Mecanum Pushing Power vs Tank vs Holonomic

Hey All. I’m trying to compile a quick table for comparing pushing/pulling power between different kinds of drivetrain configurations. I have been looking all over, and can only find the following:

ASSUMPTION / DEFINITION: Each motor/gearbox module provides N amount of force when driven. N is assumed to contain all of the inefficiencies of each motor/gearbox combo.

Tank Drive - A Tank Drive driving straight forward provides 4N force. It cannot strafe. If N = 1 unit, Tank Drive provides 4 units of power.

Holonomic (4 Omni wheels 45-degree offset) - Omni wheels driving forward in a “diamond” configuration each provide (sqrt(2)/2) N power because of the 45-degree offset. All 4 wheels together provides ((4 (sqrt(2)/2)) N) power. If N = 1 unit, the best case Holonomic provides 2.83 units driving forwards, backwards, and laterally. However, when driving diagonally (worse case) with only two wheels driving, Holonimic provides only 2N power; if N = 1 unit, worse case Holonomic provides 2 units power.

Mecanum - Straight forward/backward power = ???, Strafing power = ???, Diagonal Power = ???

Can someone point me to resources to help me complete this for Mecanum? Or, help me correct what I’ve already written?

I’ve found tons of “Mecanum can’t push at all” and “Mecanum provides 70% power” without any experimental data or reasoning to back any of it up. I’m hoping to get some kind of data without me having to go hook up a spring scale on a robot with mecanum wheels myself. I’m actually hoping someone else has done this, and is not just going to provide me with conflicting, “Mecanum wheels provide more/less force than xxxx” statements that I’ve already found online.

Thanks a TON!
-George

I think it would be easiest to just rig up a chassis (because you don’t actually care about the motors for traction, the “chassis” only really needs to be the a frame, the wheels[locked so they don’t rotate], and dummy mass) and measure how much force it takes to drag the chassis from various angles.

If you want the theory:
Assuming you have strong enough motors to provide resistance, you’re looking for the traction of a drive train. I’ll also assume that all the configurations weigh the same (including bumpers) because that matters too.
Tank it’s easy because the material resists force in all directions (for all the drive types you have another layer which is the wheel material, I’m going to recognize but ignore it). So the traction of constant.
Omni wheels slip (allow movement with negligible force) parallel to the axis of rotation of the wheel (that is to say: sideways), so each wheel will only contribute some percentage of it’s maximum according to the dot product of the applied force vector and a vector perpendicular to the wheel’s axis of rotation. Do the math for all 4 (or 3, or 10, or whatever) wheels and sum the results and you have your traction. So finding the traction involves some trigonometry but isn’t too bad because of nice right angles.
Mechanum wheels are conceptually similar to omni wheels, they allow for slip along some vector, but their traction vector isn’t perpendicular to the axis of rotation, which is to say they’re always slipping. Because things aren’t at right angles this makes the math annoying, and I would need a whiteboard in front of me to even get started. It seems like a solvable problem though, and if you want to go through the trig then knock yourself out.

If you want to go another layer deep in the traction rabbit hole then also consider that the maximum force a wheel can display is not only due to the traction, but also how much of the robot it’s lifting. So incorporating the geometric factors of where each wheel is and there the Center of Gravity is would be important for accurately modeling something like a West Coast Drive with the outer 4 wheels replaced by Omnis for added maneuverability.

Try these:

Some of the links seem to be broken after the CD migration but it looks like the PDFs are still there.

2 Likes

Wow, what a treasure trove. Thanks very much!

-George

It’s really more about resistance to being pushed than it is about pushing power, in terms of drivetrain tradeoffs here.

1 Like

That’s outside the scope of this question. I don’t necessarily want to hear stats about the vehicle survivability in a front-end crash when asking someone about a car’s power train. I don’t mean to be too snippy here towards you personally, but it’s frustrating that there are hundreds of posts about mecanum wheels on CD and when anyone asks a serious question about speed and pushing power this is what always derails the conversation. No, it’s NOT more about resistance to being pushed, especially when I’m asking about pushing power. When asked about offensive/defensive tradeoffs, THAT is the best time to bring this up. Sorry again, not meaning to call you out specifically, but I spent days searching for this answer and it was frustrating to see this over and over and over again in threads that ALMOST answered my question.

-George

1 Like

Who says we won’t have mobile goals that need to be dragged around again, like 2001, 2002, and 2004? Pushing power can matter.

1 Like

Some thoughts on this…

The definition here for Tank Drive is very simplistic and doesn’t really match what we see in competition. Based on the way everything is worded, it implies 4 gearboxes (2 per side), with motors attached. In reality, teams typically have 1 gear box per side, doubling the number of motors in that gearbox. By running chain/belt from that gearbox to other wheels, they introduce additional inefficiencies in the system that drive trains that only have direct-driven wheels do not have. Those inefficiencies may be minor and not have a big impact on what you’re trying to do, but they should at least be recognized and explained away if so.

For a holonomic drive… many teams that consider such an approach wouldn’t put the wheels at 45 degrees to their “forward” motion. It introduces complexities to their robot geometries that may not be desired. Instead, they would opt for a “slide drive” - wheels on the sides, with strafing wheels on the front/back. With such a setup, it’s easy to adjust your power by adding more motors to the forward/back set of wheels, potentially matching a tank drive for pushing power if you wish.

Finally, it’s important to remember that, while rules of thumb work out pretty well, trying to put specific numbers on a robot is difficult. This can work out well when designing your robot, giving you guidance for your design choices. But it can break down if you try to characterize other robots using this based on pit scouting. Watching performance in the matches is going to give you a better idea.

In the end, I like to think that if there is a pushing match, the defensive robot has already won - the offensive robot is spending time and energy pushing, instead of scoring. It doesn’t take a huge amount of pushing power to get in the way and slow down the offense enough to impact the game (in most games).

That’s a good point. Is there any way to characterize inefficiencies in chain/belt power transfer systems? Let’s say I have two systems:

  1. On both sides of a given robot, the rear wheels of a tank drive are being direct-driven by a motor/gearbox, and chain is being used to transfer power to the front wheels.
  2. A motor/gearbox is providing power to a central shaft, and that shaft is then driving individual chain/belt systems to the front and rear wheels.

I assume that the second system is less efficient than the first - but is it characterizably so? In other words, is it less efficient enough that it should be considered during the design of the drivetrain, or should it just be understood that “there will be some loss, but we probably won’t notice it”?

And, on the same topic, would the same answer apply if I decided to use gears instead of chain/belt?

That’s very insightful, thank you. I had never connected the shape and design of an H-Drive (slide drive) design with that of a 4-wheel-45-degree holonomic drive, even though it’s clear to me now that the designs are perfectly interlinked.

-George

Generally speaking, I don’t worry about the power losses - everything teams typically do is fairly efficient (Google can give you some states for belt/chain efficiency if you really want to dig into it), and ultimately there are other concerns that weigh in more:

  • 1 gearbox + chain and sprockets will generally be less weight than 2 separate gearboxes
  • By connecting all the wheels together on each side of the drive train, you ensure full power is available to each wheel. This is important so you don’t have any “wasted power” if your robot tips up off its front/back wheels
  • An additional gearbox is typically going to be more expensive than connecting the wheels together.

Using a gear drive, some of that doesn’t apply, depending on the exact setup - number of gears, aluminum vs steel, etc. You also have to take into account the need for precision manufacturing and your team’s capabilities - with COTS parts like Vex’s Bearing Blocks, you can get away with less precision for belts and chain, using the blocks themselves for final adjustment… that would be difficult to do with gears.

Another thing to consider is encoder placement. Placing an encoder on your gearbox and then chaining to the wheel can introduce some slop that you don’t want in your measurement - putting the sensors directly on the wheel shaft gives you more accurate results.

Also, when talking about tank drive, it’s important to consider not only pushing power, but all of the power requirements. Having just 4 wheels (one in each corner) introduces a lot of scrub, making turning difficult. This will drain the battery quickly and brown out the robot like you wouldn’t believe. That’s why most teams will run 6 or 8 wheels - they drop the center wheels, ensuring one side of the robot is always off the ground. This effectively reduces the wheel base, reducing scrub and the amount of power needed to turn.

1 Like