Truly omnidirectional omniwheels?
 

I've looked at omniwheels a lot lately(brainstorming for next season and all), and I can't help but notice that if you need to move an omniwheel diagonally, it wouldn't work too well. Is this true? It seems to me like potential omnidirectional drivetrains are stopped because omniwheels wouldn't be able to work for their design. Has anyone designed an omniwheel with the edge wheels swiveling so that they can move in any direction? I know that if the little wheels went in the same direction as the overall wheel, it wouldn't work, but that can easily be stopped from happening with a slight design alteration.

So... would it be possible to make an omniwheel like that? It'd probably be expensive, but it would open up a lot of spiffy drive trains. Imagine a hexagon drive train(the result of brainstorming at one in the morning). Or a triangle with two wheels per side. The angles would usually stop it, but with the augmented design of omniwheels, they could be given the ability to work.

I only joined my team this season, and I was one of those, "Come and work on whatever needs to be done" people, so I typically ended up drilling holes in things. Thus, my knowledge of engineering from this season is... limited. Hence, I've come to the ChiefDelphi forums, because from what I've heard, that's where people know stuff. So tell me- is my idea good, or am I just spouting nonsense?

Re: Truly omnidirectional omniwheels?
 

the omniwheels taht i know of are either 90 degrees or on a 45 degree angle and yes if u get pushed at 90 degrees or 45 pending on which wheel you have its easy to get pushed i do not know if there are any omniwheels that swivel though im having trouble picturing how well that would work but thats all i can tell ya

Re: Truly omnidirectional omniwheels?
 

If you move a single omniwheel that way, it might be hard. But when you get them on a robot in a set of three or four, it's relatively simple. Imagine a four-wheeled omni drive set up as a plus sign. Set the wheels along the Y-axis driving one way, and the ones on the X-axis another. The robot should move diagonally (at least as far as I can remember for being up ten minutes >_<)

Re: Truly omnidirectional omniwheels?
 

I've done some omni-wheel research with my VEX kit. It is possible to make an omni-wheel based robot with 3 wheels, 4 wheels, 6 wheels... I guess you could make one with 5, or any other number, but I don't think there are big advantages to doing so.
What Billfred is saying is correct. It takes a little trig to make a 3-sided omni bot work, but it is possible: http://www.cs.cmu.edu/~pprk/

Re: Truly omnidirectional omniwheels?
 

No, I know about kiwi drive and all that. I'm just thinking of certain types of drive trains that would be really fun to build with a better kind of Omniwheel(since my team apparently wants to do holonomic next year and I want to have a more unique drive train than that). There's this design I've been thinking of that would be hexagon shaped, with one wheel at every side parallel to the side(not on the corners) It could be an octagon too, but that would take too many motors. But the omniwheel design is the important part. You know those rollers that are often under teacher desks or projector trays? Just put a smaller version of those into the slots where the sideways wheels usually go. Make a small modification so that it can't rotate enough to make itself go the same way as the larger wheel, and bang, you've got an omniwheel that can move more than just forwards and sideways on its own.

I think it would be great fun to develop. Except for the mill operator. He got mad at me when I mentioned I was thinking up an omniwheel that would take longer to make. Anyone else agree that this kind of omniwheel would be spiffy as anything for oddly shaped drive trains? Even for holonomic drives, I can imagine that the little wheels being about 45 degrees off from their usual range of movement. These little dealies might save some power normally spent moving omniwheels the wrong way.

By the way, does anyone have links to videos of mechanum and holonomic drives, specifically with the wheels showing? I haven't gotten to see either in action, and I'm curious as to their real maneuverability.

Re: Truly omnidirectional omniwheels?
 

This link still works i believe! It shows a prototype of our Kiwi drivetrain sometime before the start of the season back in 2002. We made those omni-wheels ourselves.....but they are very similar to the aluminum style that AndyMark produces today.

http://stuweb.ee.mtu.edu/~alkrajew/FIRST/kiwi.mpg

Re: Truly omnidirectional omniwheels?
 

Maybe a crab aka swerve drive is what you are looking for. 4 independent wheel/motor modules that can each rotate 360 degrees (together or independently, depending on the design). Search around and you'll find some pics.

Re: Truly omnidirectional omniwheels?
 

Sanddrag is right, if you want truly omni directional motion you either have to do crab drive like team HOT has done in the past or do a ball drive(there is a picture of it somewhere on the forums) which functions as a mouse ball does except with motors instead of the roller sensors. Other than that i know of no other way to be truly omni.

Re: Truly omnidirectional omniwheels?
 

I believe the technokats at one point made just for fun a ball drive. It was a robot that had two beach balls as wheels. They had two motors hooked up to each beach ball in a contact method. That way they could get the beach balls to roll in any direction, thus moving the robot. Clever, but probably not exactly what you're looking for. ;)

Re: Truly omnidirectional omniwheels?
 

Actually, the wheels were those garden globes you see from time to time. I believe some sort of grippy coating was applied as well.

(If you search hard enough, you'll also note that their ball drive saw action in 2003. Turns out that a team at one of 45's events had a robot busted beyond repair, so they pressed it into service to let them compete again. It even beat 45's robot in one match.)

Re: Truly omnidirectional omniwheels?
 

If you are doing this, simply pick up some transwheels from Kornylak, http://kornylak.com/wheels/transwheel-4000.html. That's what Team 95 used (4202KU's, since they have two rolls of rollers and thus roll smoothly) for their omnidirection test chassis (that we ended up not using for weight reasons).

Re: Truly omnidirectional omniwheels?
 

A standard omni-wheel, like the AM Trick wheels, should work for what you are trying to do. Keep in mind that when moving, unless going directly in the direction of an omni wheel, the entire wheel is turning (as well as the passive rollers).
Additionally a "holonomic" system can have any amount of wheels. A kilo-drive (kiwi drive) is a 3 wheeled variant, etc. Holonomic just basically equates to a vector-based omnidirectional system. It has a full range of 360 deg motion, and the direction of your force can be changed almost instantaneously (as soon as the motors accelerate). Holonomics suffer in other respects though. They are inefficient in terms of drawing max potential from your motors (often resulting in a low torque and/or speed). The exact loss of potential depends on the amount of wheels in the system. Additionally they require all of your wheels to remain in contact with the floor at all times to work properly. That often means that you are unable to travel up ramps, stairs, etc
A "ball drive" acheives the same functions as a holonomic drive. The motion of the "wheels" is based upon a vector of the forces applied by the two motors on the "wheel" (the "x and y axis motors"). So once again, you suffer an potential loss in how much you can get from your motors. Ball drives do not have the same problems with inclines as a holonomic drive does. Ball drives have a much smaller contact area with the floor, and often are made of a lower traction material (but if you have money/weight I'm sure you can use a higher traction ball), resulting in ball drives being relatively easy to push.
A "mecanum drive" (Jester Drive as dubbed by 357) is another omni-directional variant. This thread has links to explain the exact mechanics of a mechanum drive. Mecanum drives, at least those in FIRST, typically have rollers placed at a 45 deg angle to the wheel. Mechanum once again suffer in losses of the maximum potential of the motors, but they typically have solid traction and can climb inclines with the same ease as a "skid steering" system, if not better.
"Swerve" drives are yet another variant of an omni-directional system. They function by having the drive wheels (typically 4) physically rotate to match the intended direction of travel. Teams 71, 118, 1261, and many others have enjoyed tremendous success with this drive style. The "wasted potential" appears again in the form of the "Steering motors" cannot be used elsewhere on the robot and do not contribute to the power of the drivetrain. These systems typically require a low traction wheel to turn properly, so the traction issue emerges again. Unlike the ball, holonomic, and mechanum systems, a swerve drive cannot instantaneously change direction, it requires time for the wheels to change direction. The biggest problem with this system is often its weight and complexity.
"Crab drives" are a system that has a 2nd set of wheels/drive motors that will move the robot directly side-to-side, in addition to its standard drivetrain. This is also typically fairly complex, and does not offer true omnidirectional motion.
I'm sure that there are omni drive systems I have forgotten, but I have done my best to describe the advantages and dis-advantages of each.

Re: Truly omnidirectional omniwheels?
 

The complete ball drive can be found in this photo with a closeup in this photo. I can't find any threads about it right now and I don't know very much about it, save where to find these images :). You may wish to ask some of the Technokats Alums - Clark Gilbert, Austin Butler, Greg McCoy, and Kyle Gilbert come to mind - about this unique system.

Re: Truly omnidirectional omniwheels?
 

The best thread to read right now is probably this one:
http://www.chiefdelphi.com/forums/sh...t+ball+dr ive , especially the summary on the first page from Andy Baker. I'll look for the video I posted a long time ago and re-uploda it.

Re: Truly omnidirectional omniwheels?
 

Speaking of ball drive, I may have just figured out a way to reinvent the wheel. Think of a stepper motor. Instead of a traditional cylindrical rotor, use a spherical one. Like a golf ball with magnets instead of dimples. Now, use two stators that are 90 degrees apart from each other such that you can rotate the sphere about the x axis and about the y-axis. Voila, compact ball drive.

Now if you'll excuse me, I need to find a patent lawyer.

Re: Truly omnidirectional omniwheels?
 

Not likely gonna need a patent for those hemispherical wheels we were working on them this summer: http://www.chiefdelphi.com/forums/sh...ad.php?t=40770 but take a look at what we came up with its a cool design

Re: Truly omnidirectional omniwheels?
 

I don't think you guys are understanding my point... I know about the other omnidirectional drives(thanks for the links, however, I like information). My point is a way to make other omnidirectional drive trains by using another design of omniwheel. That... was my intended purpose for the topic, to ask whether another kind of omniwheel could be made that was more... well, omni. I know that all the other drives are successful, but those innovation awards are really appealing, and I think that another omnidirectional drive train could be made with my idea. I've already said it a few times- rotating rollers instead of just the wheels. Basically lets you have the miniwheels have their x axis rotate, with a little modification so that it can't rotate into the same direction as the wheel itself. This would allow there to be traction for the wheel going forward, and allow it to go any other direction without traction. That would allow wheels to face 45 degrees away from each other, , or any other degree, but not oppose each other. That would let there be more drive trains with opposing wheels. Which means more creativity.

Re: Truly omnidirectional omniwheels?
 

What about an omni wheel but instead of having a bunch of wheels around the main one, you just have a bunch of ball bearings stuck into it.

I'm not sure if something like this exists yet, I don't see why it shouldn't.

I see a problem with this though. If you were able to push an omni wheel in any direction, then it wouldn't be able to provide any traction when you rotated it with a motor. It would just spin in place.

Now, someone mentioned something about "the wheels on the bottom of teacher's desks."
I assume that you meant something like these:
http://www.photoflex.com/photoflex/p...300/caster.jpg
Those are called "casters", just for future reference.


EDIT: I just saw something about limiting the casters' rotation as so that they don't allow the wheel to spin in place. Ignore the second paragraph in this post.

Re: Truly omnidirectional omniwheels?
 

Hey, i remember seeing your team at FLR. Where did you get the inspiration for your omni wheels?

I <3 OMNI wheels!

Re: Truly omnidirectional omniwheels?
 

Yeah, casters are what I've been talking about. Although with further thought, I've realized that if you have the rotation limited and the caster needs to move that way because the robot is moving in such a way that it wouldn't spin the other way... might have to do something with springs so that the wheels move to default when off the ground. Of course, ball bearings were my first idea, but then I realized the whole 'robot not moving' thing, so I switched my idea into the (heavily modified)casters plan. Probably not traditional casters, though... might want the axis in the center.

And FLR was awesome. Team 229... can't say as I can recall that one, and you have the 1126 site in your profile. Link me to a pic so I can remember. 1126 was awesome, of course. I remember- you guys CRUSHED the opposition. It was so fun to watch! Your shooter was pretty danged great. We(the Duct Tape Bandits) couldn't say we were spiffed to lose, but we had the six hour bus ride home on the AWESOMEST BUS EVER to make us feel better. Seriously, we had a 42 inch widescreen HDTV, strobe lights, lasers, a fog machine, ceiling mirrors, fiber optics, subwoofers, hardwood flooring, sideways leather seating(with tables and cupholders), those things with the electricity where you touch them and all the electricity is attracted to your hand... it was the greatest bus ever. Most definitely my most fun experience in a while.

Also, the inspiration for the omniwheels came when I was bored enough to try to think of an innovative drivetrain, so I was thinking up a hexagon design, until I realized it wouldn't work with omniwheels because the little wheels wouldn't be facing the proper way, and if you moved them so they faced the right way, they wouldn't be facing the right way for moving other directions. So I decided I'd think of a way to make a better omniwheel that would ACTUALLY be omni, instead of just forward and sideways. This was all at one in the morning, of course... took a nap soon as I got home, woke up at 11, didn't feel like doing homework... all that great stuff. You can think up crazy stuff at one in the morning. Best time for innovation.

...yeah. So, with modded casters(methinks they'd need some reinforcements to hold a hundred pounds of metal moving at high speed- I'm thinking some sort of box protection design with stronger sides for holding the wheel so the axel doesn't break under stress), omniwheels can be made to be MUCH more omnidirectional for weird drive trains like the hex drive train I want to do.

...I need some sort of CAD program for my computer so I can draw this. Or I can use one of my dad's programs. He has Python. Would that work for CAD drawings?

(I'm just posting all this stuff here instead of talking with my team because my team probably wouldn't listen to me-not very popular with people because I have problems censoring myself in conversation and am hyperactive)

All right. More opinions on my(hopefully now clarified) idea?

Re: Truly omnidirectional omniwheels?
 

well, i am no longer on 1126. i stepped up to college this year and am on 229. Clarkson University robotics. The reason why i asked where ya got the inspiration for those wheels, i that i made ones in 04' that looked similar. heres a picture.

also here is a close up of the wheels.


There is more info on www.gosparx.org too

Re: Truly omnidirectional omniwheels?
 

Erm... I'm afraid I can't see the resemblance between those omniwheels and my idea. They look like regular(multicolored) omniwheels to me... what's making them different from the normal kind?

Re: Truly omnidirectional omniwheels?
 

i was stating to the ones made on this years robot of yours.

http://robotphotos.org/v/first/finge..._7564.JPG.html

Re: Truly omnidirectional omniwheels?
 

No, it is not true.

Omniwheels move diagonally just as easily as they move forward. Perhaps more easily, since the motor doesn't need to be turning as quickly. They are called omniwheels because they indeed move omnidirectionally.

Re: Truly omnidirectional omniwheels?
 

Ah, yes. Very similar to our design. Great minds think alike. Unfortunately, I didn't know what an omniwheel was before this year. So I wasn't involved in that aspect... which means I don't get to qualify as a great mind. Curses.

And there are pictures of our robot online? Kickin! Those are some danged good pictures too... although it does give the inherent stalker-ish aspect, considering there's a picture of me in there... ah well.

Re: Truly omnidirectional omniwheels?
 

...I can't visualize omniwheels working perfectly at a 60 degree angle from the direction they're suppose to roll...

Re: Truly omnidirectional omniwheels?
 

When the robot is moving in the direction of rotation, the wheel rotates and the rollers are stationary on their axes. When the robot is moving perpendicular to the direction of rotation, the rollers rotate and the wheel is stationary on its axis. Any vector in between these produces rotation on both the rollers' axes and the wheel's axis.

Re: Truly omnidirectional omniwheels?
 

I don't know about your team, but our motors use an autolocking system so that it's harder for us to be pushed. Thus, whenever a motor isn't receiving instructions, it locks itself up, so if certain motors are moving(say, two matching sides in a hex drivetrain design) while others aren't, the wheels wouldn't be rolling to go along with the miniwheels. Omniwheels wouldn't be very omni in that situation. And that's the point of my idea.

Re: Truly omnidirectional omniwheels?
 

I've messed around with this idea in order to replicate something I saw in a music video.

A company called Animusic ( www.animusic.com ) makes this sweet midi driven animations... I've always wanted to re-create a scene out of one of the animations... so I'm going for the 3 slightly humanoid looking robots on the "Starship Groove" animation... anyhow back to the story...

The spherical track motor works exactly like a linear track motor except for the fact that the micro-electromagnets are arranged within a sphere made of a non-ferrous material. For my design, this sphere would fit into a matching hemisphere, very slightly larger than the sphere. This, in effect creates a fully functional low-load-bearing ball and socket joint.

OK thats the overview, now heres the problems with it:

There would have to be a processor dedicated to each joint. The design requires at least 3 (4 if it is going to be stable) pulse width modulation channels. It would never be FIRST legal since the victor's update rates are over 26 times too slow for updates. The PWM period would most likely be something like 500ns or so, because of the high instability of intersecting lines of magnetic flux. :ahh:

The joint itself would have to be encoded for position somehow, perhaps with a hall cell array or something like that. The updates for the servoed position would have to be done every 500ns to keep up with the pulse train, the high update rate is needed to overcome the problems with the intersecting flux lines.

So in a nutshell, here's how it is:
1. Would have to be based on a DSP running around 100MHz for a clock

2. It will eat TONS of current

3. Has to be run at a very very high voltage if you want to get much speed
or load capacity out of it.

4. Will require one UART for positional data input will need 4 channels
of simultaneous high speed PWM, run through a mux chip of some kind
in order to drive the hundreds of micro-electromagnets inside of the
sphere.

5. Will need a super high speed latch chip in order to retrieve data from all
those hall cells.

6. Driver board will have to have a power transistor or mosfet for every
single electromagnet.

7. All parts within about 6 inches of the thing will have to be non ferrous.

8. And oh yeah, the programming will be extremely hard (servoing in 2
dimensions)

:yikes: Did I lose anyone?

Granted, this is a design which would work for a positional as well as a speed based spherical track motor. But, like I said, my application required positional. It would work for motion purposes as well, though. I'm putting this on my "Things to do before I die" list.

-Q

Re: Truly omnidirectional omniwheels?
 

I originally thought the same thing but I think he means that they wouldn't move if the driveshaft isn't rotating.

Re: Truly omnidirectional omniwheels?
 

That's true of any wheel, so I don't really think it's a valid concern about omniwheels in particular.

Re: Truly omnidirectional omniwheels?
 

You can't have the full wheel rotation coupled with the roller rotation to move omnidirectionally if the full wheel can't roll. That's the point there.

Re: Truly omnidirectional omniwheels?
 

the hex drive train that you're talking about would definitely work with traditional omni wheels. you're right about the wheels not being able to free-wheel when the motor is unpowered, but this is a good thing. you want the rollers to only free-wheel in a direction that has another powered wheel opposing it, so that the powered can control motion in that direction. if the rollers swivel, then the drive will be difficult to control, and will have virtually no traction (how much traction does a desk chair get?).

i'm sorry if this doesn't make sense; omni directional drive trains are difficult to explain without delving into vectors trigonometry.

Re: Truly omnidirectional omniwheels?
 

I guess I still don't understand the point. Are you trying to make an omniwheel that rolls freely in every direction? A standard omniwheel does that. Are you trying to make an omniwheel that rolls freely in every direction even when it's prevented from turning around its main axis? That would work exactly against your reason for locking the drive shaft in the first place.

Re: Truly omnidirectional omniwheels?
 

I think he wants an "omni"wheel that will roll freely in every single direction except forward/backward. (Thus, the motors will still drive the wheel along one axis.)

With traditional omniwheels you can push them only sideways if the driveshaft isn't moving. With the wheels he's talking about you could still push the wheel at a 45 degree angle with forward being 0 degrees.

It wouldn't have very much traction though, but I think that's the point.

Re: Truly omnidirectional omniwheels?
 

Yeah, that's what I'm going for. Of course, a more senior team member is telling me I'm dumb, so most likely it won't ever happen. Of course, I can still dream.

The locking of the motors would still have its intended effect because from almost any angle there would be wheels going forward or close enough forward so that the wheels wouldn't let it roll. Or something akin to that.

And Roger, I'm afraid I can't... really visualize your statement of the problems. Someone needs to invent a machine that visualizes thoughts and sticks them on a screen. To be used for purely scientific purposes.

Re: Truly omnidirectional omniwheels?
 

They have... and i believe it's called microsoft Paint. ;) I can't visualize very well how this differs from a normal omnidrive except that it has 6 wheels... and powering 6 wheels will be a challenge since you dont have 6 identical motors.

Re: Truly omnidirectional omniwheels?
 

Are you talking about the brake/coast jumpers on the victors? You can set the jumpers on the speed controllers to short the motors with a neutral command (127), this causes them to brake. You can also set them to coast on neutral, which lets the motors free spin. The jumper block is right below the PWM cable port on the victor.

Re: Truly omnidirectional omniwheels?
 

Use 2 motors total, and use two of them to power all 6. We this year had 4 wheels, two motors. One wheel was direct drive from the gearbox, the other was 1:1 ratio chained with the other wheel that was direct driving.