Swerve Drive Rotation Control

If we were to do 1+4 &2+3 steering, would you suggest 1+3 &2+4 Drive? I cant really think of too much of a difference with which pair is driven. I would probably go with 1+3&2+4 because it frees up the front for game piece manipulation and such.

You pretty much need to have each wheel be powered independently if you want to get decent benefits out of using a swerve drive. Steering is somewhat less important (a small amount of slippage is fine) so you can usually tie pairs of wheels together for steering. Diagonal is best.

Crab Drive involves tying all wheels together for steering (i.e. all wheels will face the same direction). 118 did this in 2008 and all wheels shared power. They way they dealt with rotation was by putting their manipulator on a rotating turret. The chassis of the robot didn’t rotate at all during a match if I remember correctly.

I don’t really understand why you need to power each wheel independently if you want to get decent benefits. Could you please elaborate?

The reason we were thinking of driving two wheels together is because our team has very little access to cnc and advanced manufacturing techniques. If we were to drive all wheels independently we would have to use 4 pre-made gear boxes. By chaining wheels togeather we have half the weight and half the complexity.

Well let me ask, what kind of benefits does your team want to get out of a swerve drive as opposed to a normal tank-style drivetrain?

What we really want to get out of a swerve drive is the ability to have omni-directional movement. Our team used mecanum wheels for many years before switching over to a wcd this past year. Although we really like the traction and pushing power we gained, we still would like the ability to have omni-directional movement. With the swerve drive having both aspects it seemed like a next step for our team.

I guess what I was trying to ask was why would you drive each wheel individually if you steer two together?

If all you’re looking to do is have a robot which can drive back and forth and strafe left/right, then you don’t need individual power on each wheel. I just suggested that if you wanted to do more complex things like move and turn at the same time.

Steer 1+2 together, and 3+4 together.

It just makes more sense that way (to me), because if 1+3 and 2+4 are steered, and you steer them outwards and drive them forward, …well, it won’t do much good.

Here is a good example: http://team221.com/upload/818-layout.jpg

You may also want to study images of 118’s drive systems…

I see a lot of people mention this as a factor for going swerve over say mecanums or an omni drive but some of the real reasons with going WCD is finishing early, having a more simple drivebase to program/drive on, and overall having a more reliable system. Don’t fall into the swerve trap to get “more traction”

Most of the time, pushing matches can be avoided with just better driving, which boils down to finishing earlier and giving more stick time to the drivers.

This is exactly right. Also if anybody is thinking of going independent swerve there is a chance that some of your wheels leave the ground in pushing matches causing you to lose 25% - 50% of your power.

1+2 and 3+4 vs 1+3 and 2+4 will get you the exact same thing. Consider the game though. If you have to bring a game piece up the middle of your robot then 1+2 and 3+4 is probably the wrong choice. However I highly recommend 1+4 and 2+3: 1717 used if for 2 years on their swerve and it allows you to turn very efficiently.

To respond to your original question, the simplest solution is to tie all four modules together for steering. You are going to have to provide some method of getting all the wheels aligned but then you should be good. We have made independent steering when we wanted to achieve something special. For instance, strafing in an arc for Rack and Roll. When discussing power, most teams will use some form of transmission built into the wheel module as external drives can be less efficient. The downside of crab steering designs is the weight and space needed for the system and the complexity of programming and feedback. Above all else, the best crab is only as good as it’s drivers so practice is essential. Lot’s and lot’s of practice.

Although simple, you will lose the ability to change your robots orientation in relation to the field, which may pose a problem depending on the game and how the rest of your robot is built.

I’m not going to tell you what is best for your team, but think about the different pros and cons of connecting the halves of the robot and how that effects your steering/turning/driving etc. Also, as said before, consider how your chain paths effects your robots design/ability to play the game. What you prototype now might be modified for your competition robot because its not optimal for the game.

Quasi-related: Good for you and your team! Building a swerve is a cool thing. They are a big learning experience for all involved regardless to implementation on a competition robot. :slight_smile:

Not necessarily. If the drive power is tied 2-and-2 on the sides, or each module is individually powered, it shouldn’t be terribly difficult to program a mode where the swerve drive turns like a tank drive. As a matter of fact, the OP was asking about this very power setup.

I was referring to Al’s statement about chaining the steering of all the wheels together. Unless you’re saying they don’t power one half and do a skid-ish turn of some sort, which is something I’ve never heard of. Could you please elaborate?

EDIT: As I stated in my post after this, I was over-thinking the concept and making it out to be something far more complicated. I am now embarrassed.

Exactly what I’m talking about. You can use a 2-and-2 power arrangement to simulate a skid-steer. You’d only do it as one mode of driving; normal operation would be to drive like a swerve. But for a quick reorientation, a skid motion can help. (Turrets are also an option, but that’s another dose of mechanical complexity.)

EDIT: It was pointed out to me that I knew exactly what you mean, and was instead imagining some completely different method of turning. Original content removed for pride’s sake.

It’d be nice to list exactly which maneuvers you give up as you tie more and more degrees of freedom together.

[EDIT: The paragraph that was originally here was all kinds of wrong-- I’m going home.]

Thanks for reminding me. We have used a descending foot, pneumatically operated to lift two wheels off the floor to make orientation changes. In some designs the foot became an AndyMark omni wheel(s).

I’ll take a rough empirical stab, and then let the Ethers of the world codify the Unified Field Theory of Swerves.

First off, some assumptions. Most types of swerves can pull off turn in place/turn while translating by scrubbing wheels ala a tank drive. A “scrub-less” turn is in theory more efficient (though in practice you might find more overshoot makes it harder to control without software help). In many cases (long robots + sticky wheels), scrubby turns turn into hoppy/breaker popping turns. You might need to rotate your wheels to be along the “wide” axis for scrubby turn in place to work well. So there is a very gray line separating maneuvers an arrangement can vs. cannot do.

Here are some common swerve steering arrangements:
A. All four wheels independently
B. Front/rear or left/right wheels driven together
C. Diagonal pairs of wheels steered together
D. All wheels steered together

Here are some common swerve driving arrangements:

  1. All four wheels driven independently
  2. Front/rear or left/right wheels EDIT:steered together
  3. All wheels driven together

(Note: You could end up driving diagonal pairs of wheels together, but I cannot think of a single arrangement where this has ANY advantages over another arrangement with a similar motor allocation…hence I omit it for readability)

All possible combinations can do strafing in any arbitrary axis (f/b, l/r, or anywhere in between).

Now let’s play bingo!

A1: You can pull off any maneuver - you are holonomic (assuming unlimited pod rotation and unlimited rotation speed). The “Unicorn” drive.

A2: You incur more wheel scrub and would need to rapidly rotate individual pods 180 degrees during compound maneuvers (such as translate + rotate). Worst case is forward/backwards + rotate if the front/back pods are driven together, and strafe + rotate if the left/right pods are driven together.

A3: A lot of scrub any time you aren’t driving in a straight line or turning in place, but you can still do some low-speed rotation while translating.

B1: You give up car/snake steering during strafes and perfect turn in place. You can still turn differentially while translating if the drive geometry supports it.

B2: If the same pair of wheels is driven and steered: You get a tank drive in one axis (f/b or l/r) and scrubby “snake” steering in the other (since the inside and outside wheels of the car are being driven at the same speed). If opposite pairs of wheels are driven and steered (example, steer f/b, drive l/r): You get great snake steering in one axis, almost no ability to turn while translating in the other, and scrubby turn in place.

B3: You can in theory get differential turn in place with scrub, and some measure of snake steering at shallow wheel angles.

C1: You get perfect turn in place, but can only rotate while translating via differential steering.

C2: You get perfect turn in place (through reversing motors), but no effective rotate while translating.

C3: You can’t turn. Period.

D1: You turn just like a wide or long tank drive (and can switch between the two).

D2: You are a wide OR long tank drive that can also strafe (but cannot rotate while doing so).

D3: You can’t rotate at all. Time to build a turret. See: 118, 2007-2008. 148, 2008.

In conclusion, arrangements A1, A2, B2, C1, C2, and D3 seem to be the most worthwhile to me…
A1 is the Unicorn…you get everything at a high motor/sensor cost.
A2 is still pretty versatile if you want to switch between single axis “car steer”, rotationless strafing, and decent turn in place.
B2 removes 4 motors from A1 (2 from A2) and sacrifices scrubless turn in place.
C1 preserves quick and easy turn in place while only requiring two turning transmissions/sensors.
C2 is great if you never want to rotate + translate at all.
D3 theoretically can be done with only 2 motors…as long as you don’t need to rotate.

Feel free to add to the notes! I’m just I’ve missed plenty of (counter) examples to my above generalizations.


Nice write-up.

Our team has done 4 wheel independently drive, Independently steered, continuous rotation, co-axial swerve for the last 3 years. It has been quite an experience. So many little problems to solve. Do not underestimate the difficulty of getting a swerve to the point that it gives a competitive advantage. The risk of failure is very high. Our team after allot of debate decided that if we were going to take on the challenge and invest all the time effort and MONEY to develop swerve then we did not want a design with limitations. Thus we developed the unicorn version. Our first year we completed the mechanical but the software wasn’t quite right. Our first regional was a little embarrassing. The robot had some moments when it had what I refer to as a grand mal seizure. The team called it the happy dance. We did solve the software issues for our next regional. And it performed well except we ran into the locking pin and Jag issue with the window motors we used for steering. From our first year the most important thing we learned was the difficult of repairing and servicing the system. While the wheel modules were independent, they were built into the frame. For our second year, We redesigned each wheel module to be total independent. If we have a problem We remove 4 bolts, disconnect 3 pairs of wires and remove the unit. Reverse the process with a pre-calibrated spare run a test and get on to the next match. With tying wheels together you are going to have a much more complex servicing task and you will most likely miss matches. It’s a service nightmare in the pits. After our experience I would highly recommend a modular approach with independent modules. Take the time to perfect it.
One advantage with unicorn drive is that you can choose the location of the center point that the robot rotates around. In 2010 we had a mode were the robot rotated around the center point of the soccer ball making aiming easier. 2011 we used the same concept to rotate around the pole base to align the mini bot. Think carefully before you go down a swerve path with linked wheels. Team 1717 has posted their solution. This is a link to our solution.