Concerning Proliferation of COTS Swerve

(note, quite long)

As a student part of an australian team, I recently attended the Southern Cross regional event. From 2021 and earlier, there has only ever been one swerve team at an event at any one time. This year 7 of the 27 attending teams had swerve drive, 5 of which utilised a model of SDS mk 3 or 4 module. An important note is that my team is one of the two that developed our own, the other team had a swerve based on ours.

There is a tread following our developement of swerve drive if one is interested, but our modules are comparatively low tech compared to the edge chamfered CNC machined 6061 plates of the flipped SDS Mk 4. I observed a fairly noticable difference in performance between team’s swerve drives, not only in ours against theirs but even within the very similar COTS swerve using teams. 4613 is one of the five teams to use SDS swerve, and they were a standout more performant driving machine compared to the others that used almost identical drivetrains. Even our 3D printed and roughly routed modules were able to keep similar pace to them. I put this down to 4613 pre-existing experience with their modules as they used them in 2021, and similarly our intimate involvement in the development of our modules led to alot of optimisations and experience that just simply made everything work better.

When we started development of our swerve modules mid 2020, only one other team in austraila had really tried it but with limited success. Then in 2021 it was quite demotivating to see that a team had just purchased a very functional solution, devaluing the fact that at that point we had just got our drives working which would have been the first conventional swerve in Australia. The fact is that basically every single high budget team in my country has purchased an off the shelf solution only one season later to what had once been a hallmark of technical achievement within and outside of FIRST.

My involvement in the development of our modules is probably the single richest experience of the engineering process i’ve ever had, and it’s really disappointing to me that many teams are sacrificing that experience for their students in order to gain a competative advantage. COTS swerve is also too expensive for most non-university or high budget school teams, due to the nature of the high tolerance, quality material and custom component assemblies that they are. Therefore it is not simlar to spending a reasonable ~400USD to outfit a KOP drivetrain with NEO motors for a performance boost, but a very significant resource barrier. This effectively states that a team must either put in considerable time, effort and inginuety to develop their own designs or simply be a major institution.

I assume for many teams it may not be viable to develop an in-house design which is fair enough. But I would think those who are able to and do put the stated months in should expect a mechanism that reflects their efforts compared to another. Any complex, unique combination of software and hardware that a team develops should reasonably be able to provide a competative advantage, but when positioned necto to just having a significant lump sum at a team’s disposal it brings in to question why it is worthwhile. The reason it is worthwhile is for the student’s engineering process and learning experience mentioned before, but again that seems to be commonly sacrificed for a better chance at the blue banner.

I think we have here a fairly complex problem, one that is quite close to me personally. Nothing any of the swerve vendors is doing is illegal, immoral etc. nor do I have anything against them. From my understanding the practice of selling a team’s swerve design mostly started with a way of being able to develop and manufacture a swerve drive ahead of kickoff, as by becoming a vendor for swerve it becomes a COTS part, a respectable way of following the rules to achieve a not unreasonable goal. The problem arises when teams actually buy the things en masse. Now a significant portion of teams have Thrifty, Andymark, SDS, WCP and other types of modules.

The KOP drivetrain is a ground level solution to an essential game challenge ensuring that a team wanting to compete can have a reliable, non-season ending foundation to build their robot onto. For many I imagine it was an introduciton to many engineering concepts such as gear reductions or belt drives. I do not think COTS swerve can be considered similarly. The KOP drivetrain can be interated upon, improved with brushless drive or omni wheels in the corners. It can be of a variety of sizes and shapes with differing gear reductions, wheel quantity or wheel sizes depending on a season or team’s goals. Drivetrain code can be obtained, but I take it most first year teams would learn heaps about the RoboRio, Joysticks, Radio and all other electronic components in the process of implementing software and electrically wiring up their machine. All COTS swerve is effectively a fixed, high performance solution with no wiggle room. A team can purchase modules, bolt them to some tube and run the included drivetrain code and have learnt very little. This lack of customisation of not a fault of the vendors, nor is it the focus of this argument point, but the fact that a team wouldn’t learn much more from watching videos of the drive being implemented compared to doing it themselves. A team with the bank to buy COTS swerve almost certaintly knows the groundwork of a robot so the bolt on, plug in, download and run solution teaches little. Swerve code can be done from scratch with purchased modules, but that does not cover the huge lost knowledge of the mechanical and electrical aspects of design. Working without known gear ratios, what encoder to use and how to implement it, manufacture of complex parts, CAD design, motor selection, bearing selection and material selection are just a few of the critical engineering challenges lost to purchasing swerve even if you do code from scratch.

To summarise, I believe there is a significant problem with the proliferation of COTS swerve (and other COTS game solutions) in the form of lost student experience, knowledge and self achievement that is often sacrificed for a more performant machine. This is componded by the fact that if that team bought swerve and mine can’t develop it, then there is only one real course of action if I want to keep pace. It is no doubt that good swerve is an advantage over good tank drive especially in the wide field collect and precision lineup games recently. When I heard that 118 was using swerve this year, I was excited to see what they pulled off. It turneds out, like so many others, the drive was not unique. This is the same with other very capable teams such as 148 and 1678 if i’m not mistaken have gone that route, because it just make so much sense.

I would like to reiterate, I have absolutely no beef with any team that use COTS swerve or solution or makes a COTS swerve or solution. The decision was probably a very reasonable one.

This is a issue I do take fairly personally, so am very interested to hear other team’s or people’s perspectives, reasonings or thoughts. Unfortunately as I am leaving in my last year of being a student in F.I.R.S.T, I am seeing less of the Inspiration and Recognition that I entered too just four years ago.

(Edit:) I would like to add that I do not consider our swerve drive worse off, less significant or other similar element due to other teams purchasing swerve. Myself particularly as well as the team am very happy with the performance, durability and overall achievement and plan to continue improving.

Somewhat paradoxically there is a robotic sport that has a solution to this problem, combat robotics. In FRC the engineering problem is the field, unchanging and thus a solution can be purchased. In Combat Robotics the engineering problem is the opponent machine, infinitely variable with not perfect solution. The most expensive CNC machined, generatively designed, aircraft alloy constructed machine are often handily defeated with a garage built mild steel machine with the right mind behind it. While there are always meta designs, differing arenas change that game up. Note this is more for true open combat robotics, not the type you see on TV. Down in the lower weight classes a new material, specially made motor controller, truly custom inertial control system, overall metabreaking robot design, or even just a really well made box can deservedly win or loose based on the efforts and skills of the builder. There’s alot more I and R of S and T in that.

20 Likes

What about those who did custom gear boxes? didn’t Vex, AM and banebots ruin the “learning experience” of designing and manufacturing those? Naw, what these COTS allow teams to do is focus on actually designing a robot that plays the game. You still learned a lot from your experience. why are you judging other’s priorities when it comes to their experience?

48 Likes

There are many threads on the merits of COTS parts/assemblies vs custom.

While some teams have the ability to make things custom, many teams do not. The COTS revolution has allowed teams that have had dreams of higher level robot functionality to actually be able to achieve those dreams.

It’s great that you were able to develop your own drive solution and learn from it. Don’t let what others do diminish how you feel about yourself. Also don’t look down on other teams for using premade parts/assemblies. They may not have access to the manufacturing tools required to make their ideas a reality.

Somethin somethin a rising tide lifts all boats.

29 Likes

If you have a limited budget and education is your main goal FRC is not a program tailored to your needs. The “I and R of S and T” applies not only to the team members but to the public too, and going sideways looks really really cool. We don’t act like it most of the time, but at it’s heart FRC is for public consumption.

If we’re to means test here, having access to the machinery to make a home cooked swerve is no small feat. For many years my team worked out of a school with a ballet studio instead of a machine shop. The cost of a decent mini mill far outpace that of even two sets of COTS swerve modules, and that’s assuming you have mentors that can teach machining.

5 Likes

The proliferation of COTS swerve introduces some level of “pay to play”, but not to the point that I think custom swerve designs, or well practiced tank drivers can’t compete. The teams that are running highly optimized MK4 swerve drives were already going to be great at this game whether they used a swerve drive or the KOP chassis.

Ultimately this “problem” is solved through game design. The moment FIRST introduces field elements that inhibit the performance of COTS swerve modules, you’ll see a large percentage of swerve teams return to using tank drive.

35 Likes

To everyone coming out wholeheartedly in favor of, or in opposition to the original post:

The reason this is controversial is because different people and different teams have conflicting experiences that cannot be reconciled with a single solution.

Please try to imagine yourselves in an alternative scenario.

Please make the effort to bring nuance to this discussion.

17 Likes

Our team is using our own custom swerve with 3D printed forks and other gears this year, for the first time ever. No doubt, we are glad we did vs. WCD that we had gotten down to a science all those previous years.
While we experienced great success so far this season, we can see a difference with the other swerve bots using SDS.
Our plan is to go with SDS in the offseason, and focus on making our own custom modifications to it. Or continue iterating our custom ones to make it better.
I believe that as COTS vendors provide better solutions each and every year, the level of play rises overall. Unfortunately, doing FIRST is expensive (I really know this). We focus our efforts on creating the XY axis during the offseason, and the Z axis during build season, to try and create a successful robot annually. There are so many opportunities to focus on the design parts because the game changes every year. That’s the part that keeps the creative opportunities going.

12 Likes

I can understand the feel bad about having taken the time to design your own swerve from the ground up, and then only a year later having a COTS solution come in and provide a comparable advantage to other teams.

That being said, the metaphorical cat is out of the bag here, and to not use something that could give you a competitive advantage is doing a disservice to your students as well. One could make the argument that it is the height of arrogance to say that I will not use someone else’s design that is better than mine, simply for the reason that it wasn’t your own.

When it comes to the lost learning opportunity that is really something that is up for the students to decide if they are losing out or not. We had the discussion within our team if we wanted to do this and the majority of our students said that they wanted to use COTS swerve.

9 Likes

In 2012, a student team and I had worked very hard to create an autonomous system. It allowed sending of auto files that were executed sequentially. Real time selection of the files could be done from the driver station. Over the next 4 years trapezoidal profiling was added, along with gyroscopic heading control, velocity control, and field-relative odometry.

Those are now things that are now baked in to WPILIB, along with swerve control, trajectory planning, and system profiling. So a competitive advantage that we worked very hard to create disappeared.

I felt the same way you did initially. I was very jealous of WpiLib teams as we hand developed our swerve code this year (starting back in October of last year). We also aren’t a high resource team. We scrimped to be able to afford the thrifty swerve - which is an excellent option design based on one of those high resource teams that has trickled down to the masses.

But FIRST isn’t designed to be fair. It never was. You have NASA engineers and professionals on some teams, and biology teachers as the sole mentor on others. You have teams with 100 members, and teams with 1. You have teams with huge budgets who literally HAVE to fly everywhere they go (ask 359 if they like having to fly to every event), and teams who have lived on the yearly 1718 grant to make their registration fee.

That’s what FIRST is, and when you come to grips of it you’ll be much happier. Enjoy your own achievements and successes, and applaud others for theirs. That is the heart of Gracious Professionalism (when it isn’t weaponized).

55 Likes

There’s a fundamental difference when comparing the KOP Chassis and COTS gearboxes and swerve drives: the KOP Chassis and COTS gearboxes are not perfect. This is immediately obvious when you consider that out of thousands of teams, the majority don’t use the KOP chassis and many have at least one custom gearbox on their robot. If you include timing belts and chains in your definition of gearbox, that number definitely swings to the majority. There’s a huge solution space in FRC that relies on custom drivetrains and gearboxes.

However, when you see swerve drives at competition, it’s becoming increasingly rare to see custom solutions. Teams that pursue custom swerves almost universally do it worse than they would have had they just bought modules off the shelf. The benefit of making your own swerve drive is purely academic. Is that to say a custom swerve can never outperform an SDS MK4i? No, but certainly the number of teams that could do it without impacting development time is in the low single-digits.

Personally, I’m a big fan of COTS products that raise the floor and allow students to learn more and get creative with their designs. Configurable planetary gearboxes, the massive selection of hex-bore gears from Vex, a variety of gussets, smart CAN motor controllers, and most of WPIlib falls into this category. Where the OP and I would agree, I think, is that products that instantly outperform all custom solutions via raw performance and decreased development time are unfair to the competition. In general, I consider any product that a driven student cannot “beat” in 4 years to fall into this category.

There is no game design that will stop people from pursuing swerve. 2008 might come to mind, until you remember that 148 won 2 events on the 1st alliance with a custom crab drive. I truly cannot think of a game where a swerve drive will perform worse than a tank drive. There are games where it is different shades of better, and swerve drive inherently has more complexity and risk to it, but most teams are not doing strategic analysis when picking their drivetrain beyond validating their pre-existing thoughts.

All that said: I really don’t think COTS swerve drives are all that bad. Why? Because they inevitably don’t actually perform significantly better than tank drives in my experience. The increased development time and hundred minutiae that teams have to deal with when using a swerve makes up the difference in the amount of practice time teams get. I can point to teams like 4414 and 1678 and say that a swerve drive has certainly increased their performance, but I can also look at 973 and 254 and ask what a swerve drive would tangibly improve for their past robots. Ultimately COTS swerve stops young designers from pursuing solutions that are out of their reach and gives programming teams many more opportunities to learn and optimize that they may not have with a tank drive or a custom swerve. I’ve seen too many terrible custom swerve drives to be annoyed at the existence of COTS options. Similarly, LimeLight allows overzealous programmers to skip significant amounts of wasted time pursuing custom solutions and focus instead on tuning the filters and using the data to actually do things on the robot, without putting a huge cap on performance. Before LL, many programmers were buying Jetsons and trying to implement a custom solution that “works, but we didn’t have time to test it” after burning a week of practice on something that doesn’t benefit anyone else on the team.

19 Likes

This debate has been had over and over again. I tend to agree with the commenters that encourage more COTS solutions, and explain how it is more inspirational. (Such as what @wilsonmw04 has already commented)

That said, I think I can add some value by explaining our inspirational process with the:

the edge chamfered CNC machined 7075 plates of the flipped SDS Mk 4i

Our team started our swerve drive journey in 2019, where we did develop a custom swerve drive. We attempted to make something low cost that would work well…and well…it was…rough. I think we learned a lot, and it did sorta work, but it didn’t run well. Our designers did get some cool experiences, and our programmers had some fun,

For the 2021 “season” we decided that rather putting significant effort into the at home challenges, we would instead focus on swerve again. This time, we decided to go with a COTS solution and purchased the TTB swerve modules. We built a really cool mini swerve drive. We did NOT run it in the at home challenges, and it was simply for us to learn from. We spend some extra time designing our electrical board on this robot, so it was still a fun project for our CAD team. We decided to challenge our programming team, however, and have them program the swerve drive based on the papers from Ether. All in all, I think the project was very cool, and our programming students learned a lot from this project. After this swerve drive was successful and we were happy, we decided we would be ready to run a Swerve drive in a real season…if the game allowed.

The 2022 game did indeed look like a great fit for swerve, and we decided to go for it. We decided that while a great learning tool, the TTB swerve modules weren’t as battle tested as other options, and we were nervous running them for a regular season, so we purchased the MK4i modules. Our software team rewrote the code on our TTB swerve bot to utilize the WPI lib swerve functions, but I think our programming team had an edge because they understood the theory. In the time they had this season, we have one of the most impressive autonomous setups our team has every created, using the path planning tools to do some really cool things. The night before our first competition, they tuned a 3 ball autonomous, and it worked! All in all, I think our students got a lot out of the purchase of the COTS swerve modules.

Did it help us at our week 1 competition? I’d say it did help us, but I don’t think it was the only reason our robot did well. The stuff on top of the drive train is what actually set us apart, and the swerve drive just gave us a boost. There was at least 1 KOP chassis that was able to put up more cargo than we were on a regular basis at our competition.

Hopefully you find that a helpful perspective.

9 Likes

I think you answered your own question. Since FIRST is abut providing experiences that lead to technical growth as well as inspiring students to pursue STEM after high school, I would say that your experience is a success story.

We developed our own in-house swerve modules back in 2016 that we first used in 2017. Even though COTS modules are available, and purchasing those modules would save us a lot of valuable build-season time that it takes each year to fabricate a new set of our own modules, we have chosen to stick with our design rather than switching to COTS. We continue to refine the design and the students learn a lot about the engineering process from these modifications. They also learn a lot about project management by managing the extra resources that are needed to fabricate these modules each year compared with re-using a COTS module that could be assembled prior to kickoff.

Our students appreciate the challenge of maintaining our design and the growth that comes with it. Other teams chose to challenge themselves in other areas of the robot design and select COTS modules instead. Neither answer is the “right” answer here. As long as the students are being challenged and inspired, both paths can lead to the success of the students involved.

8 Likes

My opinion: You have just learned a very valuable lesson about how engineering works, both in academia and in industry.

You will likely face experiences where you work your rear end off, pouring all your heart and mind into the creation of something new. You’ll be proud of it, You’ll sell it, you’ll use it to do good things. It’ll be the talk of the town.

Then, out of nowhere, someone will come by and knock your product’s socks off. Maybe they make it faster. Maybe they make it cheaper. Maybe they improve it, or mass manufacture it. But it’s inevitable: If you make something good, someone else will come along to make it better.

We all stand on the shoulders of giants, directly or indirectly. Your team’s swerve stands on the shoulders of everyone who was involved in the design, everyone who designed the tread and the wheels you used, everyone who designed the electronics and controls and motors… lots of folks.

Today’s COTS swerve options stand on those shoulders, plus the thousands of hours of effort teams efforts to show what works and what doesn’t, and prove the value of what they did. Your team is a part of the shoulders which support today’s COTS industries.

How does one move forward in an environment like this?

For some companies, they turn to patents, intellectual property protections, licenses, etc. There’s some merit to this. But the key is it only delays the inevitable.

Ultimately, you gotta keep evolving and advancing. Don’t focus on what you’ve done. Focus on what you can do next.

IMO - there’s a deeper truth here. Ego Death is a recurring theme in many philosophies and religions. Human individual accomplishments are finite, and pale in comparison to what groups of people can accomplish. Look to your individual accomplishments not as things to be cherished for themselves, but rather as stepping stones toward a greater goal.

Concretely, in FIRST, the I and R are the greater goal - the S and T of today are just the stepping stones to get there. Rather than focus on the stone you’re on… look to the next stone. The alive-Ego mindset stares at the stone under your own feet. The Ego-less mindset looks to the next stone.

22 Likes

Depends where your priorities are. Is the student (engineering) experience measured in building a holisticlly competitive machine? Is it measured in nuts and bolts engineering? Something even more granular like stress analysis on custom fasteners?

Adapt to the changing meta, eyes forward, raise the floor, etc. etc.

This may all be a little rich coming from one of the few holdouts for custom proven competitive (:grimacing: cant think of more eloquent words, I do not mean to be brazen) modules. But believe me off the shelf functionality frees up a lot of resources, just as you described with the KOP chassis. We have come a long way since the age of drill motors.

If you are concerned about what you can control, your experience, there is nothing stopping you from taking your experience to the max with your own modules, or mechanism design, or custom fastener stress analysis. You will be better in the real world job market if you can bring a swerve module to an employer and say “I saw this project through in highschool”.

If you are worried about the community as a whole, as has been mentioned before, this thread exists in different clothes from years and years ago, when some people had concerns about loss of students experience from COTS gearboxes. Believe me, what we have now compared to 15+ years ago is WAY more impressive to the greater engineering world.

Finally, there are a lot of COTS solutions that are not perfect, especially for your ultra specific application. I encourage you to pivot that energy and iterate on exiting design work, prove you can understand the designs that came from another person’s head and work with those existing concepts to solve your specific problem. If I was hiring out of a pool of applicants for my engineering team that skill is very important.

Edit: it is possible I am off base with my last paragraph, if someone from the community is in the professional engineering management position and wants to correct/validate that last paragraph I would appreciate it. I am in the earth sciences and only do recreational engineering.

5 Likes

I think that the COTS solutions are so to speak a simplification or streamlining of the design process rather than a full blown solution. Its like the everybot where you start with A solution and then can build and iterate on that. The greatest thing about FIRST imo is that it is never the same, there is no one size fits all solution. Sure there are things that get you close or even 90% of the way there , BUT there is still design and engineering that needs to be done on the teams part to incorporate/make a COTS solution truly effective. Even with Swerve drives, manufacturing aside, integration is THE HARDEST part about any design and making everything fit and work is harder than anything I’ve ever done professionally and in FIRST. I have been on teams with money and resources and I have been on teams with hand drills and dreams. No matter which way you shake it there is still a challenge to incorporating COTS solutions into your designs. COTS parts offer a single solution in a program where there are so many different problems to tackle. At the end of the day you can buy every COTS part there is but that does not make it a robot.

2 Likes

Disagree, depending on the students and how they view FRC.

FIRST is, to some extent, pay-to-win. You have to deal with that and get what you can out of the program anyway.

I tend to agree that the proliferation of COTS swerves is unfortunate, but for almost for the opposite reason - I think a lot of teams that are purchasing these drives are ultimately not getting a competitive advantage out of them and are wasting their money, because using a swerve drive well is quite difficult (even with the amount of open-source code available these days).

If a team is getting a competitive advantage out of their swerve drive, then someone on that team has done a fair bit of work to get it to that point. They did not just “bolt it on and go.”

47 Likes

Instead of just saying my opinion, I would like to share the experience of my own team.

We have been competing since 2011, and have used a KoP chassis in every single year except 2022. This includes the 2021 at home challenges. During 2021, our team decided to finally make the move towards swerve. We had talked about it for years, but there was no good way for us to make that move. We don’t have access to any CNC mills or routers that can cut aluminum, and don’t have any sponsors that will machine parts for us. On top of that, we only started using CAD to lead our design process in 2019. Manufacturing custom modules was totally out of the picture.

The COTS modules were expensive, but the students organized and participated in lots of fundraising events to make sure the team had funds, and we ended up having the budget to order a set.

Resulting from that, the CAD team was able to get the experience in designing a custom chassis (we had never built a custom one before, only KoP) and the fabrication team was able to assemble the custom chassis, which required more than just cutting down C channel to size and dropping bolts in predrilled holes.

Our programming team had to learn how to program the robot to drive in teleop, figure out how to control it in autonomous, and attempt to control it with vision tracking systems.

In the case of my team, the COTS swerve modules enabled us to try something we had wanted to for a while, and allowed the students to learn many new things. I don’t think it took away any experiences, since we would have never been able to build our own kit. I am personally very proud of our students for what they have learned and accomplished, and even though they didn’t design it themselves, I think they are very happy with their experiences in using the modules.

12 Likes

So we are in a similar position - we spent a couple of years developing our own modules, tested them at an offseason. It was a great experience on many fronts, and we certainly could go down the path of continuing to develop them (and we probably will). But we bought SDS modules because as COTS, we can re-use them next year. The idea of building all new swerve modules + spares each season was going to cost way more than just buying a set, so that’s where we landed. So we spent more time on our climber and intake this year. It’s a tradeoff of time vs money for all teams.

My philosophy is development will now take place offseason - climbers, intakes, shooters, arms, drivetrains etc. Build what we can, buy what we need. I do see and get your point, but the times are a changing, and the best we can do is adapt.

My perspective on this is a little bit different. I was a rookie programmer in my freshman year during the offseason when my team was doing COTS swerve for the first time. I was also the only programmer on my team. This was early 2021, so COTS swerve was just becoming more popular and although the WPILib libraries definitely existed, swerve code was not yet “plug and play.”

When my team bought our SDS MK3s, I had never coded an FRC robot, I didn’t know how to use libraries, I had some understanding of Java syntax, and we were transitioning to command based for the first time. My school offers limited options for freshman math classes, so I was taking geometry and had no understanding of the vector math or trigonometry that happens behind the scenes in swerve.

I’m lucky enough to have an incredible programming mentor who, despite not being familiar with swerve, helped me find the resources I needed to understand it before programming it myself. We have a team rule, or actually more of an agreement, where we are only allowed to use code written by somebody else if we can explain it and all of the libraries it uses thoroughly. I spent weeks during practice and outside of practice reading every whitepaper and comprehending every example I could find online. This taught me the structure of a robot project, how to work with WPILib, math beyond what I’ve learned in classes even up until now, and it gave me the ability to program regular subsystems on our robots as well.

I don’t know where I would be in terms of programming and general robot understanding without the opportunities that COTS swerve gave me. This project was probably the largest singular thing that brought me into FRC and gave me interest in robots. It was also the most rewarding project I have ever worked on. It took from April until November of this past year to get it competition ready, and could not have been more worth it when we attended our first event as a team with swerve this past weekend. I see COTS swerve as something that creates opportunities for new swerve teams as opposed to detracting from the accomplishments of older ones.

22 Likes