5687 developed a differential swerve last year during the pandemic. It was in large part driven by one of our programming students, who did an amazing job developing the controls for it. You can find a link to our CAD on a CD post we made last year: 5687 2021 CAD release
Before I get into any details about the swerves, I should probably share some of the lessons we learn while developing them. Differential swerves require a lot of work to develop, and most teams would not see any benefit from them. We build heavy robots that are usually right at max weight. Even still our swerves are traction limited. Given how powerful the brushless motors are, if your robot is not at max weight a traditional swerve is probably traction limited as well, so the extra torque is wasted.
One of the major down sides of differential swerves is that turning becomes difficult to control at high speeds. When turning, the robot is essential a differential - add differential swerves to that, the the difference is speed of the fastest motor on the swerve located on outside of the turn, and the slowest motor located on the inside or the turn become very large, and it becomes impossible to turn rapidly at high speeds.
However, there are some benefits. We definitely notice the added traction when pushing (again, we are at max weight), and the acceleration is impressive. Also, each motor draws little current - under normal operation a little over 10 amps, and underload they usually run at 20amp. Additionally, each motor is operating in a more efficient region of their power curve. Lastly, because the load is shared between two motors, these modules have held up very well. We never had to replace or even repair a module all year (more on this later).
Would we recommend developing a differential swerve? For most teams no. I don’t think there is much of a performance advantage (unless you really need the torque), and there are significant disadvantages. Additionally, they take a lot of time to develop properly and that energy that would go into fussing with the swerves is probably better spent elsewhere. However, if you have some students that really want to spend a few years tinkering with mechanical drive systems or explore complex control systems, then it offers a great learning opportunity.
About our swerves. Our priorities in designing them were:
- Keep COG as low as possible
- Keep top below the standard bumper height of 7”
- Maintain a 1.5” ground clearance
- Use as many COTS and 3D printed parts as possible
- The translation (wheels) ratio is 6.47:1 and the rotation (steering) ratio is 9.2:1
We have gone through about 5 interiations on the design, and will probably do one or two more this offseason. Last year we were constantly breaking components, and each iteration was more robust, until the current interaction, which seems to have held up well. But, that is what made it such a great project to do while we were remote during the pandemic. Below are some CAD images: 
Here is a section view:
And here is a picture of our fist iteration (I don’t have a picture of our current iteration right now):
Please let us know if you have any questions.
