Swerve Drive vs Mecanum Wheel drive?

[Jefferson]

Originally Posted by Ether

Horizontal pushing "transfers some of the weight" to the rear wheels, even if the front wheels do not leave the ground.

Right. If fox46 had worded it that way, I wouldn't have a problem with it.

Originally Posted by Ether

I believe the point being made was that with a chained drive, all motors on each side are still providing force... And that force will be distributed to the wheels with the greater traction (the rear wheels), once the wheels with the lesser traction (the front wheels) have reached their traction limit.

As soon as any of the wheels leave the ground, though, you've lost at least some of your ~150 lb weight, making it more likely that you are limited by the traction of your wheels, not the power you can deliver to them. And if your wheels are off the ground, you've likely lost the pushing match already.
I guess the larger point here is: the primary advantage of an omnidirectional drivebase is you can avoid pushing matches all together by beating the opponent to the position. At that point, if you don't want to move, you lock the wheels in a position so that the robot won't roll in any direction. If you do want to move, you move sideways away from the opponent, a move the tank drive can't make.

[fox46]

We've not been buldozed much this season, even by very well built 6 CIM tank drives.

6 CIM, 2 speed, 2" wide natural rubber roughtop wheels? If you're going to champs I would love to conduct an experiment to prove or disprove this statement.

As soon as any of the wheels leave the ground, though, you've lost at least some of your ~150 lb weight, making it more likely that you are limited by the traction of your wheels, not the power you can deliver to them. And if your wheels are off the ground, you've likely lost the pushing match already.

Where did your weight go then?? Weight doesn't just disappear- it is transferred. Furthermore, Ff = Fn * u + Fhys. So even if you are transferring your weight to only two rear wheels, your tractive force will remain the same.

For someone who doesn't not like omnidirectional drives, your post has a very different tone. Everything you say in here is stated as absolutes, and it's just not that clear cut.

Not true- my post infers that if you a) have capable drivers and b) have a control system capable of making their task intuitive then you can indeed utilize the maneuverability in such a way as to overcome the defences of a skidsteer/tank drive.

only late last year incorporated a gyro into the driver-controlled code. The gyro improved the performance of the robot, not the ease of driving.

So how do you differentiate between the performance of the robot and ease in driving- do they not go hand in hand? I think you just proved my point.

This is true of any drivebase, not just omni drives. A well built drivebase does nothing if the drivers don't have practice time on the sticks.

It takes far less time to master a skid steer drive train than an Omni or swerve. The telepresence aspect of it is ^2 in any Omni/swerve vs skid steer.

[efoote868]

Originally Posted by fox46

It takes far less time to master a skid steer drive train than an Omni or swerve. The telepresence aspect of it is ^2 in any Omni/swerve vs skid steer.

In my experience, it takes almost no time to "master" field-centric control. It doesn't get much easier than point your joystick where you want the robot to go. While Ether has shown field-centric control can be adapted to tank drives, it isn't as natural a control as it is for omnidirectional drives.

[Jefferson]

Originally Posted by fox46

6 CIM, 2 speed, 2" wide natural rubber roughtop wheels? If you're going to champs I would love to conduct an experiment to prove or disprove this statement.

I don't know the detailed specs on the opposing robots.
We will be in St. Louis, but I have no interest in setting up an "experiment" where you attempt to buldoze our robot.

Originally Posted by fox46

Where did your weight go then?? Weight doesn't just disappear- it is transferred. Furthermore, Ff = Fn * u + Fhys. So even if you are transferring your weight to only two rear wheels, your tractive force will remain the same.

At least some of the weight is transferred to the opposing robot.

Originally Posted by fox46

So how do you differentiate between the performance of the robot and ease in driving- do they not go hand in hand? I think you just proved my point.

Additional functionality.
The robot operates in a robot-centric mode normally, When the driver initiates crab control, the gyro is reset and the robot is "field-centric" from the robot's orientation at that point. This gives the driver a twist input while in crab. If the twist input isn't used, it is the same drive code that we used for years before. If anything, it gives the driver one more thing to think about. It is, however extremely effective once mastered.

[Ether]

Originally Posted by efoote868

While Ether has shown field-centric control can be adapted to tank drives, it isn't as natural a control as it is for omnidirectional drives.

I put field-centric "Gyro Guided Skid Steer" on a practice bot back in 2011.

It was a hoot to drive. Try it some time.


efoote868 is right, though: the mecanum/omni "Halo Auto-Rotate" (Halo-AR) version is noticeably smoother.

[Citrus Dad]

Originally Posted by Alan Anderson

Prejudice like that shouldn't be part of scouting. I think the appropriate thing would be to make your assessment based on what each robot actually does, and how well, rather than what you think its design is capable of.

The assessment is based in large part on a fairly quantitative measure of scouting assessment data. (In fact we are probably the most empirically data-driven scouting team (for good or bad) in the the state now that I've seen scouting data from most of the top teams.) There are probably exceptional robots, and as Mike mentioned, we focus on offensive ability regardless of drive for our first choice. However, we observe in California that successful implementation of mecanum drives by third tier robotic teams is such a rarity that we focus on tank drives. A team that successfully uses mecanum drive probably will be an alliance captain and therefore not available as a 2nd pick.

[lcoreyl]

Let me address some of the "mecanum has power loss" type comments:

in the forward/backward direction, assuming the same motors/gearbox efficiency & ratio, same wheel diameter, weight, etc: swerve, mecanum, and tank have the same power/acceleration/speed.
Assuming they all have the same wheel material, CofM, bumper configuration, etc, then in a pushing contest tank wins, followed by swerve (which gets closer to even as CofM/weight transfer issues are minimized). mecanum is a definite 3rd; however, this has nothing to do with the fact there are rollers (since they don't roll in this case), it is that they reach max force of friction before the others. Also, this does not equate to tank or swerve always having an "easy" time pushing mecanum around, nor does it mean that mecanum must suck at pushing, it just means that mecanum shouldn't be selected if pushing was a main criterion of your design.

moving at any other angle relative to the robot mecanum will have power loss due to turning motor power down/off in the code (such as in 45 degree when only 2 motors are driving), and roller friction losses. So in this ranking it goes swerve, mecanum 2nd, then definitely last tank (since it can't strafe at all). the significance of the differences matters on what the game is, and what your goals within the game are.

[lcoreyl]

Originally Posted by Abhishek R

In my opinion only, I feel there will always be exceptions, but mecanum just doesn't have any appealing aspects. Whenever I see a competitive team with mecanum I just wonder what they could have done with tank, butterfly/nonadrive (I've heard this is called TexCoast Drive now...) or swerve. The idea of using rollers on wheels is very unappealing to me.

zero appealing aspects? How about the tiny fraction of time to design and implement compared to the other holonomic drives (some with “unappealing” rollers) which receive your blessing? I’m not sure how giving such a strong opinion without mentioning any characteristics or tradeoffs is useful.

Originally Posted by fox46

I like mechanum, Omni and swerve drives- it makes it real easy to bulldoze through them and get to where you want to go on the field!

by this logic, if robot A has 6WD and bulldozes robot B with 6WD, then robot A has just proven their own drivetrain sucks. There’s more to it than this.

Originally Posted by fox46

Secondly, don't even bother with Omni if you aren't using a gyro and accelerometer to take some of the work away from the driver. DC motors always run better in one direction than another and to try and pull off one of these drivetrains without some sort of logic to make sure it is doing what the driver tells it is just setting yourself up for an extremely difficult to drive robot.

Yes, sensors make driving mecanum easier, but you can be successful without them. For example, a mentor from 2996, Cougars gone wired, told me their mecanum drive encoders and gyro were not working in Utah. They still have a shiny blue banner they brought home, however.

Originally Posted by fox46

Swerve drive- People think that with a swerve drive you have the tractive capabilities of a tank with the maneuverability of Omni- you do not. Each wheel has (at least) one motor. When you start pushing against something, all your weight transfers to your rear wheels leaving your front wheels spinning in the air. Hence, in a shoving match, typically you are only using half your available drive power.

How about discussing the actual tradeoffs of swerve vs. tank tractive capabilities instead of saying tipping to 2 wheels is somehow inherent in swerves design? “typically” you are using only half power? Really?

Originally Posted by fox46

Furthermore, the lag time associated with steering your modules is painful. Any lag time slows down your ability to utilize your maneuverability to its maximum advantage and if you don't utilize it to its maximum advantage you WILL be outgunned by a skid steer/tank.

You make me want to create a verb form of hyperbole.

Never built or driven a swerve, but wouldn’t lag depend on how it’s built and programmed? Are you saying it’s impossible to build one with acceptably small lag? In watching swerve used in competitions, I can’t say I’ve noticed lag as a “painful” problem. Hopefully someone with good swerve experience can speak on this.

[lcoreyl]

Originally Posted by Alan Anderson

a simple roller lock to improve the maximum forward torque before losing traction

I'm assuming you mean a lock to keep the roller from rolling? In that case, if you were to lock mecanum wheels (not rollers) from turning then try to push the robot forward/backward, or left/right, the rollers already won't roll* so the lock won't change anything*. worst case would be 45 degrees, but still with no lock you have 2 wheels resisting the push, and this would be difficult for another robot to capitalize on.
The decrease in traction comes from the fact that the mecanum wheel translates torque to a force at an angle to the plane of the wheel, which means it must necessarily create more force than a regular wheel (with the same input torque) and therefore would reach the maximum force of static friction before a regular wheel (with the same wheel material).

Somewhere on CD awhile ago, I saw someone put forth the idea of a mecanum wheel that is 20 or 30 degrees between roller axle and the plane of the wheel. This would give a larger maximum friction force with all other things being equal at the cost of some strafe performance. I wish I had the resources and manufacturing skills to create wheels like this to test...
The other option would be to change the roller angle on the fly, but at that point, it's probably easier and more effective to just build an octocanum or swerve.

*OK, roller compliance and axial free play have some effect on this, but IMO these effects are not FRC significant enough that I would spend time designing a locking mechanism.

EDIT: since I just spent a bunch of posts addressing other posts that seemed to blur fact and opinion, I'll add that this is just my hypothesis for how a lock would behave, and it includes a couple untested assumptions about locked rollers, and I'm open to being disproved!

[Jefferson]

Originally Posted by lcoreyl

Never built or driven a swerve, but wouldn’t lag depend on how it’s built and programmed? Are you saying it’s impossible to build one with acceptably small lag? In watching swerve used in competitions, I can’t say I’ve noticed lag as a “painful” problem. Hopefully someone with good swerve experience can speak on this.

It's not a problem at all. We actually used the Denso window motors to turn the modules for years until we were limited to two. We have since switched to the Andy Mark 9015. We actually limit these motors to a max 75% output and the hardware ramp on the Jags to limit battery drain.

[Abhishek R]

Originally Posted by lcoreyl

zero appealing aspects? How about the tiny fraction of time to design and implement compared to the other holonomic drives (some with “unappealing” rollers) which receive your blessing? I’m not sure how giving such a strong opinion without mentioning any characteristics or tradeoffs is useful.

Okay, let me preface this with a few things. First, I don't consider the benefits of mecanum over other holonomic drives, I dislike them equally. Second, I currently drive an H-drive 50% of the time (we're using a version of nonadrive this year), so I feel I have a bit of credibility to what I was saying. I stated a few reasons earlier here and in the omni v mecanum thread as well.

The *consensus* is that mecanum and omni wheels can be pushed around because they have rollers on them as well as already being a wheel in the inherent nature of a ... wheel. This is typically considered their drawback. My favorite drive train are the slide drives because they seem to do just that, they make the field your ice rink and you bounce effortlessly of of other robots and the field. This has been great in previous years due to the existence of safe zones, but now we forward to 2014. No safe zones, and now you find yourself unable to get to your preferred shooting location because you're getting knocked around. That's when the stability of a traction system found on many butterfly and octacanum drives comes in. I have seen brakes on more teams this year than I have before (2848, 1523, 118, etc.).

At this point, people like to continue to make the claim that if you have a mecanum drive train, you shouldn't be in any pushing matches anyway, you should just outmaneuver your tank drive opponent. This is true up to a point. You will meet teams that will match your every step with a regular skid steer system, and then not only will you not be able to get around them, you won't be able to do much until the aggressor basically decides to leave you alone.

There are times for omnidirectional drives, but I never see a use for a pure mecanum or omni system.

[XaulZan11]

Originally Posted by Abhishek R

No safe zones, and now you find yourself unable to get to your preferred shooting location because you're getting knocked around. That's when the stability of a traction system found on many butterfly and octacanum drives comes in.

Or just have a huge sweet spot, the ability to shoot while moving (or being moved) in any direction and fantastic human player loading. This year 33 has provided a text book example of how to take advantages of your drive while masking the weakness. I'm not sure if the pure omni drive was worth it but they certainly maximized it.

[Abhishek R]

Originally Posted by XaulZan11

Or just have a huge sweet spot, the ability to shoot while moving (or being moved) in any direction and fantastic human player loading. This year 33 has provided a text book example of how to take advantages of your drive while masking the weakness. I'm not sure if the pure omni drive was worth it but they certainly maximized it.

The majority of teams are unable to score in the high goal while being stationary. Scoring while moving is a more difficult task, and even being able to inbound reliably is surprisingly hard for a lot of teams from what I've seen. The Killer Bees are one of the best, and it's hard to justify something by using one elite team as an example. Also, those features such as being able to shoot while moving, or having a large spot to inbound effectively do not show a certain drive train to be better than another, they just show the robot's manipulative capabilities to be better.

[fox46]

by this logic, if robot A has 6WD and bulldozes robot B with 6WD, then robot A has just proven their own drivetrain sucks. There’s more to it than this.

I don't follow you...

They still have a shiny blue banner they brought home, however.

How many blue banners have been won with skidsteer and how many with Omni/mechanum?

How about discussing the actual tradeoffs of swerve vs. tank tractive capabilities instead of saying tipping to 2 wheels is somehow inherent in swerves design? “typically” you are using only half power? Really?

There are enough other people discussing the tradeoffs on CD, I shouldn't have to regurgitate them.

Yes REALLY! Tipping to two wheels is inherent in every design! This is what allows for the wonders of wheelies! This includes cars during acceleration, ATVs, bobcats- most any wheeled or tracked vehicle exerting a force above the level of the surface it is driving on transfers weight to the rear. I would think this is common knowledge. If you have four motors, one driving each wheel of a swerve drive, as you reduce the normal force on your front two wheels during a pushing contest, then YES you are reducing the tractive power of your machine as the front wheels and hence two of your drive motors lose their effectiveness. In a skidsteer/tank- your wheels and drive motors are all daisy chained together so whatever power is sent to one wheel on one side is sent to all the wheels on that side- if the front wheels are in the air, all your tractive power is sent to your rear wheels thus maintaining the use of all your drive motors and 100% of your tractive power.

http://en.wikipedia.org/wiki/Weight_transfer

Never built or driven a swerve, but wouldn’t lag depend on how it’s built and programmed? Are you saying it’s impossible to build one with acceptably small lag?

I'm not saying it's impossible- nothing is impossible. I'm saying your time and driver's cognitive limits would be better spent improving whatever goes on top of the robot and its use rather than sinking your time and resources on being able to move sideways when defensive play is a reality. I'll remember this discussion every time I see a swerve drive do its awkward little "wiggle" when it changes direction to avoid a defensive machine.

[XaulZan11]

Originally Posted by Abhishek R

The majority of teams are unable to score in the high goal while being stationary. Scoring while moving is a more difficult task, and even being able to inbound reliably is surprisingly hard for a lot of teams from what I've seen. The Killer Bees are one of the best, and it's hard to justify something by using one elite team as an example. Also, those features such as being able to shoot while moving, or having a large spot to inbound effectively do not show a certain drive train to be better than another, they just show the robot's manipulative capabilities to be better.

I agree 100%. My point was to provide an example of how teams can maximize their drivetrain. I wasn't providing an overall assessment of omni drives.

Compared to a team like 469 that can plow through teams to get to their smaller sweet spot near the goal, 33 needs the large sweet spot as they cannot depend on getting to a specific location and staying there under defense. 33 was smart enough to understand this and put a lot of time and effort in ensuring their shooter had a huge sweet spot and could shoot fade aways or while being pushed sideways. The same applies to human player loading: 33 cannot depend on getting right next to the human player (and staying there) so they went with the largest possible catch radius. The biggest weakness to an omni drive, the inability to resist being pushed, was down played by 33 because they can still accomplish the game's task without needing to be at one small, specific location.