FTC Swerve

[SndMndBdy]

Over the past couple of months, the mentors of 3419 participated in the NYC Corporate Challenge, where companies in New York City compete in a FIRST robotics game. The original Corporate Challenge was in 2019, where the teams competed in a Deep Space tournament. Last year’s event was cancelled because of the pandemic, and this year teams built robots to compete in “at home” version of the 2020/2021 FTC game, Ultimate Goal.

We decided to build a swerve drive for our robot, which we probably overkill for the game, but an fun engineering challenge nonetheless. Doing swerve on an FTC robot presents some unique constraints:
-Robots are only allowed to use 8 motors total, so a typical 8-motored swerve was out of the question as it wouldn’t allow for any motors for other mechanisms.
-Robots must fit in an 18" cube, so space is very limited
-We didn’t have access to our machine shop because of COVID, so our machining capabilities were limited. The only things we had access to were hand tools and 3D printers.

Here are some pictures of the module:

Whole Module:

Leg

Motion Transfer

To overcome the motor limitations, we used a servo ( 2000 Series Dual Mode Servo (25-4, Super Speed) - goBILDA ) to orient the drive. We were concerned that the servo wouldn’t have the speed or torque perform well, but it turned out that it worked just fine. The servo is in continuous rotation mode and drives the rotation through a 4:1 gear ratio, which means that the wheel rotates at 47RPMs This is quite a bit slower than we’re used to for an FRC robot, but it didn’t have a significant adverse affect on the driving quality. The stall torque at the wheel was 16.7 kg-cm, which would seem like not enough for an FRC robot, but for the much lighter FTC robot, it was sufficient.

To keep the module as small as possible, we used a wheel that was only 2.5" diameter. We struggled to find a COTS wheel that we liked of this size, so instead we 3D printed our own. We printed a tread pattern into the wheel, which worked quite well on the FTC floor, which is similar to a gym mat and has a nice squish and grip to it that interfaces with the printed tread well. This allowed us to fit the entire module in a 3.625" x 5.375" footprint, and only 6" tall.

The module used bevel gears from Gobilda ( 2:1 Ratio Bevel Gear Set (8mm REX Bore Pinion Gear) - goBILDA ) which are very small and well-suited for this exercise.

The other gears are 3D printed. We used a herringbone gear in an attempt to make the gears as strong and smooth as possible for a 3D printed gear. The gears transferred the torque and held up just fine without any visible wear. The one down side of this approach is that it made assembly very difficult because you can’t slide these gears into place axially like you would a normal spur gear; rather, they need to placed where they should go and then you have to slide the shaft through them, which was quite tricky in this small assembly.

Overall we were very pleased with the results. It led to swerve drive which was very smooth and easy to drive. I’m happy to answer any questions that anyone has.

[wgorgen]

Looks great! Nice to see another FTC swerve design. Looks like your design was inspired by SDS’s design. Do you have any pictures of the robot and how well this packaged into the “box”? I’d love to see it in action!

Are you affiliated with an FTC team or was this just an fun design challenge for your FRC team?

I’m honestly shocked that swerve for FTC has not stuck. It seems so strange that Mechanum drive is the meta for FTC when swerve offers so many advantages over mechanum (not the least of which is the smoother ride).

One of your club’s FTC teams, 5795 developed a swerve for the 2016-2017 season that was inspired (at least in part) by 1533’s home-grown swerve design. After FTC 5795 did their swerve drive in 2017 I think the team members that had put in all the effort to develop it were disappointed that the team did not use it again in subsequent seasons. There was definitely some room for further optimization. The initial testing showed that much higher speeds were possible, but the steering servo was not quite strong enough to keep the modules pointed in the right direction under max acceleration and impacts with the wall, etc. So they slowed it down quite a bit.

Still, the swerve offered a lot of advantages during auto as well as teleop.

Team 5975 developed a swerve in 2018 that was an improvement over 5795’s design (what are the odds that two teams with such similar numbers would do swerve in back to back seasons).

There was even a differential swerve developed by FTC 11115 during the off season, but as far as I know, it was not used for an actual competition robot despite the speed and agility that it clearly demonstrates.

Oh well, maybe we will see FTC swerve take hold at some point in the future.

[s-neff]

@wgorgen

There was even a differential swerve developed by FTC 11115 during the off season, but as far as I know, it was not used for an actual competition robot despite the speed and agility that it clearly demonstrates.

Diffy swerve still burns all 8 motor slots, unless you get casters involved like 11115’s prototype. Which looks sweet - but still has downsides.

[Peyton_Yeung]

@wgorgen

I’m honestly shocked that swerve for FTC has not stuck. It seems so strange that Mechanum drive is the meta for FTC when swerve offers so many advantages over mechanum (not the least of which is the smoother ride).

I think mecanum for FTC makes sense for most teams due to the simplicity/size. The benefit of swerve drive in this scale is that it allows the drive encoder to accurately track distance traveled along the floor (The increased traction is mostly a non-issue in most FTC games). With the advent of odometry and field localization, many mecanum drives can get accurate field positioning eliminating the swerve drives major bonus in my opinion.

[cadandcookies]

@wgorgen

I’m honestly shocked that swerve for FTC has not stuck. It seems so strange that Mechanum drive is the meta for FTC when swerve offers so many advantages over mechanum (not the least of which is the smoother ride).

On the opposite side, it is completely unsurprising to me that swerve hasn’t caught on. At least here in MN, most teams are operating on a 4-5 year lifespan, with little to no institutional knowledge or previous resources. It takes time to spin up to designing a swerve no matter how you cut it, much less designing one within the unique constraints of FTC, already acknowledged in this thread.

Additionally, it’s worth noting that while mecanum is the standard in FTC, the vast majority of teams I’ve observed are just cargo culting it-- if you watch them actually drive in a match or in auto, they’re driving a tank drive, maybe with some nervous attempts at strafing, if they even remember they can do that at all.

It’s also worth noting that 5975 locked their swerve and replaced it with mecanum when they couldn’t fix the issues with their steering servos (it’s been a bit since I talked to Karter, but that was what I remember). Sidebar on Cybots, they also had one of the single coolest drivetrains I’ve ever seen, which was a manually machined sprung tread drive. That thing was a work of art.

This all being said, swerve in FTC is super cool, and it’s awesome to see another stab at it like in the OP.

[SndMndBdy]

Thanks for all the positive comments.

I can completely understand mechanum over swerve in FTC. It’s just way simpler and takes way less resources in FTC, even if the outcome isn’t quite as good as swerve. And this is coming from a huge proponent of swerve in FRC!

In answer to the other questions from above:
-We are not normally associate with an FTC team. Just FRC team 3419, the Hunter College High School Rohawks. We built this FTC robot because the NYC Corporate Challenge was “at home” this year and the scale of an FTC robot made more sense for that.
-I’ll post some pictures and videos of the swerve on the actual robot once the competition is complete at the end of this week. I don’t want our competition to know too much of what we’ve been up to!

[Ethan.goBILDA]

I think the last undiscussed point on ftc swerve is module swivel time, in many ftc games you’re looking for very precise line up between a game element and a field objective (skystone and relic recovery were the big examples of this). Those generally require small scale adjustments, and waiting the little bit for modules to pivot every time can outweigh the other advantages it has (this can turn into 0.5s per cycle really easily). That ~50rpm pivot speed is fairly standard with other swerve drives we’ve seen in ftc. And while I do think this design solves the specific issue Karter (5795) had with his swerve (lots of backlash) on 393s, it doesn’t solve the slowing down of cycles.

[Aaron_Li]

Still a bit confused about the motor…

Does it function in a way similar to the 550, housing spins with shaft

[Ethan.goBILDA]

No it’s a standard brushed motor. So I guess you could say it’s an inrunner. Just the shaft and internal components rotate.

[SndMndBdy]

The motor that drives the wheel is this: HD Hex Motor No Gearbox - REV Robotics . It’s a typical brushed DC motor. It is not an out-runner motor like you might be referring to ( NEO 550 Brushless Motor - REV Robotics ). The confusing part might be the fact that the picture at the top of the post doesn’t have a belt in it, but in reality there is a belt that connects the two pulleys at the top of the picture.

The motor that rotates the wheel is the Gobilda servo ( 2000 Series Dual Mode Servo (25-4, Super Speed) - goBILDA ). Just the output shaft/spline rotates on that.

[Aaron_Li]

If i’m understanding correctly, it’s 2 motors per swerve module, which is a servo motor to orient wheel, and the hex motor to drive. This still means 8 motors… unless the servo doesnt count towards the limit?

[aidanm]

The limit is 8 “normal” motors and 12 servos.

[Peyton_Yeung]

First off this is a great looking module so thanks for sharing!

With regards to seeing more FTC swerve drives:

If you think about an analog between FTC and FRC, the number of available motor ports is 1/2 (8 vs 16) but the robots still require a similar number of DoF to play similar games.

Example:
Ultimate Goal (FTC) vs Ultimate Ascent (FRC)
DT: 4 motors vs 4 motors
Floor Pickup: 1 motor vs 1 motor
Shooter Wheel: 1 motor vs 1 motor
Auxiliary Functions: 2 reserve motors vs 10* reserve motors

*I think the blue PDP had 20 slots vs the current 16

The fact that there are more available slots means you can drop another 4 in a drivetrain without having to worry about about running out of motor ports.

Lastly, I think that the rise in FRC swerve drive usage is driven by access to COTS modules and with the single degree of freedom limit on COTS parts I doubt we’ll see an increase in FTC usage.

[troy_dietz]

@cadandcookies

the vast majority of teams I’ve observed are just cargo culting it-- if you watch them actually drive in a match or in auto, they’re driving a tank drive, maybe with some nervous attempts at strafing, if they even remember they can do that at all.

This goes for FRC and even industrial applications too, annoyingly enough. Mecanum can act like decent swerve -perhaps a bit slower- with the right driver. Might as well be Lobster Drive at that point.