We built a 6-CIM Kiwi Drive. Criticisms please!

[asid61]

The thing is, having already made a "kiwi" drive makes the next year's drivetrain much easier. Maybe not day 1, but defeinitely much earlier design-wise. Whenever doing a new drivetrain, be it WCD or kiwi or swerve, it should always be prototyped in the offseason. Doing radically new things during the season tends to have problems in my experience.

Personally, I always say "play for the endgame". This is true for many competitions that I've seen. By "play for the endgame", I mean that selecting the most effective drivetrain and practicing with that will turn out better (eventually) than using easier drivetrains (like kitbot) and being unable to modify effectively.

To OP:
This looks super clean! The creativity aspect of using 3 wheels in an equilateral shape makes this worth it IMO. New ideas get us places. Even if we don't like those places very much lol.

Now, about the kiwi drive:
-Does wheel slippage make for lower efficiency? Have you tested the electrical current levels against a 4-wheel or 6-wheel drive for the same speed?
-Is programming very tricky? Is it easy to implement with a different wheel spacing?
-How much does the basic chassis weigh, not including electronics (with or without motors)?
-Did you have tipping problems due to the 3-wheel design?
-Did the pointyness help you very much?
-What kind of traction did you see with this?

[Ether]

Originally Posted by asid61

The creativity aspect of using 3 wheels in an equilateral shape makes this worth it IMO.

The fielded design did not have equilateral placement of the wheels. That's why I asked about inverse kinematics.

[Nate Laverdure]

Man, the questions keep getting harder to answer. I love it.

Originally Posted by Andrew Schreiber

I'm going to cheat and ask the same questions from both a competitive robot perspective and a team growth perspective.

That's fine. In my mind I mix the two anyway. Plus this gives me an opportunity to display my answers in a table:

[Nate Laverdure]

Originally Posted by Ether

Did you try all the suggested ways to mitigate [gyro drift]?

I believe so. The "calibrate only when compressor is off" suggestion is an excellent one that I hadn't heard before. I understand that the quality of the KOP gyro varies considerably from batch to batch, and I think we got a good one.

Originally Posted by Ether

Would he be interested in posting his equations for discussion on this forum?

I will ask. He's going to be my intern this summer, so I can assign him this task as a special project

Originally Posted by Ether

I don't think the WPILib has any functions that would be usable for your 3-wheel asymmetric kiwi arrangement.

Excellent catch! We accounted for the asymmetry in the control design.

[cxcad]

Why not build a tex-coast drive? From what I read and understand is that you do not need fancy machinery to build it. I believe 624 did it without CNC.

[Caleb Sykes]

Originally Posted by Nate Laverdure

I believe so. The "calibrate only when compressor is off" suggestion is an excellent one that I hadn't heard before. I understand that the quality of the KOP gyro varies considerably from batch to batch, and I think we got a good one.

I will ask. He's going to be my intern this summer, so I can assign him this task as a special project

Excellent catch! We accounted for the asymmetry in the control design.

Would you mind giving the relative positions/orientations of the wheels, I am trying to determine your theoretical max velocity in any given direction. My guesses are (back center wheel at (0,0) and oriented east/west):
front left: position = (-16,-25)in oriented 22.5 degrees west of north
front right: position = (16,-25)in oriented 22.5 degrees east of north

EDIT: changed front and back to reflect Nate's response.

[Kevin Leonard]

Competitive success and learning about engineering do not have to be separate ideas.

There's a student on our team right now who is disgruntled. He thinks we focus too much on competitive success, and he believes that building the same traction drivetrain each year doesn't help students as much as building something crazy, like a kiwi drive, that may not contribute as much to competitive success.

I'm going to try to address both your situation and my teammate's situation simultaneously, so bear with me.

In my humble opinion, students working with a kit-bot chassis can learn just as much, if not more as students working on a kiwi drive- however these might be different things.
There is so much to learn about a six-wheel drivebase that you can't learn from a kiwi drive. Factors involved in pushing matches, like current draw and breaking traction, as well as the advantages and disadvantages of different gearing ratios. I've learned near everything I know about these topics from my involvement in FRC on a team that uses nothing but traction drivebases.

And as you said, your shooter didn't always work when you wanted it to. By investing time into your kiwi drivebase, you lost time to work on your scoring mechanisms, which ended up unreliable. Tons of engineering hours can be put into making these other systems great. Reliability is much more difficult to engineer into a system than teams like 1114 and 254 make it seem, and having some time to work out the kinks in your subsystems- whether through prototyping or through practice time- is what makes these teams great.

And despite what some people might say- it is possible to be competitive with a non-traction drivebase in the right game if you so choose. The two teams that come to mind are Team 2052, KnightKrawler from Minnesota and Team 1425, Error Code Xero from the Pacific Northwest.
2052 uses a mecanum drive most years, and 1425 uses some sort of kiwi drive (correct me if I'm wrong). These teams are successful not because of their drivebases, but because their scoring mechanisms are so well-tuned and their drivers so practiced. 2052's drivers in 2013 were among some of the best in FRC.

What would I suggest for next year? There is no situation that has yet occurred in FRC where a 6-wheel traction drive is a bad idea, and the kitbot is a good and affordable drivetrain that can provide many lessons in engineering.

However, if you intend to stick with a Kiwi drive, figure that out Week 1 and have it built by the end of week 1. Then work on your scoring mechanisms and have your drivers start practicing with driving Week 1.

And good luck.

[Nate Laverdure]

Originally Posted by asid61

Have you tested the electrical current levels against a 4-wheel or 6-wheel drive for the same speed?

What's interesting about omni and mecanum drives is that the rollers have small but significant friction (especially when you apply an axial thrust load!). When you arrange 4 omni wheels in a rectangle, you get a drivetrain that prefers to travel in the cardinal directions and dislikes traveling in arbitrary, off-cardinal directions. With a triangular design you don't have this issue, but with a hexagonal design you would. When it comes to the number of drive wheels on a kiwi-bot, any odd number greater than 2 would work!

Originally Posted by asid61

How much does the basic chassis weigh, not including electronics (with or without motors)?

Originally Posted by asid61

Did you have tipping problems due to the 3-wheel design?

Nope. The wide-set front wheels ensured that a very small amount of weight was cantilevered outside of the supported triangle. Also the robot's CG was extremely low even when carrying a ball. Bumpers were slammed to the bottom of the bumper zone. I just wish I could have gotten the battery lower and less exposed.

Originally Posted by asid61

Did the pointyness help you very much?

Actually, the flatness did. The chassis shape gave us an extremely wide, flat front that we used for ball corralling and a blocking-style defense.

Originally Posted by asid61

What kind of traction did you see with this?

Wheel static COF in the direction of travel is supposedly 1.0 (compare to the KOP HiGrip). We did not test this. We also chose this design path with the understanding that we would avoid pushing matches and attempt to avoid defenders instead. Even with an open field, I find that you can do a lot of high-quality defense without needing to touch your opponent. Unless they're really aggressive, most drivers will choose to avoid defenders instead of barreling straight through them.

[Nate Laverdure]

Originally Posted by inkling16

Would you mind giving the relative positions/orientations of the wheels, I am trying to determine your theoretical max velocity in any given direction. My guesses are...

BOLD ITEMS ARE EDITED TO ADDRESS CONFUSION:
back middle: position = (0,0) oriented west
front right: position = (-15.9,-18.5)in oriented 30 degrees west of north
front left: position = (15.9,-18.5)in oriented 30 degrees east of north

There's 120 degrees between the directions of each neighboring wheel.

[Ether]

Originally Posted by Nate Laverdure

The wide-set front wheels ensured that a very small amount of weight was cantilevered outside of the supported triangle.

Originally Posted by Nate Laverdure

back left: position = (-15.9,-18.5)in oriented 30 degrees west of north
back right: position = (15.9,-18.5)in oriented 30 degrees east of north

To facilitate dialog and minimize confusion, are the wide-set wheels front or back?

[Nate Laverdure]

Originally Posted by Ether

To facilitate dialog and minimize confusion, are the wide-set wheels front or back?

C'mon Ether. You know an omnidirectional robot has no front or back.

[Caleb Sykes]

Originally Posted by Nate Laverdure

Close.

Well, in one direction at least.

Okay, I was thinking that the back wheels were at 22.5 degrees because the 80-20 mounting plates on the back end and on the 45 degree piece looked similar. That makes for a much simpler problem.

This makes me curious though, has anyone ever done a kiwi drive with the wheels not all oriented at 120 degree angles relative to each other?

[Nate Laverdure]

Originally Posted by inkling16

Okay, I was thinking that the back wheels were at 22.5 degrees because the 80-20 mounting plates on the back end and on the 45 degree piece looked similar. That makes for a much simpler problem.

One's a 60-deg plate and the other is a 75-deg plate. In the "rolling chassis" photo the bright white gusset plates are 3D-printed for rapid assembly, and they were replaced later with waterjetted aluminum.

[Ether]

[edit]OBE. I see you edited your prior post[/edit]

[Kingland093]

The drivetrain looks good!

I would recommend having lots of driver training. My team had a similar issue as you guys in which we lost a lot of time early on in the build season and had very little time for driver training (about 2 hours total) and that killed us at our regional. To me, it's not as much what you're driving but how you drive it.

Hope this helps