Mechanical design for motorized spherical caster wheels

Design goal is to have a mobile robot that operates on 3 large casters, essentially 2 to 4 inch diameter steel ball bearings, that are motorized. No other mechanism would touch the surface. The robot should thus be able to move in any XY direction on a flat surface, with steering being achieved by varying the speed and rolling direction of these wheels. The robot has no designated "front" side, so it does not need to (and should not have to) bodily turn, in order to move off in any given direction.

Conventional wheels or tracks are not the preferred approach.

Looking for suggested mechanical layouts of multiple rubber wheels, pressing down onto the steel ball from within the castor housing, to drive the ball in any direction. A single wheel on a stepper, rotated around the vertical axis using a sail-winch servo, is one approach under consideration. Would this be ideal, or are there any serious flaws in this approach?

Alternatively, is there any other suggested method of driving such a steel ball in any arbitrary direction under electronic control?

You could drive the sphere on two pairs of such wheels, with each pair driven by one motor. This would give you two axis control of the sphere. I.E. you can drive forwards or sideways.

Or you could use three wheels and three motors just like this robot:

This allows you to spin the sphere about the vertical axis, if that's helpful.

The good thing about this solution is that the wheels both drive the sphere, and bear the weight.

• I've seen the second idea before. If a video helps the OP, take a look at youtube.com/watch?v=bI06lujiD7E – Sidd Singal Oct 24 '12 at 22:54
• I see one key issue with this mechanism: Moving along any straight path would be sub-optimal, due to drag. Rotating the ball in the horizontal plane isn't needed, AFAICS. – Anindo Ghosh Oct 25 '12 at 3:21
• @AnindoGhosh - What drag? Notice that the omniwheels have tiny wheels on their circumference so that there is no drag when they move sideways. – Rocketmagnet Oct 25 '12 at 4:56
• @Rocketmagnet Minimal drag would be when the freewheeling movement is perpendicular to the wheel, but motion at an angle would cause some drag. – Anindo Ghosh Oct 25 '12 at 6:21
• This is interesting, but isn't it heading towards a more complex, and expensive, implementation than a single driving wheel at precisely the zenith, that itself can be rotated around the vertical axis? – Anindo Ghosh Oct 25 '12 at 6:22

What about two perpendicular rubber sheathed shafts? This would have the benefit of neatly separating out the x and y directions into two components.

This could be thought of as an inversion of the function of an old analogue mouse.

From the comments, it is clear that in order to rotate the steel balls in 100% x or y, the rollers on the opposing axis must be removed from contact with the balls.

There is also the assumption that the castor housing has bearings to keep the steel ball in position.

• I believe a configuration like that would have trouble moving the robot. If only the x shaft was rolling the ball would be braked by the y shaft. If you see this video: youtube.com/watch?v=bI06lujiD7E you can see that the wheels actually have small rollers enabling the ball to move perpendicular to the wheel. Kinda like these omnidirectional wheels: en.wikipedia.org/wiki/Omni_wheel – Nis Oct 24 '12 at 19:22
• @Nis That makes sense. I suppose that when only one is moving, the other could be allowed to spin freely, or removed from contact. That makes it more complex though. – S. Albano Oct 24 '12 at 20:48
• You can't remove the roller as the robot would only have one point of contact with the ball. It would tip over. Spinning the other roller freely wouldn't work either as the rotation of the ball would be perpendicular to the way the roller could rotate. – Nis Oct 24 '12 at 21:03
• @Nis I was picturing needing bearings to hold it in position, even in the case of two contact points. In the description, the OP stated that there would be a castor housing which appeared would do this. As there are three castors, it would not tip over. – S. Albano Oct 24 '12 at 23:21
• @Nis yes, tipping over isn't possible, it's 3 castors providing a wheeled tripod. – Anindo Ghosh Oct 25 '12 at 3:23