Hello Everyone,
I am currently looking at different types of gearboxes and how to optimize them. This has led to my looking at the various types of overall drive systems. I believe I understand what skid and crab drive are, but I am clueless about what a swerve drive is.
Can someone please explain what these (and any unnamed) drive types are?
Thanks,
indieFan
“crab” and “swerve” are two names for the same drive type
*Originally posted by RogerR *
**“crab” and “swerve” are two names for the same drive type **
These two terms are common nicknames for an omni-directional drive.
The term Crab, comes from… well, a crab. If you have ever seen a crab walk, you know that crabs walk sideways. This makes sense since omni-directional drives can go sideways like a crab.
The term swerve is attributed to the agile motion of omni-directional drives. You never really know where they are going next, they just swerve all over the place very gracefully… most of the time anyway :yikes:
*Originally posted by WakeZero *
**These two terms are common nicknames for an omni-directional drive. **
there was a large debate (over the summer, i think) in which it was determined that crab and swerve drive lack the three degrees of freedom required to be a true ‘omni-directional drive’; they only have two, rotation and x-direction.
a killough(sp?) platform or ‘holonomic drive’ has three degrees of freedom, and therefore is a ‘true’ omni-directional drive.
the first FIRST killough platform (i believe) was built by team 857 in 2002, and was called the ‘kiwi-drive’.
*Originally posted by RogerR *
**there was a large debate (over the summer, i think) in which it was determined that crab and swerve drive lack the three degrees of freedom required to be a true ‘omni-directional drive’; they only have two, rotation and x-direction. **
Bah!
*Originally posted by RogerR *
**there was a large debate (over the summer, i think) in which it was determined that crab and swerve drive lack the three degrees of freedom required to be a true ‘omni-directional drive’; they only have two, rotation and x-direction.a killough(sp?) platform or ‘holonomic drive’ has three degrees of freedom, and therefore is a ‘true’ omni-directional drive.
the first FIRST killough platform (i believe) was built by team 857 in 2002, and was called the ‘kiwi-drive’. **
If we can get ahold of (raise the money to buy) quality omniwheels we are going to use a four-wheel Killough variant (wheels 90 degrees apart instead of 120 degrees apart).
Its been worked on, and should actually be mechanically simpler and much more robust than last year’s drive. Of course, we will only go ahead if it is consistant with our strategy, which will be decided come January.
The programming will be tough, but thats not my deptartment (i’m just the lead designer). Control should be a breeze. If we manage to get the solid-state gyro online enough to give us a consistant heading so that pushing the joystick north, regardless of the facing of the robot, will cause the robot to move north. and same with all the other directions. With a twist-stick joystick, we can add translational and rotational movement all into one hand . . . it is a good plan, I think. We will see. The biggest problem will be to make the chips and drills run at the same speed not only at 100% throttle, but at all the various percentages in between. Like I said, we will see how things go.
*Originally posted by Frank(Aflak) *
**If we can get ahold of (raise the money to buy) quality omniwheels we are going to use a four-wheel Killough variant (wheels 90 degrees apart instead of 120 degrees apart).Its been worked on, and should actually be mechanically simpler and much more robust than last year’s drive. Of course, we will only go ahead if it is consistant with our strategy, which will be decided come January.
The programming will be tough, but thats not my deptartment (i’m just the lead designer). Control should be a breeze. If we manage to get the solid-state gyro online enough to give us a consistant heading so that pushing the joystick north, regardless of the facing of the robot, will cause the robot to move north. and same with all the other directions. With a twist-stick joystick, we can add translational and rotational movement all into one hand . . . it is a good plan, I think. We will see. The biggest problem will be to make the chips and drills run at the same speed not only at 100% throttle, but at all the various percentages in between. Like I said, we will see how things go. **
Getting the Chips and Drills to run the same all the time is just in our gearing. You have many options to that.
-Pat
Alright… after thinking about it, I guess my definition of an omni-directional drive would be:
A driving mechanism that allows 360 degree directional movement independent from the frame.
My thoughts are that you NEVER can or need to go in more than one direction at a time… so being able to do them all at the same time is a pointless criteria.
Therefore, under this definition, I consider Swerve Drive an omni-directional drive :yikes:
*Originally posted by WakeZero *
**Alright… after thinking about it, I guess my definition of an omni-directional drive would be:A driving mechanism that allows 360 degree directional movement independent from the frame.
My thoughts are that you NEVER can or need to go in more than one direction at a time… so being able to do them all at the same time is a pointless criteria.
Therefore, under this definition, I consider Swerve Drive an omni-directional drive :yikes: **
The definition that was given in that thread, however, required the robot to move in the X, Y, and rotational directions simultaneously. And, as for “never” going in more than one direction, did you ever see the Kiwi Drive, or even the video? There’s some awesome stuff you can do when spinning and going at a diagonal.
That being said, however, swerve is more than enough mobility for just about anybody. There are advantages to swerve over Killough, and vice-versa, but you can do well with either.
when I say that i will be tough to make the chips and drills to run at the same speed . . . I do not mean at full power. However, as I understand these things, if you were to take a chip and a drill, gear them so their free RPM’s are equal, and then increase the voltage from zero to twelve at the same rate, the different motors will have different rates of increasing . . . ness . . . . osity . . . . or something. I mean, that, while you can sync the motors at 12 volts, they won’t nescessarily be synced when you are giving them half power. Does this make sense? So, our base will run a straight line at full speed, but will pull to one side at half power and stuff. This isn’t that large a problem, but its something to think about.
Speed, Power, And, Maneuvarubalty
SPAM seems to do well with Tank treds
Start, With, Attacking, My, Pproblems
Speed, Width, Agility, Manurvibilty, Power
And i just made these up for Swamp
A Crab drive seems Great because you can point all your power in any direction, But omni wheels only a fraction at the most half in any given direction
*Originally posted by Frank(Aflak) *
**when I say that i will be tough to make the chips and drills to run at the same speed . . . I do not mean at full power. However, as I understand these things, if you were to take a chip and a drill, gear them so their free RPM’s are equal, and then increase the voltage from zero to twelve at the same rate, the different motors will have different rates of increasing . . . ness . . . . osity . . . . or something. I mean, that, while you can sync the motors at 12 volts, they won’t nescessarily be synced when you are giving them half power. Does this make sense? So, our base will run a straight line at full speed, but will pull to one side at half power and stuff. This isn’t that large a problem, but its something to think about. **
Hi Frank,
The best options is to gear the motors such that the torque-speed curve lines are parallel for both motors. Then, for the motor with more power (ie the one further away on the torque-speed graph), you want to multiply the voltage applied by a constant less than one to bring the curve inward to the weaker motor. For example, the weaker motors would be used fully (PWM range -127 to 127) while the stronger motors would not be used fully (something like -108 to 108). This does waste power, but believe me, with four powered motors geared to an appropriate gear ratio and with high-traction wheels (see below), you will not have a power problem. Ensuring the same torque-speed curve using my method above will also help a LOT with control problems. There will still be some variance from motor to motor and over time and with varying heat conditions: If I were to do it, I would actually place encoders on each wheel, and use feedback control to ensure the robot is going in the commanded direction… But I don’t know if FIRST rules allow purchasing of encoders. I know there are teams that make their own with the optical sensors. But to start with, ignore the encoders idea, and if things don’t work well enough, you might consider it.
Now, about wheels, I have a lot of experience designing omniwheels, Here is a design I worked on last year (I’ve also attached a picture of final product):
Exploded view of wheel
Engineering Drawing
Engineering Drawing of Lexan “Hub”
Engineering Drawing of Roller (small rollers around the wheel)
CAD View of Robot w/ Wheels
The wheels have lexan hubs and delrin or aluminum rollers with rubber Buta-N o-rings wrapped around the outside of the rollers. The coefficient of friction between wheel and ground of this particular design is 1.3 for a felt-like material. I suspect it is at least 1.5 (probably closer to 2.0) on a FIRST carpet.
You can machine very similar wheels with manual mill & lathe, but you will need a rotary table to do the lexan pieces. It is better if you have access to CNC. You will also need to make larger wheels, probably about double the size of these wheels (which are 2" diameter). You can order O-Rings from McMaster or try and find them at a hardware store.
If you are really interested, you can actually read our mechanical design documentation at:
Cornell Robocup 2003 Mechanical Design Documentation
There may be some things of use in the documentation that I did not include here.
Good luck! And let me know if you have any questions.
*Originally posted by Tytus Gerrish *
**A Crab drive seems Great because you can point all your power in any direction, But omni wheels only a fraction at the most half in any given direction **
Not entirely true… with a 4 wheel omniwheel design you can focus at minimum 1/2 your total power in a direction. This is considering that 2 motors power the X direction and 2 motors power the Y direction. Simple trig tells us the maximum we can focus (in a direction of 45 degrees) is 0.70 of all 4 motors. For the 3 wheel omnidirectional design, due to symmetry, you will always exert 0.57 times the total power in any direction. It is possible to increase this to 0.66 times the three motors if you orient the wheels parallel to the Y instead of perpendicular, however this is at the trade-off of losing the ability to control your heading (making this option not really feasible, unless you didn’t care what your heading was such as with a saucer-shaped wedge robot).
Keep in mind that with the 4-wheel omnidrectional platform you can have your Y-axis geared for high-speed and your X-axis geared for high-torque without motor fighting.
As said above, with robots drive systems there are three variables to control. Heading, Y-axis velocity, and X-axis velocity.
Tank drive gives control over heading and y-axis velocity, but you can only manipulate one of those at any given moment.
2-wheel / caster drive gives a little bit more control over the two, and you can even do both at the same time for turns up to 45 degrees. However, you cannot make a 90 degree or more turn while maintaining Y-axis velocity.
Swerve/crab drive gives the illusion of omnidirectionibility due to the ability to change the robot heading without changing robot orientation. Still, you only have control over robot heading and Y-axis velocity. Although, you have the ability to manipulate both at the same time.
3-wheel/4-wheel holonomic drive systems have true capability to manipulate all 3 at once, as Jeff said above. It’s possible to change robot heading/orientation while maintaining a constant velocity in a certain direction.
*Originally posted by Jnadke *
**3-wheel/4-wheel holonomic drive systems have true capability to manipulate all 3 at once, as Jeff said above. It’s possible to change robot heading/orientation while maintaining a constant velocity in a certain direction. **
I still think this is an unnecessary requirement in the definition of an omni-directional drive :rolleyes:
Guys, this debate is basically just about terminology. Everyone has their own.
Here is mine (and 229s):
Omni-Wheel - A wheel that rotates normally and functions as a wheel, but can also be moved perpendicular to it’s rotation. Like the wheel Patrick posted above. Many teams use these to aid in turning. (Hotbot 2001)
** Omni-Drive** - A drivetrain that consists of multiple “sets” of omni wheels in different (usually at perp.) directions. This alows the robot to move in any direction simply by powering the different wheels at different speeds. (think Metal-In-Motion 2002, which was a 6 wheeled variety. A 4 wheeled variety can be seen here .) I would consider 857’s kiwi-drive to fall under this category.
Crab Drive - This type of drivetrain has 2 sets of drivetrains set perp. to each other. Instead of using omni-wheels and having all wheels on the ground, it somehow actuates things so that only one set is on the ground at a given time. (Good examples of this include WPI 2003, 358 2003, TRIBE 2002).
Swerve Drive - The individual wheels can “spin” changing the heading of the robot. It can move in any direction. (think Wildstang 2003/2002, Chief Delphi 2001/2002) There are many variations on this.
2-Wheel Swerve- Two wheels “spin” like a normal swerve drive (See above) the other wheels are either casters, or omni-wheels (see definition above). (see Chief Delphi 1998/2000).
Articulated-Steering - Some, or all of the wheels will turn like a car to aid in turning. 2 cool examples of this include Hotbot and Thunderchickens 2003.
Skid-Steering, or Tank Drive - This is the most standard type of drivetrain. All the wheels/tracks/whatever are pointed in one direction. The robot turns by reversing one side, while powering the other side forward. Examples of this include… almost everyone. When a robot in this configuration is turning, some energy is lost to “side load” because the robot is essentially dragging some of it’s wheels across the floor sideways. Some teams use dropdown casters/skids to aid in turning.
This is just my terminology for things. It isn’t always the most “technically correct” terminology, but this is what I’ve developed over 4 years in this program. shrug it works for us.
Perhaps this will benefit, someone…
John
*Originally posted by WakeZero *
**I still think this is an unnecessary requirement in the definition of an omni-directional drive :rolleyes: **
om·ni·di·rec·tion·al ( P ) Pronunciation Key (ahm-nee-die-rek-shu-nal)
adj : not directional; being in or involving all directions
Let’s apply this to radio waves. Sure, with a rotator a directional antenna can signal in a certain direction well, and has the ability to change directions, but can it signal in multiple directions at once? No.
As the definition says, the crab/swerve/synchro drive is directional. It cannot move along other axises without first changing the direction of the wheels. True holonomic drives have this instantaneous ability to move in other axises than oriented. Some more reading. A suitable analogy would be if I cold move really fast, say 1/4 the speed of light. I could say I teleported myself. To the average person, they’d believe it, but the educated person knows better. This analogy is suitable because the fast and powerful motors we have allow us to turn the swerve wheels very fast. If we did not have such powerful motors, the limitations would become aparent.
Everyone is entitled to their own opinion, I’m just telling you the dictionary and accepted definition. Nobody said yes couldn’t mean no.
*Originally posted by patrickrd *
**Hi Frank,The best options is to gear the motors such that the torque-speed curve lines are parallel for both motors. Then, for the motor with more power (ie the one further away on the torque-speed graph), you want to multiply the voltage applied by a constant less than one to bring the curve inward to the weaker motor. For example, the weaker motors would be used fully (PWM range -127 to 127) while the stronger motors would not be used fully (something like -108 to 108). This does waste power, but believe me, with four powered motors geared to an appropriate gear ratio and with high-traction wheels (see below), you will not have a power problem. Ensuring the same torque-speed curve using my method above will also help a LOT with control problems. There will still be some variance from motor to motor and over time and with varying heat conditions: If I were to do it, I would actually place encoders on each wheel, and use feedback control to ensure the robot is going in the commanded direction… But I don’t know if FIRST rules allow purchasing of encoders. I know there are teams that make their own with the optical sensors. But to start with, ignore the encoders idea, and if things don’t work well enough, you might consider it.
Now, about wheels, I have a lot of experience designing omniwheels, Here is a design I worked on last year (I’ve also attached a picture of final product):
Exploded view of wheel
Engineering Drawing
Engineering Drawing of Lexan “Hub”
Engineering Drawing of Roller (small rollers around the wheel)
CAD View of Robot w/ WheelsThe wheels have lexan hubs and delrin or aluminum rollers with rubber Buta-N o-rings wrapped around the outside of the rollers. The coefficient of friction between wheel and ground of this particular design is 1.3 for a felt-like material. I suspect it is at least 1.5 (probably closer to 2.0) on a FIRST carpet.
You can machine very similar wheels with manual mill & lathe, but you will need a rotary table to do the lexan pieces. It is better if you have access to CNC. You will also need to make larger wheels, probably about double the size of these wheels (which are 2" diameter). You can order O-Rings from McMaster or try and find them at a hardware store.
If you are really interested, you can actually read our mechanical design documentation at:
Cornell Robocup 2003 Mechanical Design DocumentationThere may be some things of use in the documentation that I did not include here.
Good luck! And let me know if you have any questions.
- Patrick
We used a similar design to those (More like Wildstang 2000) and we couldn’t get the o-rings to last longer than about 10 minutes once you start spinning the wheels on the carpet. We ended up machining the rollers and putting a knurl on them which lasted a lot longer
*Originally posted by Tytus Gerrish *
**A Crab drive seems Great because you can point all your power in any direction, But omni wheels only a fraction at the most half in any given direction **
not quite.
say you have the wheels like so:
. . _
. .[_]
you only get .5 of possible max power if you are moving directly verticle or horizontal. however, you go diagonally (45 deg. from horizontal) you will get full power.
To PatrickRd: Thanks for the offer of help with the omniwheels, however we don’t have access to a machine shop and I think the best way to go is to buy our wheels. I found this source (kornylak) that looks perfect, about 5 inch diameter, polyurethane rollers (wooo!), made of aluminum. Only downside is the wheels we would need would total to 516 USD. We can probably do that . . . . if we raise 6large in time to apply for the NASA grant again.
pics:
http://www.omniwheel.com/images/omniwheel/interoll-fullx.jpg
The wheels pictured have 3 rollers, you need two wheels back-back to provide for 360 degrees of rotation.
Here are schematics:
http://www.omniwheel.com/images/omniwheel/interroll-web.jpg
linkies:
omniwheel
I think that this is the best way to go for long-term durability and for overall funcionality (any rollers we make probably will not be as efficient as these.
downside, is again, 64 dollers per wheel. But we will be able to swing it. I asked Kornylak is they would like to sponsor us, so maybe we can swing it for cheaper. Who knows. These are what I’m drawing around as of now.
edit: These wheels are also ideal because they are nearly circular, so there will be smooth weight transfer from one roller to another and it will make for a smoother ride than, say, the lexan/disc wheel you pictured above. It is also probably lots easier on the wheel and the gearbox.
I’m reading that documentation, its pretty cool. Especially your nod to strongbad: “Eating one battery . . . . Eating Five Batteris”
It was so serious up until then, too. I guess you college students just didn’t have it in you to stay that way, huh?
*Originally posted by Jnadke *
om·ni·di·rec·tion·al ( P ) Pronunciation Key (ahm-nee-die-rek-shu-nal)
adj : not directional; being in or involving all directions
Meriam-Webster
Main Entry: om·ni·di·rec·tion·al
Pronunciation: "äm-ni-d&-'rek-shn&l, -“nI-, -(”)dI-, -sh&-n&l
Function: adjective
Date: 1927
: being in or involving all directions
Same as your definition above.
So, if you put boxes on both a killough and a swerve drive, both boxes could move in ALL directions without changing the orientation of the box. Since this fits the definitions above, both killough and swerve are omnidirectional.
Yes, you can be more technical and argue that swerve can’t do them all at once… but like I said before, what is the point? You can’t physically be going in two directions at the same time, so why make that a criteria? :yikes:
Again, these are my thoughts. I don’t however want this turning into a flame war; we should all have the ability to reason and debate without getting angry or start calling each other names.
Oh, one more thing.
*Originally posted by Jnadke *
Nobody said yes couldn’t mean no.
George Bool kind of did. The above statement is what I blame for the continued digression of moral law in the world. It’s the kind of thinking that says, “In his mind, there is nothing wrong with killing someone; therefore, it’s ok and we should be more understanding” :ahh:
None of the proposed drive systems are omnidirectional because they don’t involve the z-axis.
