Mecanum vs. Swerve Drive

[DampRobot]

I guess it's my turn to be "that guy"...

I'd like to humbly propose that in 90% of cases, omnidirectional drives are unnecessary or even detrimental to the overall robot. Teams like 1640, 148, and 1717 notwithstanding, I often believe that a team would be much smarter to build a more complex manipulator and go with some sort of simple, tested tank drive (even the kit bot on steroids) than build a complex drivetrain and run out of time for a good manipulator.

Yes, swerve and mecanum drives are more maneuverable, and this can have advantages (as attested to on this thread). But that maneuverability requires a fair amount of driver practice to fully utilize. Be honest, how much of the time did the average mecanum robot spend strafing? Any?

Even if you are able to drive this type of drivetrain to its full potential, it takes a lot of time to build and program. The average swerve drive probably has three or more times the number of discrete non-COTS parts as a tank drive, so it would take a lot more time to make. It's not impossible, it's just a large investment of time, time which could be used making a better manipulator. Also, it takes a lot more time to program than you might suspect. Don't take my word for it, ask Adam from 973 or Ether (or just search for their posts about it). This goes for mecanum too, to a lesser degree.

I won't go into the reasons why a swerve or mecanum drive might be better or worse in terms of defense. Other people are better qualified to argue that than me.

Hopefully I haven't discouraged you. No matter what summer project you chose, if you can actually follow through and make it happen, it will be an awesome experience for you and your team. And, if you want to build an omnidirectional drivetrain in the offseason, go for it. The offseason's the right time for this sort of thing, and even if all you've learned that you're never building/programming/driving an omnidirectional drivetrain again, you should call it a success.

[vishnum]

One of the guys on our team has machined mecanum brakes which would prevent the rollers from spinning and in effect creating a tank drive. We haven't gotten around to actually testing them yet so I'm not sure if they work.

[LeelandS]

There are many other threads discussing this matter, the forum search feature will most likely satisfy all your curiosity. Here is the general consensus in a nutshell (with a few modifications).

Mecanums are fine. Don't let anyone tell you otherwise. Mecanums get a bad rep around here for some reason I can't fathom. But Mecanum drive is fine. If a robot on Mecanums fails to perform, it is often because either the rest of the robot didn't live up to the needs for the game, or the team didn't invest enough time into developing the drive system, not because the wheels are cursed. Mecanum does require a fair bit of programming and tuning work to get everything right, but when used properly, they can create a wonderful and incredibly frustrating (for the other teams) drive system. At FLR, one of the lead scoring robots in the field was on Mecanum. They evaded defense like it wasn't there and became the first overall pick in the draft, cut off in the Semi-finals by some very well placed strategic defense. No, you won't be able to push a robot with a strong drive train, but if you're using Mecanum and your strategy is to push, your issues run deeper than the robot.

Swerve is fantastic. It's maneuverable, and it doesn't sacrifice the traction that Mecanum wheels give up. It's a very fancy drive train, allowing for multiple driving styles (see 1640 for a notable example) that allow for versatility in different situations. The issue is, Swerve really is every bit as difficult as its reputation lends to. There are no cutting corners on a good swerve drive; it requires a full development cycle that can span through multiple seasons before being considered ready to put on the field, and even then, without a refinement in design and manufacturing resources (see the Swerve development of 1640 and 1625) it will be heavy and costly. In the long run, the pay-off of Swerve seems to be very good. I haven't seen it through yet, but if you look at the teams who have spent time refining a Swerve drive and continuing to improve upon it... Well, they are generally teams most people are unhappy going up against. Teams like 16 and 1717 have become dominant in FRC, partly because their drive system is just better than all else's. It's a great drive system, and its benefits aren't insignificant, but if you really wan't it to play out well, you need to be ready to invest countless man hours and dollars into the project.

I think that more or less sums it up. It's 1AM, so my head is a little loopy, but this is more or less a general take on Mecanum vs Swerve. Mecanum good, but hard. Swerve better, but really hard. It's a personal team thing if it's worth it or not.

[wireties]

We have used both mecanum and swerve. The mecanum is easier to build provided you pay attention to a few design rules. All 4 wheels need to be in the same plane thus excellent tolerances on the frame design and/or mounting the wheels with shocks/springs of some sort is essential. Keep the mass above each wheel about the same thus symmetrical distribution of robot components (their mass) is critical (don't forget the battery). Finally we had much better performance with encoders on each wheel. If you design the robot keeping these principals in mind the software is pretty simple.

We built mecanum because the students loved to drive it. You will hear many biased opinions about mecanum. But I would have to agree that when playing against bots at the Einstein level the disadvantages are more serious. At the regional level, mecanum is fine and it is fun. A good driver (with some practice) can move around or spin around all but the top tier tank drive setups.

Swerve drive is awesome but quite complicated. It requires more parts, tighter tolerances, more motors etc. Many others in this thread make good points about the complexity. But it is awesome when you get it working! We stopped making swerves when our expert machining mentor became ill and couldn't help with the manufacture.

Either is a worthy off-season effort. Good luck!

[SoftwareBug2.0]

Originally Posted by LeelandS

Mecanums get a bad rep around here for some reason I can't fathom.

Here's a reason why mecanum drives get picked on: It's hard to scout defensive potential well. And mecanum wheels are a proxy for being bad at defense.

At a regional, most of the teams that would make the best defenders are playing offense during the qualifications. And here's why I think that is: Suppose a team can score 25 pts if they go offense, or reduce the opposing alliance's score by 25%. So then they should play defense if and only if 25% of the opponent's expected score is >25 pts. Once you get to eliminations, there will be better opponents, so playing defense will make more sense.

So in figuring out who the best defenders are you end up doing a lot of guesswork. You can try to keep track of pushing matches won or lost but that's about it because so many don't play and defense during quals. So then you're left with pit scouting. Not all mecanum-drive robots are easy to push around, but if you're picking a robot to play defense (or counter-defense) then the scouts have to be sure that the robot you pick won't get pushed around by an average kitbot.

So from this you get teams, like mine, who have used a formula for picking the third robot that looks approximately like this:

1) Eliminate robots likely to be bad at pushing (such as mecanums or omnis)
2) Find max( [average auto pts scored] + [average climb pts scored] )

Mecanums are useful, and using them can improve your robot's performance. However, remember that the teams in eliminations aren't the 24 best teams. They're the top seeds plus whoever those teams think would give their alliance the best chance to win.

Now to the original question, mecanums or swerve: Why not holonomic?

[Koko Ed]

Originally Posted by LedLover96

We do have reason to believe that they we one of the reasons we failed to make eliminations at all in 2012, mostly because there is a stigma against them, and, if you are not first round pick, then there is a good chance you will be skipped second round.
!

This happened to us this year.
There are teams that have policies of not picking teams with Mechanums no matter how good they have performed that weekend.

[Clem1640]

Both swerve and mecanum are equally agile when executed well.

Mecanum is easier to execute well and requires fewer resources ($s, mass,...) and less learning or insitutional knowledge to do so. The rollers, however, do compromise traction.

Swerve provides agility without compromising traction, but it is resource intensive. 1640 uses swerve and we are very pleased with the results, but this has come at a considerable investment in learning. We also have to deal with having a finished drive-train rather late in the build season due to long fabrication and assembly times. Control software is also non-trivial. So there is a price to pay for this traction.

It is clear that the investment and cost of swerve has imposed limitations on other systems and capabilities on our robots.

So, like all real design decisions, this one is about what is right for you and your situation. All good designs are good compromises.

More serve info is available at the following link to "swerve central":
http://wiki.team1640.com/index.php?title=Swerve_Central

[Gdeaver]

If for a moment you do not focus on the competition value of complex drive trains, but look at the FIRST mission to inspire students, go for it. Our students learn so much about design and fabrication from the swerve project. It is a complex project and requires a team to develop a design build process. This same process can also serve a team well in all other design build missions. The knowledge gained from going for a swerve project can be huge. Our team recently discussed manufacturing and selling some of the critical parts of our swerve module. Teams would have to make and purchase the rest. There would have to be some interest to make a run. Interest.

[Taylor]

Whichever way you decide to go, don't try to reinvent the wheel (pun intended). There are many resources out there - crab/swerve COTS modules, gearboxes built into chassis, programming code shared from others - that make life so much easier for you.

[cmrnpizzo14]

Originally Posted by LeelandS

The issue is, Swerve really is every bit as difficult as its reputation lends to. There are no cutting corners on a good swerve drive; it requires a full development cycle that can span through multiple seasons before being considered ready to put on the field, and even then, without a refinement in design and manufacturing resources (see the Swerve development of 1640 and 1625) it will be heavy and costly.

I would assume 1126 can attest to that? I loved being on your alliance at FLR in 2012 but watching you try and drive up onto a bridge was at times quite painful.

[LeelandS]

Originally Posted by cmrnpizzo14

I would assume 1126 can attest to that? I loved being on your alliance at FLR in 2012 but watching you try and drive up onto a bridge was at times quite painful.

Exactly...

In 2012, when 1126 tried Swerve drive out, it was significantly less than stunning. The development cycle was much shorter than it should have been, spanning about the Fall prior to the 2012 season. The prototype, while more or less functional, was definitely not field ready. It would still suffer technical malfunctions and the software control was incomplete. This ended up hurting an otherwise decent robot, making consistent driving a challenge and, of course, those bridges were a hurdle (though that was also a result of the robot being top heavy).

While I presume the mileage may vary for most teams, I would probably target a satisfactory swerve drive at having a 2 year development time:
-Begin development over the summer following a given season
-Span that through the Summer and Fall (though development will probably slow down in the Fall as the team starts to focus on season preparation) -----Don't use it in the coming game
-Resume development in the following Summer
-Enter your prototype Swerve into an off-season event
-Refine based on off-season performance
-Determine if you are satisfied with putting that Swerve drive on the field. If not, repeat for another Summer.

That's just my ball park. Given the complexity of Swerve, I would probably not be comfortable with less than that. Higher resource or more well staffed teams who can iterate their physical design more quickly will probably have their development cycle cut down significant as opposed to a team such as mine who doesn't have significant resources.

Even after that initial development cycle to even get the Swerve on the field, it's important to keep iterating it. Reducing weight, reducing cost, augmenting reliability, simplicity and robustness are important as it allows your Swerve drive to keep a competitive edge. Equally as important as the design of the physical aspect, refinement of programming is just as important, to keep the Swerve drive driving properly.

Originally Posted by SoftwareBug2.0

Here's a reason why mecanum drives get picked on It's hard to scout defensive potential well. And mecanum wheels are a proxy for being bad at defense.

The only reason I can see not picking Mecanum for a third robot is if the team is trying to push with them. If a team builds a robot on Mecanum wheels and is pushing as part of their strategy, then I really wouldn't want them for a third robot. While most people equate "defense" to "pushing", to assume that pushing is the only way to play defense is a huge error. While yes, pushing is the most obvious way to play defense, a robot on Mecanum wheels with a creative driver and strategy can easily create traffic for any robot. While they won't be a mobile wall, it's still defense. It is more work on the scouting end, but it can be a difference maker. Especially since most teams at regionals who end of being picked as a defensive robot don't really have all-star drive trains.

Originally Posted by SoftwareBug2.0

Now to the original question, mecanums or swerve: Why not holonomic?

I've actually wanted to experiment with Holonomic for a while now. I think it's never really involved in these discussions because it's a fairly forgettable drive system. It doesn't really get any exposure because there are few FRC teams with Holonomic drive in the spotlight. I remember 1501 was on Kiwi Drive (the 3-wheel Holonomic variant) in 2010, and I remember seeing one team use it in 2011, but I don't think I've seen one since, though I'm sure there have been some.

[bbradf44]

Originally Posted by jbsmithtx

Prepare for the worst.

Nearly everyone on CD will tell you that mecanum wheels are completely awful, and have no pushing power.

I have to disagree with you there. This past season my team used mecanum for the first time. Yes they are very costly but it all panned out in competition. And as far as pushing power, we pushed a robot twice our size and just at Max weight across the field on several occasions

[apples000]

If you're considering swerve, here's some advice. Before you decide to go ahead, figure out some basic stuff about the drive train. Make a simple CAD drawing and figure out where the big stuff is going to go and how the modules will be steered and driven. Make sure you spend time to figure out how the control system will work with the swerve. You need to figure out how the steering feedback will work, and if you are doing independent steering/drive, you need to do some math(see Ether's swerve kinematics whitepaper). Finally, try to find a good machining mentor. It helps a TON to have a machinist help make the modules. If you get an experienced machinist to help out, you will avoid making silly mistakes, and you'll get your modules done faster, cheaper, and with stricter tolerances. But if swerve is something your team is really interested in, it will pay off HUGE! My old team attempted to develop a swerve drive during the build season, and we ended up in 45th place in our regional. We qualified for cmp through the chairman's award, refined our drive train during the extra 5 weeks, and once we had all of our problems solved, we ended up on Einstein that year!

[Nemo]

I disagree that programming a mecanum drive takes a significant amount of extra time.

Mecanum pseudo code:
M1 = x + y + z
M2 = x - y - z
M3 = x - y + z
M4 = x + y - z

6WD pseudo code:
M1 = y + x
M2 = y + x
M3 = y - x
M4 = y - x

Mecanum drives are nice and simple, both mechanically and in terms of software. I disagree with claims that in order to drive well, it needs suspension, closed loop control, or other extras.

[jbsmithtx]

Originally Posted by bbradf44

I have to disagree with you there. This past season my team used mecanum for the first time. Yes they are very costly but it all panned out in competition. And as far as pushing power, we pushed a robot twice our size and just at Max weight across the field on several occasions

We used them and loved them as well. I was simply stating the common CD belief that mecanum wheels are awful at defense. We too did the same thing, and held some of the best teams at our regionals to only 2 cycles. We likewise bullied teams pushing them across the field... I personally think mecanum wheels are a great and easy option, provided you build them right.