View Full Version : Mecanum Differential Drive
efoote868
22-11-2013, 22:33
I've been toying around with an idea for a while, and I'd like to share it with the FRC community, and maybe get others to improve on it or (if they feel like a challenge :) ) implement it.
My idea combines some of the present implementations of 6 wheel tank drive along with the transmission in the Toyota Prius, to improve upon some of the perceived weaknesses in mecanum drives - namely motor distribution and shifting.
I'll try to describe it as best I can here, although if it doesn't make sense a crude paint drawing may be in order.
First, start with the standard shifting tank drive. Each side has an independent shifting transmission which directly powers two or more wheels. When driven together, the robot can move forward and backward. When driven against one another, the robot can spin clockwise and counter clockwise. This accounts for 2/3rds of the motion in mecanum drives.
Second, take the planetary gearbox used in the Prius. It takes an input from a gas motor as well as an input from an electrical motor, and combines the two into power for the wheels. Now, instead of using the shifting transmission to directly drive the wheels in the mecanum drive, use the output of the shifting transmission as the first input of a planetary gearbox. Use another motor as the second input to the planetary gearbox.
Since each wheel has an independent motor, the left/right motion of the robot can be controlled by spinning the auxillary motors against each other.
Foreseeable issues include the problem of back-driving the motor at each wheel. This will have to be designed out mechanically with a brake or worm gear. There will also be an issue mechanically with weight, complexity, as well as the number of motors required for this design. Finally, controlling this will take lots of thought and optimization.
Another observation that I've had while thinking about this is that each wheel doesn't require a planetary gearbox, just one on each side. While this will allow the robot to move sideways with less complexity, it doesn't allow all motors to push together for a higher top speed.
Thoughts?
mman1506
22-11-2013, 22:57
Dude you forgot encoders, it would win champs if it had encoders.
Motor distribution and shifting aren't really the primary drawbacks of a mecanum drive; it's the decrease in traction that gets you. There's little purpose in shifting to a lower-gear with a mecanum drive train when your pushing ability is traction-limited.
As for your idea, I'm having a hard time visualizing what exactly it is you're suggesting, or how it'd result in a noted improvement over the standard one motor and one gearbox per wheel.
efoote868
23-11-2013, 00:53
Motor distribution and shifting aren't really the primary drawbacks of a mecanum drive; it's the decrease in traction that gets you. There's little purpose in shifting to a lower-gear with a mecanum drive train when your pushing ability is traction-limited.
As for your idea, I'm having a hard time visualizing what exactly it is you're suggesting, or how it'd result in a noted improvement over the standard one motor and one gearbox per wheel.
One could incorporate locking mecanum wheels to gain more traction, but the limitation this addresses is top speed / acceleration time to top speed. I haven't heard of a mecanum setup pushing 16 feet per second or higher in FIRST, which seems to be a fairly regular occurrence with tank drive.
http://i34.photobucket.com/albums/d120/eyeseesall/mecanumhalf_zps0e4ffbd5.png
Output of shifting gearbox A is used as one of the inputs to planetary gearboxes B and C. The other input to B and C are independent motors. The output of the planetary gearbox goes to each wheel.
In this setup, the annulus, the planets, and the sun gear are all utilized, instead of the traditional fixed annulus.
One could incorporate locking mecanum wheels to gain more traction, but the limitation this addresses is top speed / acceleration time to top speed. I haven't heard of a mecanum setup pushing 16 feet per second or higher in FIRST, which seems to be a fairly regular occurrence with tank drive.
Acceleration in FRC (with the exception of 2009's wonky drive surface/wheels) has never been traction-limited, so this is purely a function of the gear ratio people are using on their drives. There's nothing about a shifter which would provide any significant benefit to a mecanum bot; a well-designed mecanum drive is geared for high-speed at all times, because the ability to push is something you've already compromised on.
Your setup adds two gearboxes and two motors to the assembly, but I don't see how you get a significant benefit in return. When moving forward, you'd still have two CIMS pushing each side (as the auxiliary motors won't be doing anything, if I understand correctly), so you don't have any more effective, power-wise, than the standard setup.
efoote868
23-11-2013, 01:33
Acceleration in FRC (with the exception of 2009's wonky drive surface/wheels) has never been traction-limited, so this is purely a function of the gear ratio people are using on their drives. There's nothing about a shifter which would provide any significant benefit to a mecanum bot; a well-designed mecanum drive is geared for high-speed at all times, because the ability to push is something you've already compromised on.
A well designed two speed gearbox has a significant advantage in acceleration. DC Motor power is on a curve, and the well designed gearbox will keep the motor close to peak power. Without a shifting gearbox, 16 fps in FRC provides nearly no advantage to 12 fps or 8 fps.
Your setup adds two gearboxes and two motors to the assembly, but I don't see how you get a significant benefit in return. When moving forward, you'd still have two CIMS pushing each side (as the auxiliary motors won't be doing anything, if I understand correctly), so you don't have any more effective, power-wise, than the standard setup.
No, the auxiliary motors would be used in this setup as well. In the picture presented, 8 motors would be driving this robot when it is going full forward. Alternatively, 1 motor could be in gearbox A, which would allow for 6 motors driving full forward in mecanum drive.
The alternative configuration is that planetary gearbox C does not have an auxiliary motor, which would mean that only 4 motors are driving forward.
No, the auxiliary motors would be used in this setup as well. In the picture presented, 8 motors would be driving this robot when it is going full forward. Alternatively, 1 motor could be in gearbox A, which would allow for 6 motors driving full forward in mecanum drive.
In that case, I think you'd almost certainly do better to just plop the extra motors on the gearboxes in a standard configuration.
The gear ratio spread on most shifters (usually ~2.5:1 for a small spread) is such that a mecanum in the low gear will have little use if the high gear is in a reasonable range.
Shifters are useful for tank drives because they allow the same drive train to have high mobility and high pushing power. A mecanum drive will never have the latter to begin with. I've seen many drives set for ~16 f/s without use of a shifter; it's not a requirement to run at that speed effectively. For what it's worth, this thread is the first time I've ever seen any criticisms leveled at mecanum purely regarding mobility - if you build it right, you have a very agile robot.
If you're dissatisfied with mecanum as an omnidirectional solution and are willing to cope with a moderate increase in complexity, octanum and/or nonadrive are the direction to look in; these give you the full benefit of gear shifting, because you do not have low-traction wheels in contact with the ground in your lower gear.
In this setup, the annulus, the planets, and the sun gear are all utilized, instead of the traditional fixed annulus.
How do you plan on doing this?
This is similar to how any automatic transmission works, but I don't see a way that you could manufacture these parts.
efoote868
23-11-2013, 11:05
The annulus would have to be machined for this special purpose, but I think if it had teeth on the outside (3rd image down (http://woodgears.ca/reader/walters/planetary.html)) it would make powering it much easier than needing 3 concentric shafts.
MichaelBick
23-11-2013, 11:57
Lots of misinformation in this thread.
The only team that I have heard of using shifters for better acceleration is 33. Everybody else uses their second speed as a "pushing speed". 16fps has clear advantages. If you have a higher top speed then you can get to your destination faster. They way we got around the acceleration problem was by adding more motors(something you have already done). Also, our low gear is traction limited(if it wasn't teams like 254 with 6 CIMs would blow their main every time they hit another robot)
Locking mecanums does not improve forward traction. To improve forward traction on a mecanum you would have to change the roller material.
Also, planetaries are extremely heavy(especially ones big/strong enough for drive). For all the weight of the planetary, you might as well put in another shifter(which would also be easier to fabricate because you could put it COTS).
This doesn't even start to mention the fact that you are removing all the benefits of mecanum. Mecanums are easy and relatively light. With this, you are making it heavy and extremely hard to manufacture. You might as well run a swerve.
The only team that I have heard of using shifters for better acceleration is 33. Everybody else uses their second speed as a "pushing speed". 16fps has clear advantages. If you have a higher top speed then you can get to your destination faster. They way we got around the acceleration problem was by adding more motors(something you have already done). Also, our low gear is traction limited(if it wasn't teams like 254 with 6 CIMs would blow their main every time they hit another robot)
This is true.
Locking mecanums does not improve forward traction. To improve forward traction on a mecanum you would have to change the roller material.
This is false; if you draw a free-body diagram for the rollers in contact with a ground on a Mecanum drive train, you'll see that the effective coefficient of friction compared to a traction wheel of the same material is reduced by sqrt(2)/2, because the friction force on each roller acts parallel to its axis. That is to say, if you apply the same force in the forward/backward direction to a mecanum drive and a traction drive, the induced friction force on the rollers needed to keep the robot stationary is greater on the mecanum drive by a factor of sqrt(2), since the friction force from each roller is pointing at a 45-degree angle (it must, to keep the rollers on opposite sides from spinning).
Also, planetaries are extremely heavy(especially ones big/strong enough for drive). For all the weight of the planetary, you might as well put in another shifter(which would also be easier to fabricate because you could put it COTS).
This doesn't even start to mention the fact that you are removing all the benefits of mecanum. Mecanums are easy and relatively light. With this, you are making it heavy and extremely hard to manufacture. You might as well run a swerve.
This is dead-on. Large increases in weight and complexity must yield large benefits to be justified. For the amount of added weight/complexity in this design, there are a number of other proven designs with much greater added benefits.
AdamHeard
23-11-2013, 12:42
This is false; if you draw a free-body diagram for the rollers in contact with a ground on a Mecanum drive train, you'll see that the effective coefficient of friction compared to a traction wheel of the same material is reduced by sqrt(2)/2, because the friction force on each roller acts parallel to its axis. That is to say, if you apply the same force in the forward/backward direction to a mecanum drive and a traction drive, the induced friction force on the rollers needed to keep the robot stationary is greater on the mecanum drive by a factor of sqrt(2), since the friction force from each roller is pointing at a 45-degree angle (it must, to keep the rollers on opposite sides from spinning).
He was talking about locking the rollers so they can't rotate, in which case traction is equivalent.
He was talking about locking the rollers so they can't rotate, in which case traction is equivalent.
He said that "locking the rollers on a mecanum drive does not increase forward traction." I took that to mean, "does not increase forward traction c.f. a mecanum drive with un-locked rollers," which is not true for the reasons I outlined; locking the rollers gives you an immediate traction boost by a factor of sqrt(2).
efoote868
23-11-2013, 14:00
16fps has clear advantages. If you have a higher top speed then you can get to your destination faster.
In FRC, 16fps for a single speed has no advantages to 12fps for a single speed. By the time you accelerate to the top speed, you're already clear across the field.
They way we got around the acceleration problem was by adding more motors(something you have already done). Also, our low gear is traction limited(if it wasn't teams like 254 with 6 CIMs would blow their main every time they hit another robot)
You can't use 6 cims in present mecanum drives.
Also, planetaries are extremely heavy(especially ones big/strong enough for drive). For all the weight of the planetary, you might as well put in another shifter(which would also be easier to fabricate because you could put it COTS).
Controlling a robot through 4 independent shifts on a mecanum drive would be difficult.
This doesn't even start to mention the fact that you are removing all the benefits of mecanum. Mecanums are easy and relatively light. With this, you are making it heavy and extremely hard to manufacture. You might as well run a swerve.
8 years ago if I told you that in the future, mecanum wheels would be available for purchase to every team at a reasonable cost, you would have called me crazy.
Not all innovation leads to acceptance. Just ask Dean.
There IS one drivetrain with mecanums that COULD have 6 CIMs applied, though not during strafe maneuvers. 1322 built a 6-wheel mecanum back in 2010 (http://www.chiefdelphi.com/media/photos/34594) (2008 version (http://www.chiefdelphi.com/media/photos/32175)). If you powered the central omniwheel driveshaft (and one of the two wheels on it), you could apply a 5th and 6th CIM. I leave how to do that up to the reader.
MichaelBick
23-11-2013, 14:11
In FRC, 16fps for a single speed has no advantages to 12fps for a single speed. By the time you accelerate to the top speed, you're already clear across the field.
You can't use 6 cims in present mecanum drives.
Your calculations are totally dependent on weight. Acceleration on a 90lb robot is fine. Just look at 254 to see very sucessful drivetrains that have been geared fast(16-18 fps).
While you can't run a 6 CIM mecanum, you are ignoring the fact that at this weight and complexity you might as well make an octocanum or swerve.
Controlling a robot through 4 independent shifts on a mecanum drive would be difficult.
Teams do it all the time on butterfly. Mecanum is no different.
8 years ago if I told you that in the future, mecanum wheels would be available for purchase to every team at a reasonable cost, you would have called me crazy.
Honestly I have no idea what will be available in 8 years, so it isn't worth speculating about.
efoote868
23-11-2013, 14:23
While you can't run a 6 CIM mecanum, you are ignoring the fact that at this weight and complexity you might as well make an octocanum or swerve.
I'm not ignoring that, I even stated it as a problem in the OP.
Teams do it all the time on butterfly. Mecanum is no different.
Actually, thinking along these lines, has any team tried a 4-shifter mecanum with locking rollers? Seems to me it might be a nice alternative to octanum; it'd offer pretty much identical functionality with comparable complexity/weight.
MichaelBick
23-11-2013, 16:26
Actually, thinking along these lines, has any team tried a 4-shifter mecanum with locking rollers? Seems to me it might be a nice alternative to octanum; it'd offer pretty much identical functionality with comparable complexity/weight.
Complexity and weight isn't comparable whatsoever. Look at 3928 butterfly and 973's offseason butterfly. Substitute in mecanums and you have a octocanum. Their drives are significantly less resources than a WCD even and almost as light.
Complexity and weight isn't comparable whatsoever. Look at 3928 butterfly and 973's offseason butterfly. Substitute in mecanums and you have a octocanum. Their drives are significantly less resources than a WCD even and almost as light.
Shifters aren't that much heavier than normal toughbox nanos (~3-4lbs extra, total), and a locking mecanum system doesn't add much weight either if you build it intelligently.
In terms of moving parts, you've got four extra cylinders on a locking mecanum system, yes, but you remove the need for an extra belt, axle, and pulleys.
With the probable advent of 4'' mecanums from VexPro (which one could run straight from a CIM with direct-mounted gears and no gearbox), octanum does get quite a bit lighter, admittedly.
BBray_T1296
23-11-2013, 17:07
What about a swerve drive with mecanum wheels? haha nevermind...
What about a swerve drive with mecanum wheels? haha nevermind...
Nah, that's not good enough, you've gotta put octanum modules on the swerve modules.
With locking plates on the mecanums, of course.
BBray_T1296
23-11-2013, 17:36
Nah, that's not good enough, you've gotta put octanum modules on the swerve modules.
http://s18.postimg.org/fd2d4sbt3/drivetrain.jpg
Their drives are significantly less resources than a WCD even and almost as light.
Care to expand?
MichaelBick
23-11-2013, 19:52
They both use the same basic design:
https://scontent-b-lax.xx.fbcdn.net/hphotos-frc1/995428_567383483318423_1129116488_n.jpg
https://scontent-a-lax.xx.fbcdn.net/hphotos-prn2/1070093_567383509985087_1854889069_n.jpg
Essentially they have gotten butterfly down to the bare minimum causing it to be extremely easy to make(easier than a WCD) and just a bit heavier.
They both use the same basic design:
https://scontent-b-lax.xx.fbcdn.net/hphotos-frc1/995428_567383483318423_1129116488_n.jpg
https://scontent-a-lax.xx.fbcdn.net/hphotos-prn2/1070093_567383509985087_1854889069_n.jpg
Essentially they have gotten butterfly down to the bare minimum causing it to be extremely easy to make(easier than a WCD) and just a bit heavier.
With current COTS hardware, it's very hard to just swap out those omni-wheels with mecanum wheels without any changes; the 12t pinion on the 72t gear direct-mounted to the wheel relies on a 4'' wheel to work.
Incidentally, 4464 is making a slightly-modified version (we've made some material/layout changes to accommodate our specific machining capabilities) of that second type of module (actually, we cribbed most of our design choices from RoboWranglers - they had one of their modules on display at Worlds this year) as an offseason project.
AdamHeard
23-11-2013, 20:06
They both use the same basic design:
https://scontent-b-lax.xx.fbcdn.net/hphotos-frc1/995428_567383483318423_1129116488_n.jpg
https://scontent-a-lax.xx.fbcdn.net/hphotos-prn2/1070093_567383509985087_1854889069_n.jpg
Essentially they have gotten butterfly down to the bare minimum causing it to be extremely easy to make(easier than a WCD) and just a bit heavier.
Our variant of this was trivially different (biggest difference was using 2" tall tube versus 3") and I can confirm it was easier to manufacture than our WCDs. There just aren't a lot of parts.
MichaelBick
23-11-2013, 20:28
With current COTS hardware, it's very hard to just swap out those omni-wheels with mecanum wheels without any changes; the 12t pinion on the 72t gear direct-mounted to the wheel relies on a 4'' wheel to work.
The omni wheel is 6" dia and the traction 4" dia. I don't see the problem
The omni wheel is 6" dia and the traction 4" dia. I don't see the problem
I am quite sure that omni wheel is 4'', with a 72-tooth VexPro gear on it; vexpro does not sell gears that are that size proportional to a 6'' omni.
MichaelBick
23-11-2013, 20:46
I am quite sure that omni wheel is 4'', with a 72-tooth VexPro gear on it; vexpro does not sell gears that are that size proportional to a 6'' omni.
VexPro doesn't sell a traction wheel smaller than 4" either
BBray_T1296
23-11-2013, 20:46
I am quite sure that omni wheel is 4'', with a 72-tooth VexPro gear on it; vexpro does not sell gears that are that size proportional to a 6'' omni.
But VexPro is going to unveil (http://www.chiefdelphi.com/forums/showthread.php?threadid=122129) its new products soon, which should include this tiny devil (http://farm6.staticflickr.com/5496/10876821676_8c2b1a1569_c.jpg)
But VexPro is going to unveil (http://www.chiefdelphi.com/forums/showthread.php?threadid=122129) its new products soon, which should include this tiny devil (http://farm6.staticflickr.com/5496/10876821676_8c2b1a1569_c.jpg)
Yeah, I know, I mentioned it earlier. That should make octanum quite a bit easier to realize.
VexPro doesn't sell a traction wheel smaller than 4" either
It is very easy to make your own small traction wheels out of delrin and plaction tread; I've seen several teams that did this.
DampRobot
23-11-2013, 20:49
Our variant of this was trivially different (biggest difference was using 2" tall tube versus 3") and I can confirm it was easier to manufacture than our WCDs. There just aren't a lot of parts.
Are they really quicker/easier to make? At this point you essentially do a COTS everything WCD, whereas the butterfly, you have to make the plates for the wheels and mill the tubing. I'd also expect that WCDs are a lot harder to get wrong than butterfly drives, at least for most teams.
Of course, basically all drives in FRC (except possibly swerve) are to the point where they work quickly and well 90% of the time. It's funny to think just a few years ago just driving around the field was considered something to be proud of.
AdamHeard
23-11-2013, 20:53
Are they really quicker/easier to make? At this point you essentially do a COTS everything WCD, whereas the butterfly, you have to make the plates for the wheels and mill the tubing. I'd also expect that WCDs are a lot harder to get wrong than butterfly drives, at least for most teams.
Of course, basically all drives in FRC (except possibly swerve) are to the point where they work quickly and well 90% of the time. It's funny to think just a few years ago just driving around the field was considered something to be proud of.
It was a loaded statement I suppose, as we haven't run a WCD with a COTS gearbox yet. That certainly brings it to be more comparable.
If we do WCD again next year, we'll be sticking with our bearing blocks. If we also purchased COTs blocks it makes the drives pretty comparable.
With our router in house the butterfly side plates are very fast to make.
AdamHeard
23-11-2013, 20:59
I know for sure that those are the sizes of the wheels(6" mecs, 4" tractions). We are pretty good friends with 973, and as Adam said, they are an running almost identical drive to 3928.
Sorry buddy, but it's 4" Omni w/ a 64T gear and a custom plastic wheel at ~ 2.4" Tread diameter. :(
For the sake of completeness, the pulley ratio from omni to traction is 18:30.
Sorry buddy, but it's 4" Omni w/ a 64T gear and a custom plastic wheel at ~ 2.4" Tread diameter. :(
For the sake of completeness, the pulley ratio from omni to traction is 18:30.
Oh, it's a 64t? Do the 72t gears have ground clearance problems, or did you just want it to run faster? I ask because I had been previously advised that 12t pinion on a 72t gear with a 4'' wheel ends up at 16 fps and works well.
I mean omnis, but looking back, it does look like you are correct. I'll delete the post just so it doesn't confuse anyone
Fair enough, I'll do the same.
AdamHeard
23-11-2013, 21:08
Oh, it's a 64t? Do the 72t gears have ground clearance problems, or did you just want it to run faster? I ask because I had been previously advised that 12t pinion on a 72t gear with a 4'' wheel ends up at 16 fps and works well.
I believe, but am not certain, that 3928 ran 12:72 (15.5 fps free speed).
We ran 12:64 (17.4 fps free speed) because we like things faster in California.
The extra ground clearance is a bonus. Even with the 64T a lot of crud was sucked into the gears.
I believe, but am not certain, that 3928 ran 12:72 (15.5 fps free speed).
We ran 12:64 (17.4 fps free speed) because we like things faster in California.
Cool, thanks for the info.
4464 currently has 12:72 for our offseason butterfly design, which is basically that but on a standard c-base frame and with delrin side-plates (the only CNC machining we have ready access to is a lasercutter at UMD). Might consider changing to 12:64 in later iterations if the ground clearance is a problem.
Aren_Hill
24-11-2013, 20:57
I believe, but am not certain, that 3928 ran 12:72 (15.5 fps free speed).
We ran 12:64 (17.4 fps free speed) because we like things faster in California.
The extra ground clearance is a bonus. Even with the 64T a lot of crud was sucked into the gears.
We had no issues with the gears being that close to the ground, for reference
And 973 was driving underweight which helps a fair amount.
Also Californians are weird.
-Aren
vBulletin® v3.6.4, Copyright ©2000-2017, Jelsoft Enterprises Ltd.