Shifting Swerve Drive
The design of a two-speed swerve drive was a challenge I took on after the 2020 competition season was canceled and would quickly become the most difficult design challenge I had faced at the time. As far as I know only one other shifting swerve drive had ever been manufactured and run in competition previously by team 1717 in 2011 and 2012. Since then more powerful brushless motors became available to FRC teams and shifting became less ubiquitous. Despite this the additional acceleration, the potential for a swerve drive to more effectively play defense (single-speed swerves generally geared too high), my desire to create a swerve design that the team could call its own, and of course the potential for learning made this a great quarantine project. Overall the entire process from idea to functional drivetrain took around a year due to COVID-19 restrictions. The design went through 2 standalone design iterations before I landed on a practical solution.
First Iteration
The first iteration had a lot of issues directly related to many of the fundamental challenges with a shifting swerve drive. This first iteration was meant to identify many of these dependencies and constraints as I had not actually designed a swerve drive from scratch before. In contrast with 2910’s Mk.2 design I decided to try an in tube mounted design rather than a top mounted design as I wanted to explore for myself some of the differences. This forced me to use a much smaller 3” wheel rather than the 5” pneumatic wheels. Furthermore, in order to get a functional spread between low and high gear, the pulley had to be much larger in order to support larger gears expanding the total size of the module and increasing its weight. Beyond this mounting the piston upright like a typical shifting gearbox forced the motors apart, further increasing the footprint of the module. Though technically functional, these modules would not be practical to manufacture mostly due to the tight tolerances in the custom dog shifter, the relatively large pulley, and the overall size and weight increase of the module would not have been worth the addition performance gained from the shifter over our already proven modified Mk. 2 modules.
Second Iteration
Having identified many of the constraints and central problems associated with designing a shifting swerve module I wanted to see if they would change or if new problems would arise if I tried a top mounted module. This eliminated the glaring issues with the smaller wheel and gear ratios allowing me to focus instead on the issues with the shifting mechanism and the overall size and weight of the module.
This iteration was also designed in a zoom call over the course of a few days where some of my teammates were exposed to the level of complexity and the problem solving skills necessary to take on such a design challenge.
Third Iteration
The third iteration of this module would be the first to be manufactured and eventually run on our 2021 robot. To start the main pulley and wheel assembly are identical to those in the Mk.2 modules, which greatly reduced the amount of machining required as our previous modules could be converted. Moving the shifting gear stack from the intermediate shaft that runs through the pulley to the extended motor shaft of the Falcon 500 motor allowed for the use of an off-the-shelf ball shifter increasing the manufacturability and reliability of the mechanism. Instead of a piston mounted vertically, a lever mechanism allows for the piston to be mounted on the underside of the motor plate reducing the vertical footprint of the module and allowing the motors to be closer together enabling more realistic gear ratios. For controls integration a stack of two photo-interrupters and a flag on the piston allows the robot to reliably detect when each module shifts enabling the robot to shift while being controlled autonomously without losing its position. The module features a magnetic absolute encoder to ensure the robot never loses the angle of the wheels even if power is lost for a brief moment during a match. The overall dimensions of the modules are only about a 1/2” larger than the Mk.2 modules in every dimension and the weight is comparable as well. To see these modules in action check out the FRC 2021 tab and watch our reveal video.
Fourth Iteration
The fourth iteration of this module is the one that is featured our 2022 robot. This iteration primarily focused on optimization by reducing weight, part count, size, and machining time. This was very important going into the 2022 season because even though the V3 modules were functional, we needed something more reliable and easier to repair. Some major changes were the removal of the absolute encoder as a viable software solution was found and a 3D printed gearbox cover (not shown). Because we would have to re-machine our modules anyway as per FRC rules, this module was designed completely from the ground up to optimize for weight. Beyond these changes I think the biggest takeaway from this iteration is as an example of an (almost) optimal design. The progression of these iterations shows how an optimal solution is often not the most complex solution, and how it takes a lot more effort to reach a simple solution than a complex one.