Swerve: Belts or Chain?

[Qbot2640]

We've got a very young and energetic team (including a freshman programmer), and a new milling machine...understanding that this process may take several years, we are ready to start experimenting with swerve. Our plan is to make four modules and start playing with the programming and implementation.

A couple questions:
(1) The 15 tooth miter gears I'm finding on most robot supplier websites seem way too small to me. It seems to me, as I start to draw it all out, that a larger driven gear and a smaller pinion would make the geometry better - and accomplish a part of the speed reduction in the process. What experiences have others had here?

(2) I'm leaning toward using belts, but notice that most systems use chain. I definitely have a bias toward belt - I'll admit that, but what are the true pros and cons (other than the obvious, that is)?

[Botwoon]

Where have you seen chain driven swerve modules recently? I haven't seen them in pictures/videos for at least a few years now, and every one I've seen in my (relatively short) FIRST career has been either gear or belt driven.

[Gdeaver]

A true swerve design discussion would be a very long ramble. This is what our team has done if you haven't seen it, we published our design.
http://wiki.team1640.com/index.php?title=Swerve_Central

[Qbot2640]

Originally Posted by Botwoon

Where have you seen chain driven swerve modules recently? I haven't seen them in pictures/videos for at least a few years now, and every one I've seen in my (relatively short) FIRST career has been either gear or belt driven.

My apologies for not being more specific...I am referring to the module itself, attaching the driven bevel or miter gear to the wheel. I do intend to use belts for the motor and steering.

Originally Posted by Gdeaver

A true swerve design discussion would be a very long ramble. This is what our team has done if you haven't seen it, we published our design.
http://wiki.team1640.com/index.php?title=Swerve_Central

Thanks for the link...I have seen...but have not studied it yet.

[Answer42]

Originally Posted by Qbot2640

We've got a very young and energetic team (including a freshman programmer), and a new milling machine...understanding that this process may take several years, we are ready to start experimenting with swerve. Our plan is to make four modules and start playing with the programming and implementation.

A couple questions:
(1) The 15 tooth miter gears I'm finding on most robot supplier websites seem way too small to me. It seems to me, as I start to draw it all out, that a larger driven gear and a smaller pinion would make the geometry better - and accomplish a part of the speed reduction in the process. What experiences have others had here?

(2) I'm leaning toward using belts, but notice that most systems use chain. I definitely have a bias toward belt - I'll admit that, but what are the true pros and cons (other than the obvious, that is)?

Why not avoid bevel gear altogether? One of the oldest names in the swerve business, team 16, uses slip rings to allow continuous rotation without the use of coaxial shafts. This gives them a huge efficiency advantage because they don't have to deal with the power losses of bevel gears. Ultimately it's up to your team to decide what is most important to you. But if I were to design a swerve right now I would build on all the information the the neutrinos, pwnage, and the bomb squad have released about their swerves recently. In particular the cim as an axle style that the neutrinos pioneered is one of the simplest and most compact ways I've ever seen a swerve done.

[pwnageNick]

Just to clarify before the discussion gets more in depth, is your team specifically looking into a coaxial system?

This isn't coaxial, but I'll link you some of our info and CAD model if you would like to take a look.

http://www.chiefdelphi.com/media/photos/40299?

-Nick

[Electronica1]

Are you talking about something like this?

[aldaeron]

Great thread.

I appreciate all the resources folks have posted. I have looked at swerve drives off an on over the years and never had the resources to build one. I think we might have enough interest from the kids to do a swerve as an off season project.

I apologize if the answer to this is in one of the documents (I didn't see it during my quick skim). I have wondered why teams would run one drive motor and one steering motor per module (like http://www.chiefdelphi.com/media/photos/40299? or http://www.team221.com/order.php?cat=3). This seems like a waste of weight to me. The concept shown briefly in http://www.youtube.com/watch?v=CAJBC-DDL9w and commercialized as the Revolution Swerve (
http://www.team221.com/viewproduct.php?id=89) seems like a better idea. The reason it seems like a better idea to me is that you can have a single steering motor set and a single drive gearbox.

Why do teams use 4 total gear reductions instead of one central gearbox and a chain to transmit power to all 4 wheels? With a 72T or 80T 20 DP gear attaching to 4 CIMs you could make one big gearbox in the center, perhaps even add shifting for a second overall gear ratio.

Why do teams use 4 steering motors? This would seems to make it hard to steer straight forward (as mentioned in one paper previously linked). With 1 or 2 motors for steering and a worm gear based gearbox and a chain the wheels would be mechanically synced (minus any stretching of the chain). I realize that you would need two steering chains (i.e. 2 sets of 2 wheels chained together) to be able to rotate the bot around its center. By using less motors and therefore less controllers this would seems to save weight

I realize that belts can also be used for power transmission between wheels, but chain has a much higher overall load rating in tension and is easier to make very, very long.

These answers may be obvious to teams that have done swerve before, but are a mystery to us.

Thanks!

-matto-

[Tyler2517]

I think what you are referring to is a crab drive.
The biggest difference is that with all 4 wheels being steered together you can no longer have true holonomic motion. With a 4 wheel swerve you can turn all 4 wheels perpendicular to your point of rotation. Causing for 0 scrub when turning with your pivot point being anywhere(large banking turns).
Also using that much chain running around your robot is a nightmare to design around, It also adds up in weight after a while.

The ability to have full holonomic motion is the number one reason our team went with the swerve drive.
We use belts for all of our stagings other then the bevel and the chain in the lower pod. The reason we did not switch the chain out to a belt in the pod was that in the final staging a 9mm belt would not transmit egnuf force with out breaking or wearing.
Here is ours. Our only problem was the encoders slipping and one getting destroyed after we received a catapult shot to the encoder assemble. As always thank you 1640,2471 for cad and good design examples. http://www.chiefdelphi.com/forums/sh...d.php?t=124664 A

[Chris is me]

Originally Posted by aldaeron

Why do teams use 4 total gear reductions instead of one central gearbox and a chain to transmit power to all 4 wheels? With a 72T or 80T 20 DP gear attaching to 4 CIMs you could make one big gearbox in the center, perhaps even add shifting for a second overall gear ratio.

4 gearboxes allows for independent control of each wheel.

Why do teams use 4 steering motors? This would seems to make it hard to steer straight forward (as mentioned in one paper previously linked). With 1 or 2 motors for steering and a worm gear based gearbox and a chain the wheels would be mechanically synced (minus any stretching of the chain). I realize that you would need two steering chains (i.e. 2 sets of 2 wheels chained together) to be able to rotate the bot around its center. By using less motors and therefore less controllers this would seems to save weight

To put it bluntly, if your team can't handle putting 4 independent modules in the same direction, your team definitely can't handle the software complexity of a well tuned swerve drive.

Without being able to independently power wheels or turn modules independent of one another, a swerve base has no ability to rotate, only translate.

Assuming a four wheel swerve:
All wheels linked, all steering linked - robot can translate in any direction only, but no rotation.

Pairs of wheels linked, all steering linked - robot can translate in any direction. Robot may be able to rotate the same way a skid steer robot does, depending on which wheels are linked and if turning scrub can be overcome.

Pairs of wheels linked, opposite corner steering linked - This allows for translation in any direction, and rotation without translation, but not rotation with translation.

All wheels independent, all corners steered independent - This configuration allows for any combination of rotation and translation at the same time.

With the wide open motor rules, etc. the 4 wheel independent swerve design has gained a lot of popularity among those whom have mastered the swerve drive. I would not go to a lesser option for the sole reason that you don't think your team can handle independent swerve - if your team can't handle independent swerve, it likely can't handle ANY swerve. Not saying there aren't good reasons to make other choices, though.k

[Answer42]

Depending on the game crab drive can actually be an adequate swerve design. For example 118 used crab drive with a fully independent upper turret for their manipulator in both 2007 and 2008. they had no control of their chassis orientation but it didn't matter because they could turn their manipulator to face whatever way they wanted. Having one giant drive gearbox is a pretty neat advantage. I'm sure a revisit of their "v-6" gearbox with modern motor counts would be beastly. 148 also used crab drive in 2008. Since they had no manipulator (besides a pole to knock the ball down) they didn't need to worry about orientation. Just some thoughts.

[Andrew Schreiber]

Originally Posted by Answer42

Depending on the game crab drive can actually be an adequate swerve design. For example 118 used crab drive with a fully independent upper turret for their manipulator in both 2007 and 2008. they had no control of their chassis orientation but it didn't matter because they could turn their manipulator to face whatever way they wanted. Having one giant drive gearbox is a pretty neat advantage. I'm sure a revisit of their "v-6" gearbox with modern motor counts would be beastly. 148 also used crab drive in 2008. Since they had no manipulator (besides a pole to knock the ball down) they didn't need to worry about orientation. Just some thoughts.

Fun fact, revisiting with current motor allotments would require a bit of caution. Recall that our battery supplies a finite amount of power at any given point in time. I know 125 has both blown our main breaker and browned out our CRIO this season with our 6 motor drive. We've seen current draws dropping as low as 4v out of the battery (before the CRIO shut down since the 24v regulator stops functioning properly at 6).

[Tyler2517]

We had problems this year when we were geared on the high side for swerve.
When trying to go from a stand still to full power in a different direction we would draw a lot of current and brown out our CAN bus killing us for the match. We traced this to 2-5 year old battery. We got new batterys and implemented a new code feature that would off set the motors from each outher by .005 of a second allowing the magnetic fields to establish in the motors before the next ones would start.
CAN+Swerve...... Be carefully.

[aldaeron]

Originally Posted by Answer42

Depending on the game crab drive can actually be an adequate swerve design. For example 118 used crab drive with a fully independent upper turret for their manipulator in both 2007 and 2008. they had no control of their chassis orientation but it didn't matter because they could turn their manipulator to face whatever way they wanted. Having one giant drive gearbox is a pretty neat advantage. I'm sure a revisit of their "v-6" gearbox with modern motor counts would be beastly. 148 also used crab drive in 2008. Since they had no manipulator (besides a pole to knock the ball down) they didn't need to worry about orientation. Just some thoughts.

This is what I was going to ask about. It seems lighter to make a "crab" (4 mechanically linked wheels with linked steering) with a rotating (turret) manipulator. To me this makes a lot of sense for control as well since the driver and co-driver can move each piece (drive base, manipulator) independently.

For those who did a swerve this year - do you think a "crab" with turret manipulator would have worked well? Seems like it would be great for passing and strafing for a shot (co driver could align to goal while driver is moving away from defense)

Has anyone measured the efficiency of a swerve or crab? There are a lot of interfaces for power transfer!!!

Thanks!

-matto-

[nuclearnerd]

Originally Posted by Chris is me

Pairs of wheels linked, all steering linked - robot can translate in any direction. Robot may be able to rotate the same way a skid steer robot does, depending on which wheels are linked and if turning scrub can be overcome.[

I can see a lot of advantages of this configuration. It would be easier to build and program, it would give a robot the ability to strafe without sacrificing traction, and it would be no worse at pivoting than a standard tank drive (though not as good as a 4-wheel-independant swerve). Do you know of any team that has tried it?

One catch - as the wheels turn 90 degrees to strafe, the wheel pairing pattern would need to change to allow differential steering. You might still need 4 independent drive motors (need to think about this a bit). Even so, it would be simpler to implement the steering.