pic: REV 2 of GBX-116, Drill press swerve drive

[asid61]

The completed GBX-116 swerve drive, with encoders, hardware, and a few extra bits I forgot to add in and that were recommended.

This boasts only 9 “machined” or modified components. Of those, only three require decent tolerances and only two of those three require more than just cutting to length (the caster box and top piece).

Fully “machinable” with only a drill press, a hacksaw, a file, a 0.01" graduated ruler, a scribe, and some rookies. You’ll also need a vise to hold stuff while hacksawing and to bend some polycarbonate for the encoder mount.

EDIT: Bevel gear tolerances are almost impossible to achieve on a drill press. If you do not posses an x-y table for your press, I would highly recommend either building a good jig for the bevel gears and measuring the error or buying an x-y table.

Specific tools:
-Drill press
-Hacksaw
-Files
-Drill bit set, both numeral and fractional if possible.
-Scribe (AKA sharp metal pen thingy)
-Allen wrench/ key set (male hex wrenches)
-1.125" drill bit, reamer, or step drill. Holesaw not recommended unless you are certain you can get good holes for bearings to fit into with minimal play. Does not need to be perfect, but within 0.01" diameter tolerance hopefully.
-5/8" and 3/4" drill bits or holesaws. No tolerance on these, but the 3/4" needs to drill a hole for the cim boss. So more like 0.751"+.
-Wrench set, preferably ratcheting
-Pliers
-Deburring tools (countersinks work really well for small holes)
-0.01" graduated ruler, 6"+ long
-Solid square
-Rookies

Recommended tooling:
-Calipers. Digital or dial, doesn’t matter. These are fantastic for getting hole locations correct and making stuff the right size. Maybe $15 on ebay. Get metal, not plastic.
-Bandsaw. Useful to have around, makes cutting the 2x1 a breeze and gets right-angle cuts.
-x-y table for drill press or a mini-mill, and a wiggler or edgefinder or some sort. Makes those hole locations perfect.
-A jig to drill the bearing holes properly. This way, you just cut the jig to be able to cut the two bearing holes in the caster box at almost exactly the same place. A jig or an x-y table would be perfect.

Weight as shown is 7.19lbs plus chain. A four-module chassis is 34.7lbs plus gussets and rivets. Not too shabby, but you could probably cull a lot of weight by lightening this up or using 0.1" wall Vex 2x1 instead of the 1/8" wall stuff in the CAD.

CAD is available here: https://drive.google.com/file/d/0BzU34PeNVT0QX0JQeHhiaXdaVlE/view?usp=sharing

I will make a parts list and drawings available eventually. I might be getting my own mill soon, so that will bring a chance to try and make it myself (using the mill only as a drill press, of course). Estimated cost should be around $300-$400 per module, but that’s including the BaneBots gearbox, encoders, CIM, etc. Most teams will probably spend a lot less.

More inpute would be nice. Again, anybody with experience working to tolerance with a drill press would be very welcom in terms of general advice.

[Arpan]

I again can’t express my happiness with this. Finally, swerve for the masses!

[mrnoble]

Is there some kind of Chief Delphi award? There should be, and this should win it for 2014. Thank you so much.

[Arpan]

Is there some kind of Chief Delphi award? There should be, and this should win it for 2014. Thank you so much.

Seconded! Still not sure that you could get good enough tolerances to mesh miter gears with a drill press, though…
Maybe you could 3d print a guide that clamps over the part?

[Oblarg]

Seconded! Still not sure that you could get good enough tolerances to mesh miter gears with a drill press, though…

We did it for our roller intake last year, though it was sloppy and inefficient and I wouldn’t recommend it for a drive.

[Joe_G]

I would agree that you are unlikely to achieve adequate tolerances for your bevel gears with a drill press. I would say that a non-coaxial swerve is a much better choice for a team aiming to minimize machining resources.

[Tyler2517]

You might have some problems on the yoke(the part holding the wheel) taking large amounts of side loads might cause damage. (ie. you getting rammed) -1640 has a little triangle piece that goes in their side plates that help.

The better the mitter gears’s bores are machined the better your efficiency will be this will become noticeable on the field if it is out of a decent tolerance. If you machine gears the placement funky you will see problems. If the shafts for the mitter/gears in the yoke are not straight from the machining processes you will have some problems as well mainly a clean mesh on the mitter gears.

I do not see how you are handling thrust loads from the mitter gears. I also don’t see how you handle the thrust loads from the yoke to the main tube.

We ran a single main tube like yours last year it was a 2X3 1/4" thick peace machined down to 1/8" on 3 side the bottom still being 1/4" We did do a lot of weight reduction on the box but we still saw deformation in the box at the end of the season. We think this was due to minor drops when placing it on the field on the cart and such. Remember that sometimes the largest loads you will see will not be in the field but in the pits or shop when something goes wrong and it falls 4 inches on the the shop floor off of its blocks.

Using thinner material will be a death wish to the module remember you are supporting the whole weight of the robot on those 4 tubes plus any shock loads.

I like the idea that its a drill press swerve but making it work on a drill press is not the easyist thing to do. My teams rookies could not make that fresh into the shop you may at the very least be having to use your most experienced team members to do this.

The tolerances needed for a good mitter gear mesh are much higher then most of your frc tolerances. Yes the bearing will help some of it but not all a very well located bore and very straight bore is needed i would recommend when machining the mitter gear bore in the yoke to lock the movement of the x/y and use parallels. Do a large number of tool changes to bore the first side and then the second if you can so that the shaft is straight.

Also the weight seams a bit light ours was 7lbs-5 ounces with heavy weight loss and no gears and only a small thing of #25 chain.

[asid61]

Thank you for all the advice!

On bevel gear meshing, I’m using the 12 pitch 15 tooth miter gears from Vex, so I thought that they might be okay with 0.01" tolerances. Your experiences tell otherwise, so I’ll design a non-coaxial version as well.

Thank you Tyler for the advice. That’s exactly what Iw as looking for here.
Thrust loads from the miter gears are acting on the bearings, but the bearings face outwards such that the flange side of the miter gear is taking the loads.

Box deformation was something I though would be minimized by the use of 1/8" tubing over 1/16", but I want to test the deformation first before upgrading to 3/16" wall or 1/4" wall tubing. Perhaps the lightening on yours affected it in unforseen ways.

The bottom module can be made into 1/4" wall tubing with hardly any effects, so I’ll do that.

The weight is low because I’m used to optimizing weight in my designs; as a result I have thinner wall tubing here. If this was more than 9lbs, I would have scrapped it. That’s just too much, even for something made on a drill press.

[magnets]

I have some serious concerns about the strength of this swerve. To begin with, the 3/16th wall rotating module looks really weak, especially at the top corners. I’d be very concerned with this part “parallelogramming”.

Secondly, your spring wave washer isn’t going to be able to hold the bevel gear in the correct place axially, and you don’t have a way to deal with bevel gear thrust loads.

Finally, you have a very weak bearing setup for axial loads on the rotating module. The lower of the three hex bearings will take all of the load, and will most certainly be destroyed.

Also, have you ever tried running a chain on a 16 tooth sprocket that spins at 5300 rpm? It’s terrifying. There’s a reason teams use gear reductions!

[asid61]

I have some serious concerns about the strength of this swerve. To begin with, the 3/16th wall rotating module looks really weak, especially at the top corners. I’d be very concerned with this part “parallelogramming”.

Secondly, your spring wave washer isn’t going to be able to hold the bevel gear in the correct place axially, and you don’t have a way to deal with bevel gear thrust loads.

Finally, you have a very weak bearing setup for axial loads on the rotating module. The lower of the three hex bearings will take all of the load, and will most certainly be destroyed.

Also, have you ever tried running a chain on a 16 tooth sprocket that spins at 5300 rpm? It’s terrifying. There’s a reason teams use gear reductions!

I know about the 3/16" thickness. I’m changing it to 1/4", and I’m adding a couple spacers from McMaster for strength.

The wave washer diesn’t actually take much load at all. Bevel gears are inclined to turn away from the axis of the mating bevel gear, so the wave washer just spaces it away when not driving. When driving, the bevel gear is forced against the inner race of the hex bearing.
Vex bevel gears have a small diameter boss on the back end which sit inside the diameter of the inner bearing race on the 3/8" hex bearings. If that actually becomes a conern, then it’s easy to just add a thrust washer there. Can somebody explain why having the bevel gear boss on the bearing inner race won’t work?

What do you suggest for the lower bearing? Keeping in mind that these are 3/8" hex bearings with a relatively large inner race, I’m not sure it will be a problem.

I did not consider chain speed. Thank you for pointing that out. Belts could actually be used in this application, so I’ll switch to those. Alternatively, a cimple box or a vex 1-stage gearbox could do it too.

Most of these issues can be solved just by switching to non-coaxial.

[Jared]

This is a really cool idea and a neat design. I have a few comments/suggestions to make it stronger.

I know about the 3/16" thickness. I’m changing it to 1/4", and I’m adding a couple spacers from McMaster for strength.

If you put some type of brace like 1640 does so that the module has 4 sides, it would likely get a lot stronger. I can’t see if you’ve done this already, but if you have a lower dead axle that is a 3/8" bolt and you put spacers on it, you’ll add a ton of stiffness to the rotating piece.

The wave washer diesn’t actually take much load at all. Bevel gears are inclined to turn away from the axis of the mating bevel gear, so the wave washer just spaces it away when not driving.

I think you’ll probably be okay here, especially if the distance the gear can move too close is small.

When driving, the bevel gear is forced against the inner race of the hex bearing.
Vex bevel gears have a small diameter boss on the back end which sit inside the diameter of the inner bearing race on the 3/8" hex bearings. If that actually becomes a conern, then it’s easy to just add a thrust washer there. Can somebody explain why having the bevel gear boss on the bearing inner race won’t work?

We’ve run a setup with a worm gear driven by three CIMs which had no thrust washer and saw no wear on the bearing or the worm. However, some teams here who have run a swerve (IIRC it was 2517?) have commented that they saw issue due to thrust loading a bearing. I would ask other teams and see if they got away with this.

What do you suggest for the lower bearing? Keeping in mind that these are 3/8" hex bearings with a relatively large inner race, I’m not sure it will be a problem.

The two bearings in the what looks like a 1 x 2 aluminum box will hold the 3/8" drive shaft pretty rigidly. The first concern is that there are only three total bearings. Whenever you push on the side of the rotating piece, you’re loading the lowest hex bearing in the module. Bearings are really bad at these torques. It’s like doing a West Coast drive style bearing block, but with only one bearing. Adding that second bearing changes it from trying to bend the shaft to the side, to an axial load, that the bearing can handle much better.

All that said, I would look into using a different type of bearing/bushing. All swerves that I’ve ever seen that are set up like this have two bearings/bushings that go on some round stock that sits on top of the rotating piece. Usually, these bearings are either huge (1" ID or so) ball bearings, IGUS bushings (see 1640’s swerve), Silverthin bearings (>1" ID, see 118’s old swerve), or some custom bushing (see revolution swerve from 221 robot systems).

The strength of this bearing should also be determined by what the field looks like. If there’s a bump/step like in 2012, 2010, or 2004, or if there’s something you can bump the wheel into, like the corner of the pyramid, you may want something very strong. If it’s an open field, this may be less of a concern.

Also, I’d recommend going to a larger OD thrust washer, or just using a thrust bearing.

I did not consider chain speed. Thank you for pointing that out. Belts could actually be used in this application, so I’ll switch to those. Alternatively, a cimple box or a vex 1-stage gearbox could do it too.

Most of these issues can be solved just by switching to non-coaxial.

We’ve run chain this fast in a shooter prototype. It’s scary, but not impossible.

This is an awesome design, and it’s surprisingly light compared to some more complicated swerves.

[stufflikethat]

Only one question. What do you see the wheel base of this robot looking like. The current layout stops the wheels from getting as far to the corners as possible.

[GeeTwo]

It’s wonderful to see that someone’s going after this idea! Given that bevel gears are essential to this working, my thoughts on this line were to go OTS on this, using something similar to one of AndyMark’s LJ bevel boxes (am-2796 or am-2622), each of which already include a 2:1 reduction. I’d love to see my vision of a “dizzybot” become reality using something like this. The “dizzybot” would have the full capability of a “unicorn swerve” that is software-constrained to prevent slipping wheels, but with only three control pins needed from the 'RIO:
Control #1 - steer all the wheels relative to the base frame (all of them always point the same direction)
Control #2 - spin the wheels against the floor (all at the same speed, always)
Control #3 - spin the superstructure relative to the base frame (since the base frame ideally never rotates)

The simplest implementation I can imagine would be to put all of the electrical system on the “superstructure”, and extend three shafts and sprockets (or gears) down to the base structure.

[asid61]

Only one question. What do you see the wheel base of this robot looking like. The current layout stops the wheels from getting as far to the corners as possible.

Notice the angled sides of the 2x1 on the top? Those allow the module to be mounted into a corner. It should add some strength to the chassis as well, although you do lose some chassis space with this design.

[asid61]

It’s wonderful to see that someone’s going after this idea! Given that bevel gears are essential to this working, my thoughts on this line were to go OTS on this, using something similar to one of AndyMark’s LJ bevel boxes (am-2796 or am-2622), each of which already include a 2:1 reduction. I’d love to see my vision of a “dizzybot” become reality using something like this. The “dizzybot” would have the full capability of a “unicorn swerve” that is software-constrained to prevent slipping wheels, but with only three control pins needed from the 'RIO:
Control #1 - steer all the wheels relative to the base frame (all of them always point the same direction)
Control #2 - spin the wheels against the floor (all at the same speed, always)
Control #3 - spin the superstructure relative to the base frame (since the base frame ideally never rotates)

The simplest implementation I can imagine would be to put all of the electrical system on the “superstructure”, and extend three shafts and sprockets (or gears) down to the base structure.

I considered using the LJ bevelbox, but I decided against it due to the $130 price tag. For that much, you could just buy a $50 cross slide table from HF or Grizzly and use that to position the bevel gear holes.

That sounds like an interesting swerve design. I don’t think it would work perfectly though, as the individual wheels spin and rotate at different speeds for different maneuvers. So either you have to use a microcontroller between the cRio (or rRio) and the swerve modules, or you have to sacrifice the pins.

[GeeTwo]

The idea here is that you rotate all of the wheels to point in the same direction at the same time, and go at the same speed all the time - mechanically linked, perhaps driven off a single chain. The “base drive” system effectively only translates, never rotates, so there is effectively no wheel slippage on a flat surface. Then, a “superstructure”, including the manipulators and the entire electrical system, rotates relative to the “base drive”. Note that it would be reasonable to provide high and low gears using a single shifter, since there is already a single mechanical axle (drive for the chain) for all drive energy. The base drive would essentially work like one big swerve module with multiple wheels.

[asid61]

The idea here is that you rotate all of the wheels to point in the same direction at the same time, and go at the same speed all the time - mechanically linked, perhaps driven off a single chain. The “base drive” system effectively only translates, never rotates, so there is effectively no wheel slippage on a flat surface. Then, a “superstructure”, including the manipulators and the entire electrical system, rotates relative to the “base drive”. Note that it would be reasonable to provide high and low gears using a single shifter, since there is already a single mechanical axle (drive for the chain) for all drive energy. The base drive would essentially work like one big swerve module with multiple wheels.

That’s essentially a crab drive. I believe 148 in 2008 used a very similar system.

[Kevin_Sheridan]

That’s essentially a crab drive. I believe 148 in 2008 used a very similar system.

A better example is 118’s 2007 robot https://www.youtube.com/watch?v=h3l6Xqts4to or 2008 robot https://www.youtube.com/watch?v=CAJBC-DDL9w, both of which feature crab drive and a rotating super structure

[asid61]

A better example is 118’s 2007 robot https://www.youtube.com/watch?v=h3l6Xqts4to or 2008 robot https://www.youtube.com/watch?v=CAJBC-DDL9w, both of which feature crab drive and a rotating super structure

Oh yes, that’s the one I was thinking of, not 148. Thank you.

[matthewdenny]

Would it be possible to see this from a few more angles? especially from underneath?