is still a large challenge. Wheel speed must be closely controlled, a suspension is highly desirable, and weight must be distributed carefully.

[Alan Anderson]

Originally Posted by Jared341

As far as omni vs mechanum, the dynamics are identical. The only benefits of mechanum are that it is often easier to mount the wheels and motors parallel to the frame.

With perfectly free-spinning rollers, this is true. However, with suitable attention to the tightness of the roller axles, one can in theory "tune" a mecanum system for better forward traction at the expense of sideways traction.

[JVN]

JVN's Swerve Drive manifesto...
I would only build a Swerve Drive if all of the following are true:

1. The game is such that a swerve drive would provide a DISTINCT advantage.
2. The team can execute the design, fabrication, and execution (programming) of the swerve drive without removing resources from other areas of the team/robot which would BETTER benefit the team.
3. The driver is capable of handling the swerve such that he/she can achieve a DISTINCT advantage in match play.

Some teams also may just decide to build one for "cool factor". More power to them, but that's now how 148 rolls.

-John

[Rich Kressly]

Originally Posted by JVN

JVN's Swerve Drive manifesto...
I would only build a Swerve Drive if all of the following are true:

1. The game is such that a swerve drive would provide a DISTINCT advantage.
2. The team can execute the design, fabrication, and execution (programming) of the swerve drive without removing resources from other areas of the team/robot which would BETTER benefit the team.
3. The driver is capable of handling the swerve such that he/she can achieve a DISTINCT advantage in match play.

Some teams also may just decide to build one for "cool factor". More power to them, but that's now how 148 rolls.
-John

Swerve Drive Manifesto, my foot! Seems a lot like a strong Life Manifesto. "Never over complicate things and know thine own skills and resources well." I'm big on trying to make sure the machines and people get along well; in a perfect world enhancing each others abilities.

Thanks for including the higher level thinking here JVN. I'm a fan of the way you guys roll.

[Al Skierkiewicz]

Chris you should have come and talked to me at IRI. We do not choose crab drive before the game is announced and we proto-play the game. In our tests, we took one of our old crab drive robots and attached a trailer to it. It was clear after our trials that no one was considering the shift in center of rotation that the weight of the trailer introduced. Crab allowed a more controlled turn with the trailer attached which was deemed better for defense and ball pick up. We later found with practice that crab could be used effectively in scoring by wrapping the robot/trailer around an opponent while our alliance partner scored. It was easy to block two quadrants of movement on an opponent.
Let me say again...We do not choose crab or any function for the robot until we play the game and see what is needed. The Whole team (except for field build) brainstorms from the minute the satellite transmission goes to the First logo and we spend days at it. We have groups of people play the game as a robot to see what the interaction is and what affects scoring.
There are disadvantages to crab. The modules take a lot of space within the base. We have managed to get ours as small as can be so that they can be placed near the edges of the robot. Still the wheels will never be within three inches of the outside edge due to the rotation. Crab uses more motors, we have in the past used CIM in all four wheels, or two CIM and two FP or two FP and two drill motors. We usually use two Globe for steering but software needs to know where the steering is at so a special mount is required for feedback and position sensing. We never allow rotation much beyond 360 degrees so mechanical stops are needed as a backup to position sensing. All the hardware for construction of the crab eats into the weight budget. Crab only puts four wheels on the floor when more might be better.
And above all it requires practice, practice, practice. The first step is the realization for the driver that all of their life experience is out the window. They know how to control a bike, a big wheels or a car by steering. Crab now gives them the ability to move like they do in a hallway at school. Translating that motion takes practice, and a fair amount of software help to integrate the hand controller with robot function.