How do we know we're ready to swerve this year?

So in the 2016 off-season we built a swerve-drive, and I told my team that if it were 100% perfect by the end of December, we could put it on the list of possible drive trains for the year.

By the end of December, it was largely, but not entirely, there–they had missed the target spec.

When the Steamworks game reveal showed a flat field with little pegs and gears you had to put on them, and little wiffle balls to put into a boiler, they insisted that we use the swerve drive. I overruled them–they hadn’t made spec–and my mentors threatened to walk out if I held to that overruling, so against my judgment and warning we used the swerve drive…and had the single worst season of FRC that 1551 has ever endured, worse even than our rookie year, worse than our second year where we barely moved but had a sick end game. We made every alliance we were on worse.

That exact same swerve drive probably would have been mostly fine for Power Up, because a 4’ platform is a much, much more forgiving target than a tiny little spring on a peg. Probably. But we went 9-2 in quals in Power Up with a kitbot tank drive.

Swerve is, essentially, a very expensive way of punching the KISS principle in the face. And sometimes, maybe, that’s necessary–but a lot of the time it simply isn’t.

My advice is as follows: if you are unsure enough to have to ask here, don’t risk your entire season on doing it just because you have it. My follow-up advice is: possibly reevaluate when you see the game…if line-up is more like Steamworks and less like Power Up, lean harder toward “maybe in 2021,” while if it’s more like Power Up than Steamworks, maybe weight the possibility of swerve just a little bit more (but not too much more).

5 Likes

Why was this the case?

Because when we needed to go forward eighteen inches and to the right three to place or pick up a gear, we’d go forward eighteen-ish-but-not-really inches and to the right three-ish-but-not-really inches and then fail to pick up or place a gear. Over and over and over and over and over again.

At Buckeye we pinned the wheels in place, put stickers on the rear ones to make them slick, and ran it as a tank. It was terrible, but better.

1 Like

I would tend to fall under the category of “If you have to ask if you’re ready, you’re probably not ready.”

That being said, my time in FRC both as a student and a mentor has taught me that head strong and stubborn students can accomplish a lot of what they set their minds on.
As a student, I had started a sorta-Senior capstone / sorta personal project to machine Swerve in house based on the Revolution Swerve modules, and made it about as far as having one module fully machined. My team never actually used swerve until 2016, when we bought (well, two sets) of AndyMark Swerve n Steer and have been using them ever since.

It’s weird to say it, but I’d say Swerve is about as accessible as it’s ever been, with the plethora of COTS options (Swerve n Steer, 2910 Modules, etc.) and the variety of programming libraries for it. If you have a set up and running, and you think you can reasonably service it at a competition, (Like for instance, what happens if you have an entire module fall off the robot? We’ve had that happen before. What happens if a student drops the robot on a corner and shatters an encoder? We’ve also had that happen before.) I’d say add it to the list of drivetrain considerations for the next season.

Do not, however, go into the 2020 season dead set on using your hard built swerve. It may be an option, but it is never the only option.

2 Likes

@Sweeeeeeerve We were in a similar situation 2 years ago. During the summer and fall of 2017, 2910 built our first swerve robot. We got it working well enough that we thought it would be viable for the 2018 season. There were small hiccups during the season, but overall it worked out very well for us. No one on our team had experience with swerve prior to the 2018 preseason. If you have the right group of people it’s definitely possible to have a great first season with swerve.

Proven COTS modules definitely derisks it. A well driven swerve is a lot of fun and will get some attention. Since you already have it working, if you feel like you have the capacity and resources and think it will provide a competitive advantage for next years game I would say go for it.

Disclosure: I am associated with Swerve Drive Specialties

What would you mean by small hiccup?

Editing your phrase to make it a universal truth for all of software engineering ever.

You’ve got a few months to dial it in and you don’t even have to put your robot in a bag! I’d go for it.

This is good advice! Unless you don’t want to have fun.

3 Likes

From our point of view, it is acknowledging that our team could no longer compete at the desired level, nor advance our program without swerve drive. Competition has advanced such that FRC matches are now won or lost in a matter of seconds.

Time spent jockeying a robot on tank drive into position to score a game element is far greater than the time a swerve drive bot consumes gliding into position, already aligned, to score and then clears the space moving onto the next game piece.

Mechanum wheels are a partial answer to this, but you give up so much in defense and pushing power. The fastest mechanum bot is at best 2/3 as fast and far less assertive versus the other bots on the field.

You’re ready when, because of cumulative experience, you are tired of knowing your beat (baring a dead bot on the opposite alliance) before the start of the of the next match.

To actually answer your questions on the thread :).

There’s an inherent assumption in UnofficialForth’s statement: Robot performance is the key decision-making measure of success. This is probably true for most teams, but not all.

The bottom line, as I see it, is that the first year of doing anything involves risk. Building anything you haven’t built and tested before involves risk. By making it as similar as possible to things you’ve already done, or making calculated, precise changes, you can mitigate risk.

But, my honest question, is risk mitigation something your team wants? Or do you want to do something wild and crazy, learn from it, and say “It doesn’t matter if we win, we learned something?”

Choosing robot performance, or choosing learning outcomes - both are valid ways to prioritize design choices. I know what I’d choose, but I’m not on your team. If you like to bite off risks and don’t care if the result comes out sketchy, then go for it!

On the other hand, if you want to ensure good robot performance, take it slow and methodical, and prove to yourselves you can do it, before you put it “into production” in a real match. In this case, I would say “proven” means exactly what UnofficialForth mentions - can it outperform your other choices? This is where you get creative on setting up representative match scenarios, running time trials, etc.

Keep in mind that doing the validation to the nth degree involves knowing exactly what the 2020 game is, and would probably require a good chunk of time with two fully built drivetrains to do back-to-back trials… AKA, you shouldn’t expect to get a perfect answer either now, nor within the first week or two of build season. AKA, be willing to make some educated guesses.

All this being said. s_forbes is probably right. You’ve got some proven success already, and a bit of extra time this year. Assuming you’ve got the discipline and are willing to push a bit, and everyone keeps the mindset of “this is new, we’ll have to work at it”, this sounds like a good year to give it a shot.

Edit: Unless 2020 ends up being a water game. Then don’t do swerve.

2 Likes

Don’t do swerve unless you can pull off a tank drive plus a turret.

If you’re reading this thread to figure out if you’re ready: you’re not. If you’re ready, you’ll know.

Has there ever been a team in FRC history that had everything they need to win a world championship, and lost it due to having “only” a tank drive?

Go with the choice that has a far smaller chance of ruining your season.

9 Likes

If swerve is done correctly, AND tank drive isn’t done correctly, then you’re correct.

Unfortunately, for 90% of the teams out there, the odds of the first half of that assumption being correct are actually negative. For the second half to be correct, probably about 20%, with another 30% just needing more driver practice because they spent 7 weeks building and 1 week programming.

For the tank-drive robots with good drivers, they don’t jockey in. They’re in and out. Some even do drivebys if their mechanisms allow it.

For swerve robots that don’t have good drivers, they… uh… don’t… glide… ANYWHERE. They skitter and dodge somewhat controllably.

Personally, I think you need to head for Silicon Valley, and watch 254’s tank drive demonstrate gliding into position already aligned to score and then clearing the space. “254” and “swerve drive” in the same sentence only works if “not” or “never” is also included.

While I 110% agree with the rest of your post, you must not have heard the news.

3 Likes

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.