Mecanum vs Poor Man's Mecanum

So people on here seem to hate mecanum. This isn’t a hate on mecanum thread.

For those that don’t know, poor man’s mecanum is 4 omni wheels mounted at 45° angles to emulate mecanum rollers.

So this past season I have worked with both mecanum and PMM. To me PMM looks like it works smoother and your bot doesn’t have a preferred “front”. It moves just as fast in any direction. Mecanum on the other hand should only be driven sideways as long as it’s to adjust (keyword right there) your sideways positioning. Not to travel across the field. It is slower and less smooth.

Other than how it travels, the benefits of PMM are: cheaper (mecanum wheels cost a lot), rotation (the wheel set up makes spinning easier).

Mecanum: can use on kit chassis (have to fabricate your own for PMM)

Is there anything else that mecanum is better at?

First of all, what you’re calling “poor man’s mecanum” is generally known as kiwi drive. The major advantage of mecanum over kiwi is that a mecanum robot can use a standard frame with two front-back drive rails whereas kiwi needs a more complicated chassis that’s harder to mount mechanisms to. Another big advantage is that if you find you’re not using the strafing effectively you can switch the mecanum wheels for regular traction wheels without having to change the chassis.

In theory, kiwi shouldn’t turn any better than mecanum (mecanum already turns fine anyway). It is cheaper though, since mecanums are cheaper than omnis.

1 Like

Kiwi is 3 omnis in a triangle. 4 in a rectangle is X drive.


I’d just like to add that PMM (I usually just call it holonomic, but ik that’s a more general term) doesn’t have to have all wheels mounts at 45 degrees (which seems to be most popular on VEX bots with this drive), but you can also just have one wheel parallel to each side of your frame. As long as you can make the motors fit, it’s a very convenient way to package an omnidirectional drive system.

Disclosure, I’ve only done this on small robots, so I have no experience in how well this would scale up to FRC. My first and biggest concern is that you would have only 2 motors of power in each axis, which would be a big yikes if (when) you get into a pushing match.

Example – please excuse the wiring

Kiwi and Meccanum drives both have the same problems with traction:

a) Traction makes it hard to go where you want: Each wheel has to have good traction with the ground in order for their push vectors to sum correctly. Lose traction on one wheel (because of weight shift, or a floor obstacle) and the vehicle no longer behaves predictably.

b) Traction makes it hard to stay where you are: Since each wheel can only provide friction in one direction, your total friction with the carpet is < 1/2 of what it could be with regular wheels.

There are ways to reduce the impact of a).

  • The DJI robomasters kits add suspension to better balance the loads on the wheels.
  • Kiwi bots often use 3 wheels instead of 4 so that no wheel ends up unloaded.
  • In 2017 team 1986 famously used 8 mecannum wheels, 2 in each corner, so that each pair could “push off” eachother, while lifting roughly the same amount of normal force. (edit - at least I think that was the idea)

AFAIK, there is no way to mitigate the impact of b).


I’ve known it to be called holonomic, but same thing.

Any drivetrain that separates rotation from translation is holonomic.


Oh really?
Thank you.

This is why I generally hate FRC terminology. None of these are names known in the industrial or robotics worlds, and are essentially all just what individual teams referred to their drivetrains as. The “kiwi drive” name is just because that’s what 857 called it in 2002.

Each of these are variations of holonomic drivetrains (although there is a conversation to be had regarding if the 4-motor variations are actually holonomic).

Thank goodness that Mecanum wheels had a technical name established by patent, or else we might all know the drivetrain as the Jester Drive (from 357).

Not quite. To truly fit the definition of holonomic, the drive shouldn’t have more input degrees of freedom than the output degrees of freedom. To put it another way, a swerve drive is decidedly not holonomic, as it has 8 controllable degrees of freedom (assuming 4 modules - 1 steering and 1 wheel each) and still only produces 3 output degrees of freedom (Vx, Vy, θ).


While we did call it kiwi, I believe there is a foundation/company that does something with kiwi drive and has a kiwi bird with 3 legs as a logo.

Also, we (857) still have a kiwi prototype in storage.

I always called it Killough drive. We should rename it to “no-name” drive.

Per wiki, Killough is the same concept but without using omni wheels.

Killough is the proper name, but I was under the impression that killough was without omniwheels

Mecanum wheels are also known as Ilon wheels after their inventor, not the company he worked for.

Maybe instead of calling them all these incorrect things we should just stop talking about them altogether.



Another difference which has not been noted here between Mecanum and what OP refers to as PMM: With the same motors, gearing, and wheel size, Mecanum accelerates 41% better than PMM, while PMM has a 41% higher free speed than Mecanum.

The big advantage of Mecanum over PMM, imho is that (now that both omni and mecanum wheels are available as COTS items), construction of the mecanum chassis is much easier; there is no need for any 45° chassis members. The KoP drivetrain actually has mount points for four TB-micro to support mecanum. You could also build a chassis using versaframe or WCP COTS components without any 45° gussets to be mecanum.


A wrinkle on PMM I experimented with a couple of years ago is one I called Pakuni Drive†, named after the simian aliens from Land of the Lost and their attribute of walking forwards like humans, but running sideways like my little brother Vernon when he was 18 months old. The idea is that I toed in the wheels ~22.5° rather than 45°. (It turns out that the acute angle on a 5-12-13 right triangle is closer to 22.62°, so that’s what I really did.) Anyway, this gives a forward acceleration which is 2.4x better in the forward direction than the lateral (strafe) direction, but a free speed which is 2.4x better in the lateral. My intention was to start the robot accelerating in the “forward” direction, then rotate the robot 90 degrees to “shift” to a speed 2.4x as fast for longer (cross-field) runs. My attempts with an arduino and no gyro or other feedback nor radio input were unsuccessful, so I’m putting the concept out there in case anyone wants to try this with better control systems.

† My earlier name for this before inspiration hit was roto-shifter, as will be explained in later sentences.


If you really want to be accurate with your acronyms, both mecanum and PMM are subclasses of NPP* drive

[*] not participating [in] playoffs


Is that just because they move sideways?

Or because they do it badly?