pic: Usable Caster Drive Design



With this design, I wanted to create a caster drive that wasn’t massive, or enormous, or impossible to manufacture, so that maybe something similar could eventually be used on a competition robot.

I have also wanted to try using Onshape for a while now, so I took the opportunity to kill two birds with one stone. My experience with Onshape was a positive one in general, but it had some blemishes. I might go into more detail in a later post.

Here is a link to the CAD model. Because it’s in Onshape, you should be able to view it without downloading any software.

1 Like

Is the only difference between a traditional swerve and this being that the wheel is not inline with the axis of rotation?

Is this something anyone’s done before? because I have a similar concept to this involving the motor in module cad I post a few days ago

That’s the principal difference.

Is this something anyone’s done before? because I have a similar concept to this involving the motor in module cad I post a few days ago

I have been told that this type of dive system has been used in industry in a few applications. I think there is a related patent out there somewhere. I’ve been thinking about making an FRC size caster drive for for several years, but as far as I know, no FRC team has built one.

I love your packaging of this design, its so compact - it looks like you’re using the underside faces of the bolt heads to retain the main bearing and gear for azimuth rotation?

How well would this module (when implemented in a set of 4) stand up to the load of a fully weighted FRC robot and the rigors of competition?

Seems pretty straightforward to give the caster a suspension if you wanted to.

Well, there goes my afternoon.

Does this not still rotate around the Y axis of the bevel gear? I don’t see how increasing the distance in the x axis between the wheel shaft and bevel gear shaft does anything other than increasing the load on the rotation gear/motor since the wheel can’t pivot around it’s own forks the way a bicycle does because it is bolted to a flat plane in the frame.

Edit: because the steering shaft is still 90° to the floor.

Actually, the pivot on this module is still perpendicular with the ground, but bypasses the center of the wheel, so the castor angle isn’t quite relevant here.

The design and motivations are explained in this thread as well as this thread

TL;DR: a conventional swerve drive still needs to reorient modules for small movements, whereas a castor swerve can produce lateral movement perpendicular to the wheel through steering. This allows for very precise and instantaneous movement.

After reading that thread I see what the caster affect could theoretically do, but if you laid out a grid on a flat floor and had the robot drive forward then translate left or right without stopping, would the caster drive pass through less grids or just cut the corner slightly compared to a traditional swerve.

How is this any different that powering your drive wheels while rotating the swerve module at the same time, would both robots not move through almost the exact same arc.

This isn’t actually the intended use case. In fact, in such a use case, a conventional swerve module would likely be able to make a tighter turn, as the castor swerve also has to apply steering torque when accelerating perpendicularly with the robots motion.

The use case where this has an advantage is the one I mentioned. When the robot is stationary, a castor swerve can trivially move a few millimeters in any direction instantaneously, whereas a conventional swerve has to point the wheels in the correct direction first.

I’d bet that castor swerves would be at a terrible disadvantage in FRC, where precise movement has very little value, and where the “turn 90 degrees without stopping” use case is far more common. (when juking defense of course)

Thank you. Yes the inner race of the turning bearing is pressed and bolted onto a lip that’s machined into the top of the turning gear.

It’s very hard to say for sure, considering that there are no examples that I can turn to for reference. But I suspect that it is over engineered if anything.

That’s true.

It would certainly be a cool feature to add, but I’ve already designed and posted a crazy caster drive, so for this one I wanted to go back to basics, and design something with a minimal amount of machined parts. I might come back and design a suspension version, but not until I’ve seen one built and deemed suspension worth while, or I’ve become very bored.

The design has 10 parts that require machining.

1 large turning gear needs milled out.
3 milled plates one on top and two on bottom
1 small turning gear just needs faced down on a lathe.
2 bevel gears need bored and broached.
2 hex shafts need their ends turned round.
1 VP output shaft needs shortened.

Besides the 3D printed encoder gear, the remaining parts are COTS.

The advantage a true swerve has over crab is that you can do fancy movements by putting different angles on the wheels. Would caster swerve lose that functionality? Wouldn’t pointing one wheel at the different angle than the others and then bringing back require dragging the wheel sideways on the ground? I’m not quite sure whether that’s right (haven’t done the math), but in the extreme case when the 4 wheels are oriented like an X (top view), putting the wheels back to “drive straight” configuration obviously requires overcoming scrub.

Maybe it would be forced to behave like crab. Or maybe it would be like WCD skid steer and work anyways. What do you think will happen?

Someone please build one! :smiley:

That’s a very elegant and simple design, great work on all the packaging! I have drawn up several versions of swerve drive modules (just as concepts) and always put the goal to minimize the number of parts as the priority, but they still always have too many for my liking. Looks like you’ve done a much better job of it. The tucked away low profile of the whole module is a nice result.

Well, hop to it!

Nope. Say you point all of the wheels “in”, and want to go forward. The rear wheels begin to rotate “inward” and drive forward. The outer wheels rotate the same way, but initially, drive “backwards”. Once they achieve 90 degrees, they begin driving “forwards” again as they continue to rotate into place. The wheels themselves are always travelling in the direction they’re pointed. No scrub.

If you want to do a sharp 90 degree direction change on a standard swerve, you need to translate, stop, rotate module, stop, translate.

If you want to do a sharp 90 degree direction change on a caster swerve, you need to translate, stop, then do a coordinated module rotation + translation that smoothly blends into pure translation.

How much caster (offset) do you have?

One potential drawback of any caster swerve is that your support polygon moves well inside your frame perimeter in your direction of motion (depending on how much offset you have). That’s an argument for keeping the caster to a minimum.

So in a race on an L shaped path, assuming the exact same wheel and modual rotation speed the caster swerve would theoretically win by a very small amount?

If you’re racing, you’d never take an L-shaped path to begin with. :slight_smile:

I have not used Onshape before seeing this model, but it appears to have a measuring tool that you can use. Looks like ~2.6"

Yup. The race you’re most likely to see in FRC is “I’m going forward and now I want to go left as fast as possible.” For a conventional swerve, this time is only bounded by traction. For a castor swerve, this time depends on how much torque your motors can produce.