Valid Configuration? Kiwi with mechanum wheels

[jkelleyrtp]

I've been going through a thought experiment for a while of a new holonomic style.
In the classic kiwi setup, the front two omni wheels are angled inwards like this with the goal of resolving movement vectors in x,y,ω. I have experience with this kind of robot but none with mechanum wheels. I think that mechanum wheels drive similarly to 4 wheel holonimic systems but will only move in the left in right directions if the forward component vectors are eliminated.

So the goal would be to replace the front two wheels in kiwi with forward facing mechanum wheels like so:


Here is my very poorly done vector diagrams of what I think will happen. I do realize that the angle of the rollers on the mechanum wheels will change the component vectors and have to be addressed in chassis shape, but for now I'm only looking for feedback on the idea.

[GeeTwo]

See this thread for comments on a drive of this type.

One of the curious things that came out of my q&d analysis was that this can travel much more effectively to the left/right (that is, parallel to the axes of the mecanum wheels) than forward/reverse (parallel to the axis of the omni). More tonight.

[Ari423]

This idea will work fine, but why do it? The only thing I can think of is you save the money on one wheel, gearbox, and motor. I can't think of any other benefits. Some negatives are less driving power and less ground contact. A drive like this without actuation will make going over obstacles very hard. If you do want to do this, I would take a look at 1114's notes from this year on how to do a good kiwi drive.

[jkelleyrtp]

Originally Posted by Ari423

This idea will work fine, but why do it? The only thing I can think of is you save the money on one wheel, gearbox, and motor. I can't think of any other benefits.

It wouldn't be its own drive system, but a dead simple way to do "ocatocanum kiwi."


In a system like this, you would only need to force down the rear omni to put yourself on the appropriate strafe wheels. The forward mechanums would be off the ground just a tiny bit and chained to each side's gearbox.
Therefore you would only need to add one motor, omni, and piston (or one versadrop) and have a butterfly style drive as opposed to this.

[Anthony Galea]

Originally Posted by Ari423

This idea will work fine, but why do it? The only thing I can think of is you save the money on one wheel, gearbox, and motor. I can't think of any other benefits. Some negatives are less driving power and less ground contact. A drive like this without actuation will make going over obstacles very hard. If you do want to do this, I would take a look at 1114's notes from this year on how to do a good kiwi drive.

Another reason could be ease of building. At least to me, it seems a lot easier to build (and less errors possible) with only having to work with 90 degree angles instead of 60 degree angles. Also, you take advantage of a rectangular frame instead of a triangular frame to build off of, which is nice in games where you have to fit into a certain space instead of having a set perimeter size restriction.

[DanielPlotas]

Originally Posted by jkelleyrtp

The forward mecanums would be off the ground just a tiny bit and chained to each side's gearbox.

Would the mecanums have to be off the ground? If they're chained to the traction wheels and the same diameter (as appears in the render), would they not be moving the same speed? Or is there another reason for them to not be touching that I'm just missing?

[z_beeblebrox]

Originally Posted by jkelleyrtp

It wouldn't be its own drive system, but a dead simple way to do "ocatocanum kiwi."


In a system like this, you would only need to force down the rear omni to put yourself on the appropriate strafe wheels. The forward mechanums would be off the ground just a tiny bit and chained to each side's gearbox.
Therefore you would only need to add one motor, omni, and piston (or one versadrop) and have a butterfly style drive as opposed to this.

Why couldn't you just use 3 omni wheels instead? You still have 3 DoF movement without the complexity/ weight/ size/ cost of the mecanum wheels.

[jkelleyrtp]

Originally Posted by DanielPlotas

Would the mecanums have to be off the ground? If they're chained to the traction wheels and the same diameter (as appears in the render), would they not be moving the same speed? Or is there another reason for them to not be touching that I'm just missing?

More in an effort to consolidate traction than anything else. If you place them close enough to another traction wheel though, it shouldn't impact your pushing performance.

Originally Posted by z_beeblebrox

Why couldn't you just use 3 omni wheels instead? You still have 3 DoF movement without the complexity/ weight/ size/ cost of the mecanum wheels.

The vectors don't resolve the same way meaning you couldn't have a perfect left-right strafe. It would be more circular movements, similar to H drive with only the front wheels touching.

[Ari423]

Originally Posted by jkelleyrtp

It wouldn't be its own drive system, but a dead simple way to do "ocatocanum kiwi."


In a system like this, you would only need to force down the rear omni to put yourself on the appropriate strafe wheels. The forward mechanums would be off the ground just a tiny bit and chained to each side's gearbox.
Therefore you would only need to add one motor, omni, and piston (or one versadrop) and have a butterfly style drive as opposed to this.

I'm not sure that will be as easy or work as well as you want it to. To get the traction wheels off the ground while keeping the mecanums on the ground, the mecanums will need to be on the front corners of the robot. If you leave them there, they are exposed to being damaged by anything you drive into. If you build an extended front perimeter, it will take time to design and build the front perimeter. At that point, you might be cutting off your nose to spite your face.

I do see the advantages over normal kiwi drive, but I don't really see them over mecanum. And if you want to do octocanum, I feel like it would be easier to do it the usual way.

[z_beeblebrox]

Originally Posted by jkelleyrtp

The vectors don't resolve the same way meaning you couldn't have a perfect left-right strafe. It would be more circular movements, similar to H drive with only the front wheels touching.

Why not? In this force diagram(not to scale), the torques from the front wheels cancel the torque from the back one. The center wheel creates a sideways force while the fore-aft forces from the front wheels cancel.

[GeeTwo]

This is certainly a "boutique" configuration, only useful for a limited set of design requirements. I'll leave out the "articulated" version except to say that I'd be more comfortable articulating all three holonomic wheels rather than just the omni. If I did articulated, I might also consider doing some sort of PTO manipulator with the omni wheel to optimize motor use. Here are some of the design requirements this would meet:

1. You require being able to drive in ay direction, even if it means giving up traction. This either means that you think defense wil be light, or that you will be fast enough to get around it.
2. You would rather have only three drive systems (otherwise you'd do mecanum or octanum).
3. You require a rectangular chassis, either because of your technical limitations, or because of constraints imposed by the game and/or your game strategy.

Recycle Rush is probably the only FRC game so far where teams were likely to meet enough of these criteria to want to do this. The large rectangular game pieces and the lack of interaction with opposing robots could easily meet two of these requirements.

If you do decide to build a robot to this design, the inverse kinematics are the same as for kiwi, except that the angle for the two wheels not on a coordinate axis are 45 degrees rather than 30/60. Ideally, you'd want to have the center of gravity along the axis of the omni wheel, trailing the axes of the mecanum wheels internally by 45 degrees, and the same distance from each wheel's contact patch. For this, we can adjust the inverse kinematics from Ether's kiwi paper, adjusting an angle, and letting the omni be wheel 1.

Originally Posted by Ether's Kiwi inverse kinematics, adjusted for meckiwi (italics)

X_j, Y_j, & Z_j are the 3 joystick axes

• forward = Y_c = -Y_j
• strafe right = X_c = X_j
• rotate clockwise = rw = Z_j

• wheel1 = X_j + Z_j
• wheel2 = -0.707 X_j + 0.707 Y_j + Z_j
• wheel3 = -0.707 X_j - 0.707 Y_j + Z_j

[GeeTwo]

Originally Posted by z_beeblebrox

Why not? In this force diagram(not to scale), the torques from the front wheels cancel the torque from the back one. The center wheel creates a sideways force while the fore-aft forces from the front wheels cancel.

To obtain a "pure strafe", the wheels at the top of the image would have to apply torque with out actually rotating. It can be done, but it would not just be standard FRC hook up of controllers, motors, gearboxes, and wheels.

[EmileH]

Originally Posted by z_beeblebrox

Why not? In this force diagram(not to scale), the torques from the front wheels cancel the torque from the back one. The center wheel creates a sideways force while the fore-aft forces from the front wheels cancel.

This year 1658 used a similar drivetrain design, except with them having two rear omni wheels for strafing instead of one, and they had the strafe wheels on the front of their bot. A video explaining the drivetrain is here.