Mecanum vs. Swerve Drive

[Madison]

I think it might be an interesting exercise for people to share video of their idea of a good mecanum drive. I suspect most FRC participants have never seen an above-average implementation of mecanum drive in action.

A mecanum drive is, in some ways, a lot like fly-by-wire controls in an aircraft. It requires closed-loop software assistance or it doesn't really work at all.

[Siri]

Originally Posted by Madison

The conventional wisdom has been that implementing mecanum has become easy. The wheels are available as COTS parts, the code is shared freely, etc. The reality has been, though, that there's a significant performance discrepancy between the few good implementations and the piles and piles of really terrible ones. That suggests that mecanum really isn't that easy to get right and we should probably be putting it into the same category as WCD or swerve -- perhaps still not in terms of effectiveness, but certainly in terms of knowledge required to do it right.

While I agree, I'd venture this is the case for essentially all DTs. You don't need closed-loop to drive tank any more than you need absolute encoders for mecanum, but there's still a world of difference between the way, say, 11 drives tank and the way xxxx does. Or 25 for the former, to be more colloquial. (I just really like watching 11 this year)


And yes, somebody, please feed me video. <<pretend it's a monitor

[Madison]

For the sake of sparking discussion and because my programming team did an awesome job, I've found some video of our 2011 robot testing. This was the 'octocanum' drive; it could lower traction wheels when it needed to push its way through defense.

http://www.youtube.com/watch?v=sM8cixsE5fo

http://www.youtube.com/watch?v=Jr-eglZBAHQ

[efoote868]

Originally Posted by Siri

Chassis orientation is controlled independently of translational direction. e.g, Just push in the defender (because they can't push you), spin around them, and then spin around again until you're lined up the way you want. Like this or this. Or after a brief pin. Or before not being pushed at the unprotected feeder. Or after rocking an FCS blocker.

Mecanums can move in exactly the same way as swerve drives, and can do it without having to wait for their wheels to turn. In any holonomic drivetrain, teams can use a gyro throughout the match to keep track of their change in angular position and implement something called field centric drive.

I don't understand why this is a challenge exclusive to mecanum drivetrains, or maybe I'm not understanding your point.

[Siri]

Originally Posted by efoote868

Mecanums can move in exactly the same way as swerve drives, and can do it without having to wait for their wheels to turn. In any holonomic drivetrain, teams can use a gyro throughout the match to keep track of their change in angular position and implement something called field centric drive.

I don't understand why this is a challenge exclusive to mecanum drivetrains, or maybe I'm not understanding your point.

My point is the same as Taylor's. Swerves can and do control chassis orientation simultaneously with translational path. I have to say, it pays off big. There's nothing physical that limits mecanum drivetrains' ability to do the same. (On an open field) The phenomenon is that they just plain don't do it. I'm wondering why.

For instance, Madison's videos are very cool. I bet that really worked out for them in Logomotion matches. My question is why, with so many out there, do so few mecanums even attempt to drive this way. Is it the standard control strategy? (Madison mentioned good programming) Is it practice? Is it coaching? (As Taylor said) Do they see less benefit in it for some other reason?

[cadandcookies]

Originally Posted by Madison

I think it might be an interesting exercise for people to share video of their idea of a good mecanum drive. I suspect most FRC participants have never seen an above-average implementation of mecanum drive in action.

2052. 2052. 2052.

http://www.youtube.com/watch?v=4vlC3EqF_rc

Starting at ~2:30.

I had the pleasure of watching their driver pull moves with their mecanum drive I've never seen any other team do at the MSHSL Championships last weekend.

I can't speak for any of their code or their physical setup, but their driver really knows how to make that mecanum drive sing.

Also, my statement was more along the lines of "swerve is easy if you do it right." I would know how difficult it is when you do it wrong. We did it wrong in 2011 and couldn't move for our first regional.

[Nemo]

Why is it necessary to do fancy stuff (turning while traveling in a straight diagonal line, etc) in order to have a "good" mecanum drive? The best year for mecanum was 2011. All a mecanum drive needed to do that year was drive >90% of the time as a tank drive and strafe occasionally if that made it easier to line up to hang a tube.

Our team used a gyro in a proportional control loop to maintain the robot's orientation while it wasn't turning. I concede that I wouldn't want us to run a mecanum drive without that piece of instrumentation, because without it the robot's orientation drifts too much. Regardless, that's still not very difficult to pull off. It's a ways short of PID loops on every motor and field centric control, and it's much simpler than programming a swerve drive.

[Siri]

Originally Posted by Nemo

Why is it necessary to do fancy stuff (turning while traveling in a straight diagonal line, etc) in order to have a "good" mecanum drive? The best year for mecanum was 2011. All a mecanum drive needed to do that year was drive >90% of the time as a tank drive and strafe occasionally if that made it easier to line up to hang a tube.

I think it's less about being fancy and more about ditching the Wile E. Coyote syndrome:

Originally Posted by Taylor (for the Coyote explanation)

*Watching matches, I'm often reminded of the cartoons where Wile E. Coyote is trying to outrun a rocket. He could simply dive to the side and the rocket would fly by, but that never occurs to him.

Mecanums, even (especially) those driving in tank, will attract bad situations more than their high-traction counterparts. These situations can often be avoided or disengaged from more easily via mecanum attributes than tank-driving-mecanum attributes. At the same time, they can also set up a lot of plays that don't benefit (and actually exploit) tanks, but they're exceedingly rare. So there's a ton of Coyotes running around who are utilizing all the negatives of their choice and few of the positives.

For what it's worth, swerves could also spend most of the time driving tank and just swerve when "needed". There's nothing wrong with tank drive. But if you're going to make another DT, be it mecanum or swerve, whose physics provides such starkly different competitive advantages disadvantages (vis-a-vis tank), why wouldn't you leverage the advantages?

[pntbll1313]

Originally Posted by sentientfungus

2052. 2052. 2052.
I had the pleasure of watching their driver pull moves with their mecanum drive I've never seen any other team do at the MSHSL Championships last weekend.

I can't speak for any of their code or their physical setup, but their driver really knows how to make that mecanum drive sing.

We posted a video with some footage if you'd like to see. Hopefully this can at least quiet down some of the mecanum haters
http://www.youtube.com/watch?v=Ogy65hEPIXk

It's a 7 minute video but I'd suggest watching the whole thing as there are some cool moves in the middle and the end.

[Clem1640]

Originally Posted by Siri

I think it's less about being fancy and more about ditching the Wile E. Coyote syndrome:

I agree that it is not about being fancy but about being effective.

Swerve-Drive really introduces a new paradigm to the issue of moving from point A to point B. It requires new thinking. It benefits from new tactics. It is not really an incremental change, but it can feel this way.

It took us some time to catch on to this. We continued to suffer from the Wile E. Coyote syndrome even after the solution was in our hands.

[MichaelBick]

Originally Posted by Siri

I think it's less about being fancy and more about ditching the Wile E. Coyote syndrome:



Mecanums, even (especially) those driving in tank, will attract bad situations more than their high-traction counterparts. These situations can often be avoided or disengaged from more easily via mecanum attributes than tank-driving-mecanum attributes. At the same time, they can also set up a lot of plays that don't benefit (and actually exploit) tanks, but they're exceedingly rare. So there's a ton of Coyotes running around who are utilizing all the negatives of their choice and few of the positives.

For what it's worth, swerves could also spend most of the time driving tank and just swerve when "needed". There's nothing wrong with tank drive. But if you're going to make another DT, be it mecanum or swerve, whose physics provides such starkly different competitive advantages disadvantages (vis-a-vis tank), why wouldn't you leverage the advantages?

While I've never done a swerve, I actually believe many swerves would benefit from using a tank like "mode". A tank mode is much easier to learn and is much better for long distances. You can still switch to swerve when needed, to take advantage of it's benefits(small adjustments, getting around defense, etc.)

[Siri]

Originally Posted by MICHAELABICK

While I've never done a swerve, I actually believe many swerves would benefit from using a tank like "mode". A tank mode is much easier to learn and is much better for long distances. You can still switch to swerve when needed, to take advantage of it's benefits(small adjustments, getting around defense, etc.)

We used to have one. It's driver-dependent (as is everything), but none of our drivers have ever wanted it, 4 having had swerve drive bases. It's not particularly useful given the rest of the controls' freedom, and while the mode itself isn't hard to learn, switching between them is a needless (and rather annoying) thought "jump". There's no inherent cognitive difficulty about driving holonomic long distances, whether opposed to other swerve maneuvers or tank.

Our drive has evolved away from "modes" more and more every year. We've found very little use in that control framework itself when it comes to swerve; not using it is one of the major reasons the drive is so cohesive now. The seamlessness is a major strength of the drivetrain. Our only 2013 mode is a radial spin for aligning the climb; everything else is like playing Call of Duty. Before, drivers would largely ignore all but 1-2 modes (plus an occasional game-specific), and none used tank.

TL;DR: do what your drivers want, but don't be surprised when swerve + modes = no.



pntbll, I moved to your new thread. Cool video!

[MichaelBick]

Originally Posted by Siri

We used to have one. It's driver-dependent (as is everything), but none of our drivers have ever wanted it, 4 having had swerve drive bases. It's not particularly useful given the rest of the controls' freedom, and while the mode itself isn't hard to learn, switching between them is a needless (and rather annoying) thought "jump". There's no inherent cognitive difficulty about driving holonomic long distances, whether opposed to other swerve maneuvers or tank.

Our drive has evolved away from "modes" more and more every year. We've found very little use in that control framework itself when it comes to swerve; not using it is one of the major reasons the drive is so cohesive now. The seamlessness is a major strength of the drivetrain. Our only 2013 mode is a radial spin for aligning the climb; everything else is like playing Call of Duty. Before, drivers would largely ignore all but 1-2 modes (plus an occasional game-specific), and none used tank.

TL;DR: do what your drivers want, but don't be surprised when swerve + modes = no.



pntbll, I moved to your new thread. Cool video!

Drive mode was probably the wrong word. One setup that I've thought of is where one joystick controls direction and speed, the other joystick controls your turn radius, and another button that activates turning around the center of the robot. This is the exact same setup as cheesy drive, except you also have the ability to move in any direction. This allows you to switch between tank and swerve, without switching modes. This is completely theoretical, so I honestly have no idea how well it would work, but the idea is that there must be some way to incorporate the benefits of both drives, without having to switch modes.

[Siri]

Originally Posted by MICHAELABICK

Drive mode was probably the wrong word. One setup that I've thought of is where one joystick controls direction and speed, the other joystick controls your turn radius, and another button that activates turning around the center of the robot. This is the exact same setup as cheesy drive, except you also have the ability to move in any direction. This allows you to switch between tank and swerve, without switching modes. This is completely theoretical, so I honestly have no idea how well it would work, but the idea is that there must be some way to incorporate the benefits of both drives, without having to switch modes.

Ah. This is roughly what we do, if I understand correctly (though I admit I don't see what it has to do with tank control). One stick controls translation: crab forward, left, back...any of the 360 degrees. The other stick controls chassis orientation, at any of the 360 degrees.* (We used to have a control more along the lines of "turn radius" when it was a snake mode.) I'm not sure what you consider to be "tank" control and its drive benefits in this sense, though. I thought you meant tank control as opposed to arcade--one stick per drive side (forward is FF, back BB, turn left BF, right FB)--but this doesn't sound it. If what you describe as the first stick is "tank", what's swerve, and why is tank easier over long distances than whatever swerve is?


*I don't play COD, but I'm told this is analogous to move & look, for any gamers out there.

[MichaelBick]

Originally Posted by Siri

Ah. This is roughly what we do, if I understand correctly (though I admit I don't see what it has to do with tank control). One stick controls translation: crab forward, left, back...any of the 360 degrees. The other stick controls chassis orientation, at any of the 360 degrees.* (We used to have a control more along the lines of "turn radius" when it was a snake mode.) I'm not sure what you consider to be "tank" control and its drive benefits in this sense, though. I thought you meant tank control as opposed to arcade--one stick per drive side (forward is FF, back BB, turn left BF, right FB)--but this doesn't sound it. If what you describe as the first stick is "tank", what's swerve, and why is tank easier over long distances than whatever swerve is?


*I don't play COD, but I'm told this is analogous to move & look, for any gamers out there.

The drive style is similar in some ways, but also very different. The drive I'm describing is a combination of your snake and swerve. Imagine a non-field centric drive, where your "swerve" stick controls direction(robot centric) and the other stick controls how big your turn is. I find that tank is really good for long distances because you keep your smallest side going through holes in the defense.