2084 Mecanums with Worm Drive

Video here of us driving and balancing with Mecanums and using AndyMark’s 16:1 worm gearboxes.

Old video, so don’t expect our final robot to look like this!

[strike]One thing struck me about that drivetrain design: mechanical efficiency.

With the mecanums being no better than 70% efficient in the forward direction (inherently due to the rollers), the worm gear being about 80% efficient (an educated guess assuming some things about the worm’s lead angle and lubrication), and the chain mesh being perhaps 98% efficient, aren’t you looking at overall mechanical efficiency (in the forward direction) of only around 55%?[/strike]

You can obviously climb a bridge effectively and move around on the ground (smooth floor, at least), so maybe it’s not such a big deal.

If you are using the word “efficiency” in the usual sense of power_out/power_in, the only loss in efficiency of mecanum wheels compared to standard wheels in the forward direction would be due to roller motion caused by roller axial free play or compliance of the flooring surface or the roller material or flexing of the vehicle. I don’t think this amounts to a 30% loss, but I don’t have any test data to prove it. In the ideal case (no roller axial free play, no roller bearing friction, no squeezing/stretching of the carpet or vehicle), mecanums have the same efficiency in the forward direction as standard wheels.

Efficiency hasn’t been an issue with our robot, so even though it does lose ~45% of the CIMs’ outputs, we don’t think we’re going to be in any pushing matches. Even if we were, other robots would be doing the pushing and the gearboxes don’t like that.

I’m excited to see more teams using worm drivetrains,

Can your robot be pushed with mecanum wheels?

I’d be a little frightened of a non-backdriveable drivetrain. I’d be especially worried about using the WormBox with its standard nylon wormwheel. At the very least, I’d put the robot on some carpet and shove until the wheels gave way to convince myself I wasn’t going to shear teeth in a shoving match. I’d really want to run a robot into your robot at a decent clip to make sure the shock load on those gearboxes isn’t going to destroy them.

I should probably have been more precise: that’s not an efficiency loss in the wheel, [strike]but only about 70.7% of the torque available at the wheel can be used for forward motion,[/strike] because of the sideways component of the wheel’s friction force. (I suppose this assumes equal frictional coefficients in all directions.)

Edit: The reference to “forward motion” only applies in the limiting case where the wheel is about to slip, and even then, the relationship above is not quite right. (See below.)

The Andymark WormBox is backdriveable from what I’ve seen. They had a test chassis at IRI last year, and was quite capable of being backdriven.

That’s a friction dependent limit, though. If your wheels aren’t slipping, then you’re getting 100% of the gearbox’s output torque applied to forward motion. Thus, your “efficiency”/pushing force calculation above is bunk. The drivetrain’s forward pushing ability is limited to EITHER 80% of the theoretical output torque of the gearbox or 70.7% of the available friction force, as calculated from mu and the normal force. The two limits are unrelated, so multiplying those two numbers together is meaningless.

We’ve done a lot of stress testing and haven’t found any problems. The gearboxes are backdrivable, but it’s very hard to do.

I will retract that expression—you’re right, that’s not valid for the case where the wheels are not slipping, which is most of driving around.

I was mistakenly conflating it with the limiting torque for mecanum wheel slip in forward motion, at which point you’ve got all of the available torque at the wheel spinning it, but not all of the frictional force in the longitudinal direction that you’d expect with a conventional wheel. (This is relevant when you’re figuring out if your motors will stall before your wheels slip.) And even in that case, as Ether and Kevin noted, it’s better expressed as a mechanical efficiency and a geometry constraint. They end up multiplied (or more accurately, divided) in the equation, but they’re not part of the same quantities.

But even if the gearboxes are able to handle it, there probably is not enough frisction.

Not enough friction for what? If a robot tries to push us sideways, the wheels won’t turn.