View Full Version : Searching For Swerve / Holonomic Drive Idea
Michael Niemiec
02-04-2016, 16:02
A mentor was describing a drive module that uses a ball instead of a wheel that is controlled and rotated to move the robot around. The module would be similar to a Holonomic / Swerve module. Both the mentor and I are unable to find the thread/ post where this idea was discussed. If someone recognizes this idea please link the post/thread where I can find out more.
Thanks in Advance.
I'm guessing it's something very different from TechnoKat Ball Drive (http://www.chiefdelphi.com/media/photos/15197) but similar in concept.
Michael Niemiec
02-04-2016, 16:12
Yes, this is what I had in mind.
Thank you very much!
Yes, this is what I had in mind.
Thank you very much!
Wait seriously? I just like bringing up that robot whenever I (remotely) can, but you're welcome.
Michael Niemiec
02-04-2016, 16:20
Yes, now that I see it, It is exactly what we were talking about. I would like to do something with it in the off season maybe. I would not use a kickball like that, but maybe a small steel ball of some sort. It seems easier and less expensive than a full blown swerve module.
trumpthero786
02-04-2016, 16:35
Yes, now that I see it, It is exactly what we were talking about. I would like to do something with it in the off season maybe. I would not use a kickball like that, but maybe a small steel ball of some sort. It seems easier and less expensive than a full blown swerve module.
I feel like steel would be very slippery on carpet. Traction would be difficult. I'm not sure I understand how something like this would function like a Swerve drive. If the balls are not powered, how is this any better than a rear wheel drive, omniwheel drivetrain? It does however look like something that would be a very fun project.
Michael Niemiec
02-04-2016, 16:42
I feel like steel would be very slippery on carpet. Traction would be difficult. I'm not sure I understand how something like this would function like a Swerve drive. If the balls are not powered, how is this any better than a rear wheel drive, omniwheel drivetrain? It does however look like something that would be a very fun project.
I wonder if a layer of Gaffers tape or similar would help with the traction, because I agree steel would slip. If the traction would work out, it would be like a cheaper swerve drive option. But if the traction problem couldn't be solved, a omniwheel drive would be better.
I think this would be cool to build, but hard to implement but holonomic/ swerve drives intrigue me and I am always trying to think of a new way to do them.
I feel like steel would be very slippery on carpet. Traction would be difficult. I'm not sure I understand how something like this would function like a Swerve drive. If the balls are not powered, how is this any better than a rear wheel drive, omniwheel drivetrain? It does however look like something that would be a very fun project.
Releveant (http://www.chiefdelphi.com/forums/showpost.php?p=142413&postcount=10) posts (http://www.chiefdelphi.com/forums/showpost.php?p=142476&postcount=30) about mechanics, but largely it's "simpler" and far cheaper to build than a full swerve drive. Arguably, the effectiveness now a days is questionable, but certainly something worth looking back into since there's work around for it.
I feel like steel would be very slippery on carpet. Traction would be difficult.
True.
I'm not sure I understand how something like this would function like a Swerve drive. If the balls are not powered
They look like they are are powered.
But as shown it does not have the same mobility as a swerve*.
*"swerve" meaning independent steering and drive on all wheels.
They look like they are are powered.
But as shown it does not have the same mobility as a swerve*.
[/SIZE]
Just because there's only two balls, right? If there were four 'ball modules' then it should have the same mobility as a swerve drive from my understanding. Or are there issues with the ball modules that limit their mobility?
Peyton Yeung
02-04-2016, 17:23
Having revamped the 2003 ball drive (https://www.youtube.com/watch?v=IZEFp8iaAcU) while in high school, the trickiest part about getting it to drive in my opinion is balancing the normal force on the ball by the drive rollers. Too much force and the ball won't move while too little and it won't strafe.
Alan Anderson
02-04-2016, 20:36
Yes, now that I see it, It is exactly what we were talking about. I would like to do something with it in the off season maybe. I would not use a kickball like that, but maybe a small steel ball of some sort. It seems easier and less expensive than a full blown swerve module.
What kickball are you referring to? The TechnoKats Ball Drive robot does use hollow steel balls, coated with a thin layer of polyurethane rubber to provide traction. You definitely don't want to kick one.
http://www.bedbathandbeyond.com/store/product/rome-industries-stainless-steel-gazing-ball/3167100?skuId=18156326 is similar to what was used.
Alan Anderson
02-04-2016, 20:43
But as shown it does not have the same mobility as a swerve.
I'm trying to figure out what it's missing that makes it not as mobile as a swerve. It can rotate at any time by applying different speeds to the left and right sides, and it can move in any direction at any time by varying the forward/backward and left/right drive motors. Rotating while traveling in a straight line is difficult, but only because it takes a lot of coordination, not because physics keeps it from being possible.
Rotating while traveling in a straight line is difficult, but only because it takes a lot of coordination, not because physics keeps it from being possible.
I would think rotating while driving straight with any drivetrain would be fairly difficult...
I would think rotating while driving straight with any drivetrain would be fairly difficult...
1717, 973, 16, and other swerve teams have done it before. It's possible, but as said, requires an absurd amount of coordination.
Roboshant
02-04-2016, 23:52
I believe this is a hemispherical omnidirectional gimbaled drive or a HOG drive.
I believe this is a hemispherical omnidirectional gimbaled drive or a HOG drive.
No, according to its name a HOG drive uses a tilting hemisphere that spins in one axis (parallel to the hemisphere's circle), whereas these use a full sphere that can spin in 2 axes.
Swerve (and mecanum) technology has progressed a lot since 2003, so today there are much simpler and better ways of making a true omnidirectional drive compared to this pseudo-omnidirectional drive.
Andrew Schreiber
04-04-2016, 13:28
1717, 973, 16, and other swerve teams have done it before. It's possible, but as said, requires an absurd amount of coordination.
I've also done it numerous times with omni wheel based drives[1]. With a gyro it's fairly simple actually. Ether has posted enough of the math at this point that it's not bad at all. If you prefer to derive it yourself it's some basic trig.
[1] https://www.youtube.com/watch?v=TW6fR-bQFgI In this video the driver is simply commanding the direction to go (field centric) and a rotation rate. The robot figures out everything else. No coordination needed, this was, I think, Brando trying it out for the first time.
Alan Anderson
04-04-2016, 13:59
Swerve (and mecanum) technology has progressed a lot since 2003, so today there are much simpler and better ways of making a true omnidirectional drive compared to this pseudo-omnidirectional drive.
I'm still trying to find out what makes people call the ball drive something other than a "true omnidirectional" fully holonomic system.
I'm still trying to find out what makes people call the ball drive something other than a "true omnidirectional" fully holonomic system.
Since only the front wheels can impart a sideways force on the robot, the robot will turn (w/ CoR between the back wheels). If all 4 wheels were balls, it would be a true-omnidirectional drive. As it is now, the robot drives more like a car, where the front of the robot can go sideways, while the back of the robot doesn't.
Andrew Schreiber
04-04-2016, 14:11
Since only the front wheels can impart a sideways force on the robot, the robot will turn (w/ CoR between the back wheels). If all 4 wheels were balls, it would be a true-omnidirectional drive. As it is now, the robot drives more like a car, where the front of the robot can go sideways, while the back of the robot doesn't.
3 Wheels, you don't need 4.
Alan - I believe this system would have a difficult time translating and rotating at the same time, but, as I haven't built one or played with one I can't confirm this without putting in a bit more thought. But I think that's what folks are getting at. But I think that's based on my mental model of the wheels being the same as an omni wheel and I'm thinking that's not right.
pwnageNick
04-04-2016, 14:52
oh boy have I got something for you.
Remind me to revisit this thread in a month when all the craziness of the current season is over. I've got a fun project from a few years ago to share.
Since only the front wheels can impart a sideways force on the robot, the robot will turn
You can compensate for that by "tanksteering" the bot in the opposite direction. But that compensation is based on dynamics (forces) not simple kinematics.
With a true swerve it is straightforward to solve the inverse kinematics to get straight-line motion while simultaneously turning. Just make the direction command field-centric.
IronicDeadBird
04-04-2016, 15:49
I heard of a recently released design called the BB 8 don't know much about it but it could be worth looking into.
Alan Anderson
04-04-2016, 16:08
Since only the front wheels can impart a sideways force on the robot,...
Are we talking about the same thing? The front wheels on the TechnoKats 2003 "ball drive" 'bot are omniwheels, and can impart only forward/backward forces. The balls are the only thing that can produce a sideways force.
Chris Endres
04-04-2016, 16:27
I feel like steel would be very slippery on carpet. Traction would be difficult.
Well, purposeful metal-on-carpet is illegal in FRC (according to R6), if that is what you were referring. I would go for a hard rubber ball if I would design this, probably with a ~ 5" diameter.
Are we talking about the same thing? The front wheels on the TechnoKats 2003 "ball drive" 'bot are omniwheels, and can impart only forward/backward forces. The balls are the only thing that can produce a sideways force.
Oh then I was looking at the robot backwards. Make that the back wheels (the balls). The point is with this configuration, the drivetrain cannot travel in all three dimensions (x, y, r) at the same time. You could do it using three wheels set up with two balls in the front (or back) and at least one omni wheel facing horizontal (perpendicular to the front).
I heard of a recently released design called the BB 8 don't know much about it but it could be worth looking into.
https://40.media.tumblr.com/d58ba18c7b490bc6f4a3fbb2597d88c8/tumblr_nzqeahO4Br1rj8mcho1_500.jpg
https://40.media.tumblr.com/d58ba18c7b490bc6f4a3fbb2597d88c8/tumblr_nzqeahO4Br1rj8mcho1_500.jpg
Fire game!
Would not mecanums be easier to implement than a swerve drive and a ball drive? As an offseason project I like swerve drives more if you have the resources to do it, or mecanum if not.
Would not mecanums be easier to implement than a swerve drive and a ball drive? As an offseason project I like swerve drives more if you have the resources to do it, or mecanum if not.
Depends on the ultimate goal of the project - if you're strictly trying to develop some omnidirectional drivetrain, than yes, mecanum would be far cheaper. But if you're looking into it from a learning standpoint and wanting to push the envelope, than developing some sort of ball drive would probably teach a lot more than a "conventional" drive.
pwnageNick
04-04-2016, 21:04
Also a traditional mecanum drive isn't much of a mechanical or design challenge. It is slightly harder than a standard tank, purely because of the 2 extra gearboxes.
Also I think we can all agree swerve is a more advantageous drive if the resources available to you allow you to use it.
What about a robot that's just one ball, like an oversized Sphero?
RyanShoff
04-04-2016, 22:30
I thought this would make a good bot:
http://www.gizmag.com/spherical-drive-system-omnidirectional-electric-motorcycle/24095/
I would think rotating while driving straight with any drivetrain would be fairly difficult...
If i understand what you guys are talking about correctly it really doesn't take that much coordination at all.
All the driver has to do input the the desired direction of rotation and then match that speed of rotation with the rotating of whatever control axis controls the forward/backwards, and strafing.
For example, a driver controlling the robot with 2 joy sticks one of which controls rotation by moving left and right, and the other controls all of the strafing commands. If the driver wants to move the robot in a straight line all the while rotating to the left, all he must do is move the rotation joystick to the left, while at the same time rotating the strafing joystick in a clockwise direction with the same rotation speed as the robot itself. By slowing down the speed that the strafing stick is being rotated the robot will veer to the left and by speeding it up the robot will veer to the right.
If i understand what you guys are talking about correctly it really doesn't take that much coordination at all.
All the driver has to do input the the desired direction of rotation and then match that speed of rotation with the rotating of whatever control axis controls the forward/backwards, and strafing.
For example, a driver controlling the robot with 2 joy sticks one of which controls rotation by moving left and right, and the other controls all of the strafing commands. If the driver wants to move the robot in a straight line all the while rotating to the left, all he must do is move the rotation joystick to the left, while at the same time rotating the strafing joystick in a clockwise direction with the same rotation speed as the robot itself. By slowing down the speed that the strafing stick is being rotated the robot will veer to the left and by speeding it up the robot will veer to the right.
Something akin to this (https://www.youtube.com/watch?v=p9WHMssEF4U&list=PL51FE6FC40790DD2B&index=2) basically. It's certainly do able, just takes a lot of work and a change in how you would intuitively drive.
Something akin to this (https://www.youtube.com/watch?v=p9WHMssEF4U&list=PL51FE6FC40790DD2B&index=2) basically. It's certainly do able, just takes a lot of work and a change in how you would intuitively drive.
I would have the operator control the rotation, and use field-centric control for the driver. That simplifies the controls greatly.
Alan Anderson
05-04-2016, 12:10
...with this configuration, the drivetrain cannot travel in all three dimensions (x, y, r) at the same time.
Why not?
Each ball is independently driven in the forward/backward direction, which yields control of y and r. They are driven sideways together, which would happen anyway since they're in the same line in that direction, and that controls x. The omniwheels just go along for the ride in x, and give the same forces as the balls in the y direction (and, by extension, r).
I've driven it. It spins in place. It drives and strafes. It moves diagonally very well. It can go in nice arcs whichever direction it is traveling or facing. With extreme care and very slow and halting motions, it has even made its way in a nearly straight line while doing a full rotation. It acts very much like a four-wheel Mecanum or angled-omniwheel drivebase.
Why not?
Each ball is independently driven in the forward/backward direction, which yields control of y and r. They are driven sideways together, which would happen anyway since they're in the same line in that direction, and that controls x. The omniwheels just go along for the ride in x, and give the same forces as the balls in the y direction (and, by extension, r).
I've driven it. It spins in place. It drives and strafes. It moves diagonally very well. It can go in nice arcs whichever direction it is traveling or facing. With extreme care and very slow and halting motions, it has even made its way in a nearly straight line while doing a full rotation. It acts very much like a four-wheel Mecanum or angled-omniwheel drivebase.
Having driven it, you probably know better than I do. I would have imagined that if you only get a sideways force from the back two wheels, you would rotate rather than strafe (the same way you would turn if your horizontal wheel in an H-drive was all the way at the back), but I could be wrong. If it drives like you are saying, then I guess it is fully omnidirectional.
Peyton Yeung
05-04-2016, 12:36
Having driven it, you probably know better than I do. I would have imagined that if you only get a sideways force from the back two wheels, you would rotate rather than strafe (the same way you would turn if your horizontal wheel in an H-drive was all the way at the back), but I could be wrong. If it drives like you are saying, then I guess it is fully omnidirectional.
If you look at the video I posted about it, all the strafing done is just done by the ball. I don't recall correcting the rotation as I strafed. The rotation while strafing is pretty minimal at worst.
Alan Anderson
05-04-2016, 13:01
I would have imagined that if you only get a sideways force from the back two wheels, you would rotate rather than strafe (the same way you would turn if your horizontal wheel in an H-drive was all the way at the back), but I could be wrong.
I don't know why you would imagine that. The pair of drive balls doesn't simply apply a fixed-orientation thrust. Each of them moves at a specific speed in a specific direction. When strafing directly to the side, that velocity is exactly sideways. There is no forward or backward motion of either ball. Unless they slip on the carpet, the robot simply does not rotate. Since the omniwheels are also not turning during a pure sideways strafe, they too act to keep the robot direction constant.
I think you're failing to recognize that the balls are completely controlled in both x and y directions. Unlike omniwheels or Killough cagewheels or Mecanum wheels, there is no direction in which they "slip".
vBulletin® v3.6.4, Copyright ©2000-2017, Jelsoft Enterprises Ltd.