Mecanum wheels under varying loads [Archive]

howe

25-01-2015, 17:33

Has anyone experienced problems with mecanum wheels when the weight distribution changes? For example: picking up bins and or containers which would heavily load one pair of wheels and lessen the load on the others if the payload is cantilevered. Thank you.

Robot Sans

25-01-2015, 17:47

I have seen mecanum not perform to well in situations where the center of mass is closer to one wheel than the others but it is something that can be fixed with clever coding, use of encoders and a accelerometer. A more mechanical solution would be to add a pair of wheels closer to the load, which could be free spinning and even the weight difference.

Kevin Ray

25-01-2015, 17:51

If you do a search, you'll find that this topic has been covered ad nauseum. But to give you the Reader's Digest version, yes, they will react to disproportionate loading. It is probably the biggest drawback to this drive system. It can be compensated for by the driver and somewhat in programming. You do have to ensure, however, that all wheels are in constant contact with the ground (or platform).

Caleb Sykes

25-01-2015, 18:12

A more mechanical solution would be to add a pair of wheels closer to the load, which could be free spinning and even the weight difference.

Can you describe in more detail this method? I am having trouble visualizing what you mean.

Dr.Gusta

25-01-2015, 18:15

Can you describe in more detail this method? I am having trouble visualizing what you mean.

A set of idler wheels...So you have your regular 4 wheel mecanum drive but you add non powered omniwheels of the same size in between the two. That is what I got from it at least. I dont get the mechanics of how it would help though.

Ether

25-01-2015, 18:16

A more mechanical solution would be to add a pair of wheels closer to the load, which could be free spinning and even the weight difference.

Are you suggesting weight-bearing non-powered wheels? What kind of wheels?

howe

26-01-2015, 01:25

If you do a search, you'll find that this topic has been covered ad nauseum.

I searched ad nauseum before posting my question with no result.

Please point me to the threads or at least give me some key words for my search.

Thanks

dellagd

26-01-2015, 01:28

I searched ad nauseum before posting my question with no result.

Please point me to the threads or at least give me some key words for my search.

Thanks

'Ad nauseum' is just a latin term for 'continued until its sickening' (until nausea)

GeeTwo

26-01-2015, 02:08

We haven't tried to climb onto platforms, but 3946 has experience with mecanums that were WAY out of balance. Our competition robot last year wasn't too bad, but our proto'bot, made mostly of 2"x4"s, only put about 10# of force on the left front wheel (total weight in excess of 100#). Our drivers learned to compensate well enough that a casual observer at practice would not have suspected how out-of-balance we were. I demo'd "Woodie" at my office with absolutely zero drive practice, and impressed my co-workers (mostly scientists and engineers) with the capabilities of mecanum - I could execute all sorts of translation as well as rotation.

howe

26-01-2015, 09:06

'Ad nauseum' is just a latin term for 'continued until its sickening' (until nausea)

Duh, I just don't know what possessed me to type "ad nauseum" into the search bar BEFORE I even posted the question, foolish me.

Now seriously, how about showing some gracious professionalism and post the urls of the threads where the topic of shifting load has been discussed previously? I have gone through several pages searching simply under mecanum in addition to other key words I thought appropriate. It would be nice to know how others have modified their code or driving style to compensate as mentioned in some of the responses.

Thanks,
howe

Tom Line

26-01-2015, 09:58

http://www.chiefdelphi.com/forums/showthread.php?threadid=133086

http://www.chiefdelphi.com/forums/showthread.php?threadid=132855

http://www.chiefdelphi.com/forums/showthread.php?threadid=121317

Google search terms:
mecanum wheel load uneven site:chiefdelphi.com

I usually have much better luck with Google's search than Chief Delphi's.

dellagd

26-01-2015, 12:11

Ive also been curious about 'perfecting' a mecanum drive system by taking the load on each wheel and programming the drive to account for the uneven loading. All papers I found only referenced the geometric center of the base, not the center of gravity. I have a pretty good understanding of how they work, and I feel like there should be a way to incorporate the different normal forces into the reverse kinematics, but unfortunately I've come up short on figuring out the calculation myself.

So I'll ask too: Has anyone ever done a write up on the reverse kinematics with uneven loading or even written some code to deal with it? This seems like something Ether would have addressed in the past, but I can't find anything for it.

Robot Sans

26-01-2015, 12:17

Are you suggesting weight-bearing non-powered wheels? What kind of wheels?

Yup, thats what i was thinking. They would have to be low friction but not omni-directional. Plastic wheels would be ideal.

Ether

26-01-2015, 12:21

Has anyone ever done a write up on the reverse kinematics with uneven loading or even written some code to deal with it? This seems like something Ether would have addressed in the past, but I can't find anything for it.

assuming each wheel is being driven at the kinematically correct speed, any deviation from the kinematically predicted behavior, due to uneven loading, would mean that the bot is no longer behaving "holonomically" -- there is wheel scrub -- and this would require an analysis of the forces involved.

dellagd

26-01-2015, 12:30

Interesting. I'm guessing doing the complete analysis would be pretty complicated if there aren't any papers about it. I wonder if a good estimation could be made that works well enough in practice?

Alan Anderson

26-01-2015, 12:39

Ive also been curious about 'perfecting' a mecanum drive system by taking the load on each wheel and programming the drive to account for the uneven loading.

Varying the normal force "load" on each wheel has two primary effects. One is a subtle change in the wheel speed vs motor voltage. That can be compensated for using closed-loop control of wheel speed. The other is a significant change in the maximum thrust that the wheel can apply before slipping. That one is not so easy to account for, though I suspect monitoring the motor current can provide a way to do traction control and turn the wheel slip problem into a maximum robot acceleration problem.

Ether

26-01-2015, 12:58

I'm guessing doing the complete analysis would be pretty complicated

Yeah I think so. Alan's post#16 does a pretty good job explaining why load affects mec behavior and possible ways to deal with it.

Other options to deal with it involve using actual vehicle motion sensors (e.g. gyro, encoders on follower wheels, etc) to make corrections to the driver's commands.

Ether

26-01-2015, 13:08

Yup, thats what i was thinking. They would have to be low friction but not omni-directional. Plastic wheels would be ideal.

Why not omni? and what effect do you think those wheels might have on the mec's kinematics?

Daniel_LaFleur

26-01-2015, 13:15

Why not omni? and what effect do you think those wheels might have on the mec's kinematics?

Why not a ball caster?

howe

26-01-2015, 13:42

Chinske4296

27-01-2015, 11:46

We currently have a mecanum setup with omni wheels in between.

Caleb Sykes

27-01-2015, 12:09

We currently have a mecanum setup with omni wheels in between.

Are the omni-wheels powered?

Ether

27-01-2015, 12:31

We currently have a mecanum setup with omni wheels in between.

in between how? A, B, C, D, or something else?

gpetilli

27-01-2015, 13:55

in between how? A, B, C, D, or something else?

Also, are all the wheels at the same frame height or is the "in between" wheels dropped (by 1/8")?

pafwl

27-01-2015, 22:35

With the unpowered omni in the middle I am guessing that you may have issues getting over the humps.

Are you using Andy Mark or VEX?