Team 8096 is proud to present our take on a differential swerve drive. It’s been in the works as a fun project for a while, and after being pushed to release it during the F4 awards stream yesterday, we decided to just do it.
The actual speeds and ratios of the V2 module that we’re releasing here (V1 exists, but the design was fundamentally flawed and we dumped it before it got completed) are extremely fast, but they are close to the realm of reasonable and we’re already fiddling with them to make them better for the V3 module. We have an overall drive ratio of 2.33 to one (which is far too high and a big point of the redesign) for a theoretical max speed of 33 fps. However, since controlling a diffy swerve is a bit different from controlling a regular swerve, we would run both the falcons at about 75% power. This is because the actual drive speed of a diffy swerve is controlled by the average of the two inputs and the turning speed is half the difference. So, running the two inputs at 75% power when we don’t want to steer allows us to have a 50% difference in turning speed before we change our drive speed. At this 75% power, we would in theory run at about 25 fps, and with that 50% difference in power, we would get a steering speed of about 130 rpm. Again, these ratios have some problems and are being worked with for V3.
With regards to the physical dimensions of the module, it is an 8.5 inch square, which is a bit bigger than is optimal (we realized just how much space it was taking up on our cadathon robot) that is 6 inches tall. It weighs about 6.3 pounds, which is neither horrible nor great, and has a 3.5 inch wheel, which makes our ground clearance very short at about 0.6 inches. We did get a 1:1 on axis absolute encoder on the module however. Right now the design calls for a CANcoder, but it could be easily modified to accept most encoders.
As for manufacturing, Lake Shore Drive is much simpler than most other diffy swerve designs we saw. There are only 6 non-COTS parts that can’t be made with a basic FDM 3d printer (all of the prototyping we’ve done so far has been on my Ender 3) and hand tools, and 2 of those 6 (the top and bottom plates) could feasibly be done in 2.5D on a router. The other 4 are the actual bevel gears, which are a bit more complicated to machine, but they could in theory be done on a 3-axis mill in only 2 setups. Because of this, and the fact that we decided to use bushings on the rotation bevels instead of bearings, Lake Shore Drive becomes shockingly cheap. The aluminum required for an entire chassis’ worth of modules can be purchased for less than $500, and combined with the small number of actual parts used, we estimate it at about $200 per module without sensors or motors, which is not bad all things considered.
Link to the grabcad workbench with the cad files: A Fond Farewell to GrabCAD Workbench - GrabCAD Blog