Crab drive steering
 

Last year, my team built a coaxial crab drive with all the wheels chained to gether for steering and separate gearboxes for left and right. We are in the midst of modifying and improving our design.

If we were to chain 2 modules together to steer, would it be best to link fronts and backs, rights and lefts, or opposing corners? What are the bnefits of different set ups?

Last year, our frame was 27 inches square, I think we did this thinking it was better for swerve drive, but is this necessary? We should be able set up the swerves in rectangle and account for different angles if necessary in programming, right?

Is it better to use an optical encoder or a potentiometer for sensing the direction of your modules? One problem/inconvenience we had is that the wheels would have to be pointed forwards when you turned on the robot otherwise the wrong direction was forward for the joysticks (we were using an encoder).

Finally, I ws wondering if there was an advantage to using belts to steer instead of chain becaue belts don't stretch. Would this be of any advantage?

Any advice would be welcomed.

Re: Crab drive steering
 

To start, you had a really slick drive base this year that was a good start with swerve/crab as well as with the machining resources you've gotten nailed down. And its even better that you're focusing on improving it.

Which wheels you steer together depends on the motion you want. If you want to turn around a point in front or in back, you need the sides chained together, and vice versa. However, the "unicorn" drive as it has been dubbed here is certainly preferable in its unlimited ability. This is where you steer all four (or three) independently. For either case, you will need to work through some more advanced kinematics/dynamics to figure out the heading and speed for each wheel.

With four wheels in a rectangle and the center of mass in the geometric center of your wheel base, you get the same torque canceling as from a square base. Lots of teams do this with no extra math involved to drive correctly. Also, now that you have one dimensions longer than the other, you have the ability to turn if the wheels are in line with the short side, because the wheel base is wider than it is long. I believe you had some issues turning at first with your square base at first so you put in two omnis, which gives you less traction. Square bases are usually just for true crab where you have no need to turn the wheel base at all (manipulator is turreted or nonexistent).

The kit optical encoders (and all others) are incremental encoders, meaning they only measure change. This is not terribly useful if you have two separate steering motors you need to point in the same direction, say, at the beginning of a match. Absolute encoders are the way to go here, since they give an absolute heading of the module(s). Potentiometers and magnetic encoders both work for this. Mag encoders have no gap in readout.

Many teams have also used belts for steering because, as you said, they don't stretch, and they also give more precise motion transfer, useful for steering modules precisely. They are however more expensive and probably heavier.

Re: Crab drive steering
 

No.

Right. The equations can be found here.


You could use an absolute encoder:

Austria Microsystems AS5030 absolute magnetic encoder.

US Digital MA3 MA3 absolute magnetic encoder

Cherry AN8

Vishay 981 HE

Re: Crab drive steering
 

The wheelbase and trackwidth do figure into the inverse kinematics.


Not all optical encoders are incremental. There are absolute optical encoders.

Re: Crab drive steering
 

When calculating the omega cross r term yes (for a unicorn drive, in which case you're already doing that math anyway, you just need to plug in a different r vector for wheel position), but for standard crab, you can drive in a straight line in any direction with all four motors at full speed with the wheels in the orientation I described. I haven't rigorously proven it, but my own intuition regarding the torque exerted by each wheel, it shouldn't be an issue.



Good to know. I've only ever seen/worked with incremental ones, and absolute clearly have their advantages. I still like mags better though :P. However, both have their strengths.

Re: Crab drive steering
 

What this means is that you can drive all wheels from a central "powerbox" in any direction on a crab drive without necessarily having a rotationally symmetric wheel base. You just have to make sure there is no net torque.

Re: Crab drive steering
 

This is correct. The context was a bit ambiguous your earlier post.

Re: Crab drive steering
 

Thank you everyone for your help so far. I will look into the different sensors and the agorithms posted.

I guess I know that intuitively that turning is easier with a wide base, but what is the reason behind that?

Also, I believe you are right that belts are more expensive, but according to Cyber Blue 254 in their belt vs chain presentation, belts are actually lighter and more efficient.

Re: Crab drive steering
 

Because with a wide base there is less scrubbing of the wheels.

Think of a vehicle that is 3 feet wide and only 6 inches long. When it is skid-steer turning, there is almost no scrubbing of the wheels, because each wheel is moving in the plane of the wheel.

Now think of a vehicle that is 3 feet long and only 6 inches wide. When it is skid-steer turning, each wheel is moving sideways (perpendicular to the plane of the wheel).

Here's a sketch.

Re: Crab drive steering
 

Also very important, the wheel can actually exert a significant tangential force. Note how in the wide configuration, the wheel is nearly aligned with the direction of motion, so when you turn you get a larger component of the force.

Re: Crab drive steering
 

Absolute encoders are the way to go in my opinion. The version from US Digital previously mentioned is very good.

I particularly like it for co-axial systems as the deadband is sufficiently small that you can control the modules through 360 degrees or rotation in a direct measurement application. (sufficient enough for a competition robot anyways)

US Digital Abs Encoder on a co-ax swerve bot The encoder is linked to a tensioning sprocket and is below the chassis rail....attached via a helical beam coupling for anti-backlash and mis-alignment tolerance.

As for steering, I like linking the front two wheels and the rear two wheels. Our standard setup is to link drives on the left and right and steering on the front and back. This gives you immediate access to tank steer, car steer, monster truck steer and regular crab modes. You don't get diamond steer or King Krab abilities....but I'd argue that they have limited uses for competition.

Re: Crab drive steering
 

For measuring steering angle on swerve, absolute encoders should be used. Ether listed a few. For 2011 we used the Vishay 981's. They performed well. Automation Direct has a plastic beam coupler that is inexpensive and allows for some misalignment and will help protect the encoder. In 2010 we used the Cherry AN8's. They are a true non-contact encoder and also worked well. The absolute encoders that have been listed are analog output. They output a voltage between 2 voltage end points. These end points are dependent on the analog break out's 5 volt regulator. The .5 and 4.5 volts are common end points. If the supply voltage is above or below 5 volts these endpoints will move. Keep this in mind when programming the 0 - 360 transition. The US Digital are 0 - 5 volts. Their output gets sloppy at the rails (0 - 5). That coupled with their susceptibility to electrostatic discharge are the reasons we did not use them. Pay attention to the electrostatic rating of these sensors. A discharge can mess up some of these sensors and if your encoder goes your robot will look like it is having a Grand Mal seizure. Choosing a sensor that is automotive qualified is a good idea. There are also absolute sensors that output a 10, 12 or 14 bit reading over a SPI buss. We are looking at some of these for this year. The advantage is that on some of these the the 0 - 360 transition can be programmed leading to simplified calibration procedures. Last, we paid under 30$ for our sensors in 2010 and 2011. These same sensors are in short supply and the price has more than doubled. Must be a Japan issue from the earth quake. If they will be needed in 2012, then keep an eye on the availability and price this fall.
As to swerve, Our team believes that independent 4 wheel drive and 4 wheel independent steering modules are the only way to go. I don't have time to go in to a rant why we believe this but this link will show what we have done for swerve.
http://wiki.team1640.com/index.php?t...II_Drive_Train

Re: Crab drive steering
 

Please define what you mean by these two terms.

Thanks.

Re: Crab drive steering
 

Monster truck steer = front and rear steering going in opposite direction similar to a monster truck.

Diamond steer = all wheels turning 45 degrees with respect to forward allowing the vehicle to drive about it's own center axis.

Re: Crab drive steering
 

I have no direct experience with "Monster Trucks", so your definition is a bit unclear to me.

Do you mean something like Figure 1?

or Figure 2?

or something else?

Re: Crab drive steering
 

EDIT: This is just a guess, but from what he said I think he means the type in figure 1. (Not A)

Re: Crab drive steering
 

What's "Type A" ??

Re: Crab drive steering
 

In our situation, because two wheels are linked, Ackerman angles cannot be achieved....so we're most like Figure 1. Because left and right drive are independent we do slow down the inside wheels with respect to turning angle.

A real monster truck is probably most like Figure 2.

Re: Crab drive steering
 

It might be useful to facilitate possible future discussion in other threads if you could assign a unique name to your steering method.

Did you determine the relationship between the steering angle and the skid-steer compensation empirically or by analysis?

Re: Crab drive steering
 

Our algorithm was purely empirical. We set up a linear relationship between angle and "detuning" of the inside wheels.

More important to me isn't that we have Quadrasteer (formerly Monster Truck)...it's that you can access tank, crab and Quadrasteer from the same steering setup. Our control scheme uses two sticks on a PS style gamepad.

Tank:
Left stick up/down is driver side throttle
Right stick up/down is pass side throttle

Quadrasteer:
Left stick up/down is all wheels throttle
Right stick left/right is steering

Crab/swerve:
Left stick up/down is all wheels throttle
Right stick left/right is all wheels steer angle

Each of these modes is access by holding a top or trigger button....I prefer to have it setup as momentary.

Re: Crab drive steering
 

I'd be curious to know how your empirically determined speed reduction compares to a rough analysis I just did. See attachments.

I assumed that the center of rotation of the vehicle will be halfway between the points A and B.

Multiply the "outer" wheel speed by the fraction at the end of the analysis to give the "inner" wheel speed. L is the vehicle wheelbase, W is the trackwidth, and Ø is the steering angle, clockwise from straight ahead.