Experimental Ball Drive

[Sepsis900]

Howdy All. My team is planning on trying out an idea called a "ball-drive" for this year's competition, and i wanted to utilize this forum's engineering expertise, to see if this can somehow be made into a viable drive system, or if this is just a dumb idea that will embarass us at regionals. The thing is designed to work like a mouse-ball, except instead of providing input (like a mouse does) based on the ball's movement, the ball is driven by several wheels (probably small omni-wheels), that rotate it around the X, Y, and Z axes.

The frame will be circular for this application, and will have 5 caster wheels around the edge. The ball itself will be a hollow metal ball coated in either rubber or another grippy material. It ought to look something like the picture attached. (http://www.chiefdelphi.com/forums/at...tid=3629&stc=1 )

My opinion is that the drive is not the best idea, and here's why:
-The area of contact between the ball and the floor is very small, leading to the problems of:
---movement. If we accelerate too fast, the ball will start slipping, and our traction disappears
---rotation. turning 120 pounds of robot with such a small area of contact (directly in the center) is difficult at best and impossible at worst
---carpets. If the floor of the FIRST arena is made of carpet and the ball of rubber, you get almost no traction at all (because the carpet moves under the ball), which is compounded by the problems listed above. (can be remedied by gratuitous use of velcro)
-The robot won't be able to go up inclines. If the front end is on an incline and the back end on the floor, the ball is off the ground and not providing power.
-If the center of gravity is not directly over the area of contact with the floor, there is more thrust on one side of the COG than the other, which leads to unwanted spin.

The use of a fast, maneuverable robot was usually advantageous in past competitions, and we're shooting for that this year. The team suggested the ball because it has the potential of providing multi-directional thrust and rotation at the same time, like a holonomic drive. But the idea of a single ball is a lot simpler than everything involved in said holonomic drive.
We're talking to professional engineers on what their opinion of the ball-drive is, and considering a lot of the problems and benefits ourselves, but any advice or opinions you guys can contribute will be greatly appreciated!
Thanks!

[sanddrag]

Originally Posted by Sepsis900

(can be remedied by gratuitous use of velcro)

Not legally it can't. At least not in prior years.

Anyway, it has been done before and worked remarkably well http://www.chiefdelphi.com/forums/pi...&quiet=Verbose

http://www.chiefdelphi.com/forums/pi...&quiet=Verbose

[Sepsis900]

Interesting. That's a pretty neat design. I'll make sure to post the pictures on our team forum, thanks!

[KyleGilbert45]

Ahh yes, the ball drive. I would search the forums here for "technokat ball drive".

Here are some threads that you may be interested in.

http://www.chiefdelphi.com/forums/sh...t+ball+dr ive

http://www.chiefdelphi.com/forums/se...earchid=493678

Mark Koors was the engineer that designed the ball drive. He doesn't post too often, but if you have any questions that you cannot answer from the above threads [which are full of information] you may be able to give Andy Baker a shout. If you want any additional images, you may be able to get ahold of my brother, Clark Gilbert . Most of the past Technokat images are from him, so I'm sure he has quite a few saved.

[sanddrag]

I'm not an expert on patent laws but make sure whatever you do doesn't cause any problems because the T-kats bot/design is patented.

[Andy Baker]

Originally Posted by sanddrag

I'm not an expert on patent laws but make sure whatever you do doesn't cause any problems because the T-kats bot/design is patented.

This just means that someone besides Delphi should not use this design with the intent of making money from it. FIRST teams can still use this idea to create a drive base (I have already PM'ed Sepsis900 to offer advice). Delphi management simply wanted to protect their rights by filing for a patent covering this specific design, not to keep other FIRST teams from using it.

Further development is encouraged and FIRST teams could make this drive system better.

If someone takes this idea and develops a more elegant iteration to improve performance, then they will have an invention of their own. Then, they can create a patent and seek commercial advantages.

Andy B.

[Clark Gilbert]

I'm in the process of uploading numerous pictures and one small video. I'll add the link when I'm done.

Index of files:
http://web.ics.purdue.edu/~cagilber/files/BallDrive/

Movie:
http://web.ics.purdue.edu/~cagilber/...e/DSCF1884.AVI

Prototype from around January 2003.
http://web.ics.purdue.edu/~cagilber/...ive/Prototype/

[jdiwnab]

This thread might also intrest you. It is more about creating a bot that is compleatly round. Similer, but I think more interesting and harder
http://www.chiefdelphi.com/forums/sh...ad.php?t=31115

[Sepsis900]

The 2-ball (snicker) system is way better than our original idea. More stability, and just as maneuverable. Here's what Andy said was the main disadvantage of the drive:

The main disadvantage is what I will try to describe as a "second level of traction loss". Consider a wheeled-driven robot. If the robot is pushing against an immovable object, then the wheels will most likely spin (the wheel torque results in a tangental wheel force that is lower than the robot weight x coefficient of friction, but that is a lesson for another day). When these wheels spin, the robot has lost its ability to move because of a loss of wheel traction with the carpet. In a ball drive system, there are two points of contact and EITHER could lose their traction. The small wheel can lose its traction with the ball, and the ball could lose its traction with the floor. The opportunity for traction loss in two places results increased efficiency losses in this style of drive system.

I think the wheel-to-ball traction loss can probably be remedied by using larger rubber wheels, which would provide a bigger contact area between the wheel and the ball, while still only giving the ball reasonably small resistance while rotating around the point of contact (the other wheel turning would rotate the wheel along this axis). The ball-to-ground problem could be fixed by using a tread pattern on the ball, which is simple: it only involves epoxying mountain-bike-like-tread-patterned rubber sections instead of smooth ones. Comments?

[Stu Bloom]

Believe me, Andy knows his stuff ...

While you can increase your traction (coefficient of friction) by selecting various materials and tread patterns/orientation, you cannot escape the laws of physics. You will either lose traction at some point, or burn out your motor(s). I believe Andy's point was that you double the number of locations where you have to consider that issue (in his ball-drive design).

It certainly would be an interesting project to tackle though. Please keep us posted on your progress.

[Sepsis900]

of course, losing traction or burning out your motors is a potential problem with any drive system. I'm just trying to delay the "losing traction" part, since that seems to be the major problem Andy pointed out

[greencactus3]

i know ball drives are all cool and holonomic but whats wrong with using a "Traditional" omniwheel setup? so much simpler and less places where friction is needed to drive stuff,

[Travis Covington]

Speaking of ball drive... Last I heard, Andy Baker has still yet to see the vehicles in I-Robot!

Help me out here, If he still hasn't seen the cars/trucks etc in the movie and their ball drive propulsions systems, please bug him until he does.

This post should fill my quota this month.