The in-wheel swerve shifter!

[Greg Woelki]

Originally Posted by asid61

So they need human input then? I swear I saw autoshifting versions.

I don't have time to look into autoshifting versions right now, but if there are hubs that shift by them selves at certain loads, then they wouldn't work very well even in a test drive train because they would shift under totally different loads than they would experience in this application. My back-of-the-hand estimation of the torque on a bike wheel in low gear puts it at dozens of ft lbs, much different than the maximum of 8.3ft lbs per wheel in a swerve drive on a 154lb robot with 4in, 1.3 coefficient of friction wheels. Additionally, even if it did shift at appropriate loads, the load on each wheel would be different not only in pushing matches and collisions, but also normally because of the robot's weight distribution and the different amounts of power each drive motor runs at during different maneuvers, so you would end up with modules in different gears sometimes.

[Chris is me]

Originally Posted by asid61

So they need human input then? I swear I saw autoshifting versions.
I'm opening up the CAD so I can see what happens when you slow down.

EDIT: yup, the dog bounces back and forth like crazy while rubbing against the shifting elements when you slow down, which would completely destroy the teeth on the mating elements. I'm going to find a solution to that.
DOUBLEEDIT: So what would happen is this:
1. The dog slows.
2. The dog is pushed into the other gear by the obtusely angled faces.
3. The obtusely angled face there pushed the dog back.
4. Repeat steps 2-3 until the dog contacts the flat of the teeth and cannot shift out of its current gear.
5. The gear that it is in is forced to slow down slightly until the dog bounced back into the other gear.
Repeat until slowed. There would be a lot of bouncing, but nothing too high load except for the rubbing at the end. Another thing to test under load, although this is a much more serious issue.

Of course, that doesn't matter if you are doing something at a constant 100% speed like loading a spring or climbing, but for drive this seems like a problem.

This is what I was trying to talk about earlier. As the dog oscillates during slow down / direction changes, the gears are being alternately loaded almost all of the time, though the duration of each individual load condition might not be that high. This is going to resolve as inertia that prevents the CIM from fully slowing down as quickly as a CIM under no load, unless there's a very significant delay between shifting halves.

Another condition that might behave oddly is what happens if you shift under load, i.e. to begin a pushing match. While the dog is "in between" gears, your wheels offer essentially no rolling resistance. What if a wheel is spinning the opposite direction you're trying to drive it? It seems like no matter how long the shift window is, you're going to be dealing with either or both of these problems to varying degrees.

I ain't trying to hate on your design, just trying to figure out if there are ways around these issues.

[Seth Mallory]

Since the S3X hub has an indicator chain for shifting you could mount your shifter up in the swerve drive above the pivot bearing. What you use for a shifter is up to your imagination. One problem is that the 4 wheels would shift at different time do to load on the gears and meshing of the gears. It should be ok for telly op but not for autonomous shifting. The Sturmey Archer hubs that have automatic shifting would not work well for our robots. They would tend to shift each time you slow down. I do not think this hub would work for competitions but it could lead things in a interesting dirrection.

[asid61]

Originally Posted by Chris is me

This is what I was trying to talk about earlier. As the dog oscillates during slow down / direction changes, the gears are being alternately loaded almost all of the time, though the duration of each individual load condition might not be that high. This is going to resolve as inertia that prevents the CIM from fully slowing down as quickly as a CIM under no load, unless there's a very significant delay between shifting halves.

Another condition that might behave oddly is what happens if you shift under load, i.e. to begin a pushing match. While the dog is "in between" gears, your wheels offer essentially no rolling resistance. What if a wheel is spinning the opposite direction you're trying to drive it? It seems like no matter how long the shift window is, you're going to be dealing with either or both of these problems to varying degrees.

I ain't trying to hate on your design, just trying to figure out if there are ways around these issues.

The dog is never "in between"; it can technically be in both gears at once. However, in that position, there will always be a force that will shift it one way or another into drive.

Hm. I'm looking at the cad again, and I'm seeing something interesting. The dog will actually very quickly "get stuck" in both gears if the dog is moving slower than the gear/hub it is driving. That means that the speed will instantly drop to whatever lower speed the dog is moving at. This would occur once every 1/3 of a rotation maximum, so at ~1400rpm shaft speed there would be 70 of said speed collisions each second. I would want to include part wear in loaded tests to see how much this affects driving. Code would have to allow for taking more time to decelerate, or I could just live with the collisions; that's what shifters and gears do usually when stopping.
Thinking about it some more, and looking at the cad again, it looks like the dog would just shift into the original gear whenever it reached the intermediate position.

Because the cim is putting force in the opposite direction when you shift (reverse cim direction), it will actually shift into the other gear because there is much less resistance that way instead of locking up. I wouldn't go into low gear unless I was in a pushing match anyway or had no load on the robot.

Or a stupid solution: Turn the excess speed into weird snake turns so that you don't have to deal with decelerating at all! Just keep all cims at 100% or -100% constantly. And when you're sitting in place, you can just turn in place!
I don't even want to think about the battery load that would entail.

In the pushing match situation you describe, there would be no pushing match. It would be another robot pushing you in the direction you wish to go, only faster than what the swerve can go at. If the wheel is actually spinning in the opposite direction for some reason, then I would flip the wheel around. More interesting code situiations.

No, I appreciate the input. It would suck to make this, then see it fail horribly and explode when I try to stop.

This swerve would have to have an onboard controller like an XMOS to do all this detection.

Something that's really similar to this is the zeroshift for F-1 racers, but it uses manual shifting with multiple dogs. The angled teeth are the same though.

[asid61]

Originally Posted by Seth Mallory

Since the S3X hub has an indicator chain for shifting you could mount your shifter up in the swerve drive above the pivot bearing. What you use for a shifter is up to your imagination. One problem is that the 4 wheels would shift at different time do to load on the gears and meshing of the gears. It should be ok for telly op but not for autonomous shifting. The Sturmey Archer hubs that have automatic shifting would not work well for our robots. They would tend to shift each time you slow down. I do not think this hub would work for competitions but it could lead things in a interesting dirrection.

The point of this was to have the shifter be compact, not require pneumatic tubing and not take up space outside the swerve module. I want the cim to be mounted above the module instead of being offset by chain.

[asid61]

Out of curiosity, would the loads experienced by the dog when stopping be the same as the loads on a normal gear or dog when stopping? The dog gets locked in between gears in this until the side it is spinning with spins with the dog whenever you declerate, just like in a normal gearbox slowing down.