pic: Coaxial Swerve Drive Module with 2-speed Ball Drive and Nitrile Tread

I don’t see how not having a thrust bearing reduces the life of the gears. It may affect the life and efficiency of the 0.375" bearing sitting above it though.

We’ve done something similar with our climber (3 CIMs on a 10:1 worm reduction) with the worm resting against the inner race of a normal radial bearing, and we saw no issues.

I can confirm what Bryce2471 is saying we saw reduced life and over time reduced efficiency due to us not having the thrust bearings between the mitter gear and the radial bearing.

Of the gears or of the bearings? Also, did you do a comparison with thrust bearings to see if life improved? I don’t see how the gears would have any additional loads on them due to a lack of thrust bearing.

The tops of our bevel gears deteriorated from where they were rubbing the bearings at. We swaped them out for milled down bevels with the thrust bearings and have not had any problems.

That makes sense. I think that the original poster might have room to put a thrust washer in there, which hopefully will help.

I’m not sure exactly why, it’s just something I have observed. I think it could be because a thrust bearing that rests on the outer race of a radial bearing provides a wider base of support for the miter that keeps it from flexing the shaft it sits on. You’re correct about it affecting the life of the bearing above that supports the gear.


My team has been designing drivetrains this past off-season and I was very intrigued by your guys’ design. I was wondering if you could tell me the name of the student who designed it and perhaps a way to speak with him/her, because my team is very curious and may want to ask some questions. Thanks a ton!

You could ask on here you would have a much larger base of responses.

A lot of low quality, second hand responses aren’t as good as a few high quality, straight from the source responses.

It’s possible that the responses are high quality, and that anything wrong could be confirmed/dispelled by the OP.

Although, you could probably also reach out to ‘Kevin Ainsworth’ via PM.

I’m not sure how often Kevin checks CD so I’ll let him know there are some people interested in some more information. I do know that the team has been working on this design for a couple months now, since IRI ended.

I would say feel free to ask any questions you have here and I’m sure there will be some informed CD community that will probably know the answer. I tend to check daily and can either hop on and correct anything said or provide information or get the answer you’re looking for.


Sorry, but I wasn’t sure of you saw my questions, and I do want to see how you got this right.

Also, are those standoffs going to be okay? They seem quite thin.

While I had nothing to do with this (awesome) design I think I can take a few of these.

To me it actually doesn’t look like there is room to flip it without increasing the bellypan footprint, which may have been a greater design concern than COG or expandability for them.

There is simply no need for such a powerful motor in this application, 1640 has been using the even less powerfull RS-540s in this application for years with no problem. Personally I would use the AM-0912 (which has power between a 540 and 550) because you can use 4 of them and that frees you up to use any of the BB motors on the rest of your robot for things that may require the power of an RS-775

Not sure on this one, it certainly looks like there is enough room for an idler encoder gear down there https://imgur.com/IwbQ4JL maybe mounting issues, gonna have to wait for the OP on this one.

I would have actually said most notably literally everything else lol, the miter gears can be purchased from VexPro but there are certainly some other complicated parts on here.

I know most teams make at least 1 spare module (if not 4) that way they simply swap the entire module out if anything happens to a competition module. The problem module can then be repaired at home under controlled conditions

Now I have a few questions:

Why the hollow drive axle?

According to the CAD the shifter gears are steel but reference aluminium VexPro gears, what’s up with that?

Amazing looking drive, great work, I hope I get to check it out in person at champs!

Ah, that makes sense. Although I believe the current motor rules allow 4 banebots style motors, regardless of whether they are RS-550 or RS-775. I will double check that.

Also, I mentioned the bevel gears because although they are direct copies of the Vexpro bevel gears, they are machined out of 7075 in the cad.
Also, another question: Why are the standoffs so thin? Will they be able to support the weight?

The standoffs aren’t taking the robot weight just the gearing. They’re plenty strong for that.

Thanks for all the great feedback, some of these questions were already answered so sorry if I repeat them. Also, sorry for the long posts but we would like to make sure the students questions are given an educational answer. I would like to thank Aren Hill for his inspiration for our “In Wheel” swerve as well as this unit.

We have been recording our theoretical to actual numbers for a few years now and 82% is our average actual speed to theoretical. So the working of “actual” should be changed to “expected”.

  1. We considered flipping the CIM due to reasons already explained and a guess that since the field has been flat for the last few years we are expecting a field obstruction this year. We can always reconfigure later. A second motor on each wheel was considered also, see our previous discussion on using 8 MiniCIM’s. The 3" OD wheels would probably be overpowered in a pushing match but the extra motors would be helpful for acceleration. You might just see dual motors from us in the future if it fits the game and if something like current monitoring is effective enough against blown fuses.

  2. Already answered, but no need for a RS775. Aren Hill used a RS395 on the first “In Wheel” swerve. We have been wondering what happens when our 2014 swerve bot gets pushed up on two wheels and the wheels are not pointed in the direction of the push. Do the steering motors have enough power to change direction? It seems like the edges of the wheels could be digging into the carpet and requiring extra force to rotate. On our off-season list of tests to perform.

  3. The ratio between the Vex Versa gear isn’t 1:1 to the modules steering gear so a 360 degree absolute encoder can’t be used. The two plastic gears are 1:1 so a 360 degree absolute encoder can be used. An incremental encoder could be used but adds another place for human error. We like the robot to not depend on a human to set the wheels at the beginning of a match.

  4. Miter gears are purchased from Vex and are 4140.

  5. If designed properly the units should last a full season without repair.
    If repair is needed they will be bolted onto the chassis and can be changed in ten minutes. We bring at least one extra unit to competition in case a swap is needed. Last year our mechanic Jose and I changed out a steering box on our robot between rounds in finals on Curie. We had previously swapped in a Banebots gearbox from our practice bot at Chicago. During our debugging of our first swerve drive ever we burned up about a dozen steering motors. Mainly due to the rotation stops we used so as not to pull CIM motor wires out. We reused the pinion gears and this particular pinion had been installed/removed one too many times. The motor shaft spun inside the pinion and we lost one wheel of steering. Took about ten minutes for Jose to install a new one while I ran around grabbing the replacement parts. We had students dedicated to our swerves from the very beginning of last year to clean, inspect, repair, etc.

Steering box question was already answered correctly, hidden in view shown.
We had field centric software installed by IRI last year with the successful integration of the Kauai Labs gyro. There was a great thread on this a little while ago. “Best gyro for frc.” Big props to our students Bennett and Duffy. Bennett for coding the entire swerve software, about 10 iterations, and to Duffy for spending the better part of the season fine tuning our teams first gyro ever used.

Great point, I never thought of it that way. I think that being able to overdrive the outside wheel is important because in a high speed strafing turn the fastest wheel is the limiting factor. If just one wheel is traveling the outside radius it will limit overall speed. The other three wheels can power you through the turn.

  1. Vex 4140 pre-hard steel miter gears
    Not sure I fully agree with the needle bearing vs radial bearing arguement
    The radial bearing can take over 100lbs of thrust force.
    I get nervous when the balls are too small in diameter
  2. Easier said than done, especially with the two speed design
  3. Silverthin 4 point contact bearing with 1100lbs of dynamic thrust load
    Only 1/4" cross section but should hold up, we are supporting completely.
    We were going to use a cross roller bearing but Aren Hill informed us of these lower cost bearings
  4. Lower standoffs support plates, upper standoffs support shifter. Non support the weight of the robot
  5. 12:60 (low) or 28:44 (high) CIM to ball shifter, with 12:30 for 2nd stage, to 1:1 miter to wheel

There was a tear in my eye last year when on a Saturday we had two lathes, two CNC’s and six Bridgeport Prototrak’s (2D CNC) going at the same time. All students except for the CNC machines. Our program is unique in that we teach machining to all students that want to learn and we have enough machines for kids to use at the same time.

Radial bearing will not hold up as long as a thrust bearing, true.
The OD of a needle thrust bearing being too large to fit is the reason we went with a radial bearing. Sometimes you have to make sacrifices.

The axle is hollow for weight. It needs to be steel since the axle runs through needle bearings (with thrust bearings). An aluminum axle on needle bearings would wear too much.

The two ball shifter gears are aluminum in our model. The two CIM gears are steel to reduce wear. We like to run a steel against an aluminum gear, we keep the smaller gear steel for weight. This greatly reduces the wear and therefore efficiency loss of the gears over the season. We’ve found the aluminum on aluminum ceramic coated gears will destroy each other over time.

Thanks for the clarification.

Couple of follow-up questions:

  1. You are using the same 82% factor for both hi and lo speeds. I would have expected them to be somewhat different. Does your data show any correlation between gear ratio and actual-to-free ratio?

  2. Do you have any actual data on swerve drives, or is your 82% number based on all non-swerve drives?

Average top speed after “full” acceleration? Or after how many seconds from a dead stop?

Sorry for the nitpick, there’s a reason for such a specific question, and either a forthcoming thread or a paper for a ‘better rule of thumb’.

  1. Easier said than done, especially with the two speed design.

Challenge excepted! lol

In all seriousness, I’ve considered CADing a light weight ball shifting swerve module for quite a while now, so I’ll try my hand at it, and see what I come up with.

I still curious about one more thing:
Have you done a BOM for this design? If so, what was the cost per module?

  1. We haven’t tested low speed since 2012, we really only cared about top speed and time to travel a distance up until now so that we can get an approximate speed. I therefore can’t give an accurate real world relationship between gear ratio and free speed to theoretical speed. From a standing start to 20 feet our 2012 was 69% efficient in low and 50% efficient in high. These numbers are misleading because they aren’t max speed they include acceleration time. We haven’t been very scientific about our testing and I can only give an average or a specific value for a given drive train on an unknown date in an unknown condition. You have got me thinking and tire wear, gear wear, bearing wear, etc can all effect the efficiency. I really don’t even want to relate my data to efficiency for that reason, I think there are too many variables that can effect this calculation based on how you got to your theoretical value.

  2. Our 2014 swerve bot started with 5" OD tires and they are 4.6" OD now, that is a 88% to 81% efficiency variation. I am not sure what diameter they were when we tested the speed so I can only say the 2014 swerve bot was somewhere between 81% and 88%. I also can’t prove we were at a constant speed the entire 20 feet.

Does a smaller wheel allow for a higher speed due to less torque required to drive the wheel (torque required at a given radius) so do they offset each other?

Does gear wear help at first to reduce friction and then hurt later when the gears start to wear out?

What effect does the gear ratio have like you asked?

What is the real world loss when an extra stage of gearing is added?
Should be around 4% per theory.

Our 2013 8WD had 40mm OD/28mm ID bearings has the least rolling resistance I can imagine (unless we attempt air bearings someday). But I can’t test efficiency accurately because the aluminum Vex gears are half gone. It was only 70% efficient after testing it against the 2014 swerve bot but I know that it was much, much faster when it was brand new. We also calculate speed with nitrile tread at a compression factor. A nitrile tread that measures 4.25" at the OD is probably more like 4" at the base of the tread so
who knows who takes that into account in their calculations.

I guess in the end I can’t trust a stated efficiency, there are too many variables that someone can make a mistake on when calculating their percentage. I care about real world performance. I just want to be able to predict the top speeds and time to travel a distance accurately.

I know that there are spreadsheets that have been developed for this and I think they are good tools to use.
I trust real world times over theoretical calculations.

I hope you can gather this because I would be very interested in the data.
Maybe a standing start to 20 and 40 feet.
This would show acceleration and usable times to cover almost the entire field.

For our 2014 swerve bot: (free speed calculated = 13.3FPS)
2 seconds to accelerate from a standstill to 20 feet=10 FPS
1.7 seconds to cover 20 feet already at full speed=11.76 FPS
So it only added around .3 seconds to cover 20 feet from a standstill.

For our 2013 8WD dual CIM: (free speed calculated = 17.69FPS)
2.2 seconds to accellerate from a standstill to 20 feet=9.09 FPS
1.6 seconds to cover 20 feet (not sure if it hit full speed) =12.5 FPS
This bot is worn out, has bent chassis rails (scrubbing due to angles wheels).
Acceleration suffers but time to cover the distance is reduced.
I wish I had numbers for this one when it was new.

These tests were an average of three runs with stopwatches.
I would like to use the encoders and data logging to get more accurate results.

Bench racing is fun! And there is always a different way to do something.
Cost per module is probably pretty high, no data is available yet.
Looks like only half the parts have costs associated to them.